# ONGOING RESEARCH IN JAWED FISH IMMUNITY: STRUCTURAL AND FUNCTIONAL STUDIES AT THE PROTEIN AND CELLULAR LEVELS

EDITED BY : Monica Imarai and Brian Dixon 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 978-2-88963-731-7 DOI 10.3389/978-2-88963-731-7

#### About Frontiers

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

#### Frontiers Journal Series

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

#### Dedication to Quality

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

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

#### What are Frontiers Research Topics?

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

# ONGOING RESEARCH IN JAWED FISH IMMUNITY: STRUCTURAL AND FUNCTIONAL STUDIES AT THE PROTEIN AND CELLULAR LEVELS

Topic Editors: Monica Imarai, Universidad de Santiago de Chile, Chile Brian Dixon, University of Waterloo, Canada

Knowledge of jawed fish immune systems obtained in the last 15 years has been mostly obtained through the sequencing of genomes of different fish species and from the use of high-throughput techniques such as transcriptomic analysis and RNA sequencing, which has allowed characterization of immune gene responses at the transcriptomic level. Although these have been important tools for exploring the complexities of the immune responses of fish species, the next generation of knowledge requires the use of new tools and methods capable of unveiling the diversity of immune cells and molecules in jawed fish, the network of interactions and responses, and the mechanisms leading to immune protection against pathogens.

For more than 10 years, many laboratories have been engaged in developing antibodies against key molecules of jawed fish immune systems in several different species because the lack of sequence conservation of most immune genes makes it impossible to use the large panel of antibodies against mammalian molecules for recognition of fish molecules. The numerous genes encoding cytokines and receptors have now been used to make recombinant proteins available, allowing studies of the complex network of cytokine and receptor functions central to fish immune responses. From many laboratories, new and interesting knowledge about immune cells, their functions and interactions can be studied with the availability of new tools and methods have begun to provide a clear understanding fish immunity at the protein and cellular levels.

This Research Topic gives a comprehensive overview of the current knowledge of jawed fish immune responses with a particular emphasis on structural and functional studies at the protein and cellular levels.

Citation: Imarai, M., Dixon, B., eds. (2020). Ongoing Research in Jawed Fish Immunity: Structural and Functional Studies at the Protein and Cellular Levels. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-731-7

# Table of Contents


Chi Zhang, Shuangshuang Feng, Wenting Zhang, Nan Chen, Abeer M. Hegazy, Wenjie Chen, Xueqin Liu, Lijuan Zhao, Jun Li, Li Lin and Jiagang Tu


Ali Reza Khansari, Joan Carles Balasch, Eva Vallejos-Vidal, David Parra, Felipe E. Reyes-López and Lluís Tort

*60 Fish Lymphocytes: An Evolutionary Equivalent of Mammalian Innate-Like Lymphocytes?*

Giuseppe Scapigliati, Anna M. Fausto and Simona Picchietti

*68 Modulation of Innate Immune-Related Genes and Glucocorticoid Synthesis in Gnotobiotic Full-Sibling European Sea Bass (*Dicentrarchus labrax*) Larvae Challenged With* Vibrio anguillarum

Felipe E. Reyes-López, Johan Aerts, Eva Vallejos-Vidal, Bart Ampe, Kristof Dierckens, Lluis Tort and Peter Bossier


Beatriz Abos, Itziar Estensoro, Pedro Perdiguero, Marc Faber, Yehfang Hu, Patricia Díaz Rosales, Aitor G. Granja, Christopher J. Secombes, Jason W. Holland and Carolina Tafalla

*136 Behavioral Fever Drives Epigenetic Modulation of the Immune Response in Fish*

Sebastian Boltana, Andrea Aguilar, Nataly Sanhueza, Andrea Donoso, Luis Mercado, Monica Imarai and Simon Mackenzie

*150 Effects of Experimental Terrestrialization on the Skin Mucus Proteome of African Lungfish (*Protopterus dolloi*)*

Ryan D. Heimroth, Elisa Casadei and Irene Salinas


Tiehui Wang, Yehfang Hu, Eakapol Wangkahart, Fuguo Liu, Alex Wang, Eman Zahran, Kevin R. Maisey, Min Liu, Qiaoqing Xu, Mónica Imarai and Christopher J. Secombes

*250 Perspective on the Development and Validation of Ab Reagents to Fish Immune Proteins for the Correct Assessment of Immune Function* Brian Dixon, Daniel R. Barreda and J. Oriol Sunyer

*Patricia Pereiro1†, Alejandro Romero1†, Patricia Díaz-Rosales1 , Amparo Estepa2 , Antonio Figueras1 \* and Beatriz Novoa1 \**

*<sup>1</sup> Instituto de Investigaciones Marinas, Consejo Superior de Investigaciones Científicas (CSIC), Vigo, Spain, 2 Instituto de Biología Molecular y Celular (IBMC), Universidad Miguel Hernández, Elche, Spain*

#### *Edited by:*

*Brian Dixon, University of Waterloo, Canada*

#### *Reviewed by:*

*Irene Salinas, University of New Mexico, United States Francesco Buonocore, Università degli Studi della Tuscia, Italy*

#### *\*Correspondence:*

*Antonio Figueras antoniofigueras@iim.csic.es; Beatriz Novoa beatriznovoa@iim.csic.es These authors have contributed* 

*†*

#### *Specialty section:*

*equally to this work.*

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

*Received: 21 August 2017 Accepted: 18 October 2017 Published: 02 November 2017*

#### *Citation:*

*Pereiro P, Romero A, Díaz-Rosales P, Estepa A, Figueras A and Novoa B (2017) Nucleated Teleost Erythrocytes Play an Nk-Lysin- and Autophagy-Dependent Role in Antiviral Immunity. Front. Immunol. 8:1458. doi: 10.3389/fimmu.2017.01458*

With the exception of mammals, vertebrate erythrocytes are nucleated. Nevertheless, these cells are usually considered as mere carriers of hemoglobin. In this work, however, we describe for the first time an unrecognized role of teleost red blood cells (RBCs). We found that Nk-lysin (Nkl), an antimicrobial peptide produced by NK-cells and cytotoxic T-lymphocytes, was also expressed in flatfish turbot (*Scophthalmus maximus*) erythrocytes. Although the antiviral role of Nkl remains to be elucidated, we found a positive correlation between the transcription of *nkl* and the resistance to an infection with Rhabdovirus in a teleost fish. Surprisingly, Nkl was found to be present in the autophagolysosomes of erythrocytes, and therefore this higher resistance provided by Nkl could be related to autophagy. The organelles of RBCs are degraded through autophagy during the maturation process of these cells. In this work, we observed that the blockage of autophagy increased the replication of viral hemorrhagic septicemia virus in nucleated teleost erythrocytes, which suggests that this mechanism may also be a key process in the defense against viruses in these cells. Nkl, which possesses membrane-perturbing ability and was affected by this modulation of RBC autophagy, could also participate in this process. For the first time, autophagy has been described not only as a life cycle event during the maturation of erythrocytes but also as a pivotal antiviral mechanism in nucleated erythrocytes. These results suggest a role of erythrocytes and Nkl in the antiviral immunity of fish and other vertebrates with nucleated RBCs.

Keywords: teleost, erythrocytes, red blood cells, autophagy, autophagolysosome, Nk-lysin, granulysin

## INTRODUCTION

Erythrocytes, or red blood cells (RBCs), are the most abundant cells in the blood of vertebrates, and their primary function is to transport oxygen and carbon dioxide around the body. The enucleation of the erythroblast during erythropoiesis in mammals represents an evolutionary specialization that allows the increase of hemoglobin levels and enhances their flexibility and ability to traverse through capillaries (1, 2). Due to the anthropocentric vision that we unconsciously apply to our research, we may consider these cells as empty bags carrying hemoglobin without any other particular function. However, vertebrate evolution suggests that RBCs represent much more than that. Apart from mammals, vertebrate erythrocytes are nucleated and therefore possess the ability to modify their transcriptome and, in turn, their proteome. Any gene expressed even at low levels will achieve high proportions in the organism since RBCs are the most numerous blood cells in vertebrates. Little is known about the potential immune function of RBCs. Most of the investigations were carried out in mammals, and enucleated RBCs have only been implicated in some immune activities that are mainly mediated by hemoglobin (3–11). Although the information about the immune capabilities of nucleated erythrocytes is almost nonexistent, it was previously suggested that non-mammalian (fish and birds) erythrocytes possess the ability to specifically detect pathogenassociated molecular patterns (PAMPs) and participate in the immune response (12–16). In the case of teleost fish, there is still a lack of understanding of the immune function not only of erythrocytes but also of other immune cell populations such as natural killer and dendritic cells. This lack of understanding is mainly due to the absence of specific cell markers.

Red blood cells suffer a natural maturation process in which their organelles, such as mitochondria, endoplasmic reticulum, and peroxisomes, are degraded *via* autophagy (17, 18). In mammals, RBCs are released to the circulation in a complete mature stage; however, in teleost fish, a variable percentage of non-mature erythrocytes can be observed (19, 20). Autophagy is a highly conserved cellular selfdegradative pathway in which cytoplasmic materials (e.g., misfolded or aggregated proteins, damaged organelles and/or intracellular pathogens) are engulfed into double-membrane bound vesicles for proteolytic degradation (21). This mechanism is also used by cells to obtain energy in response to starving conditions and during critical developmental processes (21). During the autophagy process, the autophagosome containing the cytoplasmic material for degradation fuses with a lysosome to form the autophagolysosome, where the lysosomal hydrolases degrade the enclosed materials (22). In addition to the role of the autophagy in the maintenance of the cell homeostasis, this process is implicated in the defense against intracellular pathogens, including viruses (23–25). Viral recognition by endosomal toll-like receptors (TLRs) or cytoplasmic viral nucleic-acid sensors can mediate the induction of autophagy for viral degradation in autophagolysosomes (26–29). Autophagy can also activate other innate and adaptive immune responses to fight against the virus (25). Moreover, autophagy is known to play a key role in the defense against Rhabdoviruses that affect both mammals and fishes (30–32). Nevertheless, this process has not been previously associated with the antiviral defense in erythrocytes.

In this work, we investigated the antimicrobial peptide (AMP) Nk-lysin (Nkl, orthologous to human granulysin), which has been considered to be produced by natural killer cells (NK-cells) and cytotoxic T lymphocytes (CTLs) and stored in cytolytic granules together with perforin and granzymes (33, 34). Surprisingly, Nkl was found in the autophagolysosomes of turbot RBCs. Our results also indicate that Nkl is involved in the resistance against viral hemorrhagic septicemia virus (VHSV) in turbot, and therefore, we hypothesize that autophagy might be the mechanism linking Nkl to VHSV resistance. Indeed, the blockage of autophagy in erythrocytes favored the viral replication in these cells and also affected the levels of Nkl. This suggests that fish erythrocytes play an active role against VHSV mediated through autophagy and involves Nkl. These data open the door to further investigations on the implication of erythrocytes and Nkl in the immunity of fish and other vertebrates with nucleated RBCs.

#### MATERIALS AND METHODS

#### Characterization and Phylogenetic Analysis of Turbot Nkl

The complete open reading frame (ORF) of the turbot *nk-lysin* (*nkl*) gene was obtained from a previous 454-pyrosequencing of turbot tissues (35) and confirmed by sequencing using specific primers (Table S1 in Supplementary Material). A local blast against the turbot genome (36) was conducted to identify other potential *nkl* genes and to determine the number of exons/introns constituting the turbot *nkl*.

The presence of signal peptide was analyzed with the SignalP 3.01 server (37) and the presence of specific domains with SMART 4.02 (38). The three-dimensional (3D) structure of turbot Nkl was predicted using I-TASSER server (39) selecting the model with the best C-score and viewed by PyMOL.3 An alignment between several Nkls/granulysins protein sequences from fish, birds, and mammals was conducted using the ClustalW server (40). A phylogenetic tree was drawn using Mega 6.0 software (41) and selecting the model of protein evolution that best fits a given alignment according to the ProtTest 2.44 server (42). Sequence similarity and identity scores were calculated with the software MatGAT (43) using the BLOSUM62 matrix. The GenBank accession numbers of the sequences used in this section are listed in Table S2 in Supplementary Material.

#### Fish and Virus

Juvenile and adult turbot (average weight 2.5 and 125 g, respectively) were obtained from a commercial fish farm (Insuiña S.L., Galicia, Spain). Prior to experiments, fish were acclimatized to the laboratory conditions for 2 weeks. When necessary, fish were euthanized *via* MS-222 overdose (500 mg/L). All the experimental procedures were reviewed and approved by the CSIC National Committee on Bioethics under approval number ES360570202001/16/FUN01/PAT.05/tipoE/BNG.

Viral hemorrhagic septicemia virus (strain UK-860/94) was propagated in the Epithelioma Papulosum Cyprini (EPC) cell line (ATCC, CRL-2872) at 14°C in MEM (Gibco) supplemented with 2% FBS (Gibco), and 100 µg/mL Primocin (InvivoGen). The virus stock was titrated into 96-well plates according to established protocols (44, 45). VHSV aliquots were stored at −80°C until use.

#### *nkl* Expression Plasmid, HEK-293 Cell Line, and Cell Transfection

The expression plasmid pMCV1.4-*nkl* was synthesized by ShineGene Molecular Biotech, Inc. (Shanghai, China) using the pMCV1.4 plasmid (Ready-Vector, Madrid, Spain) and the nucleotide sequences encoding the turbot Nkl mature peptide.

<sup>1</sup>http://www.cbs.dtu.dk/services/SignalP/.

<sup>2</sup>http://smart.embl.de/.

<sup>3</sup>http://www.pymol.org.

<sup>4</sup>http://darwin.uvigo.es/software/prottest2\_server.html.

The plasmid was cloned by transforming One Shot TOP10F' competent cells (Invitrogen) and purified using the PureLink™ HiPure Plasmid Midiprep Kit (Invitrogen).

Human HEK-293 cells (ATCC CRL-1573) were grown in Eagle's Minimum Essential Medium (Gibco) supplemented with 100 µg/mL primocin (InvivoGen), 1× non-essential amino acids (Gibco), 1-mM sodium pyruvate (Gibco), and 10% FBS. The cells were incubated in a 5% CO2 atmosphere at 37°C.

Recombinant Nkl was produced by transfection of 6 µg of the plasmid pMCV1.4-*nkl* into HEK-293 cells at 70–80% confluence (T-25 flask) using the XtremeGENE HP DNA Transfection Reagent (Roche) according to the manufacturer's instructions. The same process was conducted with the corresponding empty plasmid pMCV1.4. Forty-eight hours after transfection, the supernatants were collected, filtered by 0.22 µm, and stored at −80°C until further use.

#### Anti-Nkl Polyclonal Antibody Production and Validation

Emini surface accessibility scale (45), Kolaskar and Tongaonkar antigenicity scale (46), and Bepipred Linear Epitope Prediction (47) methods were used to predict the best Nkl antigen binding regions of antibodies. Based on this information, two peptides were chosen (RSLEINIDDQEQVC and CLFYPKQEESQTE). To obtain the anti-Nkl polyclonal antibody (New England Peptide, Gardner, MA, USA), rabbits were co-immunized with both synthetic peptides. Blood was collected before injection (preimmune serum) and 30 days after the immunization (polyclonal antibody).

The anti-Nkl polyclonal antibody was validated by western blot (WB). For this, 15 µL of the supernatants from HEK-293 cells transfected with pMCV1.4-*nkl* or pMCV1.4 were mixed with 1× NuPAGE LDS Sample Buffer (Invitrogen) and resolved in a 4–20% Mini-PROTEAN TGX™ gel (Bio-Rad) (with and without 2-Mercaptoethanol and heat treatment 5 min at 95°C), and transferred to a nitrocellulose membrane (Bio-Rad). The membrane was blocked for 2 h with 3% (w/v) bovine serum albumin (BSA) in TBST buffer (20-mM Tris, 0.5-M NaCl, 0.1% Tween 20) and incubated for 2 h with the rabbit anti-Nkl polyclonal antibody (dilution 1:500 in 1% BSA-TBST buffer) at room temperature (RT). After three 10-min washes with TBST, membrane was incubated with a goat anti-rabbit-HRP antibody (Sigma) (dilution 1:10,000) for 1 h at RT, washed again, and revealed by chemiluminescence detection with Luminata™ Forte Western HRP Substrate (Millipore), and visualized with the ChemiDoc XRS + system (Bio-Rad).

#### *nkl* Constitutive Expression in Different Tissues and *In Vivo* Induction after VHSV Challenge

To examine the constitutive expression of *nkl*, 11 different tissues (peritoneal exudate cells – PEC–, blood, head kidney, trunk kidney, spleen, gill, liver, intestine, heart, brain and muscle) were obtained from three adult healthy fish. PECs were obtained as previously described (48).

The modulation of *nkl* was also analyzed after an *in vivo* VHSV infection. A total of 50 juvenile turbot were divided into two groups. The first group (*n* = 25) was intraperitoneally (i.p.) injected with 50 µL of a VHSV suspension containing 2 × 106 TCID50/mL. The second group was injected with the same volume of culture medium. The head kidney and spleen were removed from five turbot at 1, 2, 3, and 7 days post-infection (dpi), constituting five biological replicates for each tissue and sampling point. These samples were processed for the analysis of *nkl* expression.

#### Correlation between *nkl* Transcription Level and Resistance to VHSV

*nkl* expression was analyzed in the head kidney samples from four turbot families showing different mortality rates after VHSV infection. Two VHSV-resistant (1 and 4) and two-susceptible families (2 and 3) were previously described by Diaz-Rosales et al. (49). Five animals of each family were analyzed before (naïve) and 24 h after the VHSV challenge conducted by Díaz-Rosales et al. (49). The expression of *nkl* was analyzed by quantitative polymerase chain reaction (qPCR).

Additionally, the correlation between the constitutive expression level of *nkl* gene in blood before a viral infection and the resistance to a VHSV challenge was determined. Approximately 20 µL of blood were extracted from the caudal vein of juvenile turbot using a heparinized syringe and cells were processed for the analysis of *nkl* expression. One week after the blood extraction, the turbot were i.p. injected with 100 µL of a VHSV suspension (3 × 107 TCID50/mL). Mortality was recorded for 15 days and the size and weight of each turbot were also registered. The correlation between *nkl* mRNA levels in blood before the infection, the day of death, size and weight was determined using the Spearman's rho correlation test.

#### *nkl* Expression in Blood Cells and *In Vitro* VHSV Replication

The expression of *nkl* and the replication of VHSV in blood and in purified erythrocytes were analyzed by qPCR. Blood was taken from the caudal vein of three adult turbot using a heparinized syringe. Erythrocytes were purified in a Percoll (GE Healthcare) 51% gradient by centrifuging at 400× *g* for 30 min at 4°C without brake. Total blood cells and purified erythrocytes were adjusted to 108 cells/mL in MEM (Gibco) supplemented with 2% FBS (Gibco) and 100 µg/mL Primocin (InvivoGen). Cells (250 µL) were distributed onto 24-well plates. A proportion of these cells were infected with VHSV (104 TCID50/mL) and the remaining wells were maintained as controls. Cells were incubated at 15°C and collected 2, 3, 5, and 7 dpi for the quantification of *nkl* transcripts and VHSV glycoprotein gene by qPCR.

#### Chloroquine (CQ) and Rapamycin Treatments

For qPCR analysis, erythrocytes were purified, seeded, infected with VHSV, and maintained as mentioned above. A proportion of the erythrocytes were incubated with CQ (25 µM; Sigma-C6628) or rapamycin (RAP) (5 µM; Sigma-R0395). Cell samples were taken 1, 2, and 3 dpi for the quantification of *nkl*, *becn1* and *atg5* transcripts and VHSV glycoprotein gene.

For confocal microscopy analysis, total blood cells were incubated with CQ (25 µM) or RAP (5 µM) for 3 days. Cells were fixed and immunostained as described below.

## Effect of the DNA Vaccine Encoding the VHSV Glycoprotein (pMCV1.4-G)

The modulation of the *nkl* gene and the protein abundance and distribution of Nkl under vaccination and/or infection conditions were evaluated by qPCR and flow cytometry. Twelve adult turbot were intramuscularly (i.m.) injected with 50 µL of a DNA vaccine (2 µg/fish) encoding the G glycoprotein from VHSV (pMCV1.4-G860) (50), whereas the other 12 fish were injected with the same amount of the empty plasmid (pMCV1.4). One month after vaccination, 6 fish from each group were i.p. injected with 50 µL of VHSV (2 × 106 TCID50/mL) and the remaining six fish were injected with cell culture medium. The head kidney, spleen, and blood were taken at 48 h post-infection (hpi). Cell suspensions from the head kidney and spleen were prepared by passing the tissue through a 40-µm nylon mesh in phosphatebuffered saline (PBS). Blood was collected from the caudal vein using a heparinized syringe and diluted in PBS. All the samples were divided into two groups to be analyzed at the same time by qPCR and flow cytometry.

#### RNA Extraction, cDNA Synthesis, and Real-time qPCR Analysis

Total RNA from the different tissue samples was extracted using the Maxwell® 16 LEV simplyRNA Tissue kit (Promega) with the automated Maxwell® 16 Instrument in accordance with instructions provided by the manufacturer. The cDNA synthesis was performed with the SuperScript II Reverse Transcriptase (Invitrogen) using 0.5 µg of RNA and following the manufacturer indications, except for the blood cell samples, in which case the cDNA synthesis was conducted with SuperScript III Reverse transcriptase (Invitrogen) using 0.1 µg of RNA.

Gene expression profiles were determined using real-time qPCR. Specific qPCR primers were designed using the Primer3 program (51) and their amplification efficiency was calculated using seven, fivefold serial dilutions of cDNA from unstimulated turbot with the threshold cycle (CT) slope method (52). The identity of the amplicon was confirmed by sequencing. Individual qPCR reactions were conducted in 25-µL reaction volume using 12.5 µL of SYBR GREEN PCR Master Mix (Applied Biosystems), 10.5 µL of ultrapure water (Sigma-Aldrich), 0.5 µL of each specific primer (10 µM), and 1 µL of fivefold diluted cDNA template in MicroAmp optical 96-well reaction plates (Applied Biosystems). All reactions were performed using technical triplicates in a 7300 Real-Time PCR System thermocycler (Applied Biosystems) with an initial denaturation (95°C, 10 min) followed by 40 cycles of a denaturation step (95°C, 15 s) and one hybridization–elongation step (60°C, 1 min). No-template controls were also included on each plate to detect possible contamination or primer dimers formed during the reaction. An analysis of melting curves was performed for each reaction. Relative expression of each gene was normalized using the eukaryotic translation *elongation factor 1 alpha* (*ef1a*) as reference gene, which was constitutively expressed and not affected by the experimental treatments, and was calculated using the Pfaffl method (52). Primer sequences used for the quantification of *nkl*, *becn1*, *atg5*, and VHSV glycoprotein transcripts are listed in Table S1 in Supplementary Material.

#### Immunofluorescence Assays, Flow Cytometry, and Confocal Microscopy

The head kidney, spleen, total blood cells, and purified erythrocytes samples were obtained as previously described from adult turbots. Cells were fixed with 2% paraformaldehyde during 15 min at 4°C. After washing, the cells were blocked by incubating for 1 h in PBS with 0.1% saponin (Sigma) and 2% of BSA (Sigma). Then, cells were incubated overnight at 4°C with the preimmune serum or with the rabbit anti-Nkl polyclonal antibody in staining buffer (PBS with 0.1% saponin and 0.1% BSA) (dilution 1:250). Cells were then washed and incubated with the secondary antibody Alexa Fluor® 488 goat anti-rabbit IgG (Molecular Probes-Life Technologies; 1:1,000) for 1 h at RT. Samples were washed and resuspended in PBS. The expression of Nkl was analyzed using a FACSCalibur flow cytometer (BD Biosciences) in dot plots of relative size (forward-light-scatter, FSC) and complexity (side-light-scatter, SSC) in linear and logarithmic scale. FL1-H histograms were used to compare the fluorescence levels emitted by samples labeled with the anti-Nkl antibody and the preimmune serum. The percentage of positive fluorescent events and the intensity of fluorescence (median) were registered.

Cells were adjusted to 106 cells/mL and distributed onto 24-well plates with 12-mm glass coverslips and incubated at 15°C for 2 h before the fixation. Cells were fixed with 2% paraformaldehyde for 15 min at 4°C. After washing, the cells were blocked by incubating for 1 h in PBS with 0.1% saponin (Sigma) and 2% of BSA (Sigma). Then, cells were incubated overnight at 4°C with the corresponding primary antibody in staining buffer (PBS with 0.1% saponin and 0.1% BSA). Cells were then washed and incubated with the secondary antibody for 1 h at RT. Nkl was stained using the rabbit anti-Nkl polyclonal antibody (1:250) and the secondary antibody Alexa Fluor® 488 goat anti-rabbit IgG or Alexa Fluor® 546 goat anti-rabbit IgG (Molecular Probes-Life Technologies) (1:1,000), depending on the experiment. Autophagy activity was analyzed using a rabbit anti-LC3A/B polyclonal antibody (Cell Signaling; 4108S) (1:200), or a mouse anti-LC3B monoclonal antibody (Nanotools; 0231-100/LC3-5F10) (1:20) for co-localization assays. The Alexa Fluor® 546 goat anti-rabbit IgG and Alexa Fluor® 488 goat anti-mouse IgG (1:1,000) were used as secondary antibodies, respectively. The immune detection of VHSV in blood samples was performed using the mouse anti-N VHSV monoclonal antibody 3E7 (1:1,000) (53) and the secondary antibody Alexa Fluor® 635 goat anti-mouse IgG (Molecular Probes-Life Technologies) (1:1,000). All samples were stained with a DAPI solution (Molecular Probes-Life Technologies) for nuclear localization and mounted using ProLong Antifade Reagents (Life Technologies). LysoSensor blue DND-167 reagent (Molecular Probes-Life Technologies) was used to stain the acidic lysosomal vesicles in live cells. Confocal images were captured using a TSC SPE confocal microscope (Leica) using the LAS AF software (Leica). The 3D reconstructions were performed using the Image Surfer5 software.

#### Transmission Electron Microscopy (TEM) Images

Blood samples were fixed overnight with 2% glutaraldehyde in 0.1-M cacodylate buffer at pH 7.4. Then, samples were washed and incubated with 1% tannic acid in 0.1-M cacodylate buffer at 4°C for 1 h and washed again. The cells were centrifuged at 1,000× *g* for 5 min and included in 1% agarose blocks, which were sectioned in 1 mm3 pieces and washed with 0.1-M cacodylate buffer. The sections were incubated for 1 h with 1% osmium tetroxide at 4°C and washed and dehydrated in increasing concentrations of ethanol. After the dehydration, samples were embedded in Epon resin and 65–85 nm sections were prepared using the ultramicrotome and mounted on metal grids. Ultrathin sections were stained with 50% uranyl acetate in methanol and lead citrate prior to observation with the JEOL JEM-1010 transmission electron microscope (Electron Microscopy Unit of CACTI, University of Vigo, Spain).

5http://cismm.web.unc.edu/software/.

#### Statistical Analysis

Both qPCR expression results and flow-cytometry fluorescence data were represented graphically as the mean/median + the standard deviation of the biological replicates. To determine significant differences, data were analyzed with the computer software package SPSS v.19.0 using the Student's *t*-test or ANOVA as appropriate. For the correlation analysis, Spearman's Rho correlation coefficient was calculated. Differences were considered statistically significant at *p* < 0.05.

#### RESULTS

#### Turbot *nkl*

The complete coding region of the turbot *nkl* gene was deposited in GenBank under Acc. No. KU705506. The characteristic saposin B (SapB) domain of the saposin-like proteins (SAPLIP) family was identified (**Figure 1A**). The 3D structure was constructed with a moderate confidence value using Nkl from pig as a template (TM score = 0.582) (**Figure 1B**). On the other hand, the gene structure (exon/intron organization) was conserved among teleost fish (**Figure 1C**) and compared with other vertebrates.

A multiple alignment of several Nkls/granulysin amino-acid sequences from fish, birds, and mammals revealed six cysteine

FIGURE 1 | Characterization of turbot Nkl. (A) The complete coding region of turbot *nkl* (444-bp long) encodes a protein of 147 residues. The signal peptide is underlined and the SapB domain is highlighted. (B) 3D structure of turbot Nkl using the pig protein as a template (TM score = 0.582). The tertiary structure comprises six α-helices. (C) Structure of the turbot *nkl* gene and alignment between the coding region and the corresponding genomic sequence. This gene contains five exons and four introns. The 5′ and 3′ UTRs are represented as gray boxes, the CDSs of the exons as white boxes, and introns as solid lines. The length (bp) of the CDSs and introns is also reflected in the figure. CDSs, coding DNA sequences; Nkl, Nk-lysin; SapB, saposin B; UTRs, untranslated regions.

residues that were well conserved among the different species (Figure S1A in Supplementary Material). A phylogenetic tree showed two main clusters, one of them containing teleost Nkl and the other one containing avian and mammalian sequences (Figure S1B in Supplementary Material). As expected, an identity/ similarity matrix (Table S3 in Supplementary Material) revealed that turbot Nkl shares the highest scores with the other flatfish species, but when it was compared with sequences from birds and mammals, the identity ranked between 16 and 20% and the similarity between 36 and 45%.

#### *nkl* Transcription Level Related to Antiviral Response

The constitutive expression of the *nkl* gene was determined in different tissues from healthy turbot. *nkl* transcription was detected in all the tested tissues but the highest expression levels were detected in immune tissues (peritoneal exudate cells –PEC– followed by the spleen and head kidney) (**Figure 2A**). When turbot were infected with VHSV, the *nkl* gene was overexpressed in the two main immune organs in fish, head kidney (**Figure 2B**) and spleen (**Figure 2C**), suggesting an antiviral response.

The expression of *nkl* was analyzed in head kidney samples from fish belonging to VHSV-resistant and VHSV-susceptible families, before (naïve) and after VHSV challenge. The resistant families (1 and 4) showed a significantly higher expression of *nkl* than the susceptible ones (2 and 3) before the viral infection. Interestingly, 24 h after VHSV challenge, a significant increase in *nkl* transcription was only observed in susceptible families (2 and 3) (**Figure 2D**).

We used a non-destructive method (blood sampling from caudal vein) to determine the *nkl* transcription level before infection in 35 turbot and, after VHSV challenge, individual mortality was registered. The correlation between *nkl* mRNA level in blood cells, turbot size and weight, and the day of death after VHSV infection was analyzed using the Spearman's rho correlation test. Mortalities started 4 days after infection, reaching 90% cumulative mortality at the end of the experiment (15 days after infection). A significant positive correlation between *nkl* constitutive transcription in the blood and the day of death was observed (*R* = 0.438; *p* = 0.008) (**Figure 2E**): turbot with higher constitutive expression of *nkl* survived longer after infection than those with low expression level. Turbot size and weight were also correlated with the day of death, but no relationship between size or weight and the level of *nkl* transcripts was detected (Table S4 in Supplementary Material). Therefore, the basal transcription of *nkl* is a size/weight-independent factor, and its level is correlated with the resistance to VHSV infection.

#### VHSV Glycoprotein Inducing Long-Lasting Effects in the Levels of Nkl

The expression levels of *nkl* in the head kidney, spleen, and blood were analyzed by qPCR 1 month after the injection of a highly efficient DNA vaccine encoding the VHSV glycoprotein (pMCV1.4-G) (50) or the corresponding control empty plasmid (pMCV1.4) under both healthy and VHSV-infected conditions (48 hpi) (**Figure 3**). In parallel, flow-cytometry analysis of cell populations from the three different tissues was also conducted (**Figure 3**). Surprisingly, 1 month after vaccination, there was an increase in the transcription of the *nkl* gene detected in the head kidney and blood samples (**Figures 3C,I**), but not in the spleen (**Figure 3F**). Nevertheless, after a VHSV challenge, the expression of *nkl* increased only in the head kidney from non-vaccinated individuals but decreased in blood samples from vaccinated fish (**Figures 3C,I**). To detect the presence of Nkl at the protein level, we designed and used an anti-Nkl polyclonal antibody. After confirming the specificity of the antibody by WB (Figure S2 in Supplementary Material), flow-cytometry analysis was conducted. Differences among vaccinated and non-vaccinated turbot (in the absence of infection) were only observed in the spleen, with an increase in the percentage of Nkl-positive cells but a reduction in the median value of fluorescence per cell (**Figures 3D,E**). After the VHSV challenge, Nkl increased in the head kidney from nonvaccinated fish but decreased in vaccinated turbot (**Figure 3B**); however, there were no differences in the number of Nkl-positive cells (**Figure 3A**). In the blood, the percentage of Nkl-positive cells decreased after infection in both vaccinated and non-vaccinated fish (**Figure 3G**), but no significant differences in the level of Nkl were detected in these cells (**Figure 3H**).

### Nkl Distribution in the Head Kidney and Blood Cells

Flow cytometry was conducted on head kidney cells and total blood samples (**Figures 4A,C**). Nkl-positive cells gated in the FL1-H histogram were represented in FSC/SSC density plots (**Figures 4B,D**). In the head kidney, the positive cells were clustered in a heterogeneous population showing low size and complexity (**Figure 4B**). In blood samples, the fluorescence histogram revealed the presence of two clear populations (**Figure 4C**) with different positions in the FSC/ SSC density plot (**Figure 4D**). One population corresponded to cells with the lowest fluorescence level and small size and complexity. Erythrocytes are the most abundant cell type in this population. The other population consisted of cells that had a higher fluorescence, were larger in size, and were essentially the white cell population. A clear significant difference in the fluorescence level was detected among both populations. Fluorescence values for erythrocytes were 168 ± 16.5 and 605 ± 52.4 for leukocytes (**Figure 4E**).

To visualize the expression of Nkl in these cells, the distribution of the Nkl peptide was analyzed by confocal microscopy of the head kidney and blood cells. In the small spherical cells with a high nucleus/cytoplasm ratio, probably corresponding to CTLs and the hypothetical NK-cells, the cytoplasm was completely stained with the anti-Nkl antibody (**Figures 4F,G**). Turbot erythrocytes were also positive for Nkl-immunostaining (**Figures 4F,G**).

## *nkl* Expression in Erythrocytes

The unexpected presence of Nkl in erythrocytes was analyzed in detail. Confocal images of erythrocytes immunostained with the preimmune serum or the anti-Nkl antibody confirmed the specificity of the polyclonal antibody (**Figures 5A,B**). Almost all erythrocytes were found to be Nkl-positive and this peptide was mainly expressed in a large, spherical cytoplasmic structure; however, a few small Nkl-positive spots were also observed (**Figure 5C**). The mRNA expression of *nkl* in this cell type was lower compared with that of total blood cells (**Figure 5D**), suggesting that the

FIGURE 2 | Tissue distribution of the *nkl* expression, induction after a VHSV challenge in lymphoid tissues, and relation between nkl transcription and resistance against VHSV. (A) Constitutive expression of *nkl* in different tissues from healthy adult turbot. Normalized expression values are represented as the mean of three individuals plus SD. (B,C) Modulation of *nkl* gene expression in the head kidney (C) and spleen (B) after VHSV infection. Data are expressed as fold-change regarding the values obtained in controls at each sampling point. Graphs represent the mean of five biological replicates plus SD. Significant overexpression (*p* < 0.05) are represented with asterisks. (D) Normalized expression of *nkl* gene in head kidney samples from resistant families (1 and 4) and susceptible families (2 and 3), before and after VHSV infection. Graph represents the mean of five biological replicates plus SD. Letters (A,B) indicate significant differences (*p* < 0.05) between the families in naïve conditions. Asterisks indicate significant differences (*p* < 0.05) between the naïve and VHSV-infected condition in each family. (E) Correlation between *nkl* transcription level in blood cells (before infection) and the day of death after VHSV challenge in juvenile turbot. The Spearman's Rho correlation test showed a significant (*p* = 0.008) but moderate (*R* = 0.438) positive correlation between both variables. VHSV, viral hemorrhagic septicemia virus.

expression of this gene is much higher in other *nkl*-expressing cells present in blood (leukocytes). The FL1-H fluorescence profile of purified erythrocytes showed a well-defined peak (**Figure 5E**) corresponding to a homogeneous cell population (**Figure 5F**).

To further characterize the structure of the cytoplasmic vesicles, a study using Transmission electron microscopy (TEM) was conducted (**Figure 5G**). At low magnification, a large, spherical cytoplasmic structure was observed in almost all erythrocytes, although in some cells other smaller vesicles were also found. These structures showed a double-membrane surrounding electron-dense structures, which probably correspond to cellular organelles such as mitochondria, Golgi apparatus,

and endoplasmic reticulum. LysoSensor staining revealed the acidic nature of these spherical structures in the cytoplasm of erythrocytes (**Figure 5H**), and immunostaining with a rabbit polyclonal anti-LC3 antibody showed that autophagy is also occurring in these structures (**Figure 5I**). These data suggested that these Nkl-containing structures probably correspond to autophagolysosomes.

#### Nkl Involvement in Autophagy

To fully elucidate whether these Nkl-positive vesicles correspond to the LC3-positive structures, we conducted a co-localization analysis after *in vitro* stimulations of erythrocytes with the autophagy inhibitor CQ and the autophagy activator RAP. The execution of autophagy involves the participation of numerous proteins; however, during the final steps of autophagy, only microtubule-associated protein light chain-3 (LC3) is known to exist in mature autophagolysosomes (54).

Although the spherical Nkl-positive structures were also labeled with the rabbit polyclonal anti-LC3 A/B antibody (**Figure 5I**), in the co-localization studies, the mouse monoclonal anti-LC3B antibody resulted in LC3B-positive punctate structures that were dispersed in the cytoplasm of untreated erythrocytes (**Figure 6A**) whereas Nkl was strongly detected in the autophagolysosomes. Therefore, LC3B and Nkl did not co-localize. Interestingly, the incubation of the erythrocytes with CQ and RAP completely modified this pattern after 24 h. In the CQ-treated cells, LC3B was now confined to the autophagolysosomes, and the Nkl signal disappeared (**Figure 6A**). This is probably because CQ raises intravesicular pH (55) and the SAPLIP show markedly increased activities at acidic pH (56) because of their pH-dependent conformational properties (57). On the other hand, the autophagy activator RAP also affected the distribution of LC3B. In this case, although numerous LC3B-positive puncta were also observed in the cytoplasm, the higher fluorescent signal was found in the

were gating at 300-9910. (B) Distribution of head kidney cell populations using SSC-H vs. FL1-H density plots. The FSC-H/SSC-H position of Nkl-positive stained cells is boxed. (C) Fluorescence profile of blood samples stained with preimmune serum and polyclonal antibody. Positive cells for Nkl were gated at 262-9910. (D) Dot plot representing the position of the two Nkl-positive cell populations (red: cell population enriched in erythrocytes or red cells; green: cell population enriched in white cells). (E) Mean fluorescence value registered in blood cell populations. Asterisks indicate significant differences (*p* < 0.05). (F,G) Confocal microscopy images of a head kidney and a blood sample, respectively, showing white and red cells stained with the anti-Nkl polyclonal antibody. Green: Nkl. Blue; DAPI. Scale bar, 10 µm. FSC, forward-light-scatter; Nkl, Nk-lysin; SSC, side-light-scatter.

autophagolysosomes where a strong co-localization of LC3B and Nkl was detected (**Figure 6A**). A detailed 3D reconstruction of the erythrocytes incubated with RAP showed that Nkl seems to surround the LC3B signal in the autophagic structures (**Figure 6B**).

## Erythrocytes Showing Antiviral Activity That Depends on Autophagy and Nkl

*In vitro* infection of total blood cells and erythrocytes revealed another interesting finding. Erythrocytes were found to be positive for VHSV infection using immunofluorescence staining

FIGURE 5 | Analysis of Nkl in erythrocytes by confocal microscopy, qPCR and flow cytometry, and the characterization of the cytoplasmic structures containing Nkl. (A,B) Immunocytofluorescence of purified erythrocytes stained with the rabbit preimmune serum (A) or the anti-Nkl polyclonal antibody (B), and the secondary antibody Alexa Fluor 488 goat anti-rabbit IgG. Nuclei were stained with DAPI. Scale bar, 10 µm. The nonspecific fluorescence of the preimmune serum was not registered in samples stained with the polyclonal antibody. (C) Nkl was mainly detected in a large, spherical structure. Moreover, small puncta were also present in the cytoplasm. Scale bar, 10 µm. (D) Normalized expression of *nkl* in total blood cells and in purified erythrocytes. Bars represents the mean plus SD of 3 individual samples. (E) FL1-H fluorescence profile of purified erythrocytes stained with preimmune serum and polyclonal antibody. A specific signal of the antibody was clearly registered. (F) FSC/SSC density plot of FL1-H positive cells gated between 262-9910 log scale. (G) Description of the Nkl-positive vesicles by TEM. Almost all turbot erythrocytes present a large, double-membrane bound structure in the cytoplasm, although a few smaller ones are also observed (H) LysoSensor Blue staining of turbot erythrocytes. The spherical structures containing Nkl are LysoSensor-positive acidic vesicles. Scale bar, 10 µm. (I) Immunocytofluorescence of purified erythrocytes stained with the rabbit anti-LC3A/B polyclonal antibody, and the Alexa Fluor 546 goat anti-rabbit IgG as secondary antibody. Nuclei were stained with DAPI. Scale bar, 10 µm. Nkl, Nk-lysin; qPCR, quantitative polymerase chain reaction; TEM, transmission electron microscopy.

(**Figure 7A**), although flow cytometry revealed that VHSVpositive erythrocytes were less than 1% after 24 h. VHSV replication was analyzed both in total blood cells and erythrocytes by qPCR detection of the viral glycoprotein (G). Viral replication was higher in erythrocytes compared with total blood cells, and the viral detection increased over time (**Figure 7B**). Opposed to

the VHSV replication, the time-course experiment revealed that, both in total blood cells and erythrocytes, the transcription of *nkl* decreases in a time-dependent, infection-independent manner (Figures S3A,B in Supplementary Material).

The effect of CQ and RAP during *in vitro* VHSV infection in erythrocytes was also studied. CQ and RAP modulated the mRNA levels of autophagy-related genes (Figures S3C,D in Supplementary Material) and, interestingly, the substances affected the transcription of *nkl* in an opposite manner. CQ treatment damped the time-dependent reduction of *nkl* transcription, whereas RAP significantly reduced the level of *nkl* transcription after 24 h (**Figure 7C**).

FIGURE 7 | VHSV infecting and replicating in erythrocytes. (A) Confocal images of double-immunofluorescence staining of VHSV nucleoprotein −N− and Nkl in blood cells. The virus was stained with the mouse anti-N VHSV monoclonal antibody 3E7 and the secondary antibody Alexa Fluor 635 goat anti-mouse IgG, and Nkl with the rabbit anti-Nkl polyclonal antibody and Alexa Fluor 488 goat anti-rabbit IgG. Nuclei were stained with DAPI. Red: N-VHSV, Green: Nkl, Blue: DAPI. Scale bar, 10 µm. (B) The replication of VHSV in erythrocytes and total blood cells after an *in vitro* infection was measured by qPCR detection of the VHSV glycoprotein −G− gene. Bars represents the mean plus SD of three individual samples. (C,D) Effect of chloroquine and rapamycin in the transcription of *nkl* (C) and in the VHSV replication (D) in erythrocytes. Bars represents the mean plus SD of three individual samples. Nkl, Nk-lysin; qPCR, quantitative polymerase chain reaction; VHSV, viral hemorrhagic septicemia virus.

While there were no significant differences in the detection of the VHSV G gene in those erythrocytes incubated with RAP compared with the control, autophagy blockage with CQ favored viral replication, which is reflected through the higher detection of the VHSV G gene in the cells incubated with this compound (**Figure 7D**). This result indicates that autophagy is an important antiviral defense mechanism in these cells.

#### DISCUSSION

Nk-lysin is an AMP involved in the destruction of bacteria, fungi and parasites. Nevertheless, only a few publications suggest a potential role of Nkl in the antiviral immune response, probably because an Nkl/granulysin ortholog gene has not been identified in the mouse or rat model species (58). Nk-lysin was found to be overexpressed after viral challenge in chicken (59, 60) and fish (61–63), and some evidence of its antiviral effect were reported (63–65). In this work, we observed that turbot *nkl* transcription is positively correlated with VHSV resistance. We conducted numerous *in vivo* and *in vitro* experiments both using an expression plasmid encoding Nkl (pMCV1.4-*nkl*) or the recombinant protein, but we did not obtain any significant difference in mortality or viral replication after an infection with VHSV. Therefore, we were not able to provide any evidence of direct antiviral activity. This is probably due to the fact that Nkl does not work as a typical AMP in the antiviral context, and it needs to be confined into the autophagolysosomes.

Although Nkl was always assumed to be present only in the cytolytic granules of NK-cells and CTLs (33, 34), this does not seem to be the case in teleost fish. Flow-cytometry analysis and confocal microscopy unexpectedly revealed that this peptide is expressed in turbot erythrocytes. In these cells, fluorescence was confined to one or to a few vesicles in the cytoplasm, which were also LysoSensor- and LC3B-positive. TEM images revealed that these structures are double-membrane bound compartments containing cellular organelles. These data confirmed that these vesicles correspond to autophagolysosomes. Is Nkl contributing to the self-degradation of the RBCs organelles as a part of the maturation process? Nk-lysin, as a member of the SAPLIP family, possesses membrane-perturbing ability (66), and therefore it could be intervening in the disruption of the biological membranes of cellular organelles. Moreover, other members of the SAPLIP family, such as the saposin peptides, are important in the correct resolution of autophagy due to their specific roles in the degradation of the glycosphingolipids present in biological membranes (67, 68). Except for mammals, vertebrate erythrocytes are nucleated. During the first embryonic stages in mammals a transient population of nucleated erythrocytes is present in the bloodstream, whereas in late fetal periods and postnatal life only enucleated RBCs are found in the circulating blood (69). This early presence of nucleated RBCs in embryos could represent an evolutionary reminiscence from non-mammalian vertebrates.

Autophagy is also an important cellular mechanism in the clearance of viruses (23–29) and represents a key process in the defense against Rhabdoviruses in mammals and fish (30–32). Nevertheless, in erythrocytes, autophagy has been always relegated to function in the degradation of cellular organelles during the maturation of these cells. In turbot, it was previously reported that VHSV, which is a highly pathogenic virus affecting turbot (70), can infect primary cultures of blood leukocytes (71). However, in this work we observed that VHSV has also the ability to infect and proliferate in erythrocytes during an *in vitro* infection. Inhibition of the autophagic process using CQ increased the viral replication, indicating that autophagy is an important antiviral mechanism in RBCs. Nevertheless, we should take into consideration that CQ also affects the antigen presentation *via* major histocompatibility complex (MHC) class I and class II (72, 73), although the relevance of this process during an *in vitro* infection in erythrocytes is probably negligible.

Currently, autophagy is an emerging field of study and, although the main components of the autophagy machinery are well known, only a few articles have examined the relation between AMPs and autophagy. These publications are mainly focused on the generation of neo-AMPs with bactericidal activity from cytosolic proteins in the mycobacteria-containing autophagolysosomes (74–76). Ren et al. (77) found that a peptide derived from the human cathelicidin could activate caspase-independent apoptosis and autophagy in colon cancer cells. Interestingly, it was also reported that human granulysin induces the cleavage of Atg5 in the complex formed with Atg12 although no effects in autophagy were observed (78).

It seems that Rhabdovirus-induced autophagy is mediated, at least in part, by the viral G glycoprotein (30, 32, 79). In this work, we observed that a DNA vaccine encoding the VHSV G glycoprotein (50) induces long-lasting effects on the levels of Nkl. Because autophagy is an important mechanism in the generation of innate memory cells (80, 81), the persistence of changes in the levels of Nkl 1 month after vaccination could indicate that this peptide is also important in the "trained immunity" or innate memory. Therefore, autophagy could be the process linking the levels of Nkl and the resistance to VHSV. Nevertheless, more investigation is needed to fully understand how Nkl levels determine the antiviral state of teleost fish and if erythrocytes actively contribute to this process.

Two major conclusions can be extracted from our work: (A) Teleost erythrocytes have an active antiviral role that is mediated by autophagy and (B) this is the first time that an AMP, Nkl, is associated with the autophagic mechanism. The results that support these conclusions have been unrecognized until now: the correlation between *nkl* expression and the resistance to viruses; the presence of Nkl in nucleated fish erythrocytes; the relation between Nkl and autophagy; and the implication of autophagy in the antiviral response of erythrocytes. Taken together, these results can change the preconceived ideas that we have about vertebrate immunity, opening new doors to combat diseases that, in the case of fish, seriously affect the aquaculture industry and focusing more on cells and processes previously not considered.

#### ETHICS STATEMENT

All the experimental procedures were reviewed and approved by the CSIC National Committee on Bioethics (Protocol no. ES360570202001/16/FUN01/PAT.05/tipoE/BNG).

#### AUTHOR CONTRIBUTIONS

AE, AF, and BN conceived and designed the study, and analyzed the data. PP, AR, and PD-R performed the experimental procedures and data analyses. PP, AR, AF, and BN wrote the manuscript. All the authors reviewed the manuscript.

## ACKNOWLEDGMENTS

Antonio Figueras and Beatriz Novoa especially want to dedicate this article to Prof. Amparo Estepa, one of the coauthors of this article, who unfortunately passed away at a very young age before it was published. We want to remember her intelligence, vitality and charm and above all we thank her for being our friend.

## FUNDING

This work was funded by the Spanish Ministerio de Economía y Competitividad (grant number AGL2014-53190-REDT and grant number AGL2014-51773-C3), Agencia Estatal Consejo Superior de Investigaciones Científicas (CSIC) (grant number PIE201230E057), and Consellería de Economía, Emprego e Industria (GAIN), Xunta de Galicia (grant number IN607B 2016/12).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/article/10.3389/fimmu.2017.01458/ full#supplementary-material.

#### Pereiro et al. Antiviral Role of Erythrocytes

#### REFERENCES


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

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

# Microrna mir-214 inhibits snakehead Vesiculovirus replication by Promoting iFn-**α** expression *via* Targeting host adenosine 5**′**-Monophosphate-activated Protein Kinase

*Chi Zhang1,2, Shuangshuang Feng1 , Wenting Zhang3 , Nan Chen1 , Abeer M. Hegazy1,4, Wenjie Chen2 , Xueqin Liu1 , Lijuan Zhao2 , Jun Li 2,5,6, Li Lin1,2,6\* and Jiagang Tu1,7\**

#### *Edited by:*

*Brian Dixon, University of Waterloo, Canada*

#### *Reviewed by:*

*Alison Kell, University of Washington, United States Yong-An Zhang, Institute of Hydrobiology (CAS), China*

#### *\*Correspondence:*

*Li Lin linli@mail.hzau.edu.cn; Jiagang Tu tujiagang@mail.hzau.edu.cn*

#### *Specialty section:*

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

*Received: 18 August 2017 Accepted: 28 November 2017 Published: 11 December 2017*

#### *Citation:*

*Zhang C, Feng S, Zhang W, Chen N, Hegazy AM, Chen W, Liu X, Zhao L, Li J, Lin L and Tu J (2017) MicroRNA miR-214 Inhibits Snakehead Vesiculovirus Replication by Promoting IFN-α Expression via Targeting Host Adenosine 5*′*-Monophosphate-Activated Protein Kinase. Front. Immunol. 8:1775. doi: 10.3389/fimmu.2017.01775*

*1Department of Aquatic Animal Medicine, College of Fisheries, Huazhong Agricultural University, Wuhan, China, 2Guangzhou Key Laboratory of Aquatic Animal Diseases and Waterfowl Breeding, Guangdong Provincial Key Laboratory of Waterfowl Healthy Breeding, College of Animal Sciences and Technology, Zhongkai University of Agriculture and Engineering, Guangzhou, China, 3Key Laboratory of Prevention and Control Agents for Animal Bacteriosis, Institute of Animal Husbandry and Veterinary, Hubei Academy of Agricultural Sciences, Wuhan, China, 4Central Laboratory for Environmental Quality Monitoring (CLEQM), National Water Research Center (NWRC), Cairo, Egypt, 5School of Biological Sciences, Lake Superior State University, Sault Ste. Marie, MI, United States, 6 Laboratory for Marine Fisheries Science and Food Production Processes, Qingdao National Laboratory for Marine Science and Technology, Qingdao, China, 7 Hubei Engineering Technology Research Center for Aquatic Animal Diseases Control and Prevention, Huazhong Agricultural University, Wuhan, China*

Background: Snakehead vesiculovirus (SHVV), a new rhabdovirus isolated from diseased hybrid snakehead, has emerged as an important pathogen during the past few years in China with great economical losses in snakehead fish cultures. However, little is known about the mechanism of its pathogenicity. MicroRNAs are small noncoding RNAs that posttranscriptionally modulate gene expression and have been indicated to regulate almost all cellular processes. Our previous study has revealed that miR-214 was downregulated upon SHVV infection.

results: The overexpression of miR-214 in striped snakehead (SSN-1) cells inhibited SHVV replication and promoted IFN-α expression, while miR-214 inhibitor facilitated SHVV replication and reduced IFN-α expression. These findings suggested that miR-214 negatively regulated SHVV replication probably through positively regulating IFN-α expression. Further investigation revealed that adenosine 5′-monophosphate-activated protein kinase (AMPK) was a target gene of miR-214. Knockdown of AMPK by siRNA inhibited SHVV replication and promoted IFN-α expression, suggesting that cellular AMPK positively regulated SHVV replication and negatively regulated IFN-α expression. Moreover, we found that siAMPK-mediated inhibition of SHVV replication could be partially restored by miR-214 inhibitor, indicating that miR-214 inhibited SHVV replication at least partially *via* targeting AMPK.

conclusion: The findings of this study complemented our early study, and provide insights for the mechanism of SHVV pathogenicity. SHVV infection downregulated

**20**

miR-214, and in turn, the downregulated miR-214 increased the expression of its target gene AMPK, which promoted SHVV replication via reducing IFN-α expression. It can therefore assume that cellular circumstance with low level of miR-214 is beneficial for SHVV replication and that SHVV evades host antiviral innate immunity through decreasing IFN-α expression *via* regulating cellular miR-214 expression.

Keywords: snakehead vesiculovirus, microRNA, miR-214, interferon, replication, adenosine 5**′**-monophosphateactivated protein kinase

#### INTRODUCTION

MicroRNAs (miRNAs) are a class of small (~22 nt) noncoding RNAs that posttranscriptionally degrade and/or suppress translation of target mRNAs through base pairing between the "seed sequences" (2–8 nt at the 5′ end) of miRNAs and the target transcripts (1–5). Host miRNAs, typically binding to the 3′ untranslated regions (UTRs) of target transcripts (5–10), have been reported to play important roles in the regulation of virus replication (1, 5, 7, 10, 11). Moreover, the regulatory roles of miRNAs in virus replication were even utilized by viruses to promote their replication (6, 9). Therefore, understanding the roles of miRNAs in virus infection is helpful for understanding the mechanisms of virus pathogenesis.

Snakehead vesiculovirus (SHVV) is a fish rhabdovirus isolated from diseased hybrid snakehead in 2014 in China (12). It has caused high mortality to cultured snakehead fish these yeas. Up to now, the study about the mechanism of its pathogenicity is limited. SHVV belongs to the genus *Perhabdovirus*, family *Rhabdoviridae* (13). Its genome is an ~11 kb negative-sense RNA molecule that encodes five proteins: nucleoprotein (N), phosphoprotein (P), matrix protein (M), glycoprotein (G), and RNA-dependent RNA polymerase protein (L) (12). Our previous study has revealed that SHVV infection downregulated miR-214 (14), and in turn, miR-214 could inhibit SHVV production by targeting viral N and P (15). However, it is unclear whether miR-214 can regulate SHVV replication *via* targeting host factors that are required for SHVV replication. MiR-214 has recently been observed to be upregulated by Vibro harveyi, and the upregulated miR-214 inhibited the production of inflammatory cytokines by targeting host myd88 (16). Consequently, miR-214 played important roles in regulating pathogens infection.

Adenosine 5′-monophosphate-activated protein kinase (AMPK) is a heterotrimeric serine/threonine kinase (17), which is considered as pivotal regulator of host cellular metabolism *via* sensing cellular energy status (18). When the energy levels in cells decrease, AMPK is activated through phosphorylation by an upstream kinase (18). Activated AMPK thereby downregulates anabolic processes that consume ATP and upregulates catabolic processes that synthesize ATP (18). Given the role of sensing changes of cellular energy status, it is not surprising that AMPK plays an important role in virus infection (19). However, growing evidences have revealed that viruses can modulate the activity of AMPK, and in turn, AMPK affects virus infection by regulating cellular autophagy or innate immunity (20, 21). Here, we reported that AMPK was a target gene of miR-214, negative regulator of IFN-α expression, and positive regulator of SHVV replication. Moreover, we determined that miR-214 could inhibit SHVV replication by promoting IFN-α expression *via* reducing AMPK expression. This study provided information for understanding the molecular mechanism of SHVV pathogenicity and a potential antiviral strategy against SHVV infection.

#### MATERIALS AND METHODS

#### Cells and Viruses

Striped snakehead (SSN)-1 cells were maintained at 25°C in minimum essential medium (MEM) (HyClone, USA) supplemented with 10% heat-inactivated fetal bovine serum (FBS) (Gibco, New Zealand), penicillin (100 µg/ml), and streptomycin (100 µg/ml). SHVV was isolated from diseased hybrid snakehead fish and stored at −80°C.

#### Reagents and Antibodies

The miR-214 mimic, miR-214 inhibitor, negative control (NC) mimic, and NC inhibitor were purchased from GenePharma (Shanghai, China). Their sequences were previously described (15). Two siRNAs for AMPK (accession number: MF989224) were synthesized from GenePharma (Shanghai, China). The sequences of the first one were: 5′-CCUCCAGUAUCAAGAUCUUTT-3′ (forward) and 5′-AAGAUCUUGAUACUGGAGGTT-3′ (reverse); the sequences of the second one were: 5′-GGACACGCCCAUU AUUAAATT-3′ (forward) and 5′-UUUAAUAAUGGGCGUGU CCTT-3′ (reverse).

The antibodies against G protein of SHVV and AMPK were produced and stored in our laboratory. The antibody against β-actin was purchased from Bioss Biotechnology Co., LTD. (Beijing, China). The secondary antibody donkey anti-rabbit IgG antibody was purchased from Gene Co., LTD. (Shanghai, China).

#### Plasmids

The luciferase reporter plasmid pmirGLO-AMPK was constructed by amplifying the miR-214 target sequence (~200 nt) in the 3′ UTR of AMPK and cloning into vector pmirGLO with primers listed in **Table 1**. The plasmids pmirGLO-AMPK-MUT1 and pmirGLO-AMPK-MUT2 were generated by PCR mediated mutations into plasmid pmirGLO-AMPK using primers listed in **Table 1**. The expression plasmid p3XFLAG-CMV-14-AMPK was constructed by amplifying the open reading frame of AMPK gene and cloning into vector p3XFLAG-CMV-14 using primers listed in **Table 1**.

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


#### Transfection

The mimics, inhibitors, or plasmids were incubated with TransIntroTM EL Transfection Reagent (TransGen Biotech, China) in 500 μl Opti-MEM medium (Invitrogen, USA) for 30 min at room temperature. The incubated samples were then put onto the SSN-1 cells. At 6 h post of transfection, the medium was replaced by 1 ml of MEM and continued incubation at 25°C.

#### Dual-Luciferase Reporter Assay

The dual-luciferase reporter assay was performed as described previously (15). In brief, SSN-1 cells were co-transfected with NC mimic, miR-214 mimic, NC inhibitor, or miR-214 inhibitor, together with the luciferase reporter plasmids using TransIntroTM EL Transfection Reagent (TransGen Biotech, China). At 24 h post of transfection, the *Renilla* and firefly luciferase activities were measured, and the data were expressed as relative firefly luciferase activity normalized to *Renilla* luciferase activity.

#### Virus Infection and Titration

Virus infection and titration experiments were performed as previously described (15). In brief, SSN-1 cells were incubated with SHVV for 2 h, the inoculum was then removed and the cells were washed twice with PBS followed by adding MEM medium with 5% FBS. At 24 h post of infection (poi), the supernatants were collected for virus titration by 50% tissue culture infectious dose (TCID50), and the cells were harvested for the detection of viral mRNA or host miR-214 by qRT-PCR with primers listed in **Table 1**.

### Quantitative RT-PCR of Viral mRNA, Host IFN-**α** mRNA, and miR-214

Total RNAs were extracted from cells with TRIzol reagent (Invitrogen) according to manufacturer's instructions. The detection of viral mRNA, host IFN-α (accession number: MF989225) mRNA, and miR-214 was performed by qRT-PCR as previously described (15). Two sets of data were normalized using the 2−ΔΔCt method. For the detection of viral mRNA, data was normalized to the level of β-actin in each sample, while for the miR-214 detection, the expression level of miR-214 was calculated after normalization to 5S rRNA.

#### Western Blotting

Western blotting was performed as previously described (15). In brief, the extracted proteins were transferred onto a nitrocellulose membrane (Biosharp, China), which were blocked with 5% skim milk in tris-buffered saline with tween 20 (TBST) at 4°C overnight, followed by incubation with the primary antibody of SHVV protein (1:1,000) or β-actin (1:1,000) for 2 h at room temperature. The membranes were then washed three times with TBST and then incubated with IRDye 800CW conjugated donkey anti-rabbit antibody (1:10,000) for 1 h at room temperature. The signal intensity was then determined using Odyssey CLx (LI-COR, USA).

#### Statistical Analysis

All statistical analyses were performed using Graphpad Prism 5.0 (GraphPad Software, CA, USA). The statistical significance of the data was determined by Student's *t* test, and *P* < 0.05 was considered statistically significant. For data sets in which multiple comparisons were being made, the Student's *t*-test was corrected by using false discovery rate.

## RESULTS

#### The Effect of miR-214 on the Transcription, Translation, and Production of SHVV

MiR-214 has been indicated to inhibit the replication of several human and mammalian viruses, including human cytomegalovirus (HCMV), murine cytomegalovirus (MCMV), and herpes simplex virus 1 (HSV-1) (22). In the same vein, our previous study has suggested that miR-214 inhibited SHVV replication (15). In this study, we further evaluated the effect of miR-214 on the transcription, translation, and production of SHVV at different time point poi. SSN-1 cells were transfected with miR-214 mimic, NC mimic, miR-214 inhibitor, or NC inhibitor, followed by SHVV infection. At 3, 12, and 24 h poi, the cells and supernatants were collected in order to detect viral G mRNA, G protein, and viral titer by qRT-PCR, western blot, and TCID50, respectively. As shown in **Figure 1**, viral G mRNA expression was not significantly altered at 3 h poi. However, at 12 and 24 h poi, overexpression of miR-214 significantly reduced, whereas miR-214 inhibitor increased, G mRNA level (**Figures 1A,D**). The expression of viral G protein was under detection at 3 and 12 h poi. At 24 h poi, it was apparent from **Figures 1B,E** that G protein expression was decreased by about 50% or increased to about 2.5-fold when the cells were transfected with miR-214 mimic or miR-214 inhibitor, respectively. Similar to G mRNA and G protein, the viral titers were reduced by transfection of miR-214 mimic and increased by transfection of miR-214 inhibitor (**Figures 1C,F**). Taken together, these findings demonstrate that miR-214 inhibits SHVV replication.

## miR-214 Promotes IFN-**α** Expression

Our previous study has revealed that miR-214 promoted IFN-α expression during SHVV infection by targeting SHVV P protein, an IFN-α antagonist (15). In this study, we further investigated the role of miR-214 in IFN-α expression at different time point post of SHVV infection. SSN-1 cells were transfected with miR-214 mimic, NC mimic, miR-214 inhibitor, or NC inhibitor, followed by SHVV infection. At 3, 12, and 24 h poi, IFN-α mRNA was detected using qRT-PCR. As shown in **Figures 2A,B**, overexpression of miR-214 significantly increased, whereas miR-214 inhibitor reduced, IFN-α mRNA at 12 and 24 h poi. Moreover, we found that overexpression of miR-214 inhibited SHVV replication and increased IFN-α mRNA in a dose-dependent manner (**Figure 2C**). In addition, we found that overexpression of miR-214 promoted poly (I:C)-induced IFN-α mRNA (Figure S1 in Supplementary Material). Our data suggest that the promotion of IFN-α expression by miR-214 could be the cause of its inhibition of SHVV replication.

## miR-214 Targets the 3**′** UTR of AMPK mRNA

In addition to targeting viral P gene, it's speculated that miR-214-mediated inhibition of SHVV replication and promotion of IFN-α expression could also be caused by targeting host genes. To determine host target genes of miR-214, high throughput transcriptomic sequencing of SSN-1 cells transfected with miR-214 mimic or NC mimic has been performed. The results showed that overexpression of miR-214 resulted in 1,301 upregulated genes and 1,613 downregulated genes (data not shown). Based on association possibility with virus replication, six downregulated genes were selected for further validation

FIGURE 1 | The effect of miR-214 on the transcription, translation, and production of snakehead vesiculovirus (SHVV). Striped snakehead (SSN)-1 cells were transfected with negative control (NC) mimic, miR-214 mimic, NC inhibitor, or miR-214 inhibitor, followed by SHVV infection. The cells and supernatants were harvested at various time points. (A,D) G mRNA in SSN-1 cells was measured using qRT-PCR. β-actin was used as the internal control. (B,E) G Protein was determined by western blot. β-actin was used as the internal control. The integrated optical densities of the protein bands were measured using Image-Pro Plus 6.0. The values of the G protein bands were normalized to that of β-actin. The values of the G protein bands in cells transfected with NC mimic or inhibitor were set as 100, respectively. (C,F) The SHVV titers in the supernatants were measured using TCID50. All the data are representative of at least two independent experiments, with each determination performed in triplicate (mean ± SD). The \* and \*\*, respectively, indicate statistically significant differences (\**P* < 0.05; \*\**P* < 0.01).

as the internal control. All the data are representative of at least two independent experiments, with each determination performed in triplicate (mean ± SD). The \* and \*\*, respectively, indicate statistically significant differences (\**P* < 0.05; \*\**P* < 0.01).

TABLE 2 | Six downregulated genes from transcriptomic sequencing data.


*AMPK, adenosine 5*′*-monophosphate (AMP)-activated protein kinase; STAT, signal transducers and activators of transcription; MMP9, matrix metallopeptidase 9; EIF4E, eukaryotic translation initiation factor 4E; NLK, nemo-like kinase; STAT3, signal transducer and activator of transcription 3.*

using qRT-PCR as follows (**Table 2**): signal transducers and activators of transcription (STAT), matrix metallopeptidase 9 (MMP9), eukaryotic translation initiation factor 4E (EIF4E), nemo Like Kinase (NLK), signal transducer and activator of transcription 3 (STAT3), and AMPK. Among them, AMPK was the most downregulated gene (**Figure 3A**). To further determine the effect of miR-214 on AMPK expression, SSN-1 cells were transfected with 60 and 120 pmol of miR-214 mimic, NC mimic, miR-214 inhibitor, or NC inhibitor, followed by the detection of cellular AMPK protein using specific AMPK antibody (Figure S2 in Supplementary Material). We found that transfection of 120 pmol of miR-214 mimic significantly reduced, while miR-214 inhibitor increased, cellular AMPK protein expression (**Figure 3B**), suggesting that AMPK was probably a target gene of miR-214.

Using Miranda software, two putative binding sites of miR-214 were identified at the 3′ UTR of AMPK mRNA (**Figure 3C**). To further confirm whether AMPK was the target gene of miR-214, we first constructed a dual-luciferase reporter plasmid pmirGLO-AMPK containing the wild-type sequence of the 3′ UTR of AMPK. Based on the plasmid pmirGLO-AMPK, we generated two mutant plasmids pmirGLO-AMPK-MUT1 and pmirGLO-AMPK-MUT2, in which the miR-214-targeted sequences were mutated (**Figure 3C**). These plasmids were subsequently transfected into SSN-1 cells with miR-214 mimic, NC mimic, miR-214 inhibitor, or NC inhibitor, respectively. Significant reduction in luciferase activity was observed in cells co-transfected with miR-214 mimic and the plasmid with wildtype AMPK 3′ UTR, whereas significantly increased luciferase activity was detected when transfected with miR-214 inhibitor (**Figure 3D**). However, the luciferase activity was not significantly altered when miR-214 mimic or inhibitor was co-transfected with the mutant plasmids harboring miR-214 seed-region mutated sequences (**Figure 3D**). These results indicate that AMPK is a target gene of miR-214.

#### SHVV Infection Upregulates AMPK

To study the effect of SHVV infection on AMPK, SSN-1 cells were infected with SHVV and the cells were harvested at 3, 12, and 24 h poi. The G mRNA, AMPK mRNA, and miR-214 were determined by qRT-PCR. Along with the increase of G mRNA, miR-214 was steadily decreased at 12 and 24 h poi (**Figure 4**). This result was consistent with our previous study, in which SHVV infection downregulated miR-214 (14). In addition, AMPK mRNA was significantly increased at 12 and 24 h poi (**Figure 4**), suggesting that SHVV infection upregulated AMPK possibly *via* downregulating miR-214.

#### Knockdown of AMPK Inhibits SHVV Replication and Promotes IFN-**α** Expression

In order to understand the role of AMPK in SHVV infection, SSN-1 cells were transfected with siAMPK or siNC, followed by SHVV infection. Transfection of siAMPK significantly reduced the mRNA and protein levels of AMPK compared to that transfected with siNC (**Figures 5A,B**). The effect of AMPK on SHVV replication was further evaluated by detecting viral G mRNA, G protein, and viral titer at 24 h poi. As shown in **Figures 5C,D**, the viral G mRNA and protein were reduced to less than 10% in siAMPK transfected cells than in siNC transfected cells. Similarly, the viral titer was decreased more than 10-fold in siAMPK group than in siNC group (**Figure 5E**). It can thus be suggested that knockdown of AMPK, similar to the overexpression of miR-214, inhibited SHVV replication.

In addition to affecting SHVV replication, siAMPK increased IFN-α mRNA about 10-fold (Figure S3A in Supplementary Material). To further confirm the effects of AMPK on IFN-α expression, SSN-1 cells were transfected with plasmid p3XFLAG-CMV-14 or p3XFLAG-CMV-14-AMPK, followed by SHVV infection. At 24 h poi, the cellular IFN-α mRNA was measured by qRT-PCR. The results showed that overexpression of AMPK reduced IFN-α mRNA level (Figure S3B,C in Supplementary Material). These findings suggest that AMPK negatively regulates IFN-α expression.

### Suppression of Cellular miR-214 Can Restore siAMPK-Mediated Inhibition of SHVV Replication

In order to figure out whether miR-214-mediated inhibition of SHVV replication was caused by targeting AMPK, SSN-1 cells were transfected with siNC, siAMPK, or siAMPK with miR-214 inhibitor, followed by SHVV infection. The cells and supernatants were collected at 24 h poi. The viral G protein and viral titer were determined. As shown in **Figure 6A**, siAMPK reduced G protein level to about 1% compared to that in siNC group. However, addition of miR-214 inhibitor restored the G protein level to 15%. Similarly, the viral titer was significantly decreased by siAMPK, which was partially restored by the transfection with miR-214 inhibitor (**Figure 6B**). Overall, these findings indicate that miR-214 inhibits SHVV replication at least partially due to its targeting AMPK.

## DISCUSSION

Host miRNA has emerged as both responsive factor and modulator of virus infection. In detail, host miRNA expression is commonly altered in response to virus infection, and the altered miRNA in turn modulate virus infection (1, 6, 9–11, 22). Santhakumar et al. have revealed that miR-214 was downregulated upon several human and mammalian viruses infection, and in turn, miR-214 inhibited the replication of these viruses, suggesting that miR-214 acted as a broad antiviral miRNA (22). Our previous study has revealed that miR-214 was downregulated upon a fish rhabdovirus SHVV infection (14), and overexpression of miR-214 inhibited SHVV replication *via* targeting N and P genes of SHVV (15). As the genomes of these miR-214-inhibited viruses share little sequence similarity, it is speculated that miR-214 might target

5′-monophosphate-activated protein kinase (AMPK). Striped snakehead (SSN)-1 cells were infected with SHVV and the cells were harvested at 3, 12, and 24 h poi. The G mRNA, AMPK mRNA, and miR-214 were determined by qRT-PCR, β-actin was used as the internal control for G mRNA and AMPK mRNA, while 5S rRNA was used as the internal control for miR-214. All the data are representative of at least two independent experiments, with each determination performed in triplicate (mean ± SD). The \* and \*\*, respectively, indicate statistically significant differences (\**P* < 0.05; \*\**P* < 0.01).

host factors that are required by multiple viruses. Thereby, the aim of this study is to identify miR-214-targeted host factors and the related antiviral mechanism.

The mechanisms of miRNA-mediated regulation of virus replication have attracted much attention these years. Growing evidences have demonstrated that miRNAs could inhibit virus replication by targeting host factors that were critically important for virus replication (23–32). For example, host eukaryotic translation elongation factor 1A1 (EEF1A1) can interact with NS3 and NS5 proteins of Japanese encephalitis (JEV) to form a complex that is essential for JEV replication, miR-33a can target EEF1A1 and reduce its expression, thus suppressing JEV replication (31). In addition, miRNAs can also target host factors that positively or negatively regulate type I interferon expression or the following signaling (6, 9, 33–44). In the current study, we found that miR-214, the important cancer development regulator (45, 46), inhibited SHVV replication by regulating host IFN-α expression (15) (**Figure 2**). Many miRNAs have been identified as type I interferon regulators, including miR-373 (6), miR-466l (34), miR-155 (36), miR-15b (38), miR-526a (39), miR-223 (41), and miR-146a (42–44). Here, miR-214 was identified as a novel type I interferon regulator.

FIGURE 5 | Knockdown of adenosine 5′-monophosphate-activated protein kinase (AMPK) inhibits snakehead vesiculovirus (SHVV) replication. (A,B) Striped snakehead (SSN)-1 cells were transfected with siNC or siAMPK, the AMPK mRNA (A) and protein (B) in SSN-1 cells was measured at 24 h post of transfection using qRT-PCR or western blot. β-actin was used as the internal control. The integrated optical densities of the protein bands were measured using Image-Pro Plus 6.0. The values of the AMPK protein bands were normalized to that of β-actin. The values of the AMPK protein band in cells transfected with siNC was set as 100. (C–E) SSN-1 cells were transfected with siNC or siAMPK, followed by SHVV infection. G mRNA (C) in SSN-1 cells was measured using qRT-PCR at 24 h poi. β-actin was used as the internal control. G Protein (D) was determined by western blot at 24 h poi. The integrated optical densities of the protein bands were measured using Image-Pro Plus 6.0. The values of the G protein bands were normalized to that of β-actin. The values of the G protein bands in cells transfected with siNC was set as 100. The SHVV titers (E) in the supernatants were measured using TCID50 at 24 h poi. All the data are representative of at least two independent experiments, with each determination performed in triplicate (mean ± SD). The \* and \*\*, respectively, indicate statistically significant differences (\**P* < 0.05; \*\**P* < 0.01).

FIGURE 6 | Suppression of cellular miR-214 can restore siAMPK-mediated inhibition of snakehead vesiculovirus (SHVV) replication. Striped snakehead (SSN)-1 cells were transfected with siNC, siAMPK, or siAMPK with miR-214 inhibitor, followed by SHVV infection. (A) The G protein was determined by western blot at 24 h poi. β-actin was used as the internal control. The integrated optical densities of the protein bands were measured using Image-Pro Plus 6.0. The values of the G protein bands were normalized to that of β-actin. The values of the G protein bands in cells transfected with siNC was set as 100. (B) The SHVV titers in the supernatants were measured using TCID50 at 24 h poi. All the data are representative of at least two independent experiments, with each determination performed in triplicate (mean ± SD). The \* and \*\*, respectively, indicate statistically significant differences (\**P* < 0.05; \*\**P* < 0.01).

Adenosine 5′-monophosphate-activated protein kinase has been extensively studied as a pivotal regulator of cellular energy metabolism (47). Recent studies have revealed that AMPK was involved in the regulation of virus replication (18). Inhibition of AMPK severely attenuated HCMV replication, suggesting that AMPK was required for HCMV replication (48, 49). In the current study, knockdown of AMPK inhibited SHVV replication, suggesting that AMPK was beneficial for SHVV replication (**Figure 5**). However, activation of AMPK has been reported to restrict hepatitis B virus (HBV) production, suggesting that AMPK was disadvantageous for HBV replication (50). Therefore, AMPK played different roles in different viruses infection. Although our study demonstrated that miR-214-mediated inhibition of SHVV replication was at least partially due to targeting it's target gene AMPK, the broad antiviral property of miR-214 was probably not caused by its targeting AMPK because AMPK not only promoted but also restricted some viruses replication (50). In the current study, in addition to AMPK, five other host factors, including STAT, MMP9, EIF4E, NLK, and STAT3, have also been identified as potential target genes of miR-214 (**Figure 3A**). Therefore, identifying the host target gene of miR-214 that was responsible for the broad antiviral property of miR-214 needed to be investigated further.

Recently, AMPK has been indicated to regulate type I interferon expression. Inhibition of AMPK was observed to suppress IFN-β induction (51). In the current study, knockdown of AMPK promoted, whereas overexpression of AMPK inhibited, IFN-α expression (Figure S3 in Supplementary Material). Moreover, our study revealed that knockdown of AMPK promoted poly (I:C)-induced IFN-α expression (Figure S4 in Supplementary Material). These findings suggested that AMPK could regulate type I interferon expression. Our previous study has indicated that viral P protein of SHVV inhibited IFN-α expression, and miR-214 could target the P gene and thus suppressed P-mediated inhibition of IFN-α expression (15). As AMPK was also identified as a target gene of miR-214, It's speculated that miR-214-mediated regulation of IFN-α expression might also be due to targeting AMPK. Taken together, our studies suggested that miR-214 promoted IFN-α expression by targeting not only viral P gene but also host AMPK gene. Despite these promising results, further studies are needed to investigate how P and AMPK regulated host IFN-α expression.

## AUTHOR CONTRIBUTIONS

JT and LL designed the research. CZ, SF, WZ, NC, AH, WC, and LZ performed the experiments, contributed to the data collection and statistical analysis. JT, LL, XL, and JL finalized the paper writing.

## ACKNOWLEDGMENTS

This work was jointly supported by National Natural Science Foundation of China (31602195, 31572657), Natural Science Foundation of Hubei province (2016CFB233), and China Postdoctoral Science Foundation (2016M600604).

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/articles/10.3389/fimmu.2017.01775/ full#supplementary-material.

## REFERENCES


inhibition of autophagy. *FEBS Open Bio* (2017) 7(6):811–20. doi:10.1002/ 2211-5463.12221


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

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

# Understanding Viral dsrna-Mediated innate immune responses at the cellular level Using a rainbow Trout Model

*Sarah J. Poynter <sup>1</sup> and Stephanie J. DeWitte-Orr <sup>2</sup> \**

*1Department of Biology, University of Waterloo, Waterloo, ON, Canada, 2Department of Health Sciences, Wilfrid Laurier University, Waterloo, ON, Canada*

#### *Edited by:*

*Monica Imarai, Universidad de Santiago de Chile, Chile*

#### *Reviewed by:*

*Bertrand Collet, Institut National de la Recherche Agronomique (INRA), France Alison Kell, University of Washington, United States*

*\*Correspondence:*

*Stephanie J. DeWitte-Orr sdewitteorr@wlu.ca*

#### *Specialty section:*

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

*Received: 02 February 2018 Accepted: 05 April 2018 Published: 23 April 2018*

#### *Citation:*

*Poynter SJ and DeWitte-Orr SJ (2018) Understanding Viral dsRNA-Mediated Innate Immune Responses at the Cellular Level Using a Rainbow Trout Model. Front. Immunol. 9:829. doi: 10.3389/fimmu.2018.00829*

Viruses across genome types produce long dsRNA molecules during replication [viral (v-) dsRNA]. dsRNA is a potent signaling molecule and inducer of type I interferon (IFN), leading to the production of interferon-stimulated genes (ISGs), and a protective antiviral state within the cell. Research on dsRNA-induced immune responses has relied heavily on a commercially available, and biologically irrelevant dsRNA, polyinosinic:polycytidylic acid (poly I:C). Alternatively, dsRNA can be produced by *in vitro* transcription (ivt-) dsRNA, with a defined sequence and length. We hypothesized that ivt-dsRNA, containing legitimate viral sequence and length, would be a more appropriate proxy for v-dsRNA, compared with poly I:C. This is the first study to investigate the effects of v-dsRNA on the innate antiviral response and to compare v-dsRNA to ivt-dsRNAinduced responses in fish cells, specifically rainbow trout. Previously, class A scavenger receptors (SR-As) were found to be surface receptors for poly I:C in rainbow trout cells. In this study, ivt-dsRNA binding was blocked by poly I:C and v-dsRNA, as well as SR-A competitive ligands, suggesting all three dsRNA molecules are recognized by SR-As. Downstream innate antiviral effects were determined by measuring IFN and ISG transcript levels using qRT-PCR and antiviral assays. Similar to what has been shown previously with ivt-dsRNA, v-dsRNA was able to induce IFN and ISG transcript production between 3 and 24 h, and its effects were length dependent (i.e., longer v-dsRNA produced a stronger response). Interestingly, when v-dsRNA and ivt-dsRNA were length and sequence matched both molecules induced statistically similar IFN and ISG transcript levels, which resulted in similar antiviral states against two aquatic viruses. To pursue sequence effects further, three ivt-dsRNA molecules of the same length but different sequences (including host and viral sequences) were tested for their ability to induce IFN/ISG transcripts and an antiviral state. All three induced responses similarly. This study is the first of its kind to look at the effects v-dsRNA in fish cells as well as to compare ivt-dsRNA to v-dsRNA, and suggests that ivt-dsRNA may be a good surrogate for v-dsRNA in the study of dsRNA-induced responses and potential future antiviral therapies.

Keywords: innate immunity, double-stranded RNA, type I interferon, antiviral, rainbow trout

## INTRODUCTION

Long dsRNA molecules (>40 bp) are immunomodulatory nucleic acids that can induce interferon (IFN) and an antiviral state across vertebrate species (1, 2). dsRNA is an important pathogenassociated molecule pattern (PAMP) produced by viruses; as demonstrated by the sheer number and diversity of receptors in the cytoplasm, endosome, and surface used by host cells to detect dsRNA (3). In vertebrates, dsRNA is a potent inducer of the type I IFN response, which produces a broad-spectrum antiviral state. Viral dsRNA is sensed in the cytoplasm by a wide range of receptors, such as retinoic acid-inducible gene-I (RIG-I), RNA helicase A/DHX9, and melanoma differentiation-associated protein 5 (MDA5), toll-like receptor 3 (TLR3) in the endosome, and class A scavenger receptors (SR-As) on the cell's surface (1, 3, 4). When dsRNA is sensed in a cell it triggers a signaling cascade through various adaptor proteins, such as interferon-β promoter stimulator 1 or TIR-domain-containing adapter-inducing interferon-β (TRIF), activating transcription factors, such as interferonregulator factor (IRF)3/7, and causing the production of IFN. IFNs are secreted from the cell and signal in an autocrine and paracrine fashion *via* their cognate receptor, interferon-α/β receptor, to initiate the Janus kinase and signal transducer and activator of transcription signaling pathway resulting in the expression of a group of genes containing an interferon-sensitive response element known cumulatively as interferon-stimulated genes [ISGs (5, 6)]. In rainbow trout (*Oncorhynchus mykiss*), there is an incredibly large repertoire of type I IFNs, as many as 22 members have been identified. Type I IFNs in fish have an intricate naming system, teleost type I IFNs are subdivided into group 1 or group 2 based on the number of cysteine residues and further into subgroups (a–f) based on phylogenetic analysis (7). These naming systems have no relation to the alpha or beta system used in mammals (7). IFN1 is a type I IFN, belonging to group I and subgroup a (7) and is used in this study as a representative transcript indicative of type I IFN expression. IFNs in fish stimulate expression of a panel of ISGs including molecules from the IFN signaling pathway such as IRF3/7 and antiviral effectors that limit viral infection, including Myxovirus resistance 1 (Mx1), viral hemorrhagic septicemia virus (VHSV)-induced gene (vig)-1, vig-3, and vig-4 (6, 8, 9). vig-4 is a VHSV-induced gene; the deduced protein contains tetratricopeptide repeat motifs and shows similarities to the ISG56/IFIT1 family of ISGs (9, 10). This study chose vig-4 as a representative transcript indicative of ISG expression because it has been used as a representative ISG in previous studies (11–14) and is upregulated more strongly than other ISGs, such as Mx1 (15). The dsRNA-induced accumulation of ISG proteins produces a protective antiviral state in rainbow trout cells against various viruses (15, 16).

DsRNA sensed by a host cell in a natural system would be produced by viral infection, described here as v-dsRNA. As viruses replicate, dsRNA is produced as a by-product of replication, a genomic fragment, or transcribed from DNA by host proteins (17–20). The potent immune stimulatory nature of dsRNA makes it a candidate molecule for antiviral therapies and vaccine adjuvants, as well as for use in type I IFN studies. Unfortunately, v-dsRNA is difficult to collect from viruses in quantities useful for experimental scenarios and likely impossible for industrial applications. In the late 1960s, a synthetic form of dsRNA, polyinosinic:polycytidylic acid (poly I:C), was identified as a potent IFN-inducer and was considered a "viral mimic" (21–23). Poly I:C is clearly different from dsRNA produced by a virus; poly I:C lacks sequence variation and natural structures, contains a range of lengths and one strand contains exclusively a modified inosine nucleotide (24). Owing to these differences, poly I:C is not sensed the same nor does it induce responses exactly the same as *in vitro* transcribed (ivt-) dsRNA (24–27). In plasmacytoid dendritic cells only ivt-dsRNA was able to stimulate IFN-α production, poly I:C did not (24). In rainbow trout cells, ivt-dsRNA induced a faster, stronger IFN1 and IFN2 response compared with poly I:C even when poly I:C was of much longer lengths (15). In addition, in mice, poly I:C is recognized by MDA5 whereas ivt-dsRNA and v-dsRNA activated RIG-I (25, 28). For TLR3, human TLR3 but not teleost TLR3 has a much higher affinity for poly I:C than ivt-dsRNA (25, 29).

The current state of research regarding responses to dsRNA largely relies on the use of poly I:C; however, studies of individual receptors are shifting toward ivt-dsRNA, likely for the ease of controlling length (25, 27, 30). Length has been shown to influence the magnitude of immune response in cells and dsRNA receptor types show length requirements and specificities (15, 25, 27, 29). For example, longer dsRNA molecules have been shown to induce a strong IFN response (27), and RIG-I has been shown to sense dsRNA molecules under 1,000 bp in length, while MDA5 senses lengths greater than 1,000 bp (28, 30). The effect of dsRNA sequence on IFN induction is also poorly studied; there were no detectable sequence motifs for MDA5 activation identified from vaccinia virus-derived dsRNA (31). One example of sequence dependence is the cytoplasmic dsRNA receptor oligoadenylate synthetase (OAS) that requires a 4 bp-specific motif for binding (32). Few studies have looked at the antiviral response induced by v-dsRNA, and any studies that do exist have all used mammalian models. Specifically, v-dsRNA derived from encephalomyocarditis virus, vaccinia virus, and reovirus induced potent IFN responses in Vero, HeLa, and murine embryonic fibroblasts, respectively (28, 31). ivt-dsRNA is an alternative source of synthetic dsRNA that retains some features of v-dsRNA and can be produced on a larger scale. To the best of our knowledge, there have been no studies directly comparing a v-dsRNA and an ivt-dsRNA molecule of matched length and sequence, therefore it is unknown if ivt-dsRNA induces a comparable immune response to v-dsRNA.

Rainbow trout were used in this study as a model fish species for their importance in aquaculture and the existing knowledge base of the rainbow trout type I IFN and antiviral response (15, 33). In rainbow trout cell lines, ivt-dsRNA or poly I:C induces type I IFN and an antiviral state and similarly whole rainbow trout pretreated with poly I:C also showed decreased susceptibility to a fish virus (15, 34). Three aquatic viruses were used in this study: chum salmon reovirus (CSV), which has a segmented dsRNA genome that consists of 11 segments between 3,947 and 783 bp (35), infectious pancreatic necrosis virus (IPNV), which is a non-enveloped *Aquabirnavirus* with a bisegmented dsRNA genome (36) and VHSV, which is an enveloped rhabdovirus with a negative-sense ssRNA genome (37). All three viruses readily infect rainbow trout cells, including RTG-2, a rainbow trout gonadal cell line (38, 39). Mammalian reoviruses have previously been used as a source of dsRNA of different lengths and total genomic dsRNA has been used as an immune stimulus (14, 28, 40). VHSV and IPNV both represent important disease in the fish aquaculture industry and ecology as they have wide host ranges and can cause large die-offs of fish (36, 37, 41).

This study compares IFN-mediated responses induced by three forms of dsRNA: v-dsRNA, ivt-dsRNA, and poly I:C. The dsRNA is this study was delivered extracellularly. In a viral infection, the dsRNA would be intracellular during its production and released to the extracellular space in the case of cell lysis where it could be recognized by neighboring cells. In the case of a dsRNAbased therapy the dsRNA would be delivered to the cell surface and not to the cytoplasm. Because the dsRNA was delivered extracellularly, the surface receptor for these dsRNA molecules was investigated. The ability of v-dsRNA from aquatic viruses to induce IFNs, ISGs, and an antiviral response was quantified. Length- and sequence-matched v-dsRNA and ivt-dsRNA were compared with poly I:C for their ability to induce IFNs, ISGs, and mount an antiviral state against IPNV and VHSV. The results from this study provide valuable insight with regards to how fish cells respond to viral dsRNA as opposed to poly I:C, with applications for novel dsRNA-based therapies.

## MATERIALS AND METHODS

#### Cell Culture and Virus Propagation

Two rainbow trout cell lines were used in this study to measure dsRNA-mediated responses: RTG-2, derived from rainbow trout gonad (42) and RTgutGC, derived from rainbow trout intestine (43). Epithelioma papulosum cyprinid (EPC) and Chinook salmon embryonic cell line (CHSE-214) were used for viral propagation. All cell lines used in this study were obtained from N. Bols (University of Waterloo, Waterloo, ON, Canada). All cell lines were grown in 75 cm2 plastic tissue culture flasks

Table 1 | Primers used for *in vitro* transcription of dsRNA and qRT-PCR.

(BD Falcon, Bedford, MA, USA) at room temperature in Leibovitz's L-15 media (HyClone, Logan, UT, USA) supplemented with 10% v/v fetal bovine serum (FBS; Fisher Scientific, Fair Lawn, NJ, USA) and 1% v/v penicillin/streptomycin (P/S) (10 mg/mL streptomycin and 10,000 U/mL penicillin; Fisher Scientific). All dsRNA treatments were delivered extracellularly by addition of dsRNA to cell culture media.

#### Virus Propagation

Viral hemorrhagic septicemia virus-IVb (strain U13653) was propagated on monolayers of EPC (44) cells; CSV and IPNV were propagated on CHSE-214; and all viruses were propagated at 17°C (45). Virus containing media [L-15 with 2% v/v FBS (Fisher Scientific)] was collected 4–7 days post-infection, filtered through a 0.45 µm filter (Nalgene, Rochester, NY, USA) and kept frozen at −80°C. The 50% tissue culture infective dose (TCID50)/mL values were estimated according to the Reed and Muench method (46). The origin of the viruses used in this study has been described previously (47).

#### Polyinosinic:Polycytidylic Acid

High-molecular weight (HMW) poly I:C (InvivoGen, San Diego, CA, USA) stocks were prepared at 1 mg/mL, and low-molecular weight (LMW) poly I:C (InvivoGen) stocks were prepared at 10 mg/mL, both were diluted in phosphate-buffered saline (PBS) (HyClone), and aliquots were stored at −20°C. Before use, aliquots were heated to 55°C for 15 min and then allowed to cool to room temperature for 20 min.

#### Synthesis of *In Vitro* Molecules

*In vitro* transcribed dsRNA molecules were produced as previously described using the MegaScript RNAi kit [Fisher Scientific (15)]. All primers used for synthesizing dsRNA, **Table 1**, had the T7 promoter sequence added to the 5′ end,


*Forward (F) and reverse (R) primer sequences (5*′*–3*′*), length of product (bp), annealing temperature (Ta), and accession number of source sequence are provided. For dsRNA, the length of the final dRNA product is provided.*

TAATACGACTCACTATAGGGAG. The following molecules were prepared: the full-length CSV segment 6 (2,052 bp); a 300 bp internal segment of CSV segment 1 (CSVseg1); a 200 bp segment of rainbow trout GAPDH, rainbow trout Mx3, and an internal segment of the VHSV G gene (3,851–4,048 bp) that has been previously described (15). Where needed, nucleotide distribution was calculated using Genomatix: DNA sequence toolbox.1

#### Extraction of v-dsRNA and Isolation of Segments

To extract the CSV viral genome, CSV was propagated as described earlier; after complete destruction of the monolayer, the viruscontaining media was collected, and cell debris was pelleted by centrifugation at 3,000 × *g* for 5 min. The supernatant was mixed with poly ethylene glycol BioUltra 8000 to a final concentration of 10% w/v (Sigma-Aldrich, St. Louis, MO, USA; catalog number: 89510) and sodium chloride to a final concentration of 0.6% w/v and mixed on a Corning LSE Digital Microplate Shaker at 1,400 RPM 4°C overnight [Corning, Tewksbury, MA, USA (50)]. The solution was then centrifuged at 17,000 × *g* in a Sorvall Legend Micro 17 microcentrifuge for 20 min (Fisher Scientific). The resulting pellet was resuspended in 100 µL PBS overnight at 4°C, and the RNA was extracted using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) as per the manufacturers' instructions. To isolate segments of the genome, total genomic CSV dsRNA was run on a 1% agarose TAE gel containing GelGreen (1:10,000 dilution; Biotium Inc., Fremont, CA, USA). The band of interest was cut out and purified using the QIAquick gel extraction kit (Qiagen, Hilden, Germany) and quantified using a NanoDrop Lite Spectrophotometer (Thermo Fisher Scientific). The v-dsRNA preparation was validated as pure dsRNA by two methods: acridine orange stained gel to confirm only red stained nucleic acids were visible, and differential nuclease degradation using RNase A and RNaseIII to ensure degradation by RNaseIII alone. These methods for visualizing dsRNA have been described previously (51). 20 ng of resulting products was re-quantified by gel densitometry to confirm the NanoDrop readings. The matched ivt-dsRNA molecule was also gel purified and quantified as above for consistency.

#### Labeling of dsRNA

The CSV seg1 ivt-dsRNA was labeled using the Ulysis Alexa Fluor 546 Nucleic Acid Labeling Kit (Fisher Scientific) as previously described (15). Unbound fluorophores were removed using a BioRad p30 spin column (BioRad, Hercules, CA, USA).

## Competitive Binding Assay

RTgutGC cells were seeded at a density of 1 × 105 cells/well on glass coverslips in a 12-well tissue culture plate. After attaching overnight cells were pretreated with 200 µL of L-15 containing LMW, HMW, poly I, or poly C at 100 µg/mL, total native dsRNA at 40 µg/mL, or control L-15 alone for 30 min. After this incubation, 1.25 µg (5 µg/mL) of labeled CSVseg1 dsRNA was added to the well with 50 µg/mL of DEAE-dextran. DEAE-dextran is used in this context as a method of ensuring dsRNA reaches the cells in quantities sufficient for detection by fluorescence microscopy (1). Six hours posttreatment cells were washed 3× with PBS, fixed for 10 min with 10% neutral buffered formalin (Fisher Scientific), nuclei were counterstained with 10 µg/mL 4′,6-diamidino-2-phenylindole (DAPI; Fisher Scientific) and mounted on coverslips with SlowFade Gold mounting medium (Fisher Scientific) for visualization. Images were captured using an inverted fluorescence microscope (Nikon Eclipse TiE with Qi1 camera) and analyzed using Nikon NIS elements. A blinded third-party researcher performed the fluorescence quantification measurements. Three images were captured of each treatment for each independent replicate. Alexa Fluor 546 fluorescence intensity was measured by automatic selection of the area surrounding DAPI stained nuclei, total number of cells within the image were counted, and the intensity/cell was calculated. Intensity is presented as percentage of the *in vitro* only control with no pretreatment.

#### Cell Protection Assays

RTG-2 cells were seeded at 1 × 104 cells/well in 96-well plates and allowed to attach overnight in regular growth media. For each condition, there were triplicate wells, and for each assay there was an untreated/uninfected control and an untreated/infected control. Cells were treated with the indicated concentration of dsRNA for 3 or 6 h at 20°C in 50 µL of media containing L-15, 1% P/S, and 2% FBS. When dsRNA concentration was 0.01 nM, 6 h was selected to ensure a productive antiviral state before infection; 3 h was found to be insufficient (data not shown). After dsRNA pretreatment, the virus was added directly to the media to produce the necessary multiplicity of infection (MOI) (10 for VHSV, 0.2 or 0.02 for IPNV). Cells were incubated at 17°C for 4–7 days until desired accumulation of cytopathic effect occurred. Cells were rinsed 2× with PBS and a 5% v/v alamarBlue solution in PBS was added (Invitrogen). Cells were incubated at room temperature for 1 h in the dark, and fluorescence was measured using a Synergy HT plate reader (BioTek, Winooski, VT, USA). Data are presented as percentage of control, uninfected cells.

#### RNA Extraction, cDNA Synthesis, and qRT-PCR

5 × 105 cells/well of RTG-2 cells were plated in a 6-well plate and allowed to attach overnight. Media were removed, and cells were treated with dsRNA at the noted concentration in full-growth media, or with media only. After the indicated incubation time, media were removed, and RNA extracted with TRIzol as per the manufacturers' instructions (Thermo Fisher Scientific). Possible contaminating DNA was removed using the TURBO DNA-free kit (Fisher Scientific). cDNA was synthesized using iSCRIPT (BioRad), and 1 µg of RNA. qRT-PCR has been previously described for the primers used in this study (15). A no-reverse transcriptase (NRT) control was used to identify genomic contamination, a peak earlier than 35 cycles in the NRT control was considered excessive DNA contamination, no samples showed a peak earlier than this threshold. Melting curve analysis was completed (65–95°C with a read every 5 s) to determine primer

<sup>1</sup>http://www.genomatix.de/cgi-bin/tools.

specificity, only single peaks were produced. In addition to the NRT control, primers were designed to span introns for both β-actin and IFN1; as such genomic contamination would produce a much larger product, which would be evident in the melting curve analysis. This technique could not be performed for vig-4, as it is predicted to be intron-less. In addition, there are two predicted copies of vig-4 in the rainbow trout genome, the primers used in this study would amplify both variants; however, all sequencing performed identified only the published vig-4 sequence (NM\_001124333.1).

#### Statistical Analysis

All data presented were derived from at least three independent experiments. Data were graphed and statistically analyzed using GraphPad Prism version 7.00 for Windows, GraphPad Software, La Jolla, CA, USA.2 Statistical analyses were completed using a one- or two-way ANOVA with Tukey multiple comparison test,

2www.graphpad.com.

alpha = 0.05, *P* value < 0.05 considered significant; qRT-PCR data were log2 transformed before analysis. Data points labelled with the same letter do not have a statistically different average.

## RESULTS

#### dsRNA From Different Sources Bind to the Same Surface Receptor

RTgutGC cells, previously shown to bind poly I:C by SR-As (49), were treated with fluorescently labeled ivt-dsRNA, and punctate cell-associated binding was observed (**Figure 1A**). Cells were then pretreated with an excess of poly I:C, either high molecular weight (HMW) or low molecular weight (LMW), total CSV v-dsRNA, poly I or poly C (**Figures 1A,B**). Both sizes of poly I:C and the total v-dsRNA significantly blocked binding of the labeled ivt-dsRNA (**Figure 1B**). poly I is a competitive ligand for SR-A binding whereas poly C is a molecule of similar chemical structure that is non-competitive for SR-As. poly I significantly blocked ivt-dsRNA binding while poly C did not (**Figure 1B**).

considered significant.

## v-dsRNA Elicits a Length-Dependent, Protective Type I IFN Response

RTG-2 cells were treated with 10 ng/mL of total CSV v-dsRNA, and significant induction of an IFN (IFN1) and an ISG (vig-4) was observed at the transcript level using qRT-PCR at 24 h posttreatment (**Figure 2A**). IFN1 transcript production peaked at 3 h and was no longer measurable by 24 h and vig-4 production peaked at 24 h (**Figure 2A**). Previously, concentrations lower than 10 ng/mL of ivt-dsRNA have been shown to induce IFN and ISG transcripts in RTG-2 cells (15). To test the ability of v-dsRNA to establish a functional antiviral state, RTG-2 cells were pretreated for 3 h with 10 ng/mL of total v-dsRNA and then infected with VHSV (MOI: 10) or IPNV (MOI: 0.2; **Figure 2B**). There was significant protection observed against both viruses when cell viability was measured using alamarBlue.

To look at the effects of v-dsRNA length on the immune response, portions of the CSV genome grouped as large, medium, and small segments were isolated, and RTG-2 cells were treated with 0.05 nM of each molecule for 3 h (**Figures 3A,B**). The concentration used was chosen because in previous studies this concentration demonstrated length effect differences between ivt-dsRNA molecules. In addition, molar amounts were used to measure the effect of length as opposed to number of molecules (15). It should be noted that these v-dsRNA molecules are different sequences as well as lengths. There was a length-dependent response seen in the RTG-2 production of IFN1 and vig-4 transcripts, with the long segments inducing significantly more IFN1 and vig-4 transcripts than the short segments. For both IFN1 and vig-4, the medium segment fell in the middle range of long and short, and there was a significant difference between long and medium-induced vig-4 production. A corresponding antiviral assay against VHSV and IPNV was completed similarly to the total v-dsRNA assay described earlier (**Figure 3C**). While both long and short v-dsRNA protected cells significantly from viral infection, there were no significant differences between the two lengths.

#### Length- and Sequence-Matched v-dsRNA and ivt-dsRNA Induce IFNs and an Antiviral State Similarly

To compare the IFN response induced by v-dsRNA and ivt-dsRNA directly, v-dsRNA from the CSV genome (segment 6, 2,052 bp) was isolated, and a matching ivt-dsRNA molecule of the same length and sequence was synthesized to match (**Figure 4A**). Cells were stimulated with 0.01 nM of v-dsRNA or ivt-dsRNA, respectively. This is a lower molar amount of dsRNA than other assays in this study, due to the difficulty of isolating usable quantities

into long (2,690–3,947 bp), medium (2,052–2,242 bp), or short (1,317–1,395 bp) segment groups. (A) 20 ng of the isolated segments was run on a 1% agarose gel stained with ethidium bromide for visualization. (B) RTG-2 cells were treated with 0.05 nM dsRNA for 3 h, and IFN1 and vig-4 transcripts were measured by qRT-PCR, normalized to β-actin, and presented as values relative to an unstimulated control. (C) RTG-2 cells were pretreated with 0.05 nM dsRNA for 3 h and then infected with VHSV at a multiplicity of infection (MOI) of 10 or IPNV at an MOI of 0.2. After 4–7 days a fluorescent indicator dye, alamarBlue, was used to measure cell viability, and data are presented as the percentage of an untreated, uninfected control. Data represent at least three independent replicates and were analyzed statistically by one-way ANOVA, alpha = 0.05; a *P* value < 0.05 considered significant.

of a single v-dsRNA segment. There was significant induction of both IFN1 and vig-4 transcripts by both molecules, and no significant difference was observed between the ivt-dsRNA and v-dsRNA molecules (**Figure 4B**). This trend continued with the antiviral assays, where both ivt-dsRNA and v-dsRNA protected cells against IPNV and VHSV-induced cell death, but there were no significant differences in the protection between the two (**Figure 4C**).

For comparison purposes, a 0.01 nM HMW poly I:C (average length 3,000 bp) control was included for both qRT-PCR and antiviral assays. There was significantly less induction of IFN1 and vig-4 transcript production from the HMW poly I:C at the 3 h time point (**Figure 4B**). With VHSV infection, all three dsRNA molecules protected similarly; however, with IPNV at an MOI of 0.2 ivt-dsRNA and v-dsRNA provided complete protection (with viability levels similar to uninfected control cells) but poly I:C did not. By an MOI of 0.02, all three dsRNA molecules provided the same amount of protection (**Figure 4C**).

## Molecules of the Same Length but Different Sequence Had Similar Effects

To test the effect of sequence on the magnitude of the immune response by RTG-2 cells, three length-matched ivt-dsRNAs were made with different sequences, including different source material (two rainbow trout genes and one viral gene), and different nucleotide composition (**Figure 5A**). IFN1 and vig-4 transcripts were significantly induced following treatment with 0.05 nM dsRNA for 3 h, but no significant differences between the three sequences were detected (**Figure 5B**). This matched the antiviral assays for both VHSV and IPNV, where protection was seen but no significant differences between molecules (**Figure 5C**).

## DISCUSSION

While poly I:C is an attractive dsRNA to use in studies of the dsRNA-mediated immune response, it is not biologically relevant and as such is either too potent (21) or not potent enough (52) when modulating the innate immune response. We hypothesized that ivt-dsRNA would be a better molecule for modulating innate antiviral immune responses, as it has natural sequence variation and defined length. The goal of this study was to test whether ivtdsRNA was a comparable surrogate for v-dsRNA and to better understand v-dsRNA and ivt-dsRNA-mediated innate antiviral immune responses using a rainbow trout cell model. This is the first study to investigate v-dsRNA and its effects in an aquatic vertebrate system. The use of the two difference cell lines helped provide answers to two separate questions, (1) whether the molecules are bound by a common scavenger receptor and (2) whether the molecules were inducing similar antiviral responses. To answer the first point, RTgutGC was used, as this cell line has been shown previously to express functional SR-As, surface receptors

for dsRNA (49). RTG-2 was used to answer the second point as it has previously been shown to respond to low concentrations of dsRNA and is permissive to infection by both aquatic viruses used to test the establishment of an antiviral state (15, 16, 38).

The dsRNA in this study was delivered extracellularly; they were added directly to the media instead of being transfected into the cell. In a viral infection, dsRNA would be intracellular during its production and released to the extracellular space during a lytic infection where it could be recognized by neighboring cells. In the case of a dsRNA-based therapy, the dsRNA would be delivered to the cell surface and not to the cytoplasm. If dsRNA-based therapies are viable, they need to be able to bind to the cell's surface, thus the mechanism of recognition of ivt-dsRNA compared with v-dsRNA at the cell surface is important to discern. Poly I:C (both LMW and HMW) and v-dsRNA effectively blocked the binding of labeled ivt-dsRNA molecules in RTG-2 cells, suggesting that all three dsRNA molecules do indeed bind SR-As on the cell surface. This is consistent with previous studies in mammals that have shown ivt-dsRNA, poly I:C, and viral dsRNA are bound by SR-As (1, 53–56). A study in mouse splenocytes found HMW poly I:C but not LMW was bound by SR-As, and therefore both LMW and HMW poly I:C were tested in this study for their ability to block binding (54). In this study, no differences were observed between LMW and HMW poly I:C's ability to block ivt-dsRNA binding, suggesting SR-A binding capabilities to dsRNA in mice differ from rainbow trout.

Next, the ability of v-dsRNA to induce the IFN pathway and antiviral response was tested. v-dsRNA induced IFN1 and vig-4 transcripts with similar kinetics to what was previously reported with ivt-dsRNA (49) in that IFN1 transcript peaked at 3 h and the ISG vig-4 accumulated over time. This antiviral response effectively reduced virus-induced cell death for two important fish pathogens, VHSV-IVb and IPNV. This is the first study in fish to demonstrate v-dsRNA as a stimulant of type I IFN and an inducer of an antiviral state against aquatic viruses.

This is not, however, the first study to use the reovirus genome as a source of v-dsRNA. Previously, the total reovirus genome as well as isolated segments has been used as immune-inducing molecules in mammalian cells (14, 28, 40). One of the first studies using v-dsRNA as an immunostimulant was in 1967 when an isolated reovirus genome was injected into rabbits and induced IFN (57). The cellular response to the reovirus genome has been demonstrated in mouse embryonic fibroblasts and human embryonic kidney cells (HEK293), both cell types producing IFN-β following v-dsRNA treatment (14, 28). Reoviruses are used for these types of studies due to the relative abundance of dsRNA produced during infection. Current studies are underway to optimize methods to isolate dsRNA from viruses of other genome types.

Figure 5 | ivt*-*dsRNA of the same length and different sequence induced similar levels of IFN1 and vig-4 transcripts and protective antiviral states against viral hemorrhagic septicemia virus (VHSV) and infectious pancreatic necrosis virus (IPNV) in RTG-2 cells. Three *in vitro* transcribed dsRNA molecules of the same length, but different source sequences were used to test the effects of sequence on innate immune response. (A) Genomatix was used to calculate the nucleotide distribution within the three sequences, and the percentage of each nucleotide is shown. (B) RTG-2 cells were treated with 0.05 nM dsRNA for 6 h, and IFN1 and vig-4 transcripts were measured by qRT-PCR, normalized to β-actin, and presented as values relative to an unstimulated control. (C) RTG-2 cells were pretreated with 0.05 nM dsRNA for 3 h and then infected with VHSV at a multiplicity of infection (MOI) of 10 or IPNV at an MOI of 0.2. After 4–7 days, a fluorescent indicator dye, alamarBlue, was used to measure cell viability, and data are presented as the percentage of an untreated, uninfected control. Data represent three independent replicates and were analyzed statistically by one-way ANOVA, alpha = 0.05; a *P* value < 0.05 considered significant.

Similar to ivt-dsRNA, v-dsRNA induced IFN pathways, and in some cases an antiviral state, in a length-dependent manner. The long v-dsRNA molecule induced more IFN1 than short and long v-dsRNA induced more vig-4 transcript than medium and short. At the level of an antiviral state, there were no differences between long and short v-dsRNA induced protection for either VHSV or IPNV infection; however, long v-dsRNA did appear to provide protection similar to control in the VHSV infection model while the short molecule did not (**Figure 3C**). These differences are indeed subtle, which is unsurprising as the long segment is only 2× the length of the short, whereas previous studies of length have shown an effect with molecules with a 6× and 10× difference in length in RTG-2 cells (15). It can then be hypothesized if shorter v-dsRNA molecules were used a greater difference in antiviral assays would be observed. It should be noted that while the molecules are of different lengths they are also of different sequences, and this could be a confounding variable in this study. While future studies may address this issue by digesting native dsRNA genome fragments or ligating fragments together, this study aimed to focus on native molecules that were as unmodified as possible. Even so, evidence from this study and previous work suggest sequence does not play a role in levels of IFN induction (31, 58).

Next, a direct comparison between length- and sequencematched v-dsRNA and ivt-dsRNA was performed, and no significant differences were observed between the immune gene transcript induction and antiviral state established by v-dsRNA and the matched ivt-dsRNA. HMW poly I:C, however, even with a much longer average length, was not as effective at inducing IFNs or ISGs and required a lower MOI of IPNV to protect cells similarly to v- or ivt-dsRNA. These data suggest that indeed ivtdsRNA but not poly I:C could be used as a surrogate for v-dsRNA. Indeed, previous work has shown that poly I:C induced IFN and ISG kinetics differently from ivt-dsRNA in rainbow trout cells (15). These differences continue between fish and humans with regard to poly I:C. Human TLR3 responds very strongly to poly I:C, whereas this was not the case for a fish (*Takifugu rubripes*) TLR3, which responded most strongly to one length of ivt-dsRNA compared with poly IC or other lengths of ivt-dsRNA (29). *In vivo*, poly I:C has had mixed effects as an antiviral therapy, being an effective protection mechanism against red-spotted grouper necrosis virus in sevenband grouper (59), but in zebrafish (*Danio rerio*) infected with VHSV poly I:C was able to delay symptoms but only prevented mortality in 5% of fish (60). Clearly, there is room for increasing the efficacy of dsRNA-based antiviral therapies in fish past the protection that poly IC can provide.

It should be noted that there are differences between v-dsRNA generated *in situ* with the v-dsRNA isolated for this study. One difference is the lack of dsRNA-associated proteins in the extracted v-dsRNA. Viruses use many mechanisms to hide dsRNA, one method is the production of proteins that bind dsRNA and effectively hide it from host receptors (61). Host cells also have a number of dsRNA-binding proteins that could modify dsRNA availability and potency (21). Extracting v-dsRNA through a column or phenol/chloroform extraction would remove these proteins. These dsRNA-associated proteins may have effects on how the cell senses and responds to v-dsRNA. Other modifications that could be influencing the cellular response to dsRNA that would survive extraction include methylation and 5′-tri- or diphosphates (14, 62). Mammalian reoviruses and poly I:C both have a 5′-diphosphate whereas ivt-dsRNA have 5′-triphosphate termini; both termini are able to activate RIG-I; however, the implication of this difference in rainbow trout is harder to elucidate due to the lack of RIG-I in this fish species (14, 63–65). Interestingly long ivt-dsRNA (>200 bp), which would lack a 5′-triphosphate was still able to activate RIG-I in murine embryonic fibroblasts (28). The cap status of the dsRNA molecules may also influence the host response; this has been best studied in terms of RIG-I and MDA5. The addition of an M7G cap partially reduced the RIG-I stimulatory properties of dsRNA, whereas 2′-*O*-methylation entirely abrogated RIG-I activation (66). In terms of MDA5, mutant viruses lacking 2′-*O*-methyltransferase induced higher MDA5-dependent type I IFN expression (67). In this study, v-dsRNA likely has a cap, as reoviruses put a 5′ cap on the positive strand of the genomic segments; in comparison, the *ivt*-dsRNA and poly I:C would not have a m7G cap (14). Kato et al. (28) found that capped *ivt*-dsRNA was still able to induce IFN-β production, unfortunately these molecules were not compared to uncapped molecules to determine if this had any positive or negative effects on stimulatory properties. Studies are currently underway to explore the role of different cap modifications on the host innate immune response.

Based on the assumption that ivt-dsRNA can act as a surrogate for v-dsRNA in inducing IFN at the cellular level, and the v-dsRNA vs. ivt-dsRNA comparison performed used molecules with the same sequence, it follows to test whether nucleotide sequence makes a difference. From the results of a group of three ivt-dsRNA containing host or viral sequences it appears as though sequence does not significantly affect ivt-dsRNA's IFN inducing capabilities. This is congruent with the current literature that suggests dsRNA sequence is not a major influence on dsRNA receptor binding, as is length and structure (31, 58). Although there is evidence that RIG-I ligands generally have a uridine- or adenosine-rich ribonucleotide sequence, and OAS has a 4 bp-specific sequence

#### REFERENCES


motif, it is unclear if other receptors have any preference or sequence requirements, or whether there are any sequence preferences for fish dsRNA sensors (32, 68). Sequence may play a role in an antiviral state when RNAi is considered; however, in this system, there is an overwhelming IFN-mediated response that likely masks any RNAi effects (69).

Overall, this study sheds light on the dsRNA-mediated immune response in rainbow trout cells. The findings suggest that v-dsRNA produced by an aquatic reovirus is an IFN-inducing PAMP. There were no significant differences between a v-dsRNA molecule and length- and sequence-matched ivt-dsRNA molecules with regards to inducing IFN and ISGs and antiviral state, suggesting that ivt-dsRNA may be useful in studies of IFN and in future antiviral therapies for fish. This study sought to perform functional studies where possible, using antiviral assays to explore the biological relevancy of transcriptional quantification results. These findings contribute to a better understanding of the differences between dsRNA from different sources, which can help facilitate the production of more biologically relevant dsRNA-based therapies.

#### AUTHOR CONTRIBUTIONS

SP performed experiments and contributed to experimental design and writing of manuscript. SD-O contributed to experimental design, funding of project, and writing of manuscript.

#### ACKNOWLEDGMENTS

The authors would like to thank Brian Dixon and the members of the DeWitte-Orr and Dixon laboratories for valuable discussions, Julia Pacosz for her assistance with blind measurements, Shreya Jalali for assistance with preliminary sequence studies, and Niels Bols for providing cells, viruses, and grandfatherly support.

#### FUNDING

Funding for this project was provided to SD-O by an NSERC Discovery grant.


and LGP2. *Proc Natl Acad Sci U S A* (2007) 104(2):582–7. doi:10.1073/pnas. 0606699104


disease in Atlantic salmon. *Dev Comp Immunol* (2011) 35(11):1116–27. doi:10.1016/j.dci.2011.03.016


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

*Copyright © 2018 Poynter and DeWitte-Orr. 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.*

#### *Edited by:*

*Brian Dixon, University of Waterloo, Canada*

#### *Reviewed by:*

*Magdalena Chadzin´ska, Jagiellonian University, Poland Paul Craig, University of Waterloo, Canada Mathilakath Vijayan, University of Calgary, Canada*

#### *\*Correspondence:*

*Felipe E. Reyes-López felipe.reyes@uab.cat; Lluís Tort lluis.tort@uab.cat*

*† 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: 22 December 2017 Accepted: 06 April 2018 Published: 02 May 2018*

#### *Citation:*

*Khansari AR, Balasch JC, Vallejos-Vidal E, Parra D, Reyes-López FE and Tort L (2018) Comparative Immuneand Stress-related Transcript Response Induced by Air Exposure and Vibrio anguillarum Bacterin in Rainbow Trout (Oncorhynchus mykiss) and Gilthead Seabream (Sparus aurata) Mucosal Surfaces. Front. Immunol. 9:856. doi: 10.3389/fimmu.2018.00856*

*Ali Reza Khansari† , Joan Carles Balasch† , Eva Vallejos-Vidal, David Parra, Felipe E. Reyes-López\* and Lluís Tort\**

*Department of Cell Biology, Physiology and Immunology, Universitat Autònoma de Barcelona, Bellaterra, Spain*

Fish have to face various environmental challenges that may compromise the efficacy of the immune response in mucosal surfaces. Since the effect of acute stress on mucosal barriers in fish has still not been fully elucidated, we aimed to compare the short-term mucosal stress and immune transcriptomic responses in a freshwater (rainbow trout, *Oncorhynchus mykiss*) and a marine fish (gilthead seabream, *Sparus aurata*) to bacterial immersion (*Vibrio anguillarum* bacterin vaccine) and air exposure stress in skin, gills, and intestine. Air exposure and combined (vaccine + air) stressors exposure were found to be inducers of the cortisol secretion in plasma and skin mucus on both species in a timedependent manner, while *V. anguillarum* bacterin exposure induced cortisol release in trout skin mucus only. This was coincident with a marked differential increase in transcriptomic patterns of stress- and immune-related gene expression profiles. Particularly in seabream skin, the expression of cytokines was markedly enhanced, whereas in gills the response was mainly suppressed. In rainbow trout gut, both air exposure and vaccine stimulated the transcriptomic response, whereas in seabream, stress and immune responses were mainly induced by air exposure. Therefore, our comparative survey on the transcriptomic mucosal responses demonstrates that skin and gut were generally more reactive in both species. However, the upregulation of immune transcripts was more pronounced in gills and gut of vaccinated trout, whereas seabream appeared to be more stress-prone and less responsive to *V. anguillarum* bacterin in gills and gut. When fish were subjected to both treatments no definite pattern was observed. Overall, the results indicate that (1) the immune response was not homogeneous among mucosae (2), it was greatly influenced by the specific traits of each stressor in each surface and (3) was highly species-specific, probably as a result of the adaptive life story of each species to the microbial load and environmental characteristics of their respective natural habitats.

Keywords: mucosal immunity, cortisol, gene expression, skin, gills, gut

#### INTRODUCTION

Fish are living in a microbial-loaded environment involving an intense interaction of their mucosal surfaces with microbiota and therefore various immune responses in these surfaces. The diversity of the potential environmental or imposed stressors (i.e., changes in temperature, photoperiod, pH, oxygen saturation, population density, pathogen load, and virulence) biases the efficacy and time course of the mucosal immune responses in a species-specific manner (1–4). Thus, defensive responses in fish show great interspecific diversity and agglutinate the mucosalassociated structures in a common mucosal immunity framework (5, 6). When interacting with mucosal interfaces, exogenous bacteria and viruses skew the immune responsiveness depending on each surface. Pathogens, such as *Vibrio anguillarum*, are able to adhere preferentially to fish integument (7), modifying the thickness, quality, and secretory pattern of skin immune defenses which, in turn, vary depending on the interspecific susceptibility to diseases, pathogen virulence, and environmental toxicity (8).

The mucosal-associated lymphoid tissues (MALTs) comprise the skin-associated lymphoid tissue (SALT), the gill-associated lymphoid tissue (GIALT), the gut-associated lymphoid tissue (GALT), and the recently described nasopharynx-associated lymphoid tissue. Common features of these MALTs resemble those of mammals and include the following: (i) a copious mucus layer that actively barriers pathogen adherence and agglutinates (9); (ii) secreted antimicrobial proteins (such as lysozyme, lectins, complement proteins, histones, and defensins), antibodies (*igm* and *igt*/*z* isotypes), immune mediators (cytokines and chemokines), and enzymatic disruptors (mainly proteases, peroxidases, and phosphatases); and (iii) interposed myeloid and lymphoid immune cells (including mast cells, dendritic-like cells, macrophages, neutrophils, and B and T lymphocyte families), natural killer cells (NK/ NCC-like), epithelial phagocytic cells, and immune-associated cells such as thrombocytes and erythrocytes (6, 8, 10–12).

Skin-associated lymphoid tissue is the largest and most functionally diverse mucosal surface. Teleost skin is an extensive metabolically active non-keratinized multilayered integument that produces a complex glycoprotein-based mucus cuticle (8, 13) and also plays a crucial role in communication, sensory perception, locomotion, respiration, and osmoregulation (14, 15). Fish SALT harbors a more diverse repertoire of innate humoral components than the mammalian one, including bacteriolytic molecules such as lysozyme, complement components, lectins, proteolytic enzymes, C-reactive protein, interferons, and immunoglobulins (16). A whole cast of resident leukocyte families complete the immunological properties of SALT (8). The GIALT system is highly similar to that of skin and consists of interposed mucus-secreting cells, antimicrobial peptides, and resident leukocyte populations (11). Fish GALT lacks the mammalian Peyer's patches, but presents intraepithelial lymphocytes that include T cells and some B cells located among epithelial cells. M-cell analogs and dendritic-like cells have also been described, as well as plasma cells, granulocytes, macrophages, and neuroendocrine cells inhabiting the epithelium or distributed in the lamina propria (17, 18). All these plethora of resources enable fish to defend from external agents, although it is not yet known how cells from gut recognize pathogenic bacteria among the commensal ones. But, if fish did not have this recognition system, the immune response in intestine would be active all the time because there are millions of commensals interacting with the epithelial cells.

Environmental or aquaculture-related insults couple the mucosal defensive reactivity with the activation of fish hypothalamic– pituitary–interrenal (HPI) and sympathoadrenomedullary (SAM) stress axis (19). As in the case of mucosal immune system, fish react differently depending on the stressor and the influence of the immune response on survival remains species specific (20). Plasmatic cortisol is a well-known indicator of stress situation experienced by fish (21) and also a recurrent mediator of bidirectional immunoendocrine regulation (22). In acute stressed fish, cortisol is secreted within several minutes up to 1 h into circulation (23). The release of cortisol from the head kidney modulates the leukocyte-mediated response and negotiates the onset, lag, and efficacy of immune reactivity. This may influence the mucous adherence and virulence of some pathogens (24) and may destabilize the host–microbiota interaction in favor of opportunistic pathogens (17). Little is known about the mechanisms of cortisol secretion in the mucosal tissues, but it has been suggested that cortisol levels in skin mucus correlate with those of plasma (24, 25) and may modulate specific tissue receptors and cytokine expression in mucosae (26). In this way, it has been reported that the stress-mediated increment of mucus-producing cells in mucosal surfaces induced a reduction in the number of parasites in mucosae (6, 27). Thus, fish mucosal barriers are thought to act as sensors playing a significant role in monitoring stress.

Thus far, the effects of stress on the immune system have been described mainly in systemic compartments including blood, head kidney, liver, and spleen. From these results, it has been implicitly assumed that the physiological stress response is similar among different fish species (21, 28). Moreover, little attention has been paid on the interaction and cross-modulatory effects between endocrine and immune systems among different fish species under stress situations. In fact, it has been recently reported that the combination of stress hormones and pathogen antigens could differentially induce a species-specific response (29). On the other hand, at the local response level, few studies have addressed the effects of stressful stimuli on the fish mucosal immune system. To date, several investigations have focused on the acute (25) and chronic (30) stress effects in mucosal tissues, but no study has elucidated the modulatory effect of different types of stressors (biotic, abiotic, and the combination of them) on mucosal tissues.

Although most fish show a generalized stress reaction *via* activation of primary and secondary responses (31), there is a specificity on the pattern and magnitude of the response that may be affected by not only environmental factors (such as temperature and salinity) but also the nature of the stressor. Our hypothesis was that fish respond qualitatively similar to stressors but that this response can be significantly modulated by both genetic background and environmental conditions. Therefore, we focused our work in the differential response between the two species.

In this study, we describe the short-term (1, 6, and 24 h) effect of a biotic stressor (*V. anguillarum* bacterin bath), an abiotic stressor (air exposure), and the combination of both stressors in physiological indicators (plasmatic and skin mucus cortisol) and SALT, GIALT, and GALT mRNA abundance (stress- and immunerelated genes). These treatments were selected as similar handling procedures may be also often present when fish are subjected to vaccination in the aquaculture industry. This study was carried out using two commercial relevant species that inhabit in two distinct milieu: rainbow trout (*Oncorhynchus mykiss*; a freshwater teleost) and gilthead seabream (*Sparus aurata*; marine teleost). We aimed not only to clarify the role of mucosal immunity in the overall immune response of these two species under stress situations but also to show how the mucosae of aquatic vertebrates react to stressors of different nature.

#### MATERIALS AND METHODS

#### Experimental Animals

Juveniles of rainbow trout (mean weight: 130 g) and gilthead seabream (mean weight: 65 g) were obtained from local fish farms (TroutFactory and Aquicultura els Alfacs, Spain) and acclimatized for 3 weeks at the Universitat Autònoma de Barcelona fish facility (AQUA-UAB) in conic tanks (2.0 m3 total capacity) with water pump, recirculating chiller cooling system, sand filter, and biofilter. Fish were maintained at a photoperiod of 12L:12D and at their respective environmental temperature (15°C for trout; 20°C for seabream). Fish were fed a commercial pellet (Skretting) at 1.5% of total body weight/day. Water quality indicators (dissolved oxygen, ammonia, nitrite, and pH) were analyzed periodically. These conditions were maintained also for the experimental tanks. The experiment complied with the Guiding Principles for Biomedical Research Involving Animals (EU2010/63), the guidelines of the Spanish laws (law 32/2007 and RD 53/2013), and authorized by the Ethical Committee of the Universitat Autònoma de Barcelona (Spain) for the use of laboratory animals.

#### *V. anguillarum* Bacterin

An inactivated, formalin-killed *V. anguillarum*, serotype O1, O2α (the most pathogenic serogroup), and O2β, all with relative percentage survival ≥ 60% (Icthiovac® VR, Hipra) was utilized as a source of antigen.

## Experimental Design

For the experiment, fish were placed in 300 l conic tanks with the closed recirculating system provided with water pump, sand filter, and biofilter. The temperature (15°C for trout; 20°C for seabream) and photoperiod (12L:12D) were set accordingly. Fish were divided into three groups and maintained in eight independent tanks. (1) *Vaccinated (v) group*: 48 fish were vaccinated by immersion (1 min) with formalin-killed *V. anguillarum* bacterin according to manufacturer's instructions (Hipra). Immediately after, fish were rinsed in a cleaned water cube to discard the vaccine excess. Fish were then equally distributed (*n* = 12) in four tanks, avoiding cross-contamination for vaccine. (2) *Vaccinated and stressed (v* + *s) group*: 24 h after vaccination, 24 fish randomly selected from the vaccinated group were stressed (acute air exposure stress, 1 min) and returned to two separated tanks. (3) *Stressed (s) group*: 24 non-vaccinated fish were maintained out of water, stressed (acute air exposure stress, 1 min), and returned to two separated tanks. Control fish (*n* = 24) were mock-vaccinated (water vaccine-free immersion) in the same conditions as the vaccinated group, returned to two different separated tanks, and sampled after 24 h. Concerning time course utilized for the vaccine group, it should be stated that the preliminary data did not show any effect of vaccine immersion after 1 h and 12 h post vaccine (hpv) (data not shown). Therefore, we decided to begin sampling 24 h after bath vaccination. "Time 0" for vaccinated and vaccine + stress groups represents 24 hpv, whereas in the stress group represents the initial point of the experiment. Fish (*n* = 8) were randomly sampled from the two separated tanks per treatment at 1, 6, and 24 h post-stress (air exposure) from each experimental group (control, v, v + s, and s) and sacrificed by overanesthetization in MS222 (200 mg/l).

#### Skin and Tissue Sampling

Rainbow trout and gilthead seabream skin mucus was sampled according to Xu et al. (32). After blood sampling, skin tissue samples (upper lateral line area behind the dorsal fin, left side, and roughly same size) were carefully taken to avoid muscle contamination. Gills (first lamella from both sides) were also sampled. For gut analysis, the body cavity was opened laterally, and midgut and hindgut were removed using a sterile scalpel and forceps. These harvested intestine sections were open longitudinally and feces and mucus carefully removed with forceps. Samples from all fish were immediately frozen in liquid nitrogen and stored at −80°C for further assays.

## Quantification of Cortisol in Plasma and Skin Mucus

Cortisol level was measured by radioimmunoassay (33), and the radioactivity was quantified using a liquid scintillation counter (Scintillation Counter Wallac 1409; PerkinElmer). Anti-cortisol antibody was used for the assay at the final dilution of 1:4,500. Antibody cross-reactivity with cortisol was 100%, and the lower detection limit of the assay was 0.16 ng/ml. Cross-reactivity with other steroid hormones varied from 1.6% for corticosterone and was inferior to 0.7% for other tested steroids.

## IgM Detection in Skin Mucus

Levels of IgM in rainbow trout and gilthead seabream skin mucus at 1, 6, and 24 h post-stress were determined by ELISA according to Cuesta et al with modifications (34). Rainbow trout skin mucus samples were 1/4 diluted in PBS + 10 mM EDTA. 50 µl/well was added and incubated at 4°C onto Maxisorp microplates (Thermo Fisher Scientific) in duplicate. The unbound antigen was removed by washing twice with 200 µl/well of PBS. Possible sites with no antigen bound were blocked with 100 µl/well non-fat milk 5% in PBS for 1 h at room temperature (RT) and washed twice with PBS. Antibody mouse anti-trout IgM 1.14 mAb (1/1,000 dilution in PBS) and anti-seabream IgM mAb (1/100 dilution in PBS determined by Western blot) (Aquatic Diagnostics Ltd., UK) were used as primary antibodies to detect the presence of IgM on skin mucus. Samples were incubated with 50 µl/well of primary antibody for 1 h at RT, followed by three times washing with 200 µl washing buffer (PBS + 0.15% Tween 20). Samples were incubated with 50 µl/well of goat-anti-mouse IgG conjugated with HRP (1/4,000 dilution in PBS). The microplate was washed five times with 200 µl/well of washing buffer, and 50 µl/well of Ultra-TMB (3,3′,5,5′-tetrametilbenzidine; Thermo Fisher Scientific) was added as a substrate. After incubation for 7 min at RT, 50 µl/well of H2SO4 (2 M) was added as stop solution and absorbance was determined at 450 (0.1 s) nm with a microplate reader (Victor3; Perkin Elmer). All samples were evaluated in duplicated.

#### Isolation of RNA and cDNA Synthesis

Total RNA was isolated from individual fish samples using TRI reagent (Sigma) according to manufacturer's instructions. The RNA pellet was dissolved in autoclaved milli Q-water and immediately stored at −80°C until use. The RNA concentration was quantified by a NanoDropND-2000 spectrophotometer (Thermo Fisher Scientific). Total RNA (2 µg) was used as a template to synthesize cDNA using High-Capacity cDNA Reverse Transcription Kits (Applied Biosystems) according to manufacturer's instructions and immediately stored at −20°C until use.

#### Quantitative Real-Time PCR

Fish mucosal samples including skin, gills, and gut were analyzed using real-time PCR. The analysis included the evaluation of stress and immune-related genes (*lysozyme*, *c3*, *igm*, *hsp70*, *cox2*, *il1β*, *tnfα*, *il6*, *tgfβ1*, and *il10*). We tested several housekeeping candidate genes in rainbow trout (*ef1α* and *βactin*) and seabream (*18s*, *ef1α*, and *rpl27*) to elucidate which one had less variation. *β-Actin* (for rainbow trout) and *18s* (for seabream) were included on gene expression analysis. Specific primers used for rainbow trout (**Table 1**) and gilthead seabream (**Table 2**) are indicated. Primers were designed with Primer-Blast. The primer secondary structure and annealing specificity was checked with OligoAnalyzer (version 3.1) and Primer-Blast software, respectively. The undesirable PCR products appearance was previously verified by single peak in the melting curve for each primer set. The primer amplification efficiency was determined in all mucosal surfaces included in our study. Real-time PCR reactions were performed with iTaq universal sybr green supermix (Bio-Rad Laboratories) using 1:20 and 1:10 cDNA dilution made for genes of interest in rainbow trout and gilthead seabream, respectively. Primers for all genes were used at a final concentration of 500 nM. The thermal conditions used were 3 min at 95°C of pre-incubation followed by 40 cycles at 95°C for 30 s and 60°C for 30 s. All the reactions were performed in duplicate using CFX384 Touch Real-Time PCR Detection System (Bio-Rad Laboratories). Quantification was done according to the Pfaffl method (35) corrected for efficiency of each primer set obtained for each mucosal surface evaluated. Values for each experimental condition were expressed as normalized relative expression against those of the housekeeping gene *β-actin* and 18s for rainbow trout and seabream, respectively. Results are expressed as average of values obtained for the same treatment and time points evaluated.

## Statistical Analysis

The statistical package for social science (SPSS, v20) software was used for the analysis. The Generalized Linear Model was utilized considering the stressors and time dynamics as a two betweensubjects factor. This model is a more flexible statistical tool than the standard general linear model in terms of types of distribution and different covariance structure of the repeated measures, does




not require homogeneity of variance, and it admits missing values. After the main analysis, appropriate pair-wise comparisons were carried out. Differences in all data were considered statistically significant if *p*-values < 0.05 among groups.

#### RESULTS

#### Plasmatic and Skin Mucus Cortisol Level

In order to evaluate whether the application of air exposure, *V. anguillarum* bacterin, or the combination of both stressors induce a stress response systemic (plasmatic) and local (skin mucus), cortisol levels were evaluated. In trout, plasmatic cortisol levels augmented after air exposure (126.32 ng/ml) 1 h post-stress (**Figure 1A**; red line) and decreased at 6 h post-stress (49.57 ng/ml). The highest cortisol concentration was registered in the vaccine + air exposure group (159.85 ng/ml) at 1 h post-stress (**Figure 1A**; orange line). A slight decrease, although still higher than the control group, was also observed at 6 h post-stress in the vaccine + air exposure group (135.40 ng/ml). In the vaccinated group, no variations were registered on plasma cortisol (**Figure 1A**; blue line).

The same secretion pattern was observed in skin mucus in the air exposure group. Cortisol levels augmented at 1 h poststress (2.41 ng/ml), diminished at 6 h post-stress (1.34 ng/ml), and returned to control level at 24 h post-stress (0.20 ng/ml) (**Figure 1B**; red line). In the vaccine + air exposure group, an increase in cortisol was only observed at 6 h post-stress (1.99 ng/ml) (**Figure 1B**; orange line). In the vaccinated group, cortisol was significantly increased at 1 h post-stress (1.48 ng/ml) (25 h after vaccination) and recovered baseline values at 24 h post-stress (48 h after vaccination) (**Figure 1B**; blue line).

In gilthead seabream, the secretion patterns were similar but differed in magnitude. Significantly elevated plasma cortisol levels at 1 h post-stress (228.18 ng/ml) (**Figure 1C**; red line) were detected in the air exposure group, decreased at 6 h post-stress (96.40 ng/ml) and dropped to control levels at 24 h post-stress (25.71 ng/ml). In the vaccinated + air exposure group, the cortisol values were higher than control at 1 h (144.08 ng/ml) and 6 h (113.30 ng/ml) but not at 24 h post-stress (61.0 ng/ml) (**Figure 1C**; orange line). The vaccinated seabream group showed no variations (**Figure 1C**; blue line).

Cortisol levels in seabream skin mucus differed from those observed in trout skin mucus. High cortisol levels (1.89 ng/ml) were registered in the air exposure group at 1 h post-stress (**Figure 1D**; red line) and doubled at 6 h post-stress (3.91 ng/ml). In the vaccinated + air exposure group, cortisol levels augmented at 1 h (1.43 ng/ml) and 6 h post-stress (1.18 ng/ml) (**Figure 1D**; orange line). No significant variations were observed in the vaccinated group (**Figure 1D**; blue line).

The differences noted between trout and seabream skin mucus were also observed when the total amount of IgM was determined by ELISA. In trout skin mucus, the levels of IgM showed no variations after the different treatments (**Figure 1E**; left half). However, in seabream the levels of IgM in the air exposure group increased gradually, reaching a peak at 24 h post-stress (Abs450 = 0.25) compared to control (**Figure 1E**; right half). This augment was also registered at 24 h post-stress in the vaccinated + air exposure group (Abs450 = 0.17) (**Figure 1E**; right half).

Thus, our results indicate that the release of the glucocorticoid hormone in response to stressor depends on the stressor, the biological matrix (plasma or mucus), and is species dependent.

exposure and the vaccine + air exposure) or 24 h after bath vaccination (for vaccine group). Data are represented as mean ± SE (*n* = 8 per sampling time point). Significant differences are indicated by lower case in the air exposure group, by upper case in the vaccine + air exposure group, and by numbers in the vaccine

#### SALT Responses

mRNA expression levels were used to examine whether the air exposure as well as *V. anguillarum* were able to drive differences in the transcriptomic responses of stress and immune-related genes in MALTs: skin, gills, and gut. Our results show an overall significant interaction between treatment and time course at 1, 6, and 24 h post-stress in both species. In rainbow trout (**Figure 2A**), air exposure was able to enhance the transcription of *il1β*, *cox2*, and *lysozyme* in a time-dependent manner. The vaccine + air exposure treatment promoted the upregulation of genes associated with immunity and stress responses (*c3*, *igm*, *hsp70*, and *cox2*) at 1 h post-stress. The expression of pro-inflammatory transcripts (*il1β*) was also upregulated. The same effect was also observed at 6 h post-stress for *il1β* but not for *cox2*. No modulation was observed at 1 h poststress in gene transcripts associated with anti-inflammatory responses (*il10* and *tgfβ1*). However, the upregulation of *tgfβ1* was only observed at 6 and 24 h post-stress in the vaccinated group, probably linked with the upregulation also observed for *lysozyme*, *c3*, *cox2*, and *hsp70*. Overall, all genes showed a marked upregulation in a treatment- and time-independent manner.

group. Asterisk (\*) indicates significant difference of each treatment versus control (*p* < 0.05).

In gilthead seabream (**Figure 2B**), air exposure was found to induce pro-inflammatory cytokine transcripts (*il1β* and *il6*), *cox2*, and also *lysozyme* at 24 h post-stress. The upregulation of genes in the vaccine + air exposure group was also observed, though the magnitude and time course of this modulation was shown to be different compared to rainbow trout. A high and decreasing expression from 1 h post-stress to 24 h post-stress in *hsp70* and *lysozyme* was reported. The same expression pattern was observed for *tgfβ1*. An increased gene expression at 6 h post-stress was noted for *il1β*, *il6*, *tnfα*, *cox2*, and *igm*. The upregulation of *il10* was modulated in the same manner. An increase in a time-dependent manner was registered only for *c3*. Importantly, this upregulation in the vaccine + air exposure group seems to be influenced by air exposure and vaccine separately. Only in the cases of *il1β* and *il6*, the effect observed in the vaccine + air exposure group could be markedly associated with the expression registered in the air exposure group and vaccinated group, respectively. Importantly, *V. anguillarum* bacterin was able to induce expression of *il6*, *tnfα*, *tgfβ1*, *hsp70*, *cox2*, and *c3* mainly at 24 h post-stress.

In summary, a lower gene expression magnitude was observed in rainbow trout than in seabream (**Figure 2C**). In contrast to rainbow trout response, the gene expression data suggest a higher influence of the air exposure stressor and the combination of both stimuli in gilthead seabream. The similar expression of *lysozyme*, *hsp70*, and *tgfβ1* suggests that the anti-inflammatory cytokine response could modulate the expression of these immune-related genes in both species when stressed.

#### GIALT Responses

Gills showed a different gene expression pattern when comparing both fish species. In trout (**Figure 3A**), the gene transcript modulation in the air exposure group was observed at 6 h post-stress in pro- (*il1β* and *tnfα*), anti-inflammatory genes (*il10* and *tgfβ1*), and *cox2*. Particularly, the downregulatory tendency was observed in the air exposure group and the vaccine + air exposure group at 1 h post-stress, suggesting that the stress by air exposure could influence the early post-stress expression of trout transcripts. In the vaccinated group, the expression of *il1β* and *cox-2*, *lysozyme*, and *igm* were observed at 1 h post-stress. At 6 h post-stress, the expression levels of *il6* and *hsp70* were also modulated. Importantly, *il10* was also upregulated at 6 h post-stress, suggesting that its modulation could be related to the control of the pro-inflammatory gene expression profile. This suggests that air exposure and vaccine alone had a stronger effect on gene expression in trout.

In seabream (**Figure 3B**), a marked upregulation of the expression of *il1β*, *tnfα, cox2, lysozyme*, and *c3* was registered both at 1 and 6 h post-stress and after vaccine + air exposure. The upregulation of *cox2* (1 h post-stress) and *il1β* (6 h post-stress)

7 **48**

vaccine group. Asterisk (\*) indicates significant difference of each treatment versus control (*p* < 0.05).

suggests a specific gene expression effect of air exposure in seabream. The downregulation of several genes in the air exposure and vaccinated groups suggests that the stress stimuli including air exposure and *V. anguillarum bacterin* alone may suppress seabream immune response in gills.

Our results indicate that, aside from the increase in few genes in the gills of rainbow trout and seabream, both stressors separately induce immune suppression or a tendency to reduce immune and stress gene transcription in the gills (**Figure 3C**).

#### GALT Responses

Transcriptomic profile analysis of both species showed that air exposure modulated the immune- and stress-related gene expression transcripts in gut. In trout (**Figure 4A**), the upregulation of gene transcripts involved in regulatory responses (*lysozyme*, *c3, il1β*, *tnfα*, *tgfβ1*, *cox2*, and *igm*) was observed at 1 h post-stress indicating that, in gut, air exposure induces the upregulation of immune-related genes in trout earlier than in seabream. The same modulation of the pro-inflammatory genes (*il1β* and *tnfα*) was also observed in trout at 1 h post-stress in the vaccine + air exposure group, suggesting that this response could be directly influenced by the air exposure at the same time-point. *il1β*, *tnfα*, *tgfβ1*, *c3*, and *lysozyme* were upregulated in the vaccinated group after 6 h post-stress, *cox2* after 24 h post-stress, and *igm* remained downregulated.

Figure 3 | Gene expression level in gill-associated lymphoid tissue (GIALT) after a biotic (*Vibrio anguillarum* bacterin), abiotic (air exposure), or the combination of both stressors (vaccine + air exposure) at 1, 6, and 24 h post-stress. Quantitative real-time PCR quantification of specific mRNA immune and stress transcripts (*lysozyme*, *c3, igm, hsp70, cox2, Il1β*, *tnfα*, *il6, tgfβ1*, and *il10*) in (A) rainbow trout and (B) gilthead seabream gills. (C) Integrative comparison of GIALT transcriptomic responses. *β-Actin* and *18s* were chosen as housekeeping genes in rainbow trout and gilthead seabream, respectively. Different colors indicate different treatments: red (air exposure), orange (vaccine + air exposure), blue (vaccine), and gray (ctrl; control). Time zero represents the end of the acute air exposure stress (for the air exposure and the vaccine + air exposure) or 24 h after bath vaccination (for vaccine group). Data are represented as mean ± SE (*n* = 8 per sampling time point). Significant differences are indicated by lower case in the air exposure group, by upper case in the vaccine + air exposure group, and by numbers in the vaccine group. Asterisk (\*) indicates significant difference of each treatment versus control (*p* < 0.05).

Interestingly, in seabream gut (**Figure 4B**), the larger alteration was induced mainly by air exposure. The transcriptional level of immune and stress regulators (*il1β*, *il6*, *tnfα*, *il10*, *tgfβ1*, *cox2*, *hsp70*, *lysozyme*, *c3*, and *igm*) was enhanced at 6 h post-stress, indicating a higher sensitivity of this specie to air exposure. Our results suggest that in gut the air exposure stress promotes a roughly similar immunerelated gene expression modulation in both species although in a different magnitude and time-dependent manner (**Figure 4C**).

Taken together, the results show that both stressors modulate the SALT, GIALT, and GALT transcriptomic response, but such response depends on the nature of the stressor, time, and the species concerned.

11 **52**

Figure 4 | Gene expression level in gut-associated lymphoid tissue (GALT) after a biotic (*Vibrio anguillarum* bacterin), abiotic (air exposure), or the combination of both stressors (vaccine + air exposure) at 1, 6, and 24 h post-stress. Quantitative real-time PCR quantification of specific mRNA immune and stress transcripts (*lysozyme*, *c3*, *igm*, *hsp70*, *cox2, Il1β*, *tnfα*, *il6*, *tgfβ1*, and *il10*) in (A) rainbow trout and (B) gilthead seabream. (C) Integrative comparison of GALT transcriptomic responses. *β-Actin* and *18s* were chosen as housekeeping genes in rainbow trout and gilthead seabream, respectively. Different colors indicate different treatments: red (air exposure), orange (vaccine + air exposure), blue (vaccine), and gray (ctrl; control). Time zero represents the end of the acute air exposure stress (for the air exposure and the vaccine + air exposure) or 24 h after bath vaccination (for vaccine group). Data are represented as mean ± SE (*n* = 8 per sampling time point). Significant differences are indicated by lower case in the air exposure group, by upper case in the vaccine + air exposure group, and by numbers in the vaccine group. Asterisk (\*) indicates significant difference of each treatment versus control (*p* < 0.05).

#### DISCUSSION

In this study, we aimed to evaluate the isolated and combined effects of a biotic (*V. anguillarum* bacterin bath) stressor and abiotic (air exposure) on SALT, GIALT, and GALT of two different species, trout and seabream. To do so, we analyzed the gene expression patterns of several relevant stress- and immune-related transcripts (*lysozyme*, *c3*, *igm*, *hsp70*, *cox2*, *Il1β*, *tnfα*, *il6*, *il10*, and *tgfβ1*) and plasmatic and mucosal (skin mucus) cortisol levels. Overall, our results indicate that short-term stressors modify plasma and skin cortisol levels and regulate the transcriptomic response of several immune mediators at mucosal surfaces in both species. The observed changes suggest that stress disturbs the mucosal tissue and mainly enhances the mucosal immune response in a pronounced species-specific manner. Contrary to the notion that stress suppresses the immune response, we clearly show that under certain conditions, particularly short-term disturbances, stress can activate mucosal immune function.

#### SALT Responses

Once a stressor is sensed by the host, the activation of the HPI-SAM axis releases cortisol, a stress biomarker that, in turn, activates secondary and tertiary stress responses (36). It should be taken into account that an accurate description of the integrated stress mechanisms should include the role of local stress and immunoregulatory responses (31) such as those of mucosal surfaces. The mucosal immune system is considered an active immunological interface (4) in which stressors of different scale and sources may recruit HPI-SAM axis elements and impose long-term disturbances (21, 37, 38). Skin is one of the mucosal immune tissues that act as first barrier against both pathogens and stressors (6). Interestingly, recent work on the local stress response in skin involved the cortisol measurement in scales (39) and also in skin mucus (25), as the cortisol lipophilic nature makes feasible its diffusion through cell membranes. However, the mechanisms of mucus cortisol presence are poorly understood (25), but the amount of cortisol in mucosal surfaces may be a proxy for plasmatic cortisol levels or an indirect indication of local inflammation. In fact, recent research showed genes and peptides in peripheral tissues related to molecules of the stress axis (40). Therefore, such an issue is far from having a definitive answer, and more research has to be done in this matter.

In this study, cortisol levels following air exposure peaked in a time-dependent and species-specific manner. The higher cortisol levels after stress in gilthead seabream may indicate a lower activation threshold to air exposure. This effect did not take place after *V. anguillarum* exposure. This differential activation threshold would imply a higher activity of interrenal cells, as it was already proposed (41). While the cortisol levels in the vaccine + air exposure group were higher than the levels observed in the air exposure group in trout, in seabream an opposite situation was true, suggesting a higher sensitivity to the combination of stressors on HPI activation in rainbow trout.

Concerning *V. anguillarum* bacterin treatment, cortisol skin mucus levels raised only in trout. It has been shown that common fish pathogens such as *Lepeophtheirus salmonis*, *Edwardsiella ictaluri*, *V. anguillarum, Aeromonas salmonicida*, and *Pseudomonas anguilliseptica* are capable of inducing cortisol release in fish (42–45). Particularly, *V. anguillarum* is a widespread fish pathogen that adheres to the mucosal tissues and elicits a strong cortisolmediated stress response (46). Based on the results obtained in this study, it seems that the differential effect to *V. anguillarum* may either reside on a differential response at the mucosal level among the two species or the antigenic effect causing endocrine alterations. As stated before, the highest plasmatic cortisol values were registered in the vaccine + air exposure group for trout but not for seabream. Therefore, the apparently higher reactivity to the stimuli observed in trout skin mucus could help to explain the highest plasmatic cortisol levels obtained in trout enduring the combination of stressors. By contrast, in seabream it seems that cortisol increase was produced only in the air exposure treatment.

It has been reported that higher levels of cortisol and catecholamines following exposure to acute stressors may increase the number of circulating leukocytes, specifically neutrophils, and reduce lymphocyte numbers (21, 47–49). Thus, a modulation in IgM levels during acute stress responses is to be expected, although no alterations in the seric IgM levels in seabream after acute air exposure stress was reported in previous work (34). IgM is the most abundant immunoglobulin in skin mucus and provides protection against pathogens that are in close contact with outermost fish surfaces (6). In this study, no changes were observed in any of the stressors tested for IgM trout skin mucus. Data also indicate that the augment of cortisol on seabream skin mucus registered at 1 and 6 h post-stress did not modify the levels of IgM. However, at 24 h post-stress, the air exposure and the combination of the vaccine + air exposure stressors (not vaccinated) were able to increase the amount of total IgM in seabream skin mucus. The increase in IgM skin mucus could be associated with an immune protective mechanism in response mainly to the air exposure stress in seabream, reinforcing the hypothesis of a lower activation threshold to acute handling stress in seabream.

In this study, skin shows the highest transcript abundance of all three mucosal surfaces, particularly in seabream. This upregulatory response of trout skin was mainly observed in the vaccine + air exposure and in the vaccinated groups. By contrast, the increase in transcript expression in seabream skin was observed particularly in the air exposure and vaccine + air exposure groups, again suggesting a higher responsiveness to biotic stimuli in trout and to abiotic stimuli in seabream. *cox2* mRNA levels were elevated in skin of trout and at a greater extent in seabream, in agreement with the previously described increase in *cox2* after acute stress in the skin and intestine (37, 50). The upregulation of *lysozyme*, *c3*, and *igm* in trout and seabream vaccine + air exposure groups indicates that the combination of different stressors may activate the mucosal immunity. *Lysozyme* and *c3* are ubiquitously expressed antimicrobial and bactericidal components of the innate arm of the mucous immune system (51), and expression of *c3* indicates that extrahepatic *c3* also may play a role in stressmediated local mucosal immunity responses. The expression of seabream IgM on skin mucus was not correlated with the seabream expression pattern on skin. However, it is important to take into consideration that, although directly related, skin mucus and skin are considered different tissue matrices as also suggested by their distinct role in the stress response. While in the skin the modulation of gene expression takes place mostly in resident cells, in skin mucus the total protein content could be influenced not only by the skin resident cells but also by the cell trafficking and protein secretion as an outcome of stress responses. Due to the intimate contact with the surrounding environment, the provoked immune response in the skin may activate a local alert for the endocrine messengers in the mucosa to be prepared for potential challenges. It is worthy to note that during acute stress, skin is enriched in leukocytes and, as it was stated before, shortterm stress substantially increases trafficking of leukocytes to the skin in mammals and fish (52, 53), assuring the mobilization of leukocytes to skin and probably the IgM synthesis.

The inflammatory response plays a key role in the host defense activation mechanisms. Not only pro- but also antiinflammatory cytokine transcriptions were dramatically raised in seabream skin. This suggests an attempt to control/unleash a nascent inflammatory response, recruiting anti-inflammatory and wound-healing agents such as *il10* and *tgfβ1* (54, 55). Analog to mammals, fish inflammatory responses are characterized by a first wave of expression of pro-inflammatory cytokines (56–58). At later stages of inflammation, the release of a second wave of anti-inflammatory cytokines by macrophages initiates the process of recovery, which is pivotal to reduce the inflammation (57). Moreover, the excessive induction of the pro-inflammatory agents and innate immune components may not only harm the host but also impose more energy consumption (59). Therefore, this mechanism could also be related to minimize the energy expenditure in other physiological processes different from the stress response.

In sum, a differential modulatory effect affecting the mRNA abundance of relevant immune biomarkers was determined in skin. Particularly, the seabream response was characterized by a significant upregulation on genes related to immune and stress response to air exposure and the combination of vaccine and air exposure stimulus.

#### GIALT Responses

Several pathogens show a preference for gills during the adherence phase of the infective cycle. In this way, it has reported that a pathogen is able to rapidly modify the host mucus transcriptomic responses to facilitate bacterial adherence (58). *V. anguillarum* has been shown to cause serious diseases in fish gills provoking upregulation of both pro-inflammatory cytokines and their mediator molecules in trout and cod (60–62). Our results with *V. anguillarum* bacterin show similar responses among mucosae but a clear difference in terms of gene expression between species. In gill mucosa, the inflammatory response triggered by stressors showed a general suppression of transcripts in seabream but not in trout. The increase in transcription of pro-inflammatory cytokines *il1β* and *il6*, simultaneously with the *il10* increment following vaccine exposure, indicates that inflammation would be the predominant response in trout. Thus, markers of bacteriolytic responses in mucosal surfaces such as *lysozyme* and complement component *c3* (63, 64) were upregulated in trout but not in seabream, therefore, providing more arguments to the modulatory gene activation of immunity-related genes to *V. anguillarum* bacterin in trout. The same modulatory effect by *V. anguillarum* bacterin treatment in trout gills and the air exposure stressor in seabream was also observed in the *hsp70* and *cox2* transcript levels. This clear distinct direction of the gill response found between trout and seabream, in addition to the specificity of the response, may be associated with their difference in the genetic diversity and also with the environmental conditions such as temperature. Our results show that a most marked downregulation in seabream occurs after both air exposure and vaccine stressors. Compared to skin and gut, the overall lower expression values of immune- and stress-related genes indicate that the gill mucosa seems to be less responsive or more regulated after stress. Therefore, notwithstanding the upregulation of gene expression in trout (*il1β*, *il6*, and *il10*) to *V. anguillarum* bacterin, the overall response to stressors in the branchial tissue may obey to the constraints of the metabolic trade-offs between respiratory, osmoregulatory, and immune processes in such a multifunctional organ that may confine the number of resident macrophages and lymphocytes.

#### GALT Responses

Gut not only carries out the nutrient absorption but also acts as a physical and chemical barrier in which innate and adaptive immune responses are also crucial for protection (4, 10, 65, 66). Unlike the expression observed in trout skin and gills, a mixed effect was observed in trout gut depending on the gene evaluated. According to the results obtained in trout skin mucus, skin, and gills, an influence on the gene modulation (particularly *lysozyme* and *igm*) by vaccine + air exposure and vaccine groups was registered. However, in trout gut a similar expression pattern (*c3*, *il1β*, and *tnfα*) was observed in fish subjected to air exposure and vaccine + air exposure. *Lysozyme* expression in rainbow trout intestine agrees with previous results obtained in Atlantic cod (*Gadus morhua*) vaccinated against *V. anguillarum*, showing induction of antibacterial genes (61). This indicates, contrary to the expression of lysozyme found in skin, that the modulation of intestinal *lysozyme* expression by stressors may be tissue and/or species dependent. The upregulation of *cox2* is consistent with previous reports in Atlantic salmon showing increments of *cox2a* in midgut 1 h after stress (50, 67). On the other hand, impairment of intestinal functions has also been observed in mammals as a consequence of prostaglandin increment. As it has been previously described, cortisol-mediated stress responses may alter intestinal permeability (68), hence animals prevent such an increment of permeability through reduction of the prostanoid content after acute stress, which confirms the impact of prostaglandins on intestinal homeostasis (67) in connection with the expression of *cox2* induced by *il1β* expression (69). Therefore, the modulatory effect of *cox2* observed in this study for both species could either be explained as a result of inflammation or cortisol elevation in fish.

The expression of trout pro/anti-inflammatory (*il1β, tnfα*, and *tgfβ1*) cytokines was mainly enhanced by air exposure and vaccination in gut, while mRNA abundance of seabream cytokines (*il1β*, *il6*, *tnfα*, *il10*, and *tgfβ1*) was mainly induced by air exposure. As mentioned before, the balance between pro- and anti-inflammatory cytokines is crucial to control the inflammation. Previously reported gene expression levels of *il1β* in Atlantic salmon intestine decreased after 7 weeks exposure to hypoxia, suggesting that short-term and long-term stress may induce differential regulation of cytokines (70, 71). *igm* mRNA abundance was stimulated in trout and seabream after air exposure at different time points. The data suggest that *igm* can also be modulated in mucosal surfaces by abiotic stressors such as air exposure, as previously described in gills and intestine of stressed *Epinephelus coioides* and *Oreochromis niloticus*

Khansari et al. Fish mucosal transcript response

(72, 73). It has also been reported that environmental changes and also *V. anguillarum* increased the expression of pro- and antiinflammatory cytokines in the gastrointestinal tract (70, 74, 75) and also in head kidney, spleen, and liver (76–78). It is worth mentioning that regulation of *igm* appears to be repressed by *A. salmonicida* in Atlantic cod (79), indicating a different regulation of *igm* when fish are exposed to an antigen compared to a pathogen exposure. Our findings confirm the relevance of duration and type of the stressors that affect particularly seabream mucosal tissues and suggest more pronounced effects of air exposure in seabream intestine. Altogether, the results suggest that in trout gut a modulation of particular genes will be activated depending on the type of stressor, in this case biotic or abiotic. On the other hand, the expression of pro/anti-inflammatory (*il1β*, *il6*, *tnfα*, *il10*, and *tgfβ1*) cytokines in seabream gut was mainly induced by air exposure, reinforcing the relevance of the abiotic stressor effect on seabream mucosal tissues. Thus, the induction of stress and immune genes expression was coincident with high levels of plasmatic and mucus cortisol. Overall, from our findings, intestine appears to be one of the most affected surfaces by different types of stressors, and in terms of gene expression, the gut mucosa shows higher sensitivity to air exposure than to vaccine.

#### Overview of MALT Responses

Skin mucus cortisol level showed variations between species, and a clear difference was also observed in terms of stress- and immune-related gene expression. Skin and intestine appear to be the most affected surfaces after different types of stressors both in trout and seabream. When applying both stressors, skin particularly appears to be the most reactive barrier to vaccine + air exposure.

The extent to which husbandry conditions modulate mucosal immune response need to be much more investigated because of the complexity of the immune system and the interactive nature of the stress response. Thus, dealing with stressors of different features it may be problematic to predict the direction and magnitude of the response. Our results show, in general, an increased MALT response after the combination of stressors in seabream but not clearly in trout. Hence, previous studies showed higher response of innate indicators in low density than high density after bacterial exposure (80). One of the reasons that may explain why a combined stressor does not induce higher responses could be associated with the energetic load that concurrent challenges would require for such an increased response. Thus, the available energy would not be enough to meet the energetic needs.

Several reasons can be claimed to be responsible for the interspecific differences observed: one is the diversity of the species living in either marine or freshwater habitats. In fact, the differentiation of the fish population is eight times higher in freshwater than in seawater environments, which would support the differences among genomic architectures (81). Hence, the ecological characteristics of *V. anguillarum*, halophilic bacteria, may partially explain the observed differences. Outbreaks of *V. anguillarum* bacteria affect mainly marine and estuarine fish species at different salinities (usually 1–2% NaCl) and temperatures exceeding 15°C (82). *V. anguillarum* can also be found occasionally in freshwater, forming biofilms to enhance bacterial survival in an otherwise suboptimal environment (83). Therefore, freshwater fish (trout) would be more susceptible to *V. anguillarum*. Given that temperature seems to be more detrimental than salinity for *V. anguillarum* growth (7), our results suggest that cold freshwater trouts may not experience significant exposure to *V. anguillarum* in the natural environment, thus lacking an evolutionary-driven, parasite-tuned host–pathogen immune crosstalk. Therefore, the increased responsiveness to *V. anguillarum* bacterin observed mainly in trout may account in part for the upregulation of several key inflammatory transcripts that are downregulated in a marine fish such as seabream. A second reason would be related to the interaction of *V. anguillarum* with the fish microbiota, as freshwater or seawater fish can display rather different microbiomes. Thus, a data set analysis from a large collection of 16 S rRNA of diverse free-living and host-associated bacterial communities from intestines of different fish species suggests that variation in gut microbiota composition in fish is strongly correlated with species habitat, salinity, and trophic level (84). A third reason would be related to the salinity or temperature themselves. Thus, it has been shown that hyperosmotic and also hypoosmotic stress modify the immune homeostasis in catfish (85). However, in these experiments, fish were subjected to changes from their acclimated conditions, whereas in the present study, both species were well acclimated to their termopreferendum and natural salinity levels to precisely avoid potential stress biasing the data analysis.

Altogether, the interspecific differences in the regulatory responses observed under the different stressors suggest an adaptive lifetime in either freshwater or marine habitats resulting from a complex interaction between environmental conditions, microbial communities, and genomic variation that may affect the intensity and dynamics of the inflammatory and stress responses.

#### CONCLUSION

Our findings illustrate the implication and importance of the mucosal immunity in response to different stressors and provide comparative data on the transcriptomic responses of several immunomodulators in MALT tissues. In species such as trout and seabream acclimated to their adaptive thermoneutral environments and confronted to *V. anguillarum* bacterin, our results show a higher responsiveness of skin and gills immune transcripts to the biotic stressor in trout than in seabream. On the other hand, in all mucosal organs evaluated, a higher response to the abiotic stressor was observed. Our results indicate that the response of the immune system is not homogeneous among mucosae and that is greatly influenced by the type of stressor, suggesting a trade-off between suppression and enhancement of immune responses depending on the intensity and duration of the stressors in each surface. In agreement with previous report in mammals and recent reports under *in vitro* conditions, our results clearly indicate distinctive responses of rainbow trout and seabream (86, 87). Considering the greater immune-related gene expression of seabream after stress in skin and gut, it can be suggested that mucosal tissues of gilthead seabream (a marine fish) show more responsiveness than rainbow trout (a freshwater species). This differential immune response can be attributed to the species specificity of the response, genetic diversity, or environmental conditions such as type and abundance of pathogens. This microorganism diversity may undoubtedly participate in explaining the different immune responses between fish, together with the microbiota, high or low salinity or higher or lower temperatures. However, the scarcity of studies on these environmental influences does not allow us to propose a consistent interpretation of those differences. Like mammals, the impact of acute stress and the consequent immunoendocrine reaction appears to enhance or modulate rather than always suppress the response of mucosal tissues. Thus, features of the stressors (type, intensity, and duration) determine the direction of the effect on mucosal immune system. Overall, and regarding the species differences, although our hypothesis is confirmed in the sense that the response to stressors is species-specific, we also show that such specificity is more intense, since two different species such as trout and seabream show not only quantitative but also qualitative differences in their responsiveness.

#### ETHICS STATEMENT

The experiment complied with the Guiding Principles for Biomedical Research Involving Animals (EU2010/63), the guidelines of the Spanish laws (law 32/2007 and RD 53/2013),

#### REFERENCES


and it was authorized by the Ethical Committee of the Universitat Autònoma de Barcelona (Spain) for the use of laboratory animals.

#### AUTHOR CONTRIBUTIONS

AK performed the sampling, gene expression, analysis and interpretation of results, and wrote the manuscript. JB performed graphic presentation of results, manuscript design, interpretation of results, and wrote the manuscript. DP performed sampling and interpretation of results. EVV performed gene expression analysis and interpretation of results. FERL and LT conceived the study design, supervised the experiment, performed the analysis and interpretation of results, and wrote the manuscript. All authors read, corrected and approved the final manuscript.

#### ACKNOWLEDGMENTS

This research was funded by the projects AGL2016-76069-C2- 2-R of MINECO (Spain) and Targetfish (EU 7th framework) with the support of FEDER funds (European Union). The authors are members of the Xarxa d'Aquicultura de Catalunya. AK was recipient of a grant from the Ministry of Science, Research and Technology of Iran, and EVV was recipient of a fellowship CONICYT-BCH from Chile.

Academic Press (2015). p. 93–133. doi:10.1016/B978-0-12-417186-2. 00005-4


interleukin-1. *Vet Immunol Immunopathol* (2002) 87:467–79. doi:10.1016/ S0165-2427(02)00077-6


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

The reviewer PC and handling Editor declared their shared affiliation.

*Copyright © 2018 Khansari, Balasch, Vallejos-Vidal, Parra, Reyes-López and Tort. 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.*

# Fish Lymphocytes: An evolutionary equivalent of Mammalian innate-Like Lymphocytes?

#### *Giuseppe Scapigliati\*, Anna M. Fausto and Simona Picchietti*

*Dipartimento per l'Innovazione nei sistemi biologici, agroalimentari e forestali, Università degli Studi della Tuscia, Viterbo, Italy*

Lymphocytes are the responsible of adaptive responses, as they are classically described, but evidence shows that subpopulations of mammalian lymphocytes may behave as innate-like cells, engaging non-self rapidly and without antigen presentation. The innatelike lymphocytes of mammals have been mainly identified as γδT cells and B1-B cells, exert their activities principally in mucosal tissues, may be involved in human pathologies and their functions and tissue(s) of origin are not fully understood. Due to similarities in the morphology and immunobiology of immune system between fish and mammals, and to the uniqueness of having free-living larval stages where the development can be precisely monitored and engineered, teleost fish are proposed as an experimental model to investigate human immunity. However, the homology between fish lymphocytes and mammalian innate-like lymphocytes is an issue poorly considered in comparative immunology. Increasing experimental evidence suggests that fish lymphocytes could have developmental, morphological, and functional features in common with innate-like lymphocytes of mammals. Despite such similarities, information on possible links between conventional fish lymphocytes and mammalian innate-like lymphocytes is missing. The aim of this review is to summarize and describe available findings about the similarities between fish lymphocytes and mammalian innate-like lymphocytes, supporting the hypothesis that mammalian γδT cells and B1-B cells could be evolutionarily related to fish lymphocytes.

Keywords: innate immunity, innate-like lymphocytes, fish lymphocytes, innate lymphoid cells, comparative immunology

## INTRODUCTION

Vertebrate-type adaptive responses with MHC, RAG, memory, are present in only 2% of metazoans, but invertebrates can live very long protected by their innate immune defenses. Indeed, invertebrates classically defined as relying only on innate responses may live for centuries and have been found to respond to reinfection, suggesting that innate immunity mechanisms need more investigation. In a comparative immunology view, it is conceivable to speculate that leukocytes populations that emerged early in vertebrates evolution inherited and retained some invertebrate features related to antigen recognition and elimination. During evolution, genes coding for immune activities accumulated toward mammals in a form of "layers." This hypothesis proposes that evolution produced a layered immune system in which following descendants obtain predominance during development, giving rise to cell populations responsible for progressively more complex immune activities. As it is commonly thought that "ontogeny resembles phylogeny," a "layered immune system" hypothesis may

#### *Edited by:*

*Brian Dixon, University of Waterloo, Canada*

#### *Reviewed by:*

*Teruyuki Nakanishi, Nihon University, Japan Mike Criscitiello, Texas A&M University, United States*

*\*Correspondence:*

*Giuseppe Scapigliati scapigg@unitus.it*

#### *Specialty section:*

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

*Received: 14 February 2018 Accepted: 18 April 2018 Published: 07 May 2018*

#### *Citation:*

*Scapigliati G, Fausto AM and Picchietti S (2018) Fish Lymphocytes: An Evolutionary Equivalent of Mammalian Innate-Like Lymphocytes? Front. Immunol. 9:971. doi: 10.3389/fimmu.2018.00971*

give clues to understand cell functionality in vertebrates and provide knowledge for a better understanding of human pathologies.

Immune innate responses are a first-level of protection against infection and damage, exerted by cells reacting fast to non-self/ injury with their germline-encoded receptors. In mammals, there are different types of innate immune cells, besides macrophages/ dendritic cells/neutrophils, also innate lymphoid cells (ILC) have been described. ILC are classified in three groups for the expression of defined transcription factors, functional characteristics, and phenotype. Another group of mammalian unconventional or innate-like lymphocytes (mILL) has been identified with properties and functions as a bridge between innate and adaptive responses. Populations of mILL might thus rerepresent an "immune lower layer," with activities involved in maintaining gut homeostasis, in early response to intestinal infections, in autoimmune diseases and cancer, in a fast unprimed fight against infection and damage, in producing germline natural polyreactive antibodies and typical cytokine patterns. The mILL have been mainly identified as γδT cells and B1-B cells, are mainly located in mucosal tissues, and their functions and origin are still under scrutiny.

Of note, increasing evidence suggests that conventional fish lymphocytes display some developmental, morphological, and functional features in common with mILL and, very recently, these similarities have attracted attention among immunologists (1). However, studies aimed to clarify the links between fish lymphocytes and mIL are at their infancy.

This review is proposed to summarize the current knowledge on possible similarities between fish lymphocytes and mILL, and use the knowledge to raise the hypothesis that most of fish lymphocytes behave like subpopulations of mILL and, consequently, that mILL subpopulations (γδT cells, B1-B cells) could represent a "lower layer" of extant, evolutionary-related, analogs of fish lymphocytes.

#### MAMMALIAN INNATE-LIKE LYMPHOCYTES

Important players of innate immune activities are the mammalian ILC that derive from a common lymphoid precursor and play a role with effector and regulatory functions in innate immunity and tissue remodeling. The ILC do not have TcR or Ig rearranged receptors in their surface and are classified into three groups on the base of the patterns of cytokines they produce, and of transcription factors necessary to their functions. Namely, ILC1 produce IFNγ and depend on Tbet, ILC2 produce type 2 cytokines (IL-5/ IL-13) and require GATA3, ILC3 depend on RORγt and produce IL-17 and/or IL-22 (2). Also natural killer (NK) cells belong to innate lymphocytes are involved in fast innate responses and do not express CD3 or lymphocyte receptors on their surface. However, aside from the classical description of lymphocytes as cells responsible of adaptive responses, subpopulations of recently discovered mILL behave as innate immune cells with respect to the historical innate-adaptive classification (3).

The mILL are involved in maintaining gut homeostasis and in early response to intestinal infections (4, 5), in autoimmune diseases and cancer (6, 7), are able to combat non-self in an MHC-independent fashion (8, 9), produce unbiased natural polyreactive antibodies (10, 11) and typical cytokine patterns (12, 13).

The main lymphocyte subpopulations displaying innate-like activities in mammals have been identified as γδT cells (5), mucosaassociated invariant T cells (MAIT) (14), natural killer T cells (NKT) (15), B1-B cells (16), and spleen marginal zone B cells (17).

#### Innate-Like T Cells

The γδT lymphocytes are non-conventional T lymphocytes, comprising a minor T cell subset in blood and a major population of intestinal intraepithelial lymphocytes (IELs) having typical morphological features of lymphocytes with a surface germline TCR phenotype of γ+δ+ (mostly displaying repertoires Vδ1/ Cγ1 and Vγ9/Vδ2) and showing a potent phagocytic ability to both soluble and particulate antigens (18, 19). With respect to immunoglobulins and αβTcR molecules, the γδTcR displays the highest spontaneous diversity in the CDR3 region produced by VDJ recombination by using the V-chain gene. The γδT cells can develop extrathymically and independently from an antigen encounter and are active players in adaptive and innate-like immune responses such as the direct killing of infected cells, are involved in tumor immunosurveillance (20), produce molecules required for pathogen clearance (21), are spontaneously cytotoxic (22), release immunomodulatory cytokines (23), and can be activated by stress-induced molecules (MIC-A/B, ULBPs) to produce pro-inflammatory cytokines and lytic enzymes. In summary, evidence suggests that γδT cells act either as effectors and regulators (24), and represent an evolutionarily primitive T cell subset characterized by innate and adaptive immune functions. Supporting these findings, recent data also showed the presence of γδT cells subsets for which innate stimuli are more important than TcR ligation, as in the case of IL-17-producing (γδT-17) and IFNγ-producing (γδT-IFNγ) cells (25).

Other subpopulations of recently discovered mammalian innate-like T lymphocytes are the MAIT and NKT. MAIT are an innate T cell subpopulation (14), principally involved in antibacterial immunity at mucosal surfaces, and mainly present in man than in mouse (26), they display a germline TcRαβ phenotype (Vα7.2-Jα33/12/20 in humans, Vα19-Jα33 in mice) and variable but restricted TcRβ chains (5, 27). Upon stimulation, MAIT produce the regulatory cytokines IFNγ, TNFα, and IL-17, and express the receptors for IL-7, IL-12, and IL-18 (26).

The NKT are a subpopulation of αβ- and γδ-T cells differing from NK cells for the presence of CD3 and TcR, characterized by CD1d restriction and limited TcR diversity (15, 28). They are principally present in non-mucosal tissues, are involved in antitumor activity, and are of help for B cell proliferation and antibody production (29). The NKT can be further divided into two distinct subpopulations, namely, type I and type II NKT cells (30) that are preferentially located in the liver. Type I display a semi-invariant TcR (Vα14Jα18/Vβ2, 7, 8) in mice and (Vα24Jα18/Vβ11) in humans, whereas type II NKT cells exhibit a more diverse TcR repertoire.

#### Innate-Like B Cells

The B lymphocytes of mammals are now cataloged as B2, or classic, and B1, or innate. These two major sets of B cells are defined by differential presence of CD5 in their surface. The B1-B cells are further subdivided in B1a (B1) having a phenotype CD5<sup>+</sup>/IgMhigh/ IgDlow, and B-1b cells, which are CD5-negative (31). The B1-B cells produce large amounts of natural polyreactive antibody in a T cell-independent manner, are actively phagocytic and microbicidal (32), may be involved in autoimmunity (33), and are present as IgA-secreting plasma cells in the intestinal mucosa where they migrate during infections (16). Natural polyreactive antibodies produced by CD5+ B cells are germline-encoded antigen recognition molecules (class IgM, IgA, and IgG3) (11) with a limited repertoire of V-region genes, play an important role in early host defense, in autophagy/tissue remodeling and immune regulation, in recognition of pathogens and activation of the innate immune system *via* the classical pathway of complement activation (10). The B1-B cells are considered to have no memory, are present in mouse liver at fetal stages (34), whereas in adults are present in the spleen and peritoneal cavity (35, 36), where they undergo self-renewal with mechanisms that are poorly understood.

Being involved in innate activities, B1-B cells respond to stimulation *in vitro* through TLRs (from TLR1 to TLR8) (37, 38) inducing B1-B cell proliferation and differentiation into immunoglobulin-secreting cells. Also, B1-B cells show a rapid capacity to produce high amounts of the immunomodulatory cytokine IL-10 after innate activation (13).

An additional subpopulation of B cells having innate-like activities is located in the spleen pulp marginal zone and involved in producing IgM antibodies in a T cell-independent manner against pathogens circulating in blood (17).

Of particular interest is the tissue localization of innate-like B cells, which exert their activities principally in mucosal surfaces and mainly in the intestine, where the IgA produced by B1-plasma cells can be spontaneously present, reacting with the intestinal microflora (39). The mucosal intestine is also the richest site of γδT lymphocytes in adult mice and man (40), followed by the respiratory epithelium (24), and the epidermis (41). In mucosal tissues, during a possible infection the mILL displaying germline receptors can respond quickly, thus providing protection independently from adaptive responses and in the absence of antigen exposure as, for instance, in newborns (5).

#### FISH LYMPHOCYTES

The features of mILL, very briefly summarized above, appear to be remarkably similar to the features of conventional lymphocytes as they are known in teleost fish, where experimental data accumulated in decades of investigation showed the presence of T cells possessing surface αβ- and γδ-TcR, of B cells expressing three immunoglobulin types (IgM, IgT, and IgD), of lymphocyte subpopulations, and a complete set of master genes coding for lymphocyte-associated molecules (42–45). The fish lymphocytes have been shown to be functionally active *in vitro* and *in vivo* (46–52), and to produce and/or be affected by families of lymphocyte-related cytokines (53, 54).

#### Features of Fish T Cells

Two classes of T cells are present in teleost fish, displaying on their cell surface αβ- and γδ-TcR, together with TcR coreceptors, and expressing patterns of genes that clearly indicate the presence of T cell subpopulations as they are known in mammals, namely, cytotoxic (CD8), helper (CD4), and regulatory (Treg, Th17) (45, 55–57). The immunobiology of fish T cells has been the subject of extensive research addressed to investigate regulation mechanisms, expression of surface markers, and *in vitro/vivo* studies, that have been reassumed in recent reviews (42, 53, 54, 58–60). In relation with the present work, available data have shown that the distribution of T cells in fish is principally located in mucosal tissues of intestine and gills (60–66), and that activities of T cells are diverse in these tissues. In the intestine, IEL displays an *in vitro* spontaneous cytotoxic activity (65), proliferate poorly (unpublished), and perform *in vivo* RAG-driven spontaneous somatic rearrangement of a given V/C combination in the CDR3 junction length of TcRβ-chain/TcRγ-chain in the absence of antigen stimulation (64, 67). On the other hand, T cells from the gills are able to proliferate *in vitro* in response to lectins, but RAG expression is negligible (45). These observations suggest that the teleost intestine could be a site of production of T cells, whereas the gills could be a site where T cells are more committed as effectors/helper. A support to the hypothesis that the fish intestine can be a primary producer of T cells comes from data on the development of sea bass immune system, where first antibody-positive T cells are detected in the developing gut before, or at the same time, than in thymus (68, 69). However, definitive knowledge establishing precise timing and tissue of appearance of T cell subpopulations in fish is still missing (70).

The intestine of sea bass displays a high homogeneous expression of TcRα and TcRγ, a low expression of CD4, and differential expression of CD8α and of MHCII showing an increment and a decrease, respectively, toward the terminal part (65). Considering the number of T cells present in the intestinal mucosa, and that purified T cells from the intestine with a pan-T mAb showed enriched expression of RAG-1, TcRα, TcRγ, CD8α, and CD4, it appears evident that the gut can be considered the main lymphoid tissue for T cells in adult fish (43).

Data obtained on *in vitro* activity of fish leukocytes suggest the presence of an IL-2 modulated proliferation of T cells during a mixed-leukocyte reaction (48) and of a MHC-restricted CTL activity (52), suggesting that fish immune cells also display activities comparable to classical T lymphocytes of mammals.

Finally, fish do have memory T cells, identified by the IL-10 modulation of CD8- and CD4-populations responses and proliferation in immunized carp (71). Interestingly, it should be noted that mutant zebrafish engineered for lacking somatic recombination (RAG-1<sup>−</sup>/<sup>−</sup>) are still able to mount a specific protection after bacterial re-exposure (72), and that Atlantic cod lacks CD4 and MHCII in the genome but is protected during immune challenges with pathogens (73). These latter observations suggest that further research is needed in fish to better elucidate functional features of T cells, such as the phagocytic capability of γδT cells (18).

#### Features of Fish B Cells

Production of antigen-specific antibody in fish is known since almost 70 years, and research has shown that fish have B cells expressing three heavy Ig chain classes, namely, IgM, IgT/Z, and IgD, as defined by the expressed genes μ, τ, and δ, respectively (74–76), and of some Ig light chains (e.g., two in catfish, three in zebrafish, MW 25–28 kDa) (77). The IgM are tetrameric in fish (MW 450 kDa) and present systemically in body fluids, where they may be present in serum at high concentration. The IgT/ IgZ are mucosal immunoglobulins produced in a monomeric form (MW 170 kDa), although a non-covalent polymeric IgT association has been observed in trout mucus. The IgD has been studied at molecular level, it is expressed in a monomeric form with a putative MW of 150 kDa, but little is known on its physiological role in fish (78). Likewise T cells, the B cells of fish have been the subject of much research, with results reassumed in comprehensive reviews (44, 79, 80). With respect to the present work, main activities of fish B cells can be summarized as follows: (i) high content of natural serum IgM in unimmunized fish (81–83); (ii) poor increase in IgM affinity after secondary immunization (84, 85); (iii) presence of memory B cells (85); (iv) spontaneous phagocytosis (86); (v) production of pathogeninduced mucosal secretory IgT (evolutionary orthologs of IgA) (49); (vi) presence of kidney lymphocyte precursors similar to mouse spleen B1-B cells (34, 87); and (vii) presence of proliferating B cells in the peritoneal cavity (88); possible expression of TLRs (89). Interestingly, intriguing features regarding B cells have been observed in some fish species as, for instance, the lacking of

pathogen-specific IgM in gadoids after successful immunization against the pathogen (82), and a lack of the whole IgM gene in a coelacanth species (90).

#### Similarities Between Mammalian Innate-Like and Fish Lymphocytes

The principal features of mILL and of fish conventional lymphocytes have been briefly summarized above, and a comparison of possible similarities is shown in **Figure 1** for T cells and in **Figure 2** for B cells. A point of great importance to better understand the evolution of lymphocytes among vertebrates is the definition of the primary tissue(s) of origin and of tissue localization during development. Experimental evidence suggests that the fish intestine may be a primary lymphoid tissue for T cells, which can be detected there even before their appearance in thymus (68, 91–93), and where a T cell selection might be present that differs from thymic T cell selection. The possible thymus-independent origin of T cell subpopulations appears to be conserved until mammals, where γδT cells may derive from human fetal liver and the primitive intestine between 6 and 9 weeks of gestation, as proposed by investigating expression of the δTcR repertoire during human development (94, 95). Indeed, γδT cells play a pivotal role during human intestine development, since preterm infants

with intestinal barrier immaturity, and thus with a reduction in the number of IEL, may develop severe enterocolitis (96).

The importance of liver as a site of possible lymphocytes development emerges from data on the origin of B cells in mammals, where B1a cells have been found to develop in mouse fetal liver, from which they migrate in the spleen, but not in the bone marrow (34). Interestingly, in the mouse spleen, a lymphocyte subpopulation shows flow cytometric morphological features (SP cells) remarkably similar to that of lymphoid SP cells from adult goldfish and zebrafish kidney (87, 97).

The origin of B cells in fish is not clearly defined, in zebrafish the pancreas has been supposed to be a primary site on the base of B cell receptors genes rearrangement (98), whereas in the sea bass kidney a presence of IgM-producing cells has been established by IHC at 55 days post hatching (69). Although it is evident in all fish species investigated that the development of T cells precedes development of B cells, a definitive clarification of a primary site of B cell origin is missing (69).

Another similarity could be found in the transmission of immunity between the female and the developing embryo. In fish, a possible precursor process of the maternal antibody transfer to the fetus through placenta has been observed by the presence of IgM molecules and IgM gene expression in unfertilized eggs and during first embryonic stages (99). In mammals, the B1-B cells are already present at early stages in the extraembryonic yolk sac and continue their development in the fetal liver (100), with the IgM being the predominant class during late gestation and infancy (101).

Other experimental evidence suggests striking similarities between low affinity polyreactive serum natural IgM antibodies produced by mammalian B1-B cells (13) with IgM presence/ responses in fish (80). As in mammals and other investigated vertebrate species, the kinetic of primary antibody response in fish involves IgM but, at variance with mammals, in fish there is neither a class-switch secondary response nor a substantial increase in serum IgM affinity, although specific antibody titers can be observed after immunization. Of note, the protection mechanisms and specificity of antibody responses in fish are far to be fully understood, since some fish species result protected after immunization without producing specific IgM antibody (82). In addition, the IgM can be even totally absent, as discovered in a species lacking completely of IgM genes (90). The possible importance of natural IgM in fish as players in innate immunity emerges from their amount in serum, since the mean concentration of IgM in unimmunized fish (sera from five species, 7.7 mg/ml) (81) is much higher than the mean concentration of IgM in humans (1.3 mg/ml) (102). Considering that fish lack IgG, the higher concentration of natural IgM could contribute to immunity against pathogens in not yet completely understood ways, suggesting that research on natural IgM contribution in innate immunity, and the kinetic of production of specific IgM by B cells upon immunization, may give some clues to understand the physiology of natural antibodies in mammals.

The Ig secreted in/by mucosal tissues are particularly important for pathogen clearance at the boundary with external environment, and teleost fish have a mucosa-associated IgT class whose features, like the coating of intestinal commensal microbiota, precede that of mammalian-specific secretory IgA (49). Although IgT is not homologous to IgA, it is evident a convergent evolution of the two molecules, both are multimeric, predominantly produced in the mucosa, and induced by mucosal immunization (103).

Fish leukocytes express TLRs (90), show strong *in vitro* response to LPS, and respond to flagellin with TLR5 (104), and to viruses and poly I:C with TLR3 (105). Although these responses have been measured in leukocytes, it should be reasonable to speculate that fish B cells should express pathogen-specific conserved TLRs on the base of nucleotide sequences obtained from a transcriptome of head kidney, a B cell lymphopoietic tissue in fish that revealed the presence of several TLRs' gene expression (106). Given the presence of TLRs on fish B cells, a similarity becomes evident with TLRs on mammalian B1-B cells (37, 38).

Another population of fish IgM-B cells is located in the peritoneal cavity, capable of proliferate very soon after antigenic stimulation, produces polyreative antibodies, and is responsible of pathogen clearance (88). Similarly, in mammals the peritoneal B1-B cells can proliferate rapidly after antigen stimulation and can migrate in the periphery, including the intestine, to fight the pathogen (107).

#### SUMMARY

The immune defense system of vertebrates in its molecular and cellular components is remarkably conserved from teleost fish, the more ancient extant representatives of the evolutive lineage that directly brings to mammals. The knowledge on the similarities between morphological and physiological processes of vertebrates led to the use of teleost fish as an additional animal model for investigations in pathology and physiology of immune recognition, with the goal of applying results in translational research for modeling human diseases, as can be easily appreciated with the zebrafish model. Therefore, teleost fish play a fundamental role in understanding the evolution of immune responses of vertebrates, and experimental evidence suggests that some features of mammalian innate-like lymphocytes related to pathogenic conditions, such as chronic lymphocytic leukemia and inflammation could benefit from knowledge in fish lymphocytes.

The hypothesis described in this review is that younger species (mammals) retain immune defense features of ancestors (fish) that have been enriched by evolution with new "layers" of genes coding for cells and molecules, with a "lower" immune layer that in mammals might be composed of cells with innate activities, among which innate-like lymphocytes.

The experimental evidence considered in this review suggests similarities in morphology, gene expression, and functional signatures of fish lymphocytes with mammalian innate-like lymphocyte subpopulations, although much remains to be learned on the immunobiology of fish lymphocytes such as the origin/ functions of intestinal T cells/γδT cells and of B1-B cells.

Considering that MAIT are restricted to mammals, it remains to elucidate the possible presence of NKT in fish, where a clear surface phenotype identification of spontaneously cytotoxic T cells is missing. It also remains to investigate in more details in fish a precise timing and tissues of origin of αβ/γδT cells, of IgM/ IgT B cells, their transcriptomic signatures, and some functional activities like production and kinetic properties of natural polyspecific IgM and phagocytic capability of γδT cells.

Importantly, the hypothesis that subpopulations of mammalian innate-like lymphocytes, namely, γδT cells and B1-B cells could be an extant-like counterpart of fish lymphocytes has been already proposed (32) and supported by a recent publication (1). These works suggest that the origin of lymphocytes, possibly including innate-like lymphocytes, goes back to the origin of all vertebrates (1).

In conclusion, investigations on the development and immunobiology of fish lymphocytes is of great importance in comparative immunology, and possibly important for a better understanding of mammalian innate-like lymphocytes immunobiology and their involvement in human diseases.

#### REFERENCES


#### AUTHOR CONTRIBUTIONS

GS was responsible of organizing and supervising the review work. AF was responsible for manuscript writing. SP was responsible for checking the literature references and databases and organizes figures.

#### FUNDING

Experimental data from authors are from EU-funded projects: 5FP project QLK2-CT-2000-01076 FISHAID, 6FP project CT-2005- 007103 IMAQUANIM, 7FP project 311993 TARGETFISH.


origins and primordial roles of CD4+ lymphocytes and CD4+ macrophages. *J Immunol* (2016) 196:4522–35. doi:10.4049/jimmunol.1600222


(*Dicentrarchus labrax*): quantitation of gene expressing and immunoreactive cells. *Fish Shellfish Immunol* (2017) 63:40–52. doi:10.1016/j.fsi.2017. 02.002


**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 Scapigliati, Fausto and Picchietti. 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.*

# Modulation of Innate Immune-Related Genes and Glucocorticoid Synthesis in Gnotobiotic Full-Sibling European Sea Bass (*Dicentrarchus labrax*) Larvae Challenged With *Vibrio anguillarum*

#### *Felipe E. Reyes-López1†, Johan Aerts2,3†, Eva Vallejos-Vidal1 , Bart Ampe4 , Kristof Dierckens5 , Lluis Tort1 \* and Peter Bossier <sup>5</sup> \**

#### *Edited by:*

*Brian Dixon, University of Waterloo, Canada*

#### *Reviewed by:*

*Beatriz Novoa, Consejo Superior de Investigaciones Científicas (CSIC), Spain Mark D. Fast, Atlantic Veterinary College, Canada*

#### *\*Correspondence:*

*Lluis Tort lluis.tort@uab.es; Peter Bossier peter.bossier@ugent.be*

*† 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: 23 December 2017 Accepted: 12 April 2018 Published: 08 May 2018*

#### *Citation:*

*Reyes-López FE, Aerts J, Vallejos-Vidal E, Ampe B, Dierckens K, Tort L and Bossier P (2018) Modulation of Innate Immune-Related Genes and Glucocorticoid Synthesis in Gnotobiotic Full-Sibling European Sea Bass (Dicentrarchus labrax) Larvae Challenged With Vibrio anguillarum. Front. Immunol. 9:914. doi: 10.3389/fimmu.2018.00914*

*1Department of Cell Biology, Physiology and Immunology, Universitat Autònoma de Barcelona, Bellaterra, Spain, 2Stress Physiology Research Group, Faculty of Pharmaceutical Sciences, Ghent University, Ostend, Belgium, 3Stress Physiology Research Group, Animal Sciences Unit, Flanders Research Institute for Agriculture, Fisheries and Food, Ostend, Belgium, 4Biostatistics and Data Modeling, Animal Sciences Unit, Flanders Research Institute for Agriculture, Fisheries and Food, Melle, Belgium, 5 Laboratory of Aquaculture & Artemia Reference Center (ARC), Ghent University, Gent, Belgium*

Although several efforts have been made to describe the immunoendocrine interaction in fish, there are no studies to date focusing on the characterization of the immune response and glucocorticoid synthesis using the host–pathogen interaction on larval stage as an early developmental stage model of study. Therefore, the aim of this study was to evaluate the glucocorticoid synthesis and the modulation of stress- and innate immune-related genes in European sea bass (*Dicentrarchus labrax*) larvae challenged with *Vibrio anguillarum*. For this purpose, we challenged by bath full-sibling gnotobiotic sea bass larvae with 107 CFU mL−<sup>1</sup> of *V. anguillarum* strain HI 610 on day 5 post-hatching (dph). The mortality was monitored up to the end of the experiment [120 hours post-challenge (hpc)]. While no variations were registered in non-challenged larvae maintained under gnotobiotic conditions (93.20% survival at 120 hpc), in the challenged group a constant and sustained mortality was observed from 36 hpc onward, dropping to 18.31% survival at 120 hpc. Glucocorticoid quantification and expression analysis of stress- and innate immunityrelated genes were carried out in single larvae. The increase of cortisol, cortisone and 20β-dihydrocortisone was observed at 120 hpc, although did not influence upon the modulation of stress-related genes (*glucocorticoid receptor 1* [*gr1*]*, gr2*, and *heat shock protein 70* [*hsp70*]). On the other hand, the expression of *lysozyme*, *transferrin*, and *il-10* differentially increased at 120 hpc together with a marked upregulation of the pro-inflammatory cytokines (*il-1β* and *il-8*) and *hepcidin*, suggesting a late activation of defense mechanisms against *V. anguillarum*. Importantly, this response coincided with the lowest survival observed in challenged groups. Therefore, the increase in markers associated with glucocorticoid synthesis together with the upregulation of genes associated with the anti-inflammatory response suggests that in larvae infected with *V. anguillarum* a pro-inflammatory response at systemic level takes place, which then leads to the participation of other physiological mechanisms at systemic level to counteract the

**68**

effect and the consequences of such response. However, this late systemic response could be related to the previous high mortality observed in sea bass larvae challenged with *V. anguillarum*.

Keywords: cytokines, iron regulation, cortisol, gnotobiotic system, European sea bass, fish larvae, *Vibrio anguillarum*

#### INTRODUCTION

The innate immune response plays a pivotal role in the activation of the host defense mechanisms and determines the nature of the adaptive immune response (1). The components of the innate immune system are divided into physical (mucus layer, which acts as a physical and chemical barrier) (2); cellular (phagocytic cells such as neutrophils and monocytes/macrophages); and humoral factors (based on pattern-recognition specificities or effector functions) (3). In this framework, the presence of receptors able to activate pathways responsible for the cell signaling cascade are crucial to promote a pro-inflammatory reaction, modulating the innate and adaptive immune response (4). This function is developed by pattern-recognition receptors (PRRs), among others by soluble PRRs such as pentraxins (5). Pentraxins (C-reactive protein and serum amyloid protein) are a phylogenetically conserved superfamily of proteins characterized by the presence of around 200 amino acid–pentraxin domain in their carboxy-terminal region. These proteins are considered lectins acting as a non-redundant component of the humoral arm of innate immunity mediating agglutination, complement activation, and opsonization (6).

Previous reports have shown the expression of antibacterialrelated components at the time of hatching and the following weeks (7, 8), indicating that these molecules play a relevant regulatory and effector function to prepare the host against potential environmental pathogens that may be encountered during the early larval development stage when fish immunity is not fully mature. Lysozyme is a bacteriolytic enzyme that hydrolyzes the β-[1,4]-glycosidic linkage of bacterial cell wall peptidoglycans. Its expression on sea bass (*Dicentrarchus labrax*) larvae has been detected at 24 hours post-hatching (9). Another protein involved in the innate humoral response is transferrin. This iron-binding blood plasma glycoprotein controls the level of free iron in biological fluids. Transferrin has a bacteriostatic activity, a property assigned to its iron-binding function (10), thus chelating the available iron necessary for bacterial growth. Transferrin is also considered an acute phase protein (APP) which acts in the inflammatory response to remove iron from the bloodstream (11). Hepcidin, a liver-produced hormone that constitutes the main circulating regulator of iron absorption and distribution across tissues (plasma and intestine) and cells (macrophages, erythrocytes, and hepatocytes) has also been associated with antifungal and antibacterial activity through binding to cell walls (12). Taken together, these proteins share the role of being responsible for iron homeostasis and also participating in the immunity against pathogens.

A key mechanism in the initiation of the antibacterial response is mediated by the expression of pro-inflammatory cytokines involved in the upregulation of inflammatory reactions produced predominantly by activated macrophages. Among them, interleukin-1β (IL-1β) is an endogenous pyrogen produced and released at the early stage response following infections, and subsequently considered as initiator of the pro-inflammatory response in macrophages, activator of lymphocytes and also a synthesis promoter of other cytokines and prostaglandins (13, 14). Interleukin-8 (*IL-8*) is a chemokine produced by macrophages and somatic cells whose primary function is to serve as chemoattractant for neutrophils and T cells to the site of infection. It is also involved in phagocytosis and respiratory burst. Another crucial pro-inflammatory chemokine is chemokine (C–C motif) ligand 4 (CCL4), better known as macrophage inflammatory protein-1β, involved in infection. In mammals, it is produced mainly by macrophages, dendritic cells, and lymphocytes (15).

The pro-inflammatory response is strictly controlled by antiinflammatory cytokines which regulate the cytokine expression, immune cell proliferation and promote tissue repair (16, 17). The role of interleukin-10 (*IL-10*) has been widely described as a potent anti-inflammatory cytokine that regulates the expression of pro-inflammatory cytokines, contributing to the pathogen infection resolution and also reducing tissue damage caused by inflammation (18, 19). Thus, a highly regulated balance between pro- and anti-inflammatory cytokines makes a successful immune response against pathogens.

Vertebrates under stressful stimuli launch an endocrine stress response. This response comprises an immediate adrenaline response which prepares the organism for the "fight or flight" reaction by increasing plasma glucose levels and activating cardiovascular responses. In addition, glucocorticoids, in particular cortisol or corticosterone depending on the species, are released through activation of the hypothalamic–pituitary–interrenal (HPI) axis in fish or hypothalamic–pituitary–adrenal axis in other vertebrates (20–22). These plasma glucocorticoids are generally accepted as biomarkers for stress. They mediate a redistribution of energy with the ultimate goal to restore pre-stress conditions and homeostasis, acting as adaptive hormones. Failure to regain homeostasis will lead to chronic stress and maladaptation and eventually to higher susceptibility to pathogens and disease. Thus, when an organism cannot fully recover an allostatic overload is imposed thus becoming prone to detrimental effects of glucocorticoid mediated actions (e.g., immune suppression, decreased growth, impaired reproduction, increased mortality), leading to "distress" and difficulty to regain the set-points for eustress and homeostasis (23, 24). The main corticosteroid hormone in teleosts is cortisol (25) which has become the most common physiological indicator of stress in fish (22). The receptors for cortisol are the glucocorticoid receptor (GR) and mineralocorticoid receptor (MR) (26). In several fish species, including sea bass (27), two GR isoforms (GR1 and GR2) have been reported (28, 29). Importantly, the two GR isoforms differ in glucocorticoid sensitivity and affinity, being GR2 more sensitive to lower concentrations of cortisol (28, 30).

Cortisol is secreted by head kidney interrenal cells from a concatenated response involving the corticotrophin-releasing hormone (CRH) and the adrenocorticotropic hormone (ACTH), produced and secreted in the brain hypothalamic cells and anterior pituitary, respectively. Cortisol is synthesized from cholesterol by its conversion to pregnenolone catalyzed by P450 side chain cleavage, the rate-limiting enzyme in steroidogenesis. Pregnenolone is the precursor of progesterone synthesis that, catalyzed by 17α-hydroxylase, mediates the conversion to 17α-hydroxyprogesterone (17α-HP). The enzyme 21-hydroxylase catalyzes its conversion to 11-deoxycortisol, and 11β-hydroxylase is the enzyme responsible for the terminal step catalyzing the conversion from 11-deoxycortisol to cortisol (31–33). The main cortisol phase I metabolites are cortisone and 20β-dihydrocortisone. In fact, large circulating concentrations of cortisone have been detected in fish subjected to stress (34). It has been proposed that the conversion from cortisol to cortisone, mediated by 11β-hydroxysteriod dehydrogenase type 2, serves to downregulate cortisol into a non-active keto-metabolite. However, this hypothesis has been less studied to date (24).

It has been extensively described that stress and immune response are tightly connected (35, 36). Many studies have reported that a stressor induces alterations on innate immune response (37–39). Thus, it is widely accepted that cortisol inhibits the release of pro-inflammatory cytokines (40–42), probably as a mechanism for control and resolution of inflammation. Nowadays, as far as we know, there are no studies focused on host–pathogen interaction in which an association between glucocorticoid synthesis and immune response has been evaluated in larvae after a challenge with a highly pathogenic bacterium. In this study, the use of a gnotobiotic culture system allows to evaluate the specific and direct interaction between the microorganism of interest and the host, thus avoiding the intrinsic microbial community dynamics (43, 44). Therefore, the aim of this study was to evaluate the glucocorticoid profile and the modulation of stress- and innate immune-related genes in gnotobiotic full siblings in specific time points related to mortality of European sea bass larvae challenged with *Vibrio anguillarum*. The kinetics of key actors associated with the innate immune response will be useful to understand the ability of fish larvae to activate immune protective-related mechanisms against *V. anguillarum* and the endocrine-mediated response by glucocorticoid synthesis as an orchestrated systemic response at early larvae stage and under controlled microbial experimental conditions.

#### MATERIALS AND METHODS

#### Bacterial Strains and Culture Conditions

The rifampicin-resistant *V. anguillarum* strain HI 610 (O. Bergh, Institute of Marine Research, Norway), isolated by natural selection, was used in this study. The rifampicin resistance allows this strain to survive in an axenic sea bass larvae culture water environment. The bacteria were grown overnight at 28°C on 10% marine broth (Difco Laboratories) with NaCl to obtain the same salinity as the water in the fish larvae experiment (36 g L<sup>−</sup><sup>1</sup> ) and supplemented with 10 ppm rifampicin (Sigma). The bacterial suspension was then centrifuged at 1,500 × *g* for 10 min at 4°C and resuspended in distilled water added with Instant Ocean artificial sea salt (Aquarium Systems) to obtain a salinity of 36 g L<sup>−</sup><sup>1</sup> , and supplemented with 10 ppm rifampicin and kept on a horizontal shaker at 150 rpm at 16°C. The bacterial suspension density was determined spectrophotometrically (Genesys 20, Thermospectronic) at 550 nm according to the McFarland standard (BioMérieux).

#### Challenge Tests With *V. anguillarum* and Gnotobiotic Full-Sibling Sea Bass Larvae

Upon arrival, sea bass eggs were acclimatized in UV-sterilized seawater for 4 h in a cylindro-conical tank. The water temperature (16 ± 1°C) and salinity (36 g L<sup>−</sup><sup>1</sup> ) were kept constant during the experiment. The disinfection of eggs, hatching and axenity tests were performed according to Dierckens et al. (43). All larvae analyzed in this study belonged to the same full-sibling family batch. A summary of the experimental setup with the assays and time points evaluated is given in **Figure 1**. On day 3 dph, full-sibling larvae were stocked in groups of 12 larvae in 10 mL sterile screw cap vials with the addition of 10 mg L<sup>−</sup><sup>1</sup> rifampicin. Three vials (replicates) were prepared for each treatment and time point included in the study. Fullsibling larvae were challenged by bath with 107 CFU mL<sup>−</sup><sup>1</sup> of *V. anguillarum* strain HI 610 on 5 dph in a gnotobiotic system. The mortality was monitored in all vials by counting the living larvae (transparent and swimming) under a dissecting microscope at 24 hours before challenge (hbc) and at 0, 18, 24, 36, 48, 72, 96, and 120 hpc. As control group, uninfected larvae (mock-challenged) were used, and whole-body larva sampling (10 larvae per condition) was performed at the same time points mentioned above. Larvae were not fed during the experiment. Larvae were collected, snap-frozen in liquid nitrogen and kept at −80°C until analysis. Embryos were sacrificed by overanesthetization using methylsulfonatetricaine (MS-222) (Sigma) and immediately sampled. After sampling, larvae were immediately frozen in liquid nitrogen and stored at −80°C until analysis.

#### Ethics Statement

All experiments were approved by the Ethical Committee of the Faculty of Veterinary Medicine and the Faculty of Bioscience Engineering, Ghent University (no. EC2015\_02) and carried out in accordance with the recommendations of the European Union Ethical Guidelines for experimental animal care and other scientific purposes (2010/63/EU).

#### Glucocorticoid Quantification

As the pertinent literature lacks a method for analyzing a full glucocorticoid profile in a single fish larva (whole body), a recently validated ultra-performance liquid chromatography coupled to tandem mass spectrometry (UPLC–MS/MS) quantification method was followed (44). Shortly, a single sea bass larva was sampled, rinsed with ultrapure water, dried on a paper tissue, and subsequently weighed. The larva was homogenized and HPLC-gradient grade methanol (VWR) was used as extraction solvent. Purification was

done using GracePure™ SPE C18-Max 500 mg/6 mL solid-phase extraction (SPE) columns. After resuspension, UPLC–MS/MS was used to quantify the glucocorticoid profile for the active hormone cortisol, its precursors (17α-HP and 11-deoxycortisol), and phase I metabolites (cortisone, 20β-dihydrocortisone, tetrahydrocortisol, and tetrahydrocortisone).

Single whole-body larva samples (*n* = 10 per time point and experimental condition) were taken at 24 hbc from non-treated larvae, and also at 0, 1, 6, 24, 36, and 120 hpc from the non-challenged (mock-infected) and challenged larvae. Glucocorticoid quantification was carried out at 24 hbc to evaluate whether larval manipulation induced stress by handling at 24 h after finishing the stocking process. The glucocorticoid quantification was also carried out at 0, 1, 6, 24, 36, and 120 hpc to evaluate the potential capability of bacteria to induce an acute stress response (0, 1, 6, and 24 hpc) and based on previous studies conducted in European sea bass that showed the cortisol peak at 1 h after exposure to an acute stressor (27) as well as in studies conducted in larvae of other fish species such as rainbow trout (1 h post stress) (45) and red drum (1 h post stress) (46). The analysis of 24, 36 and 120 hpc time points was also used to evaluate the modulation of stress- and immune-related genes.

## RNA Isolation and Complementary DNA (cDNA) Synthesis

Total RNA was isolated from a single whole larva using TRI reagent (Sigma) according to the manufacturer's instructions with some modifications. RNA was precipitated from the aqueous solution with 2-propanol (Sigma) in presence of 10 µg of glycogen (Sigma) and incubated for 1 h at −80°C. The RNA pellet was dissolved in 10 µL of nuclease free-water and immediately stored at −80°C until use. The RNA concentration was determined using a NanoDropND-2000 spectrophotometer (Thermo Scientific), and the integrity was checked by Experion RNA StdSens analysis (Bio-Rad Laboratories). Samples with an RNA quality indicator number greater than 8.0 were chosen for gene expression analysis. Total RNA (500 ng) was used as template to synthesize cDNA using iScript cDNA kit (Bio-Rad Laboratories) according to the manufacturer's instructions.

#### Gene Expression Analysis

Real-time PCR assay was carried out to analyze the expression pattern of different stress and immune relevant genes in gnotobiotic full-sibling European sea bass larvae challenged with *V. anguillarum*. Samples (*n* = 10) were taken randomly at 0 hpc from non-treated larvae, and samples were taken both from the non-challenged (mock-infected) (*n* = 10) and challenged (*n* = 10) larvae at 18, 24, 36, and 120 hpc. Several reference candidate genes [ribosomal protein l13 (*rpl13*), elongation factor 1α (*ef1*α), and 40S ribosomal protein SA (*rpsa*)] were tested using the BestKeeper software (47). According to previous antecedents on sea bass infected with *V. anguillarum* (48), the *rpl13* was chosen in our study as reference gene because its lower variation tested upon all the samples included in our study. Specific primers used for gene expression analysis (**Table 1**) were designed with Primer-Blast. The potential primer secondary structures and primer specificity and was checked with OligoAnalyzer (version 3.1) and Primer-Blast, respectively. Real-time PCR reactions were performed with iTaq universal sybr green supermix (Bio-Rad Laboratories) using 2.5 µL of 1:40 dilution of cDNA (for genes of interest) or 2.5 µL of 1:1,000 dilution (for reference gene). Primers

Table 1 | Primers used in real-time PCR for gene expression analysis.

for all genes were used at 500 nM. RNA sample mix (prepared from 1 µL of each total RNA sample stock), cDNA synthesis mix, and MilliQ-water (same than used to prepare the mix for all primers evaluated) were used as real-time PCR internal amplification controls. The thermal conditions used were as follows: 3 min at 95°C of pre-incubation followed by 40 cycles at 95°C for 30 s and 60°C for 30 s. An additional temperature ramping step was utilized to produce melting curves from 65 to 95°C to verify amplification of a unique single product on all samples. All the reactions were performed in duplicate using a CFX384 Touch Real-Time PCR Detection System (Bio-Rad Laboratories). The quantification was done according to Pfaffl method corrected for efficiency of each primer set (49). The value for each experimental condition was expressed as normalized relative expression, calculated in relation to values of control group and normalized against those of the reference gene. The results were expressed as average of values obtained at 0, 18, 24, 36, and 120 hpc (*n* = 10 larvae per condition group and time point).

#### Statistical Analysis

Cumulative survival was analyzed by a linear regression model with Bonferroni posttest. Gene expression analysis of immunerelated genes in whole-body larva and the glucocorticoid screening was analyzed by a linear regression model with treatment,

Gene Primer sequence (5**′**–3**′**) Accession number Amplicon size Efficiency *pentraxin* Fw: 5′-AGTTTTTGCTGCTGGTGGTG-3′ Rv: 5′-GCCAAAGAGAAAAGGACGTGG-3′ EU660933.1 199 1.95 *lysozyme* Fw: 5′-TGATGCAGGTTGTTGATGTTAATC-3′ Rv: 5′-TCCATCCCCCATATTGTAGGC-3′ KJ433681.1 194 1.94 *transferrin* Fw: 5′-GCCCCCAAACACAGATTCCT-3′ Rv: 5′-CCGTCAGCACCCATACTGTT-3′ FJ197144.1 177 1.95 *hepcidin* Fw: 5′-GGAATCGTGGAAGATGCCGT-3′ Rv: 5′-CAGACACCACATCCGCTCAT-3′ DQ131605.1 108 1.86 Interleukin (*il*)*-1β* Fw: 5′-ATCTGGAGGTGGTGGACAAA-3′ Rv: 5′-AGGGTGCTGATGTTCAAACC-3′ AJ269472.1 106 2.02 *il-8* Fw: 5′-GTCTGAGAAGCCTGGGAGTG-3′ Rv: 5′-GCAATGGGAGTTAGCAGGAA-3′ AM490063.1 110 1.98 Chemokine (C–C motif) ligand 4 (*ccl4*) Fw: 5′-TCCTCGTCTCACTCTGTCTGT-3′ Rv: 5′-GACCTGCCACTGTCTTCAGC-3′ AM490064.1 197 1.95 *il-10* Fw: 5′-CGACCAGCTCAAGAGTGATG-3′ Rv: 5′-AGAGGCTGCATGGTTTCTGT-3′ AM268529.1 199 2.06 Glucocorticoid receptor 1 (*gr1*) Fw: 5′-GAGATTTGGCAAGACCTTGACC-3′ Rv: 5′-ACCACACCAGGCGTACTGA-3′ AY549305 401 1.92 Glucocorticoid receptor 2 (*gr2*) Fw: 5′-GACGCAGACCTCCACTACATTC-3′ Rv: 5′-GCCGTTCATACTCTCAACCAC-3′ AY619996 403 1.97 Heat shock protein 70 (*hsp70*) Fw: 5′-GCTCCACTCGTATCCCCAAG-3′ Rv: 5′-ACATCCAGAAGCAGCAGGTC-3′ AY423555.2 172 1.94 Ribosomal protein l13 (*rpl13*) Fw: 5′-AAGGGAGACAGCACTGAGGA-3′ Rv: 5′-TGCCAAAAAGACGAGCGTTG-3′ DQ836931.1 175 2.00 Elongation factor 1α (*ef1*α) Fw: 5′-AACTTCAACGCCCAGGTCAT-3′ Rv: 5′-CTTCTTGCCAGAACGACGGT-3′ AJ866727.1 144 1.97 40S ribosomal protein SA (*rpsa*) Fw: 5′-TGATTGTGACAGACCCTCGTG-3′ HE978789.1 79 1.93

Rv: 5′-CACAGAGCAATGGTGGGGAT-3′

time, and their interaction as fixed effects. The outcomes were log-transformed to obtain normality. Normality of the analyzed outcomes was evaluated based a graphical examination of the residuals. In all analyses, a *p*-value < 0.05 was considered statistically significant. In case of significant effects, a *post hoc* test for treatment (with Tukey correction) was performed (at each time point). The analysis was carried out with SAS 9.4 for Windows.

#### RESULTS

### Cumulative Survival in Gnotobiotic Full-Sibling European Sea Bass Larvae Challenged With *V. anguillarum*

To evaluate whether *V. anguillarum* challenge had an effect on larval viability, the cumulative survival was recorded throughout the study (**Figure 2**). Almost no variations were registered in nonchallenged larvae maintained under gnotobiotic conditions, whose cumulative survival decreased to 93.20% at 120 hpc. By contrast, a gradual decline on larval survival was observed from the moment of exposition to the bacterial pathogen (93.94% survival) to 24 hpc (88.52% survival). This decrease in larval survival ceased between 24 and 36 hpc. However, from 36 hpc onward, a constant and sustained drop in the cumulative survival was observed in larvae challenged with *V. anguillarum* resulting in a significant difference at 48 (76.31% survival), 72 (61.05% survival), 96 (39.68% survival), and 120 hpc (18.31% survival) compared with non-challenged larvae. This indicates that the challenge with *V. anguillarum* provokes a systemic failure of the larval defense systems resulting in death when cultured under gnotobiotic conditions.

#### Glucocorticoid Quantification in Gnotobiotic Full-Sibling European Sea Bass Larvae Challenged With *V. anguillarum*

To evaluate whether the challenge with *V. anguillarum* had a stimulatory secretory effect on either cortisol, its precursors, and the most important phase I metabolites, a glucocorticoid profile

was quantified in whole body of a single larva challenged with *V. anguillarum*. In addition, the glucocorticoid profile was evaluated at 24 hbc to determine whether larval manipulation induced stress by handling after finishing the stocking process (**Figure 3**). No variations were recorded on the synthesis of the glucocorticoid precursors, 17α-HP and 11-deoxycortisol; neither at 24 hbc nor in the non-challenged or challenged groups. Cortisol synthesis showed no significant differences in both experimental groups in the first 36 hpc. However, a marked time-dependent increase was observed in larvae challenged with *V. anguillarum* at 120 hpc (29.849 ± 8.102 μg L<sup>−</sup><sup>1</sup> ) compared with challenged larvae at 36 hpc (3.919 ± 0.654 μg L<sup>−</sup><sup>1</sup> ). The cortisol mean value at 120 hpc in challenged larvae was also higher than non-challenged larvae at the same time (5.099 ± 1.077 μg L<sup>−</sup><sup>1</sup> ), whose cortisol mean value remained unaltered at all time points evaluated. Because of the high level of cortisol at 120 hpc in the challenged group, we also evaluated the synthesis of phase I metabolites to obtain an even more detailed view on cortisol metabolism. In concordance with the cortisol value, cortisone showed also the highest level at 120 hpc in the challenged group (8.478 ± 1.288 μg L<sup>−</sup><sup>1</sup> ) compared with challenged larvae at 36 hpc (3.089 ± 0.148 μg L<sup>−</sup><sup>1</sup> ). This cortisone value at 120 hpc in the challenged group was also higher than in the non-challenged group at the same time point (4.621 ± 0.566 μg L<sup>−</sup><sup>1</sup> ). As in the case of the challenged group, cortisone in the non-challenged group was higher at 120 hpc than 36 hpc (2.871 ± 0.277 μg L<sup>−</sup><sup>1</sup> ). On the other hand, the synthesis of 20β-dihydrocortisone did not vary in the non-challenged group, observing only an increase at 120 hpc (1.813 ± 0.071 μg L<sup>−</sup><sup>1</sup> ) compared with 36 hpc (1.485 ± 0.036 μg L<sup>−</sup><sup>1</sup> ) in the challenged group. Tetrahydrocortisol and tetrahydrocortisone were absent in all samples, indicating that the latter were free of contamination caused by exogenous glucocorticoids from hands, water, etc. The total glucocorticoid concentration, being the sum of all glucocorticoids quantified, at 120 hpc (43.477 ± 9.299 μg L<sup>−</sup><sup>1</sup> ) increased compared with 36 hpc in the challenged group (11.365 ± 0.783 μg L<sup>−</sup><sup>1</sup> ), and at 120 hpc in the non-challenged group (14.331 ± 1.614 μg L<sup>−</sup><sup>1</sup> ). In all, this suggests an increased synthesis of cortisol at 120 hpc in sea bass larvae challenged with *V. anguillarum* whereby the glucocorticoid level is most likely correlated with the circulating concentrations of cortisone, affecting in sum to the total glucocorticoid circulating concentrations in challenged sea bass larvae.

#### Gene Expression in Gnotobiotic Full-Sibling European Sea Bass Larvae Challenged With *V. anguillarum*

The expression profile was analyzed to evaluate the time-dependent modulation of stress- and innate immune-related genes in larvae challenged and non-challenged with *V. anguillarum*. The expression of stress-related genes was evaluated in sea bass challenged with *V. anguillarum* (**Figure 4**). The glucocorticoid receptor *gr1* did not vary its expression when the mRNA abundance between the non-challenged and challenged groups was compared in all time points tested. Importantly, no modulations were either registered for *gr2*. The expression of *hsp70* was not either modulated on sea bass larvae when exposed to *V. anguillarum*, thus remaining the

expression at basal level in all time points evaluated. These results indicate that at 120 hpc cortisol did not exert any modulatory effect on the expression of gr1, gr2, and hsp70 in sea bass larvae challenged with *V. anguillarum* at those time points.

The expression of genes associated with the humoral arm of the innate immunity was evaluated (**Figure 5**). *pentraxin* showed no significant differences between challenged and nonchallenged larvae along the time points evaluated in this study. The expression of lysozyme showed no variations during the first 36 hpc; by contrast, the challenged group showed upregulation of lysozyme at 120 hpc compared with the non-challenged group. The expression of *transferrin* remained unaltered up to 36 hpc but increased at 120 hpc compared with the control group. Importantly, the expression of *hepcidin* was markedly augmented (1,027-fold increase) in the group challenged with *V. anguillarum*. In all, these results suggest the increase in the expression of genes associated with bacterial clearance at 120 hpc in sea bass larvae challenged with *V. anguillarum*.

To evaluate whether the upregulation of genes associated with antibacterial activity was correlated with the expression pro-inflammatory genes, the expression of *il-1β* and *il-8* was analyzed. According to the innate immune-related genes, the same expression pattern was observed both in *il-1β* and *il-8* observing a marked increase at 120 hpc in the challenged group. The same trend, although not significant, was observed at 120 hpc in *ccl4*. On the other hand, the expression of *il-10*, an anti-inflammatory cytokine involved in the pro-inflammatory outcome control, showed no variations during the first 36 hpc but its expression increased at 12 hpc in larvae challenged with *V. anguillarum*. Hence, these results suggest that both the pro-inflammatory response and its regulatory mechanism took place at 120 hpc in sea bass larvae challenged with *V. anguillarum*.

#### DISCUSSION

In this study, we evaluated the modulation of innate immunerelated genes and glucocorticoid synthesis in gnotobiotic fullsibling European sea bass larvae challenged with *V. anguillarum*. Importantly, significant differences in the expression of innate immune-related genes and glucocorticoid profile in gnotobiotic full-sibling European larvae were only observed at 120 hpc, coinciding with the highest difference in the cumulative survival between challenged and non-challenged larvae. Taking together, these results suggest an immune-endocrine association in response to the challenge with *V. anguillarum*. This is the first report in which such response is shown during the first developmental fish stages.

The gnotobiotic system involves absence of cultivable bacteria, thus allowing the specific study of the host–microbial interaction. This larval rearing system avoids the intrinsic dynamics of the conventional environment in terms of microbial community interactions (43). Subsequently, the use of gnotobiotic conditions provides information about the specific and direct interaction between an external bacterial stimulus and the host (50). Thus, the gnotobiotic culture system is an important alternative to solve the problems of a reproducible experimental setup by generating high inter-individual and inter-batch variability in the composition of the standing microbial community (51).

Infectious diseases are a huge problem in aquaculture, causing important economic losses due to high mortality. Vibriosis, an infectious disease whose etiological agent is *V. anguillarum*, has been considered one of the most important mariculture fish diseases (52). Several studies have reported the susceptibility of sea bass to *V. anguillarum* (53, 54). In our results, no significant variations were observed in the cumulative survival in nonchallenged groups along the study. By contrast, the challenge with *V. anguillarum* provoked a gradual decline of the sea bass larvae cumulative survival in the first hours post-challenge that then seemed stabilized at 36 hpc. However, a marked and strongest decrease in cumulative survival was also observed from 36 hpc up to the end of the experiment. The low survival of challenged larvae obtained in our study compared with higher survival observed in previous reports in sea bass larvae challenged with *V. anguillarum* strain HI 610 under gnotobiotic conditions (43) could be associated with the more limited larval genetic variability. It may be considered that all the individuals are full-sibling in our study, and hence, a higher effect is recorded as the fullsibling fish are more susceptible to *V. anguillarum*. In teleosts, it has been demonstrated that the same pathogen can differentially affect the cumulative survival on several full-sibling groups, and these differences may reside in their particular ability to mount an efficient immunological response against the pathogen (55).

The expression of genes related to the adaptive immune response has been reported from few days post-hatching onward (8, 50, 56, 57). At cellular level, it is accepted that T cells are the first lymphoid-cell type to appear during fish ontogenesis. The presence of thymocytes in sea bass was detected by immunocytochemistry primarily from 30 dph in thymus (58) and gut (59), while earlier T cells detection by flow cytometry were described from 5 to 12 dph onward (60). However, during this developmental life stage, thymus is not yet a differentiated lymphoid organ (58, 61), suggesting a very early/pre-T cells that could represent a cell type different from the large granular lymphocytes (62). On the other hand, mature B-cells and immunoglobulin M (IgM) were detected in sea bass from 50 dph onward (61, 62), although the presence of maternal IgM has been observed in the sea bass eggs and embryos (61, 63). Thus, the presence of an adaptive (and therefore a mature) immune system can take place from

50 dph onward. These antecedents would indicate that fish depend mainly on innate defense, at least during the first days post-hatching (64–68). Moreover, the presence of innate immune components such as cathepsin, lectins, and lysozyme have been detected at very early stages in oocytes, fertilized eggs, and larval stages of several fish species including sea bass (9, 69, 70).

The innate immune system is of key importance in combating infections by microorganisms in lower vertebrates, particularly under poikilothermic conditions (3). Thus, the modulation of genes associated with innate immunity seems to be of primary importance in the sea bass larvae response against *V. anguillarum*. In our study, the evaluation of genes associated with the stress and the innate immune response was based on the analysis of the most critical survival time points on sea bass larvae challenged with *V. anguillarum* (1): the beginning of a reduction on the cumulative survival and (2) the lowest cumulative survival time point. It has previously been reported that the evaluation of the first time points of infection is critical for fish immunity and, in fact, may even allow to classify phenotypes of responses based on the immune gene expression patterns (55). In the context of innate immunity, the expression of *pentraxin* was evaluated. Pentraxins are lectins present in the body fluids and are associated with the acute phase response (APR) by its role in mediating agglutination, complement activation, and opsonization (6, 71, 72). Our results showed no variation on *pentraxin* gene expression. In teleosts, the upregulation of *pentraxin* has been reported in juvenile ayu following *V. anguillarum* intraperitoneal challenge (73). However, it has been proposed that the level of pentraxin may or may not be elevated during an APR (3). Thus, *pentraxin* expression in the sea bass larvae evaluated in our study seems not to be affected by *V. anguillarum*, suggesting that other PRRs and actors of the innate immunity humoral arm could be implicated in the recognition, and therefore in the response against *V. anguillarum*.

Other proteins involved in the innate humoral innate response were also evaluated. Lysozyme is an enzyme able to hydrolyze the β-[1,4]-glycosidic bond present on peptidoglycans bacterial cell wall. The expression of lysozyme has been detected at 24 hph (9) probably as a primary defense mechanism previous to the first exogenous feeding. In our results, the expression of lysozyme was augmented at 120 hpc, indicating that probably an early recognition of *V. anguillarum* by the host is not taking place. In the same direction, the gene expression of molecules involved in the iron homeostasis regulation such as *transferrin* (with bacteriostatic activity) (10) and *hepcidin* (with antimicrobial activity) (12) were also upregulated at 120 hpc. The fight for iron availability in the bloodstream is crucial to establish the ability of the host to overcome a pathogen threat. Iron is critical for all bacteria to grow and determines the infective success. Thus, the host iron regulation is a key step directly associated with defense mechanisms mounted in response to a bacterial challenge. In our results, the upregulation of *transferrin* and *hepcidin* at 120 hpc could be directly related with one of the strategies responsible for counteracting the mortality observed at 120 hpc. Previous studies have shown that synthetic *hepcidin* induced protection of sea bass challenged against *V. anguillarum* (74). This antecedent suggests that the delayed upregulation of innate-related immune genes involved in the response against *V. anguillarum* could be one of the possible causes of the high mortality observed in our study. This response may also generate a delay in mobilizing the necessary physiological mechanisms to sequester the available iron from the bloodstream. Accordingly, in marine fish, it has been reported that *V. anguillarum* mediates the iron-uptake system by plasmid-mediated pJM1 (75) or from ferric citrate (76). This mechanism allows the bacteria to establish the infection and even can cause fish death as a consequence of septicemia (77). In the last years, increasing attention has been directed to exploit the bacterial weaknesses by modifying iron bioavailability to take advantage of the host iron-uptake system as a control method for bacterial infections (50). In our study, the delayed modulation of *transferrin* and *hepcidin* suggests that the iron-uptake regulation system is crucial for a successful infective process resulting in high mortalities in sea bass larvae.

The modulation of these host iron regulator genes is directly associated not only to their role as innate immune mediators responsible for removing iron from the bloodstream but also with their role as bacteriostatic and antimicrobial agents, acting as positive APPs (78). The APR is a prominent systemic reaction to a particular set of stimuli that may cause disturbance of the homeostasis, for instance, an infection process, resulting in the production of pro-inflammatory cytokines. In fact, Atlantic cod (*Gadus morhua*) injected with heat-killed *V. (Listonella) anguillarum* showed upregulation of *transferrin* accompanied by the expression of APR genes together with upregulation of *il-1β* and *il-8* (79). Accordingly, in our results the upregulation of humoral innate effectors (*lysozyme*, *transferrin*, and *hepcidin*) is directly related to the upregulation of both *il-1β* and *il-8* at 120 hpc. Regarding pro-inflammatory cytokines, the expression of *il-1β* after intraperitoneal (i.p.) infection with *V. anguillarum* has been previously observed at systemic level in gilthead sea bream (*Sparus aurata*) 4 h after bacterial challenge (80). An upregulation of *il-1β* has also been reported in juvenile sea bass intraperitoneally injected with *V. anguillarum* both at local (skin) and systemic level (spleen, head kidney) mainly during the first 8 h post-infection (81). The upregulation of *il-1β* was also accompanied by the expression of *il-8* but also *il-10* at local (gills) and systemic (head kidney, spleen) sites 24 h after i.p. injection with formalin-killed *V. anguillarum* (82). The joint upregulation of pro-inflammatory (*il-1β*, *il-8*) and anti-inflammatory (*il-10*) cytokines were also noted in our study at 120 hpc. The pro-inflammatory (involved in the upregulation of inflammatory reactions) and anti-inflammatory (control the pro-inflammatory cytokine response by immunoregulatory molecules) cytokines are relevant actors in the host immunity. In mammals, it has been widely described that IL-10 acts as a potent anti-inflammatory cytokine that regulates and inhibits the expression of pro-inflammatory cytokines, contributing to the normal resolution of infection and reducing tissue damage caused by inflammation (18, 19). In teleost fish, several reports indicated a role of IL-10 as anti-inflammatory cytokine and the induction of high levels of *il-10* in infected fish (55, 83, 84). A sequence analysis performed in cod has shown that *il-1β*, *il-8*, and *il-10* promoter regions contain similar regulatory domains which can explain the modulation of these genes, similarly of what has been described in mammals (82). This information opens the possibility that the same phenomena occur in sea bass and, therefore, it can be the responsible mechanism for the *il-1β*, *il-8*, and *il-10* modulation observed at 120 hpc. On the other hand, in mammals, it has been reported that hepcidin can also act as an anti-inflammatory agent inducing a signal cascade by hepcidin-activated Jak2, which phosphorylates the transcription factor Stat3, subsequently provoking an anti-inflammatory transcriptional response by negative feedback (85). Altogether, it seems clear that *V. anguillarum* modulates the expression of proand anti-inflammatory cytokines. However, it is striking that a significant regulation of innate immune-related genes takes place at 120 hpc considering the high mortality observed at the same time point. In this aspect, the non-significant differences observed during the first 36 hpc at gene expression level, can also be related to the fact that no significant differences were also observed in the cumulative survival at the same time-period. On the other hand, the non-modulation of genes associated with the immune system in the first 36 hpc could be related to the participation of maternally transferred immune factors devoted to protect early fish stages against invading pathogens before full maturation of immunological systems. Importantly, it has been previously reported that *V. anguillarum* evades the immune response of sea bass (80 g mean weight) (86). Hence, the sum of an immature immune system, the late immune response at 120 hpc together with *V. anguillarum* escape ability from the host antimicrobial defense, could explain the high mortality observed in our study. More studies are needed to elucidate whether the modulation of genes related to innate immune response begin from 48 hpc when differences in the cumulative mortality between challenged and non-challenged groups are observed.

An external stimulus present in the environment, such as a bacterial pathogen, may be sensed by the fish and eventually trigger an immune response, as has been observed in our study. In addition, it can also induce a subsequent global neuroendocrine response when the alarm messengers will reach and activate the HPI axis. This neuroendocrine signaling will induce a stress response in the organism. In teleosts, cortisol is the main glucocorticoid and the final product of the HPI axis activation in response to stressor stimuli exposure (87). Taking into account that cortisol alone provides a good but incomplete snap-shot of HPI axis activity when analyzing whole body of fish larvae, the additional analysis of its direct precursors and phase I metabolites as analyzed in our study provide a more detailed and accurate view of HPI axis effector activity. As described by Øverli et al. (88), reactions to stress vary between individuals as reflected in a proactive or a reactive coping style. As glucocorticoid quantification was done using a single larva, the glucocorticoid production elicited by a different coping style of the individual larvae could be quantified.

Previously, the role of cortisol and its effects in the immune response has been extensively analyzed (36). The previous hypothesis on the variations of immune and endocrine molecules assumes that organisms have a complex physiological machinery to respond efficiently to external stimuli. Hence, it should be taken into account that during larval stages, the full physiological and immune equipment to cope with external challenges may not be complete or they become progressively functional as timecourse progresses. Thus, most data available related to hormonal responsiveness of larvae studied so far show that fish larvae are able to secrete cortisol from very early stages, being either from maternal sources in the egg at first hours post-hatching or afterward by the own larval machinery. Thus, Stouthart et al. (89) showed increase of ACTH and cortisol in carp (*Cyprinus carpio*) 24 h post-hatching and Laiz-Carrión et al. (90) showed ACTH activity already at hatching stage. In sea bass, cortisol secretion is observed at first feeding (11 days post-hatching) and a high response at flexion (25 days post-hatching) (27). The marked increase of cortisol levels in sea bass larvae at 120 hpc (equivalent to 10 days post-hatching) could be the consequence of this physiological process at that time point of the developmental stage (27). However, the peak of cortisol observed at 120 hpc was not observed on the non-challenged group. On the other hand, the cortisol release could be the result of the starvation process, but both challenged and non-challenged groups were not fed along the experiment. Taken together, we hypothesize that the exposure to *V. anguillarum* is responsible for the high cortisol level obtained in our study at 120 hpc compared with the nonchallenged group.

Importantly, the whole-body cortisol increase observed in our study was correlated with the high concentration of cortisone observed at 120 hpc. A previous report showed high concentration levels of cortisone after subjecting fish to stress (34), indicating that the surge of cortisol metabolites is probably a mechanism to downregulate cortisol into a non-active metabolite at physiological level (24). Therefore, a similar mechanism could take place in the sea bass larvae in our study at 120 hpc. In addition, no variations were registered on the cortisol precursors 17α-HP and 11-deoxycortisol. Thus, the absence of variations of direct cortisol precursors and the elevated levels of cortisol and cortisone raise the possibility that the machinery of cortisol synthesis would have been activated between 36 and 120 hpc. Thus, the high level of cortisol and cortisone (but not the cortisol precursors) observed in our study would be the result of the activation of cortisol negative feedback mechanisms. However, the non-increase of glucocorticoid secretion observed in our study at the first 36 hpc suggests that the presence of *V. anguillarum* was not recognized as an acute stressor at the first developmental stages. A previous report showed that sea bass larvae open their mouth from day 3 post-hatching onward and importantly, *V. anguillarum* can be tracked on the intestinal track from 2 h post-exposure onward (91). Therefore, the level of cortisol during the first 36 hpc indicates that sea bass larvae do not mount an immediate acute response even when *V. anguillarum* is in close contact with the host inner environment (91). This result agrees with ongoing research on the late inflammatory response to bacteria on axenic sea bass larvae (Galindo-Villegas and Bossier, unpublished results) and on the hypothalamic response of juvenile sea bream to *V. anguillarum* bacterin in which the response of peptides such as CRH binding protein and TRH did not take place until 48 h of *V. anguillarum* exposure, suggesting a delay in the stress perception at brain regulatory level (Khansari et al., unpublished results).

The receptors for cortisol are the glucocorticoid receptor (GR) and MR (26), although GR is considered the primary receptor for glucocorticoid action in teleosts (24). In our study, the increase of cortisol observed at 120 hpc did not vary the mRNA abundance of *gr1* nor *gr2*. Accordingly, a previous antecedent on sea bass larvae showed that the increase of cortisol was not associated with the augment of the expression of gr1 and gr2 (27). In mammals, these receptors have been found in almost all cells, indicating that almost all cells are targets for glucocorticoids (92). In teleosts, these receptors have been found in many tissues including immune cells (93, 94) and can explain the regulatory role of cortisol upon multiple aspects of immune defense including the secretion of pro- and anti-inflammatory cytokines. On the other hand, in our study, no variation was observed for *hsp70* throughout the time points tested. HSP70 plays a key role on the cortisol-GR ligand binding because its function as chaperone (95). A correlation between *gr* and *hsp70* mRNA abundance has been proposed, most probably associated with enhancement of cortisol sensitivity in immune cells (96). However, several studies have reported that the augment of cortisol does not affect the *gr* nor *hsp70* expression level (96, 97). Altogether, the lack of variation in the mRNA abundance of both *glucocorticoid receptors* and *hsp70* observed in our study suggests that the basal expression of these genes is physiologically adequate to mount an endocrine response in sea bass larvae challenged with *V. anguillarum*. The participation of other mechanisms promoted by cortisol remains to be elucidated.

Although there is no direct activation of an acute stress response during the first 36 hpc, an endocrine response could be indirectly activated through the immune system due to the bidirectional communication between neuroendocrine and immune systems. The neuro-immuno-endocrine network orchestrates the response to allostatic load in fish (24). Thus, it was reported in trout (*Oncorhynchus mykiss*) that cortisol treatment downregulates the expression of pro-inflammatory related genes in monocytes/macrophages primary cell culture stimulated with lipopolysaccharide (41). In sea bream, cortisol showed the ability to decrease the expression of pro-inflammatory cytokines both in leukocytes from the head kidney (42) and also in head kidney primary cell culture (HKPCC) (40). The same immunosuppresor cortisol effect was observed in sea bream HKPCC stimulated with inactivated *V. anguillarum* (35). This evidence, in combination with the results reported in our study, suggests that the high level of cortisol observed at 120 hpc in sea bass larvae challenged with *V. anguillarum* is a response mechanism attributed to the sensing of a biological threat related to the augmented pro-inflammatory response registered at 120 hpc. Glucocorticoids are considered to be mediators of the anti-inflammatory response by generating immunosuppression associated with nuclear factor-κB (NF-κB) by induction of the inhibitor of kappa B protein, trapping activated NF-κB in inactive cytoplasmic complexes (98). It is noticeable that pro-inflammatory genes such as *il-1β* and *il-8* (both modulated by *V. anguillarum*) have a NF-κB binding site (82). Therefore,

#### REFERENCES


the increase in markers associated with glucocorticoid synthesis could be related with the induction of an anti-inflammatory response in sea bass larvae challenged with *V. anguillarum*. The increase of glucocorticoids at 120 hpc is in accordance with the expression of innate immune response genes that are associated with a pro-inflammatory antibacterial response. This antecedent, together with the upregulation of genes associated with the antiinflammatory response, suggests that in sea bass larvae infected with *V. anguillarum* a pro-inflammatory response at systemic level takes place, which then leads to the participation of other physiological mechanisms at systemic level to counteract the effect and the consequences of such response. This late systemic response could be the responsible for the high mortality observed in sea bass larvae challenged with *V. anguillarum*.

#### ETHICS STATEMENT

All experiments were approved by the Ethical Committee of the Faculty of Veterinary Medicine and the Faculty of Bioscience Engineering, Ghent University (no. EC2015\_02) and carried out in accordance with the recommendations of the European Union Ethical Guidelines for experimental animal care and other scientific purposes (2010/63/EU).

## AUTHOR CONTRIBUTIONS

The conceptualization of the experiment was performed by FERL and PB. The methodology was originally proposed by FERL, JA, and PB. The experiments were carried out by FERL. The data analysis was performed by FERL, JA, EVV, and BA. FERL, KD, LT, and PB were the responsible for the acquisition of the financial support for the projects leading to this publication. FERL and LT wrote the original draft. All the authors corrected, read, and approved the final manuscript. The study was supervised by FERL, LT, and PB.

## ACKNOWLEDGMENTS

The authors thank to J. C. Balasch for his graphical support on the experimental setup figure.

## FUNDING

This study was financially supported by the Aquaxcel EU project 0133/09/14/31 (FERL) and MINECO-Spain AGL2016-76069- C2-2-R project (LT). EVV is recipient of a grant from CONICYT-Chile Postdoctoral fellowship. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.


mRNA levels in larval red drum (*Sciaenops ocellatus*). *Gen Comp Endocrinol* (2010) 165:269–76. doi:10.1016/j.ygcen.2009.07.003


**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 Reyes-López, Aerts, Vallejos-Vidal, Ampe, Dierckens, Tort and Bossier. 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.*

# T cell receptor alpha chain genes in the Teleost Ballan Wrasse *(Labrus bergylta)* are subjected to somatic hypermutation

*Sumaira Bilal <sup>1</sup> , Kai Kristoffer Lie2 , Øystein Sæle2 and Ivar Hordvik1 \**

*1Department of Biological Sciences, University of Bergen, Bergen, Norway, 2 Institute of Marine Research (IMR), Bergen, Norway*

#### *Edited by:*

*Monica Imarai, Universidad de Santiago de Chile, Chile*

#### *Reviewed by:*

*Chris Secombes, University of Aberdeen, United Kingdom Kevin R. Maisey, Universidad de Santiago de Chile, Chile Jesus Hernandez, Center for Research in Food and Development (CIAD), Mexico*

> *\*Correspondence: Ivar Hordvik ivar.hordvik@uib.no*

#### *Specialty section:*

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

*Received: 24 October 2017 Accepted: 02 May 2018 Published: 22 May 2018*

#### *Citation:*

*Bilal S, Lie KK, Sæle Ø and Hordvik I (2018) T Cell Receptor Alpha Chain Genes in the Teleost Ballan Wrasse (Labrus bergylta) Are Subjected to Somatic Hypermutation. Front. Immunol. 9:1101. doi: 10.3389/fimmu.2018.01101*

Previously, somatic hypermutation (SHM) was considered to be exclusively associated with affinity maturation of antibodies, although it also occurred in T cells under certain conditions. More recently, it has been shown that SHM generates diversity in the variable domain of T cell receptor (TCR) in camel and shark. Here, we report somatic mutations in TCR alpha chain genes of the teleost fish, Ballan wrasse (*Labrus bergylta*), and show that this mechanism adds extra diversity to the polymorphic constant (C) region as well. The organization of the TCR alpha/delta locus in Ballan wrasse was obtained from a scaffold covering a single copy C alpha gene, 65 putative J alpha segments, a single copy C delta gene, 1 J delta segment, and 2 D delta segments. Analysis of 37 fish revealed 6 allotypes of the C alpha gene, each with 1–3 replacement substitutions. Somatic mutations were analyzed by molecular cloning of TCR alpha chain cDNA. Initially, 79 unique clones comprising four families of variable (V) alpha genes were characterized. Subsequently, a more restricted PCR was performed to focus on a specific V gene. Comparison of 48 clones indicated that the frequency of somatic mutations in the VJ region was 4.5/1,000 base pairs (bps), and most prevalent in complementary determining region 2 (CDR2). In total, 45 different J segments were identified among the 127 cDNA clones, counting for most of the CDR3 diversity. The number of mutations in the C alpha chain gene was 1.76 mutations/1,000 bps and A nucleotides were most frequently targeted, in contrast to the VJ region, where G nucleotides appeared to be mutational hotspots. The replacement/ synonymous ratios in the VJ and C regions were 2.5 and 1.85, respectively. Only 7% of the mutations were found to be linked to the activation-induced cytidine deaminase hotspot motif (RGYW/WRCY).

Keywords: T cell receptor, ballan wrasse, TCR**α**, polymorphism, somatic hypermutation, activation-induced cytidine deaminase motif, teleost

## INTRODUCTION

T cells in jawed vertebrates are generally divided into two subtypes αβ and γδ on the basis of the heterodimeric T cell receptor (TCR). The αβ T cells are most abundant in circulation and lymphoid organs, while γδ T cells are found in mucosal and epithelial tissues. In humans and mice, γδ cells represent less than 5% of the total T cell population while in birds and ruminants they constitute more than 40% of the total peripheral lymphocytes (1, 2). TCR αβ recognize peptides that are bound to major histocompatibility complex (MHC) molecules. TCR γδ recognize antigens directly, independent of MHC molecules in a manner similar to immunoglobulins (Ig). They are considered as a bridge between the innate and adaptive immune system as they use their receptor as a pattern recognition receptor (3). However, TCR γδ can also recognize phospholipids presented by CD1d molecules, suggesting that presentation by other non-classical MHC or MHC-like molecules might be possible (4). A unique TCRμ subtype first discovered in marsupials has subsequently been found in duckbill platypus (*Ornithorhynchus anatinus*), indicating that this locus was present in the last common ancestor of all extant mammals (5, 6).

T cell receptor molecules have structural and organizational resemblance to the Ig heavy and light chains. TCRα is encoded by variable (Vα) and joining (Jα) gene segments combined with the constant region (Cα) gene, like the Ig light chain. TCRβ is encoded by Vβ, diversity (Dβ), and Jβ gene segments combined with the Cβ gene, like the Ig heavy chain. Rearrangements of V, D, and J gene segments and variability generated in these junctions create an enormous repertoire of receptors with different specificities, providing versatility and diversity to the immune system (7–9). The mechanism of gene rearrangement is similar in B and T cells. V(D)J recombination is mediated by enzymes encoded by recombination activating genes 1 and 2 which recognize highly conserved recombination signal sequences (RSS). RSS are heptamer and nonamer motifs that flank the V, D, and J gene segments.

TCRα and TCRδ cDNA sequences have been reported from several teleost species, including common carp (*Cyprinus carpio*), channel catfish (*Ictalurus punctatus*), Atlantic cod (*Gadus morhua)*, rainbow trout (*Oncorhynchus mykiss*), zebrafish (*Danio rerio*), pufferfish (*Tetraodontidae rubripes*, *Tetraodontidae nigroviridis*, and *Sphoeroides nephelus*), Japanese flounder (*Paralichthys olivaceus*), Atlantic salmon (*Salmo salar*), and bicolor damselfish (*Stegastes partitus*) (10–20). The genomic organization of the TCRα and TCRδ genes in teleosts has been characterized in pufferfish, Atlantic salmon, and zebrafish. Like in mammals the TCRδ genes are linked to the TCRα genes, but the V gene segments are present downstream to the other elements in an inverted direction: Dδ-Jδ-Cδ-Jα-Cα-Vα/δ (15, 20–22). The Cα and Cβ sequences were considered to be relatively conserved due to interactions with other components of the TCR complex. However, allelic polymorphism of Cα and Cβ is widespread among teleost fish (17, 23).

Somatic hypermutation (SHM) is a key mechanism generating antibody diversity. In mammals, introduction of mutations in the recombined V(D)J gene in mature B cells is followed by selection of clones with higher affinities, typically in IgG-producing B-cells in germinal centers (24). Activation-induced cytidine deaminase (AID) deaminates cytosine (C) to uracil (U) in single stranded DNA creating U:G mismatch lesions, resulting in point mutations during SHM and double stranded breaks during class switch recombination. SHM occurs during transcription and primarily at RGYW/WRCY hotspot motifs (where G/C is a mutable position and R = A/G, Y = C/T, and W = A/T). SHM creates point mutations at a rate of 10<sup>−</sup><sup>3</sup> mutations/bp/generation, a million fold higher than the background genome mutation rate (25). SHM at A/T base pairs (bps) (typically WA/TW motifs) is generated by a mismatch repair mechanism employing polymerase η or other low fidelity polymerases (26). SHM has been detected in Ig genes of both cartilaginous and teleost fish (27–29). The impact of SHM on affinity maturation in fish needs further studies to be fully understood (29). Somatic mutations within Ig light chain genes of zebrafish were found to be overrepresented at AID hotspot motifs like in mammals. Mutations were most prevalent in the V region, but a significant number of substitutions were introduced in the C region; the mutation frequency decreased slightly with distance from the V region (30).

It was believed that TCR diversity was generated by V(D)J rearrangement and that SHM did not occur in the TCR genes, although SHM was detected in TCRα of mice and in TCRβ of HIV-infected patients (31, 32), and in TCR Cβ of two children after *in utero* stem cell transplantation (33). In the latter study, the frequency of Cβ mutations was significantly higher than in other groups of patients and healthy individuals (33). It was suggested that the lymphocytes of the babies presumably were under chronic activation. The SHM mechanism has also been found to target other genes, including oncogenes (34–36). A single cell PCR approach on lymph node germinal centers from a healthy person did not reveal SHM in T cells, in contrast to the situation in B cells where IgG clones were mutated (24). On the other hand, AID expression was unexpectedly detected in a subset of T cells in mice (37). More recently, studies of sandbar shark (TCRγ, TCRα) and camel (TCRγ, TCRδ) have shown that SHM occurs in TCRs of phylogenetically distant species (38–40), indicating a role in the diversification of the pre-immune repertoire. Accordingly, in mice it has been shown that early immature B cells are subjected to SHM, suggesting a role in B cell diversification as well as in the affinity maturation of antibodies (41, 42).

The aim of this study was to characterize the TCRα genes in Ballan wrasse and analyze Vα and potential Cα diversity within this species. Ballan wrasse has attracted increasing interest as a "cleaner fish" recently for the biological control of salmon lice in fish farms. Ballan wrasse belongs to *Perciformes*. Approximately 40% of all fish belong to this group (more than 10,000 species).

#### MATERIALS AND METHODS

#### Samples

Three adult wild fish of Ballan wrasse *Labrus bergylta* (700–800 g) were caught from fjords near Bergen, Norway. Ethical approval was not required, as the study did not involve transport or experiments on live fish. Fish were killed with a sharp blow to the head immediately after they were caught and tissue samples were stored in RNA-later solution (Ambion). Previously deposited transcriptome data (intestine) from 34 juvenile fish were provided from individuals sampled in a commercial fish farm in Øygarden, Hordaland, Norway (NCBI accession numbers: PRJNA382082 and PRJNA360275).

#### RNA Isolation and cDNA Synthesis

For transcriptome sequencing, intestinal tissues were homogenized using zirconium beads (4 mm) in a Precellys 24 homogenizer (Bertin Technologies, Montigny-le-Bretonneux, France) prior to RNA extraction. Total RNA was extracted using a BioRobot® EZ1 and RNA Tissue Mini Kit (Qiagen, Hilden, Germany). All samples were DNase treated according to the manufacturer. RNA quality and integrity was assessed using NanoDrop ND-1000 Spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA) and an Agilent 2100 Bioanalyzer with RNA 6000 Nano LabChip kit (Agilent Technologies, Palo Alto, CA, USA) respectively. The 260/280 and 260/230 nm ratios for the total RNA samples were >2.0 and the RNA integrity number >7.0 for all samples. For cDNA cloning and sequencing, total RNA was isolated from spleen and thymus using TRIzol® reagent (Invitrogen). First strand cDNA was synthesized using SuperScript™ II reverse transcriptase (Invitrogen) and an oligo dT16 primer.

#### Mapping of Intestinal Sequence Data

Raw Illumina HighSeq 2000 sequence reads deposited in the NCBI sequence read archive (SRA) database were analyzed in this study (SRA accession number: PRJNA382082). The raw FASTQ reads from individual intestinal samples originated from juvenile Ballan wrasse. Sequence adaptors were removed using Cutadapt (43, 44) with default parameters. The reads were further trimmed for low quality sequences using Sicle1 retaining reads with 40 bps minimum remaining sequence length and Sanger quality of 20. Prior to mapping, the quality of reads was investigated using FASTQC version 0.9.22 TopHat (version 2.1.1) short read aligner and Bowtie2 (version 2.2.9) was used to individually map each sample against the *L. bergylta* genome assembly (European Nucleotide Archive accession number: PRJEB13687) (44). Subsequent BAM files were further analyzed using the IGV genome browser (version 2.3.68).

#### PCR-Amplification of cDNA Fragments and DNA Sequencing

Primer construction for TCRα amplification was based on intestinal transcriptome data, genomic sequences and additional Vα sequence information obtained in the course of the present study (**Table 1**). Amplification using standard *Taq* polymerase (Invitrogen) was performed as follows: denaturation at

1https://github.com/najoshi/sickle.

2http://www.bioinformatics.babraham.ac.uk/projects/fastqc/.


94°C for 2 min, followed by 35 cycles of denaturation at 94°C (30 s), annealing at 55°C (30 s), and extension at 72°C (1 min/1,000 bps), and final extension for 10 min. Amplification using AccuprimeTM *Taq* DNA polymerase and AccuprimeTM High Fidelity *Taq* DNA polymerase (Invitrogen) was performed as follows: denaturation at 94°C for 2 min, 30 cycles of denaturation at 94°C (30 s), annealing at 55°C (30 s), and extension at 68°C (1 min/1,000 bps). DNA fragments were excised from the gel and further amplified for 5 cycles before cloning into pCR™ 4-TOPO® vector (Invitrogen). Sequencing was performed at an in-house sequencing facility using Big Dye termination chemistry (Applied Biosystems).

#### Sequence Analysis and Phylogeny

DNA/protein sequences were compared to the GenBank/EMBL databases using BLAST.3 DNA was translated into amino acid sequence using the translate tool available at ExPasy.4 Multiple alignments were performed using ClustalW.5 Phylogenetic trees were constructed using MEGA6 software and neighbor joining (NJ) and maximum likelihood (ML) matrixes with 1,000 bootstrap replicates (45).

#### Calculation of Mutability Index (MI) and Statistical Analysis

Mutability index is a measure of observed/expected number of mutations for a specific nucleotide without target bias. A mutability score of 1 represent unbiased mutation, while higher scores indicate that a specific nucleotide is selected for mutation. Relative frequency of each nucleotide was multiplied by total number of observed mutations within all sequenced clones to calculate the expected number of mutations. Observed numbers of mutations were divided by expected numbers for each nucleotide to calculate MI. Chi-squared analyses of MIs were carried out by comparing observed mutational frequencies to their expected (unbiased) mutational frequencies. *P* values <0.01 were considered statistically significant.

## RESULTS

#### Genomic Organization of the TCR **α**/**δ** Locus in Ballan Wrasse

Ballan wrasse TCRα sequences were identified by BLAST searches in an intestine transcriptome database using salmon TCRα as query (18). In the course of the present work, genomic sequence data of a heterozygous individual became available, and two scaffolds containing two allelic variants of the Cα gene in Ballan wrasse were identified by BLAST searches of whole genome shotgun data (GenBank): LaB\_20160104\_scaffold\_928 (99,234 nt) and LaB\_20160104\_scaffold\_4467 (16,780 nt). The Cα gene consisted of three exons corresponding to the Ig domain, the connecting peptide (CP) and the transmembrane (TM)/cytoplasmic (CYT) part. In total, 65 putative Jα segments

<sup>3</sup>http://blast.ncbi.nlm.nih.gov/.

<sup>4</sup>http://www.expasy.org/.

were found by manual inspection of the region upstream of Cα in scaffold 928. All putative Jα segments contained the highly conserved core motif FGXG or slightly modified versions of this, and splice sites and RSS flanking the Jα exons. The presence of J segments was further confirmed by alignment of transcriptome data with scaffold 928, using the IGV program (46). The TCRα cDNA clones characterized in this study contained 45 of the 65 identified Jα segments. A single Jδ segment and two putative Dδ segments were identified upstream of the Cδ gene. As in other teleosts, the SMG-7 gene was identified further upstream of Cδ (**Figure 1**; Table S1 in Supplementary Material). Scaffold 4467 was shorter and represented the other allele of the TCRα locus, comprising the Cα gene and 13 Jα segments (Table S2 in Supplementary Material). Several scaffolds containing Vα genes were found in the genomic sequence database, but a complete assembly of all Vα genes was not possible based on the present whole genome shotgun data.

## Sequence Analysis of TCR**α** cDNA Clones

The assembled Ballan wrasse TCRα sequence was used as a basis for primer construction, and cDNAs encoding part of the leader sequence, V/J, and Cα were amplified by PCR. In total, 79 distinct TCRα clones were analyzed from three individuals. The V gene regions of the cDNAs were sorted into four groups based on 75% nucleotide identity (Figure S1 in Supplementary Material). The translated sequences in group Vα1 showed all conserved characteristics of a V domain, while Vα2, Vα3, and Vα4 lacked the conserved cysteine (Cys) at position 26 (replaced with tyrosine); a pattern which was also seen in bicolor damselfish and olive flounder (16, 17). **Figure 2** shows representatives of

Figure 1 | Genomic organization of the T cell receptor α/δ locus in Ballan wrasse (LaB\_20160104\_scaffold\_928). Arrows show the direction of transcription. Cα is encoded by three exons shown in green, which are followed by 65 Jα segments. Light blue arrows present potential Jα segments, while dark blue represent the sequences found in cDNA clones. Cδ is encoded by three exons indicated here as Cδ1, Cδ2, and Cδ3. One Jδ segment is shown in yellow and two Dδ segments are represented in gray (Dδ1 and Dδ2). The SMG-7 gene was identified upstream of Cδ and its location is shown by the pink arrow.

identities between Ballan wrasse Vα1 and the other sequences are indicated at the end. GenBank accession numbers are: Wrasse Vα1 (MG594685), Wrasse Vα2 (MG594679), Wrasse Vα3 (MG594717), Wrasse Vα4 (MG594664), Atlantic salmon (ABO72169.1), bicolor damselfish (AAO88984.1), olive flounder (BAB82535.1), common carp (BAD88980.1), channel catfish (AAD56889.1), nurse shark (ADW95871.1), and human (AAD15154.1).

the four Vα families from Ballan wrasse aligned with TCR Vα sequences of other species. The Vα1 amino acid sequences have identity indices of 40–48% with the other characterized groups in wrasse; Vα2, Vα3, and Vα4. Among other species, Vα1 has 53% sequence identity to Atlantic salmon followed by common carp (45.7%) and olive flounder (44.7%).

Wrasse TCR Cα encodes a polypeptide of 112 amino acids. The Cα region can be divided into an Ig domain, CP, TM, and CYT part. The structurally important Cys residues in the Ig domain and CP are conserved. The TM region is the most conserved region containing the positively charged residues lysine and arginine involved in the assembly of the TCR–CD3 complex. Multiple sequence alignment demonstrated that Ballan wrasse TCR Cα has 47.3% sequence identity to pufferfish, 46.6% with salmon, 40.2% with zebrafish, and 38.6% with cod, 31% with mouse, and 26.6% with human (**Figure 3A**). The phylogenetic relationship between Ballan wrasse Cα and the orthologous molecules in other species is shown in **Figure 3B**.

## C**α** Polymorphism

The two scaffolds 928 and 4467 represent distinct Cα alleles; here named A and B. Analysis of transcriptome data from 34 farmed individuals identified A and B, and four additional allotypes of Cα named C, D, E, and F. Molecular cloning of TCRα cDNA from three wild fish corresponded to alleles A, B, and F [B1: (A/B), B4: (B/B), and B6: (B/F); **Figure 4A**; Table S3 in Supplementary Material].

## C**α** Somatic Mutations

Molecular cloning of TCRα from thymus cDNA of individual B1 showed that there was a significant number of point mutations in Cα (1.75/1,000 bps). Amplification of the first sample was done with standard *Taq* polymerase and 5,690 bps were sequenced (18 cDNA clones). In a second experiment, TCRα from spleen cDNA was amplified from individual B4 with high fidelity *Taq* polymerase. As a control, CD3ε cDNA was amplified under the exact same conditions. The frequency of mutations in Cα was 2.30/1,000 bps versus 0.62/1,000 bps in CD3ε. In total, 6,952 bps of Cα (22 cDNA clones) and 12,992 bps of CD3ε were analyzed. In a third experiment, TCRα was amplified from spleen cDNA with high fidelity *Taq* polymerase from individual B6, revealing 1.46 mutations/1,000 bps in a total of 12,327 bps (39 cDNA clones) (**Figure 4B**; **Table 2**). All cDNA clones examined were confirmed to be unique, possessing distinct V regions. Two clones were regarded to be artifacts of PCR jumping (i.e., partial extension on cDNA from one gene and final extension on another, resulting in hybrids of two alleles). The replacement to synonymous mutation ratio (R/S) was 1.85 and most replacements were conservative (i.e., the biochemical properties were not changed; hydrophobic, charged, and neutral, etc.). In total, 93% of the nucleotide substitutions were transitions (primarily A to G and T to C). Of all mutations, 37.5% were targeted at WA motifs, whereas 7.5% were targeted at AID motifs. Three WA motifs were present in WRCY (AID hotspot), but the targeted nucleotide was not at the C position (Table S4 in Supplementary Material).

carp (BAD89003.1), zebrafish (AAL29402.1), mouse (X14387.1), and human (L02424.1).

## Mutability Indices for Mono, Di, and Trinucleotides Indicate Targets for C**α** Mutation

To determine which nucleotides or combinations of adjacent nucleotides were preferentially targeted for mutation in the Cα gene, MIs for mono, di, and trinucleotides were calculated. Chisquare analysis showed that only A nucleotides were significantly targeted for mutation, 93% of which were transitions. C nucleotides were significant cold spots for mutation. MI scores showed that nucleotides were targeted in the order: A > T > G > C for somatic mutations (**Table 3**). The dinucleotide MIs revealed that AA, AT, GA, and AG were significant targets for mutation and when the analysis was expanded to trinucleotides, the dinucleotides were most frequently targeted in AAT, AAA, and AGA combinations (Tables S5 and S6 in Supplementary Material). This pattern of substitution indicates that mutations targeted particular nucleotides and combinations.

## V**α** Somatic Mutations

To study the mutation pattern in one single Vα gene, the B4-34 sequence was selected, and primers (TCR-VF/TCR-CR) were designed to amplify the VJ region (and a relatively short part of Cα). High fidelity *Taq* polymerase was used for amplification of spleen cDNA from individual B4, and the resulting PCRfragment was cloned. In total, 48 clones were analyzed. The 48 cDNA clones contained 25 different Jα gene segments (Table S1 in Supplementary Material). Out of the 48 clones, 45 were in frame. One clone had a stop codon (caused by point mutation) in the V sequence while two clones had frameshifts at the VJ junction. Alignment of the 48 cDNA clones indicated the presence of two


highlights residues which characterize each allotype. (B) Alignment and positioning of somatic replacement mutations found in TCR Cα of individuals B1, B4, and B6. Mutations are underlined in red. Accession numbers are given in Table S10 in Supplementary Material.



subgroups, highly similar to B4-34 (differing at two nucleotide positions). The subgroups were treated separately before the total number of substitutions at each position was calculated. Substitutions at positions 82 and 95 were considered allotypic differences as they had two alternative nucleotides represented by approximately 50% each (Figure S2 in Supplementary Material).

Of the total 102 substitutions, transitions (53%) were more common than transversions (47%) with a T/V ratio of 1.12. Transversions were relatively abundant in FR2 and complementary determining region 2 (CDR2) with a T/V ratio of 0.3. When there is no bias toward transition or transversion, the theoretical ratio of transition to transversion is 0.5 for random substitutions. The overall ratio of replacement to synonymous substitutions was 2.5. Replacement substitutions were more frequent than silent mutations in both CDRs and FRs, except for FR4 where silent mutations were dominant (Table S7 in Supplementary Material). SHM studies in mammals and other teleost Ig variable genes have shown relatively high T/V and R/S mutation ratios (30).

## Mono, Di, and Trinucleotide Targets in the V**α**1 Gene

The MIs for mononucleotides in the B4-34 group showed that nucleotides were preferentially targeted in the order: G > C > A~T. Of total mutations, G nucleotides were mutated 56.8%, followed by Table 3 | Substitutions and mutability index (MI) of T cell receptor (TCR) Cα mononucleotides.


*In total, 26,862 TCR C*α *nucleotides were analyzed (A* = *7109, T* = *5528, G* = *6846, C* = *7379). There were 40 point mutations. MI values were calculated by dividing observed number of mutations to expected number of mutations. The observed and expected numbers of mutations were compared by* χ*<sup>2</sup> analysis and significant differences are indicated on MI values.*

*a Statistically significant by* χ*<sup>2</sup> test (p* < *0.001).*

*bStatistically significant by* χ*<sup>2</sup> test (p* < *0.01).*

C (21.5%), and then A and T (10.8%). At G nucleotides 54% substitutions were transitions from G→A and transversions were G→T (41.4%) and G→C (10.4%) (**Table 4**). Analysis of dinucleotide MIs showed that CG and GC were the preferred targets for mutation, while AT were mutational cold spots (Table S8 in Supplementary Material). GC was found to be a significant target for mutation in human and catfish Ig heavy chain genes and zebrafish Ig light chain genes (28, 30, 47). When the analysis of the wrasse B4-34 group was expanded to trinucleotides, CG and GC dinucleotides were targeted most in GCG and CGA. Other combinations with significant MIs were AGG, GGT, and GAT (Table S9 in Supplementary Material).

#### DISCUSSION

The present study has shown that the TCRα genes in the teleost Ballan wrasse are subjected to SHM, and that this process also introduces some diversity in Cα. Similar to the situation in other teleosts (17, 23) the Cα gene in Ballan wrasse is polymorphic. Analysis of 37 fish identified 6 allotypes of Cα, each with 1–3 amino acid substitutions. Although TCR polymorphism is widespread among teleost fish it is tempting to suggest that some of the TCR Cα diversity observed in teleost cDNA pools might be a result of SHM.

The first attempt to amplify TCRα cDNA was based on a primer pair from a leader exon to the end of Cα, revealing four families of Vα genes. To analyze the SHM in a single Vα gene we selected the B4-34 as template and made the forward primer from the boundary of the leader and Vα exons. The resulting sequences were divided into two groups and treated separately to avoid overestimation of mutation frequencies (Figure S2 in Supplementary Material). The two groups might represent two highly similar V genes which have been amplified by the B4-34 primer pair.

The distribution of substitutions in TCR VJ and Cα is shown in **Figure 5**. The substitution rates were plotted against 20 bps nucleotide intervals. The first nucleotide corresponds to the third amino acid of FR1. Primers were designed from the start of the Vα gene and the first codons were, therefore, not included in the calculations. In the VJ region, CDR2 showed the highest frequency of substitutions. In Ballan wrasse TCRα the frequency of AID motifs was found to be much higher in the V/J genes (~8/100 bps) than Table 4 | Substitutions and mutability index (MI) of T cell receptor Vα mononucleotides.


*In total, 22,481 nucleotides were analyzed from B4-34 (A* = *6342, T* = *5746, G* = *4961, C* = *5432). There were 102 point mutations. MI values were calculated by dividing observed number of mutations to expected number of mutations. The observed and expected numbers of mutations were compared by* χ*<sup>2</sup> analysis and significant differences are indicated on MI values.*

*a Statistically significant by* χ*<sup>2</sup> test (p* < *0.001).*

in the Cα gene (4/100 bps), but no direct relationship was found between mutation frequencies and AID motifs in this study. In the VJ region, 6.8% of the mutations were present in AID motifs, while 9.2% were linked to WA/TW motifs, as compared to 40% in Cα. In a study of Ig variable genes it was found that SHM do occur in the absence of AID motifs and were predominantly G to C substitutions although the mutation frequency was lower than found in the presence of AID motifs (48). Mutability index analysis confirmed that nucleotides were targeted differently in the VJ and Cα region. In the VJ region, G nucleotides were targeted in CG and GC dinucleotides, and at GCG and GGG trinucleotide combinations. The mutations in Cα were primarily on A nucleotides, targeted mostly at AA (WA motif) and AT dinucleotides, indicating a mismatch repair mechanism employing polymerase η or other low fidelity polymerases (26). Replacement substitutions were dominant in both CDR1 and CDR2 and typically conservative. In a study of mouse TCR, extensive diversification by mutagenesis of CDR1 and CDR2 did not affect MHC binding (49), demonstrating that SHM of these regions is acceptable. In both Vα and Cα, replacement mutations were twice as frequent as silent mutations. Mutation frequencies in VJ versus Cα were found to be 4.5/1,000 bps and 1.76/1,000 bps, respectively. Key residues in Cα were conserved, showing that replacement substitutions had no impact on structural stability or interactions with CD3. Targeting of Cα is likely a side effect of SHM in the VJ region, like in the Ig light chain genes of zebrafish (30).

In the initial amplification and cloning of Ballan wrasse TCRα cDNAs about 24% of the clones had stop codons or were out of frame in the VJ region. The abundance of non-functional TCRα transcripts was similar in thymus and spleen. When narrowing the PCR-amplification to the B4-34 gene(s) the frequency of transcripts with stop codons or frame shifts decreased to 6.25%. Thus, it appears that there is a significant amount of non-functional transcripts in circulation, while some subpopulations of functional clones expand. In salmon, it was found that approximately 32% of TCRβ transcripts and 10% of TCRα transcripts in blood lymphocytes had stop codons or were out of frame (18, 50). In another study of salmon, about 10% of the TCRα transcripts from thymus were non-functional (20). Corresponding frequencies in rainbow trout thymus were 32% for TCRβ and 12.5% for TCRα (11, 51). In the amphibian Mexican axolotl more than 30% of the TCRβ transcripts from thymus and spleen, and 13.6% of TCRα

transcripts from thymus were sterile (52). The high fraction of nonfunctional TCRβ transcripts in ectothermic animals contrasts the situation in mammals where these mRNAs are eliminated (53, 54), although not absolutely (55). A more "leaky" system in coldblooded animals might be a result of less efficient cell proliferation and control mechanisms compared to higher vertebrates.

From the first amplification, TCRα clones which were in frame counted for 72% of the Cα mutations. As almost all replacement mutations were conservative there is no reason to believe that these are not incorporated into the functional T cell repertoire. Considering that we find somatic TCRα mutations in thymus as well as in spleen, SHM is probably involved in the diversification of the pre-immune TCR repertoire in Ballan wrasse, at the same time introducing some Cα diversity. A survey of translated TCRα ESTs in public databases revealed many amino acid substitutions in Atlantic salmon TCR Cα as well (Figure S3 in Supplementary Material). Thus, it is plausible to assume that SHM of TCR is a common phenomenon in teleost fish. However, SHM in mature T cells of wrasse cannot be ruled out either, and is an interesting topic for further research. SHM of TCR is generally believed to be restricted. Typically, an increase in TCR affinity after secondary challenge is minor compared to the affinity maturation of antibodies in mammals (56). It has been suggested that high-affinity TCR clones might be unfavorable due to prolonged binding, impairing serial interactions. On the other hand, somatic mutation of TCR genes in mature T cells has been documented (31, 32), suggesting that SHM of TCRs can occur under certain conditions, e.g., chronic activation (33). Vaccination of fish might facilitate conditions that trigger inappropriate immune responses (57, 58), but wild caught fish were used in the present analysis, implying that the results presented here are a normal situation in adult fish.

In conclusion, this study has shown that the TCRα in the teleost Ballan wrasse is subjected to SHM. The mutation frequency was highest in CDR2, although mutations were also evident in the constant part and FRs of TCRα. A high-throughput sequencing approach will be an interesting future study that can provide a more complete overview of the effects of SHM during the development of the TCR repertoire in the teleost fish.

### ETHICS STATEMENT

All work complied with relevant ethical guidelines and regulations. Ethical approval was not required as the study did not involve transport or experiments on live fish.

### AUTHOR CONTRIBUTIONS

SB and IH designed the study. SB did experimental work and data analysis. KL and ØS worked on high-throughput sequence data. IH took part in data analysis and supervised the whole study.

#### REFERENCES


## ACKNOWLEDGMENTS

We thank Dr. Lindsey J. Moore for comments on the manuscript.

#### FUNDING

The present study was funded by the Norwegian Research Council (Project number: 244396).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.01101/ full#supplementary-material.


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

The reviewer KM and handling Editor declared their shared affiliation.

*Copyright © 2018 Bilal, Lie, Sæle and Hordvik. 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.*

# Mechanisms of Fish Macrophage Antimicrobial immunity

*Leon Grayfer1 , Baris Kerimoglu1 , Amulya Yaparla1 , Jordan W. Hodgkinson2 , Jiasong Xie2 and Miodrag Belosevic2 \**

*1Department of Biological Sciences, George Washington University, Washington, DC, United States, 2Department of Biological Sciences, University of Alberta, Edmonton, AB, Canada*

Overcrowding conditions and temperatures shifts regularly manifest in large-scale infections of farmed fish, resulting in economic losses for the global aquaculture industries. Increased understanding of the functional mechanisms of fish antimicrobial host defenses is an important step forward in prevention of pathogen-induced morbidity and mortality in aquaculture setting. Like other vertebrates, macrophage-lineage cells are integral to fish immune responses and for this reason, much of the recent fish immunology research has focused on fish macrophage biology. These studies have revealed notable similarities as well as striking differences in the molecular strategies by which fish and higher vertebrates control their respective macrophage polarization and functionality. In this review, we address the current understanding of the biological mechanisms of teleost macrophage functional heterogeneity and immunity, focusing on the key cytokine regulators that control fish macrophage development and their antimicrobial armamentarium.

#### *Edited by:*

*Brian Dixon, University of Waterloo, Canada*

#### *Reviewed by:*

*Geert Wiegertjes, Wageningen University & Research, Netherlands Uwe Fischer, Friedrich Loeffler Institute Greifswald, Germany*

*\*Correspondence:*

*Miodrag Belosevic mike.belosevic@ualberta.ca*

#### *Specialty section:*

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

*Received: 09 February 2018 Accepted: 02 May 2018 Published: 28 May 2018*

#### *Citation:*

*Grayfer L, Kerimoglu B, Yaparla A, Hodgkinson JW, Xie J and Belosevic M (2018) Mechanisms of Fish Macrophage Antimicrobial Immunity. Front. Immunol. 9:1105. doi: 10.3389/fimmu.2018.01105*

Keywords: teleost, monocyte, macrophages, antimicrobial, cytokine, respiratory burst, nitric oxide, nutrient

#### INTRODUCTION

depravation

The immune systems of all vertebrates are integrally dependent on macrophage-lineage cells and while the ontogeny of functionally disparate macrophage subsets and lineages have been thoroughly studied in mammals (1, 2), they remain to be adequately defined in aquatic vertebrates such as teleost fish [previously reviewed by Hodgkinson et al. (3)]. In mammals, these functionally distinct macrophage subsets are framed by polarized extremes including the interferon-gamma (IFNγ) and tumor necrosis factor-alpha (TNFα) primed M1/classically activated macrophages; the interleukin-4 and/or interleukin-13-stimulated M2a/alternatively activated macrophages; the immune complexes or apoptotic cell-stimulated M2b/alternatively activated macrophages; and the interluekin-10 (IL-10), transforming growth factor-beta (TGF-β) and/or glucocorticoid (GC)-primed M2c/alternatively polarized macrophages (4). Depending on their respective stimulus-dependent polarization states, these macrophage subsets participate in either inflammatory/microbicidal or repair/woundhealing/immune suppression responses (4). While bony fish clearly possess functional analogs of these mammalian macrophage subsets (**Figure 1**), the molecular mechanisms governing the polarization and functionality of these respective fish macrophage populations remain to be fully defined.

The teleost fish inflammatory/M1 macrophage populations have been the best-studied and shown to rapidly kill pathogens through phagocytosis (5), production of reactive oxygen and nitrogen intermediates (6, 7), and restriction of nutrient availability (8, 9). Furthermore and akin to their mammalian counterparts, these fish M1 macrophages produce a plethora of inflammatory cytokines, chemokines, and lipid mediators (9). In a recent effort to gain insights into the alternatively polarized/

**94**

M2 fish macrophages, researchers have examined functional macrophage parameters such as *arginase* gene expression and activity (10) and their expression of immunosuppressive cytokines such as *il10* and *tgf*β (11–13). Indeed, the functional analogs of the mammalian M1/M2a–c macrophage subsets appear to be present in teleosts. However, defining the regulatory mechanisms governing the polarization of these effector populations is a far more challenging goal as gene-specific and whole genome duplication events have endowed disparate fish species with unique multicopy repertoires of those genes, which in mammals are though to dictate macrophage polarization and functionality (14).

In this review, we focus on the current understanding of the molecular mechanisms of fish macrophage antimicrobial responses to prokaryotic and eukaryotic pathogens.

## MACROPHAGE ONTOGENY

#### Macrophage Sources and Fates

Until recently, tissues macrophages were believed to arise from circulating monocyte precursors in response to tissue entry and accompanying stimuli (15). However, more recent research has challenged this notion and suggests that while mammalian blood monocytes may enter into tissues and become macrophages under certain inflammatory conditions, these events are infrequent (15–17). Instead, mammalian resident tissue macrophages are now thought to be seeded during embryonic hematopoiesis and replenish resident populations locally (16–18). While the presence of self-renewing fish tissue macrophage populations requires further investigation (19), recent reports showed that fish lacking functional *c-myb* transcriptional regulator of adult hematopoiesis, nonetheless possess tissue macrophages suggest that this process may be conserved in teleosts (20).

## Teleost Monopoiesis and the Colony-Stimulating Factor-1 Receptor (CSF-1R)

The differentiation and functionality of most vertebrate macrophages are controlled by engagement of the CSF-1R, which is expressed on committed myeloid precursor cells and their derivative populations (21–23). The *csf1r* (*fms*) genes of different vertebrate species exhibit poor sequence identities, particularly in their extracellular domains (24–26). By contrast, the catalytic tyrosine kinase domains of CSF-1Rs are highly conserved (27, 28). The divergence of the extracellular portions of the CSF-1R molecules likely reflects the selective pressure onto this receptor of diverging (and in some cases multiple) ligands of these receptors, as these exhibit low amino acid sequence conservation. The mammalian, reptilian, avian, and teleost fish CSF-1Rs all branch into phylogenetically separate clades (26), presumably reflecting the many distinct aspects of macrophage functionality across these divergent species. In turn, these differences may reflect distinct functional contributions of these respective ligands and receptor systems to the macrophage ontogeny and functionality of the evolutionarily diverged vertebrate species.

#### Colony-Stimulating Factor-1

Unlike birds and mammals that have a single alternatively spliced *csf1* gene (29, 30), many teleost fish species have two distinct *csf1* genes (*csf1.1* and *csf1.2*), which (for the most part) do not appear to undergo alternative splicing (25). Like its mammalian counterpart, the fish CSF-1 (CSF-1.1) also appears to be an important macrophage growth and differentiation factor (31, 32). Interestingly, while the mammalian CSF-1 is known for driving alternative/M2 macrophage differentiation (28), the cyprinid (goldfish) CSF-1.1 appears to facilitate the functional differentiation of inflammatory/M1-like macrophages with highly upregulated pro-inflammatory components (32). This is supported by the reports that a soluble goldfish CSF-1R (19) down-regulates macrophage pro-inflammatory responses by reducing available soluble CSF-1 (33, 34) (see section below). As teleosts possess multiple *csf1* genes and at least some fish species also encode two distinct *csf1r* genes (35), this suggests that teleost fish may have adopted more elaborate macrophage differentiation strategies to those seen in mammals.

#### Interleukin-34 (IL-34) as Possible Sources of Macrophage Functional Heterogeneity

Inflammatory (M1) macrophages produce multiple inflammatory mediators that coordinate antimicrobial responses, while the alternatively activated (M2) macrophages secrete immunosuppressive and angiogenic compounds that control the resolution of inflammation [reviewed by Zhou et al. (4) and Hodgkinson et al. (3)]. The mammalian CSF-1 induces the differentiation/ polarization of M2 macrophages (28), whereas the teleost CSF-1 elicits an M1-like macrophage phenotype (32). Notably, the IL-34 cytokine also ligates and activates the CSF-1R (36–38), regulating the development of mammalian osteoclasts (39, 40), Langerhans cells (41, 42), microglia (41), and B cell-stimulating myeloid cells (43). Recent work using the amphibian *Xenopus laevis* model indicated that frog macrophages differentiated by the *X. laevis* CSF-1 are highly susceptible to the emerging Frog Virus 3 *Ranavirus* whereas macrophages derived by IL-34 are important antiviral effectors (26, 44, 45). The antiviral roles of IL-34-derived macrophages remain to be fully elucidated in other vertebrates, and it is likely that akin to CSF-1, IL-34 likewise contributes to macrophage functional heterogeneity.

To date, there have been a limited number of studies addressing the contribution of IL-34 to the fish macrophage biology. Recent work indicates that the grouper IL-34 plays an important role in the fish immune response against *Cryptocaryon irritans* infections, as the expression of this gene was highly upregulated in the parasite-infected fish gill and skin tissues (46). This is consistent with the roles of the mammalian IL-34 in the differentiation and functionality of tissue resident macrophages and Langerhans cells (41, 42) and may reflect an evolutionarily conserved role for IL-34 in controlling the development of this macrophage-lineage cell type.

The trout *il34*, *csf1.1,* and *csf1.2* are differentially expressed in fish tissues and as well as in a number of trout-derived cell lines, suggesting disparate biological roles for these CSF-1R ligands (47). Notably, the trout *il34* exhibited high baseline tissue expression in which the authors attributes to a possible homeostatic role and that indeed could reflect the conserved role of this growth factor in tissue macrophage and Langerhans cell biology. Moreover, whereas stimulation of primary trout kidney macrophage cultures with a number of pathogen-associated molecular patterns (PAMPs) failed to elicit increases in *csf1.1* or *csf1.2* gene expression, these stimuli readily upregulated the expression of *il34* by these cells (47). Notably, the viral dsRNA mimic poly I:C elicited a particularly robust increase the macrophage *il34* expression, possibly reflecting a conserved role for the fish IL-34 in antiviral immunity, akin to the amphibian counterpart.

## Soluble CSF-1R

Cyprinid fish control their CSF-1 (and presumably IL-34) stimulation of macrophages by production of a soluble CSF-1 receptor (sCSF-1R) (33, 34, 48, 49). This soluble form of the receptor arises through alternative splicing, and is capable of ablating macrophage proliferation (48) and macrophage-mediated inflammatory responses (33, 49). The sCSF-1R is produced by mature macrophages, but not monocytes, in response to classical M2-polarizing stimuli such as apoptotic cells (34) and efficiently ablates an array of inflammatory events including leukocyte infiltration (34), macrophage chemotaxis, phagocytosis, production of reactive oxygen intermediates and recruitment of leukocytes (33). Moreover, sCSF-1R dampens fish macrophage chemokine and inflammatory cytokine expression, neutrophil recruitment while promoting the expression of the anti-inflammatory cytokine, interleukin-10 (49). It will be interesting to learn whether other fish besides cyprinids have adopted this strategy for controlling their macrophage inflammatory responses.

## MOLECULAR CONTROL OF MACROPHAGE ANTIMICROBIAL ARMAMENTARIUM

#### Pattern Recognition Receptors (PRRs) of Teleost Macrophages

During injury and/or infection, resident macrophages detect tissue damage and/or infiltrating pathogens by either extracellular or intracellular pattern recognition receptors (PRRs). The existence of immune PRRs was first proposed by Charles Janeway over 20 years ago (50). The known PRRs can be classified into five groups based on their structure and function: toll-like receptors (TLRs), C-type lectins, nucleotide-binding domain-leucine-rich repeat containing receptors (NLRs or NOD-like), retinoic acid inducible gene 1 (RIG1)-like receptors (RLRs), and absence in melanoma (AIM)-like receptors (ALRs) [reviewed by Hansen et al. (51)]. Neutrophils, monocytes, macrophages, dendritic cells (DCs), and specific epithelial and endothelial cells have PRRs (51).

In addition to specific recognition of distinct pathogen components (e.g., LPS, dsRNA, and flagellin), PRRs also detect tissue damage-associated molecular patterns (52–54). The human TLRs 1, 2, 4, 5, 6, and 10 are membrane bound while the TLRs 3, 7, 8, and 9 are located in endosomes (55). By contrast, the NLRs, RLRs, and ALRs are exclusively cytosolic (55).

Members of the TLR family share the intracellular toll-interleukin-1 receptor motifs (56). Initially identified in *Drosophila* spp. for controlling dorso-ventral patterning (57) and subsequently attributed to its anti-fungal properties (58), members of this family are now widely believed to be indispensable for immune recognition by most metazoans. Humans are currently known to have 10 TLRs (TLR 1–10) and mice possess 12 TLRs (51, 59). Birds also possess 10 TLRs, of which some are counterparts of the mammalian receptors (TLR 3–5, 7, and two forms of each TLR 1 and 2) (59). Some birds (TLR15, 16, and 21) are not found in higher vertebrates (60), and amphibians may have up to 20 TLRs (61). Bony fish possess 17 distinct TLRs, including some that are unique to fish, such as TLR 20–23 (62–64). Interestingly, not all fish have TLR4 and zebrafish TLR4 does not recognize LPS and negatively regulates NF-kB signaling (51, 65, 66). Additional research will be required to fully elucidate the function of TLRs in lower vertebrates, which will undoubtedly shed new light on the evolutionary history of these important innate immune receptors. The biology of fish TLRs has recently been a subject to several excellent review articles (67–70).

Other PRR families have also been identified in aquatic vertebrates. Although, gene synteny analyses identified a number of RLRs in birds and fish (71, 72), certain RLR genes are either absent or diverged beyond recognition. The evidence of functional conservation of fish RLRs exists (73, 74).

The NLRs were originally discovered in plants as R-proteins, which share nucleotide binding site and leucine rich repeat domains and can detect proteins delivered by pathogenic bacteria to trigger rapid activation of host defense (75, 76). The first identified mammalian NLR was the human NOD1 (also known as CARD4) by Bertin et al. (77) and Inohara et al. (78). The NOD1/CARD4 contained the typical NOD domain (also referred as the NACHT domain), which is a critical structural feature of NLRs (79–81), and NOD2/Card15 was identified searching for NOD1 homologs in genomic databases (82), and at present there are 23–34 NLRs known to exist in humans and mice, respectively.

There are several orthologs of mammalian NLRs as well as a unique NLR subfamily of receptors in bony fish (83). The first reported teleost NLRs were identified in zebrafish genome (84). Three subfamilies of NLRs were present in zebrafish, the first resembled mammalian NODs, the second resembled mammalian NLRPs, and the third was reported to be a unique subfamily of genes having similarities to both mammalian NOD3 and NLRPs (83). The existence of NLRs has been reported in grass carp (85), rainbow trout (86), channel catfish (87, 88), common carp (89, 90), orange-spotted grouper (91), goldfish (92), Japanese founder (93, 94), miiuy croaker (95, 96), and Japanese pufferfish (97). The results of these studies indicated the presence of inducible NLRs and that teleost NLRs shared the conserved structural domains with their mammalian counterparts. Studies on most of teleost macrophage NLRs primarily focused on the examination of gene expression induced by different immune stimuli and/or fish pathogens (87, 88, 92) and to a lesser extent on NLR signaling pathways in fish macrophages (98–102).

#### The Type II Interferon System(s) of Bony Fish

The classical/M1 macrophage activation corresponds to macrophage upregulation of an array of inflammatory, microbicidal, and antigen presentation components, and is linked due to Th1 biased cytokine stimulation of these cells (103, 104). Specifically, this classical macrophage activation is thought to predominantly occur in response to the type II interferon cytokine, IFNγ, which is produced by Th1 helper cells and activated NK cells (105, 106). The induction of the mammalian M1 macrophages requires the co-stimulation of cells with IFNγ and TNFα (107). Conversely, these classically activated macrophages may be generated following macrophage activation through pathogen PRRs (108). While teleost fish have numerous PRRs (83, 109), the roles of fish PRRs (see previous section) in teleost M1 macrophage polarization remains to be fully addressed.

The mammalian IFNγ cytokine has been linked to an vast array of immunological processes, and was first identified from the supernatants of PHA-activated lymphocytes (110). In addition to its modest antiviral capacities, IFNγ appears to be particularly important to vertebrate host defenses against obligate and facultative intracellular pathogens (111–115). These include several important macrophage pathogens such as *Listeria monocytogenes* (116), *Leishmania major* (117), and *Mycobacterium* (118). This underlines the importance of this cytokine to macrophage immunity (111, 119–122).

The mammalian IFNγ binds the interferon gamma receptor 1 (IFNGR1), which results in the formation of a receptor complex composed of this ligand binding chain as well as the IFNGR2 signal propagation chain, ensuing in the downstream signaling cascade (123). The assembly of this signaling complex (IFNγ:IFNGR1:IFNGR2) activates Janus kinases (Jak)-1 and -2 (124), upon which phosphorylation activates signal transducer of activation-1 (Stat1) transcription factor (125). Under certain cellular conditions, stimulation with IFNγ may also activate Stat2 (126) albeit to a much lesser extent than Stat1. Moreover, IFNγ signaling typically results in the activation and nuclear translocation of several other transcriptional complexes including ISGF3 and Stat1-p48, composed of Stat1: Stat2: IRF-9 and Stat1: Stat1: IRF-9 (126–129). IFNγ signaling occurs in temporal phases, where the first sets of interferon gamma stimulated transcripts are seen after 30 min of the initial IFNγ receptor activation, and many of the products of these mRNAs then modulate subsequent IFNγ-related (IFNγrel) signaling events within the stimulated cell (130).

Teleost fish are widely known to possess *ifng* genes (131–135) and the functional roles attributed to the mammalian IFNγs appear to be conserved to these fish cytokine counterparts. For example, the trout IFNγ elicits the expression of a number of immune genes such as *γip10*, *mhcIIb*, and *stat1* (136), *c-type lectin*, *il1b*, *ifng*, *tap1*, *tapasin*, *irf1*, *ikb*, and *junb* in the monocyte/macrophage RTS11 cell line (137). Fish IFNγ enhances reactive oxygen species (ROS) production by primary kidney phagocytes of trout (136), goldfish (138), and carp (139). The goldfish IFNγ primes kidney-derived monocyte ROS responses in a concentration dependent manner (138) and akin to its mammalian counterpart (140, 141), the goldfish IFNγ synergizes with the goldfish TNFα (138) to prime the fish monocyte ROS response. The goldfish IFNγ also induces modest but significant increases in kidney macrophage nitric oxide responses, which are further enhanced by co-stimulation with TNFα2 (138). Interestingly, the carp IFNγ elicits significant NO responses in fish kidney phagocytes only in conjunction with a high dose of LPS (139). The large yellow croaker IFNγ enhances the primary kidney phagocyte respiratory burst and nitric oxide responses and upregulates the gene expression of inflammatory genes such as *tnfa*, *il1b*, *stat1*, and *irf1* in these cells (142). Likewise, the black seabream and the zebrafish IFNγs induce the expression of *jaks*, *stats*, and interferon-stimulated genes such as *irf1* and *mx* (143, 144).

Goldfish kidney-derived macrophages stimulated with IFNγ upregulate their expression of several inflammatory genes including *tnfa* isoforms 1&2, *il1b* isoforms 1&2, *il12* subunits p35 & p40, *ifng*, *il8* (CXCL-8), *ccl1*, and *viperin* (138). Carp kidney phagocytes treated with IFNγ and LPS increase their gene expression of *tnfa*, *il1b*, and *il12*; (subunits p35 & p40) (139). Carp IFNγ also induced the expression of a CXCL-10-like chemokine (*cxclb*) and inhibited LPS-induced expression of *cxcl8* (139). Together, it would appear that for the most part, the inflammatory roles of IFNγ such as its synergism with LPS and TNFα (see below) are conserved in teleosts.

#### Functional Dichotomy of Fish Type II IFNs

In fish, Igawa et al. (132) identified two genes encode *ifng* isoforms, located next to the fish *il22* and *il26* genes, that have the exon/ intron organization of *ifng* genes of other vertebrates and possess the IFNγ signature motif ([IV]-Q-X-[KQ]-A-X2-E-[LF]-X2-[IV]). These two *ifng* sequences were initially coined IFNγ1 and IFNγ2 but following a reevaluation of vertebrate *ifng* genes and because the fish IFNγ2 possesses the hallmark features of the mammalian IFNγs, it was renamed as simply IFNγ (145). Since IFNγ1 appears to be structurally related to the mammalian IFNγ, but is missing a nuclear localization signal (NLS) motif, it has been coined as IFNγrel. The presence of multiple *ifng* isoforms have now been confirmed in siluriformes and other cypriniformes, including the identification of *ifngrel* and *ifng* in catfish (133), common carp (146), zebrafish (147), and the goldfish (138).

The siluriform IFNγrel proteins have not been functionally characterized. However, the cyprinid IFNγrels have been examined in some detail across several species. For example, freshly laid zebrafish eggs possess *ifngrel* transcripts, indicating maternal supply of these mRNAs (147). Also, while the gene expression of the zebrafish *ifng* is not detected until much later in development, the mRNA levels of IFNγrel continue to increase during the embryonic zebrafish development (147). Moreover, injection of zebrafish embryos with mRNAs encoding IFNγ or IFNγrel results in increased expression of genes typically activated by the mammalian IFNγ (147). Notably, morpholino knock-down of either *ifng* or *ifngrel* resulted in compromised *Yersinia ruckeri-*infected zebrafish embryo survival while the combined knock-down of both cytokines further decreased embryo survival (147), suggesting that IFNγ or IFNγrel confers at least partially non-overlapping immune roles.

The goldfish IFNγ and IFNγrel appear to confer distinct effects on macrophages (138, 148). For example, while IFNγ stimulation of goldfish monocytes results in long-lasting ROS priming, IFNγrel elicits a short-lived priming effect on these cells, followed by complete monocyte unresponsiveness to ROS priming by other inflammatory cytokines (IFNγ or TNFα2). Moreover, the goldfish IFNγ only modestly enhances fish monocyte/macrophage phagocytosis and nitric oxide responses (138, 148). By stark contrast, IFNγrel induced significantly greater phagocytosis, iNOS (isoforms A and B) gene expression, and nitric oxide production in goldfish monocytes and macrophages. Interestingly, these goldfish type II IFNs also elicit the expression of distinct immune genes in goldfish monocytes. Both recombinant cytokines induce goldfish monocyte Stat1 phosphorylation, however, nuclear translocation of Stat1 was only seen in cells treated with IFNγ, but not with IFNγrel. This was confirmed by more recent report, indicating that the zebrafish IFNγ and IFNγrel utilize distinct signaling pathways (143). It is interesting that while the recombinant ginbuna crucian carp IFNγ forms a dimer in solution, the recombinant IFNγrel appears to be monomeric (149, 150), akin to the functional forms of type I rather than type II IFNs. Moreover, an additional isoform of the ginbuna carp IFNγrel has been identified and shown to possess a functional NLS, which contrasts the other fish IFNγrel proteins (150). With the growing evidence indicating functional dichotomies of the cyprinid type II IFNs, it will be interesting to learn the roles of these distinct macrophage-activating factors in their target cells' antimicrobial responses to different fish pathogens.

## Fish Type II IFN Receptors

While the bony fish type II IFN ligands have become a subject of active research, the functional roles of the type II IFN receptors remain to be clearly defined. The trout IFNGR1 and IFNGR2 chains were initially identified and shown to exhibit conserved gene synteny across vertebrates (151). All fish IFNGR1 sequences have Jak1 and Stat1 binding sites, that are also required for functional mammalian IFNγ (152–154), and the expression of the IFNGR2 chain appears to be essential to the trout IFNγ-induced signaling (151).

The fish IFNγ and IFNγrel cytokines structural, functional, and intracellular signaling differences were thought to reflect the presence of distinct IFNγ receptors, dedicated to these respective moieties. As predicted, gene synteny analyses of the vertebrate *ifngr1* genes (encoding the ligand binding chain), revealed two distinct zebrafish *ifngr1* genes, located on distinct chromosomes (155). The presence of corresponding *ifngr* isoforms was confirmed in goldfish, and by means of *in vitro* recombinant protein binding studies, we demonstrated that IFNγrel (IFNγ1) and IFNγ each bound to their own cognate IFNγ receptor chains, the IFNGR1-1 and IFNGR1-2, respectively. Morpholino knock-down of the zebrafish *ifngr1-1*, *ifngr1-2*, or *ifngr2* (signal propagation chain) abolished the fish IFNγ function (156). Notably, only the knockdown of *ifngr1-1*, but not *ifngr1-2* or *ifngr2*, abrogated IFNγrel stimulation, suggesting that zebrafish IFNγ signals through a heterodimer (IFNGR1-1 and IFNGR1-2) and a IFNGR2 homodimer whereas the IFNγrel binds to homo-dimeric IFNGR1-1 and a distinct unknown receptor 2 chain. The discrepancy between these finds and our studies, which indicated IFNγ-IFNGR1-2 but not IFNγ-IFNGR1-1 interactions, could be explained in several ways. Aside from the possible species-specific differences, it may be that IFNGR1-1 binds IFNγ with lower affinity, explaining our inability top detect this interaction *in vitro* by western blot. Conversely, the presence of *ifngrel* mRNA in fresh zebrafish embryos (147), suggests that this cytokine may plays roles during zebrafish (and presumably other cyprinid fish) development. If is the case, morpholino knock down of its cognate receptor encoding gene, *ifngr1-1* may manifest in reduced IFNγ function, as an indirect consequence of the abrogated IFNγrel-mediated immune development rather than through direct IFNγ–IFNGR1-1 interactions.

It is notable that using HeLa cells transfected with the ginbuna carp *ifngr1-1* and *ifngr1-2* encoding plasmids, it was shown that the carp IFNγ isoform 1 exclusively signals through the IFNGR1-2 whereas the IFNγ isoform 2 signals through the IFNGR1-1 (149). It is well established that both the mammalian and fish IFNγ signaling requires IFNGR2 chains (123, 156) while the fish Jak and Stat proteins have significantly diverged from (and are present in multiple forms as compared to) the mammalian counterparts (157).

While all other vertebrates examined to date encode individual type II IFNs and IFNGR1 genes, it is intriguing that certain fish possess two distinct IFN gamma-receptor binding chains (IFNGR1-1 and IFNGR1-2) as well as multiple type II IFNs (148, 149, 156). This suggests that these fish have adopted very unique strategies surrounding their principal M1 macrophageactivating cytokine system(s) and it will be exciting to learn what are the functional consequences of these differences.

#### Teleost TNF**α**

The mammalian TNFα is involved in a broad array of immunological roles (158–160). The name of this cytokine stems from its discovery in tumoricidal sera of Bacillus Calmette-Guerinprimed, endotoxin-treated mice (161). During vertebrate inflammatory response, TNFα promotes the chemotaxis of neutrophils and monocytes/macrophages to the sites of inflammation (162, 163), enhance macrophage phagocytosis (164–166), primes reactive oxygen and reactive nitrogen responses (167, 168), facilitates the chemotaxis of fibroblasts (169) and the release of platelet activating factors (170–172). Mammalian TNFα confers its immune effects either as a 17 kDa soluble protein or a 26 kDa type II trans-membrane protein (173–175) and most effects are induced after binding of homotrimerized TNFα to either the TNF-R1 or TNF-R2 (176, 177).

Tumor necrosis factor-alpha orthologs, possessing the TNF family signature [LV]-x-[LIVM]-x3-G-[LIVMF]-Y-[LIMVMFY]2 x2-[QEKHL] have been identified in several teleosts (178), underlining the evolutionary conservation of this cytokine. Like its mammalian counterpart, the teleost fish TNFα is a reliable marker of fish M1 macrophages (179, 180). Most fish species possess multiple TNFα isoforms (178, 181–188). These TNFα isoforms confer pro-inflammatory effects such as enhancing inflammatory gene expression, macrophage chemotaxis and phagocytosis, and eliciting phagocyte reactive oxygen and nitrogen intermediate production (183–185, 189–194). The *in vivo* roles of TNFα during fish inflammatory and M1 macrophage immune responses have also been confirmed in zebrafish (179), sole (195) and trout (196, 197).

#### Teleost TNF**α** Receptors

Bobe and Goetz (198) were first to report the presence of a death domain-containing TNF receptor in zebrafish and coined this gene the ovarian TNF receptor (*otr*), while putative zebrafish *tnfr1* and *tnfr2* gene sequences were deposited to GenBank, with the zebrafish *tnfr1* sharing high sequence identity with *otr*. We identified the goldfish *tnfr1* and *tnfr2* cDNAs (199) and showed that the putative amino acid sequences of these goldfish receptors share many conserved regions with their respective mammalian counterparts. Goldfish TNF-R1 has a death domain with a conserved motif (W/E)-X31-L-X2-W-X12-L-X3-L and six residues that are essential to TNF-R1-mediated cytotoxicity (200).

Our *in vitro* binding studies using recombinant version of the respective goldfish proteins indicate that both goldfish TNFα1 and TNFα2 bind either TNF-R1 or TNF-R2 (199). Notably, recombinant sea bream TNFα (191), and the goldfish recombinant TNFα1, TNFα2, TNF-R1, and TNF-R2 all adopt homo-dimeric conformations and associate as dimers as opposed to the trimeric confirmations seen in the mammalian TNF ligands and receptors (199). Similarly, the grass carp TNFα ligand and TNF-R1 also associate as dimers (201). Interestingly, dimerized forms of the mammalian TNF-R1 have been observed (202–204) while the mammalian TNF receptor superfamily member, neurotrophin receptor (p75/NTR), is structurally similar to the teleost TNF-R1 and binds to the NTR ligand as a dimer (205, 206).

By studying the TNF systems of teleost fish, we may garner greater insights into the evolutionary origins of these important and evolutionarily conserved cytokines and receptors. Indeed the importance of the teleost TNFα proteins to their immune defenses is underlined by the fact that a number of diverse viral fish pathogens encode decoy TNF receptors (207–210).

#### Macrophages and Acute Phase Proteins (APPs) of Bony Fish

During inflammation, activated macrophages secrete cytokines and oxidative radicals that modulate the production of APPs by hepatic cells [reviewed by Gruys et al. (211)]. These APPs opsonize pathogens, activate complement, neutralize enzymes, and scavenge free hemoglobin and radicals.

Acute phase proteins rapidly increase in the blood early after exposure to pathogens or during early inflammatory response. For example, blood levels of C-reactive protein (CRP) may increase as much as 1000-fold and 50% increases in complement proteins and ceruloplasmin (*Cp*) have been observed. The activation of hepatocytes also results in decreased levels of serum transferrin, cortisol-binding globulin, zinc, iron, albumin, and retinol, as well as reduction of free hormones in the blood (212).

While viral infections induce modest acute phase responses (213), bacterial infections elicit potent production of these soluble mediators (211, 214–216). Upon recognition of LPS, monocytes and macrophages also produce gratuitous amounts of pro-inflammatory cytokines (214, 216–219). The termination of APP production is controlled by pro-inflammatory cytokines secreted by macrophages (220, 221).

Bony fish have fully functional repertoires of APPs, which are shared with their mammalian counterparts, as well as additional APPs that are unique to teleosts. The serum-CRP levels of salmonids have been used as indicators of stress in response to xenobiotics (222–224), and protozoan infections (225). The infection of goldfish with *Trypanosoma carassii* increased expression of *Cp*, *crp*, and *serum amyloid A* (*saa*), in the liver, particularly during the early phases of the infection (first 14 days of infection) (225). Serum amyloid-A (SAA) and a serum amyloid P-CRP-like pentraxin proteins have also been identified in salmonids (226), and goldfish (227). *Aeromonas salmonicida* infection of salmon also induced increased levels of SAA protein (226), while goldfish recombinant SAA was shown to induce increased gene expression of *il12p40* and *il1b*, and was chemotactic to primary goldfish macrophages (227). It has also been demonstrated that similar to mammals, trout CRP was capable of activating complement (228).

The salmon *saa* was shown to be upregulated in hepatocytes after their exposure to supernatants from LPS-activated macrophages, or recombinant TNFα, IL-1β, or IL-6 (229). Interestingly, while LPS stimulations increased the expression of the fish pentraxin, *A. salmonicida* infections downregulated the expression of this gene, suggesting that pentraxin may be a "negative" APP (226, 230).

A selective subtractive hybridization (SSH) study of hepatic transcripts in unchallenged and bacterially challenged trout confirmed that a fully functional, broad-repertoire acute phase response exists in teleosts (378). Furthermore, after exposure to distinct pathogens, trout produce overlapping but partially distinct profiles of APPs (231). Catfish also have a well-developed acute phase response following bacterial infections leading to a 50-fold increase in the expression of some of the genes that encode APPs (232). In zebrafish, SSH analysis revealed that zebrafish infected with *A. salmonicida* and *Staphylococcus aureus* possess overlapping as well as unique APPs to those reported in mammals (233).

#### Macrophages and Complement

During a pathogen insult or PAMPs-induced inflammatory responses, there is a significant increase in blood complement levels [reviewed by Mastellos et al. (234) and Markiewski and Lambris (235)]. Most of the mammalian complement components exist in bony fish [reviewed by Nonaka (236)]. When compared with mammals, birds, and amphibians; teleosts have a full set of complement genes with the exception of Factor D, and the absence of MASP-1 and MASP-2 (236). Thus bony fish have multiple forms of several complement components including C3 and C5 proteins (237–241).

Fish complement components have similar pro-inflammatory roles akin to those of mammals. The anphylatoxin, C5a, has chemo-attractive activity (237, 240) and trout C3a enhances fish leukocyte phagocytosis (238, 241). In addition trout C3a, C4a, and C5a has been shown to be chemo-attractive to head kidney phagocytes and PBLs, and enhance phagocytosis of kidney leukocytes (242). The teleost complement biology has been fully addressed in a review by Sunyer et al. (243).

#### ANTIMICROBIAL ROLES OF TELEOST M1 MACROPHAGES

#### Phagocytosis

Phagocytosis is the primordial defense mechanism of all metazoan organisms. During the inflammatory response monocytes/ macrophages and neutrophils, undergo phagocytosis mediated *via* phagocytic receptors or hydrophobic interactions of the phagocyte membrane and the target particles. Once activated, phagocytes release numerous preformed or newly synthesized inflammatory mediators, and are equipped with an armamentarium of antimicrobial responses primarily focused on the pathogens enclosed in the phagolysosomes. Potent antimicrobial compounds generated by activated phagocytes include degradative enzymes (proteases, nucleases, phosphatases, and lipases) and antimicrobial peptides (basic proteins and neutrophilic peptides), which mediate the destruction of phagocytosed pathogens (244–249).

#### Respiratory Burst Response

Macrophage ROS response is a hallmark of these cells' antimicrobial armamentarium and the efficacy of this response often reflects on the ability of macrophages to destroy internalized microorganisms. This response culminates from the assembly of a multicomponent enzymatic complex, the nicotinamide adenine dinucleotide phosphate (NADPH, **Figure 2**) oxidase on the plasma and phagosome membranes, resulting in the transfer of electrons from NADPH to molecular oxygen and thus the production of a superoxide anion (250). In turn, the generated superoxide anions may be converted into other antimicrobial ROS such as hydrogen peroxide (H2O2), hydroxyl radical (OH<sup>⋅</sup> ), and hypercholorus acid (251, 252). The NADPH oxidase complex has six interactive subunits including the cytosolic phagosome oxidases (p40phox, p47phox, and p67phox), and a guanosine triphosphatase Rac 1 or Rac 2, which are mobilized to the gp91phox and p22phox subunits that are located in the plasma membrane (253–258). All of these NADPH oxidase components have been identified in teleosts and fish macrophage ROS responses has been well documented in contexts of PAMP stimulation (259–262), antimicrobial responses (263–265), and recombinant cytokines stimulation such as with TNFα (183, 184, 266), IFNγ (136, 138, 148), and CSF-1 (32).

#### Tryptophan Degradation

Another hallmark of M1 macrophages is their capacity to deplete local tryptophan levels through their upregulated expression of the indoleamine 2,3-dioxygenase (IDO) enzyme (267) (**Figure 1**), which catalyzes this process (268). IDO-mediated tryptophan degradation is closely linked to macrophage antimicrobial responses but also to their immunoregulatory functions, as this tryptophan degradation results in the production of metabolites such as kynurenins ((269), **Figure 3**), which may inhibit T cell proliferation. IFNγ-stimulation of macrophages has been closely linked to inducing the mammalian macrophage IDO response (270–273).

The teleost IDO orthologs (renamed proto-IDOs) are less effective at tryptophan degradation than the mammalian IDOs (274), bringing to question whether these fish enzymes have distinct substrates. Interestingly, *Mycobacterium marinum*-challenged goldfish macrophages upregulate their *proto-ido* gene expression (275), suggesting a possible M1 role for this fish enzyme.

#### Nitric Oxide Response

Classically activated M1 macrophages possess high levels of the inducible nitric oxide synthase enzyme (iNOS/NOS2), which

catalyzes the conversion of L-arginine to L-citrulline, resulting in the production of nitric oxide (NO) (276) (**Figure 4**). As such, iNOS expression serves as a marker of M1 macrophage activation, which may be enhanced by macrophage stimulation with IFNγ, TNFα, and/or microbial compounds (e.g., LPS) (106). The parallel production of superoxide and NO can also result

in the formation of peroxynitrite (ONOO- ), which is a potent antiparasitic/antimicrobial agent (277). The immune mechanism governing the teleost macrophage inducible nitric oxide (NO) appears to be well conserved to those described in mammals.

Akin to its mammalian counterpart, the fish iNOS has putative binding sites for heme, calmodulin, flavine mononucleotide, flavine adenine dinucleotide tetrahydrobiopterin, and NADPH, indicating that this is a highly conserved enzyme (278). The fish macrophage iNOS gene is induced by antimicrobial and inflammatory stimuli such as PAMPs/pathogen recognition (10, 11, 278, 279), pro-inflammatory cytokines (138, 139, 183, 187) and cleaved transferrin products (280, 281). In turn, effective fish macrophage nitric oxide production is integral to fish antimicrobial immunity to a range of pathogens (282–287).

## Sequential Induction of Macrophage Antimicrobial Responses

While mammalian macrophages are thought to be able to undergo simultaneous ROI and NO responses (288), there are several reports suggesting that teleost (primarily cyprinid) fish mount and sequentially deactivate their antimicrobial responses (138, 184, 260, 289, 290). We are aware of only one report describing sequential mammalian macrophage production of ROS followed by NO (291). However, the interdependence of the respective mammalian macrophage respiratory burst, tryptophan degradation, and nitric oxide responses suggest that sequential regulation of macrophage antimicrobial responses is not a strategy that is unique to teleosts and may be a predetermined fail-safe component of all vertebrate macrophage antimicrobial responses.

The respiratory burst and nitric oxide responses are thought of, as two independent macrophage microbicidal mechanisms, where in the induction of one does not depend on the induction of the other (288). However, both responses may be linked to tryptophan degradation. IDO activation requires reduction of its ferric (Fe3+) heme to ferrous (Fe2<sup>+</sup>) heme and there has been some contention regarding the source(s) of electrons used toward this reduction of the IDO heme (292, 293). Interestingly, a prevailing theory suggests that the superoxide anion, derived from the respiratory burst response, is in turn shunted into this enzymatic pathway, serving as this electron source (270, 273, 294, 295). It is interesting to consider that the sequential induction of the respiratory burst response before tryptophan degradation would ensure sufficient quantities of superoxide as a substrate for IDO activity and in turn would repurpose any remaining superoxide anions that had not reacted with the pathogen, thereby also minimizing bystander host cell damage. This notion is supported by the fact that the metabolites from tryptophan degradation are potent scavengers of ROS (296, 297). This in mind, simultaneous induction of macrophage tryptophan degradation and the respiratory burst response would thus be an overall inefficient microbicidal strategy, as the ROS would be actively scavenged by tryptophan catabolites. Thus, sequentially mounting these responses (**Figure 5**) would maximize the targeted effects of the respective responses.

Notably, macrophage tryptophan degradation appears to also be coupled to production of nitric oxide. Picolinic acid, a catabolite of tryptophan degradation (**Figure 3**), synergizes with IFNγ to induce nitric oxide production in murine macrophages (298–302). Picolinic acid exerts this nitric oxide inducing potential *via* a hypoxic responsive element located in the 5' flanking region of the murine iNOS gene, while mutation or deletion of this promoter sequence impairs picolinic acid-induced gene

transcription of iNOS without affecting induction of nitric oxide synthase by LPS (300). Thus, we propose that staggering the kinetics of macrophage tryptophan degradation and nitric oxide production would ensure sufficient quantities of picolinic acid toward the synergistic induction of nitric oxide. In turn, if the respiratory burst response was concomitantly induced with nitric oxide production, then picolinic acid could not exert its nitric oxide inducing effects, as the respiratory burst creates an extremely hyperoxic microenvironment. We thus suggest that the induction of tryptophan degradation before nitric oxide production would facilitate the establishment of a hypoxic microenvironment due to the tryptophan catabolites actively scavenging reactive oxygen intermediates, permitting picolinic acid to augment nitric oxide production by macrophages.

Nitric oxide appears to be the terminal microbicidal response of vertebrate macrophages. In addition to its potent killing effects, nitric oxide is a deactivator of specific enzymes involved in macrophage cytotoxic reactions. Interestingly, NO inhibits both protein kinase C (needed for initiating the ROI response; **Figure 2**) and IDO enzymes involved in the activation of the respiratory burst and tryptophan degradation, respectively (303, 304). Moreover, nitric oxide acts as a negative feedback inhibitor of its own synthesis (305, 306). Therefore, simultaneous induction of nitric oxide, respiratory burst and tryptophan degradation responses would antagonize PKC and thus NADPH oxidase activation (**Figure 2**) and the IDO enzyme. By sequentially inducing the nitric oxide response, subsequent to the respiratory burst and tryptophan degradation responses would ensure that each of these responses would be maximally induced and terminated in a timely manner, thus maximizing these respective antimicrobial responses and minimizing off-target effects of each response.

Based on the above and as outlined in **Figure 5**, we propose that such sequential induction and deactivation of macrophage antimicrobial responses may represent an important and presently poorly explored component of macrophage defenses. As activated macrophages are highly cytotoxic, the interdependence and temporal segregation of their individual microbicidal responses likely represents an inherent way to minimize host cell damage and concomitantly to maximize pathogen elimination. For example, pathogenic microorganisms that are susceptible to ROI are rapidly killed upon phagocytosis by activated macrophages while those pathogens that are resistant to oxidative burst, are often susceptible to subsequent nutrient deprivation and/or antimicrobial attacks. Indeed, ablating the macrophage respiratory burst response while shunting the produced superoxide anion into tryptophan degradation and the subsequent utilization of the picolinic acid from this response toward NO production (**Figure 5**) would maximize the effectiveness of each respective response. This would allow macrophages to divert and target their metabolic energy into distinct, targeted and timely antimicrobial assaults.

The proposed model shown in **Figure 5** does not define macrophage activation in the context of a given individual macrophage, and indeed individual macrophages do not necessarily have to cycle through all of the above responses. Moreover, while much contention remains regarding the functionality of dipartite mammalian macrophage subsets, teleosts clearly possess macrophage sub-populations exhibiting dramatically different kinetics of activation and distinct antimicrobial capacities (9, 148, 184, 290). Notably, cyprinid kidney-derived monocyte-like cultures are considerably more proficient producers of ROS whereas the maturation of these cultures into predominantly macrophage-like cells coincides with their loss of respiratory burst capacities and a concomitant gain of significantly more robust NO responses (184). Presumably, sub-populations of macrophages with distinct antimicrobial potentials coordinate the sequential induction of macrophage antimicrobial responses *in vivo*.

It is unclear why despite considerably more rigorous investigation of the mammalian macrophage, there is more evidence of sequential macrophage antimicrobial responses in teleosts. The central M1/classical activation strategies of mammals and teleosts are best framed by their respective functional polarization by IFNγ. As described above, mammalian species possess single IFNγ molecules that are important for the activation of M1 macrophage ROI and NO responses (**Figure 1**). Intriguingly, many teleost fish possess multiple distinct IFNγ proteins, some of which appear to be potent elicitors of the macrophage ROS, but not NO responses whereas others elicit robust NO production but meager ROIs (148). Thus, we argue that these fish species may have evolved to generate multiple distinct M1 macrophage populations, here denoted as M1a and M1b (**Figure 1**). As an extension of this notion, we argued that fish may have evolved this relatively elaborate classical macrophage activation strategy in order to better coordinate, and when needed, segregate their respective macrophage antimicrobial responses.

#### ACTIVATION OF ALTERNATIVE/M2 TELEOST MACROPHAGES

#### Interleukin-4/13

M2 macrophages have 'anti-inflammatory', or 'pro-healing' phenotypes and the most extensively characterized M2-polarizing agents (sometimes called M2a) are the IL-4 and IL-13 cytokines (**Figure 1**), which are typically produced by Th2 cells, eosinophils, basophils, NK-T cells and certain macrophages subsets (307). IL-4 binds to the IL-4 receptor-alpha and either the IL-4 receptorgamma or the IL-13 receptor-alpha1 chains, culminating in Jak1, Jak3, and Stat6 downstream signaling (104). IL-13 also ligates the IL-13 receptor-alpha2 chain (104). Either of these M2 stimuli result in increased of expression/production of a number of hallmark M2 macrophage components including transglutaminase 2, prostaglandin-endoperoxide synthase, transcription factors IRF4, macrophage mannose receptor, and suppressor of cytokine signaling 1 (SOCS1), all of which are present in fish but await to be functionally linked to teleost M2 macrophages (308–313).

Teleost possess IL-4/13A and IL-4/13B genes with sequence homology to both the mammalian IL-4 and IL-13 cytokines (314). These fish cytokines are thought to have arisen from genome/ gene duplication events, and are present in distinct copies in different fish species (315). Paralogs of IL-4Rα, IL-13Rα1 and IL-13Rα2 have also been identified in teleosts (316, 317), while the recombinant fish IL-4/13A induces B and T cell expansion in an IL-13Rα-dependent manner (318, 319), suggesting that the roles of these fish cytokines possess the immune roles of their mammalian counterparts. The fish IL-4/13A and IL-4/13B are thought to play the M2/anti-inflammatory roles attributed to the mammalian IL-4 and IL-13 (320) and the trout, seabass, grass carp and goldfish recombinant IL-4/13A and IL-4/13B possess many of these anti-inflammatory roles including the upregulation of immunosuppressive genes (TGF-β, IL-10, SAP1, and SOCS3); dampening of pro-inflammatory cytokine gene expression (TNFα, IL-1β, and IFNγ); as well as elevating macrophage/ kidney phagocyte arginase gene expression and arginase activity (321–324). Notably, a true Th2 locus has been identified in spotted gar, consisting of RAD50, IL-4/13 and IL-3/IL-5/GM-CSF (IL-5) (325) while the constitutively high expression of trout and salmon IL4/13A in the thymus, skin and gill tissues have been attributed to immunological tolerance and thus a Th2-like response (320).

#### Arginase

The enhanced capacity to metabolize L-arginine marks an important paradigm between M1 and M2 macrophages and underlines the M2 macrophage. This is intuitive, as M1 macrophage armamentarium is known for its elevated iNOS enzyme, which converts L-arginine to L-citrulline and NO. By contrast, the M2 macrophage arginase enzyme converts L-arginine to L-ornithine and urea (326, 327). The tissue repair capacities of these M2 macrophages in turn reflect their production of L-ornithine, which serves as a precursor for polyamines and proline components of collagen, during tissue repair (328). Notably, the products of these iNOS and arginase enzymatic pathways serve as reciprocal inhibitors of these antagonistic enzymes, promoting the respective M2 or M1 macrophage phenotypes (329).

Mammals possess two arginase isoforms, of which the macrophage gene expression of arginase-1 is induced by IL-4 and IL-13 (330). By contrast, macrophage *arginase-2* gene expression is upregulated by IL-10 and LPS (331). Fish possess both *arginase*-1 and *arginase-2* (332) and like mammals the fish M1/M2 paradigm is outlined by respectively elevated macrophage *inos* and *arginase* genes (10, 11, 279). By contrast to the mammalian M2 macrophages, carp alternative macrophage activation results in the induction of *arginase-2* rather than *arginase-1* expression (10). The facets of fish macrophage M2 polarization and the roles of arginase-2 to in this process have been thoroughly reviewed (190, 333).

## GCs and Interleukin-10

Glucocorticoids and IL-10 stimulation of macrophages culminates in a unique regulatory macrophage phenotype, otherwise known as M2c. GCs diffuse across plasma membranes, resulting in alterations to the expression of a plethora of immune-related genes, which results in these M2c macrophage transcriptional profiles that are distinct from those seen in IL-4/IL-13-stimulated macrophages (334, 335). These M2c macrophage transcriptional changes include decreased inflammatory cytokine gene expression and dampening of ROS production. In line with the immunosuppressive nature of GCs, cortisol increases fish susceptibility to diseases (335, 336) and inhibits fish macrophage NO production (337). Moreover, the simultaneously of fish macrophage cell lines with combined pro-inflammatory stimuli and cortisol results in elevated *il10* gene expression (13), indicating that the cortisol treatment overrides the inflammatory stimuli.

The mammalian IL-10 cytokine signals through a receptor complex composed of IL-10 receptors 1 (IL-10R1) and 2 (IL-10R2), leading to downstream STAT3 activation, which results in decreased gene expression of pro-inflammatory cytokines (338). Macrophage IL-10 production may be elicited by TLR agonists, GCs, and C-type lectins (307). Fish IL-10R1 has been identified in several cyprinids (339, 340), while the IL-10R2 has been reported in salmonids (341). Consistent with the mammalian counterpart, the goldfish recombinant IL-10 down-regulates macrophage ROS responses and inflammatory gene expression (275).

#### THE MACROPHAGE BRIDGE BETWEEN THE INNATE AND ADAPTIVE IMMUNITY

In addition to their roles in early antimicrobial responses, macrophage-lineage cells are crucial to bridging the innate and adaptive arms of the vertebrate immune response. To this end, mammalian macrophages present intracellular pathogen-derived antigens to conventional CD8<sup>+</sup> cytotoxic T cells *via* the MHC I pathway (342); extracellular antigens to CD8<sup>+</sup> T cells in the context of MHCI by means of antigen cross-presentation (343) and extracellular antigens to conventional CD4<sup>+</sup> T helper cells by means of MHCII complex (344). In addition, myeloid cells may present non-protein antigens to unconventional lymphocytes, such as lipid antigens in the context of non-classical MHCI (CD1) to invariant T cells and NK-T cells (345). Moreover, macrophages readily clean up antibody-opsonized pathogens through Fc-receptor-mediated phagocytosis (346). The molecular mechanisms by which teleost fish macrophages bridge the innate and adaptive arms of their respective immune responses are by far the most poorly understood.

## Teleost Antigen Presentation

The fish (salmonid) MHCI peptide-loading complex appears to be fundamentally and functionally similar to that of mammals and the macrophage-like (RTS11) trout cell line has been demonstrated to assemble this antigen presentation complex (347–349). Moreover, trout appear to possess an alternatively spliced variant of MHCI loading glycoprotein, tapasin (349), which is believed to serve as additional regulatory mechanisms in the fish MHCI antigen presentation pathway. While some fish species such as medaka, sharks (350) and zebrafish (351) possess considerable polymorphism within their respective MHCI loci, other species such as Atlantic salmon do not have significant polymorphisms within their classical MHC I antigen processing genes (352). Interestingly, some of the other salmon MHC I assembly and antigen processing genes have been retained as functional duplicates (352). It is thought that these duplicated gene originated from the second vertebrate genome duplication event and are now providing various fish (and some tetrapods such as frogs and birds) with the potential of several different peptide-loading complexes (352).

Several teleost lineages have independently lost key components associated with mammalian antigen presentation and immunological memory including MHCII and CD4 (353–355), although these species exhibit effective immune responses, suggesting that they have evolved alternative immunological strategies for dealing with repeat infections. Moreover, recent genome assembly efforts concomitant with expression analyses have yielded the reconstruction of the evolutionary history of the MHCI (356) and MHCII (357) gene families, demonstrating that teleosts MHC loci have undergone a complex series of gene and genome duplications, culminating in extensive variation in MHC structure and diversity across these animals (358). These distinct teleost species have undoubtedly evolved distinct antigen presentation strategies coinciding with their great diversity across MHCI and II loci. Little is presently know regarding the roles of professional antigen presenting cells such as macrophages in these respective species and it will be most interesting to learn how such cells are integrated within these diverse immune systems.

Distinct fish species also possess several disparate lineages of non-classical MHCs (358), the linkage of which is now believed to have separated before the emergence of tetrapods (359). However, the roles of teleost macrophages and other professional antigen presenting cells in presenting novel antigens in the context of these molecules remain to be explored.

#### Teleost DCs

Myeloid-lineage DCs represent heterogeneous populations of professional antigen presenting cells that share a common myeloid progenitor (macrophage-dendritic cell progenitor) with macrophages and are integral to linking the innate and adaptive immune responses (360). Teleosts appear to possess functional analogs to the mammalian DCs and in particular, salmonids have been documented to possessing putative DCs. For example, salmon possess DC-like cells that express MHCII and CD83 (DC marker), are highly phagocytic and exhibit characteristic DC morphology (361). Trout also clearly possess DC-like cells expressing MHCII and other antigen presentation components, many DC markers (362) and exhibiting robust antigen presentation and lymphocyte activation capacities (363). Moreover, trout appear to possess DCs with cross-presentation capacities that express the same hallmark markers seen on the mammalian DCs specialized to antigen cross-presentation (364, 365). Similarly, the cyprinid zebrafish have been shown to possess cells expressing hallmark DC markers and displaying the capacity to present antigens and induce the proliferation of fish CD4<sup>+</sup> T cells (366).

### The Link Between Teleost Innate and Antibody Responses

It is presently not clear what roles teleost antibodies play in the opsonization of pathogens that enhance macrophage phagocytosis and the canonical Fc receptors responsible for this process in mammals have not been fully elucidated in teleosts (367). However, there are at least five distinct immunoglobulin domaincontaining multi-gene receptor families with some structural and signaling motifs seen in the mammalian Fc receptors (368). Moreover, as members of at least one of this family (LITRs) appear to play roles in phagocytosis (369, 370), it is conceivable that members of this, as well as the other receptor families may function as fish phagocytic receptors for antibody-opsonized targets.

While teleost orthologs to the mammalian Fc receptors remain elusive, teleosts are now known to encode poly Ig receptors (pIgRs) that are capable of binding to fish antibodies (371, 372) and appear to be involved in phagocytosis (373) but are not expressed on fish macrophages or B cells (371). It will be interesting to learn whether distinct subsets of fish phagocytes may acquire the expression of pIgRs immune stimuli.

It is notable that cartilaginous fish (sharks) possess IgMmediated opsonization and cytotoxicity, which is mediated by granulocytes rather than macrophages (374). Turbot macrophage phagocytosis of yeast and beads was greatly enhanced by opsonization with turbot Ig-containing serum fraction however, Ig-opsonized microsporidian spores were not taken up at a greater rate than non-opsonized spores (375). Similarly, brook trout macrophages phagocytosis of *A. salmonicida* was not enhanced when following opsonization of the bacteria by specific fish antibodies although complement-mediated opsonization significantly enhanced bacterial uptake (376). It will be interesting to learn whether the teleost macrophage apparent lack of hallmark Fc receptors reflects in the above observations or whether bony fish macrophages are capable of undergoing antibody-mediated phagocytosis under distinct conditions and through distinct molecular mechanisms.

#### CONCLUDING REMARKS

Akin to the vast heterogeneity of functionally desperate macrophage subsets observed across mammals, teleost fish appear to possess both a spectrum of functionally distinct macrophage subsets as well as a plethora of potential molecular drivers of these distinct lineages. Moreover and in consideration of the strikingly distinct teleost physiologies, evolutionary and pathogenic pressures as well different repertoires of candidate macrophage differentiation factors, these organisms may well utilize (at least partially) distinct macrophage differentiation and activation strategies. It is notable that while many fish species possess multiple isoforms of key macrophage cytokines, functional studies of these moieties have often been limited to one of the several isoforms and have addressed similarities to the mammalian counterparts whilst overlooking some potential functional differences. Indeed, distinct whole genome duplication events and the ploidy of respective fish species can be seen in disparate cytokine copy-number repertoires amongst even closely related fish species (377). These differences are exemplified in copy numbers of hallmark macrophage cytokines such as IFNγ and TNFα across distinct fish. It is generally assumed that the roles of these respective molecules are conserved to those of mammals. However, it is likely that the retention of multiple isoforms within a particular fish species and the often seen expression differences between these fish cytokine isoforms indicate non-overlapping and possibly novel roles for these respective immune mediators. A greater understanding of the mechanisms of fish macrophage antimicrobial immunity is warranted toward aquacultural applications and for the sake of fundamental research. With greater availability of both

#### REFERENCES


fish-specific reagents and genomic resources, the time is ripe for advancing our understanding of these processes.

#### AUTHOR CONTRIBUTIONS

LG, BK, AY, JH, JX, and MB participated in writing the manuscript.

#### FUNDING

This work was supported by the Natural Sciences and Engineering Council of Canada (NSERC; Grant No. RGPIN-2014-06395) to MB. LG thanks the George Washington University for financial support in the form of laboratory start-up funds. BK thanks George Washington University for teaching assistantship support. JH was supported by an NSERC PGS-D doctoral scholarship and JX by China Scholarship Council studentship and Alberta Innovates doctoral scholarship.


IL-34, MCSF1 and MCSF2, ligands of the MCSF receptor. *Mol Immunol* (2014) 53:398–409. doi:10.1016/j.molimm.2012.09.008


*Oncorhynchus mykiss*. *Dev Comp Immunol* (2009) 33:920–31. doi:10.1016/j. dci.2009.03.001


proinflammatory and proliferative activities of its mammalian counterparts, yet it functions in a species-specific manner. *Cell Mol Life Sci* (2004) 61:1331–40. doi:10.1007/s00018-004-4068-1


**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 Grayfer, Kerimoglu, Yaparla, Hodgkinson, Xie and Belosevic. 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.*

# Dysregulation of B cell activity During Proliferative Kidney Disease in rainbow Trout

*Beatriz Abos1†, Itziar Estensoro1,2†, Pedro Perdiguero1†, Marc Faber3 , Yehfang Hu3 , Patricia Díaz Rosales1 , Aitor G. Granja1 , Christopher J. Secombes <sup>3</sup> , Jason W. Holland3 \* and Carolina Tafalla1 \**

*1Centro de Investigación en Sanidad Animal (CISA-INIA), Madrid, Spain, 2 Fish Pathology Group, Institute of Aquaculture Torre de la Sal (IATS-CSIC) Castellón, Madrid, Spain, 3Scottish Fish Immunology Research Centre, Institute of Biological and Environmental Sciences, University of Aberdeen, Aberdeen, United Kingdom*

#### *Edited by:*

*Brian Dixon, University of Waterloo, Canada*

*Reviewed by:* 

*Mark D. Fast, Atlantic Veterinary College, Canada Simon Jones, Fisheries and Oceans Canada, Canada*

#### *\*Correspondence:*

*Jason W. Holland j.holland@abdn.ac.uk; Carolina Tafalla tafalla@inia.es*

*† 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: 05 March 2018 Accepted: 14 May 2018 Published: 31 May 2018*

#### *Citation:*

*Abos B, Estensoro I, Perdiguero P, Faber M, Hu Y, Díaz Rosales P, Granja AG, Secombes CJ, Holland JW and Tafalla C (2018) Dysregulation of B Cell Activity During Proliferative Kidney Disease in Rainbow Trout. Front. Immunol. 9:1203. doi: 10.3389/fimmu.2018.01203*

Proliferative kidney disease (PKD) is a widespread disease caused by the endoparasite *Tetracapsuloides bryosalmonae* (Myxozoa: Malacosporea). Clinical disease, provoked by the proliferation of extrasporogonic parasite stages, is characterized by a chronic kidney pathology with underlying transcriptional changes indicative of altered B cell responses and dysregulated Thelper celllike activities. Despite the relevance of PKD to European and North American salmonid aquaculture, no studies, to date, have focused on further characterizing the B cell response during the course of this disease. Thus, in this work, we have studied the behavior of diverse B cell populations in rainbow trout (*Oncorhynchus mykiss*) naturally infected with *T. bryosalmonae* at different stages of preclinical and clinical disease. Our results show a clear upregulation of all trout immunoglobulins (Igs) (IgM, IgD, and IgT) demonstrated by immunohistochemistry and Western blot analysis, suggesting the alteration of diverse B cell populations that coexist in the infected kidney. Substantial changes in IgM, IgD, and IgT repertoires were also identified throughout the course of the disease further pointing to the involvement of the three Igs in PKD through what appear to be independently regulated mechanisms. Thus, our results provide strong evidence of the involvement of IgD in the humoral response to a specific pathogen for the first time in teleosts. Nevertheless, it was IgT, a fishspecific Ig isotype thought to be specialized in mucosal immunity, which seemed to play a prevailing role in the kidney response to *T. bryosalmonae*. We found that IgT was the main Ig coating extrasporogonic parasite stages, IgT+ B cells were the main B cell subset that proliferated in the kidney with increasing kidney pathology, and IgT was the Ig for which more significant changes in repertoire were detected. Hence, although our results demonstrate a profound dysregulation of different B cell subsets during PKD, they point to a major involvement of IgT in the immune response to the parasite. These results provide further insights into the pathology of PKD that may facilitate the future development of control strategies.

Keywords: *Tetracapsuloides bryosalmonae*, proliferative kidney disease, rainbow trout, B cells, immunoglobulin T, immunoglobulin D, immunoglobulin M

## INTRODUCTION

Proliferative kidney disease (PKD) is a disease of major economic importance to salmonid aquaculture caused by the myxozoan parasite *Tetracapsuloides bryosalmonae* (1). Parasite malacospores are released from infected freshwater bryozoans, the invertebrate host of the parasite. Once in the water, the malacospores gain entry into the fish vascular system *via* the gills (2) and migrate to different organs, the kidney being the main focus of parasite development and proliferation (1). The teleost kidney is the equivalent of mammalian bone marrow as it is the largest site of hematopoiesis and the organ responsible for B cell development (3). In addition, it has also been reported to function as a secondary immune organ (4). When the water temperature rises above 15°C, the kidney responds to the presence of *T. bryosalmonae* extrasporogonic stages with a strong hyperplastic response leading to the regression of urinary tissues and anemia due to reduced erythropoietin production by cells within each nephron (5). Consequently, the fish are much more susceptible to secondary infections, and mortalities up to 95–100% can be reached (1). Below 15°C, the host develops a milder immune response to the parasite that is associated with fewer clinical signs and almost no mortality (6, 7).

Proliferative kidney disease has been defined as an immunopathological condition mediated by an exacerbated host leukocyte response to the parasite (7–9). This induced response seems to be mediated by lymphocytes, which increase in percentage during the course of the disease while granulocyte populations sharply decrease (7, 8). Several transcriptional studies performed during the course of PKD have revealed the regulation of an important number of genes related to Th functions such as IL-4/13A, GATA3, or IL-10 (7, 9). However, further studies on how T cells are affected by the parasite have not been performed and this should also be addressed in the future. In addition, these transcriptional studies have provided evidence that points toward a profound dysregulation of B cells in the kidney during PKD. For example, Gorgoglione et al. (9) analyzed the expression profile of a wide panel of immune molecules in rainbow trout (*Oncorhynchus mykiss*) following a natural exposure to the parasite and found that immunoglobulin (Ig) transcription was strongly upregulated in significant correlation to the stage of kidney pathology (9). Similarly, Bailey et al. (7) established that the transcription of secreted IgM (sIgM) and B cell-related genes such as Blimp1 were strongly induced after an experimental infection with the parasite at 15°C but not at 12°C (7). Finally, a recent study performed by our group demonstrated that the cytokines of the BAFF/APRIL family, known to play a major role in B cell differentiation and survival in mammals, were significantly modulated, along with their receptors, by the parasite and correlated with the transcriptional levels of different Ig isotypes (10).

Fish only express three Ig classes, namely, IgM, IgD, and IgT (designated as IgZ in some fish species) (11). As in mammals, most B cells found in central lymphoid tissues such as spleen and peripheral blood express both IgM and IgD on the cell surface (12). Similar to the situation in mammals, these cells seem to lose IgD during their differentiation to plasmablasts/

plasma cells (12, 13). In addition, as reported in humans in specific mucosal surfaces such as the upper respiratory tract (14), B cells exclusively expressing IgD on the cell surface have also been reported in rainbow trout gills (15) and catfish blood (16), although their precise role is still unknown. Finally, IgT, a teleost-specific Ig, is expressed on the surface of a distinct linage of B cells (17, 18). IgT<sup>+</sup> B cells constitute around 51% of all B cells in the intestinal mucosa, whereas they only represent around 18–27% of all B cells in central lymphoid organs such as spleen, kidney, or peripheral blood (18). This, together with the fact that IgT was the major responder to *Ceratonova shasta* (a myxozoan parasite with intestinal tropism) in mucosal compartments while IgM was the main Ig responding systemically to the parasite led the authors to hypothesize that IgT is specialized in mucosal immunity in teleost fish (18). Further studies supported this hypothesis in describing a similar role of IgT in gills (19) and skin (20) in response to *Ichthyophthirius multifiliis*, a protozoan parasite with strong tropism for both mucosal tissues. Nevertheless, evidence for IgT responses outside the mucosal compartments has also been reported suggesting that IgT might also play an important role in fish systemic responses. For example, Zhang et al. already described a similar capacity of IgM<sup>+</sup> and IgT<sup>+</sup> B cells in the kidney to proliferate and respond to *Vibrio anguillarum* (18). Furthermore, IgT responses were found, in addition to IgM responses, in the spleen of rainbow trout exposed to a systemic viral infection (21) and in the muscle of DNA vaccinated fish (22).

Although fish are able to mount specific antibody responses against a wide range of pathogens, it is generally accepted that the lack of specialized structures where B cells can closely interact with T-helper cells such as germinal centers (GCs) and lymph nodes strongly conditions the immune response generated in this animal group (23). In mammals, three different mechanisms have been described to generate antibody diversity, as the basis of a specific humoral immune response. Before exposure to an antigen, the initial generation of a broad antibody repertoire is achieved early in B cell development by rearrangement of the V, D, and J gene segments to produce Igs with unique Ig heavy- and light-chain variable regions (IGHV and IGLV) (23). A second strategy to increase the Ig repertoire is through junctional diversity, a number of different processes through which different sizes are generated in the heavy-chain sequences by imprecise V(D)J recombination. Terminal deoxynucleotidyl transferase (TdT) is one of the enzymes responsible for the generation of this junctional diversity, through the addition of non-templated (N) nucleotides to the single-strand DNA ends (24). Finally, during B cell differentiation, the genes encoding the variable domains of the heavy and light chains undergo a high degree of point mutations through a process designated as somatic hypermutation (SHM). SHM results in the increased diversity of the antibody pool after which only the cells with higher affinity are selected by follicular antigen presenting cells. This diversification is critical for the generation of an adequate specific immune protection (25) and is mediated by the enzyme activation-induced deaminase (AID), which also plays a role in class switch recombination (CSR), the mechanism through which B cells replace the constant region of the heavy-chain associated with the variable region to produce Ig isotypes with a higher affinity than IgM, such as IgG, IgA, or IgE (26). To date, no CSR has been reported in fish and although all the elements to induce the variability of the B cell repertoire have been identified, fish fail to induce substantial increases in the affinity of their specific antibodies (27).

In this study, following the transcriptional evidence that demonstrated an increase of IgM and IgT in rainbow trout kidney with increasing clinical disease (9), we have undertaken an in depth analysis of B cell and Ig responses in naturally PKD-infected rainbow trout exhibiting early to advanced clinical disease. We have demonstrated that all three Ig isotypes are increased at the protein level in the kidney in response to the parasite and that four different B cell subsets coexist in the infected kidney according to the expression of different Ig isotypes (namely, IgM<sup>+</sup>IgD<sup>+</sup>, IgM<sup>+</sup>IgD<sup>−</sup>, IgD<sup>+</sup>IgM<sup>−</sup>, and IgT<sup>+</sup> cells). In addition, the repertoire analysis of the three Ig subtypes in fish with no clinical signs of disease in comparison to fish with an evident pathology revealed significant changes in VH family usage, clonal expansion, and mutation rate that suggested the involvement of all three Igs in the response to the parasite. These results constitute the first evidence of IgD regulation in response to a pathogen in teleost fish, providing new data to further understand the role of this Ig. Nevertheless, multiple factors point to a prevailing role for IgT during clinical PKD. IgT was the predominant Ig coating the parasite; IgT<sup>+</sup> B cells were found to be actively proliferating in advanced stages of the disease, whereas the percentage of IgM<sup>+</sup> and IgD<sup>+</sup> proliferating cells was much lower; and the IgT repertoire was altered to a greater extent than that of IgM and IgD. Hence, our results show a predominant role of IgT in the immune and pathogenic response to PKD, which provides further evidence of IgT function outside mucosal compartments in fish. Furthermore, the data presented constitutes valuable information for the generation of novel treatments against PKD.

#### MATERIALS AND METHODS

#### Fish Sampling

Two groups of rainbow trout from the same egg source (50–100 g each) were used in this study, as described previously (9). For this, one group of rainbow trout was moved in early April to a commercial trout farm with a history of PKD outbreaks located in Southern England. Clinical signs of the disease were first seen in these fish in early June. One week before the sampling (late July), the second (parasite-naïve) group was moved to this fish farm, so that sampling of both groups was undertaken at a water temperature of 15–16°C. At this point, the fish that had been exposed to the parasite from April exhibited kidney pathology ranging from early to advanced clinical stages (kidney swelling grades 1–4), as established using the kidney swelling index system described by Clifton-Hadley et al. (28), whereas fish that were moved to the farm 1 week before the sampling showed no clinical disease (grade 0 kidney swelling). The presence of *T*. *bryosalmonae* in parasite-infected fish was confirmed by histological examination of posterior kidney smears and by PCR, as described previously (9). In all fish sampled, approximately 100 mg of kidney tissue was removed immediately below the dorsal fin, the kidney area associated with the onset of the clinical disease. Tissue samples were placed into 1 ml of RNA-later (Sigma, St. Louis, MO, USA), kept at 4°C for 24 h and stored at −80°C before RNA and protein extraction. Tissue sections (ca. 4 mm × 3 mm × 3 mm) were excised from the trunk kidney and fixed in 5 ml ice-cold 4% paraformaldehyde (Sigma) for 24 h at 4°C, rinsed repeatedly with ice-cold phosphate-buffered saline (PBS) (10 min at 4°C per rinse), and maintained at 4°C in 70% ethanol until processed for histological analysis.

#### Ig Immunohistochemical Detection

Kidney portions of parasitized fish fixed in 4% paraformaldehyde were processed for paraffin embedding following routine histological procedures. Thereafter, 4 μm-thick tissue sections were mounted on Superfrost Plus slides (Menzel-Gläser). Endogenous peroxidase was quenched with 0.3% hydrogen peroxide, and antigens were retrieved by heating in Tris–EDTA buffer (10 mM Tris base, 1 mM EDTA, pH 9) in a microwave oven for 5 min at 800 W and 5 min at 450 W. Thereafter, non-specific binding was blocked with 5% bovine serum albumin (BSA) in Tris-buffered saline (TBS). Slides were then incubated with specific mouse mAbs recognizing the different Igs. In the case of the anti-trout IgM (20 µg/ml) (29) and the anti-trout IgD (15 µg/ml) (30), this incubation step was performed for 1 h at room temperature (RT), whereas in the case of the anti-mouse IgT (10 µg/ml) (29) the incubation was performed overnight at 4°C. Then, a secondary anti-mouse IgG antibody conjugated with horseradish peroxidase was added and visualized with 3,3′-diaminobenzidine tetrahydrochloride chromogen (EnVision+ System/HRP, Dako). Eventually, sections were counterstained with Gill's hematoxylin, dehydrated and mounted in DPX (di-*N*-butyl-phthalate in xylene). Images were acquired with a Leica DFC320 digital camera connected to a Leica DM LS optic microscope. To establish the coating of parasites with the different Ig isotypes, the presence and absence of IgM-, IgD-, and IgT- immunoreactivity in 100 parasites was recorded from different individual fish (*n* = 6), and statistically significant differences between IgM, IgD, and IgT parasite coating were calculated using a one-way ANOVA followed by a two-tailed Student's *t*-test.

#### RNA and Protein Extraction

Kidney samples were homogenized in 1.5 ml Tri-reagent (Sigma), and tissue debris removed by centrifugation at 12,500 × *g* for 10 min at 4°C. Initially, RNA and DNA were separated from protein by chloroform phase separation. Total RNA was extracted following the manufacturer's instructions, dissolved in TE buffer (pH 8.0) and stored at −80°C until use. Purified RNA was quantified using a Nanodrop spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA) and reverse transcribed into cDNA as described previously (9).

In parallel, 2.25 ml of ice-cold isopropanol was added to the remaining phenol–ethanol supernatant, incubated at RT for 10 min and centrifuged at 12,000 × *g* for 10 min at 4°C. Protein pellets were washed twice in 3 ml of 0.3 M guanidine hydrochloride in 95% ethanol, incubated for 20 min at RT and centrifuged at 7,500 × *g* for 5 min at 4°C. Protein was resuspended in 2 ml absolute ethanol, incubated for 20 min at RT, and pellets solubilized in the presence of urea and sodium dodecyl sulfate (SDS) (400 µl per pellet containing 8 M urea, 1% SDS in Tris– HCl, pH 8.0). To facilitate solubilization, protein pellets were sonicated repeatedly on ice and debris removed by centrifugation (10,000 × *g* for 10 min at 4°C).

#### Western Blot Analysis

Protein concentration in kidney lysates was calculated using the bicinchoninic acid assay (Thermo Scientific) and BSA as a protein standard. Ten micrograms of each lysate were denatured with a 4× loading buffer containing SDS. The mixture was boiled at 100°C for 5 min and loaded onto a denaturing 10% SDS-PAGE gel (Mini-Protean TGX gels, Bio-Rad). Proteins were then transferred onto a polyvinylidene difluoride membrane (Trans-Blot Turbo Transfer Pack, Bio-Rad) using a semi-dry transfer system (Bio-Rad). Membranes were blocked in PBS containing 5% skimmed milk for the detection of IgM, IgD, and IgT and in PBS containing 2% BSA for the detection of α-tubulin. Following this blocking step, membranes were incubated with the corresponding antibody: anti-IgM and anti-IgD were used at 5 µg/ml, anti-IgT at 0.35 µg/ml, and anti-α-tubulin at 0.1 µg/ml (Abcam). Primary antibodies were incubated in blocking solution overnight at 4°C. Membranes were then washed with PBS containing 1% Tween-20 and incubated with a goat anti-mouse IgG-HRP conjugate (GE Healthcare Life Sciences) for 1 h at RT. Finally, the resulting bands were visualized using the ECL system (GE Healthcare Life Sciences). Band intensities were quantified by optical densitometry and normalized against α-tubulin using ImageJ software (National Institutes of Health).

#### Immunofluorescence and Confocal Microscopy

A double immunofluorescent detection of IgD and IgM was performed using 4 µm sections of paraffin-embedded trunk kidney samples. For this, antigen retrieval, blocking of the non-specific binding and incubation with the primary anti-IgD antibody were performed as described earlier for the immunohistochemical detection of IgD. Thereafter, sections were incubated with a secondary AlexaFluor®488 anti-mouse antibody (Life Technologies). A second blocking step was conducted by incubating the sections with TBS containing 5% BSA overnight at 4°C. Thereafter, a biotinylated anti-IgM mAb was added followed by incubation with Streptavidin AlexaFluor®647 (Life Technologies). Sections were then counterstaining with DAPI (1 µg/ml, Sigma), incubated with 0.3% Sudan black B in 70% ethanol for 10 min to remove tissue autofluorescence, rinsed with TBS and mounted with Fluoromount (Sigma). Immunoreactive cells were visualized with a confocal laser-scanning microscope (Zeiss LSM 880) with Zeiss Zen software.

To determine the percentage of proliferating IgM<sup>+</sup>, IgD<sup>+</sup>, or IgT<sup>+</sup> B cells at different stages of clinical disease, 4 µm sections of paraffin-embedded trunk kidney samples were also used. In these assays, the different mouse monoclonal antibodies directed against IgM, IgD, and IgT were co-incubated with an antibody directed against the proliferating cell nuclear antigen (PCNA), an intracellular molecule whose expression and synthesis is linked with cellular proliferation (31). Antigen retrieval and blocking of the non-specific binding were performed as described earlier for the immunohistochemical detection of the Ig isoforms. Tissues were then incubated with the corresponding primary antibody (anti-IgM 20 µg/ml; anti-IgD 15 µg/ml; and anti-IgT 10 µg/ml). Incubation with a secondary goat anti-mouse IgG1 antibody conjugated with AlexaFluor®488 (ThermoFisher) was followed by a further incubation with a mouse IgG2 anti-PCNA antibody conjugated with AlexaFluor®647 (BioLegend) and counterstained with DAPI (1 µg/ml, Sigma). Tissue autofluorescence was then blocked by incubation with 0.3% Sudan black B in 70% ethanol for 10 min, sections were rinsed with TBS and mounted with Fluoromount. Immunoreactive cells were visualized, and images acquired with a confocal laser-scanning microscope. Tissue images were then analyzed in 10 digital fields at 400× magnification of each tissue section with Zeiss Zen and ImageJ software packages. Statistically significant differences of proliferating and non-proliferating IgM-, IgD-, and IgT-immunoreactive cells between grade 0 and grade 2 parasiteexposed fish were analyzed by one-way ANOVA followed by a two-tailed Student's *t*-test. Data that failed the normality or equal variance test were analyzed with Kruskal–Wallis one-way ANOVA on ranks followed by Dunn's method (*P* < 0.05).

#### Analysis of AID and TdT Transcription

As an initial estimate of the level of expansion and differentiation of the B cell response that takes place during the course of PKD, we analyzed the levels of transcription of AID and TdT in kidney samples. For this, real-time PCR was performed in a LightCycler 96 System instrument (Roche) using FastStart Essential DNA Green Master reagents (Roche) and specific primers (shown in Table S1 in Supplementary Material). Amplification to determine the levels of transcription of the different Ig isoforms was performed in parallel in the same samples to determine whether correlations could be established. The efficiency of the amplification was determined for each primer pair using serial 10-fold dilutions of pooled cDNA, and only primer pairs with efficiencies between 1.95 and 2 were used. Each sample was measured in duplicate under the following conditions: 10 min at 95°C, followed by 40 amplification cycles (30 s at 95°C and 1 min at 60°C). The expression of individual genes was normalized to that of trout EF-1α, and expression levels calculated using the 2−ΔCt method, where ΔCt is determined by subtracting the EF-1α value from the target Ct as described previously (32, 33). EF-1α was selected as reference gene according to the MIQE guidelines (34) given that no statistical differences were detected among Ct values obtained for EF-1α in the different samples. Negative controls with no template and *minus-*reverse transcriptase (−RT) controls were included in all experiments. A melting curve for each PCR was determined by reading fluorescence at every degree between 60 and 95°C to ensure only a single product had been amplified. Statistical analyses were performed by one-way ANOVA followed by a two-tailed Student's *t*-test.

#### Repertoire Analysis

To study how clinical disease altered the IgM, IgD, and IgT repertoire, cDNAs from four representative fish with grade 0 and four with grade 2 were used. These cDNAs were amplified using a forward primer specific for a subgroup of the IGHV genes together with a reverse primer specific for IGHM, IGHD, or IGHT genes (Cμ, Cδ, or Cτ) previously designed by Castro et al. (21) (Table S2 in Supplementary Material). Mix reactions for the PCR were as follows: 1 µl of cDNA was used as template using 0.2 mM of each dNTP, 0.2 mM of each primer, and 0.03 U/μl DNA polymerase (Biotools) in 1× reaction buffer containing 2 mM MgCl2. The PCR was programmed as follows: an initial step of 95°C for 5 min followed by 40 cycles of 95°C for 45 s, 60°C for 60 s, and 74°C for 45 s and a final extension step of 74°C for 10 min. Negative controls without cDNA were also included.

An 8 µl aliquot of each PCR was mixed with 2 µl of loading buffer and loaded into a 1% agarose gel stained with SYBR Safe for visualization (Figure S1 in Supplementary Material). In parallel, 2 µl aliquots of each PCR product belonging to the same individual were pooled together. DNA concentrations were measured using the QuBit DNA quantification system (Invitrogen), and the quality was checked using an Agilent 2100 Bioanalyzer (Agilent Technologies). One library per individual was constructed with the TruSeq DNA PCR-Free Library Prep Kit (Illumina, San Diego, CA, USA) according to the manufacturer's protocol. Libraries were pooled together, and paired-end sequencing was performed on an Illumina MiSeq with a MiSeq Reagent Kit v3 (2 × 300 cycles) cartridge (Illumina, San Diego, CA, USA).

#### Sequence Analysis

Raw data were demultiplexed and sequencing adapters and barcodes were removed from the sequences by the MiSeq Analysis pipeline. A quality filter of phred base quality ≥20 was applied to reads. The first 20 nt from reverse primers used in PCRs were used as barcode for the identification of 3′ ends corresponding to the constant gene. Reads from R1 or R2 that perfectly matched a primer sequence were classified into the corresponding isotypes (IgM, IgD, or IgT) using the FASTQ/A Barcode splitter tool (http://hannonlab.cshl.edu/fastx\_toolkit/). The opposite paired end reads, corresponding to the 5′ end of PCR products, were extracted with FASTQ interlacer tool implemented in Galaxy (35). The paired forward and reverse reads were then merged using PEAR software (36) with a minimum overlap size of 12 nt. When no overlap was detected, reads corresponding to the 5′ end were retained and included together with merged reads in successive analysis.

Sequences for each isotype from the eight individuals were compared with available information from *O. mykiss* contained in the International Immunogenetics information system databases (37) using IMGT/HighV-QUEST tool (38). Immune repertoire and SHM&CSR pipelines from Antigen receptor Galaxy tool (39) were used to analyze V(D)J usage, complementarity-determining region 3 (CDR3) characteristics and SHM. To estimate the PCR polymerase and sequencing error ratio, a short fragment without polymorphism covering the first 50 bp from the reverse primer of the IgM constant region was selected. The selected region was mapped to the reference germline sequence from IMGT using BWA (40). The error ratio was calculated as the number of mismatches regarding the reference sequence.

### RESULTS

#### Ig Production at Different Stages of Clinical Disease

As observed in **Figure 1A**, IgM+ B cells were detected in the kidney of parasite-naive fish exhibiting no signs of PKD (grade 0) with a pattern similar to that previously reported for uninfected fish (41). In these animals, IgM<sup>+</sup> B cells were scattered throughout the kidney stroma. Most of these cells showed a lymphocyte-like morphology, that is, round or slightly ovoid cells with a large, round and centrally located nucleus, thus resulting in a high nucleus:cytoplasm ratio. A strong increase in IgM positivity (defined as intensity of immunoreactivity) was observed in fish that exhibited grade 1–2 swelling grade (**Figure 1A**). In these fish, the total number of IgM<sup>+</sup> B cells found in the kidney increased and these cells were mostly grouped in large cell clusters. Furthermore, the positivity of the individual cells was higher than that observed in grade 0 fish, suggesting higher levels of IgM per cell (**Figure 1A**). Interestingly, cells with a larger cytoplasm and an eccentric nucleus were more frequently observed in grade 1–2 fish, pointing to the potential differentiation of some IgM<sup>+</sup> B cells to plasmablasts/plasma cells at this stage. A larger variability in the level of IgM reactivity was observed in kidney samples in fish with the highest level of clinical disease (grade 3–4). However, in most of these fish, the reactivity was reduced to levels either similar to those found in grade 0 fish or slightly higher (**Figure 1A**). Despite this, the reactivity of individual cells remained high when compared with that of IgM<sup>+</sup> B cells from grade 0 fish (**Figure 1A**). All these results obtained by immunohistochemistry were further confirmed by Western blot analysis (**Figures 1B,C**). A significant increase in IgM reactivity was observed in both grade 1–2 and grade 3–4 kidney samples in comparison with that obtained in grade 0 kidney samples (**Figure 1C**).

Although no significant increases in the levels of transcription of IgD had been reported in the kidney in response to PKD (9), our protein analysis demonstrated a significant increase in the levels of IgD protein with increasing clinical PKD, confirmed by both immunohistochemistry (**Figure 2A**) and Western blot (**Figures 2B,C**). While a small number of IgD<sup>+</sup> B cells were found scattered through the kidney stroma in grade 0 fish, the number of IgD<sup>+</sup> B cells increased along with their individual reactivity, especially in grade 1–2 fish (**Figure 2A**). Through Western blot analysis, we established that the level of IgD expression in grade 1–2 kidneys was significantly higher than that of grade 0 kidneys (**Figure 2C**); however, these differences were no longer significant in grade 3–4 kidney samples due to the large variability (**Figure 2C**). In addition, in contrast to what occurred with IgM<sup>+</sup> B cells, the cells remained scattered, and no large clusters were observed at this stage (**Figure 2A**). The different distribution and number of IgD<sup>+</sup> B cells when compared with IgM<sup>+</sup> B cells observed in preparations from the

parasite-infected (swelling grades 1–4) and parasite-naïve (swelling grade 0) fish. Representative images for fish with grade 0, grade 1–2, and grade 3–4 are shown at 20× magnification (left images, scale bars = 50 µm) and at 100× magnification (right images, scale bars = 10 µm). IgM production in kidney samples from each fish group was analyzed by Western blot. (B) Results from two representative individuals in each group are shown. (C) Abundance of IgM was quantified by optical densitometry and normalized against α-tubulin. Mean values in arbitrary units (A.U.) + SEM are shown (*n* = 4). Asterisks (\*) denote significant differences (*P* < 0.05) between values obtained in grade 1–2 or grade 3–4 and grade 0 kidneys (analyzed by a two-tailed Student's *t*-test).

same fish strongly suggests that not all cells that were visualized are IgM<sup>+</sup>IgD<sup>+</sup> B cells, and that single positive cells for either of the two Igs should be present in the kidney. To confirm this point, a double immunofluorescence staining was performed using anti-IgM and anti-IgD in kidney sections from grade 2 fish. In these preparations, three different types of B cells were

identified according to their pattern of IgM/IgD expression (**Figure 3**). These included IgM<sup>+</sup>IgD<sup>+</sup> (that may correspond to non-differentiated B cells) and IgM<sup>+</sup>IgD<sup>−</sup> cells (possibly IgM

values obtained in grade 1–2 or grade 3–4 and grade 0 kidneys (analyzed by a two-tailed Student's *t*-test).

plasmablasts/plasma cells) as well as an IgD<sup>+</sup>IgM<sup>−</sup> population (**Figure 3**) similar to the one previously reported in rainbow trout gills (15) and catfish blood (16).

analyzed by Western blot. (B) Results from two representative individuals in each group are shown. (C) Abundance of IgD was quantified by optical densitometry and normalized against α-tubulin. Mean values in arbitrary units (A.U.) + SEM are shown (*n* = 4). Asterisks (\*) denote significant differences (*P* < 0.05) between

Figure 3 | IgM and IgD detection in rainbow trout kidney infected with *Tetracapsuloides bryosalmonae.* IgD (green) and IgM (red) were immunolabeled in grade 2 kidney sections and counterstained with DAPI (blue). Details show the presence of immunoreactive IgD+/IgM−, IgD+/IgM+, and IgD−/IgM+ B cells. Scale bar in left hand image = 20 µm; scale bars in right hand images = 5 µm.

Finally, we also studied the levels of IgT protein expression in the kidneys of fish from grade 0 to 4. Although IgT has been postulated as an Ig specialized in mucosal immunity, the presence of IgT<sup>+</sup> B cells in the kidney has already been reported (18). In this study, a considerable number of IgT+ B cells with a lymphocytelike morphology were found scattered throughout the kidney stroma in fish with no signs of clinical disease (**Figure 4A**). In fish exhibiting grade 1–2 swelling scores the positivity of IgT in the kidney was massively increased (**Figure 4A**). Although some of these cells appeared to be grouped together, no large cell clusters such as those observed for IgM were apparent (**Figure 4A**). The IgT reactivity decreased in kidney samples from grade 3–4 fish, although the number of IgT<sup>+</sup> B cells was still higher than that observed in fish with no signs of disease. As for IgM, the positivity of individual cells was higher in animals with swelling grades from 1 to 4, suggesting that some IgT+ B cells differentiate to IgT-producing plasmablasts/plasma cells. Supporting this hypothesis, the identification of IgT<sup>+</sup> cells with plasmablast morphology was evident in these fish (**Figure 4A**). All these immunohistochemistry results were confirmed by Western blot analysis (**Figures 4B,C**).

#### IgT Is the Main Ig Isotype Coating *T. bryosalmonae*

To determine whether the different Ig isotypes were coating *T. bryosalmonae*, we evaluated the percentage of parasites with strong surface positivity for the different Ig isotypes in the respective immunohistochemistry preparations (from fish with swelling grades of 1–4). As observed in **Figures 5A,B**, we found that 24 and 10% of the parasites visualized in the preparations stained with anti-IgM or anti-IgD were coated by IgM and IgD, respectively. However, approximately 87% of the parasites visualized in the preparations stained with anti-IgT were coated with IgT. Furthermore, the presence of several IgT<sup>+</sup> B cells surrounding the parasite, in close contact with it, was often evident in these preparations (**Figure 5C**). These results indicate that IgT is the main Ig isotype coating *T. bryosalmonae* in the kidney.

#### B Cell Proliferation at Different Stages of Clinical Disease

The increased expression of IgM, IgD, and IgT observed in the kidney of fish with clear signs of PKD pathology could be attributed to different factors such as increased differentiation of pre-B cells to mature Ig-expressing B cells, mobilization of B cells from other tissues or to local proliferation of B cells. Hence, to clarify this point, we performed double immunofluorescence staining with antibodies against the different Ig isotypes in combination with an antibody against PCNA, an intracellular molecule also linked to cell proliferation in fish, including salmonids (42–44). Through this methodology, we could verify that the number of IgM<sup>+</sup>, IgD<sup>+</sup>, and IgT<sup>+</sup> B cells was higher in grade 2 kidneys than in grade 0 kidneys, as previously observed by immunohistochemistry. Nevertheless, the number of proliferating IgM<sup>+</sup> B cells (PCNA<sup>+</sup>) was similar in grade 0 and grade 2 kidneys (**Figure 6A,B**), accounting for ~35 and ~46% of all IgM<sup>+</sup> B cells, respectively (**Figure 6C**). By contrast, the number of proliferating IgD<sup>+</sup> B cells (PCNA<sup>+</sup>) was higher in grade 2 kidneys than in grade 0 kidneys (**Figure 6A,B**), going from ~31 to ~50% of all IgD+ B cells (**Figure 6C**). Finally, the number of IgT+ proliferating B cells was dramatically increased in grade 2 kidneys when compared with kidneys with no visible signs of pathology (**Figure 6B**). In these tissues, almost all IgT<sup>+</sup> B cells visualized in the preparations (~91%) were proliferating (**Figure 6C**).

#### AID and TdT mRNA Levels in the Kidney Correlate With PKD Progression

As an initial approach to estimate how the Ig repertoire could be affected in response to *T. bryosalmonae*, we studied the levels of transcription of AID and TdT in kidneys affected by PKD at different stages of preclinical and clinical disease.

Our results show that from grade 1–2 the levels of transcription of AID are significantly upregulated in comparison with the levels detected in grade 0 fish (**Figure 7A**). This difference is maintained in grade 3 and grade 4 fish, although with mRNA levels slightly lower than those at grade 1–2 (**Figure 7A**). Despite this, there was a significant correlation between the levels of AID transcription and the extent of clinical disease (**Figure 7A**). To gain some insights on which Ig could be affected by the action of AID, we also evaluated the correlation between AID mRNA levels and mRNA levels of the different Igs. As shown in **Figure 7A**, AID mRNA levels significantly correlated with total IgM, sIgM, and total IgT mRNA levels, but not with IgD mRNA levels (**Figure 7A**).

The transcription of TdT was also upregulated throughout the course of clinical disease, reaching mRNA levels, at grades 3 and 4, significantly higher than those found in fish with no visible signs of pathology (**Figure 7B**). Furthermore, a significant correlation between the levels of TdT transcription and the kidney

(swelling grades 1–4) and parasite-naïve (swelling grade 0) fish. Representative images for fish with swelling grades 0, 1–2, and 3–4 are shown at 20× magnification (left images, scale bars = 40 µm) and at 100× magnification (right images, scale bars = 10 µm). IgT production in kidney samples from each fish group was analyzed by Western blot. (B) Results from two representative individuals in each group are shown. (C) Abundance of IgT was quantified by optical densitometry and normalized against α-tubulin. Mean values in arbitrary units (A.U.) + SEM are shown (*n* = 4). Asterisks (\*) denote significant differences (*P* < 0.05) between values obtained in grade 1–2 or grade 3–4 and grade 0 kidneys (analyzed by a two-tailed Student's *t*-test).

swelling grade was clearly visible (**Figure 7B**). Interestingly, the levels of transcription of TdT significantly correlated with the levels of transcription of all Ig isoforms (**Figure 7B**).

## VH Family Usage During PKD Progression

Previous reports have analyzed in depth the repertoire of IgM, IgD, and IgT in trout exposed to a viral infection (21). Thus, we

followed the protocol described in that study by Castro et al. to amplify heavy-chain rearranged transcripts (IGH V–D–J–C) using a set of isotype-specific primers for IgM, IgD, and IgT, respectively, and a set of IGHV subgroup-specific primers that can amplify all members of the 13 known IGHV groups (Figure S1 in Supplementary Material). We performed this analysis in four grade 0 fish and four fish with grade 2 pathology, given that it was at this stage that the levels of Ig protein expression peaked. All PCRs belonging to one fish were subsequently pooled and deep sequenced to obtain as much information as possible regarding the repertoire of the three Ig isotypes. Using Illumina MiSeq (2 × 300), an average of 2.4 million paired reads per sample were obtained (raw data). Total sequences were cataloged using reverse primers as barcodes for isotype identification. Afterward ~88, ~75, and ~90% of the sequences were classified as productive sequences after IMGT/HighV-QUEST analysis for IgM, IgD, and IgT, respectively (Table S3 in Supplementary Material). Within this set of sequences, IgM was the most common isotype, accounting for ~90% of the sequences, whereas IgD and IgT accounted for around 2–5% of sequences. The accuracy of the PCR and MiSeq sequencing was analyzed using a 50 bp fragment from the IgM constant region. A total of 757 Mbp from the eight samples was used to estimate an average error ratio, which was 4.01 × 10<sup>−</sup><sup>3</sup> . This value was concordant with values observed previously (21) and was considered low enough to evaluate SHM in Ig genes.

Unique sequences defined as V(D)J rearrangements associated with a specific CDR3 amino acid sequence were grouped into what has been previously cataloged as junction sequence types (JST) (23). Globally, a significant increase in number of unique sequences was identified for IgT in comparing grade 0 and grade 2 fish, whereas no significant changes were observed for IgM and IgD (Figure S2 in Supplementary Material). However, significant changes were identified for all isotypes with respect to the VH family usage.

Concerning IgM, as reported before in clonal trout (21), the rainbow trout analyzed also used a broad range of VH families whereas only a few families were not used by any of the fish studied (**Figure 8A**). In comparing the VH family usage between grade 0 and grade 2 fish, we observed an increased use of some of these families (IGHV4S1, IGHV8S7, IGHV11S1, and IGHV13S1) (**Figure 8A**).

When the IgD repertoire was analyzed, we found that, in general, IgD used a lower number of VH families than IgM (**Figure 8B**). In this case, B cells expressing IgD associated with IGHV1S1, IGHV8S1, and IGHV11S1 were significantly overrepresented in the kidney of fish exhibiting grade 2 pathology in comparison to fish with no evident signs of pathology (**Figure 8B**), although the increases detected were modest when compared with those found for other isotypes. However, this different profile of IgD compared with IgM in VH family expansion strongly suggests a certain degree of differential selection.

Similar to the IgM results, B cells expressing IgT use a wide repertoire of VH groups (**Figure 8C**). However, when we compared the IgT repertoire of grade 0 and grade 2 kidneys, we found

along with representative images from grade 2 kidneys at higher magnification (right; scale bars = 5 µm). Note that in non-proliferating cells, nuclei appear blue whereas they appear violet in proliferating cells. Mean number (B) and mean percentage (C) of proliferating IgM+, IgD+, and IgT+ cells were calculated in 10 digital fields (400× magnification) from 6 different individuals. Statistically significant differences (*P* < 0.05) in proliferating IgM+, IgD+, and IgT+ cells between grade 0 and grade 2 fish (analyzed by one-way ANOVA followed by two-tailed Student's *t*-test) are indicated with an asterisk.

that up to 11 VH families were overrepresented in B cells found in grade 2 kidneys in comparison with those present in kidneys with no evident signs of pathology (**Figure 8C**). These were IGHV1S1, IGHV2S1, IGHV4S1, IGHV6S1, IGHV6S4, IGHV8S7, IGHV9S4, IGHV10S1, IGHV11S1, IGHV12S1, and IGHV13S1 (**Figure 8C**). Overall these results indicate that, throughout the course of PKD progression, B cells bearing Igs with specific VH segments are positively selected. As many different VH families are expanded, such results point to a polyclonal activation of B cells. This is especially evident in the case of IgT-expressing cells whereas it is less pronounced for IgM and IgD.

## V(D)J Recombination Repertoire

We also analyzed the V(D)J recombination repertoire in each of the fish studied and created heatmaps showing the relative occurrence of each V(D)J recombination within the repertoire

linear regression is also shown (dotted line) to reveal the correlation between the expression of specific genes and the progression of the pathology, together with

the Pearson product-moment correlation coefficient (*r*) and the statistical significance of the correlation (*P* value), which are indicated in the plots. Pearson product-moment correlation coefficient (*r*) and statistical significance of the correlation (*P* value) between AID (A) or TdT (B) transcription with the

transcription of immunoglobulin (Ig) M, secreted IgM (sIgM), IgT, and IgD are included in the adjacent tables (right).

of each Ig isotype. These heatmaps revealed a large number of trends that were apparent only when analyzing the repertoire in the context of complete V(D)J recombinations. For IgM, we found that the VH families that were previously seen overrepresented in grade 2 fish, were mostly increased in association with IGHJ3, IGHJ4, IGHJ5, and IGHJ6 families (Figure S3 in Supplementary Material). A similar response was observed for IgD (Figure S3 in Supplementary Material), while the VH families that were significantly expanded for IgT in grade 2 fish were always associated with IGHJ1 and IGHJ2 segments (Figure S3 in Supplementary Material). Interestingly, the fact that IgT only uses IGHJ1 and IGHJ2 segments was also reported in a previous study using clonal rainbow trout (21). This specific expansion of V–J pairs for the three Igs was not so evident when V–D associations were studied in the case of IgM and IgD, as the VH families expanded could associate with any D segment (Figure S4 in Supplementary Material). However, again, the IgT expanded

families were preferentially associated with IGHD1, IGHD2, and IGHD3 (Figure S4 in Supplementary Material).

## JST Analysis During PKD Progression

Previous studies have established that JST which are found less than three to five times correspond to naïve non-expanded B cells whereas JST found more than 50 times correspond to highly expanded clones, possibly antibody-secreting cells (23). For IgM, around 80% of the JST were found only once in fish with no clinical signs indicating a broad diversity of the IgM repertoire in the kidney of these animals (**Figure 9A**). However, the JST profile obtained for grade 0 kidney samples was quite different from that previously described for rainbow trout spleen. In spleen, very few JST were found more than 20 times, in concordance to the presence of only a few antibody-secreting cells in a healthy spleen (21). In our studies, there was a considerable number of JST found more than 20 times, with some represented

Figure 9 | Clonal size distribution of junction sequence types (JST) during B cell response to proliferative kidney disease. Bar charts show the average percentage (mean + SD) of JST observed *n* times in the sequence datasets for IgM (A), IgD (B), and IgT (C) from grade 0 kidneys in comparison with grade 2 kidneys (*n* = 4). Statistical differences (*P* < 0.05) between the two groups (analyzed with a two-tailed Student's *t*-test) are shown with an asterisk.

more than 100 times. This possibly reflects the presence of plasmablasts and plasma cells previously reported in the kidney of unstimulated trout (4). Interestingly, as PKD progressed, the number of JST represented 16–100 times was significantly increased (**Figure 9A**), again pointing to an oligoclonal expansion of IgM<sup>+</sup> B cells.

For IgD, around 70% of the JST were found only once in fish with no clinical signs of disease and interestingly this percentage increased in grade 2 kidneys, as well as JST found twice (**Figure 9B**). On the other hand, the percentages of JST found in relatively small numbers (from 4 to 11 times) were significantly decreased in grade 2 kidneys in comparison with grade 0 kidneys, whereas the percentage of JST with large copy numbers (from 101 to 500) significantly increased for IgD (**Figure 9B**). Again, the differences in the JST profile in response to PKD pathogenesis between IgM and IgD strongly suggest a differential regulation of both Igs.

Similar to IgM, the JST distribution for IgT in fish with no clinical signs of disease pointed to the presence of some expanded clones, possibly IgT-secreting plasmablasts or plasma cells already present in these kidneys (**Figure 9C**). Despite this, in these fish, around 70% of the JST were only represented once, indicating a high degree of repertoire diversity (**Figure 9C**). However, with increasing clinical disease, the percentage of JST that were only represented once was significantly reduced to ~60% whereas the number of JST represented from two to eight times was significantly increased (**Figure 9C**). Interestingly, the IgT JST represented more than 100 times was reduced in grade 2 fish (**Figure 9C**). Thus, overall, our results seem to indicate that advanced clinical disease induces the expansion of a large number of IgT clones that became activated but do not fully differentiate into plasma cells.

#### CDR3 Spectratyping of Igs During PKD Progression

Taking into account that the CDR are the regions where BCR bind to their specific antigen and that the CDR3 region of the VH gene is the most hyper-variable region of the BCR genes (45), the determination of CDR3 size by spectratyping has become a powerful tool to analyze the BCR cell repertoire under normal and pathological conditions in mammals and fish (21). Given that B cell clones differ in CDR3 length, the CDR3 length distribution analysis is an estimate of the overall diversity, and any deviation from a bell-shaped Gaussian distribution is indicative of clonal expansions. These clonal expansions can be monoclonal or oligoclonal depending on whether there is a single or several expanded peak(s) (46). In this study, we analyzed the CDR3 length distribution of unique sequences for IgM, IgD, and IgT using either total sequences or sequences that corresponded to the most expanded VH families, but no significant perturbations common to all grade 2 fish were found when compared with the profiles obtained in grade 0 fish in any case (Figure S5 in Supplementary Material). However, when these analyses were performed with total numbers of JST, they also revealed the expansion of B cell subsets with different CDR3 lengths, especially in the case of IgD and IgT (Figure S5 in Supplementary Material). These results suggest that the response to PKD does not involve the clonal selection of a specific B cell subset, but rather that it involves a poly/oligoclonal activation of different B cell subsets. Interestingly, it is worth noting that IgT apparently uses larger CDR3 segments than IgD and IgM, as previously reported by other authors (17, 21) and longer CDR3 have often been associated with autoimmunity and polyreactivity (47).

#### Ig Mutation Rate During PKD Progression

Somatic hypermutation is a process driven by AID inside the GC to generate affinity maturation of antigen-selected GC B cells (48). Although fish do not contain GCs, they are known to express AID and have been reported to be capable of undertaking affinity maturation to a certain degree (27). However, whether SHM takes place extrafollicularly in fish or whether fish have primitive GCs is not yet defined and still under debate (49). In this study, we have quantified the mutation rate for IgM and IgT in IGHV4S1, a VH family preferentially expanded for both Igs and for IGHV11S1 in the case of IgD, as this was one of the VH families significantly expanded for IgD in response to PKD progression (**Figure 8**). In IgM, the percentage of total point mutations was significantly increased in grade 2 kidneys when compared with that observed in grade 0 kidneys (**Figure 10A**). Although the percentage of total point mutations also increased strongly in IgT, the differences were not significant due to a large variability among individuals (**Figure 10B**). Interestingly, in both cases, the mutations were accumulated in the CDR2 region as the disease progressed, with significant differences for both IgM and IgT, while they significantly decreased in the FR3 region (**Figures 10A,B**). Remarkably, in the case of IgT, 100% of the point mutations generated in the CDR2 in grade 2 kidneys were productive mutations (that implied a change in amino acid), whereas only a minor percentage of those accumulated in the FR3 region were productive (data not shown). For both IgM and IgT, most of these mutations were produced within the WRCY AID hotspot motif (**Figures 10A,B**), and mutations targeting this motif increased throughout PKD progression, possibly in concordance with increased mRNA AID expression (**Figures 10A,B**). No significant changes were observed, however, for the percentage of mutations within the RGYW motif, another AID hotspot. Mutations in A/T bases can also occur through the generation of mismatches introduced by error-prone polymerases such as polymerase (pol)ƞ that introduces errors specifically in WA/TA motives (50). Although the percentage of mutations in these sites was much lower than those introduced at the WRCY AID hotspot, the mutations introduced at the WA motif were significantly higher in grade 2 fish than in fish with no signs of pathology in the case of IgT (**Figures 10A,B**). A completely different SHM profile was obtained for IgD. In this case, although the percentage of total mutations was not significantly increased in grade 2 kidneys when compared with grade 0 kidneys, the percentage of mutations within the FR3 regions significantly increased along with a significant decrease in the mutations within the CDR2 region (**Figure 10C**). Similarly, in human IgD<sup>+</sup>IgM<sup>−</sup> B cells, the replacement mutations found in IgD in these cells were not concentrated within the CDRs (51). In this case, even though the higher mutation frequencies

were observed with the WA/TA motives targeted by the (pol)ƞ polymerase, their frequencies were reduced in response to PKD progression (**Figure 10C**), whereas those found with the WRCY AID hotspot significantly increased (**Figure 10C**), suggesting an involvement of AID in IgD SHM despite the fact that AID mRNA levels did not correlate with those of IgD (**Figure 7**).

#### DISCUSSION

A hallmark of adaptive immunity mediated by antibodies is the ability of B cells to generate immunological memory through which B cells can respond more rapidly and robustly producing specific antibodies upon re-exposure to pathogens. However, many parasites have developed strategies to manipulate the B cell response and escape adaptive immunity (52). Hence, although specific antibodies have been demonstrated to provide protection against some protozoan infections in mammals, it has also been shown that the B cell responses elicited by some of these parasites can be responsible for the induced pathogenesis. For example, *Trypanosoma cruzi*, *Neospora caninum*, and *Trypanosoma brucei* have the capacity to deplete B cell precursors in the bone marrow, thereby limiting the number of B cells in the periphery (53–55). By contrast, the spleen of individuals affected by many of these protozoan infections shows a marked cellular hyperplasia as a consequence of an intense B-cell response. Thus, parasites such as *T. cruzi* (56) or *Plasmodium* spp. (57) elicit strong extrafollicular B cell responses that lead to a strong hypergammaglobulinemia. Interestingly, while the total amount of antibodies markedly increases at early stages of infection with these parasites, the specific antibody titers are quite limited (56). Therefore, this polyclonal activation is a strategy used by some parasites to escape the host-specific immune response by means of diluting pathogenspecific antibodies while increasing irrelevant antibodies.

Previous transcriptional studies conducted in rainbow trout affected by PKD demonstrated an increased transcription of IgM and IgT that strongly correlated with the pathological score (9). Hence, in this study, we examined the production of IgM, IgD, and IgT at the protein level at different stages of clinical disease by both immunohistochemistry/immunofluorescence and Western blot analysis. Our results confirm a strong induction of Ig production in response to *T. bryosalmonae* in the kidney of infected fish. In the teleost fish, the kidney functions as both a hematopoietic tissue and as a secondary immune organ (4). Thus, while in the anterior kidney B cells develop and most proliferating B cell precursors are found, the posterior kidney houses significant populations of partially activated B cells and plasmablasts (4). This increased response, confirmed by both immunohistochemistry and Western blot, affected all Ig isotypes, namely, IgM, IgD, and IgT and was more robust in fish kidneys with grade 1–2 clinical swelling than in kidneys in which the disease had progressed to level 3–4. It should be noted that as all fish used in this study were sampled at one time point and fish with swelling grades from 1 to 4 were exposed to the parasite for the same period of time, the differences in swelling grades observed among individual fish might also be a result of different host responsiveness to the parasite. In any case, the repertoire analysis for all three Igs suggested that this increased expression of all three Igs is a result of a oligo/polyclonal B cell activation which might be an escape mechanism triggered by *T. bryosalmonae*, being also, at least in part, responsible for the pathogenesis associated with the parasite proliferation. These results seem in correlation with the serum IgM hyperimmunoglobulinemia that is induced by PKD infection (58).

Previous transcriptional studies had reported significant upregulation of IgM and IgT mRNA levels with increasing clinical disease, but not of IgD mRNA levels (9). However, an evident upregulation of IgD protein has been demonstrated in this study. Previous studies performed in mice have established that the basal level of IgD transcription remained constant in different B cell subsets regardless of their levels of IgD surface expression (59), leading to the hypothesis that posttranscriptional processing plays an important role in the changes in expression of IgD during B cell differentiation. Thus, it could be possible that posttranscriptional regulation also conditions the amount of IgD produced by different B cell subsets in teleosts. To date, the precise role of IgD in the immune response is still largely unknown in mammals and fish (14). Although generally attributed a minor immune role due to the fact that IgD-deficient mice are able to mount normal T-dependent and T-independent immune responses (60), the identification of IgD-secreting cells in specific mucosal surfaces in mammals and the fact that it has been conserved throughout evolution points to a relevant but still unknown role (14). In this study, we provide strong evidence of a specific regulation of IgD in response to antigenic stimulation at the protein level, as well as a clonal expansion of IgD-expressing B cells in response to PKD progression for the first time in teleost fish. Since we have demonstrated the presence of IgD<sup>+</sup>IgM<sup>−</sup> cells in parasite-exposed kidneys, it seems probable that these are the cells clonally expanded in response to the parasite. In mammals, despite the absence of a repetitive S region upstream of Cδ, a non-canonical CSR process that is still not fully understood generates cells that exclusively produce IgD after an AID-mediated deletion of Cµ (51, 61). Whether teleost IgD<sup>+</sup>IgM<sup>−</sup> cells also appear as a consequence of a non-canonical CSR is something that should be further explored. Interestingly, although the SHM profile of IgD was quite different to that of IgM and IgT, a certain degree of AID-mediated SHM was found associated with PKD disease progression. In humans, IgD<sup>+</sup>IgM<sup>−</sup> B cells show a very high rate of SHM (51), whereas equivalent population in mice show a very low mutation rate (62). Thus, the identification of IgD<sup>+</sup>IgM<sup>−</sup> cells in the kidneys of PKD-infected fish, as well as the demonstration of SHM and clonal expansion of IgD throughout the progression of the disease, challenges the hypothesis that IgD functions simply as an antigen receptor and adds to previous evidence in mammals that points to an important role of IgD in specific humoral responses (51, 63).

Despite the regulation of IgM and IgD throughout the course of PKD, our results unequivocally point to a prevailing role of IgT in the immune response to the parasite given its dominant role in parasite coating, B cell proliferation, clonal expansion, and SHM. Interestingly, already in 2002, Chilmonczyk et al. demonstrated that the lymphocytes that were proliferating in response to PKD were mostly IgM<sup>−</sup> (8). Even though it could be possible that some T cells are also proliferating in response to the parasite, their results are in concordance to the preferential proliferation of IgT<sup>+</sup> B cells in PKD. The fact that IgT is the main Ig isotype that is modulated in the kidney during the progression of PKD seems surprising given that IgT has been reported to be as a specialized mucosal Ig (18–20); however, previous reports have also demonstrated nonmucosal roles for IgT (18, 21, 22). Interestingly, in all the studies that pointed to a mucosal role of IgT a parasite model was used (18–20), therefore, it also seems plausible that IgT plays a leading role in the immune response to myxozoan/protozoan parasitic infections, given the absence of IgE in teleosts. Previous studies support this idea, since IgT was the only Ig isotype induced in the kidney upon infection with a myxozoan parasite in gilthead sea bream (*Sparus aurata*) (64). Our study also revealed several aspects of IgT not previously explored in fish. For example, we have seen that similar to IgM, expanded IgT JST sequences are present in the rainbow trout kidney, suggesting the presence of IgT plasmablasts/plasma cells in animals with no evident signs of disease. Thus, whether the kidney constitutes a survival niche for IgT plasmablasts/plasma cells as described for IgM plasma cells (4) is something worth exploring further. In addition, we have confirmed for the first time SHM for IgT, demonstrating that through the course of PKD, mutations could accumulate for IgT in the AID hotspot WRCY. The fact that AID is able to mutate fish Igs in the absence of recognizable GCs also merits additional investigation in the future.

In conclusion, we provide novel results that substantiate the upregulation of all three Ig types in response to PKD at the protein level, not seen previously by transcript studies (e.g., in regards to IgD). Additional repertoire analysis further demonstrates that the upregulation is a consequence of a polyclonal expansion of the different Ig subsets found in the kidney during PKD, given that no public or private monoclonal responses similar to those previously reported in response to viral infection (21) were identified. This study confirms an important role of IgD in the humoral response to the parasite, which includes the appearance of IgD<sup>+</sup>IgM<sup>−</sup> B cells, SHM, and clonal expansion of some IgD-expressing B cell subsets. Nonetheless, in addition to the regulated IgM and IgD responses, IgT was also regulated in the kidney in response to the parasite, playing a prevailing role that confirms the fact that IgT is an important player outside of mucosal compartments. Our results provide important novel data to further understand the regulation of Ig synthesis in teleosts and delivers valuable information to aid future intervention strategies against PKD.

#### ETHICS STATEMENT

This study was approved by Instituto Nacional de Investigación Agraria y Alimentaria (INIA) Ethics Committee (ORCEEA 2016-021).

## AUTHOR CONTRIBUTIONS

BA performed the immunohistochemical analysis of Ig production, Western blots and real-time PCR analyses. IE determined the parasite coating and performed all immunofluorescence assays and analyses. PP undertook all the analyses related to the Ig repertoires. MF and YH performed the fish sampling and

## REFERENCES


produced cDNAs and protein extracts. CT and JH designed the experiments. CT wrote the main body of the paper with contributions from JH, AG, PR, and CS.

#### ACKNOWLEDGMENTS

Lucía González, Esther Morel, and M. Camino Ordás are greatly acknowledged for technical support. The authors would also like to thank Dr. Belén de Andrés for help with the repertoire analysis.

#### FUNDING

This work was supported by the European Research Council (ERC Consolidator Grant 2016 725061 TEMUBLYM) and the European Commission under the H2020 Programme (Grant H2020-634429 ParaFishControl). IE was recipient of APOSTD/2016/037 grant by the "Generalitat Valenciana" and YH was recipient of a PhD Studentship from the Ministry of Education, Republic of China (Taiwan). JWH was supported by BBSRC grant BB/K009125/1 and SNSF grant CRSII3\_147649-1. PDR was funded by grant T1-BIO-1672 from the "Comunidad de Madrid".

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.01203/ full#supplementary-material.


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

*Copyright © 2018 Abos, Estensoro, Perdiguero, Faber, Hu, Díaz Rosales, Granja, Secombes, Holland and Tafalla. 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.*

# Behavioral Fever Drives epigenetic Modulation of the immune response in Fish

*Sebastian Boltana1 \*, Andrea Aguilar1 , Nataly Sanhueza1 , Andrea Donoso1 , Luis Mercado2 , Monica Imarai <sup>3</sup> and Simon Mackenzie4*

*<sup>1</sup> Interdisciplinary Center for Aquaculture Research (INCAR), Department of Oceanography, Biotechnology Center, University of Concepción, Concepción, Chile, 2Grupo de Marcadores Inmunológicos, Facultad de Ciencias, Instituto de Biología, Pontificia Universidad Católica de Valparaíso, Valparaíso, Chile, 3 Laboratory of Immunology, Center of Aquatic Biotechnology, Department of Biology, Faculty of Chemistry and Biology, University of Santiago of Chile, Santiago, Chile, <sup>4</sup> Institute of Aquaculture, University of Stirling, Stirling, United Kingdom*

#### *Edited by:*

*Lluis Tort, Universidad Autónoma de Barcelona, Spain*

#### *Reviewed by:*

*Magdalena Chadzin*ˊ*ska, Jagiellonian University, Poland Elz*̇*bieta Z*̇ *bikowska, Nicolaus Copernicus University in Torun*ˊ*, Poland*

#### *\*Correspondence:*

*Sebastian Boltana sboltana@udec.cl*

#### *Specialty section:*

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

*Received: 30 January 2018 Accepted: 17 May 2018 Published: 04 June 2018*

#### *Citation:*

*Boltana S, Aguilar A, Sanhueza N, Donoso A, Mercado L, Imarai M and Mackenzie S (2018) Behavioral Fever Drives Epigenetic Modulation of the Immune Response in Fish. Front. Immunol. 9:1241. doi: 10.3389/fimmu.2018.01241*

Ectotherms choose the best thermal conditions to mount a successful immune response, a phenomenon known as behavioral fever. The cumulative evidence suggests that behavioral fever impacts positively upon lymphocyte proliferation, inflammatory cytokine expression, and other immune functions. In this study, we have explored how thermal choice during infection impacts upon underpinning molecular processes and how temperature increase is coupled to the immune response. Our results show that behavioral fever results in a widespread, plastic imprint on gene regulation, and lymphocyte proliferation. We further explored the possible contribution of histone modification and identified global associations between temperature and histone changes that suggest epigenetic remodeling as a result of behavioral fever. Together, these results highlight the critical importance of thermal choice in mobile ectotherms, particularly in response to an infection, and demonstrate the key role of epigenetic modification to orchestrate the thermocoupling of the immune response during behavioral fever.

Keywords: behavioral fever, gene regulation, lymphocyte proliferation, cytokine release, epigenetic modification

## HIGHLIGHTS


## INTRODUCTION

Thermoregulatory behavior is a critical factor influencing the functional responses of individuals in aquatic environments (1). Surprisingly, it remains unknown how ectothermic vertebrates regulate thermal preference, where central thermostats are located in the brain and how variations in thermal choice impact upon molecular and cellular interactions (2–4). Behavioral fever has been shown to impact the immune response by drive specific modification of the molecular regulation (5, 6). However, this does not involve that fever has an adaptive impact, but that it can have a key incidence on the survival depending on the specific pathological conditions (7) where a balance between the killing of invading pathogens and the specific molecular/ cellular response of each tissue have to be taken into account (8). For example, it has recently been shown that some viruses carry genes that specifically inhibit the fever response to increase their growth (9, 10), or as in the case of *Neisseria meningitidis*, the increase of the temperature leadsto expression of mechanismsto specifically avoid the host immune responses that are triggered during fever (11).

In general, behavioral fever response has been described across a range of fish species highlighting the ubiquitous nature of this response. Acute variations in body temperature achieved by displacement through a thermal gradient may influence the regulation of the immune response at several levels of interaction reaching from molecular to whole organ response. In environmental settings where thermal choice behavior is limited by the environment such as extreme temperatures or constant thermal conditions acute stress responses and immune compromise are observed (5). In thermally stressed fish, for example, increased temperatures with no behavioral choice the thermal stress negatively impacts upon the glucocorticoid response (GC), innate immunity (12, 13), the oxidative stress response (14, 15), and causes reduced lymphocyte numbers and proliferation rates (16). Interestingly, under adequate environmental conditions that include a thermal choice fish are able to successfully orchestrate regulatory responses to potentiate immunity (17) or improve metabolic and growth performance (18). Behavioral fever in response to pathogen challenge has been shown in fish as a response to bacterial (tilapia), cytokine (trout), and viral pathogens (carp, zebrafish), in lizards as a response to lipopolysaccharide (LPS) of bacterial wall of *Escherichia coli* (19) and invertebrates highlighting the evolutionary importance of this response in ectotherms (20–23).

Existing evidence indicates that behavioral fever is an evolutionarily conserved response with significant adaptive value (9, 24–28), although there is a lack of scientific knowledge regarding the underpinning mechanisms influenced by thermal coupling of the immune response during behavioral fever. The mechanisms by which increased temperature affect the defense response are complex and may include epigenetic changes impacting specific immunological processes. Recent reports have documented the role of the epigenetic regulation on the immunological pathways underlying the defense response (29, 30). For example, mammal macrophages previously challenged with fungal cell-wall compound β-glucan or bacterial LPS exhibit genome-wide changes in the trimethylation of histone H3 at Lys4 (H3K4me3), monomethylation of histone H3 at Lys4 (H3K4me1), and acetylation of histone H3 at Lys27 (H3K27ac), as well as transcriptional remodulation (31). In particular, the molecular mechanisms affected or influenced by temperature choice and the translation into potentiated immune response remain unknown. In ectotherms, the influence of thermoregulation upon gene expression has been supported by correlations between temperature, regulatory response, and variation in gene expression (32, 33). However, the contribution of epigenetic regulation including DNA methylation and histone modification remains unexplored [e.g., Streelman et al. (34); Baalsrud et al. (35); and Mallard et al. (36)].

In this study, we proposed to explore the epigenetic regulatory mechanisms influenced by behavioral fever. To test this hypothesis, we used *Atlantic salmon*, *Salmo salar*, that had access either to a thermal choice through a thermal gradient or were held at a constant temperature. Experimental animals were subjected to a viral challenge, infectious pancreatic necrosis virus(IPNv). Using this dynamic experimental set up, we are able to test for an association between gene regulation, epigenetic modification, cellular response, and survival to define the impact of temperature on the response to viral infection. Our findings highlight the role of the epigenetic modifications driven by behavioral fever that contribute to the observed variation in gene expression, the variation in lymphocyte cell populations and cytokine production that ultimately impact on survival.

#### MATERIALS AND METHODS

#### Animal and Experimental Conditions

All animal experiments conformed to the British Home Office Regulations (Animal Scientific Procedures Act 1986), the animal research was authorized by the Universidad de Concepcion Institutional Animal Care and Use Committee. Thermal experiments were carried out at the ThermoFish Lab, Biotechnology Center, University of Concepcion, Concepción, Chile. *S. salar's* embryos were obtained from AquaGen S.A. (Melipeuco, Chile) in December 2015. Hatchery conditions used were as described by Boltaña et al. (18). Briefly, fish embryos were maintained in a recirculating freshwater systems (temperature = 7 ± 0.7°C) with a constant photoperiod of light: dark (LD) until hatching. Then, when 95% of the embryos hatched (i.e., 30 days posthatching), temperature was gradually incremented until reaching 15 ± 0.9°C (**Figure 1**). The fish were maintained for 9 months and fish were fed twice a day on a commercial diet (Biomar, S.A., Puerto Montt, Chile).

#### IPNv Challenge and Mortality Rates

Following the fish were challenged with IPNv. In brief, *in vivo* infection of salmon with IPNv was performed by immersion using dechlorinated water from stock tanks following protocols previously described (37). Clarified supernatant from IPNvinfected CHSE-214 cell monolayers (10 × 10<sup>5</sup> PFU/mL<sup>−</sup><sup>1</sup> ) was added to 5 L water tanks containing the fish (*n* = 60). After 120 min, fish were separated in two groups and placed in the two-different experimental thermal set up: (a) constant temperature (mean temperature 15 ± 0.9°C, no fever) and (b) temperature gradient (mean temperature 15 ± 7.4°C, fever). In parallel, 60 fishes in another tank were treated by adding 100 mL virus free cell culture supernatant to the water (mock infected), separated into two groups and placed in two-different conditions; (c) constant temperature (mean temperature 15 ± 0.9°C, control RTR) and (d) temperature gradient (mean temperature 15 ± 7.4°C, control WTR). All temperatures were recorded

represents the restricted (no fever group ΔT 0.9°C) and orange represents the wide range (fever group, ΔT 7.4°C). Experimental groups were (i) virus infected with infectious pancreatic necrosis virus (IPNv) (10 × 105 PFU/mL−<sup>1</sup> ) by immersion under constant normothermic (preferred temperature) conditions (no fever), (ii) virus infected with IPNv (10 × 105 PFU/mL−<sup>1</sup> ) by immersion in a temperature gradient (fever), (iii) control with no gradient (control RTR), and (iv) control in a temperature gradient (control WTR). (B) IPNV viral load during behavioral fever. Mean and SE of IPNV copy number per nanogram total RNA for the WP2 IPN viral segment. Box and whiskers plots registered differences in mRNA abundances between control and viral challenged individuals. (C) Indirect ELISA detection of proinflammatory cytokine release on the plasma in response to IPNv challenge. Box and whiskers plots registered differences between the unlike thermal group. Significance symbols correspond to the *p*-values for two-way ANOVA test between the different individuals. Unlike letters denote significantly different mRNA levels between the treatment.

each day at the same time of the day. Experimental groups were (i) virus infected with IPNv (10 × 10<sup>5</sup> PFU/mL<sup>−</sup><sup>1</sup> ) by immersion under constant normothermic (preferred temperature) conditions (no fever), (ii) virus infected with IPNv (10 × 10<sup>5</sup> PFU/mL<sup>−</sup><sup>1</sup> ) by immersion in a temperature gradient (fever), (iii) control with no gradient (control RTR), and (iv) control in a temperature gradient (control WTR). Three independent replicates by experimental group were carried out (*n* = 10 by replicate). Three fish by replicate (*n* = 9) were anesthetized by MSS2 and the pronephros and blood were sampled 24 h postinfection. Blood samples were individually collected from the caudal vein and centrifuged. Supernatant sera were kept frozen at −20°C. To achieve the mortality analysis, the remaining fish (*n* = 7 by replicate) were keep in the behavioral tanks and scored daily for a 10-day period; we recorded abdominal distension, exophthalmia, impaired swimming, and skin/fin base hemorrhages. To verify that IPNv was the cause of death, liver and kidney and Gills of moribund or dead fish were fixed in 10% buffered formalin for histopathological and immunohistochemical examination (data not shown). In all groups, behavioral data were recorded as described following.

#### Behavioral Studies

The experimental thermal gradient was carried out in 2.5 m<sup>3</sup> tanks (105 cm × 15 cm × 15 cm) divided with five transparent Plexiglas screens to create six equal interconnected chambers. Each screen had a hole at the center (3 cm diameter; 10 cm from the bottom) to allow fish to move freely between chambers. Three video cameras provided continuous monitoring of each tank chamber. During the experiment, temperatures were recorded for 10 s every 15 min throughout 24 h (96 recorded events). Four groups of fish (*n* = 10 for each group) were introduced into chamber 4 in the evening (6:00 p.m.) and filming began at 6:00 a.m. the next day, providing a 12-h acclimation period. The distribution of fish into the six compartments was monitored over time with video cameras and the number of fish in each compartment was

counted manually from the images captured at each successive 15 min, resulting in 96 measurements per day. Thermal gradients were achieved with a mean difference in temperature of 13.564°C between chambers 1 and 6 by simultaneously heating chamber 6 (mean temperature = 20.725 ± 0.712°C) and cooling chamber 1 (mean temperature = 7.161 ± 0.476°C). All temperatures were recorded each day at the same time of the day. The mean number of fish observed per day in each compartment + SD (*n* = 30) was registered for each experimental group (no fever, fever, control RTR, and control WTR).

#### IPNv Recovery (qPCR)

Pronephros samples were used to inspect the virus load after 24 h post-infection. Total RNA used for high-throughput transcriptome sequencing, RT-qPCR and flow cytometry analysis, was previously analyzed in order to attest and confirmed the successful completion of infection with (IPNv). To do so, an RT-PCR was performed forIPNv load estimation by targeting the virus segment VP2 region using primer WB117 and a Universal ProbeLibrary probes as previously described (38–40).

#### Pronephros RNA Extraction

Nine individuals from each challenge (WTR control, control mock, fever, and no fever) were selected for RNA extraction using 30 mg of tissue after 24 h post-infection. Total RNA was individually isolated using Ribo-Pure™ Kit (Ambion®, USA) according to the manufacturer's instructions. RNA concentration and purity were estimated using the NanoDrop 1000 Spectrophotometer (Thermo Scientific, MA, USA), while the RNA integrity number (RIN) was evaluated through the 2200 TapeStation (Agilent technologies, CA, USA) using the R6K screen tape. Samples with RIN ≥8 and 260/280 ratio ≥1.8 were used for library construction.

#### High-Throughput Transcriptome Sequencing: Library Construction and Illumina Sequencing

Total RNA from pronephros of each condition group (no fever, fever, control RTR, and control WTR) were pooled, considering three randomly selected individuals per replicate (*n* = 9 individuals by treatment). From these pools, six barcoded libraries were constructed using the KAPA Stranded mRNA-Seq Kit (KapaBiosystems, MA, USA) according to the manufacturer's instruction. Briefly, from 3 µg of pooled total RNA, mRNA was isolated and fragmented, followed by double-stranded cDNA synthesis. Later, ends were repaired and adenylated at the 30 end in order to perform the NEXTflex® RNA-Seq Barcodes (BiooScientific, TX, USA) ligation and final PCR amplification. Library validation was based on length distribution as estimated with the 2200 TapeStation (Agilent Technologies, CA, USA) using D1K screen tape and reagents (Agilent Technologies, CA, USA). Libraries with mean length peaks above 250 bp were used for sequencing and were quantified by qPCR using the Library Quantification Kit Illumina/Universal (KapaBiosystems, MA, USA) according to the manufacturer's instructions. Two biological replicates were used for each condition, and sequencing was performed with the Miseq (Illumina) platform using a run of 2 × 250 paired-end reads at the Laboratory of Biotechnology and Aquatic Genomics, Interdisciplinary Center for Aquaculture Research (INCAR), Universidad de Concepción, Chile. The *de novo* assembly sequence data are available from corresponding author on request.

#### Gene Ontology (GO-DAVID Analysis) and Interactome Analysis

Enrichment of specific gene ontology (GO) terms among the set of probes that are specific to challenges was assessed to correlate a specific set of mRNAs within a pronephros. In all GO analyses, Ensembl gene identifiers were tested using DAVID Bioinformatics Resources<sup>1</sup> (41, 42). Enrichment of each GO term was evaluated through use of the Fisher's exact test and corrected for multiple testing with FDR [pFDR < 0.05 (43)]. We applied a Bonferroni correction to account for multiple tests performed. Each gene set comprised of at least four transcripts that shared the same GO biological process or annotation term. The final GO immune-enrichment analysis was carried out with the Cytoscape 3.5.1.<sup>2</sup> Topological analysis of individual and combined networks was performed with Network Analyzer, and jActiveModules 2.2 was used to analyze network characteristics (44, 45). GO analyses were conducted with the Biological Network Gene Ontology (ClueGO, version 2.0) plugin (46) used for statistical evaluation of groups of proteins with respect to the current annotations available at the Gene Ontology Consortium.<sup>3</sup> In addition, we conducted a complementary analysis with ClusterMaker cytoscape plugin (47), using the MCL algorithm to search protein–protein interaction network modules derived from tandem affinity purification/mass spectrometry (TAP/ MAS). This approach clustered the network into modules based on PE score to indicate the strength of the node association and given a fixed set of genes with high protein–protein affinity (interactome cluster nodes).

#### Chromatin Extraction

Three individuals from each challenge group (no fever, fever, control RTR, and control WTR) were selected for chromatin extraction using Chromatin Extraction Kit (Abcam, Cambridge, United Kingdom) according to the manufacturer's instructions. For each individual, 10 mg of pronephros (12 h post-infection) was cut into small pieces and transferred to Dounce homogenizer. Then 50 µL of extracted chromatin was transferred to microTUBE AFA, and fragmented into 100–300 pb fragments using the following program: duty cycle = 10%, cycles per burst = 200, temperature (bath) = 4°C, cycle time = 90 s ON and 10 s OFF, cycles = 16. The size ofsheared chromatin was verified on agarose gel 1% electrophoresis (120 min, 60 V) before starting the immunoprecipitation step. Length of sheared chromatin obtained was between 100 and 300 bp with peak size of 200 pb. Chromatin solution was stored at −80°C until its use.

<sup>1</sup>http://david.abcc.ncifcrf.gov/tools.jsp (Accessed: December 10, 2017).

<sup>2</sup>http://www.systemsbiology.org (Accessed: December 10, 2017).

<sup>3</sup>http://www.geneontology.org (Accessed: December 10, 2017).

### Chromatin Immunoprecipitation (ChIP)

Immunoprecipitation was performed using ChIP Magnetic One Step (Abcam, Cambridge, United Kingdom) following the manufacturer's instructions. Chromatin solution was pooled and 10 μg per reaction was used. Samples were incubated with four different antibody Chip grade were used: anti-H3K27me3 (ab6002, Abcam), anti-H3K4me (ab8580,Abcam), anti-H3K4me1 (ab8895,Abcam), H3K27ac (ab4729, Abcam) (0.8 μg/well) during 120 min in a rolling shaker. After several washings, reversal of cross-links was carried out. Thisstep consisted in two incubations with Proteinase K (0.025 µg/µL), (at 60°C for 15 min and then 95°C per 5 min) followed by the release and elution of DNA using magnetic stand. Purified DNA was quantified using Nanodrop ND-1000 (Thermo Fisher Scientific, MA, USA), stored at −20°C and then sent to Omega Bioservices for Illumina High Throughput sequencing.

#### ChIP-Sequencing and Bioinformatics Analysis

Chromatin immunoprecipitation was carried out by the customer and the eluted DNA fragments were subjected to library prep using KAPA Hyper Prep Kit (KAPABIOSYSTEMS). Briefly, end-repair, A-tailing, adapter ligation, and PCR reactions were performed following the manufacturer's recommended protocols. Thermal cycling conditions used were 98°C for 45 sfollowed by optimal cycles of 98°C for 15 s, 60°C for 30 s, and 72°C for 30 s, then 72°C for 1 min and hold at 4°C. PCR cycles were determined according to the sample starting amount. The postligation cleanup and post-amplification cleanup were performed with Mag-Bind RxnPure Plus magnetic beads (Omega Bio-tek, Norcross, GA, USA) to remove short fragments such as adapter dimers. The libraries were qualified and quantified using Agilent 2200 Tapestation instrument (Agilent Technologies, Santa Clara, CA, USA). The samples were then pooled in equimolar concentrations and sequenced in 2 × 150 bp paired-end read setting on Illumina HiSeq machine (Illumina, San Diego, CA, USA). Forthe bioinformatics analysis, first the fastq reads were trimmed into 50 bp using customized perl script. Paired-end reads were pooled together and treated as single-end reads to enhance mapability. Follow, the trimmed reads were then aligned to Salmon genome ICSASG\_v2 using Bowtie2 (2.3.0). Mapping statistics can be found in Alignment.xlsx. The ChIP-seq peak call was performed by Homer(v4.8).The histone peak calling function was used with default parameters. Corresponding control samples were used as background for the peak calling. Each identified peak must has its peak height >1 reads per million or 10 reads per ten million and >4-fold higher than the input. Finally, ChIP-seq peaks were mapped to their nearest genes by Homer. For each histone modification marker, a pool of identified peaks was constructed by merging peaks from all samples (Bedtools 2.26.0). The average ChIP-seq tag density was calculated using Homer and plotted in R (3.3.2). Heatmap for ChIP-seq tag density was generated using seqMINER (1.2).

## Flow Cytometry Analysis

Salmon pronephros cells were isolated and processed for flow cytometry analysis (pronephros of each condition WTR control, control mock, fever, and no fever). Cell isolation was performed by disaggregating tissue through Falcon® cell strainers (100 µm) with DMEM media plus 10% FBS and 1% glutamax. The cells were pelleted by centrifugation at 1,200 rpm for 10 min. Maisey et al. (48) methodology was carried out with some modifications. In brief, pronephros cell was re-suspended and blocked with PBS plus 2% FBS immunofluorescence (IF) media.Cellswere incubate for 1 h at 4°C with primary antibody anti-CD4-1 (protein G- or affinity-purified Abs) (49). Then, cells were suspended again with IF media containing secondary antibody BV421 Goat Anti-Rabbit IgG Clone Polyclonal (565014, BD Bioscience, NJ, USA) and Alexa Fluor® 488 Goat Anti-Mouse IgG H&L (ab150113, Abcam, Cambridge, United Kingdom) and incubated 1 h at 4°C. Finally, cells were washed and suspended once again in 200 mL IF media prior to analysis. For auto fluorescence measurement, cells were suspended with IF containing no Ab, whereas for isotype controls, cells were treated only with the corresponding conjugated secondary Ab. A BD LSRFortessa™ X-20v flow cytometry was used to analyze sample and at least 10,000 events were recorded for each sample. Recorded events analyzed using the FlowJo software.

#### Absolute RT-PCR Validation in Pronephros

RT-qPCR was performed using the StepOnePlus™ Real-Time PCR System (Applied Biosystems, Life Technologies, NC, USA), and each assay was run in triplicate using the Maxima SYBR Green qPCR Master Mix (2×) (Bio-Rad, NC, USA). cDNA used in qPCR assays was first diluted with nuclease free water (Qiagen, Hilden, Germany). Each qPCR mixture contained the SYBR Green Master Mix, 2 µL cDNA, 500 nmol/L each primer, and RNase free water to a final volume of 10 µL. Amplification was performed in triplicate on 96-well plates with the following thermal cycling conditions: initial activation for 10 min at 95°C, followed by 40 cycles of 15 s at 95°C, 30 s at 60°C, and 30 s at 72°C. A dilution series made from known concentrations of plasmid containing the PCR inserts was used to calculate absolute copy numbers for each of the genes examined. Previously published primers were used (Table S1 in Supplementary Material).

#### ELISA Measurement of Plasma Cytokine *TNF-α*, *IL-1β*, and *IL-6*

Blood plasma was obtained from individual salmon (*n* = 9 individuals treatment; no fever, fever, control RTR, and control WTR) after 12 h post virus infection and stored at −80°C until use. To determine the detection of *IL-6*, *TNF-α*, and *IL-1β* in plasma samples of *S. salar*, indirect ELISA was performed according Morales-Lange et al. (50). Briefly, each plasma sample was diluted in carbonate buffer (60 mM NaHCO3, pH 9.6), planted (in duplicated for each marker) at 35 ng/µL (100 µL) in a Maxisorp plate (Nunc, Thermo Fisher Scientific, Waltham, MA, USA) and incubated overnight at 4°C. After, each well was blocked with 1% bovine serum albumin (BSA) for 2 h at 37°C. Then, plates were incubated for 90 min at 37°C with the primary antibody anti-synthetic epitope (diluted in BSA) of TNF-α (diluted 1:500), *IL-6* (diluted 1:500), and *IL-1β* (diluted 1:500). Later, the second antibody-HRP (Thermo Fisher Scientific, Waltham, MA, USA) was incubated for 60 min at 37°C in 1:7,000 dilution. Finally, 100 µL per well of chromagen substrate 3,3′,5,5′-tetramethylbenzidine single solution (Invitrogen, CA, USA) was added and incubated for 30 min at room temperature. Reaction was stopped with 50 µL of 1N sulfuric acid and read at 450 nm on a VERSAmax microplate reader. Primary antibodies against cytokines were produced according Bethke et al. (51). Synthetic epitope peptides were used for immunization in CF-1 mouse for *IL-6* and *TNF-α* (52) and New Zealand rabbit for *IL-1β* (53). For validation, antibody efficiency was determined by the calibration curve of the antibody against the synthetic peptide used for the immunization through indirect ELISA (54) and antibody specificity was verified by Western blot as described before in Schmitt et al. (53).

#### RESULTS

#### Behavioral Fever and the Impact of Thermal Choice Upon the Transcriptome

The behavioral analysis modeled the number of fish that were found in chambers with higher temperatures (chambers 5 and 6) as a function of the two temperature groups (control WTR and control RTR), the challenge groups (fever and no fever), and their interaction (four groups in total). The Wald test highlight that in the fever group, the number of fish found in higher temperatures (chamber 5 or 6) was significantly higher in comparison to the control group (*p* < 10<sup>−</sup><sup>3</sup> ) (**Figures 2A,B**; Table S2 in Supplementary Material). The results suggest that the IPNv challenged group with access to a thermal gradient tank (fever group) preferably chose the warm chambers (*n*°= 5 and 6) and the unchallenged fish prefer chambers 3 and 4 (thermo-preferendum, Wald test *p* < 10<sup>−</sup><sup>3</sup> ). The kinetics of the behavioral fever show that in the fever group fish perform behavioral fever throughout the day, however, the mostly fish develop fever past 12 h of infection (Table S2 in Supplementary Material). The transcriptomic analysis was evaluated using RNA-seq of salmon challenged with the birnavirus IPNv under two different thermal conditions: (1) wide thermal range (fever group) and (2) restricted thermal range (no fever). Each group was composed of nine individuals (three individuals by each replicate), resulting in one thermal profile for each group (see **Figures 3A–C**). Viral load was quantified using RT-qPCR assays for the genome segment WP2 of the IPNv (39, 40, 55). Viral loads were higher in individuals that have not developed behavioral fever, in contrast to the fever group (**Figure 1B**). Differences in the response to IPNv infection in the fever and no fever groups were significant both in transcript number (total number of differentially expressed transcript over the control, one-way ANOVA *p* < 0.01) and intensity (fold change FC > 2) (**Figures 3B,C**). After filtering for contigs that mapped exclusively to the *S. salar* genome, we considered the normalized expression levels of 4,248 transcripts that were differentially expressed in all group. We identified a specific transcriptomic signature for each group in the RNA-seq data (with 1,726 and 1,369 in fever and no fever group, respectively). Principal component analysis highlighted a significant correlation of gene expression data with each fish group (*R*<sup>2</sup> = 0.10, *p* < 0.03; **Figure 3A**). Following on we further analyzed the effect of behavioral fever on the variation in mRNA abundance for each specific transcript identified of interest.

We used a linear mixed-effects model in which residual variation in gene expression levels, after taking into account differences across the fish groups, was treated as the response variable. The temperature condition (fever and no fever) was incorporated as a fixed effect, and the significance of the thermal conditions on the gene expression was assessed based on the strength of this effect. We identified an association between inter-group gene expression variation and fever group for 2,879 genes (67.7% of the 4,248 genes we considered; false discovery rate = 10%; **Figure 3D**). Within the set of 2,879 fever associated genes, 1,726 genes (49.6% of all genes we considered) were more specifically expressed in the fever group (**Figures 3B,D**), and 1,369 genes (32.2%) were specifically expressed in individuals housed at a restricted thermal range (no fever group, **Figures 3B,C**). The results highlight a juxtaposed transcriptomic

Figure 2 | Behavioral fever in infectious pancreatic necrosis virus (IPNv)-challenged *Salmo salar*. (A) Frequency of chamber occupation in individual fish challenged with IPNv (fever vs no fever). The blue box represents fish challenged with IPNv under constant temperature. The orange box represents fish challenged with IPNv under thermal gradient tank (*n* = 10 mean ± SD, \**p*, 0.05; \*\**p*, 0.01; \*\*\**p*, 0.001). (B) Frequency of chamber occupation in individual fish challenged with IPNv in a thermal gradient tank (orange boxes) and control individuals (no infected group in thermal gradient tank, white boxes control-WTR; *n* = 10 mean ± SD, \**p*, 0.05; \*\**p*, 0.01; \*\*\**p*, 0.001). All individuals used were conditioned at constant temperature under 9-month (15 ± 0.9°C). (C) Survival percentage, the results are presented as the percentage of surviving fish after 10 days post viral challenge (mean ± SD, two-way ANOVA *F*2,39 = 14.89; unlike letters denote significantly differences).

expression and is correlated with the temperature effect on the gene expression mainly on the behavioral fever group (*p* = 0.04, *R*<sup>2</sup> = 0.10, *n* = 24). (B) Heatmaps of log2-transformed gene expression levels for rank-associated genes of thermal wide range group (fever). (C) Heatmaps of log2-transformed gene expression levels for rank-associated genes of thermal restricted range group (no fever). Values are shown after controlling for differences in means among social groups; 0 roughly corresponds to mean expression levels, red upregulated genes, blue downregulated genes. (D) Venn diagram on the distribution of mRNAs across the unlike thermal group. The overlapping expression of identified RNA transcripts from thermal group are depicted in different colors: fever (red), no fever (blue).

response between wide and restricted thermal range during development and after the viral infection highlighting the effects of thermal choice and behavioral fever upon underlying molecular processes.

To explore the biological functions associated with these genes, we incorporated a protein–protein interaction network derived from TAP/MAS ClusterMaker Cystoscope-plugin (47). We selected the modules with over representation of node–node interactions (immune enrichment analysis) that were found in each thermoregulatory range and conducted GO functional analysis with GO-DAVID (42) and the ClueGO Cytoscape plugin (46). GO-DAVID and ClueGO identified a significant enrichment of functional GO categories related to the immune response (**Figures 4A,B**and **5A,B**).We identified a strong signal for enrichment of specific immune response categories in the fever group associated gene set as a whole. Here, we observed enrichment for interferon signaling (*p* = 0.01, *q* = 0.05), B-cell activation (*p* = 0.01, *q* = 0.05), and lymphocyte activation

annotated genes). The specific genes were sorted into categories based on enrichment H3K4me1, H3K4me3, 3K27me3, and H3K27ac.

involved in immune response (*p* < 0.02; proportion of false discoveries at this *p* value threshold, *q* = 0.30) among genes highly expressed in the behavioral fever group (**Figures 4A,B**) suggesting a strong induction of innate and adaptive immune responses (**Figure 4A**). Examples of such genes (**Figures 6A,B**) include *IFNR-γ* and *IFN-γ* (a receptor for the proinflammatory cytokine INF, which is associated with neutrophil migration into injured tissue), *IL4/13*, *IL-2*, *IL1-2* (which is associated with the transcriptional response to T-cell stimulation) and proinflammatory cytokines, which include *mPGES-1* (proinflammatory signaling molecule that is positively regulated during the fever process) and *IL1*, *IL-6*, *TNF-α*, *COX-2* (**Figure 6A**) and also by cytokine release (**Figure 1C**). The ClueGo analysis reflected a coherent immunological function and fever group, specifically we found that the largest module (38% of the genes; mean *r* = 0.58) was enriched for B cell activation in the immune response, followed by specific immune-related gene subsets including lymphocyte activation involved in immune function (*p* = 0.002, *q* = 0.08), response to *IFN-γ* (*p* = 7.50 × 10<sup>−</sup><sup>6</sup> , *q* = 9.0 × 10<sup>−</sup><sup>5</sup> ) and myeloid cell activation involved in immune response (*p* = 6.31 × 10<sup>−</sup><sup>4</sup> , *q* = 3.80 × 10<sup>−</sup><sup>3</sup> ). The present results highlight that behavioral fever promote the activation of specific immunological/adaptive processes further supports the observation that behavioral fever induces a coordinated functional response at the level of the transcriptome that is juxtaposed of the registered at individuals kept at constant temperature

(no fever group, **Figures 5A,B**). Once we established a potential mechanism whereby behavioral fever potentiates the accumulation of response-specific mRNAs in the responding tissue, we sought to test the contribution of fever to survival. After 10 days post virus challenge, the behavioral fever group (fever individuals, Δ<sup>T</sup> 6.4°C) showed no clinical signs of infection and low mortality rates in significant contrast to those held under constant conditions (**Figure 2C**).

## Behavioral Fever Influences Lymphocyte Activation

We used fluorescence-activated flow cytometry (FACS) analysis to estimate the proportion of the two-main cell populations (CD4<sup>+</sup> T cells) in pronephros samples from 60 fish. The no fever group individuals had a reduced proportion of CD4<sup>+</sup> T cells (*p* < 0.05, *n* = 10; **Figures 7A,B**), whereas the group expressing behavioral fever had a significantly increased, twice, the proportion of CD4<sup>+</sup> T cells in contrast to the no fever group during the viral infection (*p* < 0.05, *n* = 10). Interestingly, the CD4<sup>+</sup> population was also significantly higher in the fever group in comparison to the non-infected group (both control group WTR and RTR, respectively, *p* < 0.05, *n* = 10). The results emphasizes the failure of the individuals that have not developed fever to increase the CD4<sup>+</sup> T cell population in contrast to the observed in the fever group individuals (*p* < 0.05, *n* = 10; **Figures 7A,B**).

immune process in pronephros of no fever individuals (GO DAVID and ClueGo Cytoscpae Plugin). (A) Upregulated genes and (B) downregulated genes. (C,D) Heat map showing chromatin immunoprecipitation-seq read density (log2 transformed) for the indicated of 1,369 RTR−/− genes referenced in salmon ENSEMBL database (down and up) at equivalent genomic regions in 5-kb windows upstream and downstream of the TSS binding sites (TSS; assumed to be at the 5 kp end of annotated genes). The specific genes were sorted into categories based on enrichment H3K4me1, H3K4me3, 3K27me3, and H3K27ac.

## Regulatory Mechanisms Underlying the Thermoregulation and Fever Behavior

Previous results and those obtained in this study reinforce our understanding of how both thermal choice and the expression of behavioral fever during infection can strongly impact the regulation of gene expression. Furthermore, in this study, we have also shown the impact of temperature and the feverresponse upon the CD4<sup>+</sup> T lymphocyte populations in salmon. Next, we investigated whether differences in gene expression observed could be influenced by genome histone modification during viral infection. To explore this hypothesis, we examined changes in histone modifications(ChIP-seq; HiSeq-Illumina 10×)from pronephros DNA. We then investigated the relationships between histone methylation and the recruitment of markers for active transcription (H3K4me3 and H3K27ac), for repressed transcription (H3K27me3 and H3K4me1), and gene expression.

To assess whether histone methylation levels might contribute to the observed gene expression associations, we checked histone methylationdatainthe1,726and1,369genespreviouslyselectedas specific gene setsforfever and no feverindividuals. We found that active transcription driven by methylation data from H3K27ac, H3K4me3 clearly distinguished between both thermal individuals (**Figures 7C,D**; Figures S1A,B in Supplementary Material). The histone methylation data distinguished upregulated and downregulated genes from specific groups of genes from each

experimental group (**Figures 4C,D** and **5C,D**). Specifically, we observed epigenetic features of transcriptomic gene setstriggered exclusively by individuals expressing behavioral fever that were associated with a specific methylation pattern (**Figure 7C**; Figure S1AinSupplementaryMaterial).Infever group, alsowasobserved an increase of the H3K4me3 a histone hallmark frequently linked with transcription initiation providing a "window of opportunity" for the enhancer activation. By contrast, individuals reared under a restricted thermoregulatory (no fever group, **Figure 7D**; Figure S1B in Supplementary Material) H3K4me3 modification was weakly detectable. In his thermal group, the enrichment of H3K27me3 also was slightly increased (**Figure 7D**; Figure S1B in Supplementary Material).

In individuals expressing behavioral fever, H3K4me3 and H3K27ac histone modificationswere highly noticeable in several GO where active transcription of genes related to the immune adaptive response occurs (**Figures 4C,D** and **7C**). Unlike H3K27ac, the repressive transcriptional marker H3K27me3 did not appear to have a specific pattern for these GO processes in fever group of individuals (**Figures 4C,D** and **7C**). In contrast to the pattern observed for the behavioral fever individuals, not-fever group shows an increase in repressed transcription activity linked to H3K27me3 modification across several biological processes related to the previous enrichment analysis for immune response (**Figure 5D**). In the individuals that have not

developed fever (no fever group) H3K4me1 modification was also evident. Thus, the observed gene expression patterns appear to be tightly related to differential histone methylation data that is significantly different under the different thermal regimes investigated. Epigenetic regulation therefore may explain, at least in part, how thermal choice and the expression of behavioral fever under infection can potentiate the immune response in ectotherms.

## DISCUSSION

The fact that the "*fever*" response to infection and injury has been maintained throughout at least 600 million years of evolution strongly suggests a positive benefit to immunity and overall survival (24, 25, 27). Our results highlight the influence of thermal choice and the importance of behavioral fever to mount a successful immune response during infectious episodes. Furthermore, our data opens a significant new avenue supporting the increasingly recognized link between increasing temperature and epigenetic regulation in fish (56, 57). Critically in this and our previous studies, we report that the impact of temperature is not observed across the entire transcriptome but focused into specific profiles. These transcriptome profiles correlate to functional differences measured in lymphocyte proliferation and cytokine release suggesting different immune response modules are influenced in a specific manner. The significant variation observed in immune response in ectotherms might partly be explained by unsuitable environmental settings resulting in non-optimal gene-temperature interactions (18). In our current design, individuals move freely throughout the thermal gradient (wide thermal range) after a pathogen challenge. As other abiotic variables (e.g., salinity or pH) may also play a role modulating response to the virus (58, 59), additional experiments should be needed to robustly test this hypothesis.

In fish, few studies of epigenetic modification have been associated with environmental effects during the development (60–63), and to our knowledge, no studies have associated the impact of epigenetic modification on the immune response. Here, we show that observed differences in the transcriptome (5,602 mRNAs) and associated GO processes in response to viral infection between the fever and no fever groups could be explained at leastin part by epigenetic modification.The histone modifications observed between the different groups were relatively low when compared with homeothermic species orin the context of human diseases such as cancer (64–66). However, the observed changes in epigenetic marks driven by behavioral fever are higher in comparison to other studies addressing the influence of environmental factors upon other fish species (67–69). By testing epigenetic modification in our "thermal choice" model and linking this to behavioral fever-associated gene expression, we have identified a regulatory role for the epigenome in fish. This regulatory step appears to be directly coupled to environmental temperature and

Figure 7 | Effects of tissue composition mediated regulation on fever behavior and temperature gene expression levels. (A) Behavioral fever individuals exhibit higher proportions of CD4+ T cells in pronephros tissue (*p* = 0.031, *n* = 60; *y*-axis shows the percentage of T-cell proportions after infectious pancreatic necrosis virus challenge). Panel (B) shows the lymphocyte population and the inset show an example data for histogram of fever and no fever individuals; *x* axis shows staining for CD4+ (helper) T cells (% of cells positive for CD4-BV421 antibody). (C,D) Box plots for chromatin immunoprecipitation-seq RPM values of fever and no fever group of genes. Box plots of adjusted (adj.) input normalized H3K4me3 reads per kilobase per million mapped reads (RPKM), (C) values at TSS (fever genes, *n* = 1,726) (D) and TSS-containing domains (no fever genes, *n* = 1,369). Box and whiskers correspond to the highest and lowest points within the 1.5× interquartile range.

impacts upon immune performance including cellular activity and the humoral response. In ectotherms, it is hypothesized that coupling of the immune response to fever promotes the survival (5, 70). However, currently there are not studies that explain the mechanism through which the immune response is increasing during the fever response. Our results show for the first time, evidence that the adaptive value of the fever response lie at the level of epigenetic–environment interaction affecting systemic and specific immune response.

Remarkably, we identified a strong and significant increase in CD4<sup>+</sup> T cells in the pronephros of infected salmon expressing behavioral fever. Although the presence of a definitive Th1 response in fish is currently under debate our data supports the development of a Th1 response. In parallel to increased CD4<sup>+</sup> T cells, *IL-6*, *Il-1β*, *IL-12*, *TNF-α*, *IFNR-γ,* and *IFN-γ* mRNAs are all significantly induced in IPNv infected salmon supporting the development of a specific Th1 response in salmon. Importantly these mRNAs are all significantly higher in IPNv-infected fish expressing behavioral fever. In mammals, the sensory vanilloid receptor 1 (TRPV1) has been shown that in addition to a thermal sensing role can also regulate the signaling and activation ofCD4<sup>+</sup> T cells and cytokine release (71, 72). Current studies in salmon suggest a key role for TRP receptors in regulating the behavioral fever response in fish (73). *IL-4* is critical to the development of Th2 responses in mammals and in the fish *IL-4 like mRNAs* have been described and named *IL4/13* (74, 75). Interestingly, *IL4-13* mRNA was upregulated in the pronephros as a result of IPNv infection, however, there were no differences between fish in the WTR and RTR environments. A similar expression pattern was observed for *IL-2* which supports differentiation and maintenance of both Th1 and Th2 states (76). Thus, it appears that behavioral fever at least within the time scales measured strongly favors the development of a Th1 type response to viral infection. Further studies will address CD8<sup>+</sup> T cells due to their significance during a viral infection (71, 72). Humoral response components, *COX2* and *PTGS2* mRNAs, representing the prostaglandin system were upregulated, however, no "fever" effect was evident. In previous studies we have shown significant increases in PGE<sup>2</sup> plasma concentrations (5) suggesting that behavioral fever is more likely to affect enzymatic activity by increased temperature rather than at the mRNA level.

The present results show that fish expressing behavioral fever, challenged with virus increase H3K4me3 on genes encoding factors involved in innate immunity and lymphocyte differentiation. The pathogenic stimulation of human monocytes also increases H3K4me3 in a group of innate immune genes (31). By contrast, before viral infection, increased amounts of H3K27me3 and reduced basal expression of innate target genes are sustained for individuals that have not or cannot develop behavioral fever. A similar pattern of histone influence on gene expression related to immune response has been observed in murine macrophages challenged with endotoxin that were previously exposed to β glucans (77, 78) which be a mechanistic link regulate LPS tolerance. It is known that epigenetic alterations lead to the priming of genes encoding host defense molecules that respond specifically to pathogens (79, 80). This systemic protective mechanism, through epigenetic modifications, likely has significant adaptive value as remodeled cells would respond more robustly to pathogens. Thus providing the individual with a selective immunological advantage (29, 81). Our findings document the role of epigenetic regulation on the immunological pathways activated in salmon expressing behavioral fever during viremia. These dynamic epigenetic elements and the observed thermal-coupling effect during fever clearly influences specific defense modules such as the Th1 response during viral infection and may play a critical role in the development of trained immune responses in fish. Our analysis highlights the key functions of the epigenetic modifications in regulating the immune response during the fever process and is likely where the behavioral modification leading to warmth-seeking initiates. We detected high viral load in pronephros 24 h post-challenge in both thermal treatment (fever and no fever individuals), additionally, we were able to recognize significant differences in the accumulative mortality. Early detection of IPNv after immersion challenge has been documented on rainbow trout (81). The observed mortality in the "no fever" group are in concordance with previous studies of fish challenged at a constant temperature, which suggests that high tropism of IPNv observed in pronephros is followed by an increase of the mortality after 7 days of virus infection (82–84). Although the beneficial or deleterious effects of fever are still debated (7–10), the present results extend our previous observations that behavioral fever in ectotherms has a positive adaptive value due to increased survival (5, 17).

In conclusion, our results highlight the close interaction between behavioral fever, immune performance particularly the development of a Th1-like response and significant epigenetic regulation at the molecular level in a model of viral infection in the *Atlantic salmon*, a species of considerable economic interest for aquaculture. Our study highlights the importance of creating environments where experimental animals are able to express a "normal" behavior therefore uncovering important underlying regulatory circuits. We propose that behavioral fever acts as an integrative signal that promotes specific epigenetic modifications that drives a protein production in responding lymphocyte populations. This, in turn, leads to increased efficacy of immune defense traits and provides a positive adaptive value to the host. Our data demonstrate the extensive plasticity of the immune response in fish and therefore by extension provides insight into other ectothermic organisms. This approach delineates thermal thresholds, in this case for *Atlantic salmon*, where the potential

#### REFERENCES


costs of adverse thermal environments are traded for improved immunological responses and therefore increased survival. The observed temperature epigenetic synergy leading to increased survival has implications toward understanding the molecular basis of disease resistance in ectotherms, and provides opportunity to understand the evolutionary consequences of behavioral fever in ectotherms.

#### Ethics statement

All animal experiments conformed to the British Home Office Regulations (Animal Scientific Procedures Act 1986), the animal research was authorized by the Universidad de Concepcion Institutional Animal Care and Use Committee.

#### AUTHOR CONTRIBUTIONS

The study was conceived by SB with important input from SM. AA and NS performed the experiments. SB analyzed the data. SB, SM, and NS performed the model simulations and provided extensive additional input. SB: funding acquisition. SB and SM drafted the manuscript with substantial contributions from all other authors.

#### FUNDING

We acknowledge financial support from Chilean Research Council CONICYT FONDAP (1510027) and FONDECYT (1150585) project, and the Strategic Investment Fund (FIE-2015-V014) from the Ministry of Economy, Development and Tourism of Chile (Cod No 201708070149).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material forthis article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.01241/ full#supplementary-material.

Figure S1 | Average H3K4me1, H3K4me3, 3K27me3, H3K27ac RPKM values of 5′ and 3′ sides, and flanking regions of for (A) fever genes TSS-containing domains (*n* = 1,726), and (B) no fever genes TSS-containing domains (*n* = 1,369) broader than 10 kb: infectious pancreatic necrosis virus challenged (red), control (blue).

Table S1 | mRNA primer sequences used for absolute RT-qPCR analysis.

Table S2 | Kinetic of the behavioral fever. The table shows the frequency of chamber occupation in individual fish challenged with infectious pancreatic necrosis virus during 24 h post-infection, the video cameras provided continuous monitoring of each tank chamber. The frequency was recorded for 10 s every 15 min throughout 24 h.


response. *Proc Biol Sci* (2013) 280(1766):20131381. doi:10.1098/rspb. 2013.1381


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

The handling Editor declared a past co-authorship with the authors SB and SM.

*Copyright © 2018 Boltana, Aguilar, Sanhueza, Donoso, Mercado, Imarai and Mackenzie. 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.*

#### *Ryan D. Heimroth, Elisa Casadei and Irene Salinas\**

*Center for Evolutionary and Theoretical Immunology (CETI), Department of Biology, University of New Mexico, Albuquerque, NM, United States*

Animal mucosal barriers constantly interact with the external environment, and this interaction is markedly different in aquatic and terrestrial environments. Transitioning from water to land was a critical step in vertebrate evolution, but the immune adaptations that mucosal barriers such as the skin underwent during that process are essentially unknown. Vertebrate animals such as the African lungfish have a bimodal life, switching from freshwater to terrestrial habitats when environmental conditions are not favorable. African lungfish skin mucus secretions contribute to the terrestrialization process by forming a cocoon that surrounds and protects the lungfish body. The goal of this study was to characterize the skin mucus immunoproteome of African lungfish, *Protopterus dolloi*, before and during the induction phase of terrestrialization as well as the immunoproteome of the gill mucus during the terrestrialization induction phase. Using LC-MS/ MS, we identified a total of 974 proteins using a lungfish Illumina RNA-seq database, 1,256 proteins from previously published lungfish sequence read archive and 880 proteins using a lungfish 454 RNA-seq database for annotation in the three samples analyzed (free-swimming skin mucus, terrestrialized skin mucus, and terrestrialized gill mucus). The terrestrialized skin mucus proteome was enriched in proteins with known antimicrobial functions such as histones and S100 proteins compared to free-swimming skin mucus. In support, gene ontology analyses showed that the terrestrialized skin mucus proteome has predicted functions in processes such as viral process, defense response to Gram-negative bacterium, and tumor necrosis factor-mediated signaling. Importantly, we observed a switch in immunoglobulin heavy chain secretion upon terrestrialization, with IgW1 long form (IgW1L) and IgM1 present in free-swimming skin mucus and IgW1L, IgM1, and IgM2 in terrestrialized skin mucus. Combined, these results indicate an increase in investment in the production of unique immune molecules in *P. dolloi* skin mucus in response to terrestrialization that likely better protects lungfish against external aggressors found in land.

Keywords: mucosal immunity, mucus, proteomics, terrestrialization, African lungfish, skin, skin-associated lymphoid tissue

## INTRODUCTION

Transitioning to life on land was a fundamental step in the success and diversification of the vertebrate lineage (1). This transition imposed multiple novel challenges to vertebrates, especially at their mucosal surfaces. As a consequence, drastic physiological, histological, and molecular adaptations took place at vertebrate mucosal barriers for successful colonization of land.

#### *Edited by:*

*Brian Dixon, University of Waterloo, Canada*

#### *Reviewed by:*

*Katherine Buckley, Carnegie Mellon University, United States Yong-An Zhang, Institute of Hydrobiology (CAS), China*

> *\*Correspondence: Irene Salinas isalinas@unm.edu*

#### *Specialty section:*

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

*Received: 18 February 2018 Accepted: 18 May 2018 Published: 04 June 2018*

#### *Citation:*

*Heimroth RD, Casadei E and Salinas I (2018) Effects of Experimental Terrestrialization on the Skin Mucus Proteome of African Lungfish (Protopterus dolloi). Front. Immunol. 9:1259. doi: 10.3389/fimmu.2018.01259*

**150**

The skin is the outermost organ of the vertebrate body. In aquatic vertebrates, such as teleost fish, the skin is a mucosal epithelium composed of layers of living cells coated by a mucus layer that is in direct contact with the environment. Teleost skin mucus contains many immune factors including innate and adaptive immune molecules that protect the host from invading microorganisms (2). In terrestrial vertebrates, the skin does not contain mucus-secreting cells and is cornified (3). This configuration is thought to help terrestrial vertebrates cope with desiccation stress and UV radiation.

African lungfish (*Protopterus* sp.) are a subclass of Sarcopterygian fish that are obligate air-breathers, and are the extant relative to all tetrapods (4, 5). Lungfish are evolutionarily unique organisms that have the ability to undergo conditional aestivation. Aestivation is a state of metabolic torpor, which is an adaptation for survival in areas that are subject to extreme environmental conditions (6–10). During droughts, as water evaporates from the rivers and food becomes scarce, lungfish detect environmental cues and turn them into internal signals that induce behavioral, physiological, and biochemical changes in preparation of aestivation.

There are four extant species of African lungfish: *P. aethiopicus, P. annectens, P. amphibious,* and *P. dolloi*, which implement different aestivation strategies to survive prolonged dry periods. The best-known method of aestivation is that of *P. annectens*. As water recedes, *P. annectens* will start to burrow into the mud while simultaneously secreting large quantities of mucus through its gills. This is the induction phase of aestivation. Once it has burrowed deep enough into the mud, it curls back on itself, leaving its head facing the opening of the burrow. As the water dissipates completely, the mucus/mud mixture coating around the lungfish body hardens forming a cocoon that protects the animal for months. Once encased, the lungfish ceases feeding and locomotive activities, has to prevent cell death, and sustain a slow rate of waste production until conditions become favorable (11), this is the maintenance phase of aestivation, where the lungfish can lay dormant for years. Upon the introduction of water, the lungfish instantly awakens from dormancy, leaves the mucus cocoon, and slowly swims toward the surface of the water for air (5). This is the final phase of aestivation, known as arousal, which is completed a week after water is again available and marks the return to a normal metabolic rate. *P. dolloi*, however, uses a different aestivation strategy as they do not burrow into the mud and do not appear to completely reduce their metabolic rate (12). During the induction phase, *P. dolloi* coil up on the surface of the mud while secreting mucus which, over time, turns into a dried mucus cocoon (13). Thus, *P. dolloi's* mode of aestivation has been coined as "terrestrialization," which is different from the complete aestivation and the full metabolic torpor observed in *P. annectens* (14, 15). Both *P. dolloi* and *P. annectens* can be terrestrialized in the laboratory setting making them ideal models to study the effects of air exposure on the vertebrate mucosal immune system (7, 9, 14).

Throughout tetrapod evolution different strategies were coopted to maintain barrier integrity and defend against external aggressors. Antimicrobial compounds are among the most important immune molecules present in the skin of all vertebrates, both aquatic and terrestrial (16, 17). Antimicrobial proteins, lysozyme, histones, S100 proteins, and immunoglobulins (Igs) have been previously identified to be important players in the vertebrate skin immune system (18–22). Given the importance of immune molecules in skin homeostasis, tissue repair, and responses to environmental insults, we hypothesize that these molecules play a critical role during the process of terrestrialization in lungfish skin and that African lungfish will increase the amount of resources allocated to skin immunity early on in the process of terrestrialization.

The goal of this study is to characterize the African lungfish skin mucus proteome in the freshwater state, as well as the compositional change in the proteome due to terrestrialization. Our results provide a first glance to the skin proteome composition of sarcopterygian fish and its role in adaptation to terrestrial life.

#### MATERIALS AND METHODS

#### Animals

Juvenile *P. dolloi* (slender lungfish) were obtained from Tropical Aquatics (FL, USA) and maintained in 10-gallon aquarium tanks with dechlorinated water and a sand/gravel substrate, at a temperature of 27–29°C. Fish were acclimated to laboratory conditions for a minimum of 3 weeks before being used in experiments. During this acclimation period, they were fed frozen earthworms every third day. Feeding was terminated 48 h before the start of the experiment. All animal studies were reviewed and approved by the Office of Animal Care Compliance at the University of New Mexico (protocol number 11-100744-MCC).

#### Experimental Aestivation and Mucus Collection

After 3 weeks of acclimation to laboratory conditions, water in the tanks were lowered to 20 cm at 27–29°C and allowed to naturally evaporate (7). As the water level lowered, the fish entered the induction phase of aestivation and began to hyperventilate and profusely secrete mucus from their gills. This mucus combined with the substrate from the bottom of the tank encased the fish in a cocoon, which hardened after 10 days in the induction phase. In order to avoid severe dehydration due to the dry climate of New Mexico, 1–2 mL of water were sprayed on the surface of the cocoon every third day (12). Mucus was collected from the skin of one *P. dolloi* individual before the beginning of the terrestrialization experiment (named free-swimming skin mucus) and from the same individual 10 days after the start of the induction phase. At this time point, the liquid mucus actively secreted from the gills (named terrestrialized gill mucus) was aspirated with a sterile plastic Pasteur pipette. Additionally, the hardened mucus cocoon surrounding the lungfish body (named terrestrialized skin mucus) was collected by peeling it off with sterile forceps. All samples were immediately frozen at −80°C until needed for protein solubilization.

#### Protein Solubilization

Mucus samples were solubilized in a protein extraction buffer (pH 7.6) made of 60 mM DTT, 2% SDS, and 40 mM Tris–HCl as previously described (23). The proteins were extracted by adding 4× the sample volume of cold (−20°C) acetone and incubating overnight at −20°C. The proteins were pelleted out and dissolved in a solution of 6 M Urea and 200 mM of ammonium bicarbonate. Protein concentrations were measured using the Pierce 660 nm protein assay (Thermo Fisher Scientific, San Jose, CA, USA).

#### One-Dimensional Electrophoresis

Proteins extracted from the mucus samples were analyzed by one-dimensional electrophoresis. 12 µL of each sample was loaded into a Mini-PROTEAN TGX precast Gel at a 1:1 ratio with 2× Laemmli Sample Buffer (Bio-Rad, Hercules, CA, USA) for a final volume of 24 µL in each well. A total of 4.56 µg of protein for terrestrialized skin mucus, 3.94 µg for free-swimming skin mucus, and 1.05 µg for terrestrialized gill mucus were loaded. The proteins were separated out by the sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) in 1× Tris buffer, run at 120 V for 55 min, then incubated for 1 h with mild agitation in Coomassie Brilliant Blue R-250 (Bio-Rad, Hercules, CA, USA). Each lane of the gel was cut to perform proteomics analyses. Due to the presence of high-intensity band in the free-swimming skin mucus sample and a high-intensity band in the terrestrialized skin mucus sample, these two bands were first excised from their corresponding lanes and solubilized separately from the rest of the lane.

#### LC-MS/MS

LC-MS/MS analysis of in-gel trypsin-digested excised protein bands or whole protein mixture-separated gel lanes (24) was carried out using an LTQ Orbitrap Velos mass spectrometer (Thermo Fisher Scientific, San Jose, CA, USA) equipped with an Advion nanomate ESI source (Advion, Ithaca, NY, USA), following ZipTip (Millipore, Billerica, MA, USA) C18 sample clean-up according to the manufacturer's instructions. Peptides were eluted from a C18 pre-column (100 µm id × 2 cm, Thermo Fisher Scientific) onto an analytical column (75 µm id × 10 cm, C18, Thermo Fisher Scientific) using (1) a 2% hold of solvent B (acetonitrile, 0.1% formic acid) for 5 min, followed by a 2–10% gradient of solvent B over 5 min, 10–35% gradient of solvent B over 35 min, 35–50% gradient of solvent B over 20 min, 50–95% gradient of solvent B over 5 min, 95% hold of solvent B for 5 min, and finally a return to 2% in 0.1 min and another 9.9 min hold of 2% solvent B (single protein gel band analysis) or (2) a 2% hold of solvent B (acetonitrile, 0.1% formic acid) for 5 min, followed by a 2–7% gradient of solvent B over 5 min, 7–15% gradient of solvent B over 50 min, 15–35% gradient of solvent B over 60 min, 35–40% gradient of solvent B over 28 min, 40–85% gradient of solvent B over 5 min, 85% hold of solvent B for 10 min and finally a return to 2% in 1 min and another 16 min hold of 2% solvent B (whole gel lane analysis). All flow rates were 400 nL/min. Solvent A consisted of water and 0.1% formic acid. Data-dependent scanning was performed by the Xcalibur v 2.1.0 software (25) using a survey mass scan at 60,000 resolution in the Orbitrap analyzer scanning mass/charge (*m*/*z*) 400–1,600, followed by collisioninduced dissociation tandem mass spectrometry (MS/MS) of the 14 most intense ions in the linear ion trap analyzer. Precursor ions were selected by the monoisotopic precursor selection setting with selection or rejection of ions held to a ±10 ppm window. Dynamic exclusion was set to place any selected m/z on an exclusion list for 45 s after a single MS/MS.

#### RNA-Sequencing and Assembly

Three different transcriptomes were used to analyze the proteomic data generated in the present study. First, the pre-pyloric spleen from an experimentally infected *P. dolloi* individual was used to generate 454 pyrosequencing (Roche) transcriptome as explained elsewhere (26). Data were assembled using Roche's GS *De Novo* Assembler. A second database was a *P. annectens* Illumina database kindly shared by Dr. Chris Amemiya and was sequenced and assembled as described in Ref. (4). The third transcriptome consisted of sequence read archive (SRA) databases from *P. annectens*, which were downloaded from the National Center for Biotechnology Information (NCBI). Sratoolkit.2.9.0 fastq-dump was used to convert the SRAs into forward and reverse paired read fastq files (27). The paired-end reads were then assembled into *de novo* transcriptomes using Trinity assembler at default parameters (28). These new transcriptomes were concatenated together into one large transcriptome, and any redundant sequences were removed using cd-hit-est at a 99% confidence level (29). The resulting transcriptome was translated into a protein database using Transdecoder-5.0.0 (28).

#### Protein Identification

The protein and peptide identification results were visualized with Scaffold v 3.6.1 (Proteome Software Inc., Portland, OR, USA), a program that relies on various search engine results (i.e., Sequest, X!Tandem, and MASCOT) and which uses Bayesian statistics to reliably identify more spectra (30). Proteins were accepted that passed a minimum of two peptides identified at 0.1% peptide FDR and 90–99.9% protein confidence by the Protein Profit algorithm, within Scaffold. Tandem mass spectra were searched against the three *Protopterus* sp. transcriptomes obtained as described above. A translated protein database to which common contaminant proteins (e.g., human keratins obtained at ftp://ftp.thegpm.org/ fasta/cRAP) were appended to each database. All MS/MS spectra were searched using Thermo Proteome Discoverer 1.3 (Thermo Fisher Scientific, San Jose, CA, USA) considering fully Lys C peptides with up to two missed cleavage sites. Variable modifications considered during the search included methionine oxidation (15.995 Da) and cysteine carbamidomethylation (57.021 Da). Proteins were identified at 99% confidence interval with XCorr score cutoffs as determined by a reversed database search (31).

#### Proteomic Data Analysis

Transcriptome IDs for each sample were taken from Proteome Discoverer. The proteins retrieved from the excised lanes were first merged with their corresponding whole band resulting in one protein list for free-swimming skin mucus, one for terrestrialized gill mucus, and one for terrestrialized skin mucus. Tblastn searches were performed against each of the three transcriptomic data sets and resulting nucleotide sequences were then used as queries for blastx searches in NCBI (32). Human orthologous genes were assigned using AmiGO2 (33, 34), and gene ontology (GO) analysis was performed using DAVID bioinformatics database (35, 36). GO analysis of the proteome was used to identify the percentage of immune-related proteins in each sample. Venn diagrams were created in R identifying unique and common protein composition between samples (37). In order to identify proteins with known antimicrobial function, we analyzed our protein results using the list of mammalian skin antimicrobial compounds provided in the review paper by Schauber and Gallo (38).

#### Histology

For light microscopy, skin samples from free-swimming and terrestrialized fish were fixed in 4% paraformaldehyde overnight, transferred to 70% ethanol, and embedded in paraffin. Samples were sectioned at a thickness of 5 µm, dewaxed in xylene, and stained using hematoxylin and eosin for general morphological analysis. Images were acquired and analyzed with a Nikon Eclipse Ti-S inverted microscope and NIS-Elements Advanced Research Software (Version 4.20.02).

#### Quantitative Real-Time PCR

Skin tissue from free-swimming *P. dolloi* and terrestrialized *P. dolloi* (*N* = 3) was collected using sterile dissecting tools and placed in 1 mL of Trizol (Ambion, Life Technologies, Carlsbad, CA, USA). Total RNA was extracted from each sample, and 1 µg of RNA was synthesized into cDNA as described in Ref. (26). The resulting cDNA was stored at −20°C. The expression levels of IgM1, IgM2,

Table 1 | Primers used in this study.


IgW1, H2A, S100-A11, and neutrophil elastase (ELANE) were measured by quantitative real-time PCR (RT-qPCR) using the specific primers shown in **Table 1**. Phosphoglycerate kinase 1 PGK1F was used as the house-keeping gene. The RT-qPCR and statistical analysis were performed as described in Ref. (26). Data were expressed as mean ± standard error, and qPCR results were analyzed by unpaired *t*-test (*p* < 0.05).

### RESULTS

#### Histological Changes in the Lungfish Skin in Response to Terrestrialization

Histological examination of freshwater and terrestrialized *P. dolloi* skin revealed that mucus-secreting goblet cells become exhausted from the process of terrestrialization. The epidermis becomes more compact with flattened keratinocytes on the surface creating a new barrier. We observed eosinophilic granulocytes in the dermis of both samples, but they were more abundant in the terrestrialized skin (**Figure 1**). These results are in agreement with previously reported changes in African lungfish skin in response to aestivation (7).

#### SDS-PAGE Analysis of Lungfish Mucus Proteins

The overall protein composition of the three mucus samples was visualized by SDS-PAGE. The protein band patterns of each sample were unique. There was a distinct band of high intensity at ~52 kDa in the free-swimming skin mucus, while there was an intense band ~15 kDa in the terrestrialized skin mucus sample not found in the other two samples. The gill mucus sample did not contain any predominant band (**Figure 2**). The two bands with high intensity (arrows) were excised and analyzed separately from the rest of the lane.

#### LC/MS-MS Proteomic Analysis, GO, and KEGG Pathway Analyses

Free-swimming, terrestrialized gill mucus, and terrestrialized skin mucus proteins from LC/MS-MS were analyzed against three RNAseq transcriptomes, two sequenced on an Illumina platform and one sequenced using a 454 pyrosequencing platform. These translated transcriptomes resulted in a total of 53,184 protein sequences in the translated Illumina database, 269,746 protein sequences in

Figure 1 | (A) Hematoxylin and eosin staining of free-swimming and (B) terrestrialized *Protopterus dolloi* skin paraffin sections. Skin sections show significant infiltration of granulocytes (red arrows) in the dermis (Der); epidermis (Epi), and goblet cells (GO). The black arrows indicate flattened keratinocytes after terrestrialization. (C) Image of a *P. dolloi* 4 days after the initiation of the induction phase actively secreting mucus from its gills.

the SRA database, and 41,351 protein sequences in the translated 454 database. The analysis returned a total of 974 proteins from the Illumina, 1,256 proteins from the SRA, and 880 proteins from the 454 database. Unique proteins were then identified using human orthologs and multiple copies of proteins were consolidated into single occurrences. As a result, 494 unique proteins were found in Illumina database, 636 unique proteins were in the SRA database, and 434 from in the 454 database. Comparing the outputted proteins from each database revealed that 298 proteins were shared when using all three databases, 213 proteins were unique proteins to the SRA database, 68 were unique to the Illumina database, and 50 were unique to the 454 database (**Figure 3A**). When analyzing the protein composition of each of the three mucus samples using the Illumina translated transcriptome we found 50 shared proteins among all three samples, 144 unique proteins in the freeswimming skin mucus, 113 unique proteins in the terrestrialized skin mucus, and 10 unique proteins in the terrestrialized gill mucus (**Figure 3B**). Analysis of the protein composition using the SRAtranslated transcriptome resulted in 44 common proteins among all three samples, 190 unique proteins in the free-swimming skin mucus, 171 unique proteins in the terrestrialized skin mucus, and 11 unique proteins in the terrestrialized gill mucus (**Figure 3C**). A similar trend was observed when using the 454 translated transcriptome with 33 shared proteins among all three mucus samples, 130 proteins unique to the free-swimming skin mucus, 101 proteins unique to the terrestrialized skin mucus, and only 14 proteins unique to the terrestrialized gill mucus (**Figure 3D**). These results suggest that the composition of the gill mucus secretion produced during the induction phase of terrestrialization resembles both the free-swimming skin mucus and the terrestrialized skin mucus proteome. Specifically, ~72.9% of the terrestrialized gill mucus proteome was also present in the free-swimming skin mucus proteome and ~71.7% of the terrestrialized gill mucus proteome were found in the terrestrialized skin mucus. As a result, only ~16.4% of all proteins present in the terrestrialized gill mucus were unique to this sample.

Gene ontology analyses using the three data sets revealed that biological processes (BPs) were enriched in free-swimming and terrestrialized skin mucus samples, but no unique GO for the terrestrialized gill mucus (Table S4 in Supplementary Material; **Figures 4A–F**). Based on the Illumina analysis, the top five most significant BPs enriched in free-swimming skin mucus were "small GTPase mediated signal transduction," "carbohydrate metabolic process," "UDP-N-acetylglucosamine biosynthetic

Figure 4 | Scatter plots of enriched gene ontology biological process (BP) terms for (A) free-swimming skin mucus and (B) terrestrialized skin mucus using the Illumina database; for (C) free-swimming skin mucus and (D) terrestrialized skin mucus using the sequence read archive (SRA) database; and for (E) free-swimming skin mucus and (F) terrestrialized skin mucus using the 454 database. The fold enrichment indicates the ratio of the expressed gene number to the total gene number in a pathway. Only the top 20 BPs according to *p*-value are shown. The size and color of the points represent the gene number and the log10 *p*-value of each pathway, respectively.

process," "cell-cell adhesion," and "galactose metabolic process." In terrestrialized skin mucus, in turn, we observed an enrichment in BPs such as "SRP-dependent cotranslational protein targeting to membrane," "viral transcription," "nuclear-transcribed mRNA catabolic process, nonsense-mediated decay," "translational initiation," and "translation" (**Figures 4A,B**). Interestingly, all data sets also contained significant BPs related to immune function that were different in the free-swimming and terrestrialized skin mucus. For instance, free-swimming skin mucus had unique proteins involved in "positive regulation of phagocytosis" and "antigen processing and presentation," whereas the terrestrialized skin mucus proteome included proteins with predicted functions in "platelet degranulation," "antigen processing and presentation of exogenous peptide antigen *via* MHC class I, TAP-dependent," "viral process," "tumor necrosis factor-mediated signaling pathway," "defense response to Gram-negative bacterium," and "cellular response to hydrogen peroxide" (Table S4 in Supplementary Material). Overall, these data suggest that the immunological processes that govern the skin immune system in lungfish differ in free-swimming and terrestrialized phases. Additionally, non-immune BPs that were enriched in the control skin mucus included "epithelial cell differentiation" and "membrane organization," while in terrestrialized skin mucus they included "translational initiation" and "platelet aggregation." These results may indicate that cellular organization and maintenance functions are enriched in the skin of free-swimming animals, while cell survival is enriched during terrestrialization.

KEGG pathway analyses showed that, overall, in free-swimming lungfish skin mucus, enriched pathways are mostly related to metabolic pathways such as "Histidine metabolism," "Amino sugar and nucleotide sugar metabolism," "Galactose metabolism," and "beta-Alanine metabolism." The proteome of the terrestrialized lungfish skin mucus was predicted to be enriched in "Ribosome" and "Biosynthesis of antibiotics" pathways (Figure S1 in Supplementary Material). KEGG pathway analyses using the Illumina and SRA data sets (Figures S1A–D in Supplementary Material) were more similar to each other compared to the 454-derived KEGG pathway analysis (Figures S1E,F in Supplementary Material). These results support previous studies that demonstrated that terrestrialization has profound effects on lungfish skin metabolism with an overall decrease in metabolic activity.

#### Changes in Lungfish Skin Immunoproteome in Response to Terrestrialization

In order to examine the allocation of immune resources in lungfish skin in response to terrestrialization, we performed manual counts within our protein lists using previously reported proteins with immune function. We observed an increase in the percentage of immune-related proteins present in the terrestrialized gill and skin mucus compared to the free-swimming skin mucus when using lists generated by the Illumina and SRA data sets but not the 454 data set (**Figure 5A**). It is worth noting that the SRA-based analysis revealed a higher percentage of immune-related proteins in the gill mucus (~25%) compared to the Illumina and 454 data sets (20% in both). Similarly, when we counted the number of immune proteins that were present in each sample we found a trend toward greater number of immune proteins in the terrestrialized skin mucus compared to free-swimming skin mucus when using the Illumina and SRA data sets but not the 454 data set (**Figure 5B**). Among the immune-related proteins, we observed a higher number of proteins with known antimicrobial activity in the terrestrialized skin mucus compared to the free-swimming skin mucus in all three data sets (**Figure 5C**). The greater number of proteins with antimicrobial activity in the terrestrialized skin mucus sample was due to a greater abundance of histones and S100 proteins (Tables S1–S3 in Supplementary Material). In support, we observed increased levels of expression of H2A and S100A11 in terrestrialized compared to free-swimming skin by RT-qPCR (**Figures 6A,B**). Moreover, gene expression analysis of neutrophil elastase (ELANE) showed a significant increase (150 fold) in the expression of this gene in terrestrialized compared to free-swimming lungfish skin (**Figure 6C**). This result supports our histological observations as well as the proteomic results (Tables S1 and S2 in Supplementary Material). Combined, our results suggest that lungfish increase immune resource allocation and undergo inflammation in the skin early on during the aestivation process.

Apart from differences in antimicrobial compound abundance, we also observed changes in the Ig proteins present in each sample. Where no Igs were detected in the gill mucus sample, we detected IgW1 long form (IgW1L) and IgM1 in the freeswimming skin mucus sample and IgW1L, IgM1, and IgM2 in the terrestrialized skin mucus sample using the Illumina and 454 but not the SRA data sets (Tables S1–S3 in Supplementary Material). In support, RT-qPCR analysis showed no significant changes in IgW1 or IgM1 expression as a result of terrestrialization but a significant increase in IgM2 expression (~3-fold) was observed in terrestrialized compared to free-swimming skin (**Figures 6D–F**). Both sigma and lambda-like light chains were found in both

mucus samples (Tables S1 and S2 in Supplementary Material). These results indicate a switch in Ig expression and secretion in the lungfish skin during the induction phase of aestivation.

#### DISCUSSION

The skin of all animals provides a first line of defense against pathogen invasion. Apart from being a physical barrier, the skin has its own unique suite of immune cells and molecules that constitute the skin-associated lymphoid tissue (39–40). As the major interface between the environment and the host, the skin is subject to several external stressors, and these stressors shifted dramatically during the vertebrate transition from water to land. Hence, we took advantage of the ability to terrestrialize lungfish in the laboratory setting as a model to study the water-to-land transition. We hypothesized that terrestrialization results in changes in the skin proteome composition and specifically, in an increased investment in production of immune molecules that will help prevent pathogen invasion and land stressors.

Previous studies have shown changes in the skin proteome composition of teleost fish (aquatic vertebrates) in response to infection (41, 42), stress (43–45), wounding (46), or dietary administration of immunostimulants (44, 45, 47). However, most of these studies adopted a two-dimensional (2-D) gel electrophoresis approach where selected spots were then analyzed by peptide fragment fingerprinting and LC-MS/MS. 2-D gel electrophoresis presents a number of drawbacks such as low reproducibility, the need for large sample sizes, and the difficulty to separate proteins with low abundance as well as very hydrophobic proteins (48). Thus, our results constitute a unique and unbiased report of all proteins present in the skin proteome of lungfish. We used three different RNA-seq databases from two different lungfish species (*P. dolloi*

and *P. annectens*) generated in three different platforms (named Illumina, SRA, and 454) in our analyses. As expected, we obtained different results depending on the database and tissue origin. We found greater number of proteins in every sample when the SRA database was used, an expected result given the coverage of proteincoding sequences in each data base (636 unique sequences for the lungfish SRA database but only 494 for the Illumina database and 435 for the 454 database according to Swissprot).

As previously reported, we observed dramatic remodeling of the lungfish skin histological organization upon terrestrialization. This tissue remodeling involves multiple processes such as flattening of the epithelial cells, decreases in the overall epidermal thickness, and loss of goblet cells. In support, our proteomic study showed enriched BPs and KEGG pathways present in the skin mucus before and during terrestrialization highlighting the ability of this vertebrate to respond to environmental stimuli and reshape the cellular and molecular composition of the skin. One potential caveat to our study is that increased amounts of cell debris may be present in the skin mucus samples of terrestrialized animals than in free-swimming lungfish, affecting the overall proteomic composition of the two samples. In any case, tissue remodeling and changes in the external microbial environment likely occur concomitantly in our model and teasing apart the immunoproteome changes that respond to one or the other stimulus is a challenging question.

Transition from water to land also imposes drastic changes in the microbial composition of the external environment. Thus, we observed that lungfish secrete unique suites of innate and adaptive immune molecules into the skin mucus as a response to air exposure. With respect to innate immune molecules, we identified greater number of proteins with known antimicrobial functions in the skin, particularly histones (HIST1H1D and HIST1H3C) and S100 proteins (S100A, S100P, and S100A6). In this study, we confirmed changes at the protein level with gene expression data for six selected genes. Overall, gene expression data supported the findings of the proteomics approach, but it is worth noting that we did not include biological replicates in the proteomics study. Previous work has shown a lack of correlation between proteomics and mRNA transcript levels (49) and therefore future studies should expand our current data sets to multiple biological replicates. GO analyses of the unique proteins present in the terrestrialized skin mucus revealed enrichment in BPs such as defense response to Gram-negative bacterium, antigen processing, and presentation of exogenous peptide antigen *via* MHC class I, TAP-dependent, tumor necrosis factormediated signaling pathway, and cellular response to hydrogen peroxide. Thus, these data suggest a requirement for the lungfish skin to increase antimicrobial defenses during the process of terrestrialization. Future studies should address the function of antimicrobial compounds triggered by air exposure.

*Protopterus* sp. express four different immunoglobulin heavy (IgH) chain classes: IgW, IgM, IgN, and IgQ. Additionally, there is extensive intraclass IgH diversification in lungfish, with IgM including three IgM genes (IgM1, IgM2, and IgM3), IgW including two genes (IgW1 and IgW2) as well as short and long forms, and IgN including also three genes (IgN1, IgN2, and IgN3) (50). So far, the immunological functions of different IgM subclasses have not been investigated. This study identified the presence of Igs in the skin mucus proteome of both freshwater and terrestrialized lungfish. Specifically, the freshwater lungfish skin proteome contained IgW1L as well as IgM1 secretory form. Secreted IgW(D) antibodies have not been previously characterized in sarcopterygian fish, but it is known that teleost secrete IgD into their mucosal secretions (51). Our findings support that secretion of IgW into the skin mucus occurs in sarcopterygian fish and that IgW1L may have specialized mucosal immune functions compared to other IgW forms in lungfish. In the terrestrialized skin mucus proteome, apart from IgW1L, we observed both IgM1 and IgM2 expression, suggesting the IgM1 expression occurs constitutively in the skin of lungfish but IgM2 expression is switched on in response to external stressors. We were not able to detect any Ig in the terrestrialized gill mucus sample, but this finding may be a result of the lower protein amounts in this sample compared to the skin mucus samples. Further studies should address the specific immunological function of IgM2 at lungfish mucosal surfaces.

#### CONCLUSION

This study provides the first characterization of the skin mucus proteome of a sarcopterygian fish, the African lungfish. We report important shifts in both innate and adaptive immune molecules

#### REFERENCES


in the skin mucus of lungfish in response to terrestrialization. Our results suggest that the transition from water to land in vertebrates imposed a need for increased investment in immune function in the cutaneous mucosal secretions.

## DATA AVAILABILITY

Proteomic data sets were submitted to ProteomeXchange (http:// www.proteomexchange.org/) *via* the PRIDE database, accession: PXD008981 and PXD008982.SRA used in this study can be found at NCBI (accession numbers SRR2027914, SRR2027978, SRR2027979, SRR2028000, SRR2028017, SRR2028020, SRR2028021, SRR2027980, SRR6291329, and SRR6291330). 454 pyrosequencing reads were deposited at NCBI (accession number SRP141470). Output tables from Scaffold containing all peptide table reports are given in Table S5 in Supplementary Material.

## ETHICS STATEMENT

All animal studies were reviewed and approved by the Office of Animal Care Compliance at the University of New Mexico (protocol number 11-100744-MCC).

## AUTHOR CONTRIBUTIONS

IS conceived the experiments. RH was responsible for the lungfish aestivation and performed histology, protein extraction, and transcriptomic data analysis. EC performed the RT-qPCR analysis. RH and IS wrote the manuscript and created the figures. All authors reviewed and revised the manuscript and approved the final manuscript prior to being submitted.

#### ACKNOWLEDGMENTS

The authors would like the thank Dr. Chris Amemiya for sharing his lungfish transcriptome and George Tsaprailis and Cynthia L. David for help with proteomics analysis.

## FUNDING

This work was supported by The National Science Foundation (IOS #1456940) and The National Institute of Health (COBRE grant P20GM103452).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.01259/ full#supplementary-material.


*annectens*, during the maintenance and arousal phases of aestivation. *Nitric Oxide* (2015) 44:71–80. doi:10.1016/j.niox.2014.11.017


calreticulin-like protein. *PLoS One* (2017) 12(1):e0169075. doi:10.1371/journal. pone.0169075


that predates the emergence of tetrapods. *Nat Commun* (2016) 7:10728. doi:10.1038/ncomms10728

**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 Heimroth, Casadei and Salinas. 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.*

# Novel insights on the Regulation of B Cell Functionality by Members of the Tumor Necrosis Factor Superfamily in Jawed Fish

*Carolina Tafalla and Aitor G. Granja\**

*Animal Health Research Center (CISA-INIA), Madrid, Spain*

#### *Edited by:*

*Brian Dixon, University of Waterloo, Canada*

#### *Reviewed by:*

*Irene Salinas, University of New Mexico, United States Antonio Figueras, Consejo Superior de Investigaciones Científicas (CSIC), Spain Steve Bird, University of Waikato, New Zealand*

> *\*Correspondence: Aitor G. Granja aitor.gonzalez@inia.es*

#### *Specialty section:*

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

*Received: 05 March 2018 Accepted: 22 May 2018 Published: 07 June 2018*

#### *Citation:*

*Tafalla C and Granja AG (2018) Novel Insights on the Regulation of B Cell Functionality by Members of the Tumor Necrosis Factor Superfamily in Jawed Fish. Front. Immunol. 9:1285. doi: 10.3389/fimmu.2018.01285*

Most ligands and receptors from the tumor necrosis factor (TNF) superfamily play very important roles in the immune system. In particular, many of these molecules are essential in the regulation of B cell biology and B cell-mediated immune responses. Hence, in mammals, it is known that many TNF family members play a key role on B cell development, maturation, homeostasis, activation, and differentiation, also influencing the ability of B cells to present antigens or act as regulators of immune responses. Evolutionarily, jawed fish (including cartilaginous and bony fish) constitute the first animal group in which an adaptive immune response based on B cells and immunoglobulins is present. However, until recently, not much was known about the expression of TNF ligands and receptors in these species. The sequences of many members of the TNF superfamily have been recently identified in different species of jawed fish, thus allowing posterior analysis on the role that these ligands and receptors have on B cell functionality. In this review, we summarize the current knowledge on the impact that the TNF family members have in different aspects of B cell functionality in fish, also providing an in depth comparison with functional aspects of TNF members in mammals, that will permit a further understanding of how B cell functionality is regulated in these distant animal groups.

#### Keywords: TNFSF, TNFRSF, fish, B cells, evolution, immunity

**Abbreviations:** Ab, antibody; AD, autoimmune disease; Ag, antigen; APRIL, a proliferation-inducing ligand; BAFF, B cell-activating factor of the TNF family; BAFFR, BAFF receptor; BCMA, B cell maturation antigen; BCR, B cell receptor; Blimp-1, B lymphocyte-induced maturation protein-1; BM, bone marrow; CRD, cysteine-rich domain; CSR, class-switch recombination; DcR3, decoy receptor 3; DR3, death receptor 3; DC, dendritic cell; DD, death domain; EDA, ectodysplasin; EDAR, EDA receptor; FasL, Fas ligand; GC, germinal center; GITR, glucocorticoid-induced TNFR; GITRL, GITR ligand; HVEM, herpesvirus entry mediator; LPS, lipopolysaccharide; LIGHT, homology with lymphotoxin, inducible expression, competing for GpD of herpes virus, that binds to the HVEM, and is expressed on activated T lymphocytes; LN, lymph node; LT, lymphotoxin; MHC, major histocompatibility complex; MZ, marginal zone; NGFR, nerve growth factor receptor; NHL, non-Hodgkin's lymphoma; NK, natural killer; PC, plasma cell; RA, rheumatoid arthritis; RANK, receptor activator of NF-κB; RANKL, RANK ligand; RELT, receptor expressed in lymphoid tissues; SHM, somatic hypermutation; SLE, systemic lupus erythematosus; SS, Sjögren's syndrome; TACI, transmembrane activator and calcium-modulating cyclophilin ligand interactor; TCR, T cell receptor; TD, thymus dependent; THD, TNF homology domain; TI, thymus independent; TLR, toll-like receptor; TNF receptor superfamily, tumour necrosis factor receptor superfamily; TNF ligand superfamily, tumour necrosis factor (ligand) superfamily; TRAIL, TNF-related apoptosis-inducing ligand; TWEAK, TNF-like weak inducer of apoptosis and proliferation.

## INTRODUCTION

In mammals, the tumor necrosis factor (TNF) ligand superfamily (TNFSF) signal through members of the tumour necrosis factor receptor superfamily (TNFRSF) to activate signaling pathways which play biological roles in development, organogenesis, cell death, and survival. Since the discovery of the first TNFSF members, TNF (1) and lymphotoxin α (Ltα) (2), more than 40 years ago, 17 additional TNFSF members and 29 cognate receptors have been identified in humans (3). With the completion of the large-scale sequencing of the human and mouse genomes, it is assumed that almost all TNFSF and TNFRSF members have now been identified in mammals (3, 4). Thus, ligands within the TNFSF which include 4-1BBL, a proliferation-inducing ligand (APRIL), B cell-activating factor of the TNF family (BAFF), CD27L, CD30L, CD40L, EDA1, EDA2, Fas ligand (FasL), GITRL, LIGHT, Ltα, Ltβ, OX40L, RANK ligand, TL1A, TNF, TNF-like weak inducer of apoptosis and proliferation, and TNFrelated apoptosis-inducing ligand (TRAIL) are key effector proteins in the orchestration of innate and adaptive immune responses [reviewed in Ref. (3) and summarized in **Table 1**].

TNFSF ligands are type II membrane-bound proteins with an intracellular N terminal and an extracellular C terminal domain. Among the 19 known ligands described in human and mouse, 11 encode for a proteolytic cleavage site that generates biologically active soluble forms (5). Within the C terminus, they contain the TNF homology domain (THD), which presents a weak degree of conservation (20–30%) between ligand members, and is typically formed by 10 β-strands (6). The THD folds into what has been called an antiparallel β-sandwich (3) exhibiting a compact jellyroll topology (4). The THD structure has the ability to assemble each TNF ligand into conical trimers which allow the ligands to bind respective receptors to initiate signaling (6).

Table 1 | Relation of TNF superfamily ligands (TNFSF) present in human indicating their standard name within the TNF superfamily and their alternative (most common) name.


*Uniprot ID links for each ligand are also provided.*

To date, 29 TNFRs have been identified in mammals, namely, 4-1BB, BAFFR, B cell maturation antigen (BCMA), CD27, CD30, CD40, decoy receptor 3 (DcR3), death receptor 3, DR6, EDAR, Fas, Fn14, glucocorticoid-induced TNFR, herpesvirus entry mediator (HVEM), LTbR, OPG, OX40, RANK, receptor expressed in lymphoid tissues (RELT), transmembrane activator and calcium-modulating cyclophilin ligand interactor (TACI), TNFR1, TNFR2, TRAILR1, TRAILR2, TRAILR3, TRAILR4, TROY, and XEDAR [reviewed in Ref. (7)]. The main feature of these TNFRs is a cysteine-rich domain (CRD) formed of three disulfide bonds surrounding a core motif of CXXCXXC creating an elongated molecule. There is an important variation in the number of CRDs among family members, from BAFFR or BCMA containing only one CRD to CD30 containing six CRDs. These receptors are type I membrane proteins, with the exceptions of BAFFR, BCMA, TACI, and XEDAR, which are type III membrane proteins, and OPG and DcR3, which are secreted (4, 8). The determination of the X-ray crystal structures of some ligands bound to the extracellular domain of their receptors has been primordial to characterize the binding mechanisms between them (9, 10). These analyses have revealed that those receptors with several CRDs adopt an elongated structure and bind at the interface between two ligand monomers, whereas single CRD receptors are more compact and contact a single ligand monomer in a trimeric ligand (5, 11, 12). Generally, one trimeric ligand engages three monomeric receptors, a key event for the activation of intracellular signaling pathways. By conducting a systematic flow cytometry-based assay, Bossen et al. elegantly demonstrated the mechanisms that regulate TNFSF–TNFRSF interactions in mouse and human (8). They found that TNFSF ligands bound from one to five different receptors, while most receptors bound from one to three ligands. Strikingly, they observed that, although containing the classical CRD structure, DR6, RELT, TROY, and nerve growth factor receptor (NGFR) did not bind to any of the TNFSF, thus suggesting that they either bind to other ligands or function in a ligand-independent manner. In this sense, it was later shown that although NGFR has the classic CRD structure, it binds a structurally different kind of ligands, the neurotrophins (13). This information regarding the different ligands that signal through each receptor in mammals is summarized in **Table 2**.

## EVOLUTION OF TNFSF AND TNFRSF MEMBERS

The appearance and further specialization of the adaptive immune response is a hallmark of the successful evolution of vertebrates. The adaptive immune system is based on the presence of recombination-activating gene (RAG)-recombined B cell receptors (BCR) and T cell receptors (TCR) on the surface of B cells and T cells, respectively, and the major histocompatibility complex (MHC). Molecular studies have shown that adaptive immunity arose early on vertebrate evolution, between the divergences of cyclostomes (lampreys) and cartilaginous fish, around 450 million years ago, by diversification and recombination of gene clusters on a span of time of 20 million years (36). This event is known as the "big bang" theory of the appearance of the adaptive immune


Table 2 | Functional relation of the tumour necrosis factor receptor superfamily (TNFRSF) and their cognate ligands (TNFSF).

*TNFRSF containing a death domain are underlined. Main immune functions known for each TNFRSF are also indicated, together with the references describing their functions in mammals.*

response, since it occurred in the very short period of time in which jawed fish appeared and is thought to be linked to genome duplication events (37). Thus, jawed fish were the most ancient animal group where all these elements were found, while jawless fish (Agnathans) seemed to have none of them (38). However, the posterior discovery of a lymphoid cell-based adaptive immune system in Agnathans, in which immune receptors recombined in a similar way to that of the BCR (39), pushed the origin of the adaptive immune response earlier in evolution. Despite this, the "big bang" theory for the origin and development of acquire immunity still prevails.

Interestingly, Collette et al. postulated that the divergence of the TNFSF and TNFRSF members parallels the emergence of the adaptive immune response (7). Since 11 out of the 19 human TNFSF members are clustered within the MHC and paralogous regions on chromosomes 1, 6, 9, and 19, the authors suggest that this disposition might be a consequence of the ancestral arrangement of a proto-TNFSF cluster, before en bloc duplication of the proto-MHC region in a vertebrate ancestor 500–800 million years ago (40). Remarkably, the different number of TNFSF and TNFRSF members found in the different species within vertebrates correlates with the number of rounds of genome duplication during evolution (4), thus supporting the idea that genome duplication created paralogous clusters. This hypothesis is further supported by phylogenetic analysis that indicates an ancient evolutionary origin of TNFSF ligand and receptor genes that precedes the appearance of vertebrates (7, 41).

As many invertebrate and vertebrate genomes are now available, the discovery of TNFSF and TNFRSF orthologs and paralogs has greatly increased. Consequently, recent phylogenetic studies on invertebrate TNFSF ligands and receptors have been key to better understand the appearance of TNF molecules in metazoans and to further support the hypothesis of their divergent evolution [reviewed in Ref. (4)]. The most primitive TNF superfamily member that has been functionally characterized is Eiger, a TNFSF homolog found on the fruit fly (*Drosophila melanogaster*) (42). Eiger binds to a cognate TNFRSF member, called Wengen, which contains an intracellular death domain, thus inducing cell death through the activation of signaling pathways similar to those activated by mammalian TNFRSFs (43, 44). A TNFSF member named MjTNF has also been characterized in another invertebrate, the marine arthropod kuruma shrimp (*Marsupenaeus japonicas*) (45). This protein contains a predicted transmembrane region and a THD, and shares 30.7% sequence identity with *Drosophila* Eiger. Two molluscan TNFSF members containing transmembrane regions and THDs were identified in the disk abalone, *Haliotis discus discus.* One was designated AbTNF-α (46) and the other AbFas ligand (47). Within the genome of the equinoderm purple sea urchin (*Strongylocentrotus purpuratus*), four different TNFSF genes have been found, identified as potential gene orthologs of TNFSF14 (LIGHT), TNFSF15 (TL1A), and two separate genes resembling EDA (48, 49). In parallel, several TNF superfamily receptors were also identified on these invertebrate organisms [summarized in Ref. (4)]. As most of the invertebrate TNFSF members are constitutively expressed and most TNFRSFs are phylogenetically related to EDAR, a major role in development and organogenesis with restricted immunoregulatory properties is foreseen in these organisms.

By contrast, in teleost fish, the first animal group comprising all the elements of the adaptive arm of the immune system, a diversification of the TNFSF and TNFRSF has occurred. To date, 13 different TNFSF and 13 TNFRSF homologs have been identified, together with new members of the TNFSF that are novel to this animal group (4, 41, 50–54) supporting the hypothesis of a diversification of the TNF superfamily with the appearance of the adaptive immune response. Interestingly, recent studies in sarcopterygian fish (African lungfish) revealed a different TNFSF and TNFRSF gene pattern to that seen in teleost (55). African lungfish is an extant representative of the closest ancestral lineage to all tetrapods that presents organized lymphoid structures which cannot be found in other fish species (56). Several TNFSF were reported in this study, but additional analyses are needed to further characterize their identity and functionality. In parallel, many TNFRSF members were identified, and although most of them were present in both teleost and lungfish, only the latter presented homologue sequences for the TNF receptors HVEM (TNFRSF14), 4-1BB (TNFRSF9), and OPG (TNFRSF11B) (55). These TNF superfamily members have been shown to regulate T cell homeostasis and activation, as well as development of lymphoid structures, such as lymph nodes (LNs) (15, 17, 25, 30, 31). These results suggest that an expansion of TNF molecules could have occurred in the African lungfish, conferring a phenotypical advantage that was positively selected, thus leading to the appearance of organized lymphoid structures, which might play a key role on T cell activation. Therefore, throughout the evolution of vertebrates, the expansion of TNFSF and TNFRSF has led to the appearance of new members involved in the regulation of novel immune functions (4, 57), which were coopted under selective pressure, being this crucial for the evolution of the adaptive immune system (7, 41).

#### ROLE OF TNFSF LIGANDS ON B CELL REGULATION

In mammals, many TNFSF ligands have been shown to play essential roles on the regulation of the functionality of B cells. These TNFSF ligands may influence all aspects of B cell biology from development, maturation, survival, proliferation, activation, and differentiation [reviewed in Ref. (58)] (**Figure 1**), thus playing a fundamental role on B cell-mediated immune responses.

Among TNFSF members, BAFF (TNFSF13B) and APRIL (TNFSF13) are probably the cytokines that seem to play a prevailing role on the regulation of B cell activity [reviewed in Ref. (59)]. These two cytokines exist as membrane-bound and soluble forms, being both forms biologically active (60). Both BAFF and APRIL bind to and signal through BCMA (TNFRSF17) and TACI (TNFRSF13B), whereas BAFF also binds to BAFFR (TNFRSF13C). These receptors are mainly expressed in B cells, and their specific activation leads to different outcomes of the B cell response (61). BAFF-mediated survival signals through BAFFR are necessary for immature B cells to become mature circulating B cells and for peripheral B cell survival (62). These signals regulate the size of the B cell compartment, especially that of conventional B2 cells, since the absence of BAFF does not affect the maturation or survival of innate-like B cells, such as marginal zone (MZ) or B1 cells (63). In fact, there is some evidence to suggest that the maintenance of the B1 B cell compartment is controlled by APRIL signaling through TACI (64, 65), which is highly expressed on the surface MZ B cells and B1 cells (66). In this context, BAFF and APRIL signaling through TACI have been shown to induce class-switch recombination (CSR) in response to thymus-independent (TI) antigens (Ags) (28).

Other TNFSF members can induce CSR on B cells in response to thymus-dependent (TD) Ags; such as, for example, CD40L (TNFSF5) and OX40L (TNFSF4) (21, 22). CD40L is expressed mainly by activated T cells during TD responses, thus mediating the co-stimulation of BCR-activated B cells, which express the receptor CD40 (TNFRSF5), usually within the germinal center (GC). As a consequence, co-stimulated B cells in the GC enhance their proliferation and undergo somatic hypermutation (SHM) to increase their affinity and CSR to switch from producing IgM to producing immunoglobulin isotypes with higher Ag affinity such as IgA, IgE, or IgG [reviewed in Ref. (21)]. In this context, part of these CD40L-induced proliferating B cells also differentiates to antibody (Ab)-secreting plasma cells (PCs), since CD40L and IL-21 synergistically induce the expression of B lymphocyteinduced maturation protein-1 (67), a transcription factor which is the master regulator of terminal differentiation to PCs (68). Moreover, BAFF and APRIL signaling through BAFFR and TACI can contribute to enhance PC differentiation triggered by CD40L (69). On terminally differentiated PCs, signaling through BCMA is highly expressed on PCs and is needed to promote their survival (70). Both BAFF and APRIL can signal through BCMA, although it shows much higher affinity for APRIL than BAFF (71). Concerning OX40L, this cytokine is expressed mainly on activated B cells while its receptor OX40 is expressed on activated CD4+ T cells. Their interaction triggers a bidirectional co-stimulation of both B and T cells during TD responses. Furthermore, cross-linking of OX40L on B cells by OX40 has been shown to greatly enhance B cell proliferation and Ig production (22).

One of the most studied TNFSF members is TNF-α, a wellknown pro-inflammatory cytokine able to promote cell death (72). However, TNF-α has also been shown to play an important role on B cell functionality (73). TNF-α binds to two different receptors, TNFR1, which is ubiquitously expressed on almost all cell types, and TNFR2, whose expression is limited to the central nervous system and the immune system, especially found on T cells (15). TNF-α expression is rapidly and strongly upregulated *in vitro* or *in vivo* in the presence of many types of Ags or inflammatory mediators (15). In addition, TNF-α is produced by T cells after TCR engagement (74) and by B cells after TI BCR cross-linking

and also after CD40 ligation by T cell-derived CD40L (75). In this context, TNF-α provides co-stimulatory signals which increase the proliferation and Ab production of B cells after Ag encounter, being very important for the polyclonal expansion needed within primary responses (15).

After BCR engagement, expression of CD70 (TNFSF7) is also induced on B cells. Ligation of CD70 with its ligand CD27 delivers signals to enhance proliferation, inhibit B cell differentiation to PCs, trigger SHM, and promote the generation of memory B cells (76). However, it has also been shown that ligation of CD70 in the presence of co-stimulatory T cell signals such as CD40L can promote B cell differentiation into Ab-producing PCs (77).

Recent studies have shown that BCR cross-linking increases the sensitivity of B cells to TRAIL (TNFSF10)-mediated cell death. It has been demonstrated that this effect can be reverted by ligation of CD40 on B cells, while B1 cells, which are involved in TI responses showed very high sensitivity to TRAIL-induced death. These data suggested that TRAIL is involved in B cell differentiation and survival at the GC reaction, and in Ab affinity maturation (78). Another member playing a similar role is Fas ligand (FasL) (TNFSF6), which induces apoptosis after ligation of its receptor (Fas) on the surface of the target cell (79). BCR activation induces the expression of Fas on the surface of B cells, making them more susceptible of FasL-mediated apoptosis. During the GC reaction, CD40 ligation protects B cells from Fas-induced apoptosis, thus contributing to the selection of B cells bearing a high-affinity BCR (80). LTβ has also been demonstrated to play an important role in the formation of GCs and also on Ab affinity maturation (81). Finally, CD153 (TNFSF8) also plays a role on B cells since the binding to its receptor (CD30) on T cells modulates B cell differentiation and CSR mediated by reverse signaling induced by CD30<sup>+</sup> activated T cells (82).

#### THE ADAPTIVE IMMUNE SYSTEM IN FISH

The adaptive immune system, characterized by an Ag-specific combinatorial immune response (36), first appeared in jawed fish. Thus, evolutionarily, cartilaginous fish (sharks, skates, and rays) are the first animal group in which the adaptive immune system, based on immunoglobulin superfamily members, namely, BCR, TCR and MHC, and RAG 1 and 2 genes are present (38).

Due to the anatomical differences between fish and mammals (i.e., humans), significant differences are found in the distribution and functionality of primary and secondary lymphoid organs, such as the absence of LN or bone marrow (BM) in fish (56, 83). The fish spleen functions as the major secondary lymphoid organ, as it happens in mammals, and since fish lack LN, the spleen has been shown as the most important tissue for Ag trapping (84).

Regarding hematopoiesis, fish do not have a conventional BM as it is described in the mammalian immune system. In cartilaginous fish, the Leydig organ and the epigonal organ are believed to be the equivalents of mammalian BM (85). Both are reticular structures that contain large numbers of immature leukocytes, including neutrophils, eosinophils, and other granulocytes, as well as lymphocyte aggregates with scattered PCs. Either one or both of these tissues have been demonstrated to be present in all cartilaginous species examined (83). The expression of RAG-1 and B-cell-specific transcription factors strongly supports a lymphopoietic role for these tissues (86). In the case of bony fish (teleost), the anterior part of the kidney (head kidney/anterior kidney) has no renal functions and has been shown to assume hematopoietic functions (87). B cell development at the anterior kidney has been proven by the expression of RAG-1/2 (88, 89), TdT (90), and the transcription factor Ikaros (91), and the posterior cellular analysis defining the B cell subsets residing within the kidney (92). As in mammalian BM (93), anterior kidney also stores Ig-secreting long-lived PCs (94).

The thymus is similar to that found in mammals, composed by a cortex and a medulla, and is responsible for the production of T cells (95). TCR cell surface expression has been shown in teleost (96) but remains to be described in cartilaginous fish. However, the expression of all the TCR genes identified in mammals (α, β, δ, and γ) has been reported in cartilaginous fish (97).

## FISH B CELLS

B cells are one of the most important elements of the adaptive immune response, since they are able to produce specific highaffinity immunoglobulins against pathogens (Abs), and also generate memory B cells which will protect the organism against future infections (98). In fish, there are important differences concerning the isotypes of immunoglobulins produced by B cells in comparison to mammals.

Cartilaginous fish B cells produce three types of immunoglobulins IgM, IgW, and IgNAR (immunoglobulin new antigen receptor) (38). In these species, IgM is orthologous to mammalian IgM, while IgW has been postulated as the orthologous of mammalian IgD (99), although its function is still unknown. IgNAR is a shark-specific heavy chain (H) homodimer which does not associate with light chains (L) (100). IgNAR is produced by a different B cell subset than that expressing IgM and although the specific function for IgNAR remains unclear, IgNAR responses have been shown to be TD and show high specificity for the Ag (101).

Teleost fish species produce three types of Igs, namely, IgM, IgD, and IgT/Z (102). The latter was first described in 2005 in both rainbow trout (*Oncorhynchus mykiss*) (designated as IgT for teleost) (103) and in zebrafish (*Danio rerio*) where it was designated as IgZ (104). Since then, it has been described in most teleost species (105, 106), while it seems absent in others such as channel catfish (*Ictalurus punctatus*) and medaka fish (*Oryzias latipes*) (107). While the most abundant B cell type in the main lymphoid organs in teleost is IgD<sup>+</sup>IgM<sup>+</sup>, as it has been described for mammalian naïve mature B cells (108), IgT<sup>+</sup> B cells constitute a different lineage in teleost, which do not co-express any other surface Ig (109). These cells are more abundant in mucosal surfaces than in the main lymphoid organs and consequently IgT-expressing B cells have been cataloged as B cells specialized in mucosal responses (109–111). Despite this, IgT responses have also been reported outside the mucosal compartments, thus the function of IgT in teleost immune responses is still largely unknown. In addition, IgD<sup>+</sup>IgM<sup>−</sup> (IgD single) cells have been identified in rainbow trout gills (112) and channel catfish blood (113) although their function is still unknown. Moreover, IgD<sup>−</sup>IgM<sup>+</sup> (IgM single) cells exhibiting an antibody-secreting cell (ASC) phenotype have been shown to inhabit in the peritoneal cavity of vaccinated rainbow trout (114). These observations suggest that the B cell compartment in fish is composed by different cell subsets which most probably differ in functionality and/or cytokine-mediated regulation.

Interestingly, teleost B cells have been shown to possess strong phagocytic and microbicidal activities (115). It was later shown that is ability is preserved in mammalian innate B1 cells from the peritoneal cavity of mice (116) strongly suggesting that fish B cells are less evolved than mammalian B2 cells and still retain features of the innate immune system, which could be consistent with a posterior appearance of specific B1 and B2 B cell lineages throughout evolution. Further evidence supporting this hypothesis has been collected from studies showing the capacity of fish B cells to respond to pro-inflammatory stimuli (117), the expression of innate B cell markers CD9 and CD63 (118) and toll-like receptors (TLRs) (119) or the synthesis of antimicrobial peptides (117).

#### TNFSF LIGANDS IN FISH: AT THE DAWN OF ACQUIRED IMMUNITY?

As mentioned earlier, Collette et al. proposed that the divergence of the TNFSF and TNFRSF families parallels the emergence of the adaptive immune system, after performing a detailed analysis of the phylogenetic relations between the TNFSF and TNFRSF members of invertebrate and vertebrate species (7). As the identification of TNFSF orthologs and paralogs in fish has been rapidly increasing in the recent years, as can be inferred from the work undertaken by Glenney and Wiens (4, 41), studying the function of these cytokines and their receptors in fish is becoming a fascinating research topic to better understand the evolution and regulation of the adaptive immune system. In this review, we illustrate the information available regarding the effect of TNFSF ligands in fish, focusing specifically on those with a presumed role on fish B cells.

## B CELL-ACTIVATING FACTOR OF THE TNF FAMILY

The homologue sequences to mammalian BAFF have been reported in many teleost fish species including rainbow trout (41), zebrafish (120), mefugu (*Takifugu obscurus*) (121), Japanese sea perch (*Lateolabrax japonicus*) (122), grass carp (*Ctenopharyngodon idella*) (123), yellow grouper (*Epinephelus awoara*) (124), miiuy croaker (*Miichthys miiuy*) (125), tongue sole (*Cynoglossus semilaevis*) (126), Nile tilapia (*Oreochromis niloticus*) (127), rock bream (*Oplegnathus fasciatus*) (128), and also in cartilaginous fish such as white-spotted catshark (*Chiloscyllium plagiosum*) (129), spiny dogfish (*Squalus acanthias*) (130), and small-spotted catshark (*Scyliorhinus canicula*) (131) (summarized in **Table 3**). Interestingly, some studies have revealed that many cartilaginous and bony fish species have more than one BAFF gene (51, 54), representing two distinct groups. In mammals, BAFF is considered the master regulator of B cell development and function. Besides having been identified in many cartilaginous and bony fish species, it has been found in all vertebrate groups including birds, amphibians and reptiles (132), suggesting that BAFF has been essential throughout evolution for the development of peripheral mature B cells.

Recombinant BAFF proteins have been generated in some of these fish species, such as, for example, the cartilaginous fish white-spotted catshark (129), and the teleost zebrafish (120), fugu (121), Japanese sea perch (122), yellow grouper (124), tongue sole (126), rock bream (128), or tilapia (127), to carry out functional studies. These studies have proven an increase on the number of leukocytes mediated by BAFF, although the authors did not clarify whether this was due to a promotion of cell survival or an increase on cell proliferation, and they did not demonstrate if the surviving/proliferating fish leukocytes were in fact B cells. In some of these studies, the authors tested the effect of recombinant BAFF on mouse splenic B cells costimulated with anti-IgM (120, 121, 124, 127, 129). In all these experiments, fish recombinant BAFF promoted the survival of mouse B cells, which indicates that the molecular mechanism underlying BAFF-mediated B cell survival is preserved throughout species, from cartilaginous fish to mammals. However, specific studies on the impact of BAFF on B cells from lower vertebrates are needed to reveal new aspects about the evolution of B cell homeostasis and activation. Recent studies from our group showed the role of BAFF on teleost B cells, using rainbow trout as a model (53, 159). In these studies, we determined that teleost BAFF recapitulated mammalian BAFF regulatory aspects on B cells. Teleost BAFF promoted the survival of B cells but did not induced significant proliferation, and also increased the levels of ASCs which consequently increased IgM secretion and upregulated the expression of surface MHC II, in a similar trend to that seen in mammals (132). The similarity between teleost and mammalian BAFF functions strengthens the hypothesis of BAFF as a key master regulator of B cell functionality throughout evolution. In mammals, BAFF is produced by macrophages, DCs, stimulated neutrophils, and at low levels by T cells, but is never produced by resting B cells (60, 160, 161). Strikingly, one of our findings was that not only myeloid cells produced BAFF but also that specific subsets of splenic and peritoneal B cells were able to produce BAFF in rainbow trout (53, 159). Interestingly, in mammals, B cells from B-cell chronic lymphocytic leukemia (162) or non-Hodgkin's lymphoma (163), as well as B cells from patients with autoimmune disorders, namely, rheumatoid arthritis (164), systemic lupus erythematosus (165), and primary Sjogren's syndrome (166) also express BAFF, which rescues them from apoptosis in an autocrine loop. Thus, our studies seemed to have revealed a primitive mechanism through which B cells would produce BAFF at the steady-state to regulate their homeostasis and function in teleost. It seems that this regulatory mechanism throughout evolution was later assumed by other immune myeloid cells, probably after the appearance of lymphoid follicles, although in mammals, it can reemerge in various B cell disorders such as autoimmune diseases (ADs) or B cell malignancies. In a parallel study, we were also able to report for the first time an upregulation of BAFF transcription in peritoneal IgM<sup>+</sup> B cells from fish immunized intra-peritoneally (i.p.) with viral hemorrhagic septicemia virus (VHSV) (114). In addition, it has been very recently shown that *in vivo* transgenic overexpression of BAFF induced an increased production of IgD, IgM, and IgZ in zebrafish (167). This implies that BAFF is not only involved on the homeostasis of peripheral B cell compartments but also plays an important role on the activation of different B cell subsets present in teleost.

## A PROLIFERATION-INDUCING LIGAND

Although APRIL is probably the most important B cell-regulating TNFSF together with BAFF, very little is known about fish APRIL. In fact, although BAFF homologs have been identified in many cartilaginous and bony fish (summarized in **Table 3**), only a few APRIL homologue sequences have been found in teleost, specifically in zebrafish, channel catfish, Atlantic salmon, rainbow trout (41), and grass carp (134) (**Table 3**). In addition, while BAFF is present in all vertebrates, APRIL is missing in cartilaginous fish, birds, and several bony fish (168). This has encouraged evolutionary immunologists to hypothesize that since BAFF and APRIL have structural and functional similarities and share receptors, the absence of APRIL on cartilaginous fish or the loss of APRIL in avian species could have been functionally compensated by BAFF. From the functional point of view, teleost APRIL has been shown to be mostly expressed in lymphoid tissues (spleen, head kidney) as well as in some mucosal tissues (skin, intestine) but at very low levels (41, 114, 134). After a bacterial or a viral challenge, the transcription of APRIL was quickly upregulated in immune tissues of grass carp (134), and recombinant APRIL promoted the survival of total splenocytes in zebrafish (169), indicating that this TNF ligand plays some role on the activation of the immune response. In addition, a study by our group showed that APRIL promoted the survival of peritoneal IgM<sup>+</sup> B cells in rainbow trout (114). Since APRIL signaling through BCMA is key for the survival of PCs (70), it is tempting to hypothesize that those teleost species in which APRIL is found to have obtained a phenotypical advantage, the long-term survival of peripheral subsets of Ab-secreting PCs, that would have been positively selected.

#### Table 3 | B cell-regulating TNFSF ligand members identified to date in fish.


TABLE 3 | Continued


*The species and the name for each species gene are underlined text refers to cartilaginous fish species. References are also annotated. N/A, non-available.*

#### BAFF- AND APRIL-LIKE MOLECULE (BALM)

Searching through the rainbow trout EST databases, Glenney and Wiens reported in 2007 the identification of an additional sequence with high similarity values with BAFF but containing a D–E loop characteristic of APRIL, which was subsequently designated BALM. This new TNFSF ligand was also found in fugu and three spined stickleback (41). This study also determined that BALM gene was mainly expressed in lymphoid tissues. In a more recent study, BALM orthologs have been found in cartilaginous fish (168) (**Table 3**), in addition to a BAFF-like sequence showing homology to BAFF, APRIL and BALM present in lampreys (168). In this study, the authors described that BALM is absent in all tetrapods, and there is a selective deletion of this gene in zebrafish. These data point to BALM as an ancestral BAFF-like ligand, which appeared early in evolution and was then lost when BAFF and APRIL acquired divergent functions. However, the presence of an ancient BAFF-like homolog in lampreys indicates that further research is needed to classify these closely related TNFSF members, and clarify their evolution and biological functions. From the functional point of view, it has been shown that teleost BALM can promote B cell survival and proliferation in rainbow trout kidney (170). Moreover, its transcription is upregulated on peritoneal lymphocytes after i.p. administration of VHSV (114), and it has also been demonstrated that there is a significant correlation between the expression of BALM and the progress of the proliferative kidney disease in rainbow trout (170). Although very little is yet known about this ligand, these data indicate that it plays a very important role in the activation of B cells on the responses against different types of pathogens in lower vertebrates.

## CD40L AND OX40L

CD40L homologue sequences have been identified in several teleost species, such as rainbow trout, Atlantic salmon, zebrafish, fugu, and spotted green pufferfish (41) and more recently in one cartilaginous fish, the small-spotted catshark (131) (summarized in **Table 3**). In teleost, CD40L is mainly expressed in spleen, head kidney, and gills from resting animals (41, 171, 172). Furthermore, T cell mitogens such as PHA and ConA upregulated the transcription of CD40L on the spleen, head kidney, and gills of Atlantic salmon suggesting these may be special locations for T-B cell cooperation during TD immune responses (172). Interestingly, the production of CD40L by T cells was significantly increased in zebrafish immunized with TD Ags, and the upregulation of CD40L *in vivo* increased the levels of serum IgM on those animals (171). This last set of data by Yong-Feng Gong et al. elegantly demonstrated that the interaction between T and B cells through ligation of CD40 elicits TD immune responses in fish. Studies on the small-spotted catshark paralleled these results, showing that CD40L was mainly expressed in the spleen, as well as in the mucosal tissues, such as gills and gut (131). The transcription of CD40L was significantly upregulated after the addition of T cell polyclonal activators to cultures of PBL *in vitro*. Altogether, these data suggest that CD40L produced by T cells plays a very important role on TD immune responses. Since fish lack LNs and do not form lymphoid follicles, it would be of great interest to study how and where T–B cell interactions take place to better understand the evolutionary origins of TD immune responses.

To date, only one sequence with certain homology to OX40L has been reported in zebrafish (41), designated as Dr\_TNF-New. This sequence was most similar to human OX40L and to *Xenopus* TNF-α. However, to the light of these results, it cannot be stated that true OX40L homologs exist in fish, and further studies are required its existence. In mammals, cross-linking of OX40L on B cells by OX40 expressed on activated CD4<sup>+</sup> T cells promotes B cell proliferation (22). Since fish B cells seem to maintain many functions of innate B1 cells, and B–T cell interaction might be less abundant than that seen in mammals, it is tempting to hypothesize that the absence of OX40L in fish may be compensated by CD40L on the development of TD immune responses.

## TNF-**α**

TNF-α is probably the most studied TNFSF ligand, due to its potential roles in development, cell proliferation, apoptosis, inflammation, and immunity [reviewed in Ref. (173)]. Hirono et al. first described a fish homolog of TNF-α in the Japanese flounder (135). Since then, TNF-α homologs have been identified in a plethora of teleost species (summarized in **Table 3**), such as Japanese flounder (135), sea bream (139), channel catfish (140), turbot (143), sea bass (144), mandarin fish (145), striped trumpeter (148), tongue sole (150), Atlantic salmon (146), goldfish (147), bluefin tuna (149), common carp (141, 142), and rainbow trout (136–138).

Interestingly, several teleost species present multiple isoforms of TNF-α. Two copies were initially found in rainbow trout (137) and four within the common carp (141, 142). Given that both species are tetraploid, the presence of multiple TNF-α isoforms was not surprising. However, with the discovery of at least two TNF-α genes within non-tetraploid fish species, such as bluefin tuna (149), orange-spotted grouper (174), zebrafish, and medaka (151), it has become clear that teleost have two different groups of TNF-α genes. Furthermore, the analysis of the zebrafish and medaka genomes (151) showed that members from the two different groups of TNF-α genes were found on different chromosomes, with conserved genes around them, thus indicating that the presence of these two groups is a consequence of a duplication event that occurred within bony fish.

Although both groups of TNF-α molecules can be found in various fish genomes, the role played by them in the immune response remains to be determined. On those species containing a single copy, it has been shown that TNF-α is ubiquitously expressed in all tissues analyzed from unstimulated fish such as Japanese flounder, sea bream, or mandarin fish (135, 139, 145). Results regarding the activation of TNF-α expression after treatment with lipopolysaccharide (LPS) were consistent between species; as its expression on the head kidney was significantly upregulated after treatment with LPS both *in vitro* and *in vivo.* Although TNF-α expression was also upregulated in response to viral or bacterial infections, some differences between species have been observed. For example, in turbot, virus induced higher TNF-α expression than bacteria in kidney cells (although the response was shorter in time) (143), while a very recent study showed the opposite result in tongue sole kidney, spleen, and blood cells from virus or bacteria immunized virus (150). In any case, macrophage-driven inflammation was activated in both studies. This suggests that although TNF regulation seems to be very similar among fish species, the interaction between host and pathogen might shift the spatiotemporal expression of TNF-α, to adapt the response to each specific pathogen. A nice example of this can be seen in the three striped trumpeter, which in response to the infection with the ectoparasite *Chondracanthus goldsmidi* significantly upregulates the expression of TNF-α not only in the head kidney but also in the gills, the attachment site of the parasite (148), thus demonstrating the adaptation of the host response to different pathogens.

Regarding the species that present two copies of TNF-α, some differences were found for such isoforms on each species. Both TNF-α1 and TNF-α2 were constitutively expressed in a number of tissues of healthy Atlantic salmon (146). However, incubation of head kidney leukocytes with bacteria upregulated the expression of TNF-α2 but not TNF-α1, suggesting different activation pathways for each TNF molecule. In goldfish, TNF-α2 showed higher expression levels than TNF-α1, and TNF-α2 was also able to activate primary macrophages (147). Something similar was observed in bluefin tuna, where TNF-α2 mRNA was significantly higher than TNF-α1 in blood leukocytes (149). In addition, the expression of TNF-α2 was increased by stimulation with B cell, T cell, or macrophage activators, while the expression of TNF-α1 was not affected. These observations led the authors to propose that while TNF-α1 is a ubiquitous cytokine involved in cell survival and apoptosis, TNF-α2 is an inducible form which may be a key regulator of the innate immune response. In mammals, TNFR1 and TNFR2 exert different functions, being the former involved in cell death and the latter in the regulation of the immune response (15). Little is known about TNFR in fish, so it would be of great interest to study whether the different TNF-α isoforms found in fish signal through different receptors.

Of special interest is the case of the rainbow trout where an additional TNF-α isoform was found (138): TNF-α1 and TNFα2 are expressed constitutively in some tissues, such as head kidney, and gills, and they induced a pro-inflammatory response on a rainbow trout macrophage cell line (137). By contrast, LPS induced a much faster, stronger, and longer upregulation of TNFα2 expression when compared with TNF-α1, pointing to the fact that TNF-α1 could be involved in cell survival and apoptosis, while TNF-α2 is an inducible form involved in the activation of the immune response. Moreover, the third form (TNF-α3) identified, showed low identity with the two forms previously described (138). Its basal expression was the lowest among the three molecules, but it was the most responsive of them against different stimuli, showing a strong upregulation of its expression at very early time points after stimulation of primary leukocytes with LPS or T cell mitogens. In addition, TNF-α3 induced the expression of many pro-inflammatory mediators and immune regulators (138). This suggests that TNF-α3 might be in fact a strong amplifier of the early inflammatory and immune responses.

In mammals, TNF-α has been proven as an important costimulator of B cells for their polyclonal expansion on primary immune responses (15). Fish B cells retain many innate immune features, such as phagocytic capacities (115) and expression of multiple TLRs (119), and their responses are likely to be TI since no GCs are observed. Thus, fish B cells responses seem to be based in polyclonal activation of Ag-reactive pools of cells (175). Since the direct effect of recombinant TNF-α has not been tested on fish B cells, we think it is paramount to study if this TNFSF ligand can improve B cell-mediated Ab responses.

#### LT**β**

Initially, homologous sequences of mammalian LTβ were found in zebrafish and fugu (142), designated as TNF-new (TNF-N). Posterior investigations identified homologs of this gene in medaka and zebrafish (151). Phylogenetic analysis of these sequences with known vertebrate sequences showed a closer relationship of these sequences to TNFSF3 (LTβ) than to TNFSF1 and TNFSF2. This was further supported with the inclusion of *Xenopus* sequences, which are available for all TNFSF1, 2, and 3 (50). Interestingly, in rainbow trout, two different isoforms were isolated, and named LT-β1 and LT-β2 (50) (**Table 3**). The authors hypothesized that the presence of two isoforms was a result of the additional genome duplication event that took place in salmonids around 400 million years ago (176, 177). LT-β1 was expressed in the spleen, head kidney, intestine, and gills, while LT-β2 was expressed only in the gills of rainbow trout (50). This might indicate a different role for each LTβ in the homeostasis of B and T cells. In mammals, LTβ has been involved in the formation of LN, organization of lymphoid structures, and the formation of GCs (81). The absence of LN or GC reactions in fish may explain why LTβ has not yet been identified in other cartilaginous or bony fish species. Nevertheless, immunological assays are required to clarify whether this cytokine is needed in fish, as well as its potential role in central and peripheral immune tissues in rainbow trout, the only fish in which the presence of LTβ has been shown to date (50).

#### FAS LIGAND

Fas ligand is a well-known TNFSF ligand that controls the extrinsic apoptosis pathway (178). In fish, homologue sequences of FasL have been reported in some teleost, namely, zebrafish (152), rainbow trout (41), Japanese flounder (153), Nile tilapia (154), rock bream (155), and Japanese pufferfish (*Takifugu rubripes*) (133) (**Table 3**). Homologue sequences have also been identified in cartilaginous fish (52). In all the species studied, FasL was highly expressed on spleen and head kidney, which are secondary lymphoid tissues in fish. In some species, such as rainbow trout and tilapia, it was also expressed on the gills and gut mucosal tissues (41, 154). Functional studies have determined that the T cell mitogen PHA was able to upregulate the expression of FasL in PHA-bound lymphocytes from Japanese flounder (153) while LPS was not able to do so, thus showing that activated T cells are responsible for production of FasL. In line with this, mammalian FasL has been widely reported to be expressed in Ag-activated CD8<sup>+</sup> and CD4<sup>+</sup> T cells and natural killer cells (80). A similar upregulation of FasL has been observed in rock bream after a viral challenge, suggesting that FasL may also play a role on the regulation of antiviral immune responses (155). On the other hand, Fas is expressed at low levels on resting B cells but is upregulated after Ag activation (79), making these cells more susceptible to FasL-mediated cell death. Thus, it has been proposed that FasL–Fas interaction controls the size of the B cell compartment in homeostasis, while during the immune response it seems to be responsible for the elimination of activated B cells after the generation of ASCs and memory B cells, and the subsequent clearance of the pathogen (179). Recombinant tilapia FasL has been shown to induce apoptosis on Fas-expressing HeLa cell lines (154), demonstrating its functionality. Unfortunately, the effect of FasL on fish B cells has not yet been addressed, so further investigations are needed to understand the impact of FasL on the B cell compartment in physiological or pathological conditions.

## TNF-RELATED APOPTOSIS-INDUCING LIGAND

TNF-related apoptosis-inducing ligand sequences have been characterized in several teleost fish, such as grass carp (157), zebrafish (152), rainbow trout (41), pufferfish (41), and mandarin fish (158) (summarized in **Table 3**). In humans, there is only a single TRAIL gene while two genes have been reported in avian species (41). Eimon et al. identified four TRAIL genes in zebrafish (41, 152). Hence, the authors proposed that the ancestral gene was duplicated before the divergence of ray-finned and lobe-finned fishes (around 400 million years ago) (152). In line with this, three genes with homology to TRAIL have been identified in fugu (133). This study also showed that each of these homologs presented a very different gene organization forming distinct groups (133). In a thorough study based on the search of immune genes in leukocytes from jawless fish, a sequence showing homology to TRAIL was identified in the inshore hagfish (*Eptatretus burgeri*) (156), pointing to a conserved role of this cytokine throughout evolution. In all the teleost fish analyzed, TRAIL was expressed mainly in spleen and kidney but also in mucosal tissues, such as gills, skin, and/or intestine, depending on the species. In zebrafish, overexpression of TRAIL has been reported to induce cell death by activation of the extrinsic apoptosis pathway machinery (152). In parallel, mandarin fish recombinant TRAIL-induced apoptosis of HeLa cells (158) although it was previously reported that the extracellular domain of the death receptors (DR) for TRAIL in zebrafish differs from those in human DR4 and DR5 (152), suggesting that cross-reactivity with mammalian DRs could only be achieved with certain fish TRAIL proteins. In mammals, TRAIL is involved in B cell differentiation at the GC reaction, and although fish lack secondary lymphoid structures and do not form GCs, it would be of great interest to investigate what is the role of TRAIL on B cells from lower vertebrates.

## CONCLUDING REMARKS

A plethora of studies in mammals have revealed that TNF ligand– receptor interactions elicit complex and divergent functions during the immune response. TNFSF ligands have the ability to induce both cell death or cell activation/co-stimulation and although some of these molecular mechanisms have been elucidated, the origin of these complex interactions and their multiple functionalities are not yet fully understood. TNFSF ligands play a key role in the immune response, and many researchers have taken advantage of this to develop new therapeutic strategies based on the modulation of TNF ligands. In fact, several anti-TNF neutralizing Abs are already available or under clinical trials for the treatment of inflammatory and ADs. Moreover, many immunologists are considering the option of using engineered TNFSF ligands as vaccine adjuvants. For instance, DNA vectors encoding for CD40L multi-trimers are being tested as adjuvants for a HIV vaccine, based on their positive effects on B cell costimulation during the vaccination process, thus generating high-affinity Abs against the virus. Hence, this is a fascinating research topic that is becoming of great importance within the field of immunotherapy. Most of the questions that have not been answered yet could be explained through the analysis of the evolution of the TNF superfamily of ligands and receptors. From an evolutionary point of view, it is widely accepted that acquired immunity first appeared in fish. This animal taxon possesses unique immune features, and recent evidences suggest that we can learn much about the evolution and functionality of TNF ligands in fish. For instance, they retain a unique TNF cytokine, BALM, which has become extinct in tetrapods, and although its function is still unknown, many evidences point to the fact that this molecule could be key to understand the molecular

#### REFERENCES


and functional divergence of BAFF and APRIL. On the other hand, fish express several isotypes of some TNF ligands (i.e., TNF-α) while mammals only express one. In fish, these ligand isotypes show different expression patterns, activation profiles, and immune functions. Thus, analyzing these isotypes could provide us with vital information about their original function, or about how and why some members of the TNF family were lost or acquired throughout evolution. Moreover, this analysis could also help to elucidate the convergence or divergence of some immune functions played by TNF ligands.

#### AUTHOR CONTRIBUTIONS

CT and AG reviewed the bibliography and wrote the manuscript.

#### FUNDING

This work was supported by project AGL2014-54456-JIN from the Spanish Ministry of Economy and Competitiveness (MINECO) and by the European Research Council (ERC Consolidator Grant 2016 725061 TEMUBLYM).


and biological activity of the recombinant protein. *Fish Shellfish Immunol* (2009) 27:423–32. doi:10.1016/j.fsi.2009.06.019


the trout primary lymphoid tissues. *Immunogenetics* (1997) 46:367–75. doi:10.1007/s002510050290


rheumatoid arthritis. *Scand J Rheumatol* (2007) 36:365–72. doi:10.1080/ 03009740701286615


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

*Copyright © 2018 Tafalla and Granja. 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.*

# Peculiar Expression of CD3-Epsilon in Kidney of Ginbuna Crucian Carp

*Ryuichiro Miyazawa1 , Norifumi Murata1 , Yuta Matsuura2 , Yasuhiro Shibasaki <sup>3</sup> , Takeshi Yabu4 and Teruyuki Nakanishi1 \**

*1Department of Veterinary Medicine, Nihon University, Fujisawa, Japan, 2Research Center for Fish Diseases, National Research Institute of Aquaculture, Minami-ise, Japan, 3Department of Pathobiology, School of Veterinary Medicine, University of Pennsylvania, Philadelphia, PA, United States, 4Department of Applied Biological Science, Nihon University, Fujisawa, Japan*

TCR/CD3 complex is composed of the disulfide-linked TCR-αβ heterodimer that recognizes the antigen as a peptide presented by the MHC, and non-covalently paired CD3γεand δε-chains together with disulfide-linked ζ-chain homodimers. The CD3 chains play key roles in T cell development and T cell activation. In the present study, we found nor or extremely lower expression of CD3ε in head- and trunk-kidney lymphocytes by flow cytometric analysis, while CD3ε was expressed at the normal level in lymphocytes from thymus, spleen, intestine, gill, and peripheral blood. Furthermore, CD4-1+ and CD8α+ T cells from kidney express Zap-70, but not CD3ε, while the T cells from other tissues express both Zap-70 and CD3ε, although expression of CD3ε was low. Quantitative analysis of mRNA expression revealed that the expression level of T cell-related genes including *tcrb*, *cd3*ε, *zap-70*, and *lck* in CD4-1+ and CD8α+ T cells was not different between kidney and spleen. Western blot analysis showed that CD3ε band was detected in the cell lysates of spleen but not kidney. To be interested, CD3ε-positive cells greatly increased after 24 h in *in vitro* culture of kidney leukocytes. Furthermore, expression of CD3ε in both transferred kidney and spleen leukocytes was not detected or very low in kidney, while both leukocytes expressed CD3ε at normal level in spleen when kidney and spleen leukocytes were injected into the isogeneic recipient. Lower expression of CD3ε was also found in kidney T lymphocytes of goldfish and carp. These results indicate that kidney lymphocytes express no or lower level of CD3ε protein in the kidney, although the mRNA of the gene was expressed. Here, we discuss this phenomenon from the point of function of kidney as reservoir for T lymphocytes in teleost, which lacks lymph node and bone marrow.

Keywords: CD3-epsillon, CD4, CD8, kidney, teleost, ginbuna crucian carp, T lymphocytes

#### INTRODUCTION

The antigen receptor complex on T-cell (TCR/CD3) consists of the disulfide-linked TCR-αβ heterodimer that recognizes the antigen as a peptide presented by the MHC, and non-covalently paired CD3γε- and δε-chains together with disulfide-linked ζ-chain homodimers. In mice and humans, CD3ε and ζ chains as well as TCRα and β chains are essential for surface expression of TCR/CD3 complex, while CD3δ and γ chains are individually dispensable (1). Dynamic change of TCR/CD3 cell surface expression has been reported in resting and antigen-activated T cells (2). CD3 chains play

#### *Edited by:*

*Brian Dixon, University of Waterloo, Canada*

#### *Reviewed by:*

*Kevin R. Maisey, Universidad de Santiago de Chile, Chile Francesco Buonocore, Università degli Studi della Tuscia, Italy*

> *\*Correspondence: Teruyuki Nakanishi nakanishi.teruyuki@ nihon-u.ac.jp*

#### *Specialty section:*

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

*Received: 02 February 2018 Accepted: 28 May 2018 Published: 13 June 2018*

#### *Citation:*

*Miyazawa R, Murata N, Matsuura Y, Shibasaki Y, Yabu T and Nakanishi T (2018) Peculiar Expression of CD3-Epsilon in Kidney of Ginbuna Crucian Carp. Front. Immunol. 9:1321. doi: 10.3389/fimmu.2018.01321*

**177**

critical roles in the various phases of thymocyte development. For instance, the removal of CD3ε led to a complete impairment of thymocyte development at the DN stage and CD3γε is the most critical for DN-to-DP transition (3). The role of CD3 chains in mature T cell activation has been also reported (4, 5). Ahmadi et al. (6) showed that co-transfer of CD3 and TCR genes into primary murine T cells enhanced TCR expression and antigenspecific T-cell function *in vitro*. Further, they demonstrated that addition of CD3 protein is effective to enhance the avidity, antitumor activity, and functional memory formation of TCR gene-modified T cells *in vivo*. In contrast, downregulation in the expression of CD3 chains in T cells and impaired immune responses have been reported in patients with malignant and/or inflammatory autoimmune diseases (7–9). These reports indicate that the critical roles of CD3 chains in mature T cell activation as well as in T cell development and signaling.

Fish CD3 genes have been reported from chondrostean fish (sterlet) and teleost [e.g., Japanese flounder, fugu, Atlantic halibut, sea bass, Atlantic salmon, and common carp (10)]. CD3 transcripts were widely expressed in teleost tissues including lymphoid tissues (e.g., thymus, head- and trunk-kidney, and spleen), mucosal tissues (e.g., gill, skin, and intestine), and peripheral blood leukocytes (PBL) (10–14). Non-mammalian vertebrates including birds, amphibians, and teleost fish possess only three types of CD3, e.g., CD3-γδ, CD3-ε, and the -ζ chain with CD3 γδ reflecting the common ancestor of mammalian CD3-γ and CD3-δ (15). The lack of interaction between chicken TCRαβ heterodimers and the human CD3 complex has been reported (16). Furthermore, there exist significant differences in the relative functions of the various CD3 chains even between mice and humans (17). In fish, however, function of CD3 chains, particularly, the role in the expression of TCR/CD3 complex remains unknown.

The obvious differences between fish and mammals are that fish lack a bone marrow and lymph nodes, and instead, the kidney is a major lymphoid organ in the teleost in addition to the thymus, spleen, and mucosa-associated lymphoid tissues (18). Teleost kidney is considered to be homologous organ to the bone marrow and lymph node in mammals (19). It has been reported that the presence of hematopoietic stem cells in the kidney of ginbuna and zebrafish (20, 21). Abundant presence of immature leukocytes or blast cells in the kidney of teleost also suggests that teleost kidney is equivalent to the bone marrow of mammals. It is well known in mammals that interactions between T cells and dendritic cells in the lymph nodes are crucial for initiating cell-mediated adaptive immune responses (22, 23) and the germinal centers are main sites for T cell-dependent immune responses (24). In fish, however, information on the tissues or sites equivalent to lymph nodes and germinal centers is limited.

In the present study, we found that CD4-1<sup>+</sup> and CD8α+ T lymphocytes from head-and trunk-kidney express Zap-70, but not CD3ε, while T lymphocytes from tissues except kidney express both CD3ε and Zap-70. Interestingly enough, T lymphocytes from the head- and trunk-kidney become positive for the expression of CD3ε after 24 h in *in vitro* culture. Furthermore, expression of CD3ε in kidney leukocytes became positive when kidney leukocytes were injected into the isogeneic recipient and migrated into spleen. These results indicate that expression of CD3ε molecule was suppressed in the kidney. Here, we discuss this phenomenon from the point of function of kidney as reservoir for T lymphocytes in teleost, which lacks lymph node and bone marrow.

#### MATERIALS AND METHODS

#### Experimental Fish

Triploid female ginbuna crucian carp (*Carassius auratus langsdorfii*) from Okushiri Island in Hokkaido (OB1 clone) weighing 20–30 g were used for the experiment. Offspring of OB1 clone were reproduced by naturally occurring gynogenesis artificially inseminated with loach sperm and were maintained in tanks with running water at 25 ± 1°C and fed twice daily with commercial pellets throughout the experiments.

## Identification and Characterization of Ginbuna CD3**ε**

To obtain the partial ginbuna CD3ε nucleotide sequences, we performed RT–PCR using primers (**Table 1**) designed using nucleotide sequence of zebrafish *cd3e* (NM\_001326401) and Japanese flounder *cd3e* (AB081751). PCR was carried out in 40-µl reaction mixtures containing Prime STAR HS, with reaction conditions consisting of denaturation at 96°C for 2 min and 30 cycles of denaturation at 94°C for 10 s, annealing at 60°C for 10 s, polymerization at 72°C for 30 s, and extension at 72°C for 2 min. The PCR products were subcloned into the pGEM-T Easy plasmid vector by using a TA-Cloning method (Promega, Madison, WI, USA). After confirming through sequencing, 5′- RACE and 3′-RACE protocols (TaKaRa Bio, Shiga, Japan) were used to obtain full-length gCD3ε sequences using the primers in **Table 1**. Nucleotide and amino acid sequence were analyzed using GENETYX-WIN version 9.0 and CLUSTALW. Similarity searches were performed using BLAST against the NCBI nonredundant protein database and the Protein Data Bank. Ig domains, CXXC motif, and immunoreceptor tyrosine-based activation motifs (ITAM) were predicted using Simple Modular Architecture Research Tool. Phylogenetic analysis was performed using molecular evolutionary genetics analysis.

## Recombinant Protein Production

The cDNA sequences of ginbuna *cd3e* were amplified and subcloned into a pET-16b vector (Novagen, Madison, WI, USA). To increase solubility, transmembrane (TM) domain was deleted within CD3ε sequence and named as CD3e–ΔTM. The CD3e– ΔTM plasmid DNA was amplified using the primers (gCD3edel-TM-F, gCD3e-del-TM-R in **Table 1**) phosphorylated with Prime STAR mutagenesis kit (TaKaRa Bio, Shiga, Japan). The construct was designated as pET–CD3e–ΔTM. The CD3e–ΔTM protein was expressed in *Escherichia coli* BL21 (DE3) pLysS cells (Novagen) that had been transformed with pET–CD3e–ΔTM. After the three chromatography purification steps, sequential His-tag affinity purification, gel filtration chromatography, and endotoxin removal, the recombinant proteins were used for immunization of rabbit.

#### Table 1 | Oligonucleotide primers used in this study.


#### Production of Polyclonal Antibody

New Zealand White rabbit were immunized with the purified recombinant CD3ε according to the standard method. Rabbits were bled by cardiac puncture under deep terminal anesthesia and the serum were purified by protein G sepharose (GE Healthcare, Piscataway, NJ, USA) according to the manufacturer's protocols and guidelines. The antibody was further purified by an affinity column, which was prepared by coupling of recombinant CD3e–ΔTM protein to NHS-activated Sepharose (GE Healthcare, Piscataway, NJ, USA) according to the manufacturer's instructions.

#### Preparation of Leukocytes

Fish were deeply anesthetized with 35 ppm ethyl-4-aminobenzoate (Benzocaine, Sigma-Aldrich, St. Louis, MO, USA), and their spinal cords were severed for euthanasia. Fish were bled from the caudal blood vessels with a heparinized syringe, and the thymus, spleen, head-kidney, trunk-kidney, gill, and intestine were dissected. To avoid contamination with blood, 10 ml of PBS with 10 U/ml of heparin (Wako Chemicals, Osaka, Japan) was injected into gill tissue through the bulbus arteriosus. All subsequent manipulations of cells were done at 4°C.

For the thymus, spleen, head-kidney, and trunk-kidney, the organs were placed on a stainless steel mesh filter (100 µm) and pressed through with 5 ml of HBSS (Nissui Pharmaceutical Co. Ltd., Tokyo, Japan) to create single-cell suspensions. For gill and intestine, the tissues were incubated with PBS containing 1 mM DTT (Wako Chemicals, Osaka, Japan) and 1 mM EDTA for 15 min after mincing with scissors. After incubation, the organs were washed and dissociated by incubating with calcium- and magnesium-free Hank's Balanced Salt Solution (CMF-HBSS) containing 0.1 mg/ml collagenase (Wako Chemicals, Osaka, Japan), 0.1 mg/ml DNase (Sigma-Aldrich, St. Louis, MO, USA), and 5% FBS for 90 min with shaking at room temperature. Dissociated organs were disaggregated by pressing through the stainless steel mesh filter into HBSS. The buffy coat from peripheral blood and leukocytes from tissues were collected by centrifugation at 400 × *g* for 5 min at 4°C. After discarding the supernatant, 1 ml of distilled water was added to cell pellet and gently mixed with a pipette to lyse mature erythrocytes. Subsequently, 9 ml of the 0.2% FBS-HBSS was added and the cells were washed twice by centrifugation. Cell concentration and viability were determined by trypan blue dye exclusion with a hemocytometer. Viability of cells was approximately 90%.

#### Cell Culture

Head- and trunk-kidney lymphocyte were suspended in RPMI1640 medium (Thermo Fisher Scientific Inc., MA, USA) supplemented with 1% ginbuna serum. The cells were seeded in 6-well plates at 1 × 106 cells/2 ml/well at 25°C with 5% CO2 for 24 h.

#### Flow Cytometry

5 × 106 cells/ml of leukocytes from the various tissues were fixed with 2% paraformaldehyde (PFA) followed by cell membrane permeabilization with 0.1% saponin for 10 min. Cells were then incubated with 1:300 anti-gCD3ε antibody or Rabbi (DA1E) mAb IgG Isotype control (CST, MA, USA) for 45 min at 4°C, washed three times, and stained with 1:500 diluted Alexa Fluor® 647 conjugated secondary antibody (Thermo Fisher Scientific Inc., MA, USA) against anti-gCD3ε antibody. The cells were then washed three times and served for flow cytometric analysis. Lymphocytes were gated on FS and SS dot plot and then analyzed using a FACS Canto flow cytometer (Becton Dickinson, NJ, USA).

For two-color immunofluorescence analysis of cell surface antigens along with T cell-specific intracellular markers including Zap-70 and CD3ε, kidney leukocytes were first incubated with mAbs against CD4-1(6D1, rat), CD8α (2C3, rat), IgM (B12, mouse), phagocyte (GB21, mouse), and thrombocyte (GB10, mouse) markers and then fixed with 2% PFA followed by cell membrane permeabilization with 0.1% saponin for 10 min. Cells were then incubated with 1:300 anti-gCD3ε antibody or 1:50 anti-hZap-70 (rabbit, CST, MA, USA) for 45 min at 4°C, washed three times, and stained with 1:500 diluted Alexa Fluor® 488 donkey Anti-Rat IgG (H + L) antibody, Alexa Fluor® 488 goat Anti-mouse IgG (H + L) antibody, and Alexa Fluor® 647 goat Anti-rabbit IgG (H + L) antibody (Thermo Fisher Scientific Inc., MA, USA). A donkey anti-rat IgG antibody was used for mAbs 2C3 and 6D1, a goat anti-mouse IgG antibody was used for mAbs B12, GB21, and GB10 along with a goat anti-rabbit IgG antibody was used for mAbs gCD3ε and hZap-70. The cells were then washed three times. Lymphocytes were gated on FS and SS dot plot and lymphocytes were then analyzed for double staining with the mAbs. Doublets discrimination was performed in FSA-H/ FSA-W and SSC-H/SSC-W dot plots with Flowjo 7 (TreeStar).

## Transcriptional Analysis of FACS Sorted Populations

Leukocytes from kidney and spleen were labeled with mAbs against CD4-1 and CD8α as described above. Dead cells were eliminated by 2.5 µg/ml of propidium iodide (Thermo Fisher Scientific Inc., MA, USA). Lymphocyte fraction of kidney and spleen leukocytes was gated and doublets discrimination was performed as described above. CD4-1<sup>+</sup> and CD8α+ cells were isolated by FACS Aria II cell sorter (Becton Dickinson, NJ, USA). Purities of FACS sorted CD4-1 and CD8α were confirmed to be more than 95% when the sorted lymphocytes were re-analyzed with the mAbs used for sorting by FACS analysis. Total RNA was extracted from 5 × 105 cells of FACS sorted cells using the ReliaPrep RNA Tissue Miniprep System (Promega, Madison, WI, USA) according to the manufacturer's protocols and guidelines. cDNA was synthesized from total RNA from each sample using a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, CA, USA) according to the manufacturer's protocols and guidelines. mRNA expression analysis was performed by Real-time PCR using a Thermal Cycler Dice® Real Time System (TaKaRa Bio, Shiga, Japan). PCR reactions were performed with 5 µl of 1:50 diluted cDNA, 12.5 µl of SYBR®*Premix Ex Taq* (TaKaRa Bio, Shiga, Japan), and 200 nM of each specific primer pair (**Table 1**) in 25 µl mixtures under the following conditions: one cycle at 95°C for 30 s, 45 cycles at 95°C for 5 s, 60°C for 30 s. Melting curve analysis showed that there was no primer dimer formation.

Target genes were amplified on the same plate with the internal control genes, *ef1a* or T cell control gene, *lck*, and the relative mRNA quantities were determined. Raw data were analyzed by the 2−ΔΔCT method (25) normalized to *ef1a* or *lck*.

#### Transfer of Kidney and Spleen Leukocytes Into Isogeneic Recipient

Kidney and spleen leukocytes of donor fish were prepared as mentioned above and were labeled with CFSE (Invitrogen) for detection by flow cytometry according to Toda et al. (26). Cell suspensions were adjusted to 2 × 106 cells/ml and labeled with 5 µM CFSE (invitrogen) for 10 min at room temperature. The reaction was stopped by the addition of an equal volume of HBSS at 4°C followed by three washes. 100 µl of 5 × 107 cells/ml of CFSE-stained kidney or spleen leukocytes were then injected into the naïve isogeneic recipient *via* caudal blood vein. Kidney and spleen cell suspension of recipient fish were prepared and the percentages of CFSE-positive cells were analyzed by FACS.

#### Western Blot Analysis

Leukocytes from kidney and spleen were lysed in 25 mM Tris– HCl (pH 7.4), 150 mM NaCl, 0.1% TritonX-100, 0.05% SDS. The extracted proteins were resolved on a SDS-polyacrylamide gel and electroblotted onto a PVDF membrane as described by Yabu et al. (27). The Membrane was blocked with blocking reagent (Block-Ace, Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) for 1 h at room temperature. The Membrane was probed with 1:300 Anti-gCD3ε Ab overnight at 4°C, washed five times, and incubated with horseradish peroxidase goat anti-rabbit IgG antibody (Sigma-Aldrich, St. Louis, MO, USA) diluted 1:10,000 for 1 h at room temperature, and washed an additional five times. Membranes were visualized by Western Lightning ECL Pro (Perkin Elmer, Inc., Waltham, MA, USA) and exposed to Hyperfilm ECL (GE Healthcare, Piscataway, NJ, USA).

#### Immunohistological Analysis

Cryosections (8 µm) were prepared as previously reported (28). For immuno-staining, frozen sections were incubated with 1:300 diluted rabbit anti-CD3ε Ab or Rabbi (DA1E) mAbIgG Isotype control (CST, MA, USA) for 8 h at 4°C, washed, and then stained with 1:500 diluted Alexa Fluor® 488 goat Anti-rabbit IgG (H + L) antibody as a secondary antibody. Sections were then washed three times and nuclei were stained with DAPI (Sigma-Aldrich, St. Louis, MO, USA). Finally, the sections were mounted with ProLong Gold anti-fade mounting medium (Life Technologies). Sections were examined by fluorescence microscopy (Olympus IX71) with a digital camera and software (Olympus DP73).

## *In situ* Hybridization (ISH)

A 1,295-bp of gCD3ε cDNA was subcloned into the pGEM-T Easy plasmid vector and then appended T7 and SP6 promoter sequence at the 5′ and 3′ terminal, respectively, by PCR reaction using primers (**Table 1**). The PCR product was used for sense and antisense RNA probe synthesized using DIG RNA Labeling Kit (Sigma-Aldrich, St. Louis, MO, USA). For the ISH of tissue sections, tissue samples from kidney, spleen, and thymus were fixed at 4°C for 12 h in 4% PFA. Cryosections (7 µm) were prepared on the slide. ISH was performed as described previously by Nagasawa et al. (29). The sections were then incubated in 0.0018% of BCIP (Sigma-Aldrich, St. Louis, MO, USA) and a NTMT solution containing 0.0035% of NBT (Nacalai tesque, Japan) at RT in the dark. After the color reaction had occurred, sections were washed with PBS. Finally, the sections were mounted with 70% glycerol. Sections were examined under BX51 microscope (Olympus, Tokyo, Japan).

#### Statistics

Results of FCM analysis were statistically compared using twoway ANOVAs, followed by Tukey's multiple comparisons tests to detect significant difference between means in the percentage of positive cells. A *p* value of <0.05 was considered statistically significant.

## RESULTS

## Ginbuna CD3**ε** Sequence Analysis

Sequence analysis of 5′ RACE and 3′ RACE PCR product revealed that ginbuna CD3ε cDNA consists of 1,499 nucleotides with a 519 bp open reading frame encoding 173 amino acid (Figure S1A in Supplementary Material). Comparison of gCD3ε with Atlantic salmon CD3ε (NM \_001123622) and human CD3ε (NM \_000733.3) revealed that the polypeptide was composed of a signal peptide sequence, CXXC motif, a Ig-like domain containing two Ig-fold cysteine, and ITAM in cytoplasmic domain (Figures S1B,C in Supplementary Material). These data indicate that gCD3ε has similar feature to that of other vertebrate CD3ε. Phylogenetic analysis revealed that gCD3ε was classified into the vertebrate CD3ε group (Figure S1D in Supplementary Material).

## Specificity of Antibody Against Ginbuna CD3**ε**

Specificity of rabbit antibody purified with antigen column using recombinant ginbuna CD3ε (rgCD3ε) was examined. Western blot analysis showed that the antibody is specific to native ginbuna CD3ε (gCD3ε) present in thymus and spleen since positive band was detected at 20–25 kDa as expected molecular weight (Figure S2 in Supplementary Material) and disappeared after adsorption with the rgCD3ε (Figure S3 in Supplementary Material). Immuno-precipitation followed protein sequencing by LC-MS/MS revealed that the antibody recognized gCD3ε (Figure S4 in Supplementary Material).

## CD4-1**+** and CD8**α+** Kidney Lymphocytes Express Zap-70, but not CD3**ε**

In the present study, we found the expression of CD3ε in the lymphocytes from thymus, spleen, intestine, gill, and peripheral blood lymphocytes but not in the lymphocytes of head- and trunk-kidney where CD4-1 and CD8α positive T cells were present (**Figure 1**). Immuno-histochemical analysis also showed that CD3ε-positive cells are present in all of tissues examined except kidney and the morphology of antibody-positive cells showed the typical feature of lymphocyte (**Figure 2**).

We previously reported the distribution of CD4-1 and CD8α positive lymphocytes in both lymphoid and non-lymphoid tissues of adult fish (30, 31). Present dual immune-fluorescence analysis revealed that CD4-1 and CD8α positive lymphocytes in the head- and trunk-kidney did not express CD3ε, while CD3ε expression was observed in other tissues, e.g., the thymus, spleen, gill, intestine, and peripheral blood (**Figure 3**). However, expression of Zap-70 was found in all tissues including head- and trunk-kidney. Majority of CD4-1 and CD8α positive lymphocytes express lower level of both CD3ε and Zap-70 in the thymus. Similar phenomenon was also observed in the spleen and PBL, although the tendency was more apparent in the thymus than spleen and PBL.

#### FACS Sorted CD4-1**+** and CD8**α+** Lymphocytes Express Transcripts of T Cell-Related Markers

Since we found that CD4-1<sup>+</sup> and CD8α+ kidney lymphocytes did not show the expression of CD3ε protein, we then examined the expression of CD3ε at mRNA level. All T cell-related genes including *tcrb*, *cd3*ε, *zap-70,* and *lck* were expressed to the similar extent in both kidney and spleen (**Figure 4**).

#### Expression of CD3**ε** mRNA in Kidney as Well as Thymus and Spleen

We confirmed the expression of CD3ε mRNA in kidney as well as thymus and spleen by ISH (**Figure 5**). CD3ε mRNA positive cells were detected as small lymphocytes in the intertubular tissue of trunk-kidney (**Figure 5C**), although the number of positive cells in kidney was fewer when compared to thymus where most of cells were positive (**Figure 5A**).

#### Presence of CD3**ε** Protein in Spleen but not Kidney Leukocytes

We further examined the expression of CD3ε protein in kidney and detected a clear band of 20 kDa in spleen but not in kidney by Western Blot analysis (**Figure 6**). No band was detected even at higher dose of samples of kidney.

## Increased Expression of CD3**ε** Gene in Kidney Leukocytes After *In Vitro* Culture

FACS analysis revealed that kidney leukocytes become positive for the expression of CD3ε after 24 h *in vitro* culture (**Figure 7A**). The number of CD3ε-positive cells was approximately 20 folds after the culture compared to that of leukocytes before the culture, although Zap-70-positive cells also two and three times increased (**Figure 7B**). However, mRNA expression of *cd3e* normalized to T cell control gene, *lck* was not different in cells before and after the culture (**Figure 7C**).

We also examined the changes of CD3ε expression after alloantigen stimulation and bacterial infection. Kidney leukocytes from fish immunized with scale allografts or infected with *Edwardsiella tarda* did not show the increase of CD3ε expression (Figures S8A,B in Supplementary Material).

## Expression of CD3**ε** in Migrated Kidney Lymphocytes Into Recipient Spleen

Since we found that kidney leukocytes express CD3ε protein after *in vitro* culture, we suspect that kidney environment suppresses the expression of CD3ε protein. We then examined the CD3ε expression of kidney leukocytes migrated into other tissues of isogeneic recipient. More than 3 × 104 of CFSE-positive lymphocytes were obtained and percentages of CFSE positive lymphocytes were 1.2–2.4% in kidney, spleen, and PBL of recipients fish (Figure S10 in Supplementary Material). CFSE-stained kidney leukocytes migrated into the spleen of isogeneic recipient showed the appearance of CD3ε expressing cells, while kidney cells migrated into the kidney of the recipient showed no expression (**Figure 8A**). Interestingly enough, CFSE-stained spleen leukocytes migrated into the kidney of the recipient failed to express CD3ε, although the leukocytes express CD3ε in the recipient spleen (**Figure 8B**).

## DISCUSSION

In the present study, we found that all the leukocytes including CD4-1<sup>+</sup> and CD8α+ T cells did not express CD3ε molecule in the head- and trunk-kidney, while lymphocytes from other tissues including thymus, spleen, intestine, gill, and peripheral blood expressed CD3ε. However, CD4-1+ and CD8α+ T cells in the kidney expressed Zap-70 as in other tissues. Furthermore, CD4-1<sup>+</sup> and CD8α+ T cells in the head- and trunk-kidney become positive for the expression of CD3ε after 24 h in *in vitro* culture. Gene expression analysis revealed that CD4-1<sup>+</sup> and CD8α+ T cells express both *cd3e* and *zap-70* together with *lck* and *tcrb*, and there was no difference between kidney and spleen or between before

and after *in vitro* culture. To be interested, expression of CD3ε of spleen lymphocytes was suppressed in the kidney of recipient, while kidney lymphocytes expressed CD3ε in the spleen of recipient in leukocyte transfer experiment using isogeneic fish. These results suggest that the expression of CD3ε protein, but not mRNA, is suppressed in kidney environment.

Present study revealed that T cell-related genes including *tcrb*, *cd3*ε, *zap-70,* and *lck* were expressed in CD4-1<sup>+</sup> and CD8α+ T cells, while FACS analysis showed that CD3ε was neither expressed in the cytoplasm nor at cell surface of lymphocytes in the kidney. Western Blot analysis also supported the results of FACS analysis. Accordingly, these results suggest that CD3ε expression is regulated at transcriptional level and CD3ε is expressed at mRNA but not protein level in ginbuna kidney. It has been reported in mammals that CD3ε and ζ chains as well as TCRα and β chains are essential for surface expression of TCR/ CD3 complex (1). Taken together, it is possible that TCRα and β chains as well as CD3ε are not expressed on the cell surface or present in cytoplasm, although it is difficult to confirm due to the lack of antibody against TCR in fish including ginbuna. Critical role of CD3ε in TCR signaling has been extensively studied in mammals (3, 32). In fish, however, little is known about the role of CD3 chains in TCR signaling and transcriptional regulation of CD3 expression to date and further studies focusing on CD3ε is required.

To be interested, CD3ε-positive cells greatly increased after 24 h in *in vitro* culture of kidney leukocytes. Furthermore, kidney lymphocytes migrated into spleen of recipient expressed CD3ε as mentioned above. These results suggest that expression of CD3ε protein in lymphocytes was suppressed in the kidney. In our study on the cytotoxicity of CD8<sup>+</sup> T cells against allogeneic target cells, effector kidney leukocytes are required to culture *in vitro* for at least 8 h before the mixture with the target cells to induce cytotoxic activity (unpublished data). Dynamic and rapid changes in the cell surface expression of TCR/CD3 complex have been reported and the cell-surface levels of the complex present a balance among internalization, recycling, and degradation of existing complexes (33). Furthermore, impaired cell-mediated immune responses due to decreased expression of the CD3ζ or CD3ε chain have been reported in many patients with malignant and inflammatory autoimmune diseases. For instance, downregulation of CD3ε but not CD3ζ expression in CD4<sup>+</sup> and CD8<sup>+</sup> T cells has been reported in patient with lung carcinomas (9). Matsuda et al. (7) have reported the decreased expression of signal-transducing CD3ζ chains in T cells from the joints and peripheral blood of rheumatoid arthritis patients. Chen et al. (8) have reported that decreased expression of both CD3ζ and CD3ε result in an increased *ex vivo* susceptibility to apoptosis of peripheral blood T cells in patients with chronic myeloid leukemia. Taken together, it is possible that the function of T cell subsets is suppressed in ginbuna kidney environment.

Zap-70 is a part of the TCR/CD3 complex and is essential for the normal development of T cells and TCR signaling. In the present study, we found that Zap-70 was expressed as protein while CD3ε was not in kidney of ginbuna by FACS analysis. In an early event in TCR activation, Zap-70 is recruited to the TCR/ CD3 complex upon activation after the phosphorylation and activation of Lck and promotes recruitment and phosphorylation of downstream adaptor or scaffold proteins (34). Zap-70 is present as protein in the cytoplasm of αβ T cells and epithelial γδ

T cells except for some γδ T cells in peripheral lymphoid tissues (35). In contrast, dramatic and rapid changes in the expression of CD3 chains on cell surface or in cytoplasm have been reported as mentioned above. Therefore, the difference in the expression between CD3ε and Zap-70 can be attributed to the difference in the role of two molecules.

Teleost kidney is an important hematopoietic organ (36) and has morphological similarities with the bone marrow in higher vertebrates (37). The kidney also serves as a secondary lymphoid organ involved in the induction and activation of immune responses (38). In mammals, the elimination of activated T cells

Figure 3 | Dual fluorescence analysis of CD3ε+, Zap-70 with lymphocyte markers in tissues. Leukocytes from spleen, kidney, and thymus were stained with the anti-CD4-1 and CD8α mAbs followed by Alexa Fluor® 488 anti-rat IgG, and stained with anti-gCD3ε Ab or anti-hZap-70 mAb followed by 647 goat anti-rabbit IgG. (A–C) Lymphocytes were gated on FS and SS dot plot. (D) Leukocytes from peripheral blood leukocytes were stained with anti-CD4-1, CD8α, IgM, phagocyte, and thrombocyte mAbs, respectively, followed by Alexa Fluor® 488 goat anti-rat or mouse IgG, and stained with anti-gCD3ε Ab followed by Alexa Fluor® 647 goat anti-rabbit IgG. Mean ± SD of more than three independent experiments are shown.

at the end of immune response is essential to maintain peripheral immune tolerance and avoid excessive immune responses. Resting mature T lymphocytes in the periphery start to proliferate and then undergo the activation-induced cell death *via* apoptosis when the T cells are activated by repeated stimulation of their TCR (39). Here, we hypothesize that kidney of cyprinid fish may play a role as the reservoir of resting mature T lymphocytes. That is, T lymphocytes in kidney are suppressed in the expression of CD3ε protein and then activated after the migration into other tissues such as spleen. This hypothesis is strongly supported by our transfer experiment of CFSE-stained lymphocytes.

Flow cytometric analysis of CD3ε protein expression in tissues has been reported in several fish species. Considerably high percentages (10–40%) of CD3ε+ cells were detected among total head-kidney lymphocytes of rainbow trout (40–42) and Japanese flounder (43). The percentages were similar or even higher in head-kidney rather than spleen in these species, although number of CD3ε+ cells was relatively lower in head-kidney, spleen, and PBLs than that in thymus, gill, and intestine, and Western blot analysis showed that head-kidney preparations appeared negative or below the detection limit (40). Present results with the lack of CD3ε protein expression in kidney in cyprinid species do not agree with the abundant presence of CD3ε+ cells in kidney in rainbow trout and flounder. Presence of species-specific differences in fish physiology including immune responses has been reported. For instance, ultraviolet B (UVB) irradiation markedly

enhanced the blood respiratory burst and cytotoxic activity in carp, although these parameters were significantly suppressed in the head kidney. In contrast, rainbow trout respiratory burst was affected only after exposure with the highest dose of UVB (44). Atlantic cod lacks the genes for CD4, MHC class II, and invariant chain involved in making and transporting MHC class II (45). However, Atlantic cod is not exceptionally susceptible to disease under natural conditions (46). Instead, Atlantic cod has a highly expanded number of MHC class I genes and unique and markedly expanded TLR genes resulting in the highest number of TLRs found in a teleost. Thus, teleost immune system is greatly diverse among species or fish groups. Accordingly, difference of CD3ε protein expression in kidney between cyprinid and other species may be attributed to the difference among species.

In the present study, we found that majority of CD4-1<sup>+</sup> and CD8α+ lymphocytes express lower level of both CD3ε and Zap-70 in the thymus, spleen, and PBL when compared to CD4-1<sup>−</sup>, CD8α−, and CD3ε+ cells. We found that there are two sIgMpositive lymphocytes in the spleen and kidney of ginbuna, sIgMlow and sIgMhigh. In our previous study, sIgM-positive lymphocytes showed moderate non-specific cytotoxicity, while CD8α+ lymphocytes exhibited high-specific killing of allogeneic target cell lines when effector donor fish were sensitized by alloantigens (26). Furthermore, we also found that sIgMlow cells expressing granzyme and perfolin genes exhibited moderate cytotoxicity against allogeneic target cell lines, although sIgMhigh did not express these genes and showed no cytotoxicity suggesting that sIgMhigh cells are B lymphocytes (unpublished data). It has been reported that activated NK cells express cytoplasmic CD3ε protein in human adult and NK cell clones established from human fetal liver express CD3γ,δ,ε complexes in the cytoplasm but not cell surface (47). Similarly, Phillips et al. (48) has reported that fetal NK cells mediate cytolytic function and express cytoplasmic

Figure 5 | Gene expression analysis of CD3ε in tissue: *in situ* hybridization. Cryostat sections were hybridized with an antisense (A = thymus, B = spleen, C = trunk-kidney) or sense (a = thymus, b = spleen, c = trunk-kidney) probe. No unspecific staining was observed (a–c). The signal of CD3ε mRNA is observed (A–C). CD3ε-expressing cells are indicated with arrow head (B,C). NT represents nephric tubule in (C) (trunk-kidney). Scale bar = 10 µm.

CD3δ,ε proteins. Furthermore, it has been reported that fish NK cells also express Fc receptor (49). Taken together with previous studies, present results suggest that sIgMlow cells among CD4-1<sup>−</sup>, CD8α−, and CD3ε+ cells are NK cells with Fc receptor.

Present study revealed that considerable numbers of CD3ε+ CD4-1<sup>−</sup>CD8α− cell populations (approx. 40%) were present in spleen of carp and goldfish as well as ginbuna. It is difficult to conclude that all the populations are NK cells. Recently, lymphocyte populations involved in innate immunity have been discovered in human and mice as "innate lymphocytes," which includes NK cells and three groups of innate lymphoid cells (ILCs) (50, 51) and "innate-like lymphocytes" including γδT cells invariant NKT (iNKT) cells and mucosal-associated invariant T (MAIT) cells (52). Intracellular expression of CD3ε has been reported in CD4<sup>+</sup> ILC1 (53), and it is well known that γδT cells, iNKT, and MAIT cells express CD3ε as innate T cells. Accordingly, we suspect that some of "innate lymphocytes" or "innate-like lymphocytes" are included in CD3ε+CD4-1<sup>−</sup>CD8α− cells in spleen. In fish, however, no information is available on these newly discovered lymphocyte-like populations except NK cells. Further study on lymphocyte-like populations involved in innate immunity in fish is required to solve the problem.

In conclusion, CD4-1<sup>+</sup> and CD8α+ T cells express CD3ε mRNA but not molecule in the head- and trunk-kidney of

Figure 8 | Expression of CD3ε in re-injected lymphocytes in recipient tissues. Donor leukocytes were stained with CFSE, and the expression of CD3ε was analyzed by FACS before injection. Twenty four hours after injection, lymphocytes from recipient trunk-kidney, spleen, and peripheral blood leukocytes were gated on FS and SS dot plot. Histograms show the percentages of anti-CD3ε Ab and CFSE double-positive cells derived from donors. Dotted lines show negative control stained with isotype antibody and black solid lines with gray shadow show anti-gCD3ε Ab-positive cells. Mean value and ±SD of more than three independent experiments are shown. Bars indicate gating. CFSE-labeled donor leukocytes from kidney (A) or spleen (B) were injected to the recipients.

ginbuna suggesting the lack of surface expression of TCR/CD3 complex. CD4-1+ and CD8α+ T cells in the kidney become positive for the expression of CD3ε after 24 h in *in vitro* culture and kidney lymphocytes expressed CD3ε in the spleen of recipient when transferred into other individuals belonging to the same clone. These finding indicate that expression of CD3ε was suppressed in kidney and suggest that teleost kidney plays a role as the reservoir of resting mature T lymphocytes, although the precise mechanism of the suppression in CD3ε expression in fish kidney remain unknown.

#### ETHICS STATEMENT

All of the experiments described comply with the Guidelines of Nihon University Rules concerning Animal Care and Use and have been approved by the Nihon University Animal Care and Use Committee (No. AP12B014).

#### AUTHOR CONTRIBUTIONS

RM and NM performed all the experimental work, with help from YM, YS, and TY. TN and RM designed the experiments and wrote the main body of the paper, with contributions from YS.

#### FUNDING

This work was supported in part by a Grant-in-Aid for Scientific Research (B) (Grant Number 16H04984) from Japan Society for the Promotion of Science (JSPS).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.01321/ full#supplementary-material.

Figure S1 | Ginbuna CD3ε sequence. (A) Ginbuna CD3ε sequence. Nucleotide and amino acid sequence of ginbuna CD3ε are shown. Predicted signal peptide, extracellular domain, transmembrane region, and cytoplasmic domain are labeled, CXXC motif and ITAM are boxed. Amino acid numbers are at right. (B) Schematic illustration of gCD3ε. Ginbuna CD3ε can be divided into Ig-like domain, CXXC motif, transmembrane region, and ITAM. (C) Amino acid alignment of gCD3ε with Atlantic salmon (NM \_001123622) and human (NM \_000733.3) CD3ε sequences. The predicted signal peptide and domains are labeled. Residues similar/identical with gCD3ε are gray/black shade respectively. Ig-fold cysteine, CXXC motif and ITAM are boxed, and gaps (–) are indicated. Amino acid numbers are at right. (D) Comparison of ginbuna CD3ε with vertebrates CD3ε, CD3γ, CD3δ, and CD3ζ. Accession number of CD3 sequences are carp CD3ε (XM\_019126514.1), takifugu (ta)CD3ε (NM\_001037982.1), taCD3γ/δ (NM\_001037983.1), taCD3ζ (XM\_011608167.1), Japanese flounder (ja)CD3ε (XM\_020094967.1), jaCD3γ (XM\_020094974.1), jaCD3ζ (XM\_020112573.1), salmon (sa)CD3ε (NP\_001117094.1), saCD3δ (XM\_014162423.1), saCD3ζ (XM\_014164569.1), chicken CD3ε (NM\_206904.1),

mouse (mo)CD3ε (NM\_007648.4), moCD3γ (NM\_009850.2), moCD3δ, moCD3ζ (NM\_001113391.2), human (hu)CD3ε (NP\_000724.1), huCD3γ (EF444965.1), huCD3δ (EF444964.1), and huCD3ζ (AK128376.1).

FIGURE S2 | Immuno-precipitation and Western blot analysis. After immuno-precipitation of the thymus and spleen protein samples with anti-gCD3ε Ab, the proteins were detected by western blotting with anti-gCD3ε Ab. Both samples show some bands around 20–25 kDa expected to be gCD3ε and single band around 55 kDa expected to be heavy-chain.

FIGURE S3 | Specificity test of rabbit serum by immune-absorption. Absorption test of gCD3ε Ab was performed using transmembrane deletion mutant (TMDM) recombinant gCD3ε protein. Western blot analysis shows no band when antigCD3ε Ab was absorbed with antigen (TMDM rgCD3ε, right), while the Ab not absorbed with the antigen shows positive band (left).

FIGURE S4 | Protein sequencing by LC-MS/MS. Protein sequencing was determined using a protein band reactive with anti-gCD3ε Ab detected by Western blot. LC-MS/MS revealed 28 amino acid residues (gray highlight) and 16.1% of residues matched with gCD3ε amino acid sequence (A). Mass spectrum and fragmentation tables of each amino acid fragments are shown in (B,C), respectively. Peptide sequencing is indicated by matching b ion (red) and y ion (blue) fragments.

FIGURE S5 | Expression analysis of CD3ε in ginbuna tissues by RT-PCR. Total RNA was prepared from peripheral blood leukocytes (PBL), thymus, head-kidney, trunk-kidney, spleen, liver, ovary, intestine, skin, and gill tissues, and used for RT-PCR analysis. *ef*-*1a* was used as an internal control. Numbers to the right indicate PCR cycles.

FIGURE S6 | Gene expression analysis of T and B cell related genes in sorted CD3ε+ lymphocytes. Spleen cells were stained with anti-CD3ε Ab as described. Lymphocytes fraction from spleen were gated on FS and SS dot plot and anti-CD3ε Ab positive cells were sorted by FACS. Total RNA was prepared from 1×106 sorted cells and used for RT-PCR analysis. mRNA expression of *cd3e, cd4-, tcrb, lck*, and *igm* in sorted lymphocytes were shown. *ef-1a* was used as an internal control. Numbers to the right indicate PCR cycles.

FIGURE S7 | CD3ε expression in tissues of other cyprinid species. Spleen and kidney leukocytes from carp (A) and goldfish (B) were stained with anti-CD4-1 and CD8α mAbs followed by Alexa Fluor® 488 anti-rat IgG, and stained with anti-gCD3ε Ab or anti-hZAP-70 mAb followed by 647 goat anti-rabbit IgG. Lymphocytes were gated on FS and SS dot plot. Mean ± SD of more than three independent experiments are shown.

FIGURE S8 | Modulation of CD3ε expression. After allo-antigen stimulation (A) or *Edwardsiella tarda* infection (B), kidney lymphocyte was stained with anti-gCD3ε as described above and analyzed by FACS. Mean ± SD of more than three independent experiments are shown. Statistical significance was calculated using *t* tests to each gene (ns, not significant; *p* > 0.05).

FIGURE S9 | FS and SS dot plots of kidney and spleen leukocytes. Lymphocytes, myeloid cells, and granulocytes were gated on FSClow SSClow, FSChigh SSClow, and FSCmed SSChigh population, respectively (A). Lymphocytes from kidney and spleen were gated on FSClow SSClow population. The percentages of anti-CD3ε pAb positive cells were shown in the histogram (B).

FIGURE S10 | Migration of donor cells in recipient organs. CFSE-labeled donor cells were detected in recipient kidney, spleen, and peripheral blood leukocytes (PBL) on the histograms. Mean ± SD of more than three independent experiments are shown.

FIGURE S11 | Effect of *in vitro* culture on the leukocytes composition. Before *in vitro* (0 h) culture, leukocytes from kidney are composed of 46.2% of granulocytes, 11.1% of monocytes, and 39.3 % of lymphocytes. Similarly, after *in vitro* (24 h) culture, leukocytes from kidney are composed of with 50.0% of granulocytes, 7.6% of monocytes, and 32.2% of lymphocytes.

#### REFERENCES


hematopoietic tissue. *Blood* (2008) 111:1131–7. doi:10.1182/blood-2007- 08-104299


**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 Miyazawa, Murata, Matsuura, Shibasaki, Yabu and Nakanishi. 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.*

*Dustin M. E. Lillico, Joshua G. Pemberton and James L. Stafford\**

*Department of Biological Sciences, University of Alberta, Edmonton, AB, Canada*

#### *Edited by:*

*Brian Dixon, University of Waterloo, Canada*

#### *Reviewed by:*

*Maria Forlenza, Wageningen University & Research, Netherlands Annalisa Pinsino, Istituto di biomedicina e di immunologia molecolare Alberto Monroy (IBIM), Italy*

> *\*Correspondence: James L. Stafford stafford@ualberta.ca*

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

*Received: 15 February 2018 Accepted: 07 May 2018 Published: 28 June 2018*

#### *Citation:*

*Lillico DME, Pemberton JG and Stafford JL (2018) Selective Regulation of Cytoskeletal Dynamics and Filopodia Formation by Teleost Leukocyte Immune-Type Receptors Differentially Contributes to Target Capture During the Phagocytic Process. Front. Immunol. 9:1144. doi: 10.3389/fimmu.2018.01144*

Phagocytosis evolved from a fundamental nutrient acquisition mechanism in primitive unicellular amoeboids, into a dynamic and complex component of innate immunity in multicellular organisms. To better understand the cellular mechanisms contributing to phagocytic processes across vertebrates, our research has focused on characterizing the involvement of innate immune proteins originally identified in channel catfish (*Ictalurus punctatus*) called leukocyte immune-type receptors (IpLITRs). These unique teleost proteins share basic structural as well as distant phylogenetic relationships with several immunoregulatory proteins within the mammalian immunoglobulin superfamily. In the present study, we use a combination of live-cell confocal imaging and high- resolution scanning electron microscopy to further examine the classical immunoreceptor tyr osinebased activation motif (ITAM)-dependent phagocytic pathway mediated by the chimeric construct IpLITR 2.6b/IpFcRγ-L and the functionally diverse immunoreceptor tyrosine-based inhibitory motif-containing receptor IpLITR 1.1b. Results demonstrate that IpLITR 1.1b-expressing cells can uniquely generate actin-dense filopodia-like protrusions during the early stages of extracellular target interactions. In addition, we observed that these structures retract after contacting extracellular targets to secure captured microspheres on the cell surface. This activity was often followed by the generation of robust secondary waves of actin polymerization leading to the formation of stabilized phagocytic cups. At depressed temperatures of 27°C, IpLITR 2.6b/IpFcRγ-L-mediated phagocytosis was completely blocked, whereas IpLITR 1.1b-expressing cells continued to generate dynamic actin-dense filopodia at this lower temperature. Overall, these results provide new support for the hypothesis that IpLITR 1.1b, but not IpLITR 2.6b/ IpFcRγ-L, directly triggers filopodia formation when expressed in representative myeloid cells. This also offers new information regarding the directed ability of immunoregulatory receptor-types to initiate dynamic membrane structures and provides insights into an alternative ITAM-independent target capture pathway that is functionally distinct from the classical phagocytic pathways.

Keywords: phagocytosis, innate immunity, immunoregulatory receptors, signal transduction, actin cytoskeleton, comparative immunology, teleost fish

### INTRODUCTION

Phagocytosis is a vital innate immune response that involves the engulfment, destruction, and removal of extracellular targets; such as microbes, necrotic or apoptotic cells, and cellular debris (1). The unique ability of phagocytic cells to recognize and engulf large particulate targets depends on the surface expression of specialized immunoregulatory receptors. Well known mammalian phagocytic receptor-types include complement receptors (1–4), members of the Fc receptor (FcR) family (1–3), and dectin-1 (5). Studies using these model immune proteins have shown that phagocytosis is a multifaceted process that tightly regulates the active capture, ingestion, and subsequent destruction of various microbial targets (1–5). Phagocytic receptors relay their interactions with extracellular targets into dynamic filamentous (F)-actin remodeling events that reshape the plasma membrane through specialized intracellular signaling events (1, 6–9). Generally, each of these phagocytic pathways requires localized phospholipid metabolism and the engagement of actin nucleation and regulatory factors that link surface receptor activation with the cytoskeletal machinery to facilitate target engulfment (1, 6–9).

Interactions between phagocytic receptors and extracellular targets are not always reliant on passive binding events; rather, phagocytes actively increase the incidence of target-binding events through the formation of unique finger-like extensions of the plasma membrane called filopodia (10–13). These membrane protrusions are composed of un-branched filaments of polymerized F-actin that vary greatly in length (1–100 µm), thickness (0.1–0.3 µm), molecular composition, and geometry (10–13). Early studies using scanning electron microscopy (SEM) revealed that mouse peritoneal macrophages formed cord-like extensions that arose from the plasma membrane to tether extracellular targets to the cell surface (14). The formation of filopodia following bacterial lipopolysaccharide stimulation of macrophages was also shown to occur through the phosphorylation of various intracellular signaling mediators (15, 16). Following the initial contact with extracellular targets, filopodia quickly retract back toward the cell body, resulting in the immobilization and tethering of targets to the plasma membrane (12–15, 17, 18). This action allows for additional surface receptor–target interactions to occur that reinforce the transduction events responsible for the temporal activation of the phagocytic process (12, 13, 18). While the ability of phagocytes to actively deploy filopodia has been demonstrated, relatively little is known about the specific intracellular molecules and receptor-types that participate in filopodial dynamics within innate immune cells. Across eukaryotes, the formation of filopodia during diverse cellular events has been reported to require several classes of protein kinases and kinase-associated molecular scaffolds, small Rho-family GTPases [e.g., Ras-related C3 botulinum toxin substrate 1/2 (Rac1/2) and cell division control protein 42 homolog (Cdc42)], various cytoskeletal elements including components of the actin polymerization machinery (e.g., myosins and formins), as well as the generation of membrane-embedded phosphatidylinositol 3,4,5-trisphosphate (PtdIns(3,4,5)P3) (1, 7, 17–22). While some of the molecular constituents required for filopodia formation in mammalian macrophages have been described, much less is known about what specific cell-surface receptor-types trigger the formation of these structures and whether the ability of specific immunoregulatory receptor-types that control membrane dynamics are conserved in basal vertebrates.

Our research has progressively established channel catfish (*Ictalurus punctatus*) leukocyte immune-type receptors (IpLITRs) as a unique comparative model to understand mechanisms of immunoregulatory receptor-mediated control of innate cellular immunity in vertebrates. IpLITR-types are co-expressed by catfish myeloid and lymphoid cells (23, 24) and sub-types of these receptors have also recently been shown to serve as surface markers on subsets of cytotoxic T cells during viral infection in catfish (25). To date, IpLITRs have only been reported in the channel catfish (23, 24) and zebrafish (*Danio rerio)* (24, 26, 27) but they are likely not exclusive to these species as we have identified IpLITR-related sequences in several other fish species from the non-redundant protein sequence databases (*unpublished data*). Overall, IpLITRs share basic structural as well as distant phylogenetic relationships with several immunoregulatory proteins within the mammalian immunoglobulin superfamily (23) and they appear to be important regulators of innate cellular responses *via* classical as well as unique biochemical signaling networks (28–30). Although our functional characterization of IpLITR-types has relied on heterologous expression of teleost proteins in mammalian cells, this strategy has allowed us to demonstrate important conserved aspects regarding IpLITR-mediated immunoregulatory signaling events and revealed some unanticipated aspects regarding the versatility of IpLITR-mediated transduction. In particular, when expressed in the mammalian myeloid rat basophilic leukemia (RBL)-2H3 cell line, we have previously shown that IpLITR 2.6b/ IpFcRγ-L activates phagocytosis using a characteristic intracellular transduction response that is reminiscent of the prototypical mammalian immunoreceptor tyrosine-based activation motif (ITAM)-dependent FcR phagocytic pathway (29). Subsequently, we also described an alternative phagocytic mechanism mediated by an immunoreceptor tyrosine-based inhibitory motif (ITIM) containing receptor called IpLITR 1.1b. This unique IpLITR 1.1b-mediated mechanism exhibited reduced target engulfment, overall; but, alternatively, this receptor sub-type featured a significantly enhanced ability to capture extracellular beads (29). While the atypical pathway requires active engagement of the actin polymerization machinery, IpLITR 1.1b-expressing cells were insensitive to pharmacological inhibitors that blocked the classic signaling components of ITAM-dependent phagocytosis. Furthermore, the ability of IpLITR 1.1b-expressing RBL-2H3 cells

**Abbreviations:** FcR, Fc receptor; SEM, scanning electron microscopy; Rac1/2, ras-related C3 botulinum toxin substrate ½; Cdc42, cell division control protein 42 homolog; PtdIns(3,4,5)P3, phosphatidylinositol 3,4,5-trisphosphate; IpLITRs, channel catfish *Ictalurus punctatus* leukocyte immune-type receptors; RBL-2H3, rat basophilic leukemia-2H3; ITAM, mammalian immunoreceptor tyrosinebased activation motif; ITIM, immunoreceptor tyrosine-based inhibitory motif; LCI, live-cell imaging; GFP, green fluorescent protein; HA, hemagglutinin; CYT, cytoplasmic tail; MEM, minimal essential media; mAb, monoclonal antibody; PBS, phosphate buffered saline; BB, blue beads; Nck, non-catalytic region of tyrosine kinase adaptor protein; Wave2, N-WASp family verprolin-homologous protein-2; Src, proto-oncogene tyrosine-protein kinase Src; Vav, proto-oncogene Vav; Syk, spleen tyrosine kinase; PtdSer, phosphatidylserine; PSR, PtdSer receptor.

to capture beads was not affected at 27°C, an incubation temperature that completely inhibited IpLITR 2.6b/IpFcRγ-L phagocytosis (29). Imaging studies of fixed IpLITR 1.1b-expressing cells showed that the presence of this receptor produced extended plasma membrane structures that appeared to participate in the capture and tethering of microsphere targets to the cell surface; a phagocytic phenotype that was not observed fpr IpLITR 2.6b/ IpFcRγ-L-expressing cells during the phagocytic response (29). However, whether or not IpLITR 1.1b was directly controlling the formation of filopodia and how these structures were utilized for target interactions could not be deciphered from fixed cells.

In the present study, we utilized a combination of live-cell imaging (LCI) and high-resolution SEM to provide detailed new information regarding IpLITR-induced plasma membrane dynamics during the phagocytic process as well as explore the potential for receptor-selective filopodia formation. Our results support the hypothesis that F-actin-dense protrusions are indeed produced by IpLITR 1.1b-expressing cells. In addition, we show that during the early stages of the IpLITR 1.1b-mediated phagocytic process these filopodia-like structures often retract after target contact to secure the captured microspheres to the cell surface. This unique target-capturing phenotype is followed by the formation of phagocytic cup-like structures at the membrane interface and, in some cases, the eventual engulfment of the immobilized microspheres. At the reduced incubation temperature of 27°C we also show that although the membrane structures had repressed mobility, dynamic filopodia structures were still generated by IpLITR 1.1b-expressing cells, which continued to facilitate sustained cell–target interactions. Conversely, no F-actin dynamics or any associated membrane activity was seen in IpLITR 2.6b/IpFcRγ-L-expressing cells; likely due to an inability of this receptor to promote or maintain F-actin polymerization events below 37°C. Overall, results from these studies show that a unique IpLITR sub-type can selectively regulate filopodia formation over a range of incubation temperatures. This also reinforces the use of IpLITRs as an alternative vertebrate model for investigating the integration of immune cell membrane and cytoskeletal dynamics during the coordinate control of the phagocytic process.

## MATERIALS AND METHODS

#### Generation of Stable IpLITR-Expressing RBL-2H3 Cells

The transfection and selection of RBL-2H3 cells stably expressing N-terminal hemagglutinin (HA)-tagged IpLITRs in the pDisplay vector was performed as described (29, 30). IpLITR 2.6b/FcRγ-L is a chimeric receptor, which contains two extracellular Ig-like domains (GenBank Accession: ABI23577) fused with the ITAMcontaining cytoplasmic tail (CYT) region of the signaling adaptor IpFcRγ-L. This chimeric receptor construct was used to examine ITAM-mediated responses transmitted by the teleost adaptor IpFcRγ-L (30, 31). IpLITR 1.1b (GenBank Accession: ABI16050) encodes the full length TS32.17 L1.1b sequence and contains four extracellular Ig-like domains and a six tyrosine-containing CYT. Transfected RBL-2H3 cells were grown at 37°C and 5% CO2 in complete culture media [minimal essential media (MEM); Sigma-Aldrich, St. Louis, MO, USA] supplemented with Earl's balance salt solution (GE Healthcare, Baie d'Urfe, QC, Canada), 2 mM l-Glutamine (Life Technologies, Inc., Burlington, ON, Canada), 100 U/mL penicillin (Life Technologies, Inc.), 100 µg/mL streptomycin (Life Technologies, Inc.), 400 µg/mL G418 disulfate salt solution (Sigma-Aldrich, St. Louis, MO, USA), and 10% heat inactivated fetal bovine serum (Sigma-Aldrich). Surface expression of IpLITRs was monitored by flow cytometry using an αHA monoclonal antibody (mAb; Cedarlane Laboratories Ltd., Burlington, ON, Canada) as described previously (29, 30).

#### SEM of IpLITR-Mediated Phagocytosis

Scanning electron microscopy of IpLITR-mediated phagocytosis was performed using antibody-opsonized 4.5-µm microsphere targets. Briefly, 3 × 105 RBL-2H3 cells expressing either IpLITR 2.6b/IpFcRγ-L or IpLITR 1.1b were plated onto a sterile 18-mm diameter #1 1/2 circular coverslip (Electron Microscopy Sciences, Hatfield, PA, USA) and cultured overnight at 37°C with 5% CO2 in a six-well tissue culture plate (Fisher Scientific Company, Ottawa, ON, Canada). The following day cells were washed with phosphate buffered saline (PBS) and then incubated in phagocytosis buffer (1:1 mixture of 1× PBS containing 2 mg/mL bovine serum albumin, Sigma-Aldrich) and 1× Opti-MEM reduced serum medium (Fisher Scientific Company) containing 9 × 105 4.5-µm target microspheres (beads; Polybead® Carboxylate YG microspheres; Polysciences, Warrington, PA, USA) opsonized with 10 µg/mL of αHA mAb (Cedarlane Laboratories Ltd., Burlington, ON, Canada) or 10 µg/mL of the isotype control mouse IgG3 (Beckman Coulter, Mississauga, ON, Canada). Antibody opsonization was performed by absorbing them onto protein A precoated microspheres (isolated from *Staphylococcus aureus*; Sigma-Aldrich) as previously described (28, 29). Plates containing cells and target beads were then centrifuged at 1,500 rpm for 1 min to synchronize cell–bead interactions and then incubated for 1 h at either 27 or 37°C. For some experiments, the IpLITRexpressing cells were pre-treated for 1 h with 12.5 µM of the F-actin inhibitor Latrunculin B (EMD Millipore; Burlington, MA, USA) prior to their incubation with opsonized beads. Cells then were fixed with 2.5% glutaraldehyde/2% paraformaldehyde in a 0.1 M phosphate buffer solution. Dehydration of cells was then performed by sequential treatments with ethanol and hexamethyldisilazane according to previously described procedures (32, 33). After dehydration, coverslips were mounted onto round metal double-sided sticky stubs and coated with an ultrathin coating of gold/plutonium *via* a Hummer 6.2 Sputter Coater (Anatech USA, Hayward, CA, USA) and then imaged using a Philips/FEI XL30 SEM microscope (FEI: Hillsboro, OR, USA). Image analysis was performed using the Scandium 5.0 software (Emsis: Muenster, Germany).

## Generation of Stable LifeAct-GFP-Expressing RBL-2H3 Cells

Cytoskletal dynamics within IpLITR-expressing RBL-2H3 stable cell lines were examined using LifeAct-GFP (a generous gift from Dr. Nicholas Touret, University of Alberta), which is a C-terminal conjugated green fluorescent protein (GFP) probe that binds specifically to F-actin molecules (34). IpLITR-expressing RBL-2H3 cells were stably transfected with LifeAct-GFP using nucleofection (Amaxa Cell Line Nucleofector Kit T, RBL-2H3; Lonza, Cologne, Germany) according to the manufacture's recommended protocol. Briefly, IpLITR 2.6b/IpFcRγ-L- and IpLITR 1.1b-expressing RBL-2H3 cells were grown to confluence (~2.6 × 106 cells) in a six-well tissue culture plate, harvested, and then washed with PBS. Cells were then mixed with 100 µL of Cell Line Nucleofector Solution T (Amaxa) and 5 µg of LifeAct-GFP plasmid. Samples were transferred into a nucleofection cuvette and transfected using the Nucleofector II Device (Amaxa) using the program designated for the RBL-2H3 cell line (Program X-001). Cells were then placed into pre-warmed selection media (complete MEM supplemented with 400 µg of G418) and incubated at 37°C with 5% CO2 until confluent. Cells were then harvested and the GFP positive cells were sorted using a FACSCanto II (BD Bioscience). The GFP positive cells were then plated and incubated until they grew to confluence prior to being stained for IpLITR expression to verify that co-expression of LifeAct-GFP did not alter IpLITR surface expression levels.

#### LCI of IpLITR-Mediated Phagocytosis

Rat basophilic leukemia-2H3 cells stably co-expressing IpLITR and LifeAct-GFP (3 × 105 cells) were plated onto 50 mm μ-dishes (Ibidi; Madison, WI, USA) the day prior to imaging. The following day, cells were washed with PBS and then incubated in phagocytosis buffer containing either 9 × 105 αHA mAb-opsonized 4.5 µm non-fluorescent beads or αHA mAb-opsonized 4.5 µm blue beads (BB) (Fluoresbrite™ Carboxy BB microspheres; Polysciences, Warrington, PA, USA) and placed into a microscope stage chamber, which was supplied with 5% CO2 and heated to 37 or 27°C. Immediately after the addition of target beads, images were collected at 10-s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope (objective 60×, 1.3 oil plan-Apochromat; Munich, Germany) located at the Cross Cancer Institute Microscopy Facility (Faulty of Medicine & Dentistry; University of Alberta). Imaging data were analyzed using the Zen software package (2011; Carl Zeiss; Oberkochen, Germany) and ImageJ (ImageJ version 1.51p; Rasband, 1997–2017).

## RESULTS

#### IpLITR 2.6b/IpFcR**γ**-L-Mediated Phagocytosis

When incubated with isotype control IgG3 beads, no internalized or surface-bound targets were observed (**Figure 1A**). Comparatively, when incubated with αHA-opsonized beads (1 h at 37°C), IpLITR 2.6b/IpFcRγ-L-expressing RBL-2H3 cells displayed a characteristic flattened morphology and had multiple internalized beads with few surface-bound targets (**Figure 1B**). Pretreatment of the cells with a selective inhibitor of actin polymerization, Latrunculin B, abrogated bead internalization; although many of the beads remained associated with the cell surface (**Figure 1C**). Not surprisingly, cells treated with Latrunculin B also had a rounded morphology (compare **Figure 1B** with **Figure 1C**). When IpLITR 2.6b/IpFcRγ-L-expressing cells were incubated with target beads for 1 h at the reduced temperature of 27°C to inhibit phagocytosis, engulfment was indeed abrogated but the target microspheres appeared to be loosely associated with the plasma membrane (**Figure 1D**). Notably, when IpLITR 2.6b/IpFcRγ-L-expressing cells were incubated at 27°C they also had an overall rounded appearance. To capture additional representative images during the early stages of the IpLITR 2.6b/IpFcRγ-L phagocytic process, SEM was performed using cells incubated with the αHA-opsonized beads for shorter time periods at 37°C (i.e., 4, 8, 16, and 32 min). Characteristic stages of IpLITR 2.6b/IpFcRγ-L phagocytosis, beginning with cell-bead contacts through to complete target internalization are shown in **Figure 1E** (panels i-iv). During IpLITR 2.6b/IpFcRγ-L-mediated phagocytosis, phagocytic cup formation (**Figure 1E**; i, beads b1 and b2) occurs after initial contact of the target beads with the cell membrane. The cup progresses as extended pseudopod-like structures (**Figure 1E**; ii, b3 and b4) around the outer edges of the beads, which then continues over the beads (**Figure 1E**; iii, b5 and b6) until the targets are internalized (**Figure 1E**; iv).

#### IpLITR 2.6b/IpFcR**γ**-L-Mediated Phagocytosis at Different Temperatures

Scanning electron microscopy provided high-resolution static images of the IpLITR 2.6b/IpFcRγ-L-mediated phagocytic process. However, to observe the dynamic membrane remodeling events that occur during cell–target interactions required the use of real-time LCI. To achieve this, IpLITR 2.6b/IpFcRγ-Lexpressing RBL-2H3 cells stably expressing the fluorescent probe LifeAct-GFP were used. This allowed us to visualize and track distinct F-actin polymerization events (green) and associated membrane dynamics that occur starting from initial target contacts through to the engulfment of individual microspheres. Importantly, stable expression of LifeAct-GFP did not reduce the surface expression of IpLITRs (Figure S1 in Presentation 1 of Supplementary Material).

IpLITR 2.6b/IpFcRγ-L mediates the internalization of αHAopsonized target beads through a series of distinctive phases of F-actin dependent plasma membrane remodeling events. These F-actin polymerization dynamics are shown in a representative LCI time-lapse video (Video S1 in Presentation 2 of Supplementary Material) and in an associated series of time-stamped static images extracted from the LCI video (**Figure 2**). In Video S1 in Presentation 2 of Supplementary Material, both brightfield images merged with LifeAct-GFP (S1a) as well as the LifeAct-GFP signal alone (S1b) are displayed with the non-fluorescent microspheres clearly visible in the brightfield panels. For the time-stamps, both the brightfield-LifeAct-GFP merged views (top panels) and the LifeAct-GFP views alone (bottom panels) are shown with the location of the target microsphere indicated with a red asterisk. During the initial stages of IpLITR 2.6b/IpFcRγ-L-mediated phagocytosis, actin polymerization (green) is clearly visible at the cell surface-target interface in what appears as a phagocytic cup-like structure (**Figure 2**, asterisk, 0–40 s). As the phagocytic process proceeds, polymerization of F-actin is visible along the leading edges of extended pseudopods (**Figure 2**; 40–70 s). The accumulated F-actin behind the bead then depolymerizes as the pseudopods seal together and the microsphere sinks into the cell

(**Figure 2**; 120–140 s). Another representative LCI time-lapse video showing IpLITR 2.6b/IpFcRγ-L-mediated phagocytic behavior is provided in Video S2 in Presentation 2 of Supplementary Material with the associated time-stamped static images provided in Figure S2 in Presentation 1 of Supplementary Material.

Next, we compared the phagocytic activities of IpLITR 2.6b/ IpFcRγ-L-expressing cells incubated at 37 vs. 27°C. For these experiments, αHA-opsonized 4.5 µm BB were used as targets to allow for simultaneous visualization of both the target beads (blue) and F-actin dynamics (green). As shown in Video S3 in Presentation 2 of Supplementary Material and the accompanying time-stamped still images (**Figure 3A**; arrowhead), a microsphere target (blue) contacts the cell membrane (between 70–120 s) and is then progressively engulfed through a series of F-actin-mediated plasma membrane dynamic events similar to the temporal events described above. From initial contact to internalization, the entire process is completed in ~400 s at 37°C (**Figure 3A**). Note: although multiple beads are present in the LCI videos, for clarity we selected target beads that could be resolved from their initial contacts with the cell through to engulfment. An additional example of the IpLITR 2.6b/IpFcRγ-L-mediated phagocytosis process at 37°C is also shown in Video S4 in Presentation 2 of Supplementary Material with the time-stamped images shown in Figure S3 in Presentation 1 of Supplementary Material. In comparison, when IpLITR 2.6b/IpFcRγ-L-expressing RBL-2H3 cells were incubated with αHA-opsonized 4.5 µm BB at the inhibitory temperature of 27°C, no apparent target contacts or F-actin-mediated membrane dynamic events were observed.

This is shown in Video S5 in Presentation 2 of Supplementary Material and the associated time-stamped images in **Figure 3B** that spans 560 s starting from the time when the targets were first introduced to the cells. Although several beads were observed in the field of view, none of these targets establish contacts with the cell membrane over the duration of the video.

#### IpLITR 1.1b-Mediated Phagocytosis

Scanning electron microscopy was also performed to examine the IpLITR 1.1b-mediated phagocytic process (**Figure 4**). Similar to our observations for IpLITR 2.6b/IpFcRγ-L (**Figure 1**), when incubated with isotype control IgG3 beads, no internalized or surface-bound beads were observed (**Figure 4A**). However, when IpLITR 1.1b-expressing cells were incubated with αHA-opsonized beads (1 h at 37°C), SEM revealed that most of the targets appeared to be firmly secured to the cell surface but often they were not engulfed (**Figure 4B**). Pretreatment of the cells with the F-actin inhibitor Latrunculin B significantly altered cell morphology and also caused the beads to remain loosely tethered to the cell surface by disorganized plasma membrane structures (**Figure 4C**). When the IpLITR 1.1b-expressing cells were incubated with targets at 27°C for 1 h, the beads remained secured at the cell surface (**Figure 4D**), but the cells displayed a more rounded appearance compared to when they were incubated at 37°C. SEM imaging was also performed at 37°C for shorter time periods (i.e., 4, 8, 16, and 32 min) and **Figure 4E** shows representative images of the temporal stages of IpLITR 1.1b-mediated target interactions starting from initial cell-bead contacts (panels i–iv) through target capture and their eventual tethering to the cell membrane and occasional engulfment (panels v–viii). Specifically, during the early stages of target interactions most IpLITR 1.1b-expressing cells produced thin elongated membrane protrusions with beads tethered at their ends [**Figure 4E**; panels i-iv, beads (b1–b5)]. Some cells also generated thicker cellular extensions (**Figure 4E**; v, vi), which also participated in bead capture (b6–b9). We also consistently observed what appeared to be membrane ruffling (**Figure 4E**; vii), which contributed to the tethering of target beads to the plasma membrane (b10, b11). At later stages, stalled phagocytic cup-like structures (**Figure 4E**; viii) could be seen interacting with multiple beads (b12, b13) on the cell surface,

Figure 3 | Live-cell imaging of IpLITR 2.6b/IpFcRγ-L-mediated phagocytosis at different incubation temperatures. Rat basophilic leukemia-2H3 cells (3 × 105 ) stably co-expressing IpLITR 2.6b/IpFcRγ-L and LifeAct-GFP were incubated at 37°C (A) or at 27°C (B) with 9 × 105 αHA monoclonal antibody-coated 4.5 µm bright blue microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope (objective 60×, 1.3 oil plan-Apochromat; Munich, Germany). Representative time-stamps in (A) were extracted from Video S3 in Presentation 2 of Supplementary Material and the time-stamps in (B) were from Video S4 in Presentation 2 of Supplementary Material. In (A), the target microsphere of interest is indicated with an arrowhead.

Figure 4 | Scanning electron microscopy (SEM) of IpLITR 1.1b-mediated phagocytosis. IpLITR 1.1b-expressing rat basophilic leukemia-2H3 cells (3 × 105 ) were incubated at 37°C for 1 h with 9 × 105 IgG3-coated 4.5 µm microspheres (A) or with 9 × 105 αHA monoclonal antibody (mAb)-coated 4.5 µm microspheres (B) prior to imaging using a Philips/FEI XL30 SEM microscope (FEI: Hillsboro, OR, USA). Cells (3 × 105 ) were also pretreated for 1 h with 12.5 µM of the F-actin inhibitor Latrunculin B prior to their incubation with 9 × 105 αHA mAb-coated 4.5 µm microspheres (C) or incubated at 27°C for 1 h with 9 × 105 αHA mAb-coated 4.5 µm microspheres (D) prior to imaging. IpLITR 1.1b-expressing cells (3 × 105 ) were also incubated at 37°C for various times (e.g., 4, 8, 16, and 32 min) with 9 × 105 αHA mAb-coated 4.5 µm microspheres and representative SEM images of the progressive stages of IpLITR 1.1b-mediated phagocytosis are shown in panels i–viii (E). Specific beads are labeled (b1–b14) as described in the results section.

and occasionally targets that were almost completely surrounded by the plasma membrane (b14). Please note that **Figure 4E**, panel viii, is the same image shown in **Figure 4B**.

#### Temporal Examination of IpLITR 1.1b-Mediated Target Interactions

Using IpLITR 1.1b-expressing cells co-transfected with LifeAct-GFP, a variety of unique F-actin polymerization-dependent membrane remodeling events could be observed. Specifically, LCI imaging shows that IpLITR 1.1b-expressing cells generated F-actin-rich filopodia-like structures that extend out from the cell surface, attached to beads, and then rapidly withdrew back toward the cell membrane. For example, as shown in Video S6 in Presentation 2 of Supplementary Material and its associated timestamped static images (**Figure 5A**; target marked with an asterisk), a representative IpLITR 1.1b-expressing cell produces a thick actin-rich extension (green) that reaches out and attaches to the target bead (**Figure 5A**; 410–480 s). After ~10 s of contact with the target, the bead is then rapidly retracted back toward the cell surface, which correlates with the disappearance of the F-actin-rich extension as shown in the time-stamped panels at 480–490 s (**Figure 5A**). Following ~100 s of sustained contact between the target bead and the plasma membrane, a second F-actin-rich pseudopod-like extension (**Figure 5A**; 600 s) can be seen crawling up and then over the outer edge of bead until it returns toward the cell surface; momentarily wrapping the target in the plasma membrane (**Figure 5A**; 600–610 s). Subsequently, the pseudopod then retracts away from the bead before rapidly disappearing with the bead now tethered at the cell surface (**Figure 5A**; 660–690 s). An alternative mode of filopodia-mediated capture of targets displayed by IpLITR 1.1b-expressing RBL-2H3 cells is also shown in Video S7 in Presentation 2 of Supplementary Material and its associated time-stamped images (Figure S4 in Presentation 1 of Supplementary Material). Here, the formation of a thin F-actin containing membrane protrusion (green) rapidly extends out from the cell surface and makes contact with a bead (Figure S4 in Presentation 1 of Supplementary Material; 60–110 s). After initial contact with the target, the membrane protrusion rapidly retracts back toward the cell surface (110–140 s), thus pulling the bead toward the cell and tethering it to the membrane. This is then followed by the transient generation of actin-dense pseudopodlike structures that appear to surround the bead (Figure S4 in Presentation 1 of Supplementary Material; 150–170 s). Another representative cell–target interaction phenotype that we only observed for IpLITR 1.1b-expressing cells using LCI was the generation of an F-actin-rich extended membranous stalk (Video S8 in Presentation 2 of Supplementary Material) that formed after initial contact with the bead (see **Figure 5B**; target bead indicated with an asterisk; at 40 s). Following target contact, this extended stalk exhibited probing behavior for ~700 s during which time there were variable levels of F-actin polymerization observed along the edges and around the surface of the bead (**Figure 5B**; 40–740 s). The plasma membrane stalk also appeared to both elongate and thicken over the course of its contact with the bead and, unlike what we described earlier (**Figure 5A**), the target remained at a distance from the cell body as it was not retracted back toward the membrane surface during the duration of the video (Video S8 in Presentation 2 of Supplementary Material). Of note, this IpLITR 1.1b-expressing cell also appeared to contain one internalized bead as well as one tethered bead in addition to the target bead we have documented in the time-lapse.

While IpLITR 1.1b-expressing cells commonly generated and used extended membranous F-actin containing structures to capture and tether extracellular targets, other dynamic interaction behaviors were also observed and representatives of these are shown in the Videos S9–S11 in Presentation 2 of Supplementary Material and their associated time-stamped images in Figures S5–S7 in Presentation 1 of Supplementary Material. For instance, IpLITR 1.1b-expressing cells were capable of generating complex membranous ruffles (Video S9 in Presentation 2 of Supplementary Material). In this example two extracellular beads are captured by actin-dense membrane ruffles generated at the cell surface (Figure S5 in Presentation 1 of Supplementary Material; red and yellow asterisks 130–270 s). After being captured, the formation of a second F-actin-rich membrane ruffle is observed, which encapsulates one of the beads (Figure S5 in Presentation 1 of Supplementary Material; yellow asterisk; 290– 320 s) along the outer edge of the cell before depolymerizing as the bead is tethered to the cell surface (Figure S5 in Presentation 1 of Supplementary Material; 340 s). We also observed situations where a thin F-actin-dense protrusion captures a target bead at its outer most end (Video S10 in Presentation 2 of Supplementary Material). As the time-lapse progresses, polymerized actin accumulates around the bead as it is contracted down onto the cell surface (Figure S6 in Presentation 1 of Supplementary Material; 0–60 s). After this initial tethering, actin-dense membrane structures begin to surround the bead (Figure S6 in Presentation 1 of Supplementary Material; 110 s) initially from the left side and then from the right side of the target (Figure S6 in Presentation 1 of Supplementary Material; 170–200 s). Over the next 300 s, the F-actin depolymerizes leading to the apparent internalization of the bead (Figure S6 in Presentation 1 of Supplementary Material; 310–500 s). Finally, as shown in Video S11 in Presentation 2 of Supplementary Material; Figure S7 in Presentation 1 of Supplementary Material, multiple cell–bead interactions could be

Figure 5 | Continued

Figure 5 | Live-cell imaging of IpLITR 1.1b-mediated target interactions. Rat basophilic leukemia-2H3 cells (3 × 105 ) stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 37°C with 9 × 105 αHA monoclonal antibody-coated 4.5 µm microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope (objective 60×, 1.3 oil plan-Apochromat; Munich, Germany). Both the brightfield-LifeAct-GFP merged views (top panels) and the LifeAct-GFP views alone (bottom panels) are shown for two representative (A,B) IpLITR 1.1b-mediated target interactions with the location of the target microsphere indicated with an asterisk. Representative time-stamps in (A,B) were extracted from Videos S6 and S8 in Presentation 2 of Supplementary Material, respectively.

seen for an individual IpLITR 1.1b-expressing cell (Figure S7 in Presentation 1 of Supplementary Material; red, yellow, and orange asterisk). As the cell moves within the field of view from the top left, one bead had already begun to be internalized (Figure S7 in Presentation 1 of Supplementary Material; red asterisks; 160 s) and another bead was actively tethered to the cell surface (Figure S7 in Presentation 1 of Supplementary Material; yellow asterisks; 200–260 s). The generation of an F-actin-rich membrane protrusion then actively extended toward and contacted a third bead (Figure S7 in Presentation 1 of Supplementary Material; orange asterisks; 290–320 s). After this initial contact, a phagocytic-cuplike structure rapidly formed around the edges of the bead (Figure S7 in Presentation 1 of Supplementary Material; orange asterisks; 370 s), which was subsequently retracted back toward the cell surface (Figure S7 in Presentation 1 of Supplementary Material; 470 s). Overall, these results provide a representative summary of the diverse F-actin mediated plasma membrane remodeling events uniquely observed for IpLITR 1.1b- but not IpLITR 2.6b/ IpFcRγ-L-expressing cells. Notably, most IpLITR 1.1b-mediated target interactions involved the formation of membranous extensions as further described below.

### IpLITR 1.1b-Mediated Target Interactions at 27°C

After documenting several unique phagocytic phenotypes for IpLITR 1.1b-expressing RBL-2H3 cells, experiments were then performed to compare IpLITR 1.1b-mediated target interactions at 37 vs. 27°C (**Figure 6**). The rationale for these experiments was based on our previously reported ability of IpLITR 1.1b-expressing cells to facilitate target interactions at lower incubation temperatures (29). Here, again, αHA-opsonized 4.5 µm BB were used to allow for simultaneous visualization of target beads (blue) and F-actin dynamics (green). As shown in an LCI time-lapse video (Video S12 in Presentation 2 of Supplementary Material) and the accompanying still images (**Figure 6A**; arrowhead), at 37°C an extracellular bead is located at a distance from the cell membrane until a series of F-actin-rich membrane structures (green) extend toward (**Figure 6A**; 350–360 s) and then contacts the target (**Figure 6A**; 370–380 s). The bead is then rapidly pulled toward the cell membrane during which time distinct actin polymerization events appear to mediate extension of pseudopods around the entire bead (**Figure 6A**; 390–400 s). The actin-rich pseudopod immediately retracts away from the bead as evidenced by the gradual depolymerization of F-actin from around the outer surface of the bead (**Figure 6A**; 410–440 s). Over the remainder of the time series, the captured bead remains tethered on the cell surface but it is not engulfed (**Figure 6A**; 450–510 s). There are two other targets in the frame that are similarly captured by this IpLITR 1.1b-expressing RBL-2H3 cell at 37°C, and their interactions are documented separately in Videos S13 and S14 in Presentation 3 of Supplementary Material and their accompanying time-stamped images (Figures S8 and S9 in Presentation 1 of Supplementary Material, respectively). In an additional representative cell–target interaction phenotype, we observed an IpLITR 1.1b-expressing cells that continuously probed and then repeatedly attempted to pull a pre-tethered microspheres away from another cell. Specifically, as shown in Video S15 in Presentation 3 of Supplementary Material, a target bead has already been captured and is visible at the top left of the video. As the time series progresses, a second cell visible in the center of the video projects an F-actin containing membrane ruffle toward this target (**Figure 6B**, arrowhead; 130–180 s), which is already tethered to the other cell (**Figure 6B**, arrow; 130 s). After the ruffle makes initial contact with this bead, it appears to then withdraw back toward the cell leaving an extended membrane structure attached to the target with a detectable F-actin-rich area at the point of contact with the bead (**Figure 6B**; 190–260 s). Subsequently, following the cells initial failed attempt to pull the bead back toward the cell, a secondary F-actin-rich membrane ruffle projects out toward the secured target (**Figure 6B**; 270–300 s). As this ruffle subsides and the F-actin depolymerizes, the cell has again failed to retract the target toward its surface, although contact with this bead still remains (**Figure 6B**; 310–490 s). When viewed in its entirety, three separate F-actin-rich membrane ruffles are actively projected toward the target in what may be repeated attempts to capture the tethered target. Ultimately, these events leave the bead attached to two separate cells. Of note, this cell also appears to have engulfed two other targets as viewed in the bottom left of the video (**Figure 6B**; 430 s).

Unlike the inhibition of phagocytic responses we observed for IpLITR 2.6b/IpFcRγ-L at lower incubation temperatures, IpLITR 1.1b-expressing cells continued to display active target capture phenotypes at 27°C. However, at this lower incubation temperature the overall activity of the F-actin mediated membrane dynamics were markedly repressed. For example, as shown Video S16 in Presentation 3 of Supplementary Material and associated time-stamped images (**Figure 6C**), an extracellular target (arrowhead) is initially contacted by the cell (**Figure 6C**; 430 s), which promotes the extension of plasma membrane around the left side of the target (**Figure 6C**; 430–470 s). This appears to promote the formation of a thin F-actin-rich structure that extends beyond the surface of the attached bead and into the extracellular space (**Figure 6C**; 480–500 s). At this stage, the membrane protrusion appears to momentarily probe the environment before collapsing back toward the cell, which coincides with the disappearance of the F-actin signal (**Figure 6C**; 520–550 s). At the conclusion of this time series, the bead remains tethered at the cell surface but it was not

microscope (objective 60×, 1.3 oil plan-Apochromat; Munich, Germany). Representative time-stamps in (A,B) were extracted from Videos S12 in Presentation 2 and S15 in Presentation 3 of Supplementary Material, respectively, and the time-stamps in (C–E) were from Videos S16–S18 in Presentation 3 of Supplementary Material, respectively. In all time-stamps, target beads are indicated with an arrowhead and in (b; 130 s) a second cell with a pre-captured target bead is indicated with an arrow.

engulfed (**Figure 6C**; 590 s). Shown in Video S17 in Presentation 3 of Supplementary Material and **Figure 6D** (time-stamped images) is another example at 27°C where IpLITR 1.1b-expressing cells first use an F-actin containing membrane ruffle to attach to a bead, which is then followed by the formation of a thin F-actin containing membrane structure that partially surrounds the tethered target. In Video S18 in Presentation 3 of Supplementary Material and **Figure 6E** (time-stamped images) a target bead is observed coming into close contact with an extended membrane protrusion, after which it is then partially engulfed as it becomes surrounded by a thin but F-actin-dense membrane structure. Additional examples of the unique activities of IpLITR 1.1b-expressing RBL-2H3 cells at 27°C are provided in Videos S19 and S20 in Presentation 3 of Supplementary Material and their accompanying time-stamped images (Figures S10 and S11 in Presentation 1 of Supplementary Material, respectively).

## DISCUSSION

Using high-resolution SEM and real-time LCI, the results of this study provide new evidence regarding IpLITR-mediated production of dynamic cytoskeletal and membrane remodeling events. In particular, we show for the first time that IpLITR 1.1b-expressing cells uniquely generate filopodia-like extensions composed of long filaments of polymerized F-actin and reveal that these plasma membrane structures are actively used for extracellular target binding and capture. Considering that very little is known regarding the ability of immunoregulatory receptor-types to induce filipodia formation in other vertebrates, including mammals, our functional studies described here set the stage for future studies targeted at understanding how the dynamic control of intracellular transduction events controlled by IpLITR 1.1b selectively contribute to diverse innate cell effector responses across vertebrates. That being said, it is important to note that our results were obtained using heterologous expression of fish immunoregulatory proteins in a mammalian cell system. Although this strategy is not directly able to inform us about the actual *in vivo* activities of IpLITRs, it is clear from our studies that these receptors can potently regulate various innate cellular responses including the selective induction of filopodia for extracellular target capture. It is likely that IpLITR-mediated responses in mammalian cells feature similar signaling components that would be present in representative fish immune cell-types. Further exploration of these mechanisms will assist in uncovering the functional versatility for ITIM- and ITAM-encoding receptors that can eventually be used to explore teleost immunoregulatory receptor networks in homologous systems.

Filopodia are dynamic membrane structures that can vary in length and thickness but rely on the cytoskeletal machinery and actin-binding proteins for their formation (10–13, 19). Filopodia also play many important physiological roles in health and disease and have been shown to participate in a range of cellular processes including cell migration, morphogenesis, neurite outgrowth, metastasis, and wound healing (10–13, 19, 35–37). For example, within the immune system, SEM analysis has shown that prior to the initiation of phagocytosis, bacteria are tethered to phagocyte surfaces by long and thin membranous protrusions (11–14, 17, 18). Importantly, these plasma membrane extensions provide phagocytes with the ability to dynamically explore their extracellular environments as the rapid elongation and subsequent retraction of filopodia assists in the active capture of microbes by increasing the functional radius available for pathogen contact beyond the circumference of the cell (1, 7, 11–14, 18). While this shows that immune cells can actively deploy filopodia to capture targets, very little is known regarding the specific receptor-types and associated intracellular dynamics that participate in the formation and regulation of these membrane structures. Previously, we have reported that IpLITR 1.1b-expressing RBL-2H3 cells displayed a unique target acquisition and engulfment phenotype associated with the formation of extended membranous protrusions (29). The results of the present study further support a role for this specific immunoregulatory receptor-type in the control of cytoskeletal dynamics and filopodia formations during the initial contact with and then capture of extracellular targets over a range of temperatures. Taken together, these findings show, for the first time, that active capture and tethering of extracellular targets to the cell surface might represent a conserved function for certain members of the IpLITR family *via* their unique ability to transmit signals that affect F-actin polymerization and associated plasma membrane dynamics. Comparatively, for IpLITR 2.6b/IpFcRγ-L-expressing cells, filopodia-like structures were not specifically used to capture targets, as sustained contact time between the plasma membrane and microspheres was required to trigger the IpLITR 2.6b/IpFcRγ-L-mediated phagocytic process. In addition, phagocytic activity and membrane dynamics were both completely abolished at 27°C in IpLITR 2.6b/IpFcRγ-L-expressing cells; likely due to an inability of IpLITR 2.6b/IpFcRγ-L to promote or facilitate F-actin polymerization events at temperatures below 37°C in RBL-2H3 cells.

Filopodia are regulated by mechanisms instigated in part by constitutive intracellular signaling events that involve a number of conserved transduction molecules (e.g., docking protein 1, non-catalytic region of tyrosine kinase adaptor protein 1 (Nck), neural Wisskot–Aldrich syndrome protein, N-WASp family verprolin-homologous protein-2 (Wave2), inverse-BAR protein insulin receptor substrate protein of 53 kDa, Cdc42, formins, fascin, and myosins) (10, 19, 38–44). Overall, constitutively generated filopodia allow phagocytes to constantly probe their extracellular environments as these membranous probes also contain phagocytic receptors located along their edges, a process that depends on an unknown mechanism for the loading of phagocytic receptors into the protrusions (1, 18, 22). Any stochastic phagocytic receptor–target interactions that may occur would facilitate the attachment of specific targets to the extended membranes. Extracellular targets would then be pulled back toward the cell surface during filopodial retractions due to the retrograde flow of actin back toward the cell body and the contractile forces generated by myosins (10). Once in close contact with the plasma membrane, newly established target–receptor interactions could activate additional intracellular signaling pathways to reinforce tethering or subsequently trigger target engulfment (7, 18).

While constitutively generated filopodia facilitates continuous sampling of the environment by phagocytes, it has been shown that these structures are also produced in responses to specific stimuli. For example, lipopolysaccharide-induced activation of toll-like receptor 4 increases the production of filopodia-like structures (15, 16). Receptor-induced filopodia formation has also been characterized in cancer cells, which use these inducible pathways to produce invasive membrane protrusions (36, 37). Termed invadopodia, these extensions are formed by the selective stimulations of tumor cell-expressed platelet-derived growth factor and epidermal growth factor (36). Upon growth factor stimulation, the local recruitment and activation of kinases, such as focal adhesion kinase and proto-oncogene tyrosine-protein kinase Src (Src), occurs early in the process of invadopodia formation that initiate phosphorylation of downstream signaling proteins (36, 37). Unlike the receptor-specific production of filopodia or invadopodia described above, IpLITR 1.1b-expressing cells were not stimulated by any known endogenous ligands. Therefore, it seems that the stable expression of IpLITR 1.1b alone was sufficient to support filopodia generation. These results suggest that IpLITR 1.1b can uniquely network with intracellular components requisite for the production of F-actin containing filopodia-like structures. Previously, we hypothesized that IpLITR 1.1b-controlled signaling events induce formation of macromolecular complexes with its CYT that pre-assemble prior to receptor engagement; effectively priming the receptor for subsequent interactions with extracellular targets (28–30). Preassociations of IpLITR 1.1b with intracellular effectors capable of modulating the cytoskeletal machinery would allow for dynamic membrane remodeling events prior to the formation of stable receptor–ligand interactions (28–30).

Our proposed mechanism for target acquisition and engulfment pathways facilitated by IpLITR 1.1b have been described in detail elsewhere (28, 29, 45) and we hypothesized that this likely requires the differential participation of the proximal and distal regions of its CYT in the recruitment and activation of select intracellular effectors (45, 46). Specifically, for IpLITR 1.1b to constitutively trigger filopodia formation in RBL-2H3 cells without ligand engagements, this receptor may exist in a primed state, facilitating its basal coupling to effectors of actin dynamics. In this model, Nck serves as a cytosolic adaptor that could couple surface expressed IpLITR 1.1b with the intracellular effector Wave2. Basal recruitment and activation of a cytoplasmic guanine nucleotide exchange factor such a proto-oncogene Vav (Vav) family proteins would then activate Rho family GTPases (47), which may then activate F-actin polymerization *via* the Nckassociated Wave2 complex to trigger the constitutive formation of filopodia. Interestingly, this model closely aligns with the short-circuited phagocytic pathway recently described for human carcinoembryonic antigen-related cell adhesion molecule (48). In addition, we have reported that IpLITR 1.1b-mediated activity is partially dependent upon the catalytic activity of Src and the spleen tyrosine kinase (Syk) (29). Although yet to be confirmed, sustained activation of these kinases in the absence of agonist stimulation would likely require pre-aggregation of IpLITR 1.1b on the cell surface. This would maintain basal Srcdependent tyrosine phosphorylation of the IpLITR 1.1b CYT region facilitating constitutive coupling of IpLITR 1.1b to select components of the cytoskeletal machinery. Importantly, we also recently showed that Nck is recruited to a consensus interaction motif located in the proximal CYT region of IpLITR 1.1b (45), which would directly bridge IpLITR 1.1b with the Wave2 complex. In mammalian cells, activation of Wave2 requires statespecific phosphorylation as well as interactions with GTP-bound Rho superfamily proteins, most commonly Rac (49). As a result, the assembly of the Nck-Wave2 complex within the proximal CYT region of IpLITR 1.1b would most likely be coupled to the recruitment of cytoplasmic guanine nucleotide exchange factors, including Vav. Our recent biochemical studies showed that Syk is preferentially recruited to the distal region of the IpLITR 1.1b CYT (45). Therefore, we suspect that recruitment and activation of Vav by Syk would provide the necessary catalyst for Rac 1/2 activation and the stimulation of actin-driven membrane protrusions *via* the Nck-recruited Wave2. Overall, this predicted model encompasses the minimal machinery required for a constitutive IpLITR 1.1b-dependent deployment of filopodia in the absence of agonist stimulation and is supported by our recent biochemical studies (27, 29, 45). Future work is required to formally establish functional roles for Nck, Syk, Vav, Rac 1/2, and Wave2 during IpLITR 1.1b-mediated triggering responses including filopodia formation. Finally, if IpLITR 1.1b is indeed basally phosphorylated and pre-associated with intracellular components linking it to F-actin dynamics, then this would in part explain why IpLITR 1.1b continues to capture targets at suboptimal incubation temperatures due to pre-assembly of these components with the receptor. The reduced plasma membrane dynamics for IpLITR 1.1b-expressing RBL-2H3 cells at 27°C are likely due to specific affects on phospholipid dynamics and membrane mobility at this lower temperature, but likely not from an inability of IpLITR 1.1b to associate with signaling complexes, which would have previously occurred prior to the cooling of the cells.

Following filopodia-mediated capture of extracellular targets, we frequently observed the generation of secondary waves of F-actin polymerization after the target was secured at the cell surface. These events may be triggered by aggregations of IpLITR 1.1b at the newly established contact sites formed between the plasma membrane and the captured target. In some cases, the immobilized targets remained firmly tethered on the cell surface and occasionally the beads were completely internalized. This phenotype is reminiscent of efferocytosis, a process responsible for phagocyte-mediated clearance of apoptotic bodies through the recognition of phosphatidylserine (PtdSer) on dying cells (50, 51). However, unlike linear filopodia that extend perpendicular to the cell surface, efferocytosis typically involves membrane dynamics that form extended but laterally moving arcs or wave-like structures that flow along the cell surface (50–52). Functionally, these structures reach out into the extracellular space to make contact with apoptotic cells and their sweeping motion augments trapping of distant targets. This brings dying cells into close proximity to the plasma membrane, where they are tethered and eventually cleared by secondary activated phagocytic processes (50, 51). Rac 1/2, Cdc42, and Wave2 have all been identified as key players during the control of efferocytosis (50, 51, 53, 54), which occurs in two discrete receptor-specific steps known as the tethering and tickling (50, 53, 55, 56) that participate in the step-wise capture and engulfment of apoptotic bodies. For example, engagement of receptors for identifying apoptotic cells, including CD36, CD14, CD68, ανβ3, and ανβ5, promotes the tethering of specific targets on macrophages (56). Uptake of targets then occurs when tethering receptors are co-engaged with the phagocytic PtdSer receptor (PSR) (56). Interestingly, incubation of the cells with PtdSer-coated erythrocytes was insufficient for both tethering and phagocytic uptake by the PSR; indicating that both tethering and phagocytic signals are required for effective apoptotic cell removal (56). In agreement with this dual mode for target capture and engulfment, our observations support a model that involves constitutive mechanisms for IpLITR 1.1b-mediated deployment of filopodia to tether targets to the cell surface. Subsequently, captured targets can trigger additional IpLITR 1.1b-dependent pathways, which may be distinct from the constitutive mechanism that regulates resting filopodia production. One example includes a CYT proximal-specific pathway involving the formation of a heterotrimeric complex consisting of growth factor receptorbound protein, growth factor receptor-bound protein-associated binding protein 2 and phosphoinositide 3-kinases (57–60) that recruits Vav to activate Rac 1/2 and then trigger the actin-related protein 2/3-dependent actin protrusions *via* the Wave complex. Unlike constitutive filopodia induction, this model would be distinct from basal Nck-mediated recruitment of the actin regulatory Wave2 complex and could be achieved *via* the localized production of phosphatidylinositol 3,4,5-trisphosphate (61). A second example suggests that the phosphorylation of protein tyrosine phosphatases at the C-terminal tyrosine residue Y542 may form a cryptic ITAM in concert with a neighboring IpLITR 1.1b present at the site of bead contact to recruit phosphorylated protein tyrosine phosphatases. This mechanism would be similar to the recently discovered pathway described for dectin-1 (62) and would require only the distal segment of IpLITR 1.1b CYT. Future studies are still required to decipher the specific mechanisms underlying the variable signaling events that control IpLITR 1.1b-mediated regulation of target capture, tethering, and engulfment. However, the results of this study combined with our previous biochemical recruitment experiments provide the necessary framework for deciphering how IpLITR 1.1b variably controls the actin polymerization machinery.

Taken together, our results show that the expression of IpLITR 1.1b, but not IpLITR 2.6b/IpFcRγ-L, specifically triggers RBL-2H3 cells to induce filopodia formation in the absence of any known immune stimuli. The receptor-specific nature of IpLITR 1.1b-indcued filopodia is clearly evident when both IpLITRexpressing cell-types were incubated at depressed temperatures. This also appears to be the first study to suggest that expression of a specific immunoregulatory receptor can promote the constitutive formation of filopodia without the need for an exogenous ligand. IpLITR 1.1b-mediated signaling also initiates secondary waves of actin polymerization events that are associated with the internalization and membrane tethering of extracellular targets, which we propose to be a distinct event from those involved in the initial generation of filopodia. These responses are likely due to the unique structure and signaling potential associated with the IpLITR 1.1b CYT; thereby allowing for diversity in the integrated control of cytoskeletal and membrane remodeling associated with IpLITR 1.1b expression. Overall, our results offer novel information regarding the ability of immunoregulatory receptors to initiate filopodia formation and provide new insights into the temporal organization of cellular events surrounding the unique transduction dynamics that regulate F-actin polymerization and membrane remodeling events.

#### AUTHOR CONTRIBUTIONS

JS and DL conceived and designed the study. DL performed the experimental procedures. DL, JP, and JS analyzed the data, wrote the manuscript, and reviewed the manuscript.

#### FUNDING

This work was supported by grants from; the Natural Sciences and Engineering Council of Canada (NSERC; grant# RGPIN-2012-341209) awarded to James Stafford; graduate teaching assistantship awarded by the Department of Biological Sciences to Dustin Lillico; an NSERC PGS-D, Alberta Innovates Health Solutions Graduate Studentship, Honorary Izaak Walter Killam Memorial Scholarship, and University of Alberta Dissertation Fellowship awarded to Joshua Pemberton.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.01144/ full#supplementary-material.

Figure S1 | Flow cytometric analysis of IpLITR surface expression levels in rat basophilic leukemia-2H3 (RBL-2H3) cells co-expressing LifeAct-GFP. IpLITR 2.6b/IpFcRγ-L (A) and IpLITR 1.1b (B)-expressing RBL-2H3 stables were transfected with LifeAct-GFP by nucleofection using the Amaxa Cell Line Nucleofector Kit T, RBL-2H3. IpLITR surface expression (FL-2) was determined by staining the cells with αHA monoclonal antibody (mAb) and a PE-conjugated secondary goat αmouse polyclonal. LifeAct-GFP expression was examined using the FL-1 intensity. Each histogram shows the staining profiles for cells stained with an IgG3 isotype control antibody (solid line), IpLITR-expressing cells stained with αHA mAb (untransfected; dotted line), and IpLITR-expressing cells co-transfected with LifeAct-GFP and stained with αHA mAb (LifeAct-GFP Transfected; dashed line).

Figure S2 | Live-cell imaging of IpLITR 2.6b/IpFcRγ-L-mediated phagocytosis. Rat basophilic leukemia-2H3 cells (3 × 105 ) stably co-expressing IpLITR 2.6b/ IpFcRγ-L and LifeAct-GFP were incubated at 37°C with 9 × 105 αHA monoclonal antibody-coated 4.5 µm microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope (objective 60×, 1.3 oil plan-Apochromat; Munich, Germany). Both the brightfield-LifeAct-GFP merged views (top panels)

and the LifeAct-GFP views alone (bottom panels) are shown with the location of the target microsphere indicated with an asterisk. Representative time-stamps were extracted from Video S2 in Supplementary Material.

Figure S3 | Live-cell imaging of IpLITR 2.6b/IpFcRγ-L-mediated phagocytosis. Rat basophilic leukemia-2H3 cells (3 × 105 ) stably co-expressing IpLITR 2.6b/ IpFcRγ-L and LifeAct-GFP were incubated at 37°C with 9 × 105 αHA monoclonal antibody-coated 4.5 µm bright blue microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope (objective 60×, 1.3 oil plan-Apochromat; Munich, Germany). Representative time-stamps were extracted from Video S4 in Supplementary Material.

Figure S4 | Live-cell imaging of IpLITR 1.1b-mediated target interactions. Rat basophilic leukemia-2H3 cells (3 × 105 ) stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 37°C with 9 × 105 αHA monoclonal antibodycoated 4.5 µm microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope (objective 60×, 1.3 oil plan-Apochromat; Munich, Germany). Both the brightfield-LifeAct-GFP merged views (top panels) and the LifeAct-GFP views alone (bottom panels) are shown with the location of the target microsphere indicated with an asterisk. Representative time-stamps were extracted from Video S7 in Supplementary Material.

Figure S5 | Live-cell imaging of IpLITR 1.1b-mediated target interactions. Rat basophilic leukemia-2H3 cells (3 × 105 ) stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 37°C with 9 × 105 αHA monoclonal antibodycoated 4.5 µm microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope (objective 60×, 1.3 oil plan-Apochromat; Munich, Germany). Both the brightfield-LifeAct-GFP merged views (top panels) and the LifeAct-GFP views alone (bottom panels) are shown with the locations of two target microspheres indicated with asterisks. Representative time-stamps were extracted from Video S9 in Supplementary Material.

Figure S6 | Live-cell imaging of IpLITR 1.1b-mediated target interactions. Rat basophilic leukemia-2H3 cells (3 × 105 ) stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 37°C with 9 × 105 αHA monoclonal antibodycoated 4.5 µm microspheres. Immediately after the addition of target beads, images were collected at 10 sec intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope (objective 60×, 1.3 oil plan-Apochromat; Munich, Germany). Both the brightfield-LifeAct-GFP merged views (top panels) and the LifeAct-GFP views alone (bottom panels) are shown with the location of the target microsphere indicated with an asterisk. Representative time-stamps were extracted from Video S10 in Supplementary Material.

Figure S7 | Live-cell imaging of IpLITR 1.1b-mediated target interactions. Rat basophilic leukemia-2H3 cells (3 × 105 ) stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 37°C with 9 × 105 αHA monoclonal antibodycoated 4.5 µm microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope (objective 60×, 1.3 oil plan-Apochromat; Munich, Germany). Both the brightfield-LifeAct-GFP merged views (top panels) and the LifeAct-GFP views alone (bottom panels) are shown with the locations of three target microspheres indicated with asterisks. Representative time-stamps were extracted from Video S11 in Supplementary Material.

Figure S8 | Live-cell imaging of IpLITR 1.1b-mediated target interactions. Rat basophilic leukemia-2H3 cells (3 × 105 ) stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 37°C with 9 × 105 αHA monoclonal antibodycoated 4.5 µm bright blue microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope (objective 60×, 1.3 oil plan-Apochromat; Munich, Germany). Target bead of interest is indicated with an arrowhead. Representative time-stamps were extracted from Videos S13 in Supplementary Material.

Figure S9 | Live-cell imaging of IpLITR 1.1b-mediated target interactions. Rat basophilic leukemia-2H3 cells (3 × 105 ) stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 37°C with 9 × 105 αHA monoclonal antibodycoated 4.5 µm bright blue microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope (objective 60×, 1.3 oil plan-Apochromat;

Munich, Germany). Target bead of interest is indicated with an arrowhead. Representative time-stamps were extracted from Videos S14 in Supplementary Material.

Figure S10 | Live-cell imaging of IpLITR 1.1b-mediated target interactions at 27°C. Rat basophilic leukemia-2H3 cells (3 × 105 ) stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 27°C with 9 × 105 αHA monoclonal antibody-coated 4.5 µm bright blue microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope (objective 60×, 1.3 oil plan-Apochromat; Munich, Germany). Target bead of interest is indicated with an arrowhead. Representative time-stamps were extracted from Videos S19 in Supplementary Material.

Figure S11 | Live-cell imaging of IpLITR 1.1b-mediated target interactions at 27°C. Rat basophilic leukemia-2H3 cells (3 × 105 ) stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 27°C with 9 × 105 αHA monoclonal antibody-coated 4.5 µm bright blue microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope (objective 60×, 1.3 oil plan-Apochromat; Munich, Germany). Target bead of interest is indicated with an arrowhead. Representative time-stamps were extracted from Videos S20 in Supplementary Material.

Video S1 | Live-cell imaging of IpLITR 2.6b/IpFcRγ-L-mediated phagocytosis. Rat basophilic leukemia-2H3 cells stably co-expressing IpLITR 2.6b/IpFcRγ-L and LifeAct-GFP were incubated at 37°C with αHA monoclonal antibody-coated 4.5 µm microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope. Both the brightfield-LifeAct-GFP merged views (A) and the LifeAct-GFP views alone (B) are shown. Target bead of interest is located within the box.

Video S2 | Live-cell imaging of IpLITR 2.6b/IpFcRγ-L-mediated phagocytosis. Rat basophilic leukemia-2H3 cells stably co-expressing IpLITR 2.6b/IpFcRγ-L and LifeAct-GFP were incubated at 37°C with αHA monoclonal antibody-coated 4.5 µm microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope. Both the brightfield-LifeAct-GFP merged views (A) and the LifeAct-GFP views alone (B) are shown. Target bead of interest is located within the box.

Video S3 | Live-cell imaging of IpLITR 2.6b/IpFcRγ-L-mediated phagocytosis. Rat basophilic leukemia-2H3 cells stably co-expressing IpLITR 2.6b/IpFcRγ-L and LifeAct-GFP were incubated at 37°C with αHA monoclonal antibody-coated 4.5 µm bright blue microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope. Target bead of interest is located within the box.

Video S4 | Live-cell imaging of IpLITR 2.6b/IpFcRγ-L-mediated phagocytosis. Rat basophilic leukemia-2H3 cells stably co-expressing IpLITR 2.6b/IpFcRγ-L and LifeAct-GFP were incubated at 37°C with αHA monoclonal antibody-coated 4.5 µm bright blue microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope. Target bead of interest is located within the box.

Video S5 | Live-cell imaging of IpLITR 2.6b/IpFcRγ-L-mediated phagocytosis at 27o C. Rat basophilic leukemia-2H3 cells stably co-expressing IpLITR 2.6b/ IpFcRγ-L and LifeAct-GFP were incubated at 27°C with αHA mAb-coated 4.5 µm bright blue microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope.

Video S6 | Live-cell imaging of IpLITR 1.1b-mediated target interactions. Rat basophilic leukemia-2H3 cells stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 37°C with αHA monoclonal antibody-coated 4.5 µm microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope. Both the brightfield-LifeAct-GFP merged views (A) and the LifeAct-GFP views alone (B) are shown. Target bead of interest is located within the box.

Video S7 | Live-cell imaging of IpLITR 1.1b-mediated target interactions. Rat basophilic leukemia-2H3 cells stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 37°C with αHA monoclonal antibody-coated 4.5 µm microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope. Both the brightfield-LifeAct-GFP merged views (A) and the LifeAct-GFP views alone (B) are shown. Target bead of interest is located within the box.

Video S8 | Live-cell imaging of IpLITR 1.1b-mediated target interactions. Rat basophilic leukemia-2H3 cells stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 37°C with αHA monoclonal antibody-coated 4.5 µm microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope. Both the brightfield-LifeAct-GFP merged views (A) and the LifeAct-GFP views alone (B) are shown. Target bead of interest is located within the box.

Video S9 | Live-cell imaging LCI of IpLITR 1.1b-mediated target interactions. Rat basophilic leukemia-2H3 cells stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 37°C with αHA monoclonal antibody-coated 4.5 µm microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope. Both the brightfield-LifeAct-GFP merged views (A) and the LifeAct-GFP views alone (B) are shown. Target beads of interest are located within the box.

Video S10 | Live-cell imaging of IpLITR 1.1b-mediated target interactions. Rat basophilic leukemia-2H3 cells stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 37°C with αHA monoclonal antibody-coated 4.5 µm microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope. Both the brightfield-LifeAct-GFP merged views (A) and the LifeAct-GFP views alone (B) are shown. Target bead of interest is located within the box.

Video S11 | Live-cell imaging of IpLITR 1.1b-mediated target interactions. Rat basophilic leukemia-2H3 cells stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 37°C with αHA monoclonal antibody-coated 4.5 µm microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope. Both the brightfield-LifeAct-GFP merged views (A) and the LifeAct-GFP views alone (B) are shown. Note: five target beads are interacted with by the cell over the duration of the video.

Video S12 | Live-cell imaging of IpLITR 1.1b-mediated phagocytosis. Rat basophilic leukemia-2H3 cells stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 37°C with αHA monoclonal antibody-coated 4.5 µm bright blue microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope. Target bead of interest is located within the box.

Video S13 | Live-cell imaging of IpLITR 1.1b-mediated phagocytosis. Rat basophilic leukemia-2H3 cells stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 37°C with αHA monoclonal antibody-coated 4.5 µm bright blue microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope. Target bead of interest is located within the box.

Video S14 | Live-cell imaging of IpLITR 1.1b-mediated phagocytosis. Rat basophilic leukemia-2H3 cells stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 37°C with αHA monoclonal antibody-coated 4.5 µm bright blue microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope. Target bead of interest is located within the box.

Video S15 | Live-cell imaging of IpLITR 1.1b-mediated phagocytosis. Rat basophilic leukemia-2H3 cells stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 37°C with αHA monoclonal antibody-coated 4.5 µm bright blue microspheres. Immediately after the addition of target beads, images were

collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope. Target bead of interest is located within the box, which was pre-captured by another cell that is visible in the top right corner of the video.

Video S16 | Live-cell imaging of IpLITR 1.1b-mediated phagocytosis at 27o C. Rat basophilic leukemia-2H3 cells stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 27°C with αHA monoclonal antibody-coated 4.5 µm bright blue microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope. Target bead of interest is located within the box.

Video S17 | Live-cell imaging of IpLITR 1.1b-mediated phagocytosis at 27°C. Rat basophilic leukemia-2H3 cells stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 27°C with αHA monoclonal antibody-coated 4.5 µm bright blue microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope. Target bead of interest is located within the box.

Video S18 | Live-cell imaging of IpLITR 1.1b-mediated phagocytosis at 27°C. Rat basophilic leukemia-2H3 cells stably co-expressing IpLITR 1.1b and

#### REFERENCES


LifeAct-GFP were incubated at 27°C with αHA monoclonal antibody-coated 4.5 µm bright blue microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope. Target bead of interest is located within the box.

Video S19 | Live-cell imaging of IpLITR 1.1b-mediated phagocytosis at 27°C. Rat basophilic leukemia-2H3 cells stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 27°C with αHA monoclonal antibody-coated 4.5 µm bright blue microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope. Target bead of interest is located within the box.

Video S20 | Live-cell imaging of IpLITR 1.1b-mediated phagocytosis at 27°C. Rat basophilic leukemia-2H3 cells stably co-expressing IpLITR 1.1b and LifeAct-GFP were incubated at 27°C with αHA monoclonal antibody-coated 4.5 µm bright blue microspheres. Immediately after the addition of target beads, images were collected at 10 s intervals for ~8 min using a Zeiss LSM 710 laser scanning confocal microscope. Target bead of interest is located within the box.


phagocytic modes for target acquisition and engulfment. *J Leukoc Biol* (2015) 98:235–48. doi:10.1189/jlb.2A0215-039RR


**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 Lillico, Pemberton and Stafford. 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.*

*Margarita Álvarez-Rodríguez1 , Patricia Pereiro1 , Felipe E. Reyes-López <sup>2</sup> , Lluis Tort2 , Antonio Figueras1 and Beatriz Novoa1 \**

*<sup>1</sup> Institute of Marine Research (IIM), National Research Council (CSIC), Vigo, Spain, 2Department of Cell Biology, Physiology and Immunology, Universidad Autónoma de Barcelona, Bellaterra, Spain*

In recent years, the innate immune response has gained importance since evidence indicates that after an adequate priming protocol, it is possible to obtain some prolonged and enhanced immune responses. Nevertheless, several factors, such as the timing and method of administration of the immunostimulants, must be carefully considered. An inappropriate protocol can transform the treatments into a double-edged sword for the teleost immune system, resulting in a stressful and immunosuppressive state. In this work, we analyzed the long-term effects of different stimuli (β-glucans, lipopolysaccharide, and polyinosinic:polycytidylic acid) on the transcriptome modulation induced by Spring Viremia Carp Virus (SVCV) in adult zebrafish (*Danio rerio*) and on the mortality caused by this infection. At 35 days post-immunostimulation, the transcriptome was found to be highly altered compared to that of the control fish, and these stimuli also conditioned the response to SVCV challenge, especially in the case of β-glucans. No protection against SVCV was found with any of the stimuli, and non-significant higher mortalities were even observed, especially with β-glucans. However, in the short term (pre-stimulation with β-glucan and infection after 7 days), slight protection was observed after infection. The transcriptome response in the zebrafish kidney at 35 days posttreatment with β-glucans revealed a significant response associated with stress and immunosuppression. The identification of genes that were differentially expressed before and after the infection seemed to indicate a high energy cost of the immunostimulation that was prolonged over time and could explain the lack of protection against SVCV. Differential responses to stress and alterations in lipid metabolism, the tryptophan–kynurenine pathway, and interferon-gamma signaling seem to be some of the mechanisms involved in this response, which represents the end of trained immunity and the beginning of a stressful state characterized by immunosuppression.

Keywords: zebrafish, immunostimulants, **β**-glucans, stress, tolerance, IFN-**γ**, kynurenine, TDO

#### INTRODUCTION

Studies on the effect of immunostimulants in fish aquaculture have been conducted in past decades (1). These immunostimulants include different pathogen-associated molecular patterns (PAMPs), such as polyinosinic:polycytidylic acid [poly(I:C)], lipopolysaccharide (LPS), and β-glucans (2). PAMPs are recognized by host cells through specific pattern-recognition

#### *Edited by:*

*Brian Dixon, University of Waterloo, Canada*

#### *Reviewed by:*

*Jorge Galindo-Villegas, Universidad de Murcia, Spain Francesco Buonocore, Università degli Studi della Tuscia, Italy*

> *\*Correspondence: Beatriz Novoa beatriznovoa@iim.csic.es*

#### *Specialty section:*

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

*Received: 23 April 2018 Accepted: 25 June 2018 Published: 09 July 2018*

#### *Citation:*

*Álvarez-Rodríguez M, Pereiro P, Reyes-López FE, Tort L, Figueras A and Novoa B (2018) Analysis of the Long-Lived Responses Induced by Immunostimulants and Their Effects on a Viral Infection in Zebrafish (Danio rerio). Front. Immunol. 9:1575. doi: 10.3389/fimmu.2018.01575*

**212**

receptors (PRRs), which are germline-encoded host sensors with a key role in innate immunity (3). These PRRs can be proteins expressed mainly in the cell membrane or endosomes of innate immune cells, such as toll-like receptors, cytoplasmic receptors, such as RIG-I-like receptors or NOD-like receptors, or secreted receptors, such as mannose-binding lectin (MBL) of the complement system (3, 4). Nevertheless, the PRRs that sense LPS and β-glucans have not been completely identified in teleosts (5, 6).

Many PRRs can share several signaling pathways, and they have the ability to influence the type, intensity, and duration of the immune response (7). Signaling through PRRs after a first immune stimulus may result in cell reprogramming, which changes and conditions the immune response triggered by a second stimulus. This reprogramming is a host adaptation that coordinates multiple receptors and reflects the plasticity of innate immunity (8).

The term "trained immunity" was coined by Netea et al. in 2011 (9) to refer to cross-protection between infections that occurs independently of T and B cells or to the enhanced non-specific innate immune protection elicited by microbial components or PAMPs. This innate memory is mainly provided by NK cells and macrophages. It has been shown that the priming (or training) of mice with several PAMPs, such as β-glucans, peptidoglycans, or flagellin, can protect against a subsequent lethal infection (10). In contrast, LPS-induced tolerance has been described as the transformation of macrophages to a state that is much less responsive to subsequent challenges with LPS. Such a response should protect organisms from hyperinflammation or sepsis (11). Nevertheless, this suppression can become pernicious when the organism has to fight against a pathogen. Both tolerance and training involve long-term epigenetic reprogramming of the innate immune system after the first contact with a particular pathogen or PAMP (12–14).

It has been demonstrated that β-glucans, which are probably the most commonly used immunostimulant in fish aquaculture, are able to induce metabolic changes that reflect a shift from oxidative phosphorylation to aerobic glycolysis (Warburg effect) during the establishment of trained immunity (15). Because increasing evidence indicates a close relation between metabolism and epigenetic reprogramming (16, 17), the effect of β-glucans on metabolism could mediate changes in the epigenetic pattern. Other metabolic pathways that are independent of glucose metabolism can play a role in the induction of trained immunity by β-glucans (18). Although the identity of the PRR responsible for recognizing β-glucans in fish remains unknown (19), mammals sense this PAMP *via* dectin-1, a C-type lectin receptor that is absent in fish genomes (20). In mammals, training immunity occurs in a dectin-1-dependent manner in macrophages (21).

The outcome of β-glucan stimulation depends on several factors, such as the administration route and fish species, and its protective effect depends on the infectious agent (19). It is also important to keep in mind that only a very small fraction of orally administered β-glucans pass through the intestine to the circulation (22). Therefore, the intraperitoneal administration of β-glucans can provide higher levels of protection, even after a single injection (19). There are many works based on studying the effect of β-glucan pretreatment on the resolution of bacterial infections in fish after either oral or intraperitoneal administration (23–27). However, investigations of resistance against viral infections after β-glucan stimulation in teleosts (28–32), or in vertebrates in general (33–35), are scarce.

The main premise that must be taken into account when considering a substance as a good immunostimulant is the absence of pathogenic, toxic, or other undesirable effects after administration to aquatic organisms. Nevertheless, the administration of a certain immunostimulant can be a stressful stimulus and produce indirect pernicious effects at the immune level. In non-mammalian vertebrates, such as fish, the relation between stress and the immune system is more evident. Neuropeptides and cytokines play roles in the immune and neuroendocrine systems and belong to the same family of molecules (36, 37). A prompt but intense stimulus can result in acute stress; therefore, long-lasting effects on several stress- and immune-related parameters could be observed. Because the head kidney of fish is a unique structure in vertebrates that performs important functions for the endocrine, nervous, and the immune systems (36), it is an ideal tissue for studying the potential effects of immunostimulants.

In this work, we analyzed the long-lived response induced by three PAMPs [LPS, poly(I:C), and β-glucans] in zebrafish at 35 days posttreatment and investigated their effects on survival after Spring Viremia Carp Virus (SVCV) infection. None of the treatments induced protection against viral challenge, and non-significant higher mortalities were even found in the PAMP-treated groups. Kidney samples were collected from nonchallenged and SVCV-challenged fish after 24 h, and microarray analyses were conducted. Special attention was paid to β-glucans, which induced a stronger and more specific transcriptome modulation in both uninfected and infected zebrafish. The transcriptome response at 35 days posttreatment with β-glucans revealed interesting changes that could be related to the stress response, immunosuppression, or tolerance. This response to β-glucans seems to be mediated by the interplay among different metabolic and immune processes, such as the alteration of lipid metabolism, the tryptophan–kynurenine pathway and interferon-gamma signaling.

## MATERIALS AND METHODS

#### Fish, Viruses, and Bacteria

Zebrafish were obtained from our experimental facility, where the fish were cultured using established protocols (38, 39) (see http://zfin.org/zf\_info/zfbook/zfbk.html). In this work, adult wild-type zebrafish were used for transcriptome and mortality analyses. Double-transgenic *Tg(mpeg:mCherry/mpx:GFP)* zebrafish embryos, in which macrophages are labeled red and neutrophils are labeled green, and wild-type embryos were used in this work for microinjection experiments. Zebrafish were euthanized using a tricaine methanesulfonate (MS-222) overdose (500 mg/l<sup>−</sup><sup>1</sup> ). For microinjection experiments, larvae were anesthetized by adding two drops of a 0.05% MS-222 solution to a Petri plate with 10 ml of water. Fish care and the challenge experiments were conducted according to the guidelines of the CSIC National Committee on Bioethics under approval number ES360570202001/16/FUN01/PAT.05/ tipoE/BNG.

Spring Viremia Carp Virus isolate 56/70 was propagated on epithelioma papulosum cyprini (EPC) carp cells (ATCC CRL-2872) containing MEM (Gibco) supplemented with 2% FBS (Gibco) and 100 µg/ml Primocin (InvivoGen) and titrated in 96-well plates. The TCID50/ml was calculated according to the Reed and Muench method (40).

#### Plasmid Construction

The zebrafish *interferon gamma 1-2* (*ifng1-2*) gene was amplified by PCR (primers presented in Table 1 in Data Sheet S1 in Supplementary Material), and the PCR product was cloned using the pcDNA 3.1/V5-His TOPO TA Expression Kit (Invitrogen). One Shot TOP10F′ competent cells (Invitrogen) were transformed to generate the plasmid construct (pcDNA 3.1-*ifng1-2*). Plasmid purifications were conducted using the PureLink HiPure Plasmid Midiprep Kit (Invitrogen) following the manufacturer's instructions.

#### Reagents

Lipopolysaccharide from *Escherichia coli* O111:B4, poly(I:C), a TDO-inhibitor [C80C91 or 6-fluoro-3-((1E)-2-(3-pyridinyl) ethenyl)-1H-indole], and an IDO-inhibitor (1-MT or 1-methyld-tryptophan) were purchased from Sigma-Aldrich (reference numbers L2630, P1530, SML0287, and 452483, respectively). β-glucans extracted from *Saccharomyces cerevisiae* were obtained from Biotec-Mackzymal. The zebrafish IFN-gamma 1-2 recombinant protein was acquired from Kingfisher Biotech (reference number RP1045Z).

#### Stimulation With PAMPs and Challenge With SVCV

Four groups composed of 92 adult zebrafish each were intraperitoneally (i.p.) injected with 20 µl of one of the following treatments: β-glucans (1 mg/ml), LPS (0.75 mg/ml), poly(I:C) (1 mg/ml), and control treatment with phosphate-buffered saline (PBS). After 35 days, half of the individuals were i.p. infected with 20 µl of a SVCV suspension (3 × 102 TCID50/ml), and the remaining fish, which served as uninfected controls, were inoculated with viral medium (MEM + 2% FBS + Primocin). For microarray hybridization, a total of 16 fish from each treatment (β-glucans-control, LPS-control, poly(I:C)-control, β-glucans-SVCV, LPS-SVCV and poly(I:C)-SVCV) were sacrificed at 24 h postinfection, and the kidney was removed, yielding four pooled biological replicates (four fish/replicate). Total RNA was extracted as described below, and RNA quality was assessed with the Agilent 2100 Bioanalyzer (RIN ≥ 7).

To determine the long-term effects of the different PAMPs on mortality after challenge with SVCV, the remaining 30 fish from each group were divided into 3 replicates (10 fish per replicate), and mortality was recorded for a period of 3 weeks.

To analyze potential short-term priming with β-glucans, a total of 48 zebrafish were i.p. inoculated with 20 µl of β-glucans (1 mg/ml) or PBS. After a rest period of 7 days, half of the fish were i.p. infected with 20 µl of an SVCV suspension (1.5 × 102 TCID50/ml), and the remaining individuals were inoculated with viral medium (control fish). The fish from each group were distributed in three biological replicates (eight individuals/replicate). Mortality was recorded for a period of 3 weeks. In parallel, the same experiment was conducted to analyze gene expression by qPCR at 24 h postinfection in kidney samples (five individual fish by treatment).

## RNA Isolation and cDNA Transcription

Total RNA was isolated using the Maxwell 16 LEV Simply RNA Tissue Kit (Promega) according to the manufacturer's instructions. cDNA synthesis was conducted with an NZY First-Strand cDNA Synthesis kit (NZYTech) using 0.3 µg of total RNA.

#### Microarray Analysis

The 4 × 44K Zebrafish Gene Expression Microarray (V3, AMADID 026437) from Agilent Technologies (Madrid, Spain) was used to analyze gene expression in the different samples (Glucans, LPS, poly(I:C), Glucans-SVCV, LPS-SVCV, and poly(I:C)-SVCV).

The labeling of 1 µg of RNA and hybridizations were carried out at the Universidad Autónoma de Barcelona microarray facility, in compliance with the Minimum Information about a Microarray Experiment (MIAME) standards (41). The signal was captured, processed, and segmented using an Agilent G2565B scanner (Agilent Technologies, Madrid, Spain) with Agilent Feature Extraction Software (v9.5) protocol GE1-v5\_95 using an extended dynamic range and preprocessing by Agilent Feature Extraction v9.5.5.1.

The results for the fluorescence intensity data and quality annotations were imported into GeneSpring GX version 14.9 (Agilent Technologies). All of the control features were excluded from the subsequent analyses. After grouping the biological replicates (four replicates per treatment), entities with an expression level between the 20th and 95th percentiles in the raw data were retained and used in the subsequent analyses. The genelevel experiment was carried out by normalizing the data *via* a percentile shift at the 75th percentile and using the median of all samples as the baseline transformation. Differentially expressed genes were identified through volcano plot filtering. An unpaired *t*-test was conducted without correction, and data were considered significant at *p* ≤ 0.01. The fold-change cut-off was set at 1.5. Raw and normalized data were deposited in the NCBI's Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/ geo/) and are available under the accession number GSE113241. The heatmaps included in this work were also constructed with GeneSpring GX version 14.9.

## Microarray Validation and qPCR

Microarray results were validated through qPCR by analyzing the expression of nine different genes in the same samples used to hybridize the microarray, specifically, in the samples Glucans-SVCV and PBS-SVCV. Those genes were *nk-lysin a* (*nkla*), *nk-lysin*  *d* (*nkld*), *interferon-gamma 1-2* (*ifng1-2*), *tryptophan 2,3-dioxygenase a* (*tdo2a*), *cyclin y* (*ccny*), *membrane-bound O-acyltransferase domain containing 4* (*mboat4*), *pyruvate dehydrogenase kinase isozyme 3a* (*pdk3a*), *nuclear receptor subfamily 1 group D member 2a* (*nr1d2a*), and *adrenomedullin 2a* (*adm2a*). The correlation between microarray results and qPCR fold-change values was analyzed using the Pearson correlation coefficient. Specific qPCR primers were designed using the Primer3 software (42), and their amplification efficiency was calculated using seven serial twofold dilutions of cDNA with the threshold cycle (CT) slope method (43). Primer sequences are listed in Table 1 in Data Sheet S1 in Supplementary Material. Individual qPCR reactions were carried out in a 25 µl reaction volume using 12.5 µl of SYBR GREEN PCR Master Mix (Applied Biosystems), 10.5 µl of ultrapure water (Sigma-Aldrich), 0.5 µl of each specific primer (10 µM), and 1 µl of fivefold diluted cDNA template in MicroAmp optical 96-well reaction plates (Applied Biosystems). All reactions were performed using technical triplicates in a 7300 Real-Time PCR System thermocycler (Applied Biosystems), with an initial denaturation (95°C, 10 min) followed by 40 cycles of a denaturation step (95°C, 15 s) and one hybridization-elongation step (60°C, 1 min). The relative expression levels of the genes were normalized using *18S ribosomal RNA* (*18 s*) as a reference gene following the Pfaffl method (43).

### Panther and Blast2GO Transcriptome Analysis

Lists of genes that were significantly up- and downregulated relative to the corresponding control were imported into the Protein ANalysis THrough Evolutionary Relationships (PANTHER) Classification System Version 13.0 (44) to classify the modulated genes into Gene Ontology (GO)-Slim Biological Processes.

Enrichment analysis was carried out using the Blast2GO software (45, 46) to detect the major GO terms that were overrepresented in relation to the entire microarray chip (*p*-values ≤0.01 for biological process and ≤0.005 for molecular function).

#### Overexpression of *ifng1-2* in Zebrafish Larvae

The expression plasmid pcDNA3.1-*ifng1-2* and the corresponding control plasmid (pcDNA3.1) were microinjected into one-cell stage zebrafish embryos using a glass microneedle incorporated into the Narishige MN-151 micromanipulator and the Narishige IM-30 microinjector. Wild-type zebrafish embryos were microinjected with 200 pg/egg (final volume of 2 nl, diluted in PBS) of the recombinant plasmid or the empty plasmid (60 embryos per treatment). Three days after plasmid injection [3 days postfertilization (dpf)], half of the larvae from each treatment were infected through microinjection into the duct of Cuvier (47) with 2 nl of an SVCV suspension (1.5 × 105 TCID50/ml) diluted in PBS and 0.1% phenol red. The remaining larvae were microinjected with the same volume of PBS + 0.1% phenol red (uninfected controls). The infections were carried out at 23°C. Larvae from each group (pcDNA3.1-Control, pcDNA3.1-*ifng1-2-*Control, pcDNA3.1-SVCV, and pcDNA3.1-*ifng1-2*–SVCV) were distributed in three biological replicates (10 larvae/replicate) into 6-well plates. Mortality was assessed during the next 5 days postinfection (dpi). This experiment was replicated three times. Samples were also collected to assess by qPCR the correct replication of the expression plasmid pcDNA3.1-*ifng1-2* in non-infected larvae (3 biological replicates, 10 larvae/replicate).

To analyze the effect of *ifng1-2* overexpression in macrophages and neutrophils, double-transgenic *Tg(mpeg:mCherry/mpx:GFP)* embryos were also microinjected with 200 pg/egg of pcDNA3.1 *ifng1-2* and pcDNA3. At 3 dpf, half of the larvae from each treatment were challenged with SVCV (103 TCID50/ml). The remaining larvae were microinjected with the same volume of PBS + 0.1% phenol red (uninfected controls). The cell morphology was analyzed in 4 dpf larvae using a TSC SPE confocal microscope (Leica), and images were processed with the LAS AF software (Leica).

#### Effect of Ifng1-2 and TDO and IDO Inhibitors on Mortality After an SVCV Challenge in Adult Zebrafish

Six groups of 24 adult zebrafish were i.p. inoculated with a 10 µl volume of one of the following treatments: recombinant zebrafish Ifng1-2 (0.003 µg/µl) co-administered with SVCV, a TDO inhibitor (680C91; 1.5 µg/µl) in combination with an IDO inhibitor (1-MT; 1.5 µg/µl) and SVCV, SVCV alone, Ifng1-2 without SVCV, a TDO-inhibitor and an IDO-inhibitor without SVCV, or viral medium alone (MEM + 2% FBS + Primocin + 0.08% DMSO). The SVCV concentration was 3 × 102 TCID50/ml, and all treatments were diluted in viral medium. The fish from each group were distributed in three biological replicates (eight individuals/ replicate). Mortality was recorded during the next 3 weeks.

## Statistical Analyses

For the survival experiments, Kaplan–Meier survival curves were constructed and analyzed with the log-rank (Mantel–Cox) test. The Pearson correlation coefficient was calculated using the SPSS software version 23.0.

## RESULTS

## Effect of Different PAMPs on the Zebrafish Transcriptome After 35 Days

The number and intensity of modulated genes in the zebrafish kidney 35 days after the administration of three different PAMPs are shown as stacked column charts in **Figure 1A**. Additionally, the genes that were significantly affected by the long-term administration of β-glucans, poly(I:C), and LPS and their corresponding fold-changes are represented in Data Sheet S2 in Supplementary Material.

The number of differentially expressed genes (DEGs) in the PAMP-treated fish compared to the control (PBS-injected) fish indicates that the effect of the immunostimulation endured over time. The response to LPS administration was the most equilibrated, with a lower number of DEGs, 64 genes in total, and similar quantities of up- and downregulated genes, 31 and 33 genes, respectively. Stimulation with β-glucans and poly(I:C) induced similar responses in terms of the number of modulated

genes, with 135 and 109 DEGs, respectively, and a predominance of downregulated genes was observed (**Figure 1A**). Nevertheless, a Venn diagram comparing the number of shared and exclusive DEGs among the three treatments revealed strikingly different gene modulation patterns (**Figure 1B**). No DEGs were common to the three PAMPs. In addition, only two genes were shared by glucans and LPS or LPS and poly(I:C), and three DEGs were common to glucans and poly(I:C). These results reflect an exclusive reprograming of the transcriptome for each PAMP.

### Differential Responses to SVCV Challenge After Previous Administration of Different PAMPs and Effects on Survival

When the transcriptome response to SVCV was analyzed at 24 h postinfection in fish that were previously stimulated with PAMPs or not (Data Sheet S3 in Supplementary Material), the number of DEGs relative to PBS-inoculated zebrafish was found to be higher when PAMP stimulation and viral infection were combined than when the fish received PAMP stimulation or SVCV infection alone (**Figure 2A**). This multiplied response was drastic in the case of the individuals pretreated with β-glucans and then infected with SVCV after 35 days. In these individuals (Glucans-SVCV), the number of DEGs relative to the PBS group was 2,486, whereas only 135 and 113 genes were modulated in the Glucans and PBS-SVCV groups, respectively. This difference implies that the response to the virus was 21 times higher in the Glucans-SVCV group than in the PBS-SVCV group and 18 times higher in the Glucan-SVCV group than in the Glucans group. The additive effect of the two stimuli was much lower in the case of poly(I:C) [946 DEGs in poly(I:C)-SVCV] and especially LPS (250 DEGs in LPS-SVCV) (**Figure 2A**).

When the transcriptome response to SVCV challenge in fish that were previously treated with different stimuli was compared with the response to infection alone (PBS-SVCV group) (Data Sheet S4 in Supplementary Material) through a Venn diagram, the number of exclusive DEGs was also high, representing 92.1% in the Glucans-SVCV group, 62% in the poly(I:C)-SVCV group, and 71.7% in the LPS-SVCV group (**Figure 2B**).

A dendrogram and a heatmap of the overall microarray data revealed two main clusters of similarity (**Figure 3A**). In the first cluster, zebrafish groups that were inoculated with PAMPs or PBS but uninfected were placed together. Interestingly, in this cluster, the PBS-SVCV group was also included, which was closely related to the poly(I:C)-injected fish. The expression pattern of the Glucans group was the most different within this cluster and formed an independent branch. Therefore, the treatment with β-glucans induced a stronger long-lasting effect on the transcriptome. In the other cluster, the groups that were stimulated with the three PAMPs and then infected with SVCV were included, with the Glucans-SVCV group being the most different.

The DEGs in all treatment groups, infected and uninfected, relative to the PBS uninfected group, were categorized according to their biological processes using the PANTHER software. In the "Immune response" category, the three PAMP-treated groups presented genes in "B cell mediated immunity"; Glucans and poly(I:C) also presented genes in "complement activation," whereas only Glucans presented genes belonging to the "natural killer cell activation" category (**Figure 3B**). After infection in the absence of previous immunostimulation (PBS-SVCV), the response seemed to be dominated by "response to interferongamma," but individuals that were previously treated with PAMPs maintained certain mechanisms triggered by the PAMP even after SVCV infection (**Figure 3B**), presenting a completely different response than the non-immunostimulated group. This finding could indicate a longer conditioning effect after the first stimulus (PAMP) than after the second stimulus (SVCV).

Despite these amplified responses to the virus in individuals that were previously stimulated with PAMPs, a reduction of

mortality was not observed (**Figure 4A**). In contrast, although the differences were not found to be statistically significant, higher mortality after SVCV infection was observed in the three groups stimulated with PAMPs.

## The Modulated Transcriptome of **β**-Glucan-Treated Fish at 35 Days Poststimulation Suggests an Immunosupressive State

Due to the larger/stronger response of the individuals treated with β-glucans, we focused our attention on this immunostimulant. Because the SVCV infections conducted in this work were performed after a long resting period (35 days after stimulation with the different PAMPs), we wanted to determine if β-glucans were able to provide protection against SVCV at 7 days poststimulation. In this case, β-glucan administration slightly but significantly increased the survival rate from 17.4 to 33.3% (**Figure 4B**), indicating that the effect of this stimulus after 7 days benefits the resolution of the disease.

To determine why β-glucans were not able to protect after a long rest period but had a profound effect on gene expression, we analyzed in more detail the modulation of the transcriptome in treated fish 35 days after i.p. injection (Data Sheet S5 in Supplementary Material). When the complete set of DEGs was taken into account, interesting clues about the effects of long-term stimulation with β-glucans were revealed (Table 2 in Data Sheet S1 in Supplementary Material). In this case, it is interesting to highlight the enriched GO terms related to steroid hormones in the comparison Glucans-SVCV vs. Glucans, such as "intracellular steroid hormone receptor signaling pathway" and "response to epinephrine," which do not appear in the comparison PBS-SVCV vs. PBS. These terms are related to the activation of stress response pathways. Indeed, "response to epinephrine" is already represented in the comparison Glucans vs. PBS. It is also worth noting that terms related to

Figure 3 | Similarity analysis of the transcriptome response at 35 days post-immunostimulation in the presence or absence of Spring Viremia Carp Virus (SVCV) infection. (A) Dendrogram and heatmap representing the overall microarray results. The samples were divided into two clusters, one containing the pathogen-associated molecular patterns (PAMP)-stimulated groups and the individuals infected in the absence of pre-stimulation, and the other including the groups that were pre-stimulated with PAMPs and then infected with SVCV. In both clusters, the β-glucans formed a separate branch, indicating that this stimulus induced the most differential response. (B) Functional classification of DEGs in the zebrafish kidney between the different treatments and the phosphate-buffered saline-treated group according to Slim Biological Process Gene Ontology Terms. The level 1 category "Immune System Process" was selected, and within that category, the level 2 category "Immune Response." After SVCV challenge, the gene categorization was conditioned by previous stimulation with a PAMP.

the antiviral responses that were observed after SVCV infection in the comparison PBS-SVCV vs. PBS (such as chemotaxis and inflammation) were not observed after infection in the comparison Glucans-SVCV vs. Glucans (Table 2 in Data Sheet S1 in Supplementary Material).

If we focus on the GO terms associated with the top 25 most up- and downmodulated genes (**Table 1**) in the β-glucan-treated fish after infection relative to non-stimulated but infected fish (**Table 1**), it is interesting to note that the GO terms enriched in the upregulated genes were related to lipid transport ("mediumchain fatty acid transport"), the catabolism of amino acids, especially tryptophan ("tryptophan catabolic process to acetyl-CoA," "tryptophan catabolic process to kynurenine," "tryptophan 2,3-dioxigenase activity," "tyrosine metabolic process," and "l-phenylalanine catabolic process"), and interferon-gamma activity ("interferon-gamma receptor binding"). Interestingly, an immunosuppressive profile seems to be present in individuals that previously received β-glucans, represented by GO categories enriched in the most downmodulated genes that were directly related to pathogen recognition ("detection of diacyl bacterial lipopeptide," "toll-like receptor 6 signaling pathway," and "MyD88-dependent toll-like receptor signaling pathway") and inflammation ("positive regulation of interleukin-6 biosynthetic process," "T-helper 1 type immune response," and "regulation of cytokine secretion") (**Table 2**).

#### **β**-Glucans Modify Lipid Metabolism by Inhibiting the Synthesis of Fatty Acids and Cholesterol Before and After Viral Infection

Lipid transport seemed to characterize the response to SVCV in individuals that were previously treated with β-glucans. This finding was consistent with the modulation of genes that were overexpressed, such as *fatty acid-binding protein 1b, liver, tandem duplicate 2 (fabp1b.2), microsomal triglyceride transfer protein (mtp), and major facilitator superfamily domain containing 2ab (mfsd2ab)* (**Table 1**). It is known that β-glucans induce changes in lipid metabolism, and these metabolic changes could condition the immune status of the host. To determine the differential expression profiles of genes encoding key proteins in different lipid metabolic pathways [fatty acid oxidation (FAO), ketolysis, lipogenesis, and cholesterol biosynthesis], specific heatmaps for each pathway were constructed (**Figure 5**).

Hormone genes that could underlie the modulation of these pathways are shown in **Figure 5A**. The genes *leptin* (*lepb*), *insulindegrading enzyme* (*ide*), and *glucagon* (*gcgb*), whose expression is potentiated by double stimulation with β-glucans and SVCV, are inhibitors of the synthesis of lipids and favor lipolysis and ketolysis (48, 49). On the other hand, an opposite effect was observed for *insulin* (*isn*), *ghrelin* (*ghrl*) and *membrane-bound o-acyltransferase domain containing 4* (*mboat4*), which endode proteins that activate lipogenesis (50, 51). In accordance with these observations, the heatmaps associated with lipid metabolism showed that Glucans-SVCV zebrafish present a clear induction of FAO (**Figure 5B**) and ketolysis (**Figure 5C**) and a significant inhibition of fatty acid syntesis (**Figure 5D**), which is reflected in the downregulation of key components of lipid synthesis, such as *sterol regulatory element binding transcription factor 1* (*srebf1*), *fatty acid synthase* (*fasn*), and *acetyl-coA carboxylase (acc)*, and genes of the cholesterol synthesis pathway (**Figure 5D**). The effect on lipid metabolism was also observed in fish that were stimulated with

(A) Kaplan–Meier survival curves after an SVCV i.p. challenge. Fish were stimulated with PAMPs or inoculated with phosphate-buffered saline (PBS), and 35 days after a single-dose administration, they were i.p. challenged with SVCV. No significant differences were observed among the different groups, although the PAMP-treated zebrafish showed a lower survival rate. (B) Kaplan–Meier survival curves after SVCV i.p. challenge. Fish were previously inoculated with β-glucans or PBS and i.p. challenged with SVCV 7 days after a single-dose administration. A slight but significant increase in the survival rate was observed in fish that were previously treated with β-glucans. Significant differences are represented by asterisks (\**p* < 0.05).

β-glucans alone, in which the inhibition of cholesterol synthesis, the downregulation of *ghrl,* and the upregulation of the fatty acid translocase *scavenger receptor class B, member 3* (*cd36*), which is essential for fatty acid import into cells and efficient oxidation, were observed.

## Effect of Ifng1-2 on Macrophage Activation and Survival After an SVCV Challenge

The transcriptome of infected fish that were previously treated with β-glucans was enriched in the immune term "interferongamma receptor binding," as a result of the high expression of the *ifng1-2* gene, among others (**Table 1**). IFN-γ is an important activator of macrophages in mammals, with immunostimulatory and immunomodulatory properties. Nevertheless, in this case, Ifng1-2 seeems to be associated to an immunosuppressive state. For that reason, we wondered if this molecule has a pernicious effect in zebrafish during an infection with SVCV or if its association with the immunosuppression depends on other factors.

An expression plasmid encoding *ifng1-2 (*pcDNA3.1-*ifng1-2)* was constructed, and its correct replication was confirmed in 3 dpf larvae (Figure 1 in Data Sheet S1 in Supplementary Material). The plasmid pcDNA3.1-*ifng1-2* and the corresponding empty plasmid (pcDNA3.1) were microinjected into double-transgenic *Tg(mpeg:mCherry/mpx:GFP)* zebrafish embryos alone or in combination with SVCV. When we compared the morphology of macrophages and neutrophils in 4 dpf larvae, it was evident that larvae overexpressing *ifng1-2* presented a different morphology than the controls for both cell types (**Figure 6A**). Both neutrophils and macrophages changed from a spherical morphology to a dendritic morphology, which was especially evident in infected and pcDNA3.1-*ifng1-2-*treated fish; in the case of macrophages, these changes probably reflect the activation of these cells.

The overexpression of *ifng1-2* in wild-type zebrafish larvae also revealed the ability to significantly increase survival after an SVCV challenge (**Figure 6B**). The survival of the empty plasmidtreated fish was 48%, whereas this rate increased to 76.7% in individuals that were previously inoculated with the expression plasmid pcDNA3.1-*ifng1-2*. This protective capability against SVCV was also tested in adult zebrafish using the recombinant protein (**Figure 6C**). In this case, the untreated fish exhibited a survival rate of 17.4%, but individuals that were inoculated with Ifng1-2 significantly increased their survival to 50%.

#### The Alteration of the Tryptophan– Kynurenine Pathway Influences the Survival Rate After SVCV Infection

As observed in the GO enrichment analysis, the kynurenine pathway of tryptophan catabolism was highly altered in the Glucan-SVCV individuals compared to the PBS-SVCV fish (**Figure 7A**). Interestingly, this pathway has a close relation with the activity of IFN-γ. A heatmap including the genes that encode some of the main proteins involved in tryptophan catabolism, such as *ifng1-2*, tryptophan 2,3-dioxygenase a (*tdo2a*), tryptophan hydroxylase 2 (*tph2*), and kynurenine aminotransferase 2 (*kyat2*), revealed that previous stimulation with β-glucans determines the response of these genes to an SVCV challenge (**Figure 7B**). The genes *ifng1-2* and *tdo2a* were two of the most overexpressed genes in Glucans-SVCV vs. PBS-SVCV fish (**Table 1**). However, after short-term glucan stimulation, only *ifng1-2* was slightly overexpressed in individuals treated with glucans and then infected and *tdo2a* was not modulated (Figure 2 in Data Sheet S1 in Supplementary Material). Although it was not included in this heatmap, it is also interesting to highlight the second most upregulated gene in the Glucans group relative to the PBS group, the *G protein-coupled receptor 35 (gpr35),* because intermediates of the kynurenine pathway are endogenous agonists of this receptor.

To investigate the importance of this pathway in mortality after SVCV, a TDO-inhibitor (C80C91) and an IDO-inhibitor (1-MT) were inoculated together with SVCV. The blockage of the kynurenine pathway with these inhibitors significantly increased the survival rate after infection (**Figure 7C**), with values that changed from 17.4% in untreated fish to 46.7% in individuals Table 1 | Top 25 most up- and downregulated genes in β-glucan-treated zebrafish compared to unstimulated fish in the absence (A) or presence (B) of Spring Viremia Carp Virus (SVCV) infection.

#### (A) Top 25-DEGs glucans vs. phosphate-buffered saline (PBS)


#### (B) TOP 25-DEGs glucans-SVCV vs. PBS-SVCV


#### TABLE 1 | Continued


that received C80C91 + 1-MT. Therefore, the blockage of the kynurenine pathway significantly reduces the mortality caused by SVCV.

#### Validation of the Microarray Data by qPCR

Correlation of the fold-change values for the comparison Glucans-SVCV vs. PBS-SVCV between microarray and qPCR showed a Pearson correlation coefficient (*r*) of 0.943 (*p* < 0.001). The expression data obtained for the selected genes are plotted in Figure 3 in Data Sheet S1 in Supplementary Material.

Table 2 | Gene ontology (GO) enrichment analysis of the top 25 most modulated genes in fish that previously received β-glucans and were then infected [Glucans-Spring Viremia Carp Virus (SVCV)] compared to untreated and infected fish (PBS-SVCV).


#### TOP 25 downregulated DEGs GLUCANOS-SVCV vs. PBS-SVCV


*BP, biological process; MF, molecular function.*

## DISCUSSION

Typically, the use of immunostimulants, especially β-glucans, in aquaculture has been conducted with the objective of increasing the non-specific defense mechanisms of the cultured fish (2). Nevertheless, several factors, such as the timing and method of administration, must be carefully considered (2). An inappropriate treatment protocol can transform the immunostimulants into a double-edged sword for the teleost immune system, resulting in a stressful and immunosuppressive state. The long-term effects of immunostimulants, even after a single administration, can produce distress situation, with a negative impact on immune competence. Chronic stress involves metabolic changes that allow the organism to face the demand for resources that is associated with stress. A prolonged immune stimulus consumes energy resources due to the cost required to generate an appropriate immune response. When this demand is prolonged over time, these resources are exhausted, leading to immunosuppression (52).

The trained immunity induced by β-glucans seems to be limited to a period of approximately 20 days. This effect was observed after *in vitro* stimulation of murine spleen-derived monocytes with this substance (53) and after *in vivo* stimulation in zebrafish (32). Moreover, a work based on long-term protection against the virus IHNV after an i.p. injection of β-glucans in trout revealed that beyond 36 days poststimulation, a higher mortality rate occurred in pre-immunized individuals (28). The genes encoding proinflammatory cytokines, such as TNF-α, IL-1β, and IL-6, which are usually found to be upregulated during cell training (15, 32, 53) were not differentially modulated in our microarray results. These findings, together with the absence of protection against SVCV, led us to consider that in this work, we are describing long-term mechanisms that could persist beyond the end of training immunity.

In this work, we wanted to analyze the long-term effects of a punctual administration of three different PAMPs [LPS, poly(I:C), and β-glucans] on the immune system of zebrafish and their consequences during a viral infection. At 35 days posttreatment, none of the PAMPs was able to provide protection after infection, and higher mortalities were even observed for the three treatments, especially for β-glucans. Moreover, the transcriptome of fish that were previously treated with PAMPs was found to be highly altered, even after this long period of time, and this effect was more pronounced with β-glucans.

This is not the first time that a negative effect of β-glucans was observed in teleosts. The *in vitro* administration of β-glucans to turbot (*Scophthalmus maximus*) and gilthead seabream (*Sparus aurata*) phagocytes revealed that high doses of this molecule directly induced the respiratory burst and rapidly led to cell exhaustion (52), which may increase disease susceptibility. *In vivo* studies in large yellow croaker (*Larimichthys crocea*) showed that individuals fed high doses of β-glucans were less resistant to a challenge with *Vibrio harveyi*, probably because high doses resulted in immunosuppression or feedback regulation (54). A recent study in rainbow trout (*Oncorhynchus mykiss*) demonstrated that overdoses of β-glucans can lead to poor immune responses, probably due to the activation of stress response mechanisms (55).

The response triggered by a stressful stimulus is mediated by the release of stress hormones, such as glucocorticoids (GCs), which can increase or suppress certain pathways of the immune response depending on the intensity and duration of the stressor (36). While an acute stress is more closely associated with eustress ("good stress"), which involves short-term challenges that result in immunoenhancement, distress ("bad stress") is caused by repeated or prolonged stress over time, which can cause immune suppression (56). In our case, in the short term (pre-stimulation with β-glucan and infection after 7 days), eustress could be responsible of the mechanisms that underlie training immunity and the slight protection observed

after infection. In the long term (pre-stimulation with β-glucan and infection after 35 days), the distress generated by the high energy cost of prolonged training immunity could explain the lack of protection against SVCV and even the tendency toward an increase in mortality.

Several factors, mainly involved in cell metabolism, could explain the lack of protection against the viral challenge. In mammals, processes such as glycolysis, the pentose phosphate pathway, glutaminolysis, and cholesterol and fatty acid synthesis, are usually upregulated after β-glucan training, and the blockade of some of these pathways (such as glycolysis, glutaminolysis, and cholesterol synthesis) inhibits training immunity (57). The synthesis of lipids in immune cells is associated with cell proliferation and the generation of inflammatory cytokines; however, FAO is more closely associated with tolerance, the suppression of the immune response, memory, and a long cellular lifespan (58). The microarray revealed that genes related to fatty acid and cholesterol synthesis are inhibited in individuals treated with β-glucans, especially after viral infection, and genes related to FAO were overexpressed. In our results, *fasn*, *acc,* and *srebp1* mRNA levels decreased in immunostimulated fish. However, it is known that *mammalian target of rapamycin* (mTOR), the key player in training immunity activation (15), promotes the synthesis of fatty acids through the induction of these three genes (59). These results suggest a shift of training immunity to a more immunotolerant profile in our experiments.

Another remarkable aspect of the altered transcriptome of glucan-treated fish was amino acid depletion, which is another tolerance indicator. Among the most DEGs in the Glucans-SVCV group relative to the PBS-SVCV group are the TDO gene *tdo2a* and *homogentisate 1,2-dioxygenase* (*hgm*), which encode enzymes involved in amino acid catabolism. Phenylalanine and tryptophan are essential amino acids that are necessary for mTORC1 activation (60–62). TDO and indoleamine 2,3-dioxygenase (IDO), which catalyze the first step of tryptophan catabolism, are the rate-limiting enzymes of the kynurenine pathway (63). TDO is mainly produced in the liver but is expressed at immune-privileged sites and in some tumors and is associated with tolerance and immune evasion (17). Therefore, because tryptophan is degraded by the action of IDO and TDO in the kynurenine pathway, these enzymes act as inhibitors of mTOR activation. Indeed, the use of inhibitors of IDO activity, such as D-1MT, reversed the inhibitory effects generated by IDO on mTOR activation (64).

The intermediates of the kynurenine pathway, due to their toxic potential, may cause tissue damage and, as a consequence, generate higher stress, which can in turn interfere with the immune response to minimize tissue damage. It was recently observed that feed supplementation with kynurenic acid (KYNA) in rainbow trout had a toxic/stress-inducing effect, and this deleterious effect was related to the dose of KYNA (65). This pathological status was manifested during a subsequent infection with *Yersinia ruckeri,* with a higher mortality rate in animals that were administered

a higher concentration of KYNA (65). On the other hand, kynurenic acid is an endogenous agonist of *gpr35* (which was the second most upregulated gene in β-glucan-stimulated zebrafish in the absence of infection), and its activation increases FAO and induces an anti-inflammatory state (66, 67). Additionally, kynurenic acid and kynurenine are endogenous agonists of the aryl hydrocarbon receptor (AhR) (68), which mediates immunosuppressive effects on the immune system (69), leading to inhibition of the promoter activity of IL-6 (70). Consistent with

our results, the depletion of tryptophan from the medium impairs the antiviral response in mammals by inducing the formation of regulatory T cells and inhibiting the formation of effector T cells (71). These effects are observed in our transcriptome results, with a downregulation of the "T-helper 1 type immune response" and "positive regulation of interleukin-6 biosynthetic process" in the comparison Glucans-SVCV vs. PBS-SVCV. The regulation of the innate responses by the adaptive immune system is known in mammals (72) and also it has been reported in zebrafish (73).

pathway and the DEGs in the comparison Glucans-Spring Viremia Carp Virus (SVCV) vs. Glucans-PBS (*tdo2a*, *kyat2*) and the comparison Glucans-SVCV vs. PBS (*tph2*). (B) Heatmap representing some of the most affected genes after β-glucan treatment. Whereas *ifng1-2* and *tdo2a* are highly expressed in the Glucans-SVCV group, *tph2* and *kyat2* have lower expression levels in these fish. A color gradient scheme representing gene modulation (red: lower expression; green: higher expression) is shown on the bottom. (C) Kaplan–Meier survival curves representing the effect of a TDO-inhibitor (C80C91) and an IDO-inhibitor (1-MT) during a challenge with SVCV in adult zebrafish. A significant increase in the survival rate was observed in individuals inoculated with the inhibitors. Significant differences are represented by asterisks (\**p* < 0.05).

The overexpression of the *ifng1-2* gene after viral infection in fish that previously received β-glucans initially contradicted the impaired response of these individuals. In fact, in our experiments, the overexpression of *ifng1-2* in zebrafish induced the activation of immune cells and was able to increase survival after an SVCV challenge. Nevertheless, a previous work reported an unexpected *in vivo* failure of Ifng1-2 to increase the resistance of zebrafish to bacterial and viral infections (74). IFN-γ is a pleiotropic cytokine, and it has been classically associated with a proinflammatory immune response (75). However, IFN-γ induces the expression of IDO, which catalyzes the same step as TDO in the kynurenine pathway (63). Therefore, the activation of the kynurenine pathway results from the induction of TDO through the action of GCs and from the induction of IDO through the overexpression of IFN-γ. This process not only depletes tryptophan from the medium but also probably interferes with the antiviral response of IFN-γ. The lack of induction of the *tdo2a* and the slight overexpression of *ifng1-2* compared to the infected controls in the short-term immunostimulation experiment would explain the increased survival of the individuals treated with glucans. Moreover, the higher viral resistance that was observed in this work when we applied the inhibitors of TDO and IDO together supports this hypothesis.

The depletion of tryptophan from the medium seems to be harmful for fish dealing with SVCV infection, as has been described for other viruses (76–78). However, it has been reported that toxic intermediates of the kynurenine pathway present bactericidal effects (76). Indeed, i.p. injection of β-glucans in *Salmo*  *salar* was able to increase survival against Gram-negative bacteria, even after a long resting period (23). Interestingly, tryptophan depletion (by the action of TDO or IDO) and IFN-γ signaling can have opposite effects in the context of infection, depending on the type of pathogen (virus or bacteria) (76, 79).

In summary, long-term immunostimulation with β-glucans does not protect against SVCV infection. This pre-stimulation seems to lead the zebrafish immune system toward a strategy of immunosuppression and tolerization, which results in pernicious effects for responding to the virus. The interplay among GCs, lipid metabolism, the kynurenine pathway, IFN-γ, and mTOR activation seems to be the mechanism behind this response.

#### ETHICS STATEMENT

Fish care and the challenge experiments were conducted according to the guidelines of the CSIC National Committee on Bioethics under approval number ES360570202001/16/FUN01/ PAT.05/tipoE/BNG.

#### AUTHOR CONTRIBUTIONS

BN and AF conceived and designed the study. MA-R and PP performed the experimental procedures and data analyses. FR-L and

#### REFERENCES


LT conducted the microarray hybridizations. MA-R, PP, AF, and BN wrote the manuscript. All authors reviewed the manuscript.

#### ACKNOWLEDGMENTS

We want to thank Judit Castro and the aquarium staff for their technical assistance.

#### FUNDING

This work was funded by the projects AGL2014-51773-C3 and BIO2017-82851-C3-1-R from the Spanish Ministerio de Economía y Competitividad, Proyecto Intramural Especial, PIE 201230E057 from Agencia Estatal Consejo Superior de Investigaciones Científicas (CSIC), and IN607B 2016/12 from Consellería de Economía, Emprego e Industria (GAIN), Xunta de Galicia. MA-R was the recipient of an FPU fellowship from the Ministerio de Educación (FPU014/05517).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.01575/ full#supplementary-material.

functional reprogramming of monocytes. *Cell Host Microbe* (2012) 12:223–32. doi:10.1016/j.chom.2012.06.006


(Cyprinus carpio) infected with *Aeromonas hydrophila*. *Fish Shellfish Immunol* (2005) 19:293–306. doi:10.1016/j.fsi.2005.01.001


immunity acquire an antiviral alert state characterized by upregulated gene expression of apoptosis, multigene families, and interferon-related genes. *Front Immunol* (2017) 8:121. doi:10.3389/fimmu.2017.00121


**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 Álvarez-Rodríguez, Pereiro, Reyes-López, Tort, Figueras and Novoa. 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.*

*Tiehui Wang1 \*, Yehfang Hu1 , Eakapol Wangkahart1,2, Fuguo Liu1 , Alex Wang1 , Eman Zahran1,3, Kevin R. Maisey <sup>4</sup> , Min Liu1,5, Qiaoqing Xu1,6, Mónica Imarai4 and Christopher J. Secombes1 \**

*1Scottish Fish Immunology Research Centre, School of Biological Sciences, University of Aberdeen, Aberdeen, United Kingdom, 2Division of Fisheries, Department of Agricultural Technology, Faculty of Technology, Mahasarakham University, Kantharawichai, Thailand, 3Department of Internal Medicine, Infectious and Fish Diseases, Faculty of Veterinary Medicine, Mansoura University, Mansoura, Egypt, 4 Laboratorio de Immunologia, Centro de Biotecnología Acuícola, Departamento de Biología, Facultad de Química y Biología, Universidad de Santiago de Chile, Santiago, Chile, 5 College of Animal Science and Technology, Northeast Agriculture University, Harbin, China, 6School of Animal Science, Yangtze University, Jingzhou, China*

#### *Edited by:*

*Ram Savan, University of Washington, United States*

#### *Reviewed by:*

*Magdalena Chadzińska, Jagiellonian University, Poland Masahiro Sakai, University of Miyazaki, Japan*

#### *\*Correspondence:*

*Tiehui Wang t.h.wang@abdn.ac.uk; Christopher J. Secombes c.secombes@abdn.ac.uk*

#### *Specialty section:*

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

*Received: 22 April 2018 Accepted: 09 July 2018 Published: 26 July 2018*

#### *Citation:*

*Wang T, Hu Y, Wangkahart E, Liu F, Wang A, Zahran E, Maisey KR, Liu M, Xu Q, Imarai M and Secombes CJ (2018) Interleukin (IL)-2 Is a Key Regulator of T Helper 1 and T Helper 2 Cytokine Expression in Fish: Functional Characterization of Two Divergent IL2 Paralogs in Salmonids. Front. Immunol. 9:1683. doi: 10.3389/fimmu.2018.01683*

Mammalian interleukin (IL)-2 is a cytokine centrally involved in the differentiation and survival of CD4+ T helper subsets and CD4+ T regulatory cells and in activation of cytotoxic effector lymphocytes. In bony fish, *IL2* orthologs have been identified with an additional divergent *IL2-Like* gene on the same locus present in several fish species. We report here two divergent *IL2* paralogs, *IL2A* and *IL2B*, in salmonids that originated from the whole genome duplication event in this fish lineage. The salmonid *IL2* paralogs differ not only in sequence but also in exon sizes. The IL-2 isoforms that are encoded have disparate pI values and may have evolved to preferentially bind specific IL-2 receptors. Rainbow trout *IL2* paralogs are highly expressed in thymus, spleen, gills, kidney and intestine, important tissues/organs in fish T cell development and function. Their expression in peripheral blood leukocytes (PBL) is low constitutively but can be upregulated by the mixed leukocyte reaction, by the T cell mitogen phytohemagglutinin and by signal mimics of T cell activation (phorbol 12-myristate 13-acetate and calcium ionophore). Both trout IL-2 isoforms promoted PBL proliferation and sustained highlevel expression of *CD4* and *CD8*, suggesting that trout IL-2 isoforms are T cell growth/ survival factors mainly expressed by activated T cells. The recombinant proteins for these two trout *IL2* paralogs have been produced in *E. coli* and possess shared but also distinct bioactivities. IL-2A, but not IL-2B, induced *IL12P35A1* and *CXCR1* expression in PBL. IL-2B had a stronger effect on upregulation of the T helper 1 (Th1) cytokine *interferon-γ* (*IFNγ*) and could sustain *CD8α* and *CD8β* expression levels. Nevertheless,

**Abbreviations:** CH, chromosome; CI, calcium ionophore; IL, interleukin; MLR, mixed leukocyte reaction; Mya, million year ago; PBL, peripheral blood leukocytes; PHA, phytohemagglutinin; PMA, phorbol 12-myristate 13-acetate; Th1, T helper 1; Th2, T helper 2; Th17, T helper 17; WGD, whole genome duplication.

both cytokines upregulated key Th1 (*IFNγ1*, *IFNγ2*, *TNFα2* and *IL12*) and T helper 2 (Th2) cytokines (*IL4/13B1* and *IL4/13B2*), cytokine and chemokine receptors and the antimicrobial peptide *cathelicidin-1* but had limited effects on T helper 17 cytokines and *TGFβ1* in PBL. They could also enhance PBL phagocytosis. These results suggest, for the first time in fish, that IL-2 isoforms may have an important role in regulating Th1 and Th2 cell development, and innate and adaptive host defenses in fish, and shed light on lineage-specific expansion, evolution, and functional diversification of *IL2* in vertebrates.

Keywords: salmonids, interleukin-2, expression, bioactivity, T cell growth factor, T helper 1, T helper 2, phagocytosis

#### INTRODUCTION

The T cell growth factor interleukin (IL)-2 was discovered in 1976 (1) and its cDNA cloned in 1983 in humans (2). It has a wide range of actions, including the ability to boost the cytolytic activity of natural killer (NK) cells and T cells, augment immunoglobulin production by activated B cells, maintain homeostatic proliferation of regulatory T cells (Treg cells), induce innate lymphoid cells (ILCs) and effector T cell differentiation, as well as influence memory T cells, effector T cells and monocytes (3–5). In T helper (Th) cell differentiation, IL-2 modulates the expression of receptors for other cytokines and transcription factors, thereby either promoting or inhibiting cytokine cascades that correlate with each Th cell development state (6). Thus, IL-2 promotes naïve T cell differentiation into T helper 1 (Th1) and T helper 2 (Th2) cells while inhibiting T helper 17 (Th17) cell development (5–7).

The production of IL-2 in mammals is tightly regulated and largely restricted to activated CD4+ T cells (6). Other populations of cells, including activated CD8+ T cells, NK cells and NKT cells, dendritic cells and mast cells have also been reported to secrete IL-2, albeit at much lower levels than activated CD4+ T cells (8, 9). At resting conditions, CD4 Th cells are the main source of the constant but low levels of IL-2. On immune activation of the T cell receptor (TCR) by antigens presented by antigen-presenting cells (APCs) and costimulatory signals, T cells (CD4+ and CD8+ T cells) start to secrete large amounts of IL-2 (3, 4). *In vitro*, IL-2 can be induced in T cells by T cell mitogens and co-mitogens, such as phytohemagglutinin (PHA), phorbol 12-myristate 13-acetate (PMA), and calcium ionophore (CI) (10).

Structurally, IL-2 is a short-chain type I cytokine with a four α-helical bundle (helices A–D) "up-up-down-down" configuration typical of this family of cytokines. In this configuration, helices are aligned in a fashion that allows interaction with its receptor chains (11, 12). There are three IL-2 receptor components, IL-2Rα (CD25), IL-2Rβ (CD122) and IL-2Rγ (CD132, also known as γC), forming three classes of IL-2 receptors. The low-affinity receptor contains only IL-2Rα, the intermediate-affinity receptor contains IL-2Rβ and γC, and the high-affinity receptor contains all three chains (3, 4). IL-2 activates three main signaling pathways, including the JAK–STAT pathway, the RAS–MAP kinase pathway, and the PI3-kinase/AKT pathway. Kinetically, IL-2 first interacts with IL-2Rα, resulting in a conformational change in IL-2 that then allows it to efficiently interact with IL-2Rβ, with γC subsequently recruited (3, 4, 11, 12).

The γC receptor chain is shared by receptors for IL-2, IL-4, IL-7, IL-9, IL-15, and IL-21 ligands that form the γC cytokine family (or IL-2 family). In mammals, each cytokine in this family has a unique private α receptor chain in addition to γC. Furthermore, IL-2 and IL-15 share IL-2Rβ as well as γC. The receptor chains of known fish γC cytokines are present in fish (13) with the apparent exception of the IL-2Rα. It has been suggested that in fish IL-2 and IL-15 may both bind to IL-15Rα (also termed CD25-like, CD25L) (14). A second *IL15*-related gene (*IL15L*) identified in fish was later found to be present in other vertebrate groups but is a pseudogene in man and mouse (15). The cow IL-15L binds to the IL-15Rα but not the IL-2Rα, as seen with IL-15 itself (15). It is noteworthy that mammalian IL-2Rα and IL-15Rα are distinctive cytokine receptor proteins that contain sushi domains at the N-terminal but lack domains typical of class I cytokine receptor proteins (13). The genes encoding these two proteins have a similar intron–exon organization and are very closely linked on human chromosome (CH)10 and mouse CH2, suggesting duplication of an ancient precursor of both genes (16). The lack of fish IL-2Rα may suggest that fish IL-2, IL-15, and IL-15L may share three receptor chains, the CD25L, IL-2Rβ, and γC.

Mammalian IL-2 functionally signals *via* either high or intermediate-affinity IL-2 receptors. In mammals, the three receptor chains are located on different CHs and differentially expressed and modulated (17). Many immune cells can respond to IL-2, but their sensitivity to IL-2 varies based on the types of IL-2 receptors expressed and the induction vs constitutive expression of the different IL-2 receptor chains. *IL2Rα* is constitutively expressed on Treg cells and ILC2 cells, whereas it is more transiently induced on activated lymphocytes (7). Although *IL2Rβ* is constitutively expressed, it is also induced in T cells by activation *via* TCR and IL-2 stimulation, albeit to a lesser extent than *IL2Rα*. *γC* is constitutively expressed in the lympho-hematopoietic lineage (17). Thus, on resting lymphocytes and NK cells, IL-2 signals *via* intermediate-affinity IL-2 receptors, whereas activated lymphocytes, Treg, and ILC2 cells additionally express IL-2Rα and therefore have both high- and low-affinity receptors. Interestingly, activated dendritic cells have been reported to express IL-2Rα and to be capable of binding secreted IL-2 and trans-presenting it to neighboring cells expressing IL-2Rβ and γC (18). Following receptor binding, IL-2 activates multiple signaling pathways to activate the expression of genes essential for effector cell function, differentiation, and T cell growth (7).

Two *IL2*-related genes, *IL2* and *IL2L*, have been described in fish (19). *IL2* was originally discovered by analysis of the fugu *Takifugu rubripes* genome sequence that also identified *IL21* as a neighboring gene as in mammals (20). An *IL2L* gene has also been discovered in several percomorph fish genomes (e.g., fugu, tetraodon *Tetraodon nigroviridis*, and stickleback *Gasterosteus aculeatus*) next to *IL2*. Both fish genes had a 4 exon/3 intron organization, as seen in mammals, but only shared 22–25% amino acid (aa) identity in the same species. *IL2* has since been cloned in rainbow trout *Oncorhynchus mykiss* (21, 22), and other species, but the bioactivity of fish IL-2 has only been reported in rainbow trout (23). The trout IL-2 recombinant protein induces expression of *interferon-γ* (*IFNγ*), the CXC chemokine *γIP*, *IL4/13B*, and to a lesser extent *IL2*, *STAT5* and *Blimp-1*, and *IL17A/F* genes in head kidney (HK) cells (21, 24, 25).

In this report, a second *IL2*-related gene, *IL2B* (with the previous one now termed *IL2A*), has been cloned in Atlantic salmon *Salmo salar*, and rainbow trout. These two *IL2* genes have also been identified in other available salmonid genomes, including coho salmon *Oncorhynchus kisutch*, chinook salmon *Oncorhynchus tshawytscha* and Arctic char *Salvelinus alpinus*, and arose *via* the salmonid whole genome duplication [WGD; (26)] event. IL-2A and IL-2B share only 39–43% aa sequence identity, suggesting that they may have changed functionally. The expression of both genes in rainbow trout was activated by the mixed leukocyte reaction (MLR), by the T cell mitogen PHA, and was synergistically induced by PMA and CI in peripheral blood leukocytes (PBL). Recombinant proteins for trout *IL2A* and *IL2B* have been produced in *E. coli* and tested functionally in PBL. Both cytokines upregulated the expression of genes involved in Th1 and Th2 pathways, sustained high-level expression of T cell markers but had limited ability to modulate the pro-inflammatory (Th17) and Treg cell pathways. They also promoted the proliferation of PBL *in vitro* and enhanced phagocytosis. This study suggests that fish IL-2 molecules are important T cell cytokines that regulate the Th1 and Th2 pathways and antimicrobial defense in fish.

#### MATERIALS AND METHODS

#### Fish

Juvenile rainbow trout were purchased from College Mill Trout Farm (Perthshire, UK) and maintained in aerated fiberglass tanks supplied with a continuous flow of recirculating freshwater at 14°C. Fish were fed twice daily on a commercial pellet diet (EWOS) and were given at least 2 weeks of acclimatization prior to treatment. All the experiments described comply with the Guidelines of the European Union Council (2010/63/EU) for the use of laboratory animals and were carried out under UK Home Office project license PPL 60/4013, approved by the ethics committee at the University of Aberdeen.

#### Cloning of Salmonid *IL2B* cDNA in Atlantic Salmon and Rainbow Trout

The *IL2B* cloning was performed in 2011 when no trout whole genome shotgun sequences (WGS) were available. BLAST (27) search using know salmonid and other fish *IL2* at the National Center for Biotechnology Information (NCBI) identified a salmon WGS contig (acc. no. AGKD04000795) that could encode for a second *IL2* (termed *IL2B* thereafter). Exons were predicted and primers were designed to the 5′-untranslated region (UTR) (sIL2BF1–2, Table S1 in Supplementary Material) and used for 3′- RACE (rapid amplification of cDNA ends) using salmon SMART cDNA prepared from PMA and CI stimulated blood samples, as described previously (28). A 0.9 kb 3′-RACE product was obtained, cloned, sequenced, and encoded the salmon IL-2B (acc. no. HE805272). A cDNA fragment of 0.75 kb was also amplified from trout cDNA using salmon primers sIL2BF1 and R1 (Table S1 in Supplementary Material) and encoded trout IL-2B (acc. no. HE805273).

#### Sequence Analysis

The nucleotide sequences generated were assembled and analyzed with the AlignIR program (LI-COR, Inc.). Homology search was performed at NCBI using the BLAST program1 (27). The gene organization was predicted using the Spidey program at NCBI. Protein prediction was performed using software at the ExPASy Molecular Biology Server2 (29) and signal peptides were predicted using the SignalP4.0 program (30). Disulfide bonding and cysteine connectivity were predicted using the DISULFIND program3 (31). Protein secondary structure was predicted using Jpred4 program4 (32). Global sequence comparison was performed using the scoring matrix BLOSUM62 within the MatGAT program, with a gap open penalty of 10 and gap extension penalty of 1 (33). Multiple sequence alignments were generated using CLUSTALW (34). The synteny of IL-2 loci was analyzed using Genomicus (database version 75.01) (35). A neighbor-joining phylogenetic tree was constructed on full-length aa multiple alignments using the MEGA7.0 software (36). The evolutionary distances were computed using the JTT matrix-based method with all ambiguous positions removed for each sequence pair.

#### Tissue Distribution of Rainbow Trout *IL2* Paralogs

Six healthy rainbow trout (~140 g) were killed and 17 tissues (blood, thymus, gills, scales, skin, muscle, tail fins, adipose fin, brain, adipose tissue, spleen, liver, heart, intestine, gonad, HK, and caudal kidney) were collected and processed as described previously (24, 25). The RNA preparation, cDNA synthesis, and real-time PCR analysis of gene expression were also as described previously (37). The primers for real-time PCR were designed, so that at least one primer crossed an intron to ensure that genomic DNA could not be amplified under the PCR conditions used. The primers for qPCR expression analysis of *IL2* paralogs are detailed in Table S1 in Supplementary Material and for other genes in Table S3 in Supplementary Material. To directly compare the expression level of the different *IL2* paralogs, a reference was

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

<sup>2</sup>http://www.expasy.org/tools.

<sup>3</sup>http://disulfind.dsi.unifi.it.

<sup>4</sup>http://www.compbio.dundee.ac.uk/jpred.

constructed using equal molar amounts of PCR product from each gene, including the house keeping gene *elongation factor-1α* (*EF1α*). The relative expression level of each sample was normalized against the expression level of *EF1α*.

#### PBL Preparation

The PBL were prepared by hypotonic disruption of erythrocytes, using a method modified from the study by Crippen et al. (38). The detailed protocol and characterization of the PBL has been reported elsewhere (39). Briefly, blood was withdrawn from the caudal vein of rainbow trout (200–500 g/fish) using a BD Vacutainer Plus blood collection tube (with Lithium heparin, BD, UK). The red blood cells were lysed by combining blood and ice-cold cell culture grade water at a ratio of 1:9 and mixed for 20 s. The osmotic pressure was brought back by adding 10× PBS (Sigma, UK). The resultant PBL preparation was kept on ice for 5–10 min to allow the cell debris to settle. The PBL were then separated from cell debris by passing through an EASYstrainer (70 µm, Greiner Bio-One, UK), pelleted by centrifugation (200 *g*, 5 min), and washed once with incomplete cell culture medium [Leibovitz medium L-15 (Life Technologies) supplemented with 100 IU/ml penicillin, 100 µg/ml streptomycin, and 1% fetal calf serum (FCS, Sigma, UK)]. The resultant PBLs were resuspended in complete cell culture medium (as above except 10% FCS) at 2 × 106 PBL/ml, ready for culture or stimulation.

## Modulation of the Expression of Trout *IL2* Paralogs in PBL

Freshly prepared PBL isolated as described above were stimulated with the T cell mitogen PHA (at 10 µg/ml), for 4, 8, 24 and 48 h. The PBL were also stimulated with PMA (50 ng/ml) or CI A23187 (100 ng/ml) alone or together for the same time points. The stimulation was terminated by dissolving the cells in TRI reagent (Sigma, UK). Quantification of gene expression was as described above. *IL2* expression was expressed as arbitrary units where the expression level in control cells at 4 h equals 1.

#### Modulation of the Expression of Trout *IL2* Paralogs in PBL and Primary HK Cells During the MLR

A MLR from three individual fish was set up to increase the magnitude and decrease the variation of the *in vitro* response (40). PBL prepared as above or HK cells isolated as described previously (24, 25, 41) were resuspended at 2 × 106 leukocytes/ ml and mixed following a loop design where an equal number of cells from 6 fish (1–6) were mixed as follows: 1/2/3, 2/3/4, 3/4/5, 4/5/6, 5/6/1, and 6/1/2. The cells from individual fish at the same density were used as control. The cells were seeded in 12-well cell culture plates at 2 ml/well or in 6-well plates at 3 ml/well. The plates were then sealed (Thermo Fisher Scientific, UK) and incubated at 20°C. The cell cultures were terminated at 4–120 h by dissolving in TRI reagent and *IL2* gene expression was quantified as above. A fold change was calculated as the average expression level of mixed cells to that of cells from individual fish at the same time point.

#### Cloning, Expression, and Purification of Recombinant Trout IL-2 Isoforms Cloning

The sequences encoding the two trout IL-2 mature peptides were amplified from cloned cDNA using the primers detailed in Table S1 in Supplementary Material. The amplified products were cloned to a pET vector (Novagen) as described previously (24, 37, 42–44). Each construct has a His-tag (MGSHHHHHHHHS) at the N-terminus for translation initiation and purification. Thus, the recombinant trout IL-2A and IL-2B were 134 aa and 129 aa, with a calculated molecular weight/theoretical pI of 15.1 kDa/5.96 and 14.5 kDa/7.80, respectively.

#### Expression and Purification of Trout IL-2 Isoforms

For each protein, a sequence confirmed plasmid was transformed into BL21 Star (DE3) competent cells (Invitrogen). The induction of recombinant protein production, purification under denaturing conditions, refolding, re-purification under native conditions, SDS-PAGE analysis of proteins, and quantification of protein concentration were as described previously (24, 37, 42). The refolding buffer contained 50 mM Tris–HCl, pH 7.0 (for IL-2B) or pH 8.0 (for IL-2A), 10% glycerol, 0.5 M arginine monohydrochloride, 0.5% Triton X-100, 0.2% PEG3350, and 10 mM 2-mercaptoethanol (2-ME). The purified proteins were desalted in desalting buffer [50 mM Tris–HCl, pH 7.0 (for IL-2B) or pH 8.0 (for IL-2A), 140 mM NaCl, 10 mM arginine, 50% glycerol, and 5 mM 2-ME] using PD-10 Desalting Columns (GE Healthcare). After sterilization with a 0.2-µm filter, the recombinant proteins were aliquoted and stored at −80°C ready for stimulation of cells.

## Modulation of Gene Expression in PBL by Recombinant Trout IL-2 Isoforms

The recombinant proteins produced above were initially added to HK cells (2 × 106 cells/ml) at 0.2–500 ng/ml for 24 h and the expression of *IFNγ* and *TNFα* analyzed. Further analysis of their bioactivity was focused in PBL using a dose of 200 ng/ml that showed good responses in HK cells. Freshly prepared PBL were stimulated with IL-2A, IL-2B, or both at 200 ng/ml for 4, 8, 24 and 48 h. The stimulation was terminated by dissolving the cells in TRI reagent, and real-time PCR analysis was conducted as described above. The expression of 75 trout immune genes, including those encoding for cellular markers, antimicrobial peptides, cytokines, and cytokine receptors, was analyzed. The primer information is detailed in Table S3 in Supplementary Material. The expression of each gene was first normalized to that of *EF1α*, and expressed as arbitrary units where one unit equals the average expression level in the control samples at 4 h. To give an estimation of constitutive expression in PBL, Δcp, the average cp (crossing point at which the fluorescence crosses the threshold during qPCR, *N* = 4) of a target gene minus that of the house keeping gene *EF1α* in control cells at 4 h is provided in Table S3 in Supplementary Material. A higher Δcp value indicates a lower expression level.

#### Proliferation Assay

Peripheral blood leukocyte proliferation was quantified by measuring BrdU incorporation during DNA synthesis in replicating cells using a Cell Proliferation ELISA, BrdU (colorimetric) kit (Sigma, UK) as per the manufacturer's instructions. Briefly, PBL from each fish in complete cell culture medium, at 4 × 105 cells/ well, were cultured in 96-well cell culture plates in the presence of 200 ng/ml of IL-2A or IL-2B. A control without IL-2 and a blank control without cells were also included. Three replicate wells were used for each treatment. The plates were then sealed and incubated at 20°C for 3 days. BrdU at 10 µM was added 20 h before fixation. The cell culture medium was removed after centrifugation (400 *g*, 5 min) and the cells fixed and DNA denatured by adding FixDenat solution. Anti-BrdU-peroxidase was then added to bind to BrdU incorporated in newly synthesized cellular DNA and detected using Tetramethylbenzidine. The color reaction was read at 450 nm using an ELISA plate reader (SoftMax Pr0 5.3). To calculate a stimulation index, the average OD450 of triplicates from each fish was first subtracted from the background value (without cell blank control). A stimulation index was calculated as the resulting OD450 of IL-2-stimulated cells divided by that of untreated samples.

#### Phagocytic Assay

Peripheral blood leukocytes in complete cell culture medium prepared above (2 × 106 cells/ml) were added to 12-well suspension cell culture plates (Greiner Bio-One) and incubated at 20°C. The fresh PBL were stimulated with recombinant IL-2A, IL-2B, or medium alone as control. Fluorescent latex beads (FluoSpheres Fluorescent Microspheres yellow green fluorescent, 1.0 μm, Life Technology) were added 24 h later at a cell/bead ratio of 1:20, and incubated for a further 3 h. The cells were harvested using 0.5% trypsin–EDTA (GIBCO) and the supernatant removed by centrifuging at 400 *g* for 3 min. Non-ingested beads were removed by centrifuging (100 *g* for 10 min at 4°C) over a 3% BSA and 4.5% d-glucose cushion prepared with FACS buffer (HBSS supplemented with 2% FCS, 5 mM EDTA, and 0.1% sodium azide). Cells were washed with FACS buffer and analyzed with a C6 Accuri Flow Cytometer, measuring at least 75,000 cells after live cell gating according to the FCS/SSC.

#### Statistical Analysis

The data were analyzed statistically using the SPSS Statistics package 24 (SPSS Inc., Chicago, IL, USA). The real-time PCR data were scaled and log2 transformed before statistical analysis as described previously (37). One-way analysis of variance and the LSD *post hoc* test were used to analyze expression data in MLR, with *p* ≤ 0.05 between mixed and control groups considered significant. For the tissue distribution of expression and other *in vitro* experiments that consisted of sample sets from individual fish, a paired samples *T*-test was applied.

#### RESULTS

## Two Divergent *IL2* Paralog Arose From the Salmonid-Specific 4R WGD

#### Cloning and Sequence Analysis of Salmonid *IL2*

In addition to the known salmonid *IL2A* in rainbow trout (trout), a second *IL2* (*IL2B*) has been cloned in both trout and Atlantic salmon (Atlantic) (Figures S1 and S2 in Supplementary Material). Analysis of Atlantic salmon WGS contig (acc. no. AGKD04000200) resulted in the prediction of an Atlantic salmon *IL2A* with a translation that differs by two aa from the prediction in the database (acc. no. EU816603, Figure S3 in Supplementary Material). Further analysis of the recently released WGS of coho salmon (coho), chinook salmon (chinook), and Arctic char (char) resulted in the prediction of both *IL2A* and *IL2B* in all the species (Figures S4–S9 in Supplementary Material). The salmonid *IL2* sequences are summarized in Table S2 in Supplementary Material.

#### Gene Organization of Salmonid *IL2*

Both the *IL2A* and *IL2B* genes, identified in the five salmonids, have a four-exon gene organization with three phase 0 introns (**Figure 1**; Figures S1–S9 in Supplementary Material). A fourexon organization is typical of mammalian *IL2* genes, and *IL2* and *IL2L* genes from other fish species (**Figure 1**). The mammalian IL-2 protein has a four helix-bundle structure (helices A–D), and this structure is predicted for each of the salmonid IL-2 molecules using the Jpred 4 program (32). Exon 1 encodes for the signal peptide and helix A, exon 2 for a large AB loop, exon 3 for helices B and C, and exon 4 encodes for a large CD loop and helix D. Lineage-specific and paralog-specific exon size differences are apparent. Mammalian *IL2* genes have a large coding region in exon 1 and a large exon 2, compared with fish *IL2* orthologs. Salmonid *IL2A* genes have a relatively smaller exon 2 (36 vs 45–48 bp) but larger exon 3 (153 vs 123–126 bp) relative to salmonid *IL2B*. The percomorph *IL2* genes also have a larger exon 3 (144–156 vs 123–123 bp) but in *IL2L* genes it is smaller (**Figure 1**). In general, salmonid *IL2A* has more similarity to *IL2* from other fish species, while salmonid *IL2B* is similar to *IL2L*.

#### Nucleic Acid and aa Sequence Analysis of Salmonid *IL2*

Each salmonid *IL2* gene has four to seven ATTTA mRNA instability motifs in the 3′-UTR, suggesting that salmonid *IL2* mRNA is unstable. Each salmonid *IL2* gene encodes for 135 −147 aa with a predicted signal peptide of 20 aa, a mature peptide of 115 −127 aa and a molecular weight of 12.8–14.5 kDa (Table S2 in Supplementary Material). The isoelectric point (pI) of the mature peptide of salmonid IL-2A is acidic (4.65–5.11) as seen in other fish species (Tables S2 and S4 in Supplementary Material). However, the pI of salmonid IL-2B is relatively high (6.79–7.87). One to four N-glycosylation sites can be predicted in each salmonid IL-2 peptide, with the site in the CD loop conserved in salmonid IL-2A (**Figure 2A**; Figure S10 in Supplementary Material).

There are six conserved cysteine residues in salmonid IL-2A but only four in IL-2B (Table S2 and Figure S10 in Supplementary Material). These cysteine residues are conserved in seven positions (C1–7) in a paralog-specific manner. IL-2A contains the first six cysteines (C1–6) potentially forming three intra-molecular disulfide bonds (C1/C4, C2/C5, and C3/C6, Figure S10 in Supplementary Material), as predicted using DISULFIND program (31). In addition to three conserved cysteine residues

and Figures S1–S9 in Supplementary Material. The human, cow, and rat *IL2* gene organization was extracted from Ensemble genes ENSG00000109471, ENSBTAG00000020883, and ENSRNOG00000017348, respectively. Other fish *IL2/IL2L* genes were extracted from the NCBI genomic sequences NC\_018903 (Fugu), CAAE01023259 (Tetraodon), and AANH01006550 (Stickleback). The aa sequence domains (signal peptide, helices A–D, AB, and CD loops) encoded by each exon are indicated above.

(C1, C3, and C5) present in IL-2A, IL-2B had a unique cysteine (C7). These four cysteine residues were predicted to form two distinct disulfide bonds (C1/C3 and C5/C7, Figure S10 in Supplementary Material).

A multiple alignment was produced using ClustalW (27) from the predicted mature peptide sequence of all salmonid IL-2A and IL-2B, other fish IL-2 and IL-2L, and selected mammalian IL-2 molecules (**Figure 2**). In general, there is high conservation in regions of the four helices but marked differences in the long AB and CD loops. There are conserved cysteine residues in eight positions (C1–8) over the alignment but none are conserved in all molecules and show lineage and paralog specificity. Mammalian IL-2 possesses three cysteine residues (C3, C6, and C8) that form a single disulfide bond (C3/C6) (**Figure 2**) important to stabilize its structure and bioactivity (45). Salmonid IL-2A and other fish IL-2, including the two common carp IL-2 isoforms, share the same six cysteine pattern (C1–6) that is predicted to form three disulfide bonds (C1/C4, C2/C5, and C3/C6) (**Figure 2B**). Salmonid IL-2B and other fish IL-2L all have four conserved cysteine residues potentially forming two disulfide bonds; however, the cysteine patterns are different. Salmonid IL-2B molecules possess C1, C3, C5, and C7, whereas IL-2L have C1, C2, C4, and C5 (**Figure 2B**).

#### Phylogenetic Tree Analysis of Salmonid *IL2*

The aa sequences of salmonid IL-2 orthologs, i.e., IL-2A or IL-2B, share high identity/similarity, e.g., salmonid IL-2B share 74.1–98.5%/80.4–99.3% aa identities/similarities. However, comparison of the salmonid IL-2 paralogs, i.e., IL2A vs IL-2B, revealed the aa sequence identities/similarities are lower, 38.9– 43.0%/59.62–67.9%, similar to values for salmonid type I and type II TNFα isoforms (with 45.2–47.5% identities) that arose from the 3R teleost-wide WGD (41), and suggested a possible 3R origin. To clarify this, a neighbor-joining phylogenetic tree was produced from a multiple alignment of salmonid IL-2, other fish IL-2 and IL-2L, and mammalian IL-2 molecules, along with the IL-2 close relatives IL-15 and IL-21 from selected fish and mammals. All fish IL-2, IL-2L, IL-2A, and IL-2B molecules were grouped with mammalian IL-2 and separated from IL-15 and IL-21 from mammals and fish (**Figure 3**), suggesting that all fish IL-2-related molecules are indeed orthologs of mammalian IL-2.

Figure 2 | Amino acid (aa) sequence alignment of IL-2 mature peptides from salmonids and selected fish and mammalian species (A) and predicted potential intra-molecular disulfide bonds (B). (A) The multiple alignment was produced using ClustalW, and conserved aa shaded using BOXSHADE (version 3.21). The four α helices (A–D) are indicated above the alignment. The conserved cysteine residues are in red and numbered at the bottom of the alignment (C1–8). The aa sequences and accession numbers are detailed in Supplementary Material: protein sequences. (B) Lineage- and paralog-specific conservation of cysteine residues in mammalian IL-2, salmonid IL-2A and other fish IL-2, and salmonid IL-2B and IL-2L is apparent. Cysteine residues potentially forming disulfide bonds are linked by green lines.

Salmonid IL-2A and IL-2B each formed an independent clade, which grouped together but was separate from other IL-2 molecules, a typical scenario for genes arising from the salmonid-specific 4R WGD (46). The two common carp IL-2 also grouped together suggesting a recent origin, perhaps from the carp-specific 4R WGD that occurred 5.6–11.3 million year ago (Mya, 44). The IL-2L, its gene only found in percomorphs neighboring the percomorph *IL2* (19) also formed an independent clade.

sequences and accession numbers are detailed in Supplementary Material: protein sequence.

#### Synteny Analysis of *IL2* Loci in Rainbow Trout and Atlantic Salmon

To clarify the origin of salmonid *IL2A* and *IL2B*, a synteny analysis was performed on the reference genome of rainbow trout, Atlantic salmon, fugu, stickleback, mouse, and chicken. The rainbow trout *IL2A* and *IL2B* were mapped to CH25 (Sequence ID: NC\_035101) and CH14 (Sequence ID: NC\_035090), respectively, while the Atlantic salmon *IL2A* and *IL2B* were mapped to CH9 (Sequence ID: NC\_027308) and CH5 (Sequence ID: NC\_027304), respectively. The *IL2* loci across vertebrates are syntenically conserved as shown in **Figure 4**. The genes (*Fat4*, *Ankrd50*, *Spry1*, *Spata5*, *Nudt6*, *Fgf2*, *Bbs12*, *Cetn2*, *IL21*, *Adad1*, *Bbs7*, *Ccna2*, and *Anxa5*) are syntenically conserved across all vertebrates with an insertion of genes (*Tlx1*, *Npm1*, *Fgf18*, *Pcgf1*, *Lbx1*, *Nanos1*, *Ap1ar*, *Prom1*, *Fgfbp1*, *Fgfbp2*, etc.) to one side of fish *IL2*. While the *IL2A* and *IL2B* are located in two homologous CH regions in both rainbow trout and Atlantic salmon, the *IL2* and *IL2L* genes in other teleosts are adjacent (19) within the syntenically conserved *IL2* loci (**Figure 4**). This chromosomal organization of *IL2* loci suggests that the divergent salmonid *IL2A* and *IL2B* arose from the salmonid-specific 4R WGD event,

Figure 4 | Gene synteny at the *IL2* loci in salmonids in comparison to other teleosts (Fugu, stickleback), mammals (mouse), and birds (chicken). The mouse and chicken *IL2* loci were analyzed using the Genomicus program (database version: 90.01). The information for salmonid and fugu *IL2* loci was extracted from NCBI reference genome sequences NC\_035101 (Trout *IL2A*), NC\_035090 (Trout *IL2B*), NC\_027308 (Atlantic salmon *IL2A*), NC\_027304 (Atlantic salmon *IL2B*), and NC\_018903 (Fugu, *Takifugu rubripes*). The stickleback *Gasterosteus aculeatus IL2* loci was extracted from Ensembl database (https://www.ensembl.org/ Gasterosteus\_aculeatus/Location/View?r=groupIV:2800000-3700000). Arrows indicate transcriptional direction. *IL2* genes are in red. The genes syntenically conserved across vertebrates are in green (upstream of *IL2*) or yellow (downstream of *IL2*), and those conserved in teleosts only are in blue.

in contrast to other teleost *IL2* and *IL2L* that were due to local gene duplication.

#### Comparative Tissue Transcript Analysis of *IL2* Paralogs in Rainbow Trout

To gain insights into the potential function of the salmonid *IL2* paralogs, their expression was examined in 17 tissues of rainbow trout. The expression of both paralogs was detectable in all the tissues albeit at different levels. *IL2A* expression was highest in the immune organs thymus and spleen: high in gills, HK, caudal kidney and intestine, and low in muscle and liver (**Figure 5**). *IL2B* expression was also highest in immune organs (spleen, thymus, HK and gills) and lowest in liver. It is noteworthy that the expression of *IL2B* was lower than *IL2A* in all tissues examined (*p* ≤ 0.05, paired samples *T*-test).

## Modulation of *IL2* Paralog Expression in Rainbow Trout PBL by PHA, PMA, and CI

The modulation of trout *IL2* paralogs by the T cell activator PHA was next investigated in freshly prepared PBL. Trout *IL2A* and *IL2B* were lowly expressed constitutively in PBL as indicated by a ΔCP of control PBL at 4 h of 18.0 and 22.5, respectively (Table S3 in Supplementary Material). The expression of both *IL2* paralogs was induced by PHA from 4 to 48 h and peaked at 24 h (224-fold increase for *IL2A* and 457-fold for *IL2B*, **Figures 6A,B**).

Signaling *via* the TCR is believed to result in various biochemical events that include a rise in intracellular free calcium and activation (translocation) of protein kinase C. These two signals also can be generated by CI A23187 and by activators of protein kinase C, such as PMA (10). Thus, *IL2* expression in PBL was further investigated after stimulation with PMA and CI alone or in combination. Trout *IL2* was induced by CI from 4 to 48 h and reached the highest induction at 48 h (225-fold for *IL2A* and 77-fold for *IL2B*) (**Figures 6C,D**). PMA alone was a relatively weak inducer of *IL2* expression. It induced *IL2A* expression 28-fold at 24 h and 35-fold at 48 h, while *IL2B* expression was modulated from 8 h and reached a 24-fold increase at 48 h (**Figures 6C,D**). However, PMA and CI could synergize to upregulate *IL2* expression. Both paralogs had markedly higher expression in samples stimulated by the combination of PMA and CI from 4 to 48 h vs that seen in samples stimulated with PMA or CI alone (6,126-fold increase for *IL2A* and 25,207-fold for *IL2B* at 48 h) (**Figures 6C,D**).

mean (+SEM) relative expression is presented as arbitrary units where one unit equals the average expression level in control PBL at 4 h. The outcome of a paired samples *T*-test between stimulated samples and controls at the same time point is shown above the bars as \**p* ≤ 0.05, \*\**p* ≤ 0.01, and \*\*\**p* ≤ 0.001. The expression levels in PMA + CI stimulated samples are significantly higher than either samples stimulated with PMA or CI alone at all the time points (*p* ≤ 0.05).

#### Modulation of *IL2* Paralog Expression by the MLR

The MLR is due to T-cell activation by alloantigens presented by APCs. Three individuals were mixed to increase the magnitude and decrease the variation of the *in vitro* response (40). Compared with PBL cultured from individual fish, trout *IL2A* expression was significantly induced by the MLR from 24 to 96 h, and *IL2B* expression from 8 to 96 h (**Figure 7**). Similarly, MLRinduced *IL2* expression was also seen in HK cells (Figure S11 in Supplementary Material).

and single PBL at the same time point is shown above the bars as \**p* ≤ 0.05, \*\**p* ≤ 0.01, and \*\*\**p* ≤ 0.001.

#### Bioactivity of Trout Recombinant IL-2A and IL-2B in PBL

Recombinant trout IL-2A and IL-2B were expressed in *E. coli* after IPTG induction and purified under denaturing conditions (Figure S12 in Supplementary Material), with extensive washing with a buffer containing 1.5% Triton X-100 to remove LPS. The proteins were then refolded and re-purified. The purified IL-2 proteins, at up to 2 µg/ml, were ineffective at inducing the expression of TNF-α1, IL-1β, and COX-2, classic inflammatory genes that are known to be upregulated by LPS (41, 42) in the macrophage cell line RTS-11. When added to HK cells for 24 h, both proteins induced the expression of *IFNγ1* and *IFNγ2* from 2 to 500 ng/ml, and *TNFα2* from 20 to 500 ng/ml (Figure S13 in Supplementary Material). The expression of *TNFα1* was again not affected, suggesting that LPS contamination is negligible. The dose at 200 ng/ml of IL-2 isoforms induced a good response in HK cells and was used for analysis of IL-2 bioactivities in PBL in a time-course experiment to investigate the early (4 h), intermediate (8 and 24 h) and long-lasting (48 h) effects of IL-2 isoforms.

#### Modulation of Th1 Pathway Gene Expression by IL-2 Isoforms in Rainbow Trout PBL

IL-12 and IFNγ are critical cytokines that initiate the downstream signaling cascade *via* their receptors and the master transcription factor T-bet to develop Th1 cells to secrete IFNγ, TNFα, and IL-2 (3). Two *IFNγ* genes (47), *IFNγ1* and *IFNγ2*, present in salmonids due to the salmonid WGD were both rapidly induced by both IL-2 isoforms, alone or in combination, from 4 h until 48 h (**Figures 8A,B**). Similarly, the expression of *TNFα2*, *IL12P40B2*, and *IL12P40C* (48, 49) was also induced from 4 to 48 h (**Figures 8E,G,H**). However, *TNFα1* and *TNFα3* expression was less responsive, with relatively higher induction levels seen at 24 h (TNFα3 only) and 48 h (**Figure 8D**; Figure S14A in Supplementary Material). Of the other TNF family members studied, *FasL* was refractory to IL-2 isoform stimulation but *CD40L* was induced to a small extent at 4 and 48 h (Figures S14B,C in Supplementary Material). Within the IL-12 family, the expression of the alpha chains, *IL12P35A2* (Δcp = 20.2), *IL23P19* [(50), Δcp = 21.8] and *IL27A* [(51), Δcp = 19.7] was low in PBL and not significantly modulated by IL-2 isoforms. The expression of *IL12P35A1* was higher (Δcp = 15.1) and induced at 4 h by IL-2A but not by IL-2B (**Figure 8F**). No induction was seen for *IL12P40B1* and *EBI3* (Figures S14F,G in Supplementary Material). *CXCL11L1* (52), a downstream effector of IFNγ signaling, was upregulated by the IL-2 isoforms at late time points (**Figure 8C**). *IFNγR2* (53) expression was induced to a small extent in IL-2A (24–48 h) and IL-2B (8–48 h) stimulated PBL; however, *IFNγR1* and *T-bet* (54) were refractory (Figures S14D,E,H in Supplementary Material). *IFNγ* can be induced by IL-15 (43), IL-18 (55), and IL-21 (37). None of these cytokines was modulated in PBL by the IL-2 isoforms except for a downregulation of *IL15* at 8 h by both isoforms and a downregulation of *IL21* at 24 h by IL-2A (Figures S14I–K in Supplementary Material).

The ability of the trout IL-2 isoforms to modulate the expression of Th1 pathway genes was generally overlapping, but subtle differences were observed. The expression of *IFNγ1*, *IFNγ2*, *CXCL11L1*, and *TNFα3* was higher in IL-2B stimulated PBL at some time points. By contrast, the expression of *IL12P35A1* was induced only by IL-2A (**Figure 8**). The modulated gene expression seen with the combination of the two isoforms was mostly IL-2B-like with the exception of *IL12P35A1* that is IL-2A like (**Figure 8**). These patterns of activity suggest that the IL-2 isoforms may signal *via* the same cell surface receptors independently. The signal intensity is likely determined by the presence

of specific combinations of receptor subunits and their affinity to the IL-2 isoforms that will be discussed later.

#### Modulation of Th2 Pathway Gene Expression by IL-2 Isoforms in Rainbow Trout PBL

The expression of Th2 cytokines *IL4/13B1* and *IL4/13B2*, and their receptor *IL4Rα1* (44) was upregulated in stimulated PBL from 4 to 48 h with no difference between IL-2B and the combination of IL-2A and IL-2B (**Figures 9B–D**). However, at some time points, e.g., *IL4/13B1* and *IL4Rα* at 8 h, and *IL4/13B2* at 48 h, the induced expression was higher in PBL stimulated by both IL-2A and IL-2B than by either alone. In agreement with previous studies showing IL4/13A expression was less responsive to T cell stimulation (24), *IL4/13A* expression was refractory to both isoforms from 4 to 24 h and showed a small decrease at 48 h in PBL stimulated with IL-2B or both isoforms (**Figure 9A**). The expression of *IL4Rα2*, the abundant *IL4Rα* isoform (Δcp of 11.4 compared with 14.4 of *IL4Rα1*, Table S3 in Supplementary Material) was also upregulated to a small extent at 4 and 8 h by the combination of IL-2A and IL-2B, at 8 h by IL-2A, and at 8 and 24 h by IL-2B (**Figure 9E**). The expression of the Th2 master transcription factor *GATA3* was maintained at 8 h after IL-2 stimulation when its expression in unstimulated PBL had decreased, but no difference was seen at other time points (**Figure 9F**).

#### Modulation of Th17 Pathway Gene Expression by IL-2 Isoforms in Rainbow Trout PBL

*IL17A/F1A*, *IL17A/F2A*, *IL17A/F3* (25), and *IL22* (56) genes were lowly expressed in PBL with a Δcp of 19.7, 21.3, 20.4, and 22.6, respectively (Table S3 in Supplementary Material). Their expression was not modulated by either isoform alone or together (Figures S15A–C in Supplementary Material). *IL17C1*, *IL17C2* (57), and *IL17D* were also lowly expressed in PBL (Table S3 in Supplementary Material) and are not described further (Figure S15D in Supplementary Material). The expression of the Th17 master transcription factor *RORγ* (58) was slightly downregulated at 8 h by IL-2A and at 48 h in the presence of both IL-2 isoforms (Figure S15E in Supplementary Material). *IL17Rα* (59) was refractory (Figure S15F in Supplementary Material).

#### Modulation of Pro-Inflammatory Gene Expression by IL-2 Isoforms in Rainbow Trout PBL

In general, the effects of the IL-2 isoforms on the expression of pro-inflammatory cytokines (*IL1β*, *IL6*, *IL8*, *IL11*, *IL34*, and *M17*) were minor (Figure S16 in Supplementary Material). Three *IL1β* are present in rainbow trout (60). *IL1β2* was lowly expressed in PBLs (Δcp = 21.8) and refractory. *IL1β1* expression was inhibited at 8 h by IL-2B but a small increase of *IL1β3* was seen at 4 h in the presence of both isoforms (Figures S16A,B in Supplementary

Material). nIL1Fm (61) expression was also increased to a small extent at 4 h by IL-2 isoforms alone and maintained to a higher level at 48 h with all three treatments (Figure S16C in Supplementary Material). *IL6* expression was refractory, while *IL8* (62) expression was inhibited at 4 h by both IL-2 isoforms alone, and *IL11* (63) was inhibited at 8 h by their combination (Figures S16D–F in Supplementary Material). The expression of *M17* (64) was increased at 8 and 48 h by all three treatments (Figure S16G in Supplementary Material), and a small increase in *IL34* (65) expression at 24 h was induced by IL-2B or both isoforms (Figure S16H in Supplementary Material). Relatively higher-level expression was seen for *M17* at 4 h and *IL34* at 24 and 48 h in PBL stimulated with IL-2B vs IL-2A (Figures S16G,H in Supplementary Material).

#### Modulation of Treg Pathway Gene Expression by IL-2 Isoforms in Rainbow Trout PBL

Small changes in expression of regulatory pathway genes were seen at some time points (**Figure 10**). *TGFβ1A* expression was increased to a small extent at 8 h by IL-2B but was downregulated at 48 h by IL-2B or IL-2B + IL-2A (**Figure 10A**). *TGFβ1B* (66) was also downregulated, at 4 h by both IL-2 isoforms (**Figure 10B**). However, *IL10A* and *IL10B* (67) were induced from 4 to 24 h by both IL-2 isoforms alone and together (**Figures 10C,D**). Interesting, the expression of the master transcription factors for Treg cell development, *FOXP3A* and *FOXP3B* (68), was increased from 4 to 48 h by all three treatments (**Figures 10E,F**).

#### Modulation of IL2 and IL2R Gene Expression by IL-2 Isoforms in Rainbow Trout PBL

*IL2A* was weakly induced at 24 and 48 h by IL-2B or the combination of IL-2A and IL-2B (**Figure 11A**). *IL2B* was also increased at 48 h by IL-2A or IL-2B alone (**Figure 11B**). The putative *IL2Ra CD25L* expression decreased over time *in vitro* but was sustained in the presence of IL-2 isoforms (**Figure 11C**). *γC1*, the broadly and abundantly expressed isoform, was only weakly induced at 4 and 8 h by IL-2B (Figure S14L in Supplementary Material), whereas *γC2*, that is less expressed constitutively (Δcp = 12.6 vs 5.6 of *γC1*, Table S3 in Supplementary Material) but more inducible by T-cell stimulants (13), was highly induced from 4 to 48 h by both IL-2 isoforms alone or together (**Figure 11D**). In contrast, the expression of *IL2Rβ1* was downregulated at 4 h by both IL-2 isoforms (alone or together) but maintained at higher levels in the presence of IL-2 when a decrease was seen in the control cells at later timings (**Figure 11E**). A similar expression pattern was observed for *IL2Rβ2* (**Figure 11F**).

#### Modulation of T Cell Marker Gene Expression by IL-2 Isoforms in Rainbow Trout PBL

The expression of T cell markers (*CD4-1*, *CD4-2A*, *CD4-2B*, *CD3ε*, *CD8α*, and *CD8β*) in PBL *in vitro* without IL-2 decreased

over time (**Figure 12**). Although there was no induction of these genes at 4 h (except for a small increase in *CD4-2B* by IL-2B), the expression of these T cell markers was maintained at almost constant levels by IL-2. Thus, the expression of CD4-1 was higher from 8 to 48 h in PBL treated with IL-2 compared with respective controls. This was also seen for *CD4-2A* and *CD4-2B*, particularly at 48 h (**Figures 12A–C**). The effect on *CD3ε* expression was less apparent but higher levels of *CD8α* and *CD8β* were seen at 48 h in PBL treated with IL-2B or IL-2A plus IL-2B (**Figures 12D–F**). The induction of *CD4-2A* at 24 h, and *CD8α* and *CD8β* at 48 h was greater in PBL treated with IL-2B or the combination of IL-2A and IL-2B.

24 h, and higher than with IL-2B for IL10B at 4 h (*p* ≤ 0.05).

#### Modulation of Chemokine Receptor Gene Expression by IL-2 Isoforms in Rainbow Trout PBL

The expression of the chemokine receptors, such as *CXCR2* and *CXCR3B* (69), was increased in PBL from 4 to 48 h by IL-2A and IL-2B treatment (alone or in combination) (**Figures 13B,D**). A small induction was also seen for *CXCR1* at 4 h in PBL treated with IL-2A alone (**Figure 13A**). *CXCR3A* expression was downregulated at 4 h by IL-2B treatment but was sustained at higher levels in IL-2 treated PBL at 48 h when it was decreased in control cells (**Figure 13C**). A small induction of the expression of *CCR7A* was seen at 8 and 48 h, and *CCR7B* at 4, 24, and 48 h mainly by IL-2A or the combination of IL-2A and IL-2B (**Figures 13E,F**). The expression of *CXCR1* and *CXCR3A* at 4 and 8 h, and *CCR7A*

and *CCR7B* at 8 h was higher in IL-2A-treated PBL. By contrast, the expression of *CXCR2* at 24 and 48 h was higher in IL-2Btreated PBL (**Figure 13**).

### IL-2 Isoforms Differentially Modulate Cathelicidin Gene Expression and Promote Phagocytosis of PBL

The phagocytic activity of PBL cultured *in vitro* with/without IL-2 isoforms was analyzed by flow cytometry using fluorescent beads (**Figure 14A**). 5% of the lymphoid cells were phagocytic in control PBL. This percentage was not affected by IL-2 treatment (**Figure 14B**). 26% of the myeloid cells were phagocytic in control PBL. This percentage was increased significantly to 37% by IL-2B stimulation (**Figure 14B**) but the increase seen with IL-2A treatment was not statistically significant (*p* = 0.07). In addition, more cells ingested more than one bead in the IL-2-stimulated samples (**Figure 14A**). Hence, the mean fluorescence intensity (MFI) of myeloid cells was increased significantly from 2.9 × 106 in control samples to 3.2 × 106 and 3.5 × 106 in IL-2A and IL-2B stimulated samples, respectively (**Figure 14C**). The MFI was not altered in phagocytic lymphoid cells after IL-2 treatment. Importantly, the expression of *cathelicidin* (*CATH*)-*1*, known to enhance phagocytic activity in rainbow trout (70) was upregulated from 4 to 48 h by both IL-2 isoforms alone or together (**Figure 14D**). In contrast, the expression of *CATH2* was downregulated by IL-2

isoforms, when used individually at 8 h or together at 4, 8, and 48 h (**Figure 14D**). The differential modulation of *CATH1* and *CATH2* expression has also been seen with other cytokines, e.g., IL-4/13 (24), IL-6 (42), and TNFα (41) in rainbow trout.

#### Recombinant Trout IL-2 Isoforms Promote Growth of PBL

Mammalian IL-2 is a lymphocyte growth factor and so we examined the proliferation of PBL after stimulation with IL-2 isoforms. BrdU incorporation was increased in PBL treated with IL-2B alone and IL-2A + IL-2B (**Figure 15**). While no significant increase was seen using IL-2A-treated PBL, an intermediate level of proliferation was present that did not differ from the IL-2B treated cells.

#### DISCUSSION

In this study, two divergent *IL2* paralogs that arose from the salmonid WGD have been characterized molecularly in five salmonid species for which a sequenced genome is available. Expression and bioactivity analysis of these two *IL2* paralogs was then undertaken in one of these species, the rainbow trout *O. mykiss*. Our results suggest that IL-2 is an important T cell factor that regulates Th1 and Th2 pathways in fish and shed light on lineage-specific expansion, evolution, and functional divergence of *IL2* orthologs and paralogs.

### Lineage-Specific Expansion, Functional Divergence, and Convergent Evolution of *IL2*

Vertebrate animals emerged from the invertebrates approximately 500 Mya *via* two sequential rounds of WGD (2R). In fish, there was a further teleost-specific WGD (3R) approximately 300 Mya, with several individual lineages having further WGD events, as occurred in the salmonids (4R) approximately 95 Mya and in carp (4R) 5.6–11.3 Mya (46, 71). Many immune genes are retained after a WGD that may have a beneficial role. Thus, a mammalian cytokine gene can have two paralogs in 3R teleosts and up to four paralogs in 4R fish, as seen with *IL1β* (60), *TNFα* (41), and *IL12* (49). The two *IL2* paralogs present in salmonids share only low aa sequence identity, a range similar to 3R paralogs (41, 49). However, our phylogenetic tree and synteny analysis clearly indicate that the salmonid *IL2A* and *IL2B* arose from the 4R salmonid WGD. The two common carp *IL2* paralogs, sharing 51% aa identity, might also have arisen from the 4R WGD event in this lineage (71). By contrast, the two *IL2* paralogs found only in percomorphs, share even lower aa identity (Table S4 in Supplementary Material) but are located at the same genomic site and clearly arose from a local gene duplication event in this lineage. Thus, the expansion of *IL2* in teleosts originated from lineage-specific pathways, local gene duplications, and 4R WGD.

Figure 12 | Modulation of T cell marker gene expression by IL-2 isoforms in peripheral blood leukocytes (PBL). Freshly prepared PBL from four fish were stimulated with IL-2A, IL-2B, IL-2A + IL-2B, or with medium alone as control. The expression of *CD4-1* (A), *CD4-2A* (B), *CD4-2B* (C), *CD3ε* (D), *CD8α* (E) and *CD8β* (F) was quantified and presented as in Figure 8. Significant results of a paired samples *T*-test between the stimulated samples and time-matched controls are shown above the bars as \**p* ≤ 0.05, \*\**p* ≤ 0.01, and \*\*\**p* ≤ 0.001. Different letters over the control bars indicate significant differences over time in the unstimulated cells (*p* ≤ 0.05). The expression levels in samples treated with IL-2B or IL-2A + IL-2B are higher than IL-2A-treated samples for CD8α and CD8β at 48 h (*p* ≤ 0.05).

Figure 13 | Modulation of chemokine receptor gene expression by IL-2 isoforms in peripheral blood leukocytes (PBL). Freshly prepared PBL from four fish were stimulated with IL-2A, IL-2B, IL-2A + IL-2B, or with medium alone as control. The expression of *CXCR1* (A), *CXCR2* (B), *CXCR3A* (C), *CXCR3B* (D), *CCR7A* (E) and *CCR7B* (F) was quantified and presented as in Figure 8. Significant results of a paired samples *T*-test between the stimulated samples and time-matched controls are shown above the bars as \**p* ≤ 0.05, \*\**p* ≤ 0.01, and \*\*\**p* ≤ 0.001. Different letters over the control bars indicate significant differences over time in the unstimulated cells (*p* ≤ 0.05). The expression levels in samples treated with IL-2A are higher than with IL-2B for CXCR1 and CXCR3A at 4 h and CCR7A at 8 h, but lower for CXCR2 than with IL-2B or IL-2A + IL-2B at 24 h, and IL-2B at 48 h (*p* ≤ 0.05).

Figure 14 | Flow cytometry analysis of phagocytosis (A), percentage of phagocytic cells (B), mean fluorescence intensity (MFI) of phagocytic cells (C), and cathelicidin gene expression modulated by IL-2 isoforms in peripheral blood leukocytes (PBL) (D). (A) Trout PBL were incubated with IL-2A, IL-2B, or medium alone as control for 24 h. PBL were then incubated with 1.0-µm fluorescent beads for 3 h and analyzed by flow cytometry. Typical results from a single fish are shown. (B) The percentage of phagocytic leukocytes in lymphoid and myeloid gates. The results are presented as the average + SEM of three fish. "\*" indicates significant differences (*p* ≤ 0.5) of a paired samples *T*-test. (C) The MFI of phagocytic lymphoid and myeloid cells. The results are presented as the average + SEM of three fish. "\*" indicates significant differences (*p* ≤ 0.5) of a paired samples *T*-test. (D) Freshly prepared PBL were stimulated with IL-2A, IL-2B, IL-2A + IL-2B, or with medium alone as control for 4, 8, 24 and 48 h, and the expression of *CATH1* and *CATH2* quantified. The results are presented as the average + SEM from four fish. Significant results of a paired samples *T*-test between the stimulated samples and time-matched controls are shown above the bars as: \**p* ≤ 0.05, \*\**p* ≤ 0.01, and \*\*\**p* ≤ 0.001. Different letters over the control bars indicate significant differences over time in the unstimulated cells (*p* ≤ 0.05).

Figure 15 | Rainbow trout IL-2 isoforms promote proliferation in peripheral blood leukocytes (PBL). Freshly prepared PBL from four fish were incubated with IL-2A and IL-2B, or with medium alone as control in triplicate for 3 days. BrdU was added 20 h before incorporation of BrdU was detected by ELISA. The data are presented as the average (+SEM) stimulation index, calculated as the OD450 of IL-2-treated cells divided by that of untreated samples. Different letters over bars indicate significant differences (*p* ≤ 0.05, paired samples *T*-test).

Although little is known about the specific function of each paralog possessed in fish, the high aa sequence divergence of *IL2* paralogs (Table S4 in Supplementary Material) within and between different lineages suggests a functional diversification. The functional diversification, neo-/sub-functionalization, can be demonstrated by changes in temporal and spatial gene expression at the transcript level or by changes of protein function as seen with the rainbow trout paralogs that will be discussed later. The functional diversification can also be reflected by diversification of aa sequence impacting secondary and tertiary structure that may affect ligand–receptor binding and signaling. The disulfide-bonding potential in IL-2 and related molecules from different lineages is also worth noting. Mammalian IL-2 forms a single-disulfide bond to stabilize its structure and is critical for its bioactivity (45). The salmonid IL-2A, and other fish IL-2 including the two carp IL-2, have three predicted disulfide bonds while salmonid IL-2B and percomorph IL-2L have the potential to form two disulfide bonds. Whether more disulfide-bonding plays an important role in stabilizing fish IL-2 isoforms at low physiological temperature in a changing environment is an interesting research area for future studies.

Mammalian IL-2 signals *via* dimeric (IL-2Rβ/γC) or highaffinity trimeric (IL-2Rα/IL-2Rβ/γC) receptors. In the latter, when IL-2Rα binds IL-2 it stabilizes a secondary binding site for presentation to IL-2Rβ. γC is then recruited to the composite surface formed by IL-2/IL-2Rβ (10). In this complex, γC contacts IL-2 helices A and D, IL-2Rβ contacts IL-2 helices A and C, and IL-2Rα interacts mainly with the long AB loop. IL-2Rα forms the largest of the three IL-2/IL-2R interfaces (11). There are marked differences in the AB loop that is encoded by exon 2, between mammalian and fish *IL2* orthologs as seen by comparison of their gene organization and aa sequence (**Figures 1** and **2**) which reveals the AB loop is smaller in fish. This difference may suggest that the two IL-2 isoforms have evolved binding preferences (affinity) to a specific IL-2 receptor (assuming multiple IL-2 receptors in fish). While *IL2Rβ* and *γC* are present in fish with two copies in salmonids (13), the *bona fide IL2Rα* (CD25) with two Sushi domains has never been discovered in teleosts. However, CD25L or IL-15Rα, with a single Sushi domain, has been characterized and shown to bind IL-15 strongly, and IL-2 to some extent (14). Furthermore, in addition to an *IL15* gene, teleosts possess an *IL15-Like* (*IL15L*) gene that is present in some mammals but pseudogenized in humans and mice (15). Mammalian (cow) IL-15L also binds IL-15Rα. Thus, the fish IL-2, IL-15, and IL-15L may all share three types of IL-2R subunits, CD25L, IL-2Rβ, and γC, with two isoforms each in some fish species. Understanding how this receptor sharing by multiple ligands (e.g., five known receptor subunits, and five ligands, IL-2A, IL-2B, IL-15, IL-15LA and IL-15LB in salmonids) helps regulate the immune response in fish is a fascinating but challenging topic to fish immunologists.

Interestingly, marked differences are present in the two salmonid IL-2 paralogs regarding the size of exon 2 and exon 3. This difference may impact receptor-binding affinity/signaling, leading to functional diversification. It is noteworthy that the two paralogs present in salmonids and percomorphs that arose *via* different evolutionary pathways (4R WGD in salmonids and local gene duplication in percomorphs) share exon 3 size diversification, with a larger exon 3 in salmonid *IL2A* and percomorph *IL2*, and a smaller exon 3 in salmonid *IL2B* and percomorph *IL2L*. This may suggest convergent evolution has occurred.

While the pI of salmonid IL-2A and IL-2 in other teleosts is acidic, that of all salmonid IL-2B is 2.14–3.04 higher. Protein pI is the pH at which the molecule carries no net electrical charge and determines the net charge at a specific environmental pH. The structure, stability, solubility, and function of a protein depend on its net charge and on the ionization state of the individual residues, both of which depend on the pH of the surrounding environment (72). The salmonids studied belong to the Salmoninae and are anadromous fish that evolved in freshwater in the Northern Hemisphere (73). Two isoforms of IL-2 with a different pI may be beneficial in coping with the changing pH of freshwater and ocean environments.

It is noteworthy that the IL-2 paralogs discovered in percomorphs and in 4R WGD teleost species (salmonids and common carp) share low aa sequence identity and appear to be fast-evolving regardless of the mode of duplication. This suggests an inherent potential for sub-/neo-functionalization in such molecules, as demonstrated here for trout IL-2A and IL-2B.

#### IL-2 Isoforms Are T Cell Factors

Highest expression of both trout *IL2A* and *IL2B* was seen in thymus, spleen, gills, kidney, and intestine, important tissues/organs in T cell development and function. While their constitutive expression level in PBL was relatively low it could be upregulated by MLR, an alloantigen-mediated T cell activation, by the T cell mitogen PHA, and by signal mimics of T cell activation (PMA and CI stimulation). Previously, *IL2A* was found to be upregulated in PBL by T cell activation with the costimulatory signal CD80/ CD86 (22) and *in vivo* in CD4-1 and CD4-2 double-positive T cells after bacterial infection (74). These expression patterns suggest that both IL-2 isoforms are T cell factors secreted by activated T cells.

Interestingly, the two *IL2* paralogs are differentially expressed and modulated. *IL2B* expression is lower constitutively relative to *IL2A* but is more inducible, as seen with stimulation with PHA and PMA + CI. Differential expression and modulation is a common feature of duplicated paralogs as seen with salmonid *IL1β* (60), *TNFα* (41), *IL4/13* (24), *IL-12* (48, 49) and *IL17A/F* (25) genes. In this study, many paralogous genes, e.g., *TNFα*, *IL-12*, and cytokine/chemokine receptors, are differentially modulated by the two IL-2 isoforms. These differential expression patterns and modulation of paralogs hint at functional diversification.

#### Potential Role in Th Cell Development and Adaptive Immunity in Fish

In mammals, when naive CD4+ T cells recognize a foreign antigen-derived peptide presented in the context of MHC class II on APCs, they undergo massive proliferation and differentiation into distinct Th cell subsets such as Th1, Th2, Th17, and induced T-regulatory cells. Each cell subset expresses a unique set of signature cytokines (75). Cytokines produced by these Th subsets play a critical role in immune cell differentiation, effector subset commitment, and in directing the effector response. Although evidence for the existence of mammalian type Th cells in fish is elusive, the major cytokine players are present (5). In this study, both trout IL-2 isoforms could upregulate the expression of signature cytokines for Th1 (*IFNγ1*, *IFNγ2*, and *TNFα2*) and Th2 (*IL-4/13B1* and *IL4/13B2*) cells, but have no effects on Th17 cytokines (*IL17A/F1A*, *IL-17A/F2A*, and *IL17A/F3*), and limited effects on Treg cytokines (*TGFβ1* and *IL10*). Furthermore, IL-2 could modulate the expression of receptors for IFNγ and IL4/13 and maintain the expression of Th cell markers (e.g., *CD4-1*, *CD4-2A*, and *CD4-2B*). Importantly, *IL12P40B2* and *IL12P40C* are induced by both IL-2 isoforms with *IL12P35A1* upregulated by IL-2A and suggests that two isoforms of IL-12 (a driver cytokine for Th1 cell development in mammals) can be produced in response to trout IL-2 that are known to have distinct bioactivities in terms of induction of *IFNγ* and *IL10* expression (71). Overall, these findings indicate that the trout IL-2 isoforms may have an important role in regulating Th1 and Th2 cell development and adaptive immune responses in fish.

## Is There a Role of IL-2 in Treg Cell Development?

Although originally described as a potent T cell growth factor *in vitro*, the main non-redundant role of IL-2 *in vivo* is now known to be the maintenance of peripheral T cell tolerance by promoting the thymic development, peripheral homeostasis, and suppressive function of Treg cells in mammals (76). Treg cells express the signature transcription factor Foxp3, which is critical for their development, lineage commitment, and secretion of IL-10 and TGF-β (75). Both IL-2 isoforms upregulate *FOXP3* and *IL10* expression to a small extent, but have limited effects on *TGFβ1*. This may suggest that the role of IL-2 in Treg cell development is somewhat conserved in fish. However, the mammalian IL-2/Treg paradigm is linked with CD25 expression and hence the high-affinity IL-2R on Treg cells, allowing Treg cells to efficiently compete with effector CD4+ T cells for the available IL-2 (7). Due to the lack of a *bona fide* CD25, a high-affinity IL-2R specific for IL-2 may not exist in fish. Instead, receptors with different affinities may be shared by at least five IL-2/15 cytokines (discussed above) in salmonids. The contribution of each of these IL-2/15 cytokines to Treg cell development in fish remains to be determined.

#### Functional Diversification of IL-2A and IL-2B

The aa sequence and exon size diversification of salmonid IL-2 hints at functional diversification. This notion was confirmed by bioactivity analysis of the two recombinant IL-2 isoforms. While in most cases the modulated gene expression by the two IL-2 isoforms was similar (overlapping) when added to cells alone or together, the upregulation of *IL12P35A1* and *CXCR1* expression was found in only IL-2A-treated samples. In several cases (e.g., *IFNγ1*, *IFNγ2*, *CXCL11L1*, *CD8α*, *CD8β* and *CXCR2*), the response seen was more IL-2B like when the two molecules were added together, with a stronger induction stimulated by IL-2B alone. However, in a few cases (e.g., *IL12P40C* and *IL4Rα2* at 4 h), a stronger response was seen when both IL-2 were present. These bioactivity patterns may suggest that each IL-2 isoform signals independently *via* a pool of receptors expressed in different cells.

It is known that rainbow trout possess one CD25L, and two IL-2Rβ and γC that potentially form four intermediate affinity receptors (IL-2Rβ/γC) and four high-affinity receptors (CD25L/ IL-2Rβ/γC). The receptor subunits are differentially expressed and modulated in different cell types (13, 14), and this will determine which functional receptors are available on a cell. It is quite possible that each IL-2 isoform evolved a preference (high affinity) for binding to a specific receptor (discussed above), as part of receptor–ligand co-evolution. IL-2B may have higher affinity receptors on the cells that expressed *IFNγ1, IFNγ2, CXCL11L1, CD8α*, *CD8β* and *CXCR2*, even though IL-2A can induce a similar response (albeit weaker) in the absence of IL-2B. Similarly, IL-2A may have higher affinity receptors on the cells that expressed *IL12P35A1* and *CXCR1*.

#### IL-2 and Host Defense

Both trout IL-2 isoforms induce the expression of Th1 (e.g., *IFNγ* and *TNFα*) and Th2 (e.g., *IL-4/13B*) cytokines in PBL and HK cells (data not shown), that are crucial for host defense against intracellular pathogens and extracellular helminthic parasites in mammals, respectively. They also modulate PBL expression of chemokine receptors important in leukocyte trafficking, and the antimicrobial peptide *cathelicidin-1*, key components of innate immune defense against microorganisms (70). Moreover, they enhance PBL phagocytosis of myeloid cells. Clearly, IL-2 isoforms have an important role in fish defense.

Interestingly, trout IL-2 isoforms induce only *cathelicidin-1* but not *cathelicidin-2* in PBL. This is a similar situation to that found with trout IL-4/13 isoforms in HK cells (24), but in contrast with trout IL-6 that induces *cathelicidin-2* but not *cathelicidin-1* (42). This may relate to the signaling pathways used by different cytokine families in relation to the transcription factors needed for cathelicidin 1 or 2 expression.

## Is IL-2 a T Cell Growth Factor in Fish?

Besides its potent T cell growth factor activity, mammalian IL-2 induces proliferation of NK cells and B cells (7). Trout IL-2 isoforms can promote PBL growth *in vitro* as shown by enhanced BrdU incorporation, although direct evidence on T cell growth was not analyzed in the current study. However, the maintenance of high-level expression of T cell markers (*CD4-1*, *CD4-2*, *CD8α*, and *CD8β*) after IL-2 stimulation at late time points, when their expression was decreased in the absence of IL-2, suggests that T cell proliferation could contribute to the enhanced PBL growth seen.

#### Conclusion

Two divergent *IL2* paralogs are present in salmonids due to the salmonid 4R WGD. The salmonid *IL2* paralogs differ not only in sequence but also in exon sizes. The IL-2 isoforms encoded have disparate pI values and may have evolved preferential binding to specific IL-2 receptors. Rainbow trout *IL2* paralogs have highest constitutive expression in thymus, spleen, gills, kidney and intestine, important tissues/organs in fish T cell development and function. While their transcript levels are relatively low in PBL, their expression can be upregulated by MLR, by the T cell mitogen PHA, and by signal mimics of T cell activation. Both trout IL-2 isoforms promote PBL proliferation and sustain high-level expression of the T cell markers CD4 and CD8, suggesting that trout IL-2 isoforms are T cell growth factors mainly expressed by activated T cells. The two trout IL-2 isoforms have shared but also distinct bioactivities. IL-2A, but not IL-2B, induces *IL12P35A1* and *CXCR1* expression in PBL. IL-2B has a stronger effect on upregulation of the Th1 cytokine *IFNγ* and sustaining *CD8α* and *CD8β* expression. Both proteins upregulate the expression of key Th1 and Th2 cytokines, cytokine and chemokine receptors, and the antimicrobial peptide *cathelicidin-1*, and enhance phagocytosis of myeloid cells in PBL. Our results suggest that *IL2* paralogs have an important role in regulating Th1 and Th2 cell development and host defense in fish.

## ETHICS STATEMENT

All the experiments described comply with the Guidelines of the European Union Council (2010/63/EU) for the use of laboratory animals and were carried out under UK Home Office project license PPL 60/4013, approved by the ethics committee at the University of Aberdeen.

## AUTHOR CONTRIBUTIONS

TW and CS contributed conceptually, to the design of the study, and wrote the first draft of the manuscript. TW, YH, EW, FL, AW, EZ, ML, and QX performed the experiments. All the authors analyzed the data, contributed to the manuscript revision, read, and approved the submitted version.

#### FUNDING

This work was funded by the Biotechnology and Biological Sciences Research Council (BBSRC, BB/N024052/1) under the Newton Fund RCUK-CONICYT Research Partnerships call. YH was supported by a PhD Studentship from the Ministry of

#### REFERENCES


Education, Republic of China (Taiwan). EW was supported financially by the Faculty of Technology, Mahasarakham University Grant Year 2018. FL was supported by a Newton International Fellowship funded by the Academy of Medical Sciences, UK (AMS, NIF004\1036). ML and QX were supported financially by the National Scholarship Council of China. This work was partially supported financially by European Commission contract No. 311993 (TargetFish).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.01683/ full#supplementary-material.


evolution of Th1-like immune responses. *Eur J Immunol* (2014) 44:1541–51. doi:10.1002/eji.201344273


IL-34, MCSF1 and MCSF2, ligands of the MCSF receptor. *Mol Immunol* (2013) 53:398–409. doi:10.1016/j.molimm.2012.09.008


**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 Wang, Hu, Wangkahart, Liu, Wang, Zahran, Maisey, Liu, Xu, Imarai and Secombes. 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.*

# Perspective on the Development and Validation of Ab Reagents to Fish Immune Proteins for the Correct Assessment of Immune Function

Brian Dixon<sup>1</sup> \*, Daniel R. Barreda<sup>2</sup> \* and J. Oriol Sunyer <sup>3</sup> \*

<sup>1</sup> Department of Biology, University of Waterloo, Waterloo, ON, Canada, <sup>2</sup> Department of Biological Sciences, and Department of Agricultural, Food and Nutritional Science, University of Alberta, Edmonton, AB, Canada, <sup>3</sup> Department of Pathobiology, School of Veterinary Medicine, University of Pennsylvania, Philadelphia, PA, United States

#### Edited by:

Geert Wiegertjes, Wageningen University and Research, Netherlands

#### Reviewed by:

Carolina Tafalla, Instituto Nacional de Investigación y Tecnología Agraria y Alimentaria (INIA), Spain Steve Bird, University of Waikato, New Zealand

#### \*Correspondence:

Brian Dixon bdixon@uwaterloo.ca Daniel R. Barreda dan.barreda@ualberta.ca J. Oriol Sunyer sunyer@vet.upenn.edu

#### Specialty section:

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

Received: 31 May 2018 Accepted: 30 November 2018 Published: 13 December 2018

#### Citation:

Dixon B, Barreda DR and Sunyer JO (2018) Perspective on the Development and Validation of Ab Reagents to Fish Immune Proteins for the Correct Assessment of Immune Function. Front. Immunol. 9:2957. doi: 10.3389/fimmu.2018.02957 Understanding of immune function in humans and model organisms, such as mice, has advanced in the last few decades because of technological breakthroughs and availability of reagents. While novel genomic technologies have helped to increase knowledge of many aspects of immunology, most developments in immunology have occurred because of the availability of antibodies to identify and sort different cell types, as well as to identify and quantify the protein products of cells. Unfortunately, many studies performed in fish make use of poorly characterized antibody reagents that may affect the conclusions of those studies. In light of this, we would like to offer some insight and discussion points based on our research experience on the strategies and techniques that are required for proper validation of antibody reagents to fish immune molecules. Our main goal is to encourage a much needed discussion in our field to foster the use of correctly validated reagents that enable the study of fish immune function.

#### Keywords: immunology, teleost fish, antibodies, assays, validation, reagents

## INTRODUCTION

Genomic technology is advancing rapidly because nucleic acids can be amplified and, thus, methods for inexpensive large-scale sequencing have been developed. In non-classical model species, like those comparative immunologists use, genomics is a key first step because specific reagents to one particular gene or protein are not usually available. While microarrays or RNAseq can determine mRNA concentrations of all genes expressed in a cell or tissue, in some cases this may not necessarily reflect transcription rates, as mRNA degradation by RNAses plays a role in determining mRNA, especially for certain genes like cytokines and chemokines with degradation motifs in their 3 ′ untranslated regions (1). Thus, tools to measure proteins and cells are required to understand immune function.

## TRANSCRIPTS VS. PROTEINS

Transcript concentrations that nucleic acid-based assays detect may directly relate to protein concentrations, but for many genes they may not, and will certainly not reflect concentrations of active protein. This is especially true for many of the genes that immunologists are interested in. For example, receptors must be translated into protein through specific channels into the endoplasmic reticulum (ER), perhaps glycosylated and folded by chaperones, then transported through the Golgi to the surface. These steps all depend on the dynamics and rates of action of several enzymes and transport molecules that cannot be assessed by examining mRNA concentrations. Cytokines must also be translated into the ER and transported to the cell surface for secretion. Many are secreted in inactive forms (glycosylated or as precursors that need to be cleaved i.e., IL1β). These processes all have dynamics that RNA based assays cannot capture. In addition, receptors or high affinity binding receptor subunits (e.g., IL2Rα) must be upregulated in a coordinated fashion for full function and, in some cases, decoy receptors bind the ligands to prevent their function and/or decoy ligands compete for the active receptor [e.g., as seen for IL1β, see (2, 3)]. Often these molecules are expressed in a different cell type or tissue than the one studied by RNA based assays, which can be addressed through inclusion of additional experimental steps but should be considered carefully when designing experiments. Another example of mismatch between mRNA concentrations and functional capacity that is very familiar to immunologists is complement. Complement component C3 must be cleaved to be activated and simply measuring C3 mRNA cannot give a clear picture of the level of complement activity at any given point in time, especially since for C3, like many teleost immune genes, there are multiple isoforms present in the genome, all of which may have different dynamics and roles (4).

For those of us that have chosen to work in mixed cell populations or at the tissue level, it is important to consider that many target genes are differentially expressed among cells (types and developmental stages) and may further exert different functions depending on which cell expresses them [e.g., consider TLRs; (5)]. Large populations of immune cells also migrate actively through various tissues further complicating assessments when single tissues are examined. Cell-cell interactions are also critical for immune functions. Even as single cell transcriptomic technologies become increasingly accessible, these approaches are unlikely to capture cell-cell interactions that are critical for immune function. Coupled to the challenge of working with outbred populations of animals, this can introduce significant variability in the resulting datasets and make it much harder to assess functional implications. RNA based assays on a whole tissue will capture and average mRNA expression in interacting cells, along with several other cell types, making understanding how each reacts to the other impossible. Also, not all transcripts make protein: MicroRNA (miRNA) regulation of many genes, including immune genes of fish [e.g., see (6, 7)], means that a measurement of the total amount of mRNA present does not accurately reflect the amount of protein that will ultimately be produced. Most critically, as new opportunities arise to understand adaptive mechanisms in fish, one common scenario continues to be seen: changes in the transcription of the IgM, IgT/IgZ, and IgD genes are being used to evaluate antibody responses. While these genes may be upregulated during immune responses, such an increase is not indicative of a specific antibody response because one cannot identify an immunoglobulin with specificity for the antigen in question from mRNA expression. Secondly, while in mammals there might be an isotype switch indicative of a specific response (e.g., from IgM to IgG), detectable using mRNA expression data, this is not possible for species like teleost fish where there is no isotype switching. Primers used in many studies to amplify fish immunoglobulin transcripts, also do not distinguish between sterile and productive transcripts. Finally, in humans, not all circulating immunoglobulins are active: immunoglobulins bearing sugar moieties with terminal sialic residues are anti-inflammatory and the structure of these carbohydrates' changes during immune responses to forms with terminal N-acetylglucosamine residues that have much more bioactivity (8). This is likely to be similar in other species; for example, rainbow trout IgM heavy chains have five potential Nlinked glycosylation sites (9) while IgT isoforms have at least two (10, 11).

Thus, caution is needed in forming conclusions solely from RNA based studies without complementing them with functional data at the protein, cellular, and/or organismal levels. RNA based assays may be useful as first experiments to focus future work on the correct cells and proteins and can provide important context as we dissect mechanisms of immune function but cannot be the final experiment used to make conclusions about immune functions, processes, and responses.

## KINETICS

Exhaustive analysis of cellular or molecular events occurring during a typical immune response is unrealistic because of funding and time constraints. However, on the other extreme, we continue to see costly decisions being made in both academia and industry based on evaluations of individual or very limited time points. Regardless of the analytical depth (molecular or cellular), no single time point can provide sufficient functional context for the effectiveness of an immune response and, thus, added emphasis should be placed on kinetics rather than the robustness of select parameters at a single time point. The number of relevant time points and the level of depth in which they should be evaluated will vary depending on the biological question. However, effects on immune competence (e.g., due to infection, environment, diet, therapy) need to consider changes to the efficiency of induction of immune mechanisms, their absolute concentrations, and whether a timely return to homeostasis has been achieved (e.g., required to minimize unnecessary tissue damage and unproductive use of metabolic energy resources). For all of these, it is paramount that molecular datasets are complemented by evaluation of functional responses (e.g., cell function, pathology, host performance).

## DEVELOPING ANTIBODIES TO ASSESS FISH IMMUNE FUNCTION

In addition to using transcript concentrations as proxy of functionality, fish immunologists need antibodies against immune molecules and cells in order to understand specifically how the fish immune system operates. Unfortunately, developing and validating such tools is slow, painstaking work, regardless of the animal model used. Most importantly, the correct standardization and validation of these antibodies is fundamental for comparison of results among different labs. Below are some key areas that require special attention during the production and validation of polyclonal (pAb) and monoclonal (mAb) antibodies to fish molecules.

#### Developing Antibody Targets

In most instances, the antibody target cannot be purified in its native form, and thus, recombinant proteins must be produced. Soluble proteins expressed in prokaryotic expression systems usually require proper refolding to acquire a more accurate conformational structure. Thus, a refolding step has proven critical for some of our prokaryotically-produced antigens in order to induce antibody responses that will recognize the native fish molecule (12). This is especially true in immune assays in which these reagents are required to recognize the native molecule (i.e., ELISA, flow cytometry). Antibodies may also be developed to proteins that are expressed in mammalian cells and then reintroduced into the host, for example, expression in rat cells and injection into rats. Antibodies may also be made to peptides as long as care is taken to ensure the protein fragment is expressed on the outside of the folded protein. In our experience, antigens produced in eukaryotic expression systems are better at inducing antibodies that recognize the native fish molecule compared to antigens made in prokaryotic expression systems. Antibodies developed to mammalian molecules that have a high degree of sequence identity to the equivalent fish proteins can sometimes be used, but must be validated very carefully using some of the methods suggested below to ensure that they do indeed bind to the correct target.

#### Immunization and Adjuvants

In our experience, the use of rabbits for the production of pAbs against fish molecules is not ideal. Rabbit antibodies appear to naturally recognize, or non-specifically bind to a significant percentage of fish leukocytes and proteins, producing false positive results, and high background, perhaps due to cross priming. This is particularly critical for flow cytometry, in which the real reactivity is confounded by the cross-reactive/nonspecific binding capacity of these rabbit Igs. Antibodies produced in guinea pigs, rats, mice, and chickens do not usually present this problem. For mAb production, the use of more than one species (i.e., mice and rats) provides alternative choices for secondary reagents when the experiment involves several mAbs to different (e.g., flow cytometry) or the same (e.g., ELISA) antigen. The choice of adjuvant may amplify the titers of non-specific or crossreactive Abs, especially in rabbits, and thus, it is worth exploring which adjuvant (e.g., oil-based vs. non-oil-based) works best for your antigen. In many cases, it is incorrect to assume that the pre-immune serum is a good control, since those animals have not been exposed to adjuvant. Serum from animals injected with adjuvant alone might be more appropriate in some cases. Control serum should be subjected to the same purification steps as immune serum in order to ensure that background reactivity that remains is consistent.

## Characterization and Validation

For pAbs, it is crucial to use antibodies affinity purified against the antigen (i.e., using an affinity column) after protein A or a similar purifying agent has been used to obtain the pure Ig fraction. It is also helpful to affinity purify polyclonal serum against recombinant proteins produced in two different systems (i.e., bacterial and eukaryotic) to minimize potential problems of non-specificity or cross-reactivity.

Without a doubt, the most critical aspect in the development of pAb or mAbs, is the strategy used to validate such reagents. Below we describe some of the most critical steps we have used for the correct validation of Ab reactivity:

#### General Validation Strategies

Since recombinant antigen is not usually produced in fish cell lines, it is likely that critical antigenic sites are lost when proteins expressed in prokaryotic or eukaryotic expression systems do not fold correctly. Therefore, it is mandatory to test whether the Abs induced by those recombinant antigens recognize the native fish molecule, as recognition of recombinant antigen by these Abs does not necessarily translate into recognition of the native molecule. In our view, one of the best strategies is to assess whether these Abs recognize the native antigen when recombinantly expressed on a fish cell line of the same or similar species used for your experiments. For example, recently produced mAbs to rainbow trout CD4-1 and CD4- 2 molecules were validated by showing that these mAbs could recognize those molecules transiently expressed on a rainbow trout cell line (13). Moreover, recognition of antigens expressed in a cell line should be checked by at least two different techniques, including flow cytometry and western. Another effective strategy to validate the correct reactivity of the Abs is to immunoprecipitate or column purify the native antigen from fish serum or leukocytes. This strategy is most valuable when combined with sequencing of the purified protein to confirm its identity, but requires that antigen is produced in significant levels and that it displays sufficient stability.

#### Validation Strategies of Abs to CDs

When producing Abs to CDs to detect specific fish leukocyte populations, additional validation steps to those described in section General validation strategies are required, to confirm that the Abs recognize the expected leukocyte subset. Thus, until more CD markers for fish cells become available, one might sort the cell subset/s recognized by the Abs (ideally by FACS) and perform RT-PCR on the sorted cells to confirm the expression of target cell-type transcripts for the target cells. Equally important, is measurement for transcripts uniquely expressed by other leukocyte subsets to rule out cross-reaction or non-specific recognition of surface molecules of unrelated leukocyte populations. For instance, if the newly produced mAb recognizes CD4, then sort the cells recognized by this mAb and check not only for the expression of CD4 transcripts but also for the expression of molecules typically expressed by other cell leukocytes (i.e., CD8 T cells, immunoglobulin, NK cell receptors, etc.). If the sorted leukocyte population only expresses CD4 T cell transcripts, then it indicates that the Ab is highly specific. The possibility exists however that we may find unexpected fish cell populations expressing CDs found only in certain leukocyte subsets in mammals, in which case further characterization of those potentially new cell subsets is needed to validate the Abs. Thus, it is important to keep an open mind and not assume that fish leukocyte subsets will express the same CDs expressed by their mammalian counterparts.

#### Validation Strategies of Abs to Fish Cytokines

In addition to the validation steps described in section General validation strategies, there are some peculiarities related to the validation of Abs raised against fish cytokines. To date, only a few antibodies to fish cytokines have been reported, and while they work very well on immunoblotting assays, further validation is required before any antibody can be used for more quantitative assays or to detect native molecules. It has been very difficult to develop sandwich ELISAs to measure specific cytokine concentrations in fish fluids, and/or to detect fish cytokines intracellularly. For instance, in the last 5 years two of us (Dixon and Sunyer labs) have attempted to produce mAbs and pAbs to over 15 different fish cytokines, and after performing all of the pertinent validation steps, Sunyer's lab could only validate with a high degree of confidence Abs to one out of eight target cytokines in tissues and fish fluids (Sunyer's personal communication) while Dixon's lab is confident that one target cytokine of seven can be detected. In some cases, the produced Abs did recognize the recombinant cytokine, but could not recognize the native fish cytokine (Dixon's and Sunyer's personal communication). In other instances, the Abs could recognize both the recombinant antigen and the native cytokine when expressed on a trout cell line, however, we were unable to detect the cytokine in any fish fluid, or intracellularly by means of ELISA, western blotting, or flow cytometry. Thus, it is apparent that some of these fish cytokines may be expressed at very low concentrations (i.e., Dixon's lab successful ELISA detects IL-1β at concentrations below 100 pg/mL in serum) or require the proper stimulation of the cell type producing them, in order to be detected. For example, detection of IL-10-producing B cells in mammals requires re-stimulation of these cells exvivo in order detect intracellular IL-10 (14). There may also be factors in the fish tissues that inhibit ELISA reactions as Dixon's group has shown that chloroform extraction enhances detection by ELISA and shows expression profiles that match qPCR and Western data in some cases—for example IL-1β in serum. Moreover, the detection of intracellular cytokines in mammals typically requires the use of brefeldin or monensin, both protein transport inhibitors, to enhance the accumulation of intracellular cytokine (15). However, the application of all of the above measures to enhance the intracellular detection of cytokines by flow cytometry is not a guarantee of success. For example, while Sunyer's lab has produced mAbs and pAbs to trout IL-10 and have successfully developed a sandwich ELISA that detects soluble native trout IL-10 (unpublished results), they have not been able to consistently assess intracellular IL-10 expression in trout lymphocytes. They have tried multiple combinations of in vitro re-stimulation cocktails and protocols, different kinetics, different concentrations of brefeldin and monensin, different cell permeabilization protocols, and different combinations of all of the above without success. However, IL-10 could be detected in the cell supernatants in many cases, which indicates a failure to detect the intracellular IL-10 specifically. Alternatively, Sunyer's lab has developed immunohistochemistry protocols to identify the cell types producing IL-10 (Sunyer's personal communication). Dixon's group can detect IL-1β in serum, cell culture supernatants, and cell extracts, although Western blots show bands of odd sizes that are difficult to reconcile in the latter case. These may or may not be background band because the true size of the native protein in vivo is unknown and is unlikely to match the predicted size based on amino acid sequences. Native IL-1β, may not even be a single size because like IL-10 and many other cytokines, it can be glycosylated at one or more sites in vivo. This effects stability and biological activity in ways that are unclear even in mammals (16), but can inhibit function (17) or enhance receptor binding (18). This glycosylation may indeed cause the problems in intracellular staining noted by Sunyer's group as the carbohydrates may block or mask the epitopes recognized by the antibodies and glycosylation may differ between extracellular and intracellular compartments. Thus, an understanding of the basic biology of each cytokine is absolutely required before one can be confident in antibody staining and quantification techniques. Unfortunately, that will take time and effort—for example Dixon's lab is now deglycosylating extracts and is sequencing bands detected by his antisera that are not the "predicted" size to see if they actually represent alternative forms of the target cytokine.

## CONCLUSIONS

A true understanding of immune system function in fish absolutely requires that we do not simply ascribe function based on transcript profiles only, but that we develop antibodies and, most importantly, validate those reagents very carefully based on a detailed understanding of the basic biology of the target molecules and cells specifically in fish. We then need to share those reagents widely and wisely to further validate them, so that we can all advance knowledge together to not only improve fish immunology academically, but also partner and profit with the industries that depend on the knowledge we produce.

## AUTHOR CONTRIBUTIONS

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

## ACKNOWLEDGMENTS

This work was supported by a National Institutes of Health Grant 2R01GM085207-05 to JOS, by a National Science Foundation Grant NSF-IOS-1457282 to JOS and by a USDA-NIFA AFRI grant# 2016-67015-24901 to JOS, an NSERC Discovery grant RGPIN-2018-05768 to DRB, an NSERC Discovery grant RGPIN-2018-04116 to BD, an NSERC Canada Research Council Chair to BD and funding to BD from the Atlantic Canada Opportunities Agency/Atlantic Innovation Fund.

## 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 © 2018 Dixon, Barreda and Sunyer. 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.