# PARASITE INFECTIONS: FROM EXPERIMENTAL MODELS TO NATURAL SYSTEMS

EDITED BY : Toni Aebischer, Kai Matuschewski and Susanne Hartmann PUBLISHED IN : Frontiers in Cellular and Infection Microbiology

#### Frontiers Copyright Statement

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

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

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

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

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

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

The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use. ISSN 1664-8714 ISBN 978-2-88945-485-3 DOI 10.3389/978-2-88945-485-3

#### 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

# PARASITE INFECTIONS: FROM EXPERIMENTAL MODELS TO NATURAL SYSTEMS

Topic Editors:

Toni Aebischer, Robert Koch-Institut, Germany Kai Matuschewski, Humboldt Universität, Germany Susanne Hartmann, Freie Universität Berlin, Germany

Image: Svenja Steinfelder.

Eukaryotic parasites (including parasitic protozoans, worms and arthropods) are more complex and heterogeneous organisms than pathogenic bacteria and viruses. This notion implies different evolutionary strategies of host exploitation. Typically, parasites establish long-term infections and induce relatively little mortality, as they often limit pathological changes by modulating host cells and downregulating adverse immune responses. Their pattern of distribution tends to be endemic rather than epidemic. Despite these seemingly benign traits, parasites usually cause substantial chronic morbidity, thus constituting an enormous socioeconomic burden in humans, particularly in resource poor countries, and in livestock worldwide. Parasite-induced fitness costs are an evolutionary force that can shape populations and contribute to species diversity. Therefore, a thorough understanding of parasites and parasitic diseases requires detailed knowledge of the respective biochemical, molecular and immunological aspects as well as of population genetics, epidemiology and ecology. This Research Topic (RT) bridges disciplines to connect molecular, immunological and wildlife aspects of parasitic infections. The RT puts emphases on four groups of parasites: *Plasmodium*, *Toxoplasma*, *Giardia* and intestinal helminths. Co-infections are also covered by the RT as they represent the most common form of parasite infections in wildlife and domestic animal populations. Within the four types of parasites the following topics are addressed: (1) Experimental models: hypothesis testing, translation and limits. (2) Critical appraisal of experimental models. (3) Natural systems: Technological advances for investigations in natural parasite-host systems and studies in natural systems. (4) The urgent need for better models and methods in natural parasite systems. Hence, the RT covers and illustrate by the means of four main parasitic infections the parasite-host system at the molecular, cellular and organismic level.

Citation: Aebischer, T., Matuschewski, K., Hartmann, S., eds. (2018). Parasite Infections: From Experimental Models to Natural Systems. Lausanne: Frontiers Media. doi: 10.3389/978-2-88945-485-3

# Table of Contents

*07 Editorial: Parasite Infections: From Experimental Models to Natural Systems*

Toni Aebischer, Kai Matuschewski and Susanne Hartmann

### SECTION 1

### EXPERIMENTAL MODELS: HYPOTHESIS TESTING, TRANSLATION AND LIMITS


Maria C. Souza, Denise M. Fonseca, Alexandre Kanashiro, Luciana Benevides, Tiago S. Medina, Murilo S. Dias, Warrison A. Andrade, Giuliano Bonfá, Marcondes A. B. Silva, Aline Gozzi, Marcos C. Borges, Ricardo T. Gazzinelli,José C. Alves-Filho, Fernando Q. Cunha and João S. Silva

*57 Behavior of Neutrophil Granulocytes During* Toxoplasma Gondii *Infection in the Central Nervous System*

Aindrila Biswas, Timothy French, Henning P. Düsedau, Nancy Mueller, Monika Riek-Burchardt, Anne Dudeck, Ute Bank, Thomas Schüler and Ildiko Rita Dunay

*70 Th2/1 Hybrid Cells Occurring in Murine and Human Strongyloidiasis Share Effector Functions of Th1 Cells* Cristin N. Bock, Subash Babu, Minka Breloer, Anuradha Rajamanickam,

Yukhti Boothra, Marie-Luise Brunn, Anja A. Kühl, Roswitha Merle, Max Löhning, Susanne Hartmann and Sebastian Rausch

*85* Toxoplasma *Co-infection Prevents Th2 Differentiation and Leads to a Helminth-Specific Th1 Response*

Norus Ahmed, Timothy French, Sebastian Rausch, Anja Kühl, Katrin Hemminger, Ildiko R. Dunay, Svenja Steinfelder and Susanne Hartmann

*97 Reciprocal Interactions Between Nematodes and Their Microbial Environments*

Ankur Midha, Josephine Schlosser and Susanne Hartmann

### SECTION 2

### APPRAISALS OF EXPERIMENTAL MODELS


Martin R. Kraft, Christian Klotz, Roland Bücker, Jörg-Dieter Schulzke and Toni Aebischer

*160 Translational Rodent Models for Research on Parasitic Protozoa—A Review of Confounders and Possibilities*

Totta Ehret, Francesca Torelli, Christian Klotz, Amy B. Pedersen and Frank Seeber

### SECTION 3

### NATURAL SYSTEMS

### 3.1 TECHNOLOGICAL ADVANCES FOR INVESTIGATING NATURAL PARASITE-HOST SYSTEMS

*180 MALDI-TOF MS Profiling-Advances in Species Identification of Pests, Parasites, and Vectors*

Jayaseelan Murugaiyan and Uwe Roesler

*189 Nematode Species Identification—Current Status, Challenges and Future Perspectives for Cyathostomins*

Christina M. Bredtmann, Jürgen Krücken, Jayaseelan Murugaiyan, Tetiana Kuzmina and Georg von Samson-Himmelstjerna


and Filippo Castiglione

### 3.2 STUDIES IN NATURAL SYSTEMS

*227 Strain- and Dose-Dependent Reduction of* Toxoplasma Gondii *Burden in Pigs is Associated With Interferon-Gamma Production by CD8+ Lymphocytes in a Heterologous Challenge Model* Malgorzata Jennes, Stéphane De Craeye, Bert Devriendt, Katelijne Dierick, Pierre Dorny and Eric Cox

*247 Long-Term Temporal Trends of* Nosema spp*. Infection Prevalence in Northeast Germany: Continuous Spread of* Nosema Ceranae*, an Emerging Pathogen of Honey Bees (Apis mellifera), but no General Replacement of*  Nosema Apis

Sebastian Gisder, Vivian Schüler, Lennart L. Horchler, Detlef Groth and Elke Genersch

*261 NF-*κ *B-Like Signaling Pathway REL2 in Immune Defenses of the Malaria Vector* Anopheles Gambiae

Suzana Zakovic and Elena A. Levashina


Oriana Kreutzfeld, Katja Müller and Kai Matuschewski

# Editorial: Parasite Infections: From Experimental Models to Natural Systems

#### Toni Aebischer <sup>1</sup> , Kai Matuschewski <sup>2</sup> and Susanne Hartmann<sup>3</sup> \*

<sup>1</sup> Unit 16 Mycotic and Parasitic Agents and Mycobacteria, Robert Koch-Institute, Berlin, Germany, <sup>2</sup> Department of Molecular Parasitology, Institute of Biology, Humboldt University, Berlin, Germany, <sup>3</sup> Department of Veterinary Medicine, Institute of Immunology, Freie Universität Berlin, Berlin, Germany

Keywords: parasite, malaria, toxoplasmosis, giardiasis, helminth infections

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

experimental models to natural systems:

#### **Parasite Infections: From Experimental Models to Natural Systems**

The aim of this research topic is to illustrate the multidisciplinary approaches of modern parasitology. The motivation to study parasites and parasitism varies. In the case of human and animal parasites, research is often motivated by the tremendous health threats and socioeconomic burden they pose. For instance, Plasmodium, the causative agent of malaria, continues to be the most important vector-borne pathogen and was responsible for more than 200 million new cases and 400,000 deaths worldwide in 2016 (WHO, 2016a). Another example is soil-transmitted helminth infections affecting 24% of the world's population, primarily school children (WHO, 2017). Many parasites are etiologic agents of infections classified as Neglected Tropical Diseases (NTDs) and continue to afflict societies with limited resources (Hotez et al., 2007). Moreover, research on parasites of wildlife can be critical for understanding animal communities and disease ecology (Gomez and Nichols, 2013; Johnson et al., 2015) and—by extrapolation—ecosystems' dynamics.

Parasites are defined by life style but reflect polyphyletic groups of protozoa, helminthes, and arthropods. In investigating these non-model organisms, studies on parasites often fall short of in-depth molecular, genetic, or biochemical analyses that characterize investigations of established laboratory-adapted organisms. This research topic (RT) collates 20 contributions by >100 authors, which we introduce here briefly by classifying them according to four sections along the path from

#### Edited by:

Kenneth Pfarr, Universitätsklinikum Bonn, Germany

> Reviewed by: Teresa Gil Carvalho, La Trobe University, Australia

\*Correspondence: Susanne Hartmann susanne.hartmann@fu-berlin.de

Received: 17 October 2017 Accepted: 12 January 2018 Published: 02 February 2018

#### Citation:

Aebischer T, Matuschewski K and Hartmann S (2018) Editorial: Parasite Infections: From Experimental Models to Natural Systems. Front. Cell. Infect. Microbiol. 8:12. doi: 10.3389/fcimb.2018.00012 EXPERIMENTAL MODELS: HYPOTHESIS TESTING, TRANSLATION, AND LIMITS

Experimental models such as laboratory mice or model cell culture systems have the great advantage to mechanistically answer biological questions. This proven strategy is used by Ngwa and colleagues who used in vitro differentiation of Plasmodium falciparum gametocytes to describe epigenetic changes due to histone acetylation that are likely relevant during parasite transmission from human host to vector (Ngwa et al.). Interaction of parasite and host factors can also reliably be studied in experimental models. Here, Dunst and colleagues identified the microfilament tethering factor moesin as an interacting partner of P. falciparum GPI purified from blood stages parasites (Dunst et al.). The authors show that absence of moesin influences neither life cycle progression nor malaria-related pathology, calling for further studies to identify GPI-binding recognition factors in the host. Analyses in mouse parasites illustrate key elements of the tool-boxes available to parasitological research in experimental model organisms.

This tool-box is also exploited in the contribution by Souza et al.. They apply a murine model of microbial sepsis to test the collateral damage by a pro-inflammatory immune response elicited by a persisting preexisting Toxoplasma gondii infection. The study on neutrophils in cerebral inflammation due to Toxoplasma infection contributed by Biswas and coworkers highlights another benefit of using laboratory mouse models, namely the broad range of antibodies available to identify and characterize specific immune cell subsets, based on signature proteins and cytokines (Biswas et al.). This allowed the authors to show that two distinct populations of inflammatory neutrophils act together to diminish parasite infiltration into the brain and reduce experimental cerebral toxoplasmosis.

Deciphering of immune cell interaction and cellular networks of host responses to parasites often rely on laboratory models. In this RT Bock and colleagues define so-called Th2/1 hybrid cells that combine lineage transcription factors and cytokine expression patterns of CD4<sup>+</sup> Th1 as well as Th2 cells as integral parts of the immune response to nematodes (Bock et al.). Along the same line, Ahmed and colleagues analyze changes of the nematode-specific T cell response in a coinfection setting with an opposing pathogen (Ahmed et al.). In the helminth field the vast possibilities of working with the free-living model nematode Caenorhabditis elegans are often exploited to extrapolate findings to parasitic nematodes. Here, Midha et al. review the current knowledge and potential new directions to study reciprocal interactions of nematodes with their microbial environment (Midha et al.).

### APPRAISALS OF EXPERIMENTAL MODELS

Cell culture and animal models for parasitic diseases often remain suboptimal, and their correlation with the infection of the host species of interest is a frequent matter of debate. Dunst and coworkers review present knowledge on the relevance of cytokines in the etiology of cerebral malaria (Dunst et al.). Although the respective murine model can only offer an incomplete assessment of this major complication of human malaria, it permits the fine resolution of molecular mechanisms involved in the development of cerebral malaria. Dunst et al. integrate findings from the murine experimental malaria model with in vitro observations and results of clinical studies to deduce a potential sequence of pathophysiological events that entails the activation of endothelial cells and leukocyte recruitment and ultimately leads to permeability of the blood-brain barrier and neuroinflammation. Acknowledging the difficulties to establish a suitable model host for human P. falciparum, Cunha and colleagues present a Chlodronate-liposome-based protocol for new world common squirrel monkeys (Saimiri sciureus) to render these hosts susceptible to experimental P. falciparum inoculations by depletion of phagocytes (Cunha et al.). This might substitute the current common practice of surgical splenectomy.

Giardia duodenalis is a cause of chronic diarrheal disease globally, but not every infected host develops symptoms. Pathogenic mechanisms, such as the breakdown of intestinal barriers, have been primarily investigated using human intestinal cell lines, such as CaCo-2 cells. Kraft and colleagues attempt a critical analysis of the value of the latter as a model of pathogenesis based on literature and own observations (Kraft et al.). They propose that this set up rather models asymptomatic colonization and may serve to screen for till date elusive nonhost, non-parasite factors precipitating disease.

An appraisal of limitations and pitfalls, underexplored virtues and promises of small rodent models is contributed by Ehret and colleagues. They review illustrative examples and discuss the translational potential of non-standard rodent resources, such as collaborative crosses of Mus musculus, humanized mice, and wildtype rodents (Ehret et al.).

### NATURAL SYSTEMS

### Technological Advances For Investigating Natural Parasite-Host Systems

A prerequisite to study cognate parasite-host systems in nonmodel situations is proper identification of parasite species. DNA sequencing at low cost is enabling encyclopedic projects like the barcoding of all life (Blaxter, 2016; Hebert et al., 2016). Similarly, mass spectrometry-based species identification has also come of age also for pathogen identification, particularly in medical diagnostics (Greco and Cristea, 2017). Murugayan and Roesler review recent advances in this approach to identify pests, parasites, and vectors (Murugayan and Roesler). In their contribution, Bredtmann and colleagues compare barcoding and MS-based approaches and discuss their combination to solve the long pending issue of species differentiation of cyathostomine nematodes that parasitize horses (Bredtmann et al.). Heitlinger et al. report on their use of multi-amplicon sequencing to assess the load of different gastrointestinal parasites in free-living spotted hyenas (Crocuta crocuta) (Heitlinger et al.). Together with long-term data on these social animals their results suggest a higher diversity of eukaryotes, which include presumed parasitic taxa, in the intestine of high-ranking animals that usually exhibit a higher Darwinian fitness.

Modeling infectious processes is a highly dynamic field, and Ribeiro et al. contributed a report of their cellular automata/lattice gas modeling strategy to simulate immune cellparasite interaction in cutaneous leishmaniasis based on ectonuclease levels expressed by different Leishmania spp. (Ribeiro et al.). In general, it is probably the combination of modeling and technological advances that holds promise for exciting research directions in wildlife parasitology.

### Studies in Natural Systems

Several contributions addressed their parasitism-related research questions directly in the natural hosts. Jennes and co-workers study the response to T. gondii infections in pigs (Jennes et al.). Toxoplasma-infected pork meat is a major source driving the epidemiology of toxoplasmosis, for example in Germany (Wilking et al., 2016). Jennes et al. propose a strategy to reduce parasite burden in pork meat based on data from an epidemiologically relevant host—parasite pair of this zoonosis.

Zoonotic parasites of livestock are extremely relevant for global health and global food security. So is the good health of honey bees (Apis mellifera), which has direct impact on pollination efficacy and honey production. Gisder and colleagues contributed a longitudinal analysis of Nosema apis and the emerging N. ceranae prevalences in honey bees of North East Germany (Gisder et al.). Although infestation with N. ceranae has been linked to alarming colony losses in bee populations Gisder et al. find no evidence for a general fitness advantage of N. ceranae in their survey data. They propose climate-dependent relative advantages of one over the other Nosema species and show in an insect host cell model that N. ceranae has higher replication capacity than N. apis when temperatures are elevated.

While bees may classify as lifestock, Anopheles mosquitoes are best known as vectors of diverse pathogens. A key pathway governing their own immune response to infection is reviewed by Zakovic and Levashina who summarize knowledge on the role of the mosquitoe's NF-kB-like signaling pathway REL2 in the response to plasmodial infection (Zakovic and Levashina). Understanding these aspects nurture the hope for new avenues in vector-centered malaria management.

### THE URGENT NEED FOR BETTER MODELS AND METHODS IN NATURAL PARASITE SYSTEMS

Arguably, instructive models to test novel malaria vaccine candidates would be of very high translational relevance. A vaccine to prevent human malaria is WHO's target by the year 2030 (WHO, 2016b). Two contributions to this RT

### REFERENCES


discuss malaria vaccines. Jaurige and Seeberger review examples, strategy and status of vaccines consisting of carbohydrate antigens against malaria, but also against toxoplasmosis and leishmaniasis (Jaurige and Seeberger). The idea of a live attenuated vaccine has a long history in malaria research. Irradiated or chemically attenuated parasites have been used but may not be the ultimate solution. In that context, Kreutzfeld and coworkers discuss the promise that genetically attenuated Plasmodium spp., which do not develop beyond the liver stages, hold with respect to inducing a protective immune response (Kreutzfeld et al.). They conclude that realizing the potential of such parasites requires earnest investment into murine models to improve their relevance for translation.

We hope that this RT sparks interest for parasitic infections, which remain a research priority in medicine, veterinary medicine, and public health.

### AUTHOR CONTRIBUTIONS

SH, TA, and KM were the three editors of the research topic and thus put together the editorial for the RT "Parasite Infections: From experimental models to natural systems."

### FUNDING

The three authors received funds by the Deutsche Forschungsgemeinschaft: GRK 2046.


**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 Aebischer, Matuschewski and Hartmann. 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.

# Transcriptional Profiling Defines Histone Acetylation as a Regulator of Gene Expression during Human-to-Mosquito Transmission of the Malaria Parasite *Plasmodium falciparum*

Che J. Ngwa1 †, Meike J. Kiesow1 †, Olga Papst <sup>1</sup> , Lindsey M. Orchard<sup>2</sup> , Michael Filarsky <sup>3</sup> , Alina N. Rosinski <sup>1</sup> , Till S. Voss <sup>3</sup> , Manuel Llinás 2, 4 and Gabriele Pradel <sup>1</sup> \*

<sup>1</sup> Division of Cellular and Applied Infection Biology, RWTH Aachen University, Aachen, Germany, <sup>2</sup> Department of Biochemistry and Molecular Biology, The Pennsylvania State University, University Park, PA, United States, <sup>3</sup> Department of Medical Parasitology and Infection Biology, Swiss Tropical and Public Health Institute, Basel, Switzerland, <sup>4</sup> Department of Chemistry and Huck Center for Malaria Research, The Pennsylvania State University, University Park, PA, United States

#### *Edited by:*

Kai Matuschewski, Humboldt University of Berlin, Germany

#### *Reviewed by:*

Richard Bartfai, Radboud University Nijmegen, Netherlands Serge Ankri, Technion—Israel Institute of Technology, Israel

*\*Correspondence:*

Gabriele Pradel pradel@bio2.rwth-aachen.de † These authors have contributed equally to this work.

*Received:* 08 February 2017 *Accepted:* 28 June 2017 *Published:* 24 July 2017

#### *Citation:*

Ngwa CJ, Kiesow MJ, Papst O, Orchard LM, Filarsky M, Rosinski AN, Voss TS, Llinás M and Pradel G (2017) Transcriptional Profiling Defines Histone Acetylation as a Regulator of Gene Expression during Human-to-Mosquito Transmission of the Malaria Parasite Plasmodium falciparum. Front. Cell. Infect. Microbiol. 7:320. doi: 10.3389/fcimb.2017.00320 Transmission of the malaria parasite Plasmodium falciparum from the human to the mosquito is mediated by the intraerythrocytic gametocytes, which, once taken up during a blood meal, become activated to initiate sexual reproduction. Because gametocytes are the only parasite stages able to establish an infection in the mosquito, they are crucial for spreading the tropical disease. During gametocyte maturation, different repertoires of genes are switched on and off in a well-coordinated sequence, pointing to regulatory mechanisms of gene expression. While epigenetic gene control has been studied during erythrocytic schizogony of P. falciparum, little is known about this process during human-to-mosquito transmission of the parasite. To unveil the potential role of histone acetylation during gene expression in gametocytes, we carried out a microarray-based transcriptome analysis on gametocytes treated with the histone deacetylase inhibitor trichostatin A (TSA). TSA-treatment impaired gametocyte maturation and lead to histone hyper-acetylation in these stages. Comparative transcriptomics identified 294 transcripts, which were more than 2-fold up-regulated during gametocytogenesis following TSA-treatment. In activated gametocytes, which were less sensitive to TSA, the transcript levels of 48 genes were increased. TSA-treatment further led to repression of ∼145 genes in immature and mature gametocytes and 7 genes in activated gametocytes. Up-regulated genes are mainly associated with functions in invasion, cytoadherence, and protein export, while down-regulated genes could particularly be assigned to transcription and translation. Chromatin immunoprecipitation demonstrated a link between gene activation and histone acetylation for selected genes. Among the genes up-regulated in TSA-treated mature gametocytes was a gene encoding the ring finger (RING)-domain protein PfRNF1, a putative E3 ligase of the ubiquitin-mediated signaling pathway. Immunochemistry demonstrated PfRNF1 expression mainly in the sexual stages of P. falciparum with peak expression in stage II gametocytes, where the protein localized to the nucleus and cytoplasm. Pfrnf1 promoter and coding regions associated with acetylated histones, and TSA-treatment resulted in increased PfRNF1 levels. Our combined data point to an essential role of histone acetylation for gene regulation in gametocytes, which can be exploited for malaria transmission-blocking interventions.

Keywords: histone acetylation, gene expression, malaria, gametocyte, transmission

### INTRODUCTION

The mosquito-borne disease malaria is the most devastating infectious tropical disease in the world, causing ∼200 million new cases and more than 400,000 casualties annually (World Malaria Report, WHO, 2016). Malaria is caused by intracellularly living protozoa of the genus Plasmodium, with P. falciparum being the causative agent of malaria tropica, the most severe form of malaria.

The complex life-cycle of P. falciparum involves an initial round of replication in the human liver and subsequent 48-h replication cycles in the red blood cells (RBCs) that are pivotal for malaria pathogenesis. The virulence of P. falciparum is attributed to its ability to efficiently evade the host immune response, which includes molecular escape mechanisms to avoid complement and antibody recognition with the latter particularly depending on antigenic variation (reviewed in Scherf et al., 2008; Recker et al., 2011; Dinko and Pradel, 2016; Schmidt et al., 2016).

Immune evasion of P. falciparum is mediated by a tightly regulated transcription program with well-coordinated sequences of gene activation and silencing caused by chromatinmediated epigenetic regulatory mechanisms (Duraisingh et al., 2005; Freitas-Junior et al., 2005; reviewed in Duraisingh and Horn, 2016). A major part of epigenetic control involves histone post-translational modifications (PTMs). Among others, these include histone acetylation and methylation, which are mediated by specialized transferase enzymes, including histone acetyltransferases (HATs), which promote DNA accessibility, as well as histone methyl transferases (HMTs) which can either act as promotors or inhibitors of DNA accessibility, dependent on the methylation site (Lopez-Rubio et al., 2007; Sautel et al., 2007). Histone PTMs can also be reversed, e.g., via histone deacetylases (HDACs) which remove the acetyl groups and thus inhibit gene expression. The genome of P. falciparum encodes five plasmodial HDACs; i.e., PfHDAC1 and 3, PfHda2 and the two type III silent information regulators PfSir2A and PfSir2B (Joshi et al., 1999; Gardner et al., 2002; reviewed in Cui and Miao, 2010) and four HATs, including the previously reported MYST and PfGNC5 (Cui et al., 2007a; Miao et al., 2010). Further, the genes coding for 10 SET (Su(var)3-9-'Enhancer of zeste-Trithorax)-domaincontaining HMTs, termed PfSET1-10 were identified (Cui et al., 2008; Volz et al., 2010).

To date, histone PTMs were particularly studied during the expression of virulence-associated clonally variant multigene families, like the var gene family, which encodes the P. falciparum erythrocyte membrane protein PfEMP1 (Lopez-Rubio et al., 2007, 2009; Petter et al., 2011; reviewed in Llinás et al., 2008; Cui and Miao, 2010; Duffy et al., 2014; Duraisingh and Horn, 2016). Only a single var gene is expressed during replication of the RBC parasites at any one time, whereas all other var genes remain silent. Switching var expression and thus PfEMP1 structure alters the antigenic type of the infected RBCs and in consequence pathogenesis of the tropical disease. The expression of var genes relies on epigenetic mechanisms that induce dynamic changes in the chromatin structure (reviewed in Duffy et al., 2012). Only the active var gene copy assumes a euchromatic state characterized by both acetylated lysine 9 and tri-methylated lysine 4 of histone 3 (H3K9ac and H3K4me3, respectively; Lopez-Rubio et al., 2007; Salcedo-Amaya et al., 2009). On the other hand, var gene silencing is linked to H3K9 tri-methylation (H3K9me3) and further involves Sir2A and B and the class II HDAC PfHda2 (Duraisingh et al., 2005; Freitas-Junior et al., 2005; Lopez-Rubio et al., 2009; Tonkin et al., 2009; Coleman et al., 2014).

Recent work has further unveiled epigenetic control mechanisms during gametocyte commitment, when the RBC parasites enter the sexual pathway to form gametocytes, thus enabling parasite transmission from the human to the mosquito vector (reviewed in Josling and Llinás, 2015). Gametocyte commitment, which is proposed to be triggered by environmental signals (reviewed in Kuehn and Pradel, 2010; Bennink et al., 2016), is closely linked to the plasmodial heterochromatin protein PfHP1. This regulator usually binds specifically to H3K9me3 to maintain the heterochromatin state, resulting in var gene silencing and suppression of gametocyte commitment (Flueck et al., 2009; Pérez-Toledo et al., 2009; Brancucci et al., 2014). Conditional depletion of PfHP1 leads to hyper-induction of gametocytes caused by the de-repression of the ap2-g gene, which encodes the AP2-G transcription factor, a member of the apicomplexan Apetala2/ethylene response factor (AP2/ERF) DNA-binding protein family (Balaji et al., 2005; Kafsack et al., 2014; Sinha et al., 2014). Besides PfHP1, PfHda2 appears to be involved in the silencing of ap2-g gene expression in non-committed parasites, probably by removing acetylated histone residues allowing for their methylation leading to the binding of PfHP1 (Coleman et al., 2014). Once the ap2-g gene is activated and AP2-G becomes synthesized, the protein acts as a transcriptional switch that controls gametocyte differentiation by activating the transcription of early gametocyte genes (Kafsack et al., 2014; Sinha et al., 2014; reviewed in Voss et al., 2014; Josling and Llinás, 2015).

Once gametocyte commitment is induced, a total of about 20% of plasmodial genes are specifically expressed. These are needed for gametocyte maturation, but also for preparing the parasite to rapidly adjust to the mosquito midgut environment and to undergo gametogenesis, after the gametocytes are taken up by the blood-feeding female Anopheles (Florens et al., 2002; Lasonder et al., 2002; Le Roch et al., 2003). A suppression subtractive hybridization study identified 126 genes that changed in expression during initiation of gametogenesis, amongst others with putative functions in signaling, cell cycle, and gene expression (Ngwa et al., 2013). However, the mechanisms underlying differential gene regulation during gametocyte maturation and gametogenesis up to date have not been investigated.

The therapeutic use of epigenetic inhibitors in treatment of cancers has been known for more than a decade, and several HDAC inhibitors like vorinostat and romidepsin have meanwhile been approved for anticancer therapy (e.g., reviewed in Schobert and Biersack, 2017; Zagni et al., 2017). Also, the antimalarial effects of inhibitors targeting HAT and HDAC enzymes have been explored in the past (e.g., Cui et al., 2007b; Andrews et al., 2008, 2009, 2012a; Chaal et al., 2010; Wheatley et al., 2010; Sumanadasa et al., 2012; Engel et al., 2015; Alves Avelar et al., 2017; Chua et al., 2017). Additional studies showed that HDAC inhibitors also exhibit gametocytocidal activities in vitro, indicating that the enzymes are essential for gametocytogenesis (e.g., Hansen et al., 2014; Sun et al., 2014; Trenholme et al., 2014).

To date, only a few studies have investigated the transcriptional changes in P. falciparum following treatment with HDAC inhibitors. An initial study reported a general deregulation of gene expression following treatment with the HDAC inhibitor trichostatin A (TSA) during the erythrocytic replication cycle with up to 60% of the genome affected (Hu et al., 2010). Follow-up experiments, though, comparing the transcription profiles in blood stage parasites following treatment with three different HDAC inhibitors, could only detect an overlap for two genes with altered expression (Andrews et al., 2012b). The effect of HDAC inhibitors on gene expression in gametocytes has hitherto not been addressed, amongst others due to the technical challenge of harvesting high numbers of pure gametocyte stages.

To unveil the potential role of histone acetylation during gene expression in gametocytes, we now have carried out a microarray-based transcriptome analysis, in which we compared the transcriptomes of gametocytes treated with TSA and of untreated gametocytes. We show that TSA-treatment results in the deregulation of 453 genes, demonstrating a crucial role of histone PMTs in preparing the parasite for human-to-mosquito transmission.

### MATERIALS AND METHODS

### Antibodies

Primary antibodies used in this study included: rabbit anti- (tetra)-acetyl histone H4 K5, 8, 12, 16ac (H4Kac4) (Millipore; note: according the manufacturer's material data sheet this antibody may cross-react with other acetylated histones like H2B); rabbit anti-H3K9ac (Diagenode); rabbit anti-PfHP1 (Brancucci et al., 2014); rabbit IgG antibody (Millipore); mouse/rabbit anti-Pfs230 (Ngwa et al., 2013; Simon et al., 2016), rabbit anti-Pfs25 (BEI Resources); mouse anti-Pf39 (Scholz et al., 2008); rabbit anti-HA (Sigma Aldrich); mouse anti-proteasome SU α5 (Aminake et al., 2012), and anti-PfActinI (Ngwa et al., 2013). Mouse anti-PfRNF1 was generated for this study (see below). For indirect immunofluorescence assays (IFAs), the following dilutions of the antibodies were used: anti-H4KAc4 (1:200), anti-H3K9ac (1:200), mouse/rabbit anti-Pfs230 (1:200), anti-Pfs25 (1:1,000), anti-PfRNF1 (1:20), anti-PfHP1 (1:300), and anti-proteasome SU α5 (1:50). For Western blot (WB) analysis the following dilutions were used: anti-H4Kac4 (1:1,000), anti-H3K9ac (1:1,000), anti-Pf39 (1:1,000), anti-PfActinI (1:200), anti-PfRNF1 (1:200), rabbit anti-HA (1:1,000). For chromatin immunoprecipitation (ChIP) assays, 1 µg of each antibody (anti-H4Kac4, anti-H3K9ac, anti-PfHP1, IgG) was used.

## Parasite Culture

P. falciparum strain NF54 was used in this study. The parasites were cultivated in vitro in RPMI 1640 medium supplemented with 10% heat-inactivated human serum as described (Ifediba and Vanderberg, 1981) and cultures were maintained at 37◦C at an atmosphere of 5% O2, 5% CO2, and 90% N2. Cultures were synchronized by repeated sorbitol treatment as described (Lambros and Vanderberg, 1979). To generate gametocytes, the cultures were kept at high parasitaemia and gametocytogenesis was induced following addition of lysed RBCs. As soon as stage I gametocytes started to emerge in the culture, the culture medium was supplemented with 50 mM N-acetyl glucosamine (GlcNac) for ∼5 days to kill the asexual blood stages (Fivelman et al., 2007). The gametocyte culture was then maintained in normal culture medium without GlcNac until immature (stage II–IV) or mature stage V gametocytes were harvested and enriched by Percoll gradient purification (Kariuki et al., 1998). In order to obtain activated gametocytes, Percoll-enriched mature gametocytes were incubated with 100 µM xanthurenic acid (XA) for 30 min, 1 or 6 h at room temperature (RT). Giemsa-staining of purified gametocyte smears was used to confirm purity of the samples. The human erythrocyte concentrate and serum used in this study were purchased from the Department of Transfusion Medicine (University Hospital Aachen, Germany). The University Hospital Aachen Ethics commission approved all work with human blood, the donors remained anonymous, and serum samples were pooled.

### Malstat Assay

To determine the antimalarial effect of the histone deacetylase inhibitor TSA (Sigma-Aldrich), a Malstat assay was performed as described previously (Aminake et al., 2011). Synchronized ring stages of P. falciparum strain NF54 were plated in triplicate in 96-well plates (200 µl/well) at a parasitaemia of 1% in the presence of TSA dissolved in 0.5% vol. ethanol (5 µM to 0.06 nM). Chloroquine, dissolved in double-distilled water, served as a positive control in the experiments. Incubation of parasites with ethanol alone at a concentration of 0.5% vol. was used as negative control. Parasites were cultivated in vitro for 72 h, resuspended, and aliquots of 20 µl were removed and added to 100 µl of the Malstat reagent in a 96-well microtiter plate. The assessment of parasite lactate dehydrogenase (pLDH) activity was obtained by adding 20 µl of a mixture of NBT (Nitro Blue Tetrazolium) and diaphorase (1:1; 1 mg/ml stock each) to the Malstat reaction, and optical densities were measured at 630 nm. Each compound was tested four times, and the IC<sup>50</sup> values were calculated from variable-slope sigmoidal dose-response curves using the GraphPad Prism program version 4.

### Gametocyte Toxicity Test

P. falciparum strain NF54 parasites were grown at high parasitaemia to induce gametocyte formation. Upon appearance of stage II gametocytes, 1 ml of culture was aliquoted in triplicate in a 24-well plate in the presence of TSA at asexual blood stage IC<sup>50</sup> (29 nM) and IC<sup>90</sup> (0.26 µM) concentrations. Incubation of parasites with ethanol at a concentration of 0.5% vol. and chloroquine (IC50), dissolved in double-distilled water, were used as negative controls. The proteasome inhibitor epoxomicin (60 nM), diluted in DMSO, served as a positive control in the experiments. The gametocytes were cultivated for 10 d with daily medium replacement. For the first 2 days of cultivation, the gametocytes were treated with the inhibitors, subsequently the medium was inhibitor-free. Every second day, Giemsastained blood smears were prepared and the gametocytemia was evaluated by determining the numbers of gametocytes of stages II to V in a total number of 1,000 erythrocytes in triplicate. The student's t-test was used to determine significant differences between TSA-treated and untreated samples.

### Macrogamete and Zygote Development Assays

Equal volumes of mature gametocyte cultures were incubated with TSA at IC<sup>50</sup> or IC<sup>90</sup> concentrations or with 0.5% vol. ethanol for negative control for 1 h at 37◦C. Subsequently, the cultures were activated with 100 µM XA and incubated for 30 min (macrogamete development assay) or 12 h (zygote development assay) at RT. An equal volume of each sample was then coated on Teflon slides and the cells were immunolabeled with anti-Pfs25 as described below. The numbers of Pfs25-positive macrogametes or zygotes, as distinguished by their round shapes, were counted for a total number of 1,000 erythrocytes, which were visualized by differential interference contrast in triplicate, using a Leica DMLS microscope at 1,000-fold magnification. Zygotes were distinguished from macrogametes by their larger nuclei through Hoechst nuclear stain 33342 (Molecular Probes). The student's t-test was used to determine significant differences between TSAtreated and untreated samples.

### Indirect Immunofluorescence Assay

P. falciparum cultures (of the wild-type NF54 strain or the PfRNF1-HA-Strep-expressing transfectant line) were air-dried on glass slides and fixed for 10 min in a methanol bath at −80◦C. For membrane permeabilization and blocking of nonspecific binding, fixed cells were successfully incubated in 0.01% saponin/0.5% BSA/PBS and 1% vol. neutral goat serum (Sigma-Aldrich)/PBS each for 30 min at RT. For labeling of PfHP1, the air-dried samples were fixed with 4% paraformaldehyde/PBS, pH 7.4, for 10 min at RT and subsequently treated with 0.1% vol. Triton X-100/125 mM glycine (Carl Roth)/PBS at RT for 30 min, followed by blocking of non-specific binding sites with 3% BSA/PBS for 1 h. Preparations were then incubated with the primary antibody diluted in 0.01% saponin/0.5% BSA/PBS for 1.5 h each at 37◦C. Binding of primary antibody was visualized by incubating the preparations with Alexa Fluor 488-conjugated goat anti-mouse or anti-rabbit IgG secondary antibody (Molecular Probes) diluted in 0.01% saponin/0.5% BSA/PBS for 1 h at 37◦C. The different parasite stages were detected by double-labeling with stage-specific antibodies, i.e., polyclonal rabbit or mouse antisera directed against PfMSP1 for the detection of asexual blood stages and Pfs230 and Pfs25 for the detection of gametocytes and activated gametocytes, respectively. This was followed by incubation with Alexa Fluor 594-conjugated goat anti-rabbit or anti-mouse IgG secondary antibodies (Molecular Probes) diluted in 0.01% saponin/0.5% BSA/PBS for 1 h at 37◦C. Nuclei were highlighted by treatment with Hoechst nuclear stain 33342 for 1 min at RT and cells were mounted with anti-fading solution AF2 (Citifluor Ltd) and sealed with nail varnish. In cases where double-labeling was not employed, counterstaining of erythrocytes was performed using 0.05% Evans Blue/PBS (Sigma-Aldrich) for 1 min. Digital images were taken using a Leica AF 6000 microscope and processed using Adobe Photoshop CS software.

### Western Blot Analysis

Percoll-enriched immature (stages II–IV) and mature (stage V) gametocytes (of the wild-type NF54 strain or the PfRNF1-HA-Strep-expressing transfectant line) were harvested as described above and erythrocytes were lysed with 0.05% saponin/PBS followed by a washing step with PBS to remove the hemoglobin. The pellets were resuspended and sonicated in NP-40 lysis buffer (50 mM Tris HCl pH 8.0, 150 mM NaCl, 1% vol. NP-40) supplemented with a protease inhibitor cocktail (Roche Diagnostics, Germany). SDS-PAGE loading buffer was then added to the lysates, heat-denatured for 10 min at 95◦C, and separated via SDS-PAGE and transferred to Hybond ECL nitrocellulose membrane (Amersham Biosciences) according to the manufacturer's protocol. Membranes were blocked for nonspecific binding by incubation in Tris-buffered saline containing 5% skim milk and 1% BSA, followed by incubation with the respective rabbit or mouse antibody for 2 h at RT. After washing, the membranes were incubated with an alkaline phosphataseconjugated anti-rabbit or anti-mouse IgG secondary antibody (Sigma-Aldrich) for 1 h at RT and developed in a solution of nitroblue tetrazolium chloride (NBT) and 5-bromo-4-chloro-3 indoxyl phosphate (BCIP; Sigma-Aldrich) for 5–30 min. Scanned blots were processed using Adobe Photoshop CS software.

### Histone Hyper-Acetylation Assay

To investigate histone hyper-acetylation caused by TSAtreatment, hyper-acetylation assays were carried out as previously described (Andrews et al., 2012b). Percoll-enriched immature (stages II–IV) and mature (stage V) gametocytes were treated with TSA at IC<sup>90</sup> concentrations or with 0.5% vol. ethanol (negative control) for 1 and 6 h at 37◦C, respectively. Protein lysates were generated, employed to SDS-PAGE and histone hyper-acetylation was analyzed by WB analysis using anti-H4Kac4 and anti-H3K9ac antibodies as described above. Immunoblotting with anti-Pf39 antisera was used as loading control. Histone hyper-acetylation was quantified from three to six different experiments by measuring the band intensities via the Image J programme. When immunoblotting with anti-H4Kac4 antibody, total acetylation bands were quantified. The related band intensity was normalized to Pf39 and compared with respect to untreated samples.

### Microarray Analysis

Total RNA was isolated from enriched immature (stages II–IV) and mature (stage V) gametocytes and gametocytes at 1 h postactivation following treatment with TSA at IC<sup>90</sup> concentrations or with 0.5% vol. ethanol (untreated control) for 1 and 6 h, respectively, using the Trizol reagent (Invitrogen) according to the manufacturer's protocol. Quality of RNA samples were assessed using a ND-1000 (NanoDrop Technologies, Thermo Scientific) and by agarose gel electrophoresis. The microarray experiments were carried out as described previously (Kafsack et al., 2012). Briefly, synthesis of first strand amino-allyl cDNA was performed using Superscript II reverse transcriptase (Invitrogen). The amino-allyl cDNA was then cleaned and concentrated using the Zymo DNA clean and concentrator-5 column (Zymo Research) followed by coupling with Cy5 dye (GE Healthcare). The reference pool consisted of a mixture RNA from asexual blood stages and gametocytes, in which synthesis of first strand amino-allyl cDNA was performed as describe above, and coupled with the Cy3 dye. Equal amounts of Cy5-labeled samples from each treatment and the Cy3-labeled reference pool were subjected to array hybridization for 17 h at 65◦C using a P. falciparum DNA Agilent microarray chip (Agilent Technologies AMADID #037237) containing the 5,363 coding genes (Bozdech et al., 2003). The arrays were scanned using the Agilent scanner G2600D (Agilent Technologies). Normalized intensities were extracted using the Agilent Feature Extraction Software version 11.5.1.1 and uploaded to the Princeton University Microarray Database (PUMA.princeton.edu) for analysis. After background subtraction, the log2 of the (Cy5/Cy3) intensity ratio was extracted. Raw intensity data have been submitted to the NCBI Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/ geo/) under accession number GSE99223. Transcript abundance of treated samples were compared to that of untreated samples. For the selection of up- and down-regulated genes, a cut-off value of greater than 2-fold for at least one of the two time-points with a consistent up- or down-regulation for both time-points was chosen. A cut-off value of greater than 2-fold for both time points was considered significant. Data were analyzed using the database PlasmoDB (http://plasmodb.org/plasmo; Aurrecoechea et al., 2009) and Microsoft Excel 2010.

### Real-Time RT-PCR

To validate the microarray data, total RNA was isolated from enriched immature (stages II–IV) and mature (stage V) gametocytes following treatment with TSA at IC<sup>90</sup> concentration or with 0.5% vol. ethanol (untreated control) for 1 h as described above. One microgram of each total RNA sample was used for cDNA synthesis using the SuperScript III First-Strand Synthesis System (Invitrogen), following the manufacturer's instructions. The synthesized cDNA was first tested by diagnostic PCR for asexual blood stage contamination using primers specific for the gene encoding the apical membrane antigen AMA-1 (Peterson et al., 1989; Narum and Thomas, 1994) and for gametocytespecificity using primers specific for the gene encoding the LCCL-domain protein PfCCp2 (Pradel et al., 2004; Ngwa et al., 2013). Controls without reverse transcriptase were also used to investigate potential genomic DNA (gDNA) contamination by using pfccp2 primers (for primer sequences, see Table S1). Primers for quantitative real time RT-PCR to confirm the upregulation of selected genes as identified by microarray analysis were designed using the Primer 3 software (http://frodo.wi.mit. edu/primer3/) and tested on gDNA in conventional PCR to confirm primer specificity (for primer sequences, see Table S1). Real time RT-PCR measurements were performed using the Bio-Rad iQ5 Real-Time Detection System. Reactions were performed in triplicate in a total volume of 20 µl using the maxima SyBR green qPCR master mix according to manufacturer's instructions (Thermo Scientific, Germany). Controls without template and without reverse transcriptase were included in all real time RT-PCR experiments. Transcript expression levels were calculated by the 2−1Ct method (Livak and Schmittgen, 2001) using the endogenous control gene encoding the P. falciparum seryl tRNAligase (PF3D7\_0717700) as reference (Salanti et al., 2003), which was confirmed not to be affected in its transcript levels by TSAtreatment.

### Chromatin

### Immunoprecipitation-Quantitative PCR

To provide evidence for a link between gene expression and histone acetylation for selected genes identified by microarray analysis, ChIP assays combined with subsequent-quantitative PCR (ChIP-qPCR) were carried as previously described (Flueck et al., 2009). Mature (stage V) gametocytes were enriched by Percoll gradient purification, treated with either 0.5% vol. ethanol (untreated) or TSA at IC<sup>90</sup> concentrations for 6 h (TSA-treated) and then resuspended in RPMI 1640 medium containing human erythrocyte concentrate at 5% haematocrit. Crosslinking of gametocyte chromatin was triggered by incubation of the cultures with 1% formaldehyde (Sigma-Aldrich) for 10 min at 37◦C and a termination of the reaction by addition of 0.125 M glycine diluted in doubled-distilled water. The RBCs were then lysed using 0.15% saponin/PBS and crosslinked nuclei harvested and separated from cytoplasmic proteins by the use of a 0.25 M cytoplasmic lysis buffer (20 nM Hepes, 10 mM KCl, 1 mM EDTA, 1 mM EGTA, 0.65% NP-40, 1 mM DTT, 1x protease inhibitor cocktail). The nuclei of 10<sup>8</sup> mature gametocytes were then pooled and the nuclei were sheared by sonication with UPH50 (Hielscher) on ice for 40 intervals with each interval composed of 10 s sonication and 50 s resting cycles to gain fragment sizes <500 bp. A total amount of 400–450 ng chromatin was incubated under rotation with 1 µg antibody (anti-H3K9ac, anti-H4Kac4, anti-PfHP1, and rabbit IgG antibody for negative control) overnight at 4◦C in the presence of 20 µl of protein A- and G-coated magnetic beads (Diagenode). After six washing steps the immunoprecipitated chromatin was eluted by adding 1% SDS and 0.1 M NaHCO<sup>3</sup> as DNA elution buffer and decrosslinked at 65◦C overnight. DNA purification was carried out using the PCR and Gel Clean Up

Kit (Macherey-Nagel). Primers for qPCR for selected genes as identified by microarrays were designed using Primer 3 software and tested on P. falciparum gDNA in conventional PCR to confirm primer specificity (for primer sequences, see Table S1). Quantitative PCR measurements were performed as described above. The amount of recovered target DNA gained from untreated and TSA-treated gametocyte samples was compared to associated input DNA sample (1:10) and depicted as percentage of input for each chosen gene.

### Recombinant Protein Expression and Production of Mouse Antisera

A recombinant peptide, corresponding to the N-terminal region of PfRNF1 (Figure S1A), was expressed as a maltose-binding protein-tagged fusion protein using the pMALTMc5X-vector (New England Biolabs). DNA was amplified by PCR using gene-specific primers (for primer sequences, see Table S1). Recombinant protein was expressed in E. coli BL21(DE3)RIL cells according to the manufacturer's protocol (Invitrogen) and isolated and affinity-purified using amylose resin according to the manufacturer's protocol (New England Biolabs). Polyclonal antisera were generated by immunization of 6-weeks old female NMRI mice (Charles River Laboratories) with subcutaneous injections of 100 µg recombinant protein emulsified in Freund's incomplete adjuvant (Sigma-Aldrich) followed by a boost after 4 weeks. Mice were anesthetized at day 10 after the boost by intraperitoneal injection of a mixture of ketamine and xylazine according to the manufacturer's protocol (Sigma-Aldrich), and immune sera were collected via heart puncture. The immune sera of three mice immunized were pooled; sera of three non-immunized mice (NMS) were used as negative control. Experiments in mice were approved by the animal welfare committee of the District Council of Cologne, Germany (ref. no. 84-02.05.30.12.097 TVA).

### Generation of a PfRNF1-HA-Strep-Tagged Parasite Line

To tag PfRNF1 with hemagglutinin (HA)-streptavidin (Strep) at the C-terminus, a 1,230 bp homologous gene fragment was amplified from P. falciparum NF54 gDNA using gene-specific primers (for primer sequences, see Table S1). Cloning was done using the pHAST vector (kindly provided by Alex Maier, ANU Canberra; Rug and Maier, 2013) with the help of the SacII/XhoI restriction sites. A P. falciparum strain NF54 culture with 5% ring stages was loaded with 100 µg of the pHAST-PfRNF1 construct in transfection buffer via electroporation (parameters: 310 V 950 µF, 13 ms; Bio-Rad gene-pulser) as described (Wirth et al., 2014). WR99210 was added to a final concentration of 2.5 nM, starting at 4 h after transfection. WR99210-resistant parasites appeared after 4 weeks. After 60–90 days of drug pressure, the respective cultures were investigated for plasmid-integration by diagnostic PCR. The gDNA of the transfected parasites was isolated using the NucleoSpin Blood Kit (Macherey-Nagel) according to the manufacturer's protocol and used as template in the diagnostic PCR to test for vector integration (for primer sequences, see Table S1; for primer location, see Figures S1B,C). Once integration was confirmed, a clonal dilution was carried out to select for single PfRNF1-HA-Strep-tagged parasite clones and one clone was used for characterization.

### Determination of Protein Expression of PfRNF1 Following TSA-Treatment

To determine if the up-regulation of PfRNF1 transcript expression following TSA-treatment corroborates with the upregulation of protein expression, purified mature gametocytes of the P. falciparum wild-type NF54 strain or the PfRNF1-HA-Strep-expressing transfectant line were splitted in two equal volumes and each was pipetted into a pre-warmed 96-well plate. One part of the culture was treated with TSA at IC<sup>90</sup> concentrations and the other part was treated with 0.5% vol. ethanol (untreated control). The samples were incubated for 24 h at 37◦C in an atmosphere of 5% O2, 5% CO2, and 90% N2. Protein lysates were generated, separated via SDS-PAGE and protein levels were analyzed by WB analysis using anti-PfRNF1 antisera or anti-HA antibody as described above. Anti-Pf39 antisera was used as loading control. PfRNF1 levels were quantified for untreated and TSA-treated samples from six (for the wild-type) and two (for the transfectant) different experiments by measuring the band intensities by Image J. The related band intensities were normalized to Pf39 and compared with respect to untreated samples.

### Parasite Sub-cellular Fractionation

Nuclear and cytosolic fractions of P. falciparum strain NF54 parasites were prepared as previously described (Voss et al., 2002). Enriched immature (stages II–IV) gametocytes were treated with 0.1% saponin in PBS to lyse RBCs and washed twice with PBS. The parasite pellet was then resuspended in cold lysis buffer (20 mM Hepes, pH 7.8, 10 mM KCl, 1 mM EDTA, 1 mM DTT, 1 mM PMSF, 1% Triton X100) and incubated for 5 min on ice. The nuclei were pelleted at 2,500 g for 5 min at 4◦C and the supernatant containing the cytoplasmic proteins was collected. The nuclear pellet was washed three times with lysis buffer and resuspended in twice the pellet volume of the extraction buffer (20 mM Hepes, pH 7.8, 800 mM KCl, 1 mM EDTA, 1 mM DTT, 1 mM PMSF, 1x protease inhibitor cocktail). Following incubation under rotation for 30 min at 4◦C, the extract was cleared by centrifugation at 13,000 g for 30 min and 4◦C. The supernatant containing the nuclear fraction was then diluted with 1 volume of dilution buffer (20 mM Hepes, pH 7.8, 1 mM EDTA, 1 mM DTT, 30% vol. glycerol). The nuclear and cytoplasmic fractions were subjected to WB using anti-PfRNF1 antisera as described above.

## RESULTS

### TSA-Treatment Affects *P. falciparum* Blood and Sexual Stage Development

The effect of TSA on the blood and sexual stages of P. falciparum was tested in vitro. Malstat assays, which measure the pLDH activities demonstrated that TSA-treatment inhibited blood stage replication with a mean IC<sup>50</sup> value of 29 nM and an estimated IC<sup>90</sup> of 0.26 µM. Chloroquine-treatment was used for positive control in the assay and resulted in parasite growth inhibition with a mean IC<sup>50</sup> value of 16 nM (**Table 1**).

We next tested the effect of TSA on gametocyte development. In this regard a culture of mainly stage II gametocytes was grown in 24-well plates in triplicate in the presence of TSA at IC<sup>50</sup> and IC<sup>90</sup> concentrations (as determined by Malstat assay) or in the presence of 0.5% vol. ethanol (untreated control) for 48 h. The parasites were subsequently cultured for another 8 days without inhibitor. The gametocytemia was determined every 2 days via Giemsa-stained blood smears. We showed that the numbers of stage IV and V gametocytes formed on day 10 were 43% less when the stage II gametocytes were treated with TSA at IC<sup>50</sup> concentrations and 67% less when treated with IC<sup>90</sup> concentrations as compared to controls (0.5% vol. ethanol and 16 nM of chloroquine) (**Figure 1A**). Treatment of stage II gametocytes with 60 nM epoxomicin as positive control resulted in complete elimination of stage IV and V gametocytes on day 10 (data not shown). We also investigated in detail the differentiation of gametocytes from stage II to stage V during incubation with TSA. TSA-treatment particularly affected gametocytes of stages II and III, but did not result in any delay of gametocyte maturation compared to the controls (Figure S2).

The effect of TSA on male gamete formation (termed exflagellation) has already been demonstrated for P. falciparum (Trenholme et al., 2014). The impairment of exflagellation by TSA is minor with IC<sup>50</sup> values of 0.22 ± 0.04 mM (**Table 1**). To determine if TSA-treatment affects the formation of female macrogametes and zygotes, gametocyte cultures were incubated with TSA at IC<sup>50</sup> and IC<sup>90</sup> concentrations or with 0.5% vol. ethanol (untreated control) for 1 h at 37◦C. The cultures were activated with 100 µM XA and further cultured for 30 min and 12 h to detect macrogametes and zygotes, respectively. Both stages were immunolabeled with anti-Pfs25 antibodies and counted for a total of 1,000 erythrocytes in triplicate. Comparative analyses demonstrated a slight reduction of macrogamete numbers, when these were treated with TSA at IC<sup>90</sup> concentration, while the numbers of zygotes decreased by 21 and 24%, when treated with TSA at IC<sup>50</sup> and IC<sup>90</sup> concentrations, respectively (**Figures 1B,C**).

### Treatment of Gametocytes with TSA Causes Histone Hyper-Acetylation

To assess the extent of histone acetylation during P. falciparum gametocyte development, two commercially available histone acetylation antibodies (anti-H3K9ac and anti-H4Kac4) were used to detect histone acetylation in gametocytes by IFA. Histone

TABLE 1 | Antimalarial activities of TSA against the P. falciparum blood and microgamete stages.


N/A, Not applicable; <sup>a</sup>published in Trenholme et al. (2014).

acetylation was detected throughout the nuclei of gametocyte stages II-V highlighted by Hoechst staining and immunolabeling of Pfs230, respectively (**Figures 2A,B**). WB analysis using the histone acetylation antibodies detected protein bands migrating at molecular weights of 15 kDa (for anti-H3K9ac) and of 11 and 13 kDa (for anti-H4Kac4). These molecular weights are in accord with the expected molecular weights of 15.4 kDa for histone H3 and 11.5 kDa for histone H4. The second protein band detected by the anti-H4Kac4 antibody might represent H2B, which has an expected molecular weight of 13.1 kDa, since the antibody also recognizes this acetylated histone as indicated in the manufacturer's material data sheet. The intensities of the protein bands increased, when the gametocytes were treated with TSA at IC<sup>90</sup> concentrations, demonstrating hyper-acetylation of the histones following TSA treatment (**Figures 2C,D**). Quantitative WB analysis, measuring the protein band intensity of total acetylated histones showed a significant increase in their acetylation levels, when the immature and mature gametocytes were treated with TSA (**Figures 2E,F**).

### TSA-Treatment Results in the Deregulation of Genes during Gametocyte Development

We next aimed to identify genes deregulated following treatment of gametocytes with TSA. Immature, mature and activated gametocytes were treated with TSA for 1 or 6 h, total RNA was isolated, and the RNA of untreated cultures was used for comparative controls. Beforehand, the purity of the cell samples was confirmed via Giemsa smears (Figure S3A). Following cDNA synthesis, the samples were applied to a P. falciparum DNA Agilent microarray chip containing DNA spots corresponding to the 5,363 coding genes of nucleus, mitochondrion and apicoplast (Bozdech et al., 2003; Kafsack et al., 2012) (Table S2). Genes with transcript levels greater (i.e., up-regulated genes) or lower (i.e., down-regulated) than 2-fold compared to the untreated control (0.5% vol. ethanol) for at least one of the two time-points combined with a consistent up- or down-regulation for both time-points were used for further analysis. Changes in transcript levels greater than 2-fold for both time points compared to the control were considered significant.

In immature and mature gametocytes a total of 219 and 214 genes, respectively, were identified by comparative transcript analysis that were more than 2-fold up-regulated in their expression levels, when these stages were treated with TSA. The up-regulations for 120 and 76 of these transcripts, respectively, were considered significant. Accordingly, transcripts of 90 and 66 genes were more than 2-fold down-regulated in TSA-treated immature and mature gametocytes with a significant downregulation being observed for 8 and 11 genes, respectively. In activated gametocytes, which were less sensitive to TSAtreatment, the transcript levels of 48 genes were non-significantly increased and were decreased for 7 genes (**Figure 3A**, Table S3). The average up-regulation values for immature, mature and activated gametocytes were in the range of 1.3 to 3.2-fold absolute changes (immature: 1 h, 2.32; 6 h, 2.58: mature: 1 h, 2.03; 6 h, 3.16; activated: 1 h, 1,31; 6 h, 2.36), and the down-regulation values ranged between 0.7 and 0.5-fold absolute changes (immature: 1 h,

0.70; 6 h, 0.49; mature: 1 h, 0.62; 6 h, 0.47; activated: 1 h, 0.72; 6 h, 0.51) (**Figure 3B**). A total of 29 genes that were up-regulated in their transcript levels were shared by immature, mature and activated gametocytes, but none of the down-regulated genes were shared by the three gametocyte stages (**Figure 3C**).

0.001; ns, not significant; Student's t-test.

For each of the identified genes, the microarray-based transcript data were compared with the transcriptomics data published in the database PlasmoDB. The comparison revealed that roughly 23 and 22% of the genes transcriptionally upregulated in immature and mature gametocytes following TSAtreatment, respectively, have their peak expression in mature stage V gametocytes and in ookinetes (Figure S4; Table S3). Genes down-regulated in immature gametocytes following TSAtreatment were mainly expressed in gametocytes of stages II and V (25 and 23%, respectively), while for the mature gametocytes 58% of down-regulated genes showed their peak expression in gametocytes of stage V. In activated gametocytes, 42% of genes up-regulated after treatment with TSA were highly expressed in ookinetes, whereas all of the down-regulated genes exhibited peak expression profiles in gametocytes of stage V (Figure S4; Table S3).

The identified genes were grouped according to their predicted functions as indicated in PlasmoDB. The in-silico analyses demonstrated that in TSA-treated immature and mature gametocytes the up-regulated genes mainly associated with functions in gene expression and transcription and in antigenic variation and cytoadherence, or they code for exported proteins. Furthermore, in immature gametocytes, genes associated with RBC invasion and proteostasis were significantly up-regulated (**Figure 3D**; Figures S5A,B). The up-regulated genes mostly included ones coding for SURFINs and RIFINs, for the merozoite surface protein (MSP) family or the rhoptry neck (RON) family and for the PHIST family (Mphande et al., 2008; Proellocks et al., 2010; Beeson et al., 2016; Warncke et al., 2016). Also, genes with functions in the mosquito-specific stages were activated, when the gametocytes were treated with TSA. Downregulated genes included ones assigned to gene expression and transcription and translation (**Figure 3D**). Furthermore, genes with assigned functions in metabolism and signaling as well as components of the cytoskeleton and the inner membrane complex were affected in their activation by TSA-treatment (Table S3).

In TSA-treated immature and mature gametocytes, several genes with more than 5-fold up-regulated transcription were identified, i.e., four genes in the immature gametocyte samples and 16 genes in the mature gametocyte samples (**Table 2**). The genes could mostly be assigned to signaling, cell cycle and DNA replication, gene expression and transcription, and proteostasis, or they included genes encoding exported proteins. Six of the genes had unknown function. One gene, encoding for the RBC invasion-related protein MSRP4 (merozoite surface protein 7 related protein 4), was identified in both immature and mature gametocytes.

To validate the microarray array data, a total 32 genes (immature gametocytes: 15 genes; mature gametocytes: 17 genes) transcriptionally up-regulated following TSA-treatment were randomly selected. The gametocyte cultures were treated with TSA at IC<sup>90</sup> concentration or 0.5% vol. ethanol (untreated

FIGURE 2 | Histone acetylation and hyper-acetylation following treatment of gametocytes with TSA. (A,B) Presence of acetylated histones in gametocytes. Acetylated histones were detected in the different gametocyte stages (GC stage II-V) via immunolabeling using rabbit anti-H3K9ac (A) and anti-H4Kac4 (B) antibodies (green). Gametocytes were highlighted with mouse antibodies against the gametocyte marker Pfs230 (red). Nuclei were highlighted by Hoechst nuclear stain 33342 (blue). Bar, 5 µm. (C,D) Histone hyper-acetylation following gametocyte treatment with TSA. Protein lysates from immature gametocytes (imGC) (C) and mature gametocytes (mGC) (D) following treatment with TSA at IC<sup>90</sup> concentrations or with 0.5% vol. ethanol (untreated control) for 1 and 6 h at 37◦C were subjected to WB analysis using anti-H3K9ac and anti-H4Kac4 antibodies. Results shown (for A–D) are representative for three to six independent experiments. (E,F) Quantification of histone hyper-acetylation following gametocyte treatment with TSA. Lysates of imGC (E) and mGC (F) were subjected to immunoblotting as described above and (Continued)

#### FIGURE 2 | Continued

histone acetylation was quantified between TSA-treated and untreated samples by measuring the band intensities via Image J for three or more different experiments; the values were normalized with the band intensities of Pf39 used as loading control (set to 1). Since the anti-H4Kac4 detected two bands in both imGC and mGC indicating that the antibody could also detect other acetylated histones, we quantified the total histone acetylation levels from both bands. \*P < 0.05; \*\*P < 0.01; \*\*\*P < 0.001; ns, not significant, Student's t-test. According the manufacturer's material data sheet the rabbit anti-H4Kac4 antibody can cross-react with other acetylated histones like H2B.

FIGURE 3 | Deregulation of gene expression following treatment of gametocytes with TSA. Immature (imGC) and mature (mGC) gametocytes as well as gametocytes at 1 h post-activation (aGC) were treated with TSA at IC90 concentrations or with 0.5% vol. ethanol (untreated control) for 1 and 6 h, total RNA was isolated and cDNA synthesized to be employed in microarray assays. Genes with a relative expression levels greater than 2-fold for at least one of the two time-points combined with a consistent up- or down-regulation for both time-points were used for further analysis. (A,B) Bar charts showing total up- and down-regulated genes in imGC, mGC and aGC (A), and mean fold change of deregulated genes in imGC, mGC, and aGC (B) at 1 or 6 h following TSA-treatment. (C) Venn diagram showing the overlap among deregulated genes in imGC, mGC, and aGC after TSA-treatment. (D) Pie chart showing the detailed number of deregulated genes in the different gametocyte samples based on the predicted function following TSA-treatment.


**20**

TABLE 2 | Genes with more than 5-fold up-regulated

 transcription

 following

TSA-treatment.

control) for 1 h and total RNA was isolated. Complementary DNA was synthesized from each sample and purity was further assessed by diagnostic RT-PCR using stage specific markers. The tests confirmed the presence of pfccp2 (gametocytespecific) transcript in all the gametocyte samples while ama1 (asexual blood stage-specific) transcript was absent (Figure S3B), confirming that the gametocyte samples were devoid of any asexual blood stage contamination. A test for gDNA contamination in sample preparations lacking reverse transcriptase using pfccp2-specific primers was negative.

Subsequently, the transcript expression levels of the 32 selected genes were measured via real time RT-PCR in TSAtreated and untreated samples. Transcript expression was calculated by the 2−1Ct method (Livak and Schmittgen, 2001) in which the threshold cycle number (Ct) was normalized to the Ct of the endogenous control gene encoding P. falciparum seryl tRNA-ligase (PF3D7\_0717700) as reference gene. For TSA-treated immature gametocytes, we demonstrated a greater than 2-fold transcriptional up-regulation for 14 out of the 15 (93.3%) up-regulated genes as identified by microarray analysis (**Figure 4A**). Accordingly, 14 out of 17 (82.4%) of the identified up-regulated genes of the TSA-treated mature gametocyte samples had greater than 2-fold increased transcript levels compared to the untreated control (**Figure 4B**), which strongly validates the microarray results.

### Genes Transcriptionally Up-Regulated Following TSA-Treatment Associate with Acetylated Histones

ChIP-qPCR assays were employed to investigate a potential link between histone acetylation and the transcriptional up-regulation of selected genes following TSA-treatment of gametocytes. Five genes transcriptionally up-regulated in TSA-treated mature gametocytes were chosen, i.e., genes encoding a putative ring finger protein (henceforth termed pfrnf1, 3D7\_0314700), a WD40-domain protein (pfwdtc1, 3D7\_1428400), the exported protein PHISTc (3D7\_0219800) and two unknown proteins (PF3D7\_0620200, PF3D7\_0926600). For control a gene transcriptionally down-regulated in TSA-treated mature gametocytes was selected, i.e., pfnep1 (PF3D7\_0821500), a gene encoding the ribosomal RNA small subunit methyl-transferase NEP1 (**Figure 5A**). For further controls, the housekeeping genes coding for arginine tRNA-ligase (PF3D7\_1218600) and seryl tRNA-ligase (PF3D7\_0717700) were chosen. Additional controls included two genes that were previously shown to associate with chromatin bound to PfHP1, i.e., ap2-g (3D7\_1222600) and var gene upsB (PF3D7\_0426000) (Brancucci et al., 2014). For each gene, primers corresponding to the promotor and coding regions were generated (for primer locations, see **Figure 5A**). ChIP assays were performed using anti-H3K9ac and anti-H4Kac4 antibodies, which were previously used to precipitate plasmodial chromatin (Crowley et al., 2011; Gómez-Díaz et al., 2017). For negative control, an IgG antibody from non-immunized rabbit was used in the assays. For positive control, anti-PfHP1 antibody was used to precipitate ap2-g and var upsB (Flueck et al., 2009). The ChIP recovery rates of these

TSA. Transcript analysis for 15 up-regulated immature (imGC) (A) and 17 up-regulated mature (mGC) (B) gametocyte genes as identified by microarray via real-time RT-PCR. Transcript expression levels were calculated by the 2 <sup>−</sup>1Ct method; the threshold cycle number (Ct) was normalized with the Ct of the gene encoding seryl tRNA-ligase (PF3D7\_0717700) as reference. Genes were considered up-regulated when the fold changes between TSA-treated and untreated sample were greater than 2-fold. Results shown are representative for two to three independent experiments.

genes following immunoprecipitation were compared between chromatin generated from mature gametocytes treated with TSA at IC<sup>90</sup> concentrations for 6 h and from untreated mature gametocytes via qPCR.

Initially, the presence of PfHP1 in the nuclei of gametocytes was demonstrated via IFA and compared to PfHP1 localization in trophozoites and schizonts. As previously described, PfHP1 as a marker for heterochromatin particularly localized to the nucleus periphery in trophozoites and schizonts (Flueck et al., 2009; Pérez-Toledo et al., 2009). Similarly, PfHP1 was predominantly found at the nucleus periphery of immature and activated gametocytes, while in mature gametocytes PfHP1 was often found in concentrated foci within the nuclei (Figure S6). Immunolabeling with NMS served as negative control and did not result in any labeling (Figure S7).

The ChIP-qPCR analysis on precipitated acetylated histones using anti-H3K9ac antibody demonstrated a higher recovery of the five genes transcriptionally up-regulated in TSA-treated mature gametocytes (**Figure 5B**). The increased recovery rates

#### FIGURE 5 | Continued

anti-PfHP1antibody and IgG as controls. The immunoprecipitated material was analyzed by qPCR to confirm specific enrichment of selected genes including pfrnf1 (PF3D7\_0314700), pfwdtc1 (PF3D7\_1428400), phistc (PF3D7\_0219800), PF3D7\_0620200, and PF3D7\_0926600. As a down-regulated gene pfnep1 (PF3D7\_0821500) was analyzed. The genes encoding arginine tRNA-ligase a-t-l (PF3D7\_1218600) and seryl tRNA-ligase s-t-l (PF3D7\_0717700) as well as ap2-g (PF3D7\_1222600) and the var gene upsB (PF3D7\_0426000) were used as controls. Primers targeting either the coding regions (cod.) or promoter regions (prom.) were used for qPCR. The values represent the proportion of chromatin recovered from the input samples. Results shown are representative for two to three independent experiments.

were detected both when primers corresponding to the promotor regions or the coding regions were used. A slightly higher recovery was also seen for the two control tRNA-ligaseencoding control genes, when the gametocytes were treated with TSA. No increased recovery rate, on the other hand, was detected, when primers corresponding to the promotor and coding regions of pfnep1 were used in the qPCR analyses. Furthermore, the genes ap2-g and var upsB were precipitated with the anti-HP1 antibody (**Figure 5B**). The recovery rates for the var gene upsB were higher than the ones for ap2-g and neither of the recovery rates changed upon TSA-treatment of the gametocytes. Generally, similar higher recovery rates were observed for the TSA-dependent transcriptionally upregulated genes, when the anti-H4Kac4 antibody was used for precipitation. No higher recoveries of pfrnf1 and pfwdtc1 were achieved, though, when the promotor regions were amplified, indicating that these promotors might associate particularly with acetylated H3 (Figure S8). Furthermore, no higher recovery for PF3D7\_0620200 was detected. The overall high recovery rates observed when using anti-H4Kac4 antibody for precipitation might be due to the fact that this antibody also reacts with acetylated H2B (compare with **Figures 2C,D**). The combined data demonstrate an association of acetylated histones with the five selected genes that were transcriptionally up-regulated in TSA-treated gametocytes.

### Treatment of Mature Gametocytes with TSA Results in Increased PfRNF1 Synthesis

In a final set of experiments, we aimed to determine, if upregulation of transcript expression following TSA-treatment also corroborates with increased protein synthesis. We selected one gene, pfrnf1, whose transcript level was up-regulated in mature gametocytes following TSA-treatment, for further investigations. In-silico analyses disclose the ring finger-domain protein PfRNF1 as a putative E3-ligase (Ponts et al., 2008). A recombinant protein corresponding to the N-terminal region of PfRNF1 was bacterially expressed and used to generate antisera against PfRNF1 in mice. IFA were performed in the different asexual and sexual blood stages of P. falciparum. Asexual blood stage parasites were highlighted by MSP-1 labeling; gametocytes and activated gametocytes were highlighted by Pfs230 immunolabeling. The IFA revealed a prominent expression of PfRNF1 in immature gametocytes with peak expression in stage II gametocytes. Here, the protein localized to the gametocyte cytoplasm and nucleus (**Figure 6A**). PfRNF1 was also detected at low levels in mature gametocytes and in gametocytes 30 min post-activation as well as in schizonts.

Expression of PfRNF1, which has a calculated molecular weight of 135 kDa, was then investigated in immature and mature gametocytes via WB analysis. A predominant protein band was detected in both, immature and mature gametocyte lysates, when immunoblotted with anti-PfRNF1 antisera, which was migrating at a molecular weight of roughly 200 kDa. An additional protein band at ∼140 kDa was also observed (**Figure 6B**). No protein bands were present in lysates of the non-infected RBC control. For loading control, immunoblotting with anti-Pf39 antisera, directed against the endoplasmic reticulum-specific protein Pf39 (Templeton et al., 1997), was performed, and Pf39 was detected in all parasite lysates. We also investigated the subcellular localization of PfRNF1 in gametocytes via WB, using cytosolic and nuclear fractions of mixed gametocyte cultures. WB revealed the presence of PfRNF1 as a 200-kDa protein in both fractions (**Figure 6C**). Antibodies against PfActin1, which is predominantly present in the cytoplasm and to a minor level in the nuclei were used as a fraction control. The purity of the fractions was further confirmed by immunoblotting with anti-H4Kac4 antibody, which labeled histones in the nuclear fraction, but which did not result in any protein band in the cytoplasmic fraction (**Figure 6C**).

To validate the protein expression data, we generated a P. falciparum transfectant line, which expresses PfRNF1 C-terminally tagged with HA-Strep (Figure S1B). Successful integration was confirmed by diagnostic PCR (Figure S1C). Subsequent WB analyses, using rabbit anti-HA antibody, detected a protein band with an approximate molecular weight of 200 kDa in lysates of immature gametocytes from the transfectant line, but not from wild-type parasites used as a negative control (**Figure 6D**). Further, no protein band was detected in lysates of non-infected RBCs. The combined WB data indicate that PfRNF1 migrates at a higher molecular weight than expected, which might be caused by PTMs. When the PfRNF1-HA-Strepexpressing parasite line was used in IFA and immunolabeled with anti-HA antibody, a similar protein expression pattern was observed as has been described above (Figure S9).

Finally, we assessed whether treatment of mature gametocytes with TSA results in up-regulation of PfRNF1 on the protein level. In this regard, mature gametocytes, both of the wild-type strain and the PfRNF1-HA-Strep-expressing transfectant line, were treated with TSA at IC<sup>90</sup> concentration or 0.5% vol. ethanol (untreated control) for 24 h, lysates were generated and PfRNF1 was detected by WB analysis using the anti-PfRNF1 antisera. Immunoblotting revealed increased PfRNF1 levels in lysates of TSA-treated gametocytes compared to the untreated control for both wild-type and transfectant (**Figure 7A**, Figure S10A). Quantitative WB analysis showed a significant up-regulation of PfRNF1 following TSA-treatment as compared to the untreated control (**Figure 7B**, Figure S10B).

### DISCUSSION

Histone PTMs are emerging as major regulatory mechanisms thought to modulate gene expression in eukaryotes. While these mechanisms have been extensively studied during the erythrocytic replication of P. falciparum, little is known about this process during parasite human-to-mosquito transmission. In this study, we aimed to determine the role of histone acetylation and deacetylation in the control of gene expression in P. falciparum gametocytes during their development and transmission to the mosquito. To achieve our goal, we used the chemical loss-offunction approach using the HDAC inhibitor TSA (inhibitor of HDACs I and II) on the P. falciparum gametocyte-producing NF54 strain. Treatment with TSA results in a mean IC<sup>50</sup> value of 29 nM. The killing of the P. falciparum asexual blood stages by TSA with IC<sup>50</sup> values in nM ranges have already been reported before for the chloroquine-sensitive 3D7 (8 nM) and chloroquine resistant strain DD2 (11 nM) (Andrews et al., 2008).

In order to determine the effect of TSA on gametocyte development and transmission to the mosquito, we treated stage II gametocytes with the HDAC inhibitor and followed gametocyte development. We show that gametocyte development was strongly affected, and here stage II and III gametocytes appeared to be more vulnerable to TSA treatment then stage IV and V gametocytes. On the other hand, TSA only moderately effected macrogamete and zygote development. In a previous study, it was also reported that microgametes were only moderately affected following treatment with a variety of HDAC inhibitors (Trenholme et al., 2014). The fact that TSA has strong gametocytocidal activities, but only exerts minor effects on gametes and zygotes indicates that histone acetylation-mediated gene regulation is important during gametocytogenesis, but not during the early phase of midgut-stage development. This is in accord with findings in P. falciparum and P. berghei that transcript required for the midgut stage formation is synthesized and stored in female gametocytes, where it is translationally repressed by binding to regulatory ribonucleoprotein complexes, like the RNA helicase DOZI (development of zygote inhibited) and the Sm-like factor CITH (homolog of worm CAR-I and fly Trailer Hitch), as identified in P. berghei, or the Pumilio/FBF (Puf) family RNA-binding protein Puf2, as was shown for P. falciparum (Mair et al., 2006, 2010; Miao et al., 2013; Guerreiro et al., 2014). A recent analysis integrating transcriptome and proteome data revealed 512 highly expressed transcripts in P. falciparum female gametocytes without corresponding protein expression, indicating large scale translational repression (Lasonder et al., 2016). Repression of the stored transcript is released and translation is initiated, once the gametocytes become activated in the mosquito midgut.

To confirm histone acetylation in gametocytes, we used two commercially available histone acetylation antibodies (anti-H3K9ac and anti-H4Kac4) to detect the acetylated histones in the nuclei of different gametocyte stages. By means of these

FIGURE 6 | Characterization of PfRNF1. (A) Localization of PfRNF1 in the P. falciparum asexual blood and gametocyte stages. Mouse anti-PfRNF1 antisera was used to immunolabel fixed samples of trophozoites, schizonts and mature gametocytes (GC) of stages II to V as well as of activated gametocytes (aGC) at 30 min post-activation (green). Schizonts were visualized by labeling with rabbit anti-MSP-1 antibody and gametocytes were visualized by rabbit anti-Pfs230 antisera (red); nuclei were highlighted by Hoechst nuclear stain 33342 (blue). Bar, 5 µm. Results shown are representative for four independent experiments. (B) Expression of PfRNF1 in gametocyte lysates of immature (imGC) and mature (mGC) gametocytes were immunoblotted with anti-PfRNF1 antibody and detected two protein bands of approximate molecular weights of 200 kDa (arrow) and 140 kDa. Lysates of non-infected red blood cells (niRBCs) were used for negative control. Immunoblotting with mouse anti-Pf39 antisera served as loading control. (C) Sub-cellular localization of PfRNF1. Cytosolic and nuclear fractions of enriched immature gametocytes were subjected to WB using anti-PfRNF1 antisera and detected a 200-kDa band (arrow) in both fractions. Mouse antibodies against PfActinI (41 kDa) and rabbit antibodies against H4Kac4 detecting acetylated histone H4 (∼11 kDa) were used as fraction controls. (D) Detection of PfRNF1-HA-Strep in gametocytes of parasite line PfRNF1-HA-Strep. Lysates of PfRNF1-HA-Strep immature gametocytes were immunoblotted with rabbit anti-HA antibody and detected a protein band of 200 kDa (arrow). Lysates of non-infected red blood cells (niRBCs) as well as wild-type (Wt) gametocytes were used as negative control. Results shown (for A–D) are representative for three to four independent experiments.

antibodies, we demonstrated that gametocyte treatment with TSA results in histone hyper-acetylation thereby leading to transcriptional activation. Histone hyper-acetylation following treatment of the malaria parasite with HDAC inhibitors have been earlier reported for P. falciparum blood stages (Andrews et al., 2008; Trenholme et al., 2014) and the asexual blood stages of P. knowlesi (Chua et al., 2017).

As a next step we aimed to investigate any deregulation of gene expression in immature, mature and activated gametocytes caused by TSA-treatment via microarray- and real-time RT-PCRbased analyses. Comparative transcriptomics between untreated and inhibitor-treated gametocytes identified 453 genes, which were more than 2-fold deregulated after 1 or 6 h following TSAtreatment. Among these 303 were more than 2-fold up-regulated, while 150 genes were down-regulated. Up-regulation of gene expression can be explained by increased euchromatin formation due to the loose contact between the negatively charged DNA and the acetylated histones, which hence counteracts gene silencing. Down-regulation of gene expression following TSA-treatment, on the other hand, which occurs in less than half of the identified genes, might be due to indirect effects. As mentioned earlier, PfHP1 was reported to bind to H3K9me3 to maintain the heterochromatin state, resulting in ap2-g repression (Brancucci et al., 2014), a process also involving PfHda2. It is postulated that the removal of acetyl groups by PfHda2 promotes histone methylation leading to PfHP1 binding (Coleman et al., 2014). Other studies have also reported the down-regulation of gene expression following treatment with HDAC inhibitors (Glaser et al., 2003; Chaal et al., 2010; Andrews et al., 2012b). Subsequent ChIP-qPCR analyses confirmed the general association of genes up-regulated in gametocytes following TSA-treatment with acetylated histones, particularly acetylated H3 and H4. The observed higher recovery rates achieved, when anti-H4Kac4 was used in the ChIP assays may be due to the facts that the antibody recognizes several lysine acetylation sites in H4 and that it additionally exhibits a minor binding to acetylated H2B.

An interesting finding in this study is the fact that the impairment of gene silencing during gametocytogenesis re-activates genes known to be crucial for blood stage replication. This was particularly observed for genes assigned to antigenic variation and cytoadherence and to RBC invasion by merozoites, or genes encoding exported proteins, which are particularly characteristic for the intraerythrocytic trophozoites. Interestingly, genes involved in antigenic variation and cytoadherence included ones coding for RIFINs and SURFINs, while only one out of 60 known var genes was identified. This might be explained by the fact that var gene activation and silencing is mostly accredited to the NAD-dependent histone deacetylases PfSir2A and PfSir2B (Duraisingh et al., 2005; Freitas-Junior et al., 2005; Tonkin et al., 2009) which are not affected following TSA-treatment. Var gene regulation is also linked to histone methylation, like H3K4me3-mediated var gene activation and H3K9me3-mediated var gene silencing (Chookajorn et al., 2007; Lopez-Rubio et al., 2007, 2009; Salcedo-Amaya et al., 2009), which further substantiate the reason why var genes were hardly identified. RIFINs are encoded by about 135 rif genes and comprise the largest family of antigenically variable molecules in P. falciparum (reviewed in Kirkman and Deitsch, 2014). H3K9 acetylation has been shown to control the expression of rif genes, which probably accounts for the reason why RIFINs were affected (Cabral et al., 2012). SURFINs are encoded by 10 surface-associated interspersed (surf) genes, which are located close or within the subtelomeric region of chromosomes. Little is known whether they are epigenetically regulated, but it is likely that they are regulated by histone PTMs (Mphande et al., 2008). Not identified by the microarray analyses were any of the genes coding for the 15 variant antigen-encoding mc-2tm genes. Furthermore, only one out of 35 STEVOR-encoding genes was up-regulated in gametocytes following TSA-treatment, suggesting that these gene families are not primarily affected in their expression levels by histone acetylation.

The majority of exported proteins regulated in their synthesis by histone acetylation include members of the PHIST (Plasmodium helical interspersed subtelomeric) family. The PHIST family comprises 89 proteins, the most of which have yet unknown functions. The PHIST protein family is characterized by a conserved domain of ∼150 amino acids predicted to form four consecutive alpha helices and have been shown to be differentially expressed during the Plasmodium life-cycle (Warncke et al., 2016). The differential expression of these proteins may suggest an epigenetic mechanism. Invasion-related genes regulated in their expression by histone acetylation include previously identified surface proteins of merozoites, e.g., members of the MSP (merozoite surface protein) and the RON (Rhoptry neck) families (Proellocks et al., 2010; Beeson et al., 2016; Lin et al., 2016). Noteworthy, all of these identified gene families, i.e., rif and stevor and the ones encoding the PHIST family as well as the merozoite invasion-related families were reported to associate with heterochromatin markers (Flueck et al., 2009; Lopez-Rubio et al., 2009; Salcedo-Amaya et al., 2009) thereby maintaining default silencing of the majority of redundant members of multi-gene families (reviewed in Duffy et al., 2012).

While the above mentioned genes appear to be silenced by deacetylations of the associated histones after the parasites have entered the sexual pathways, other genes were apparently waiting to be activated by histone acetylation, once the parasite has been transmitted to the mosquito vector. Among others, these genes encode proteins important for midgut extravasation by the ookinete, like the secreted ookinete protein 25 (PSOP25), the secreted ookinete adhesive protein (SOAP), the von Willebrand factor A-domain related protein (WARP), the circumsporozoite and thrombospondin-related adhesion protein-related protein (CTRP), the cell traversal protein for ookinetes and sporozoites (CelTOS), or chitinase (CHT1) (reviewed in Pradel, 2007; Bennink et al., 2016).

To further justify our data, we selected one gene, pfrnf1 (PF3D7\_0314700), whose transcript was up-regulated following TSA-treatment of mature gametocytes, for further characterization. PfRNF1 possesses a C-terminal RING finger domain, which shows a high homology with the human E3 ubiquitin-protein ligase Praja-1. PfRNF1 has been annotated as a potential E3 ligase in P. falciparum as a component of the ubiquitin-mediated pathway (Ponts et al., 2008). TSA-treatment caused increased transcript and protein levels of PfRNF1 in mature gametocytes. Based on our data we postulate that under normal conditions deacetylation of histones H3 and H4 down-regulates PfRNF1 expression in mature gametocytes. The regulation of expression by histone deacetylases was recently also reported for the human RING finger domain protein RNF148

### REFERENCES


(Liu et al., 2013). We therefore suspect that PfRNF1 is a potential HDAC-regulated E3 ligase involved in ubiquitin-mediated pathways during gametocyte development, which might be important for regulatory processes during human-to-mosquito transmission of Plasmodium.

Our combined data highlight the role of histone acetylation in the control of gene expression during gametocyte development and transmission from the human to the mosquito, which may be exploited in malaria transmission-blocking strategies.

### AUTHOR CONTRIBUTIONS

Conceived and designed the experiments: CN, MK, ML, TV, and GP. Performed the experiments: CN, MK, OP, LO, and MF. Analyzed the data: CN, MK, OP, LO, AR, and GP. Wrote the manuscript: CN, MK, and GP. All authors read and approved the manuscript.

### FUNDING

We thank Alex Maier (ANU Canberra) for kindly providing vector pHAST. The work was funded by grant PR905/7-1 of the Deutsche Forschungsgemeinschaft (to GP) and further supported by grant 01DR14019 of the German Ministry for Education and Research (to GP) and an ERS Start-Up grant of RWTH Aachen University (to CN). MK received a short-term fellowship by the Deutscher Akademischer Austauschdienst; CN received a Theodore von Kármán fellowship from the RWTH Aachen University. GP is recipient of a Heisenberg professorship of the Deutsche Forschungsgemeinschaft.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fcimb. 2017.00320/full#supplementary-material


of the histone variant H2A.Z at the promoter. PLoS Pathog. 7:e1001292. doi: 10.1371/journal.ppat.1001292


genes and antigenic variation in Plasmodium falciparum. PLoS Biol. 7:e1000084. doi: 10.1371/journal.pbio.1000084


cinnamate or NSAID components. Bioorg. Med. Chem. Lett. 20, 7080–7084. doi: 10.1016/j.bmcl.2010.09.096


**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 Ngwa, Kiesow, Papst, Orchard, Filarsky, Rosinski, Voss, Llinás and Pradel. 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.

# Interaction between *Plasmodium* Glycosylphosphatidylinositol and the Host Protein Moesin Has No Implication in Malaria Pathology

Josefine Dunst 1, 2 \*, Nahid Azzouz 1, 3, Xinyu Liu<sup>4</sup> , Sachiko Tsukita<sup>5</sup> , Peter H. Seeberger 1, 3 and Faustin Kamena1, 2 \*

*1 Institute of Chemistry and Biochemistry, Free University Berlin, Berlin, Germany, <sup>2</sup> Parasitology Unit, Max Planck Institute for Infection Biology, Berlin, Germany, <sup>3</sup> Department of Biomolecular Systems, Max Planck Institute for Colloids and Interfaces, Potsdam, Germany, <sup>4</sup> Department of Chemistry, University of Pittsburgh, Pittsburgh, PA, USA, <sup>5</sup> Department of Frontier Biosciences, Graduate School of Frontier Biosciences, Osaka University, Osaka, Japan*

Glycosylphosphatidylinositol (GPI) anchor of *Plasmodium falciparum* origin is considered an important toxin leading to severe malaria pathology through stimulation of pro-inflammatory responses from innate immune cells. Even though the GPI-induced immune response is widely described to be mediated by pattern recognition receptors such as TLR2 and TLR4, previous studies have revealed that these two receptors are dispensable for the development of severe malaria pathology. Therefore, this study aimed at the identification of potential alternative *Plasmodium* GPI receptors. Herein, we have identified the host protein moesin as an interaction partner of *Plasmodium* GPI *in vitro*. Given previous reports indicating the relevance of moesin especially in the LPS-mediated induction of pro-inflammatory responses, we have conducted a series of *in vitro* and *in vivo* experiments to address the physiological relevance of the moesin-*Plasmodium* GPI interaction in the context of malaria pathology. We report here that although moesin and *Plasmodium* GPI interact *in vitro,* moesin is not critically involved in processes leading to *Plasmodium*-induced pro-inflammatory immune responses or malaria-associated cerebral pathology.

#### *Edited by:*

*Anton Aebischer, Robert Koch-Institute, Germany*

#### *Reviewed by:*

*Nyssa Drinkwater, Monash University, Australia Sabrina Absalon, Boston Children's Hospital, USA*

#### *\*Correspondence:*

*Josefine Dunst dunstj@zedat.fu-berlin.de Faustin Kamena kamena@zedat.fu-berlin.de*

*Received: 09 February 2017 Accepted: 27 April 2017 Published: 16 May 2017*

#### *Citation:*

*Dunst J, Azzouz N, Liu X, Tsukita S, Seeberger PH and Kamena F (2017) Interaction between Plasmodium Glycosylphosphatidylinositol and the Host Protein Moesin Has No Implication in Malaria Pathology. Front. Cell. Infect. Microbiol. 7:183. doi: 10.3389/fcimb.2017.00183* Keywords: GPI, ERM, moesin, *Plasmodium*, cerebral malaria

## INTRODUCTION

Malaria still causes a devastatingly high number of deaths and new infections each year and is thereby a major contributor to the global burden of infectious diseases (WHO, 2016). This disease is caused by human host-adapted Plasmodium species and transmitted by the bite of an infective Anopheles mosquito. During Plasmodium infection, an immune response is mounted by the host in order to limit parasite expansion and mediate clearance. Consequently, blood stage infection is accompanied by a systemic pro-inflammatory immune response resulting in classical symptoms of mild malaria such as fever (Stevenson and Riley, 2004). However, some individuals progress to a severe course of malaria, partly owing to an imbalance in the pro- and anti-inflammatory immune response (Langhorne et al., 2008), resulting in malaria-associated mortality which is largely attributed to P. falciparum infections (WHO, 2016). One of the major complications of severe malaria is cerebral malaria (CM) which manifests with retinal abnormalities (Storm and Craig, 2014) as well as impaired consciousness or coma (Cunnington et al., 2013). The symptoms of CM are attributable to sequestration of infected erythrocytes and inflammatory leukocyte subsets, endothelial dysfunction, and inflammation (Storm and Craig, 2014), and these processes are mutually dependent and have synergetic effects (Cunnington et al., 2013). However, the precise molecular mechanisms underlying CM are not yet fully understood.

The induction of innate pro-inflammatory cytokine responses is mediated by germline-encoded pattern-recognition receptors, such as toll-like receptors (TLR), which recognize conserved microbial structures, i.e., pathogen-associated molecular patterns (PAMP) (Kawai and Akira, 2011). Among the malaria PAMP, glycosylphosphatidylinositols (GPI) are considered the main pathogenicity factor (Gowda, 2007). While GPI structure is conserved among Plasmodium species, human and Plasmodium GPI differ considerably (Boutlis et al., 2005). GPI serve as membrane anchors for certain cell surface proteins such as circumsporozoite protein and merozoite surface protein 1, and are also abundantly present free of protein attachment in membranes of pathogenic protozoa (Gowda, 2007; Gazzinelli et al., 2014). P. falciparum GPI have been found to induce the production of nitric oxide, tumor necrosis factor (TNF), and interleukin 1β (IL-1β) in murine macrophages in vitro (Schofield and Hackett, 1993; Tachado et al., 1996) and a synthetic malarial GPI glycan was demonstrated to be immunogenic in vivo (Schofield et al., 2002). Together, these findings point toward a role for Plasmodium GPI in malaria pathogenesis. Notably, Plasmodium GPI were described to be primarily recognized by TLR2 or heterodimers of TLR2/1 and TLR2/6 (Krishnegowda et al., 2005), yet TLR-deficiency did not protect mice from experimental cerebral malaria (ECM) (Togbe et al., 2007; Lepenies et al., 2008), indicating that TLR-mediated pro-inflammatory immune responses are not critical in the development of ECM. Since elucidating molecular mechanisms leading to malaria pathology might allow specific modulation of innate immune activation to prevent detrimental immune responses, this study was designed to identify potential Plasmodium GPI receptors. Using synthetic GPI affinity chromatography, we have identified the host protein moesin as an interaction partner of P. falciparum GPI and further addressed the functional relevance of this interaction in the development of malaria pathology. Moesin is a member of the ezrin-radixin-moesin (ERM) family of intracellular proteins which link actin filaments to transmembrane proteins (Louvet-Vallee, 2000) and interact with proteins involved in key signaling events, such as phosphatidylinositide 3-kinase, protein kinase A, or Rho-specific GDP dissociation inhibitors (Ivetic and Ridley, 2004; Niggli and Rossy, 2008; Ponuwei, 2016). Additionally, moesin cell surface translocation has been described upon lipopolysaccharide (LPS) stimulation in vitro (Iontcheva et al., 2004; Takamatsu et al., 2009), pointing toward a role for moesin in PAMP recognition. Moreover, we reasoned that moesin may play a role in the immune response to Plasmodium infection via its ability to interact with Plasmodium GPI as well as in malaria pathology due to its ability to modulate immunological synapse and endothelial paracellular gap formation (Itoh et al., 2002; Koss et al., 2006; Parameswaran and Gupta, 2013). Therefore, the capability of moesin to translocate to the cell surface upon Plasmodium GPI stimulation as well as the impact of moesin-deficiency on malaria PAMPmediated cytokine induction and phagocytosis of P. berghei was analyzed in vitro. Additionally, P. berghei ANKA-infected moesin-deficient mice were used as a model to study the role of moesin in the host immune response to Plasmodium and in the development of cerebral pathology in vivo. We report here that despite the interaction between Pf GPI and moesin in vitro, moesin does not translocate to the cell surface in response to malaria PAMP in human and murine macrophages. Moreover, moesin-deficiency did not impair Plasmodiuminduced cytokine responses in vitro and in vivo and did not protect mice infected with P. berghei ANKA from development of ECM.

### MATERIALS AND METHODS

### GPI Affinity Chromatography and Mass Spectrometry

GPI glycans were synthesized with a terminal sulfhydrylcontaining linker (Kwon et al., 2005; Liu et al., 2005) to be covalently immobilized on SulfoLink <sup>R</sup> coupling gel (Pierce, Rockford, IL) according to the manufacturer's instructions. Using a syringe, the column was equilibrated by washing with coupling buffer (50 mM Tris, 5 mM EDTA, pH 8.5). Mouse macrophage cell-line RAW264.7 plasma membrane fraction was prepared as previously reported (Smart et al., 1995). Plasma membrane was solubilized in coupling buffer containing 0.5% Triton X-100. Samples were then loaded in the coupling buffer and the column was incubated for 1 h at 4◦C. Non-bound and excess proteins were removed by washing the column with 3 column volumes of coupling buffer. Elution of bound proteins was carried out using 3 column volumes of coupling buffer containing 1 M mannose and 0.1% Triton X-100. Eluted protein extracts were subjected to 12.5% SDS-PAGE under non-reducing condition. After Coomassie staining, protein bands were excised, destained, and reduced prior to tryptic digestion and peptide mass fingerprinting. MALDI mass spectra were generated using a Voyager DE-STR MALDI-TOF MS system (PerSeptive Biosystems) with delayed extraction in the reflectron mode. Proteins were identified by comparison of peak lists generated from the Data Explorer application (PerSeptive Biosystems) against NCBInr (no redundant) and Swiss-prot databases using the Protein-Prospector V3.4.1 software MS-Fit (http://www.prospector.ucsf. edu).

**Abbreviations:** BMDC, bone-marrow derived dendritic cells; BMDM, bonemarrow derived macrophages; CM, cerebral malaria; DC, dendritic cells; ECM, experimental cerebral malaria, ERM, ezrin-radixin-moesin; FCS, fetal calf serum; FMO, fluorescence-minus-one; GPI, glycosylphosphatidylinositol; IFN, interferon; IL, interleukin; LPS, lipopolysaccharide; PAMP, pathogen-associated molecular pattern, PbS, Plasmodium berghei schizonts; PbSE, Plasmodium berghei schizont extract; Pf SE, Plasmodium falciparum schizont extract; PMA, phorbol 12-myristate 13-acetate; TLR, toll-like receptor; TNF, tumor necrosis factor

### Moesin-GST Expression and Purification

A pGEX-4T-3 vector (Addgene) carrying Moesin-GST as N-terminal fusion protein was transformed into a bl21de3plysS (Promega, USA) E. coli strain. A single colony was picked to inoculate an overnight culture in 100 ml of LB medium supplemented with 100 µg/ml ampicillin at 37◦C under shaking at 200 rpm. On the following day the overnight culture was diluted 1:20 in LB supplemented with 100 µg/ml ampicillin and incubated at 37◦C at 300 rpm until an OD<sup>600</sup> of 0.5 was reached. IPTG (Sigma-Aldrich) was added to the final concentration of 1 mM and the culture was incubated for additional 4 h. The bacteria pellet was harvested by centrifugation at 4,000 × g at 4 ◦C for 20 min.

For the purification, bacteria pellet was resuspended in lysis buffer (50 mM Tris, 50 mM NaCl, 5 mM EDTA, 1 µg/ml leupeptin, 1 µg/ml pepstatin, 0.15 mM PMSF, 1 mM DFP, 1 mM 2-ME, pH 8.0) and lysis was completed by sonication. Cell lysate was centrifuged at 48,000 × g for 20 min at 4◦C and the clear supernatant loaded on a GSH-agarose affinity column (Thermo Fisher). The column was washed three times to remove unbound material and moesin-GST was eluted with elution buffer (10 mM GSH in 50 mM Tris, 10 mM reduced glutathione, pH 8.0). The eluate was dialyzed against PBS and aliquots were kept at −80◦C until use.

For surface plasmon resonance measurements, moesin was released from the moesin-GST fusion protein by thrombin digestion using immobilized thrombin, agarose (Sigma-Aldrich) according to the manufacturer's instruction (Supplementary Figure 1). Briefly, purified moesin was incubated with washed thrombin agarose beads in batch for 2 h at 37◦C and the cleaved protein was recovered from the supernatant after centrifugation. Cleaved GST was removed by incubating the product from the thrombin cleavage with GSH-agarose beads in batch for 2 h at RT. Moesin without the GST tag was recovered in the supernatant after centrifugation of the beads at 10,000 × g for 10 min. Purified moesin was dialyzed against PBS and aliquots were frozen until use.

### Microarray Binding Assays

GPI microarrays were constructed as previously described (Ratner et al., 2004; Kamena et al., 2008) and covered with a FlexWell-64 (GRACE BIO-LABS, Bend, OR) layer to form a multi-well plate. Wells were then blocked with 5% milk powder in PBS for 1 h at RT followed by three washes with PBS containing 0.05% Tween-20. The wells were incubated with purified moesin-GST in PBS containing 0.05% Tween-20 and 0.5% BSA for 2 h at RT or with FITC-labeled concanavalin A (FITC-ConA) in ConA binding buffer (20 mM Tris; 500 mM NaCl; 1 mM CaCl2; 1 mM MgCl2; pH 7.4). After washing, the slides were incubated with rabbit polyclonal anti-GST antibodies for 1 h at RT. After extensive washing the slides were incubated 1 h at RT with ALEXA-Fluor <sup>R</sup> 594-labeled anti-rabbit secondary antibody (Invitrogen, Eugene. OR) at 1:1,000 in PBS containing 0.5% BSA and 0.05% Tween-20. The slides were then washed and fluorescence was revealed using an Affymetrix 427 laser scanner (MWG Biotech, Huntsville, AL).

### Surface Plasmon Resonance

GPI structure VI (**Figure 1B**) having a free thiol group was covalently attached to a gold surface CM5 chip using a Biacore T100 (GE Healthcare, Uppsala, Sweden). GPI glycan was immobilized on gold chips according to the manufacturer's protocol. Briefly, the carboxymethylated dextran matrix (CM5 chip) was activated at a flow rate of 10 µL/min using an 8 min injection pulse of an aqueous solution containing N-hydroxysuccinimide (NHS, 0.05 M) and Nethyl-N'-(dimethylaminopropyl) carbodiimide (EDC, 0.2 M). The surface was further activated with a solution of 2-(2 pyridinyldithio) ethanolamine (PDEA, 80 mM in 0.1 M sodium borate; pH 8.5) at the flow rate of 10 µL/min using a 10 min pulse. Next, a solution of synthetic GPI (50 µg/ml) containing 1 mM hexadecyltrimethylammonium chloride was flowed over the activated surface for 10 min at 4 µL/min. Remaining reactive groups on the surface were quenched by injection of a cysteine/NaCl solution (50 mM cysteine and 1 M NaCl in 0.1 M sodium acetate; pH 4.3) for 7.5 min at 10 µL/min. The reference flow cell was activated in parallel and ethanolamine was covalently attached. For K<sup>D</sup> determination between immobilized GPI and moesin, HBS-EP buffer (10 mM HEPES, pH 7.4, 150 mM NaCl, 3 mM EDTA, 0.005% v/v surfactant P20) was used as running buffer. Various concentrations of moesin (10, 2, 37.5, 75, 150, and 300 nM) were injected into the flow cell for 10 min each at 20 µL/min at 25◦C. After each sample, running buffer flowed over the sensor surface for 10 min to allow dissociation. The chip surface was regenerated for the next sample by injection of the regenerating solution (0.1% SDS, 0.085% H3PO4, 1 M NaCl and 0.1% HCl) for 1 min at 80 µL/min flow rate. The responses were calculated as the difference in response unit (RU) between analyte and reference flow cell and monitored as function of time (sensogram). Data processing and kinetic analysis was performed using the BIAevaluation software for T100 (Version 1.1.1) and graphs were plotted using Origin 8.0 (OriginLab, Northampton). Double referenced association and dissociation phase data were globally fitted to a simple 1:1 interaction model (A + B = AB).

To calculate the KD, the signal from the reference flow cell containing ethanolamine was subtracted from each value to correct for the contribution of non-specific interactions and systematic errors.

### THP-1 Cells

The human monocytic leukemia cell line THP-1 (ATCC: TIB-202) was a kind gift from Dr. Pedro Moura-Alves (Max Planck Institute for Infection Biology, Berlin, Germany) and was cultured in RPMI 1640 (Gibco, Germany) supplemented with 10% fetal calf serum (FCS; Gibco), 2 mM L-glutamine (Gibco), 100 mM HEPES (Gibco), 1 mM sodium pyruvate (Sigma-Aldrich, Germany), 1% non-essential amino-acids (NEAA) (Gibco) and 55 µM β-mercaptoethanol (Sigma-Aldrich). For differentiation of THP-1 cells into macrophage-like cells, monocytic THP-1 cells were plated at a density of 0.6 × 10 e6 cells/ml in culture medium supplemented with 50–200 nM phorbol 12-myristate 13-acetate (PMA, Sigma-Aldrich) for 72 h (Moura-Alves et al., 2014). Upon transition of THP-1 cells into adherent growth, cells were washed with PBS repeatedly and

#### FIGURE 1 | Continued

spectrometry moesin was confirmed by western blot using anti-moesin antibody (2). (B) The affinity column was generated using the full length glycan moiety of the GPI anchor. (C,D) Moesin interacts with GPI on microarray. Seven GPI glycan fragments and a polymannose control structure (depicted in D) were immobilized covalently as quadruplicate on glass slides to generate a microarray. (C) The slides were incubated with either FITC-ConA or GST-moesin. After incubation bound proteins were revealed by a fluorescent scanner. (E) Surface plasmon resonance measurement of moesin interaction with GPI glycan. Various concentrations of recombinantly expressed moesin (10, 20, 37.5, 75, 150, and 300 nM) were flown through the GPI-coupled gold chip and sensograms recorded and analyzed using the BIAevaluation software (BIAcore Life sciences) and the generated data were exported and graphs plotted using Origin 8.0 (OriginLab, Northampton) to obtain the K*D* value. Double referenced association and dissociation phase data were globally fitted to a simple 1:1 interaction model (*A* + *B* = *AB*). The GPI glycan immobilized on the sensor chip was the same as that used for the affinity chromatography (B).

rested in culture medium for 24 to 72 h. Successful differentiation was assessed in terms of expression of the cell surface marker CD11b by flow cytometry, since CD11b is not expressed on monocytic THP-1 cells (Schwende et al., 1996).

### Bone Marrow-Derived Cells

Bone marrow-derived cells were isolated from C57BL/6 (Charles River Laboratories) or WT and moesin-deficient mice from the Max Planck Institute for Infection Biology breeding facility, according to Gonçalves and Mosser (2015). For the generation of bone marrow-derived macrophages (BMDM), bone marrowderived cells were cultured in IMDM incomplete (IMDM (Gibco) containing 10% FCS and 2 mM L-glutamine) supplemented with 30% L-929 supernatant and 5% horse serum. Macrophage colony-stimulating factor (M-CSF)-expressing L-929 cells were kindly provided by Soo-Kyung Peuschel (Max Planck Institute for Infection Biology, Berlin, Germany). Medium was added at day 3 after seeding and a third of the volume was replaced at days 6 and 8 of culture. At day 10, differentiation was assessed by flow cytometry in terms of expression of macrophage surface markers CD11b and F4/80 as well as intracellular CD68 (Gonçalves and Mosser, 2015). Cells were re-plated for experiments or preserved in frozen stocks for future use.

For the generation of bone marrow-derived dendritic cells (BMDC), bone marrow-derived cells were further processed according to protocols adapted from Lutz et al. (1999), Brasel et al. (2000), Wells et al. (2005) and Madaan et al. (2014). Briefly, bone marrow-derived cells were resuspended in ACK lysis buffer (pH 7.2) and incubated for 2 min at 37◦C for erythrocyte lysis. After washing, cells were cultured in RPMI 1640 supplemented with 10% FCS, 1% NEAA, 2 mM L-glutamine, 100 U/ml penicillin/streptomycin (Gibco), 50 µM β-mercaptoethanol, and 20 ng/ml granulocyte-macrophage colony-stimulating factor (GM-CSF; Miltenyi Biotec, Germany) as well as 20 ng/ml IL-4 (Miltenyi Biotec). After 3 days of culture, fresh medium was added to the cells, and half of the medium was replaced with fresh medium at days 6 and 8. At day 10, loosely adherent cells were collected and re-plated in culture medium without IL-4 and 10 ng/ml GM-CSF for 24 h. Optionally, BMDC were LPS-primed (100 ng/ml) for 24 h to mature. BMDC differentiation was assessed by flow cytometry in terms of expression of DC surface markers CD11b, CD11c, and MHCII, as well as F4/80. Upon successful differentiation, BMDC were re-plated for stimulation experiments.

### Cell Stimulation

Cells were stimulated with 10 ng/ml LPS (LPS from E. coli 0111:B4, Sigma-Aldrich), 10 ng/ml TNF (Miltenyi Biotec, Germany), P. falciparum schizont extract diluted 1:100 in culture medium, P. berghei schizont extract diluted 1:1,000 in culture medium, or P. berghei schizonts added at a ratio of 1:10, i.e., 10 schizonts per cell. P. falciparum schizonts were obtained through a percoll gradient centrifugation of a mixed-stages culture as previously described (Rivadeneira et al., 1983). Purified schizonts were washed in PBS and subsequently sonicated to produce schizont extract. P. berghei schizont extract was generated by repeated freeze-thaw cycles and adjusted to contain the extract of 4 × 10 e6 schizonts/µl in PBS.

### Schizont Culture

For schizont enrichment, blood of P. berghei-infected mice was cultured in RPMI 1640 supplemented with 20% FCS (Gibco) and 15 µg/ml gentamycin (Gibco) with mild shaking for 20–24 h at <sup>37</sup>◦C and 80% humidity under 5% O<sup>2</sup> and 5% CO<sup>2</sup> (Matz et al., 2015). Schizont development was assessed by Giemsa-stained thin blood smear and schizonts were purified by density gradient centrifugation using 60% Percoll (GE Healthcare, UK) in PBS. Schizonts were collected from the interphase, washed repeatedly and further processed depending on experimental end-point.

### Transcript Quantification

RNA was isolated with TRIzol reagent (Ambion, Germany), glycogen (Ambion) and ammoniumacetate (Ambion) according to the manufacturer's instructions. Complementary DNA (cDNA) was synthesized from 1 µg total RNA per sample by reverse transcription using the RETROscript reverse transcription kit (Ambion) according to the manufacturer's instructions. Transcript abundance was determined by qPCR with reactions carried out on a StepOne Plus (Applied Biosystems, Germany) with Power SYBR green master mix (Applied Biosystems) and Quantitect primer assays (Qiagen, Germany) for the following transcripts: Ppib (Mm\_Ppib\_1\_SG), Tnf (Mm\_Tnf\_1\_SG), Il1b (Mm\_Il1b\_2\_SG), Il12a (Mm\_Il12a\_1\_SG), Ifng (Mm\_Ifng\_1\_SG), as well as with primers Hs\_RPL13A\_fwd 5′ -CCTGGAGGAGAAGAGGAAA GAGA-3′ , Hs\_RPL13A\_rev 5′ -TTGAGGACCTCTGTGTAT TTGTCAA-3′ (Rizopoulos et al., 2016), Mm\_Gapdh\_fwd 5′ - TGAGGCCGGTGCTGAGTATGTCG-3′ , Mm\_Gapdh\_rev 5 ′ -CCACAGTCTTCTGGGTGGCAGTG-3′ (Sato et al., 2014) (synthesized by Eurofins Genomics, Germany). qPCR was performed in technical triplicates with the following cycling conditions: 94◦C for 15 min and 40 cycles of 94◦C for 15 s, 60◦C for 60 s. For determining phagosomal degradation, P. berghei 18 s RNA transcripts were amplified as specified in Friesen et al. (2010). Melt curve analysis was included in each run to verify the specificity of each reaction. Relative transcript abundance and fold change were determined using the comparative threshold cycle (CT) method (Schmittgen and Livak, 2008), while RPL13A, Gapdh, or Ppib served as internal controls.

### Western Blot

For the detection of proteins using Western blot, samples were prepared by resuspension in sample buffer (Laemmli 2x concentrate, Sigma-Aldrich), separated by SDS-PAGE, and transferred onto Amersham Hybond P 0.45 PVDF membranes (GE Healthcare). Membranes were incubated with anti-MSN antibody (clone 38/87, Sigma-Aldrich; or clone EP1863Y, Abcam, UK) or anti-GAPDH antibody (clone 71.1, Sigma-Aldrich) as a loading control. Antibodies were detected using corresponding anti-mouse or anti-rabbit antibody coupled to HRP (Jackson Immunoresearch, UK) and ECL Western blotting substrate (Pierce, Germany). PageRuler pre-stained protein ladder (Fermentas) was used to determine the molecular weight of separated proteins.

### Quantification of Cytokines and Chemokines

THP-1, BMDM, and BMDC supernatants as well as serum of P. berghei ANKA Bergreen-infected mice were assayed for cytokines by human or mouse TNF DuoSet ELISA (R&D systems, USA) in combination with the DuoSet ELISA ancillary reagent kit (R&D systems) or by cytometric bead array for the murine cytokines TNF, IFN-γ, IL-6, IL-10, IL-12p70, and MCP-1/CCL2 by using the CBA mouse inflammation kit (BD Biosciences, Germany). Each kit was used according to the manufacturer's instructions.

### Flow Cytometry

Cells were incubated with human or mouse FcR blocking reagent (Miltenyi Biotec) according to the manufacturer's instructions prior to antibody staining. Cells were stained with the following antibodies: anti-human CD11b (ICRF44, eBioscience, Germany), anti-mouse CD11b (M1/70, eBioscience), CD11c (N418, eBioscience), CD68 (FA-11, eBioscience), F4/80 (BM8, eBioscicence), MHCII (M5/114.15.2, eBioscience), anti-Moesin (38/87, Sigma-Aldrich; EP1863Y, Abcam), anti-mouse IgG1 (M1- 14D12, eBioscience) or anti-rabbit IgG (poly4064, Biolegend; polyclonal, Invitrogen). Stainings included fixable viability dyes LIVE/DEAD fixable dead cell stain aqua (Invitrogen) or fixable viability dye eFluor780 (eBioscience). Fixable viability dyes and antibodies for cell surface antigens were diluted in PBS and added to cells for 15 min at 4◦C. For staining of intracellular proteins, cells were fixed in 2% paraformaldehyde (PFA, Sigma-Aldrich) and permeabilized using Permeabilizing solution 2 (BD Biosciences). Antibodies for intracellular antigens were diluted in PBS and added to cells for 30 min at 4◦C. For unconjugated primary antibodies, a further incubation step with a secondary fluorochrome-conjugated antibody was performed (30 min, 4 ◦C). Cells were acquired on a FACSCanto or Fortessa (BD Biosciences) in the flow cytometry core facility of the Deutsches Rheuma-Forschungszentrum and data was analyzed in FlowJo (Treestar, USA). The gating strategy employed excluded doublets and dead cells from the analysis and further gates were set based on fluorescence-minus-one (FMO) and unstained controls.

### Phagocytosis Assay

BMDM were incubated with labeled or unlabeled P. berghei ANKA Bergreen parasites expressing GFP under the HSP70 promoter (Kooij et al., 2012). Briefly, P. berghei ANKA Bergreen schizonts were isolated by density gradient centrifugation and stored in 10% glycerol (Roth) in Alsever's solution (Sigma-Aldrich) at −80◦C. Due to freeze-thawing of schizonts, merozoites were released from erythrocytes, subsequently washed with PBS and labeled according to a protocol adapted from Cambos and Scorza (2011). Briefly, merozoites were labeled with 5 µM caroboxyfluorescein diacetate succinimidyl ester (CFDA-SE; Molecular Probes, Germany) or 2.5 µM CellTrace Violet (Molecular Probes) in PBS supplemented with 0.1% FCS for 3 min at room temperature. Residual dye was diluted with 10%FCS in PBS, followed by an additional washing step with 10%FCS in PBS. CFSE- or CellTrace Violet-labeled merozoites were resuspended in BMDM culture medium and added to cells at a ratio of 1:10, i.e., 10 merozoites per one BMDM. Alternatively, P. berghei ANKA Bergreen schizonts were isolated and immediately added at a ratio of 1:10 to BMDM without prior labeling. At indicated time points, cells were collected on ice, washed in PBS, and stained for cell surface markers or immediately resuspended in acquisition buffer. Phagocytosis was analyzed by flow cytometry based on detection of CFSE, CellTrace Violet, or GFP for unlabeled schizonts. Gate settings were determined using untreated BMDM.

### Ethics Statement

All animal work was conducted in accordance with the German "Tierschutzgesetz in der Fassung vom 22. Juli 2009," which implements Directive 2010/63/EU from the European Parliament and Council (On the Protection of Animals Used for Scientific Purposes). The protocol was approved by the ethics committee of the Berlin state authorities (LaGeSo).

### Mice

C57BL/6 and NMRI mice were obtained from Charles River laboratories (Germany). Moesin-deficient (Doi et al., 1999) and corresponding wild type control colonies on a C57BL/6 background were maintained by the Max Planck Institute for Infection Biology breeding facility. SNP genotyping (Taconic, USA) revealed that moesin-deficient mice were backcrossed to the C57BL/6 background for at least six generations. Moesin-deficient mice were identified by PCR using Taq DNA polymerase (Fermentas) and standard cycling conditions with primers Msn\_fwd1 5′ -CTGAAGTCGGACAAAGATTTC CAGG-3′ , Msn\_fwd2 5′ -CATCAGTATATGAAACAGCCCCCT G-3′ , Msn\_rev 5′ -AGGTGTCTCCCAGAGATACGATTTGG-3′ (synthesized by Eurofins Genomics). Mice were kept under specific pathogen-free conditions with ad libitum diet at the Max Planck Institute for Infection Biology animal facility in a 12 h light/12 h dark cycle.

### *P. berghei* ANKA Bergreen *In vivo* Infection

Six to nine week old female Msn-/- or WT mice were infected with 10,000 P. berghei ANKA Bergreen-infected erythrocytes derived from an NMRI donor mouse by i.v. injection. Parasitemia was determined daily by flow cytometry starting day 3 postinfection. Briefly, blood of infected mice was diluted in Alsever's solution containing Hoechst 33342 (Invitrogen) and acquired on a Fortessa flow cytometer (BD Biosciences). Parasitemia was assessed by determining the fraction of GFP<sup>+</sup> Hoechst 33342<sup>+</sup> cells in the samples. Blood from a naïve mouse served as control for determination of gate settings. Cells single positive for Hoechst 33342, representing leukocytes, were excluded from the analysis. Additionally, serum was collected by retrobulbar punction at indicated time points. Mice were closely monitored for behavioral symptoms of ECM, such as convulsions, absence of touch escape, and unconsciousness (Lackner et al., 2006), and sacrificed when presenting with severe neurological impairment.

### Data Analysis

All data was imported into Prism 7.0 (GraphPad) for statistical analysis. Statistical analyses to determine the P-value of each comparison are indicated individually. No statistical analysis was performed on data sets with n < 3.

## RESULTS

### Moesin Interacts with *Pf*GPI

The GPI of Plasmodium origin has been widely described to induce pro-inflammatory responses from host immune system via pattern recognition receptors such as TLR2 and TLR4 (Krishnegowda et al., 2005). However, observations that TLR2/4 double knockout mice show no resistance to cerebral malaria (Togbe et al., 2007) prompted the question whether additional receptors might be involved in this process. In order to look for interaction partners of Pf GPI, we immobilized chemically synthesized GPI glycan (**Figure 1B**) on an affinity column and performed pull-down experiments using plasma membrane preparation from the mouse macrophage cell line RAW264.7. Proteins bound to synthetic GPI glycan were eluted from the column, subjected to SDS-PAGE and identified after trypsin digestion by peptide mass fingerprinting. The membraneorganizing extension spike protein moesin was identified as one of the most prominent GPI-binding protein, and was confirmed by western blot using anti-moesin antibody (**Figure 1A**). Other proteins were also identified but were mostly very abundant proteins that might easily result from contamination such a keratin, actin, and histones. Additionally, proteins such as CD1b were identified along with moesin as potential binding candidates. However, since moesin showed the highest probability after repeated experiments and given the postulated role in LPS signaling (Tohme et al., 1999; Amar et al., 2001; Iontcheva et al., 2004; Zawawi et al., 2010), moesin was selected for further investigations.

### Moesin Binds to *Pf*GPI on Microarray and on Surface Plasmon Resonance (SPR)

In a first attempt to confirm the interaction between moesin and GPI glycans, a carbohydrate microarray containing seven synthetic GPI glycans of different length (**Figure 1D** positions 1 to 7), as well as a cap polysaccharide from Leishmania as control (**Figure 1D** position 8) was constructed. Incubation of Dunst et al. GPI and Moesin in Malaria

recombinantly expressed GST-moesin with the carbohydrate array revealed binding of moesin to all GPI fragments but not to the cap polysaccharide. In a control experiment, the mannosespecific lectin concanavalin A bound all mannose-containing GPI fragments as well as the cap polysaccharide, but not the shortest GPI fragment lacking any mannose (**Figure 1C**). Next, we studied the kinetics of the interaction between moesin and Pf GPI using surface plasmon resonance (SPR). The full GPIglycan of Plasmodium falciparum (**Figure 1B**) was covalently immobilized on an SPR gold chip. Binding was assessed using various concentrations of purified moesin. The result showed a linear concentration-dependent binding of moesin to GPIglycan (**Figure 1E**). Strikingly, the binding constant (KD) for the interaction between moesin and GPI was 9.7 × 10−<sup>4</sup> M, which is relatively weak but is in the typical range for carbohydrate interaction with carbohydrate-binding proteins (Goldstein and Poretz, 1986; Lee and Lee, 1995). Taken together these results clearly indicated an interaction between moesin and the carbohydrate moiety of the Pf GPI anchor.

### Malaria PAMP Do Not Induce Moesin Cell Surface Translocation on Macrophage-Like THP-1 Cells

After clearly establishing the binding of moesin to Pf GPI we next sought to determine the physiological relevance of this interaction in the context of malaria. The classification of Pf GPI as one of the major malaria toxins is due to its ability to induce the secretion of pro-inflammatory cytokines from host innate immune cells such as macrophages in vitro (Schofield and Hackett, 1993; Tachado et al., 1996; Krishnegowda et al., 2005; Zhu et al., 2009). Intriguingly, moesin is primarily described as an intracellular protein (Ponuwei, 2016), yet translocation to the cell surface of e.g., macrophage-like THP-1 cells was reported to be induced by LPS stimulation (Iontcheva et al., 2004). Consequently, it was hypothesized that moesin might acts as a macrophage cell surface receptor for Plasmodium GPI. Given that GPI are abundantly present on the parasite surface (Gowda, 2007), preparations of schizont extracts were used as a stimulus containing Plasmodium GPI. In order to evaluate the responsiveness of human macrophage-like THP-1 cells to stimulation with LPS or GPI-containing P. falciparum schizont extract (Pf SE), TNF secretion was determined in cell culture supernatants by ELISA. While macrophage-like THP-1 cells secrete TNF to a great extent when stimulated with LPS, stimulation with Pf SE also results in a consistent yet less pronounced induction of TNF secretion from differentiated THP-1 cells (**Figure 2A**). Given the suitability of the stimuli applied to elicit a pro-inflammatory response, next we assessed the cell surface localization of moesin upon stimulation of human macrophage-like THP-1 cells with LPS or Pf SE by flow cytometry. Surprisingly, moesin cell surface translocation was not detected on viable differentiated THP-1 cells with antimoesin antibody 38/87 in either LPS- or Pf SE-treated samples at all timepoints tested (**Figure 2B**). This finding was further corroborated by using another anti-moesin antibody clone (EP1863Y) (**Figure 2B**). Moreover, the suitability of the staining protocol was verified by successful intracellular moesin detection with each anti-moesin antibody clone (**Figure 2C**). Taken together, these results indicate that even though macrophage-like THP-1 cells respond to both LPS and Pf SE by TNF secretion, this response is not accompanied by recruitment of moesin to the cell surface.

### Macrophage Response to Malaria PAMP Is Moesin-Independent

Since Pf GPI-induced TNF secretion was reported to be MyD88 dependent (Krishnegowda et al., 2005), another aspect to be investigated by this study was the role of moesin in Plasmodium GPI-induced signal transduction leading to the secretion of pro-inflammatory cytokines. In order to address this question in vitro, we isolated cells from moesin-deficient mice (Doi et al., 1999). Moesin-deficient mice were identified by PCR (Supplementary Figure 2A) and lack of moesin protein was validated via western blot. Surprisingly, use of anti-moesin antibody 38/87 suggested the presence of moesin in cells isolated from moesin-deficient mice (Supplementary Figure 2B), while anti-moesin antibody EP1863Y confirmed the absence of moesin from moesin-deficient mice (Supplementary Figure 2C). Since anti-moesin antibody 38/87 recognizes a second, slightly heavier protein in human THP-1 cells (Supplementary Figure 2B), and the protein recognized in murine samples corresponds to the size of the upper band detected in human THP-1 cells, it may be concluded that this antibody clone binds unspecifically to a protein other than moesin in murine samples. Consequently, anti-moesin antibody clone EP1863Y was used for subsequent experiments. Next, bone marrow-derived macrophages (BMDM) were generated from both wild type (WT) and moesin-deficient C57BL/6 mice. Successful differentiation was confirmed by flow cytometric analysis of the key mouse macrophage surface markers CD11b and F4/80, as well as intracellular CD68 (Gonçalves and Mosser, 2015). All viable WT and moesindeficient BMDM expressed both, CD11b and F4/80, and cells were also largely positive for CD68, thus verifying efficient differentiation of bone marrow-derived cells (Supplementary Figure 2D).

Upon confirmation of differentiation of WT and moesindeficient BMDM, cells were stimulated with LPS, Pf SE, P. berghei schizont extract (PbSE), and P. berghei schizonts (PbS) for 24 h and cell culture supernatants were analyzed for TNF concentrations by ELISA or CBA. Despite previous reports on markedly reduced LPS-induced TNF secretion in the absence of moesin (Iontcheva et al., 2004) or antibody-mediated moesin blocking (Tohme et al., 1999; Amar et al., 2001; Zawawi et al., 2010), TNF concentration in supernatants of LPStreated moesin-deficient BMDM was similar to WT BMDM in the experimental setting used here (**Figure 3A**). Additionally, the TNF response induced by Plasmodium-derived stimuli was not markedly different between WT and moesin-deficient BMDM. Since the TNF response induced by malaria PAMPs was generally low and variable for both WT and moesindeficient cells, potential differences in TNF secretion may not be detectable in this experimental setting. Therefore, transcript

levels of Tnf and Il1b were determined by qPCR in both WT and moesin-deficient BMDM after 4 h of stimulation under otherwise unchanged conditions. In good agreement with TNF concentration in BMDM supernatants, induction of TNF transcripts was most pronounced in LPS-treated samples (∼100-fold compared to untreated BMDM). Transcription of

the Tnf gene was also induced by Plasmodium-derived stimuli, yet to a much lower extent (∼5-fold compared to untreated BMDM) than that initiated by LPS treatment (**Figure 3B**). Nevertheless, induction of TNF transcripts was similar in WT and moesin-deficient BMDM for all conditions tested. Another Plasmodium GPI-induced cytokine is IL-1β (Schofield and Hackett, 1993), and LPS-induced secretion of IL-1β has also been demonstrated to be moesin-dependent (Zawawi et al., 2010), thus indicating a role for moesin in signal transduction leading to IL-1β secretion. However, in the present study, induction of Il1b gene transcription did not markedly differ in WT and moesin-deficient BMDM for all stimuli applied, including LPS (**Figure 3C**). Taken together, these results demonstrate that the macrophage pro-inflammatory response to malaria PAMPs in terms of TNF secretion as well as Tnf and Il1b gene transcription is not as pronounced as that elicited by LPS. Furthermore, moesin-deficient and WT BMDM responded similarly to LPS and Plasmodium-derived stimuli, suggesting that moesin is not essential in signaling processes leading to TNF secretion and Il1b gene transcription in murine BMDM.

Since it cannot be excluded that stimulation-induced cell surface localization of moesin is required for prompting other processes so far not covered in this study, we also investigated the translocation of moesin to the cell surface of macrophages upon stimulation with PAMPs in the murine system. Therefore, WT BMDM were stimulated with LPS, Pf SE, and PbSE for different periods of time and analyzed for moesin cell surface expression via flow cytometry. In line with our observations in differentiated human THP-1 cells, localization of moesin to the cell surface of viable WT BMDM was not detected with anti-moesin antibody clone EP1863Y in either LPS-, Pf SE-, or PbSE-treated samples at all timepoints tested (**Figure 4A**), while the intracellular staining control demonstrated the suitability of the protocol to detect moesin (**Figure 4B**).

### Moesin-Deficient Macrophages Display Unaltered Phagocytic Activity

An alternative site for moesin and Plasmodium GPI to interact is at the phagosomal membrane (Desjardins et al., 1994; Defacque et al., 2000) and thus moesin might orchestrate parasite recognition and/or degradation (Defacque et al., 2000; Erwig et al., 2006). Consequently, experiments aimed at investigating whether moesin is critically involved in phagocytic uptake and degradation of P. berghei merozoites or schizonts were performed.

In order to determine the impact of moesin-deficiency on P. berghei merozoite internalization, CFSE- or CellTrace Violetlabeled P. berghei merozoites were added to WT and moesindeficient BMDM for different periods of time and phagocytosis was analyzed by flow cytometry (**Figures 5A,B**). Since the two approaches revealed comparable phagocytic uptake of P. berghei merozoites by WT and moesin-deficient BMDM, unlabeled P. berghei schizonts expressing high levels of GFP [ANKA Bergreen; (Kooij et al., 2012)] were used next in order to exclude that excess CFSE or CellTrace Violet dye resulted in BMDM labeling instead of reflecting phagocytosis. Although the three approaches differed in overall detection levels of phagocytic activity, phagocytic uptake of parasites was consistently similar in WT and moesin-deficient BMDM within experiments at all-time points tested (**Figures 5A–C**). Additionally, WT and moesindeficient BMDM were incubated with P. berghei merozoites for 4 and 24 h in order to assess parasite degradation on the RNA level via qPCR. In good agreement with the data obtained by flow cytometry, P. berghei rRNA levels were similar in WT and moesin-deficient BMDM after 4 h of incubation (**Figure 5D**), thus confirming that moesin is not essential for phagocytic uptake of P. berghei merozoites. Additionally, P. berghei rRNA levels were markedly reduced after 24 h of incubation in both WT and moesin-deficient BMDM to a similar extent

(**Figure 5D**), thereby indicating that lack of moesin does not impair phagosomal degradation of parasite material.

### Moesin Deficiency Does Not Affect Dendritic Cell Response to Malaria PAMP

In line with the observations of the present study, Wu et al. (2015) reported that macrophage responsiveness is impaired due to Plasmodium-induced phagosomal acidification and that early cytokine responses to Plasmodium infection are rather DC-mediated. Given that macrophages and DCs both recognize PAMPs via pattern recognition receptors and are capable of phagocytosis (Gazzinelli et al., 2014), the interaction of moesin and Plasmodium GPI may serve a similar function like that proposed for macrophages in DCs. Consequently, bone marrow-derived dendritic cells (BMDC) were generated and successful differentiation was confirmed by flow cytometric analysis of viable WT BMDC which expressed key DC surface markers, i.e., CD11b, CD11c, and MHCII (Lutz et al., 1999), while only few cells expressed the macrophage marker F4/80 (**Figure 6A**). In line with previous reports (Madaan et al., 2014), LPS-priming induced maturation of BMDC, as indicated by a population shift to high level MHCII expression (Villadangos et al., 2005), demonstrating that the cells are properly differentiated.

Next, WT and moesin-deficient BMDC were stimulated with LPS, PbSE, and P. berghei schizonts and TNF as well as IL-6 concentrations in supernatants were analyzed by CBA. In good agreement with previous observations in BMDM, the LPSand malaria PAMP-induced response of BMDC revealed subtle and inconsistent differences between WT and moesin-deficient BMDC for all conditions tested (**Figures 6B,C**). Even though the cytokine response to Plasmodium-derived stimuli is rather low, these results indicate that the BMDC response to these pathogen-derived stimuli as measured by TNF and IL-6 secretion is independent of the presence of moesin.

### Immune Response to *P. berghei* ANKA and Infection Outcome Are Moesin-Independent

While a pivotal role of moesin-Plasmodium GPI interaction for macrophage responsiveness or functionality could not be demonstrated with the in vitro assays performed in this study, the interaction of moesin with Plasmodium GPI may be relevant in other Plasmodium-related host responses in vivo. Additionally, moesin may be involved in key events leading to the development of malaria pathology independent of its interaction with Plasmodium GPI, since moesin was described to be relevant in immunological synapse formation (Itoh et al., 2002; Parameswaran and Gupta, 2013) and in endothelial permeability (Koss et al., 2006; Yao and Tsirka, 2011). Thus, moesin-deficient C57BL/6 mice and the corresponding wild type controls were infected with P. berghei ANKA Bergreen blood stage parasites in order to investigate the potential involvement of moesin in the development of ECM. Despite the previously mentioned indications for moesin to play a role during processes leading to ECM, moesin-deficient mice were not protected from development of ECM (**Figure 7A**) concomitant with normal progression of parasite growth (**Figure 7B**). Moreover, lack of moesin did not affect the proinflammatory immune response mounted upon P. berghei ANKA

(*n* = 1), reduction in *P. berghei* 18 s rRNA indicates phagosomal degradation.

infection in terms of serum levels of TNF, IL-6 and MCP-1/CCL2 (**Figures 7C,E,F**). Noteworthy, the interferon (IFN)-γ response differed significantly (p = 0.03) between WT and moesindeficient mice at days 3 and 5 post-infection (**Figure 7D**), yet higher serum IFN-γ concentration at day 5 post-infection did not impact the overall course of infection in moesin-deficient mice.

### DISCUSSION

In the present study, the ERM protein moesin was found to interact with Plasmodium GPI in vitro and the relevance of this interaction was further investigated in the context of Plasmodium-induced pro-inflammatory responses and pathology. Moesin is a protein of the ERM family, which undergoes conformational change upon phosphatidylinositol-4,5-bisphosphate (PIP2)-mediated phosphorylation (Neisch and Fehon, 2011) and links actin filaments to transmembrane proteins (Louvet-Vallee, 2000). Thereby, ERM proteins contribute to cytoskeletal rearrangement, cellular migration, and membrane dynamics (Ponuwei, 2016). In addition to ERM proteins interacting with phosphatidylinositide 3-kinase, protein kinase A, or Rho-specific GDP dissociation inhibitors (Ivetic and Ridley, 2004; Niggli and Rossy, 2008; Ponuwei, 2016), it has been suggested that LPS-induced TNF secretion is mediated by moesin signaling through the adapter protein MyD88 (Zawawi et al., 2010).

Moesin is a membrane-associated intracellular protein, which was previously reported to translocate to the cell surface upon LPS stimulation (Iontcheva et al., 2004; Takamatsu et al., 2009) and to be constitutively present on the surface of lymphocyte subsets (Takamatsu et al., 2009). Therefore, it was hypothesized that, upon translocation to the cell surface, moesin might interact with Plasmodium GPI. In contrast to previous reports on LPSinduced moesin cell surface translocation (Iontcheva et al., 2004; Takamatsu et al., 2009), we were unable to detect moesin on the cell surface of LPS-stimulated macrophage-like THP-1 cells or BMDM. Since intracellular moesin was detected in THP-1 cells and BMDM, and given that THP-1 cells as well as BMDM and BMDC responded to LPS-stimulation with TNF secretion, we were able to demonstrate that our protocol is generally suitable to detect moesin by flow cytometry and to elicit an LPSinduced pro-inflammatory cytokine response from different cell types. Although transient or nominal cell surface translocation of moesin upon LPS or Pf SE/PbSE stimulation may not have been detected with the experimental settings used here, our results indicate permanent absence of moesin from the cell surface of macrophages in vitro.

The phagosome may be an alternative site for Plasmodium GPI and moesin to interact, since merozoite surface GPI become exposed upon schizont degradation. Moesin was described to be associated with phagosomes in murine J774 and human U937 macrophages (Desjardins et al., 1994; Defacque et al., 2000) and to be involved in phagosomal acidification (Erwig et al., 2006). However, in accordance with a previous study reporting that the rate of phagocytosis of apoptotic cells is moesin-independent (Erwig et al., 2006), moesin-deficiency did not affect non-opsonic uptake of P. berghei merozoites or schizonts in three independent experiments. Additionally, the absence of moesin did not have an impact on phagosomal degradation of P. berghei 18 s rRNA. Even though degradation of P. berghei 18 s rRNA was only analyzed once and at a limited number of time points, markedly reduced 18 s rRNA levels after 24 h of incubation in both wild type and moesin-deficient BMDM indicate that the presence of moesin is not critical for this process. Notably, the degradation of parasite

dilution in medium), and *P. berghei* schizonts (*Pb* schizonts, added at a ratio of 1:10 to BMDC), determined after 24 h incubation by CBA; (B) TNF supernatant

concentrations of stimulated BMDC, (C) IL-6 concentrations in supernatants of BMDC; Due to *n* = 1, no statistical analysis was performed.

components other than 18 s rRNA, e.g., hemozoin, was not quantified. Although Fc receptor-mediated phagocytic uptake of antibody-opsonized P. berghei merozoites or schizonts could be affected by moesin-deficiency, previously published data suggest that ERM proteins do not localize to phagosomes containing opsonized cells (Erwig et al., 2006).

Previous studies investigating the pro-inflammatory response to Plasmodium GPI demonstrated pronounced TNF secretion from BMDM in vitro using purified Pf GPI immobilized on gold particles (Krishnegowda et al., 2005; Zhu et al., 2009, 2011). Additionally, purified GPI of other protozoan parasites such as Toxoplasma gondii have been described to induce TNF secretion from murine macrophages in vitro (Debierre-Grockiego et al., 2003), indicating that protozoan GPI represent conserved PAMP. Although malaria PAMP other than GPI were present in the schizont extracts used here, which could activate pattern recognition receptors independent of moesin, we observed minor cytokine secretion and subtle induction of cytokine transcription in both wild type and moesin-deficient BMDM and BMDC. Thus, it seems unlikely that a potential contribution of moesin to Plasmodium GPI-induced signaling has been masked by activation of other pathways leading to cytokine production.

In line with our observation of a subtle induction of proinflammatory cytokines upon BMDM stimulation with P. berghei schizonts, another recent study reported that the cytokine response of BMDM is impaired upon internalization of P. falciparum- or P. berghei-infected erythrocytes or merozoites due to pronounced phagosomal acidification and consequential inactivation of endosomal TLR (Wu et al., 2015). Furthermore,

the majority of splenic macrophages isolated from P. berghei ANKA-infected mice did not produce inflammatory cytokines (Wu et al., 2015). Collectively, our data support the notion that macrophages contribute marginally to the pro-inflammatory cytokine response at early stages of infection (Stevenson and Riley, 2004). Interestingly, it was recently demonstrated that in vitro stimulation of BMDC with P. falciparum-infected erythrocytes or merozoites resulted in cytokine secretion in a dose-dependent manner (Wu et al., 2015). In contrast, the BMDC cytokine responses to Plasmodium-derived stimuli observed here were subtle and inconsistent. Since the properties of BMDC generated in vitro vary depending on the stimuli used to induce differentiation (Xu et al., 2007), these BMDC may not properly reflect the spectrum of DC subtypes, e.g. monocytederived and plasmacytoid DC (Heath and Carbone, 2013), yet cytokine responses to P. berghei ANKA infection are DC subtype-dependent in vivo (Wu et al., 2015). However, cytokine secretion and transcription were markedly induced by LPS, suggesting that BMDC were generally proficient to respond to TLR ligands. Notably, despite previous reports on considerably reduced LPS-induced TNF secretion from THP-1 cells when moesin was silenced (Iontcheva et al., 2004) or blocked by antimoesin antibody (Tohme et al., 1999; Amar et al., 2001; Zawawi et al., 2010), the LPS-induced TNF response was not affected by moesin-deficiency when compared to wild type BMDM or BMDC. Consequently, these results indicate that the LPSinduced cytokine response from murine BMDM and BMDC is moesin-independent in vitro. Although moesin likely serves similar functions in human and murine cells, further studies are needed to clarify the localization and the role of moesin in the pro-inflammatory immune response to LPS in human cells.

Moesin is the predominantly expressed ERM protein in T-cells (Itoh et al., 2002) and ERM proteins are involved in immunological synapse formation, thereby modulating T and B cell activation (Itoh et al., 2002; Parameswaran and Gupta, 2013). Furthermore, moesin is the predominant ERM protein in endothelial cells (Berryman et al., 1993) and was found to be involved in TNF-induced endothelial cell paracellular gap formation, resulting in increased endothelial permeability in vitro (Koss et al., 2006). Nevertheless, moesindeficiency did not impair pro-inflammatory cytokine responses to P. berghei ANKA infection and did not affect ECMassociated mortality in vivo. These findings point toward a dispensable role for moesin as well as the interaction of moesin with Plasmodium GPI in the immune response to P. berghei ANKA infection and in the manifestation of symptoms of ECM.

Collectively, our findings demonstrate that even though Plasmodium GPI and moesin interact, the relevance of this interaction in the context of malaria pathology could not be established. Interestingly, although moesin has been described to be the predominantly expressed ERM protein in lymphocytes (Itoh et al., 2002) and endothelial cells (Berryman et al., 1993), it cannot be excluded that moesin-deficiency may be compensated for by other proteins of the ERM family, especially taking partial functional redundancy of ERM proteins into consideration (Niggli and Rossy, 2008). Additionally, given the conflicting in vitro data and considering that TLR-deficiency did not protect mice from ECM (Togbe et al., 2007; Lepenies et al., 2008), the precise mechanisms leading to the activation of the innate immune system during Plasmodium infection, as well as the relevance of GPI in this process, need to be further investigated to elucidate the underlying mechanisms of malaria pathology.

### AUTHOR CONTRIBUTIONS

FK conceived the study. JD, NA, XL, and FK designed and performed the experimental work. JD, NA, PS, and FK analyzed and interpreted data. ST provided moesin-deficient embryos. JD and FK wrote the manuscript. All authors approved the final version of the manuscript.

### REFERENCES


### FUNDING

This work was supported by the German Research Foundation grant to FK (KA 3347/4-1), the German Federal Ministry for Education and Research (BMBF) and the Max Planck Society.

### ACKNOWLEDGMENTS

We thank Kai Matuschewski for insightful comments and fruitful discussions. We also thank Toralf Kaiser and Jenny Kirsch of the flow cytometry core facility at the Deutsches Rheuma-Forschungszentrum (Berlin) as well as the Max Planck Institute for Infection Biology breeding facility for technical assistance. We thank Pedro Moura-Alves and Soo-Kyung Peuschel for kindly providing THP-1 and L-929 cells, respectively. Parts of this work have been published previously in a PhD thesis by JD and may therefore present similarities in wording and/or content without constituting plagiarism.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fcimb. 2017.00183/full#supplementary-material

ezrin or radixin in moesin gene knockout. J. Biol. Chem. 274, 2315–2321. doi: 10.1074/jbc.274.4.2315


**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 Dunst, Azzouz, Liu, Tsukita, Seeberger and Kamena. 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.

# Chronic *Toxoplasma gondii* Infection Exacerbates Secondary Polymicrobial Sepsis

Maria C. Souza<sup>1</sup> , Denise M. Fonseca<sup>1</sup> , Alexandre Kanashiro<sup>2</sup> , Luciana Benevides <sup>1</sup> , Tiago S. Medina<sup>1</sup> , Murilo S. Dias <sup>1</sup> , Warrison A. Andrade<sup>3</sup> , Giuliano Bonfá<sup>1</sup> , Marcondes A. B. Silva<sup>2</sup> , Aline Gozzi <sup>4</sup> , Marcos C. Borges <sup>4</sup> , Ricardo T. Gazzinelli <sup>3</sup> , José C. Alves-Filho<sup>2</sup> , Fernando Q. Cunha<sup>2</sup> and João S. Silva<sup>1</sup> \*

<sup>1</sup> Department of Biochemistry and Immunology, Ribeirão Preto Medical School, University of São Paulo, São Paulo, Brazil, <sup>2</sup> Department of Pharmacology, Ribeirão Preto Medical School, University of São Paulo, São Paulo, Brazil, <sup>3</sup> Department of Medicine, University of Massachusetts Medical School, Worcester, MA, USA, <sup>4</sup> Department of Internal Medicine, Ribeirão Preto Medical School, University of São Paulo, São Paulo, Brazil

Sepsis is a severe syndrome that arises when the host response to an insult is exacerbated, leading to organ failure and frequently to death. How a chronic infection that causes a prolonged Th1 expansion affects the course of sepsis is unknown. In this study, we showed that mice chronically infected with Toxoplasma gondii were more susceptible to sepsis induced by cecal ligation and puncture (CLP). Although T. gondii-infected mice exhibited efficient control of the bacterial burden, they showed increased mortality compared to the control groups. Mechanistically, chronic T. gondii infection induces the suppression of Th2 lymphocytes via Gata3-repressive methylation and simultaneously induces long-lived IFN-γ-producing CD4<sup>+</sup> T lymphocytes, which promotes systemic inflammation that is harmful during CLP. Chronic T. gondii infection intensifies local and systemic Th1 cytokines as well as nitric oxide production, which reduces systolic and diastolic arterial blood pressures after sepsis induction, thus predisposing the host to septic shock. Blockade of IFN-γ prevented arterial hypotension and prolonged the host lifespan by reducing the cytokine storm. Interestingly, these data mirrored our observation in septic patients, in which sepsis severity was positively correlated to increased levels of IFN-γ in patients who were serologically positive for T. gondii. Collectively, these data demonstrated that chronic infection with T. gondii is a critical factor for sepsis severity that needs to be considered when designing strategies to prevent and control the outcome of this devastating disease.

Keywords: sepsis, septic shock, *Toxoplasma gondii*, coinfection, chronic disease

## INTRODUCTION

The majority of studies regarding host-pathogen relationships have focused on the interaction of a single pathogen with its host. However, humans are commonly exposed recurrently or simultaneously to multiple pathogens (Jamieson et al., 2010; Telfer et al., 2010). Understanding how previous/simultaneous infections can modify the host immune response and consequently affect the outcome of a secondary infection is crucial to designing new therapeutic strategies to control coinfections.

#### *Edited by:*

Kai Matuschewski, Humboldt University of Berlin, Germany

#### *Reviewed by:*

Frank C. Gibson, III, University of Florida, USA Serge Ankri, Technion, Israel

#### *\*Correspondence:*

João S. Silva jsdsilva@fmrp.usp.br

*Received:* 21 December 2016 *Accepted:* 23 March 2017 *Published:* 07 April 2017

#### *Citation:*

Souza MC, Fonseca DM, Kanashiro A, Benevides L, Medina TS, Dias MS, Andrade WA, Bonfá G, Silva MAB, Gozzi A, Borges MC, Gazzinelli RT, Alves-Filho JC, Cunha FQ and Silva JS (2017) Chronic Toxoplasma gondii Infection Exacerbates Secondary Polymicrobial Sepsis. Front. Cell. Infect. Microbiol. 7:116. doi: 10.3389/fcimb.2017.00116

Sepsis, a life-threatening disease associated with high morbidity and mortality worldwide, is caused by a dysregulation of the immune system to different infectious agents (Schmid et al., 2004). An uncontrolled infection induces a systemic inflammatory reaction that culminates in a cytokine storm. Such inflammatory mediators deregulate the cardiovascular system, cause vascular permeability, and lead to severe sepsis and septic shock characterized by severe hypotension, multi-organ failure, and death (Bone et al., 1997; Hotchkiss et al., 2009).

Toxoplasmosis is a highly prevalent protozoan infection, affecting approximately one-third of the world population (Robert-Gangneux and Darde, 2012). The oral route of infection favors the disruption of the intestinal epithelium caused by inflammatory stressors induced against the parasite and facilitates the spread of the parasite toward different target organs (Dubey et al., 1998; Montoya and Liesenfeld, 2004). Locally, innate cells express sensors such as Toll-like receptors (TLRs) that are indispensable to parasite recognition (Debierre-Grockiego et al., 2007; Andrade et al., 2013). T. gondii infection induces a scenario of intense inflammation, characterized by IFN-γ-producing long-lived CD4<sup>+</sup> and CD8<sup>+</sup> T lymphocytes (Gazzinelli et al., 1994; Mashayekhi et al., 2011; Hand et al., 2012). Although the parasite growth is controlled efficiently, it triggers intense tissue damage that is commonly harmful to the host (Mashayekhi et al., 2011). After the acute phase of infection, the parasite is maintained latently in the brain and skeletal muscle, leading to a chronic state (Munoz et al., 2011). This evidence suggests that long-lived T. gondii-induced lymphocytes have an intrinsic molecular programme that is promptly activated to release excessive amounts of IFN-γ after a secondary pathogenic exposure. Although clear evidence exists that sepsis modulates secondary infections (Nascimento et al., 2010), it is currently unknown whether previous infections can interfere with the sepsis outcome.

In this study, we showed that IFN-γ-producing lymphocytes induced by T. gondii parasites persist after the acute phase of infection and are deleterious during polymicrobial sepsis. Mechanistically, T. gondii infection is followed by Gata3 methylation and increased transcription of IFN-γ-related genes in CD4<sup>+</sup> T cells, thus inducing long-lived memory T cells. The partial blockage of IFN-γ prevented massive cytokine production, arterial hypotension, and prolonged host lifespan. Notably, these data mirrored our observation in patients because elevated serum levels of IFN-γ correlate with sepsis severity. Additionally, patients serologically positive for T. gondii had increased serum levels of IFN-γ compared to patients who were serologically negative. These observations demonstrate that chronic infection with T. gondii aggravates the course of sepsis and opens new avenues to design strategies to control the severity of T. gondii-sepsis coinfection.

### MATERIALS AND METHODS

### Ethics Statement

The research was approved by the Ethics Committee on Animal Experiments of Ribeirão Preto Medical School (CETEA-107- 2009) and the Institutional Animal Care and Use Committee of the University of Massachusetts Medical School (IACUC-2371- 12). The study on human samples was approved by the Human Research Ethics Committee at Hospital das Clínicas of Faculdade de Medicina da Universidade de São Paulo (research protocol n◦ . 536/2008). After obtaining written informed consent from all of the patient's relatives, venous blood samples were collected.

### Mice and Parasites

Female C57BL/6 mice were used. The low-virulent ME-49 strain of T. gondii was harvested from the brains of infected mice (Benevides et al., 2008). Since our major aim was to address sepsis susceptibility in chronically infected mice, we only used the ME-49 strain that allows mice to progress to chronic phase of infection.

### Polymicrobial Sepsis Model

Sepsis was induced using a cecal ligation and puncture (CLP) model. Two punctures with sterile 21-G needles were used to standardize the sub-lethal sepsis (Benjamim et al., 2003; Nascimento et al., 2010).

### Bacteria Count Determination and Leukocyte Migration to the Peritoneum

The quantification of the bacterial load in the blood and peritoneal exudates was performed at 6, 12, and 24 h after CLP. For these analyses, the animals were anesthetized, and blood was collected via cardiac puncture, following which the animals were euthanized in a CO<sup>2</sup> chamber. The peritoneal exudates were collected via an injection of 1.5 mL of PBS/EDTA into the peritoneal cavity. After the sample collection, 10 µL of blood or peritoneal wash without dilutions were plated on Mueller-Hinton agar (Difco Laboratories, Detroit, MI, US) and incubated at 37◦C under aerobic conditions for 24 h. Colonyforming units (CFUs) were expressed as Log2 of CFU/10 µL of blood or peritoneal wash. All procedures were performed under sterile conditions (Nascimento et al., 2010). Leukocyte migration and differential count were assessed 6, 12, and 24 h after the induction of CLP. The cells present in the peritoneal cavity were harvested via washing of the peritoneal cavity using 1.5 ml of phosphate buffered saline (PBS) containing EDTA (1 mM). The total leukocyte counts were obtained with a cell counter (Coulter AC T series analyser, Coulter, Miami, FL), and the differential count was performed using flow cytometry (BD Immunocytometry System, Franklin Lakes, NJ, USA). The results were expressed as the number of total leukocytes, neutrophils or lymphocytes in the peritoneal cavity (Alves-Filho et al., 2006).

### Histopathological Analyses

For pathological analyses, the gut was washed to remove the intestinal contents, and the ileum fragment was individually wrapped in a "Swiss roll," fixed in 10% formalin, embedded in paraffin and processed routinely for haematoxylin and eosin staining. Slides were imaged using light microscopy. The images were acquired with a digital camera (Leica DC300F, Switzerland) coupled to a microscope for histological analysis.

### Flow Cytometry

All antibodies used for flow cytometry were purchased from BD Biosciences or eBiosciences and used according to the manufacturer's instructions. For neutrophil staining, Ly6G (FITC, 1A8), CD11c (APC-Cy7, N418), CD11b (PE Cy7, M170) and F4/80 (Percp Cy5, BM8) were used. For intracellular cytokine staining, CD3 (FITC, 145-2C11), CD4 (PercP, RM4-5), CD8 (PE Cy7, 53-6.7), IFN- γ (APC, XMG1.2) and TNF-α (PE, MP6X722) were used. For transcription factors, CD3 (FITC, 145-2C11), CD4 (APC Cy7, RM4-5), CD44 (PercP, IM7) and Tbet (Alexa 647, 4B10) or CD3 (PE, 145-2C11), CD4 (APC Cy7, RM4-5), CD44 (PercP, IM7), RorγT (APC, Q31-378) and Gata3 (Alexa 488, L50- 823) were used. For memory cells, CD3 (PE, 145-2C11), CD4 (APC Cy7, RM4-5), CCR7 (Alexa 647, 3D12) and CD62L (FITC, MEL-14) were used. Briefly, tissue-isolated cells were incubated with monoclonal antibodies in buffer containing blocking antibody. For intracellular staining, the cells were harvested and plated with PMA-ionomycin and Brefeldin A (Golgi plug BD bioscience), and stained for flow cytometry analysis. After incubation, cells were fixed and permeabilized with the BD Cytofix/Cytoperm kit (BD Biosciences and eBiosciences, CA, USA). Cell acquisition was performed on the BD CANTO II cell analyser (BD Biosciences) using FACSDiva software (BD Biosciences), and data were analyzed using FlowJo software (Tree Star, Ashland, OR).

### Measurement of Cytokine and Chemokine Levels

The levels of mouse TNF-α, IFN-γ, IL-6, and KC were measured using DuoSet ELISA kits (R&D Systems).

### Induction of Colitis

Chronic dextran sodium sulfate (DSS) colitis was induced by administering 3% (w/v−1) DSS (molecular mass, 36–50 kDa; MP Biomedicals, OH) for 14 days, followed by 26 days of normal drinking water.

### Antibody Treatment

The cytokine IFN-γ was neutralized by intraperitoneal (i.p.) injection of 10 µg of purified mAb 24 h before CLP. The mouse anti-IFN-γ mAb was purified from the ascites of mice injected with an anti-IFN-γ hybridoma (XMG1.2). Controls received 500 µg of normal rat IgG diluted in PBS. The IgG was isolated from naïve rats using protein G column purification.

### Antibiotic Treatment

The broad-spectrum antibiotic treatment regimen was followed as described by Hand et al. (2012) with modifications. Briefly, each animal received a daily combination of 5 mg neomycin trisulfate (Sigma), 2.5 mg vancomycin (Sigma), 5 mg metronidazole (sigma) and 5 mg ampicillin sodium (sigma) diluted in 200 µL water via oral gavage for 2 weeks before and 2 weeks post-infection with T. gondii.

### Epigenetics Analysis

For cell isolation, CD4<sup>+</sup> T cells were negatively selected using the EasySep kit (StemCell Technologies, Canada), and DNA extraction was subsequently performed. The epigenetic analysis was performed using the PCR kit EpiTect Methyl II PCR Arrays (Qiagen Sciences, USA) to determine the methylated CpG islands and indicate the percentage of methylated DNA. The DNA in each individual enzymatic reaction was quantified using realtime PCR with primers that flanked the promoter region of interest as follows: tbx21 - Chr11, CpG-start 96975880, CpGend 96976654; eomes - Chr9, CpG-start 1183859814, CpG-end 118386331; gata3 - Chr2, CpG-start 9802719, CpG-end 9803061 and rora - Chr9, CpG-start 68501352, CpG-end 68502769.

### Blood Pressure Analysis

Arterial blood pressure was non-invasively measured by determining the tail blood volume with a pressure recording sensor and an occlusion tail cuff (CODA System, Kent Scientific, CT) 24 h after CLP. The results were expressed in millimeters of mercury (mmHg).

### Patient Samples and Experiments

The patients' blood was collected in the first 24 h after admission. The samples were processed and analyzed for cytokine production and Toxoplasma gondii serology. The levels of the cytokine IFN-γ were measured using DuoSet ELISA as indicated by the manufacturer's instructions (R&D Systems). The Toxoplasma gondii serology was performed by indirect immunofluorescence assay as indicated by the manufacturer's instructions (FLUOCON IgG/IgM WAMA, Belgium).

### Statistics

The animal survival was expressed as the percentage of surviving animals analyzed using the Mantel-Cox log-rank test (X2, chisquared). For comparison of multiple parametric data, the variance (ANOVA) tests were used followed by the post-hoc Tukey-Kramer test. Data are expressed as the means ± standard error of the mean (SEM). Statistical analysis and graphics were performed using the GraphPad Prism version 5.0 (GraphPad Software, San Diego, CA, USA).

## RESULTS

### Chronic *T. gondii* Infection Increases the Susceptibility to Polymicrobial Sepsis

Mice chronically infected with T. gondii for 40 days were subjected to sublethal sepsis induced by CLP surgery (coinfected mice). Herein, we observed increased mortality of coinfected mice compared to sublethal CLP-subjected mice (**Figure 1A**). These data indicate that chronic T. gondii infection aggravated polymicrobial sepsis, which was not due to toxoplasmosis reactivation (Figures S1A,B).

To investigate whether coinfected mice were able to control bacterial proliferation, the bacterial burden was quantified in the blood and peritoneum. We found that 24 h after CLP, coinfected mice were more efficient in controlling bacterial replication both systemically (**Figure 1B**) and locally (**Figure 1C**) compared to septic mice. Moreover, we counted the number of leukocytes recovered from the peritoneal cavity after CLP. The total leukocyte count strikingly decreased in the blood

analyses. The survival rate was evaluated until the 10th day post-CLP induction (A). These data are representative of three independent experiments (n = 10), and the statistical analysis was determined using the Mantel-Cox log-rank test. The bacterial load was analyzed in the blood (B) and peritoneal lavage (C) at 6, 12, or 24 h post-CLP induction. The results are expressed as the log of the colony-forming units (CFU) per microliter. The leukocytes from the peritoneum (D) were stained with May-Grünwald Giemsa, and the number of neutrophils (E) and lymphocytes (F) were determined using a cell counter. Data are presented as the means ± SEM for 4 animals per group in four independent experiments (ANOVA, followed by Tukey's test; \*p < 0.05; \*\*p < 0.01; \*\*\*p < 0.001). At 24 h post-CLP, histopathological features of the ileum were analyzed after staining with haematoxylin and eosin (H&E). The results are representative of three independent experiments using 4 mice per group (G).

of coinfected mice (data not shown); however, the leukocytes increased in the peritoneum of coinfected mice compared to septic mice (**Figure 1D**). The neutrophil recruitment to the peritoneal cavity was increased in coinfected mice compared to CLP-subjected mice (**Figure 1E**), which paralleled the bacterial clearance. We also observed increased migration of lymphocytes in mice 12 h after CLP (**Figure 1F**). To investigate whether intense cellular recruitment to the peritoneal cavity reflects intestinal inflammation and tissue damage, we evaluated the histological features in the intestine 24 h after sepsis induction. We found extensive neutrophil infiltrate mainly in the ileum of coinfected mice along with decreased mRNA expression of occludin, a tight junction protein (Figures S2A,B), and massive intestinal tissue damage (**Figure 1G**). These data suggest that chronic T. gondii infection promotes a microenvironment in the intestine that favors the inflammatory response during CLP. Notably, we did not find evidence of vital organ failure in coinfected mice, which was evaluated by either histopathological analysis or biochemical biomarkers for renal, hepatic, cardiac, and muscular dysfunctions.

### Chronic *T. gondii* Infection Intensifies Proinflammatory Cytokine Production during Sepsis

The major factor leading to host susceptibility during sepsis is the increased production of inflammatory cytokines, which can promote septic shock (Ebong et al., 1999). To evaluate whether the high mortality of coinfected mice mirrored the production of inflammatory mediators, we quantified the cytokines in the serum and peritoneal lavage 24 h after CLP. Notably, coinfected mice showed increased levels of IFN-γ (**Figures 2A,E**), TNF-α (**Figures 2B,F**), IL-6 (**Figures 2C,G**), and KC (**Figures 2D,H**) in both the serum and peritoneal lavage compared to the control groups. For the immuneregulatory cytokines, IL-4 was not detected and the levels of IL-10 in the peritoneal lavage were similar in all groups studied. Splenic CD4<sup>+</sup> and CD8<sup>+</sup> T cells of T. gondii-infected mice produced elevated levels of IFN-γ and TNF-α even in the absence of a secondary stimulus (**Figures 3A–D**). Although the size of the spleen from T. gondiiinfected mice was similar to naïve mice, after the induction of sepsis, T. gondii-infected mice experienced a notable reduction in spleen size along with a striking emergence of IFN-γ- and TNFα-producing CD4<sup>+</sup> (**Figures 3E–G**) and CD8<sup>+</sup> (**Figures 3H–J**) T cells into the peritoneum. These data suggest that inflammatory cells were leaving the spleen/bloodstream and reaching the peritoneal cavity. Thus, T. gondii-programmed CD4<sup>+</sup> and CD8<sup>+</sup> T cells may be recruited to the site of sepsis and are the primary source of inflammatory mediators during sepsis in infected mice.

### Long-Lived CD4<sup>+</sup> *T. gondii*-Primed T Cells Release IFN-γ and TNF-α during Sepsis

To further gain insights into the mechanisms by which chronic T. gondii infection aggravates sepsis, we assessed whether chronic infection maintained a pool of long-lived lymphocytes that act as first responders in polymicrobial sepsis. As expected, chronic T. gondii infection maintained a pool of CD4<sup>+</sup> and CD8<sup>+</sup> T cells (CD4+CD44<sup>+</sup> and CD8+CD44<sup>+</sup> T cells, respectively) independent of sepsis (Figures S3A,B), thus confirming that such cells were activated by T. gondii. Indeed, chronic infection deeply induced the transcription of IFN-γ-related genes, which

were intensified during sepsis development (Figure S3C). As long-lived T cells comprise a pool of central and effector memory cells (Wherry et al., 2003), we explored whether T. gondii exposure maintained a pool of memory T cells during its chronic phase. Chronic T. gondii infection induced an increased number of CD4+CD44highCD62L+CCR7<sup>+</sup> T lymphocytes, named here as central memory-like T lymphocytes (**Figures 4A,B**) that were converted to effector memory-like (CD4+CD44highCD62L−CCR7−) T lymphocytes after CLP surgery (**Figures 4C,D**). Herein, sepsis induction reactivated long-lived T. gondii-experienced T cells to produce IFN-γ and TNF-α in both the mesenteric lymph nodes (**Figures 4E,G**) and peritoneum (**Figures 4F,H**). To evaluate the involvement of microbiota in promoting long-lived T cells, we induced colitis with a classical intestinal stressor dextran sodium sulfate (DSS) to promote bacterial translocation followed by CLP. Surprisingly, DSS-induced bacterial translocation was unable to induce central and effector memory-like T lymphocytes (**Figures 4A–D**) and IFN-γ and TNF-A production after sepsis induction (**Figures 4E–H**). These results rule out a potential role of the microbiota in generating deleterious long-lived T cells that cross-react against released bacteria via CLP.

### Chronically *T. gondii*-Infected Mice Develop Hypotension during Sepsis

T. gondii-primed CD4<sup>+</sup> T lymphocytes are polarized toward a Th1 profile (Gazzinelli et al., 1994), and recently, it was demonstrated that IFN-γ suppresses permissive chromatin remodeling in the regulatory region of the Il4 gene (Nishida et al., 2013). To test whether the microenvironment elicited by chronic toxoplasmosis could promote epigenetic reprogramming of lymphocytes, we purified splenic CD4<sup>+</sup> T cells and investigated the repressive methylation status of their transcription factors. During the chronic phase of T. gondii infection, Gata3 was found significantly methylated compared to the naïve CD4<sup>+</sup> T splenic cells (**Figure 5A**). Additionally, repressive methylation of Th17- (Rora) and Th1-associated transcription factors (Tbx21 and Eomes) were undetectable in chronically infected mice (**Figure 5A**). The epigenetic changes observed indicate that Th1 induced expression can be due to Gata3 repression. This provides

further evidence for our hypothesis that Gata3 repression during T. gondii infection drives the immune response toward a Th1 pattern. To characterize whether CD4<sup>+</sup> T lymphocytes are driven to long-lived Th1 lymphocytes during the chronic infection, we analyzed the transcription factors expressed by the Th1 lymphocytes (Tbet). We observed that the T. gondiiinduced inflammatory milieu reprogrammed naïve CD4<sup>+</sup> T lymphocytes into long-lived Th1 lymphocytes independent of CLP (**Figure 5B**). In contrast, T. gondii-primed T cells did not express significant amounts of the transcription factors Rorγt and Gata3 with or without a secondary infection, which excludes the role of long-lived Th17 and Th2 cells in our model.

Nitric oxide (NO) production and the subsequent cytokine storm induced during sepsis are major inductors of hypotension and septic shock (Bone et al., 1997). Indeed, we detected increased NO production in coinfected mice compared to CLPsubjected mice (Figure S4). Since coinfected mice also had increased inflammatory cytokines, mainly IFN-γ, we explored whether chronic T. gondii infection could aggravate sepsis by reducing systolic and diastolic arterial blood pressures. Only coinfected mice developed hypotension through the reduction of systolic and diastolic blood pressures (**Figure 5C**). Moreover, the partial blockage of IFN-γ prevented hypotension in these mice (**Figure 5C**).

Recent data have revealed that T. gondii infection promotes long-lived IFN-γ-producing microbiota-specific CD4<sup>+</sup> T cells (Hand et al., 2012). To verify a possible role of the microbiota in controlling arterial blood pressure of coinfected mice, we treated them with broad-spectrum antibiotics to prevent bacterial translocation during the acute phase of T. gondii infection. We found that depletion of the microbiota did not prevent cytokine production and arterial hypotension (**Figure 5C**). These observations reveal that chronic T. gondii infection intensifies the plethora of cytokines during sepsis and predisposes the host to septic shock. Since the blockade of IFN-γ prevented host hypotension, we evaluated whether treatment with monoclonal anti-IFN-γ antibody could prevent the susceptibility of coinfected mice. The pretreatment with 10 µg/kg of anti-IFN-γ significantly improved the survival of these mice (**Figure 5D**).

### Positive Serology for Toxoplasmosis Increases IFN-γ during Sepsis

To determine whether the IFN-γ-mediated mechanism described in coinfected mice is also observed in coinfected humans, we collected the blood of patients with sepsis to test the serology for toxoplasmosis and to measure IFN-γ production. Our data showed that patients with more severe clinical forms of sepsis had increased IFN-γ serum levels (**Figures 6A,B**), which strongly supports that IFN-γ production contributes to sepsis severity. In addition, patients serologically positive for toxoplasmosis had increased levels of IFN-γ during sepsis compared to serologically negative patients or healthy controls (**Figure 6C**). Similarly, previous exposure to T. gondii was deleterious to the host because mortality was increased in coinfected patients compared to patients who were serologically negative for toxoplasmosis (**Figure 6D**). Collectively, our data supports the hypothesis that

patients with positive serology for toxoplasmosis are at risk for the development of severe sepsis.

### DISCUSSION

Several studies demonstrated that a variety of outcomes is possible during a secondary infection, depending on the route of infection, the type of pathogen, or even the temporal proximity (Gardner, 1981; Navarini et al., 2006; Barton et al., 2007; Gumenscheimer et al., 2007; Humphreys et al., 2008; Miller et al., 2009; Jamieson et al., 2010; Fenoy et al., 2012). In this context, the involvement of a chronic infection followed by acute infections has been poorly explored. T. gondii creates a scenario of intense intestinal inflammation featured by the high prevalence of Th1 lymphocytes. Similarly, sepsis is a multifactorial disease characterized by systemic inflammatory response syndrome (SIRS). In the present study, considering their worldwide prevalence and their inflammatory condition, we conjectured that patients serologically positive for toxoplasmosis could aggravate the outcome of sepsis by intensifying the inflammatory response. Surprisingly, the majority of patients who died of septic shock were serologically positive for toxoplasmosis, which led us to investigate the cellular and humoral mechanisms involved in the immune response in mice that were chronically infected with T. gondii during sepsis development.

Leukocytes are required to control bacterial replication during a microorganism's invasion (Reddy and Standiford, 2010; Kovach and Standiford, 2012). Interestingly, although mice chronically infected with T. gondii were more susceptible to sepsis, they had better control of bacterial proliferation by enhancing leukocyte recruitment to the site of infection, suggesting that the infection provides intrinsic factors for modulating the host immune response that can interfere with the outcome of sepsis. This phenomenon can be possible because the inflammation developed during T. gondii infection is maintained during its chronic phase and interferes with the outcome of the secondary infection caused by CLP. Interestingly, this phenomenon is specific of T. gondii, since other protozoa as Trypanosoma cruzi, Plasmodium chabaudi, or even a fungus, Paracoccidiodes brasiliensis, do not induce an increased susceptibility to sepsis in mice.

It is well known that CD4<sup>+</sup> and CD8<sup>+</sup> memory T cells are essential for the control of T. gondii proliferation, preventing the de-encystation and reactivation of the disease (Gazzinelli et al., 1992). Our findings revealed that T. gondii infection robustly induces long-lived IFN-γ- and TNF-α-producing CD4<sup>+</sup> and CD8<sup>+</sup> T cells that are maintained, reprogrammed and amplified to act against a secondary challenge of sepsis. Our data suggests that central memory-like T cells are induced and maintained in the secondary lymphoid organs during chronic T. gondii infection by expressing the surface receptors CCR7 and CD62L. After sepsis induction, such cells are reprogrammed and quickly converted to effector memory-like cells by reducing their expression of both CCR7 and CD62L, which facilitates their migration to peripheral organs where they produce robust amounts of NO, IFN-γ and TNF-α to control the bacterial burden.

presented as a percentage (D).

There are several factors that must be considered when analyzing T. gondii oral infection. During its acute phase, intestinal injury is responsible for the transient release of bacterial microbiota in the mesenteric lymph nodes, spleen, and liver, thereby triggering the activation and differentiation of memory cells that are specific not only for the parasite but also for antigens of the microbiota (Benson et al., 2009; Hand et al., 2012). Nevertheless, we showed that bacterial translocation did not influence the outcome of sepsis severity because intestinal damage induced by DSS did not induce memory cells or interfere with the host survival after CLP. Thus, the presence of longlived T. gondii-experienced CD4<sup>+</sup> and CD8<sup>+</sup> T cells was more influential in aggravating sepsis than bacterial translocation.

Oral T. gondii infection induces highly inflammatory responses that dysregulate the intestinal epithelium and cause ileitis (Craven et al., 2012). Our findings support that inflammation due to chronic T. gondii infection aggravated the intestinal tissue damage after CLP. Such characteristic features promote a synergistic cytokine effect that can lead to septic shock (Fong et al., 1989; Calandra and Glauser, 1990; Dofferhoff et al., 1991; Dinarello, 1997). Chronic T. gondii infection promoted the overexpression of IFN-γ-related genes, which were exacerbated by CLP. Indeed, the blockade of IFN-γ prevented hypotension and improved the host survival upon sepsis induction.

Recent studies have shown that epigenetic changes, such as methylation, can induce gene silencing and heterochromatin remodeling that inhibits the access of transcription factors to DNA (Jaenisch and Bird, 2003; Esteller, 2007; Gómez-Díaz et al., 2012; Fernández-Sánchez et al., 2013). Here, T. gondii infection promoted the methylation of the gata3 gene, which inhibited access to this gene and impaired the Th2 immune response, what could explain the absence of IL-4 in the peritoneum. IL-10 apparently does not take part of the immunoregulation, since it is produced equally in not infected or infected groups. In contrast, the Tbx21 and Eomes genes were not inhibited by the methylation process, which facilitated the induction of the Th1 immune response and its maintenance during sepsis. Interestingly, naïve C57BL/6 mice exhibited 47% gata3 methylation, which explains in part why such mice strains have an intrinsic susceptibility to pathogens that promote the Th1 immune response.

Since the overproduction of inflammatory cytokines is the primary cause of mortality in septic patients (Puneet et al., 2010), we reduced the robust inflammatory response using an IFN-γ monoclonal antibody. When subjects were treated with low doses of anti-IFN-γ, we observed a significant improvement in the survival rate of septic mice that were previously infected by T. gondii. These results support that although this cytokine is necessary to control chronic T. gondii infection (Gazzinelli et al., 1994), the overproduction of it may contribute to sepsis severity.

Robustness of systemic inflammation is the main factor that aggravates sepsis and predisposes patients to septic shock (Tumes et al., 2013). Our findings showed that previous exposure to T. gondii was a factor that intensified IFN-γ production and aggravated sepsis severity. Patients previously exposed to T. gondii had an increased mortality rate during sepsis compared to non-exposed patients. In this study, we proposed that serology for toxoplasmosis should be monitored in septic patients to predict sepsis severity. Because serology for toxoplasmosis is not time-consuming and laborious, it could be used to optimize the screening of sepsis severity.

In conclusion, we described that toxoplasmosis imprints intracellular signals that activate CD4<sup>+</sup> and CD8<sup>+</sup> T cells to produce IFN-γ. During infection, such cells are converted to memory-like T cells to maintain a pool of central memorylike CD4<sup>+</sup> T cells in secondary lymphoid organs. Our data suggest that during sepsis induction, central memory-like T cells specific to T. gondii are properly converted to effector memorylike T cells. Although their specificity against T. gondii is not impaired, such cells can cross-react against bacteria to control secondary bacterial infections. Nevertheless, the exacerbated systemic inflammatory response is deleterious to the host because it aggravates SIRS, leading to hypotension with consequent septic shock. Collectively, these data demonstrate that environmental features, such as previous chronic and non-lethal infections, may explain why sepsis has a broad spectrum of clinical forms.

### AUTHOR CONTRIBUTIONS

MCS, DF, AK, LB, JA, FC, RG, and JS designed the study. MCS, DF, AK, LB, GB, MD, WA, and MABS performed the mouse experiments. MCS, AG, and MB provided the samples and performed the patient's experiments. MCS, AK, and TM wrote the manuscript. MCS, LB, AK, RG, and JS edited the manuscript.

### FUNDING

This work was supported by the Coordination for the Improvement of Higher Education Personnel (CAPES) and the

### REFERENCES


National Counsel of Technological and Scientific Development (CNPq). "The research leading to these results received funding from the Sao Paulo Research Foundation (FAPESP) under grant agreement n 2013/08216-2 (Center for Research in Inflammatory Disease), from the University of Sao Paulo NAP-DIN under grant agreement no. 11.1.21625.01.0."

### ACKNOWLEDGMENTS

The authors would like to thank Cristiane Milanezi for technical support, as well as Vanessa Carregaro for providing anti-IFN-γ antibody.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fcimb. 2017.00116/full#supplementary-material


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

Copyright © 2017 Souza, Fonseca, Kanashiro, Benevides, Medina, Dias, Andrade, Bonfá, Silva, Gozzi, Borges, Gazzinelli, Alves-Filho, Cunha and Silva. 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.

# Behavior of Neutrophil Granulocytes during *Toxoplasma gondii* Infection in the Central Nervous System

Aindrila Biswas <sup>1</sup> , Timothy French<sup>1</sup> , Henning P. Düsedau<sup>1</sup> , Nancy Mueller <sup>1</sup> , Monika Riek-Burchardt <sup>2</sup> , Anne Dudeck <sup>2</sup> , Ute Bank <sup>2</sup> , Thomas Schüler 2 † and Ildiko Rita Dunay <sup>1</sup> \* †

1 Institute of Inflammation and Neurodegeneration, Otto-von-Guericke University Magdeburg, Magdeburg, Germany, 2 Institute for Molecular and Clinical Immunology, Otto-von-Guericke University Magdeburg, Magdeburg, Germany

Cerebral toxoplasmosis is characterized by activation of brain resident cells and recruitment of specific immune cell subsets from the periphery to the central nervous system (CNS). Our studies revealed that the rapidly invaded Ly6G<sup>+</sup> neutrophil granulocytes are an early non-lymphoid source of interferon-gamma (IFN-γ), the cytokine known to be the major mediator of host resistance to Toxoplasma gondii (T. gondii). Upon selective depletion of Ly6G<sup>+</sup> neutrophils, we detected reduced IFN-γ production and increased parasite burden in the CNS. Ablation of Ly6G<sup>+</sup> cells resulted in diminished recruitment of Ly6Chi monocytes into the CNS, indicating a pronounced interplay. Additionally, we identified infiltrated Ly6G<sup>+</sup> neutrophils to be a heterogeneous population. The Ly6G+CD62-LhiCXCR4<sup>+</sup> subset released cathelicidin-related antimicrobial peptide (CRAMP), which can promote monocyte dynamics. On the other hand, the Ly6G+CD62-LloCXCR4<sup>+</sup> subset produced IFN-γ to establish early inflammatory response. Collectively, our findings revealed that the recruited Ly6G+CXCR4<sup>+</sup> neutrophil granulocytes display a heterogeneity in the CNS with a repertoire of effector functions crucial in parasite control and immune regulation upon experimental cerebral toxoplasmosis.

## *Edited by:*

Anton Aebischer, Robert Koch-Institute, Germany

#### *Reviewed by:*

Emma Harriet Wilson, University of California, Riverside, United States Michal Adam Olszewski, University of Michigan, United States

#### *\*Correspondence:*

Ildiko Rita Dunay Ildikodunay@gmail.com † These authors have contributed equally to this work.

*Received:* 23 February 2017 *Accepted:* 02 June 2017 *Published:* 21 June 2017

#### *Citation:*

Biswas A, French T, Düsedau HP, Mueller N, Riek-Burchardt M, Dudeck A, Bank U, Schüler T and Dunay IR (2017) Behavior of Neutrophil Granulocytes during Toxoplasma gondii Infection in the Central Nervous System. Front. Cell. Infect. Microbiol. 7:259. doi: 10.3389/fcimb.2017.00259 Keywords: *Toxoplasma gondii*, neutrophil granulocytes, cerebral toxoplasmosis, neuroinflammation, neutrophil Infiltration

## INTRODUCTION

T. gondii is a highly successful parasite capable of crossing most biological barriers of the body (Barragan, 2002; Harker et al., 2015). The parasite can infect migratory immune cells such as dendritic cells (DCs), macrophages and neutrophil granulocytes to enter immunoprivileged sites (Barragan, 2002; Da Gama et al., 2004; Courret et al., 2006; Lambert et al., 2006). During the acute phase of infection, enterocytes secrete chemokines including monocyte chemotactic protein-1 (MCP-1) and macrophage inflammatory protein-2 (MIP-2), which recruit neutrophil granulocytes, inflammatory monocytes and DCs to the site of infection (Robben et al., 2005; Courret et al., 2006; Pittman and Knoll, 2015). We have previously detected that the CD8α <sup>+</sup> DC subset is the crucial source of interleukin-12 (IL-12), the cytokine that stimulates natural killer (NK) cells and T cells to produce the cytokine IFN-γ (Mashayekhi et al., 2011). IFN-γ is a major mediator of host resistance to T. gondii, regulating a plethora of intracellular mechanisms to eliminate the parasite (Gazzinelli et al., 1993; Hunter et al., 1994; Kang and Suzuki, 2001; Kim and Weiss, 2004; Gazzinelli and Sher, 2014). Recent studies revealed that neutrophil granulocytes, besides innate lymphoid cells, NK cells and T cells, are able to produce IFN-γ (Sturge et al., 2013). They also produce IFN-γ independently from toll-like receptor (TLR)11-induced IL-12 during acute toxoplasmosis in the periphery (Sturge et al., 2013, 2015). Our previous experiments provided evidence that unlike inflammatory monocytes, neutrophil granulocytes worsened ileal pathology upon specific depletion in the acute phase of toxoplasmosis (Dunay et al., 2010). These studies were carried out in mice, which had an intact TLR-recognition system responsible for IL-12 dependent IFN-γ production by NK and T cells. However, the relative contribution of neutrophil granulocytes to IFN-γ production was not elucidated.

The parasites can enter the CNS within immune cells, initially infecting endothelial cells of the blood-brain barrier, where they egress, and invade brain resident cells (Konradt et al., 2016). Cyst formation within neurons undergoing a stress-mediated response is followed by ongoing basal inflammation, establishing a chronic persistent infection (Parlog et al., 2014; Blanchard et al., 2015). Chronic cerebral toxoplasmosis is characterized by the activation of resident cells such as microglia and astrocytes, which carry out distinct antiparasitic functions (Strack et al., 2002; Schluter et al., 2003; Drögemüller et al., 2008; Kamerkar and Davis, 2012; Cabral et al., 2016). In addition, peripheral immune cell subsets of myeloid and lymphoid origin are recruited to the inflamed CNS (John et al., 2011). Recently, we characterized the behavior of infiltrating CD11b+Ly6Chi myeloid derived cells, which carry out a pivotal role upon cerebral toxoplasmosis (Möhle et al., 2014, 2016; Biswas et al., 2015). Several studies have highlighted the role of brain-resident cells, lymphocytes, DCs, and macrophages (Strack et al., 2002; John et al., 2011), however, the function of neutrophil granulocytes in host defense in the CNS is still uncertain.

Neutrophil granulocytes differentiating from the common myeloid progenitor in the bone marrow (BM) form the first line of defense during infection and inflammation (Soehnlein et al., 2008b). Upon reaching the site of infection, neutrophils can execute many immune functions including production of reactive oxygen species (ROS) or proinflammatory modulators (Denkers et al., 2012; Bardoel et al., 2014; Perez-de-Puig et al., 2015). The surface marker Ly6G, exclusively expressed on neutrophil granulocytes, facilitates distinction from monocytes, enabling specific functional studies. The role of neutrophil granulocytes in the periphery and the CNS has been investigated in multiple neurodegenerative diseases and infections, however, their impact on cerebral toxoplasmosis remains to be elucidated (Zhou et al., 2003; Zehntner et al., 2005; Mildner et al., 2008; Ransohoff and Engelhardt, 2012; Sturge et al., 2013; Wantha et al., 2013; Marais et al., 2014).

Here we characterized the phenotype and behavior of Ly6G<sup>+</sup> neutrophil granulocytes in chronic T. gondii infection. We detected their influx into the CNS during the acute stage of cerebral toxoplasmosis, which was associated with the upregulation of expression of certain activation markers and co-stimulatory molecules. Importantly, intracellular cytokine analysis revealed that neutrophil granulocytes form a non-lymphoid source of the pro-inflammatory cytokine IFN-γ at the acute stage of infection. Upon using a specific anti-granulocyte monoclonal antibody (mAb) 1A8, we measured a significant increase in the parasite burden. We also detected a reduction in monocyte recruitment and IFN-γ production confirming the specific contribution of neutrophil granulocytes. Neutrophils exhibited heterogeneity based on specific expression of the adhesion molecule CD62-L and the chemokine receptor CXCR4, respectively. This implies that multiple signals at the site of the inflammation can influence the functional characteristics and phenotype of neutrophils.

## MATERIALS AND METHODS

### Animals

Age and sex matched C57BL/6 wild-type (WT) mice, obtained from Janvier (Cedex, France) were used. All animal care was in accordance with institutional guidelines. Food and water were available ad libitum. Experiments were performed in accordance to German and European legislation.

### Infection

T. gondii cysts of type II strain ME 49 were harvested from the brains of female NMRI mice infected intra-peritoneally (i.p.) with T. gondii cysts 8–10 months earlier. Brains obtained from infected mice were mechanically homogenized in 1 ml sterile phosphate-buffered saline (PBS). Cyst numbers were counted in a 10µl brain suspension using a light microscope. Two cysts were administered i.p. in a total volume of 200µl per mouse as described before (Möhle et al., 2016). Control mice were mockinfected with sterile PBS. The mice were perfused intracardially with 60 ml sterile ice-cold PBS. The brains were removed for further analysis. The acute stage of infection was defined between day 10 and day 14 whereas the chronic stage of infection was starting at day 21. Therefore, to investigate the recruited neutrophil granulocytes in chronic cerebral toxoplasmosis, mice were sacrificed 4 weeks post-infection for further analysis.

### 1A8 mAb Treatment

Depletion of neutrophils was performed by i.p. administration of anti-Ly6G mAb (clone 1A8, BioXCell). Mice were injected with 500µg (as described before by Daley et al., 2008; Dunay et al., 2010) of the antibody i.p. on alternate days from days 10 to 23 post-infection. Mice were sacrificed 24 h after the last antibody treatment. Rat IgG2a (BioXCell) was used as a control to mAb.

### Cell Isolation

Brains were homogenized in a buffer containing 1M HEPES (pH 7.3) and 45% glucose and then sieved through a 70µm strainer. The cell suspension was washed and fractionated on 25–75% Percoll gradient (GE Healthcare). Isolated cells were washed with PBS and used immediately for further experiments. Peripheral blood was obtained from posterior vena cava and lysed with erythrocyte (RBC) lysis buffer (eBioscience). Subsequently, these cells were stained with the desired fluorescent conjugated antibodies (Biswas et al., 2015).

### Flow Cytometry

Isolated mononuclear cells were incubated with Zombie NIRTM or VioletTM fixable dye (Biolegend) for live/dead discrimination. Unspecific antibody binding was blocked by incubation with anti-FcγIII/II receptor antibody (clone 93). Thereafter, cells were stained with fluorochrome-conjugated antibodies against cell surface markers: CD45 (30-F11), CD11b (M1/70), Ly6C (HK1.4), MHC I H-2D<sup>b</sup> (28-14-8), MHC II I-A/I-E (M5/114.15.2), CD80 (16-10A1), and CD86 (GL1), all purchased from eBioscience, CXCR2 (SA04E1), CXCR4 (L27GF12), and CD64 (X54-5/7.1) from Biolegend, Ly6G (1A8), CD62-L (MEL-14) from BD Bioscience, then washed and fixed in 4% paraformaldehyde. Matched isotype controls were used to assess the level of unspecific binding.

For intracellular cytokine staining, single-cell suspensions (5 × 10<sup>5</sup> cells/well) were stimulated in 96-well plates in the presence of Toxoplasma lysate antigen (5µg/ml) and Brefeldin A (10µg/ml, GolgiPlug, BD Biosciences). After 6 h, cells were incubated with Zombie NIRTM or VioletTM fixable dye and anti-FcγIII/II receptor antibody (clone 93) and surface stained for CD45 (30-F11), CD11b (M1/70), Ly6G (1A8), Ly6C (HK1.4), CRAMP (R-170), washed in FACS buffer [PBS with 1% of fetal calf serum (FCS)] and fixed in 4% paraformaldehyde. Cells were permeabilized using Permeabilization Buffer (Biolegend). To measure the cytokine expression, the staining was performed with the following antibodies: IL-1α (ALF-161), TNF (MP6- XT22), NOS2 (CXNFT), IFN-γ (XMG1.2), IL-1β (NJTEN3), and IL-10 (JES5-16E3) from eBioscience in Permeabilization Buffer (Biolegend) (Biswas et al., 2015).

A total of 100,000 cells was acquired using a flow cytometer (BD FACS Canto II). Data were analyzed using FlowJo software (Version 10 Tree Star). Matched isotype controls were used to assess the level of unspecific binding.

### Detection of Reactive Oxygen Species

Isolated cells were stained for CD45 (30-F11), CD11b (M1/70), Ly6G (1A8), Ly6C (HK1.4), CXCR4 (L27GF12), and CD62-L (MEL-14) in FACS buffer for 30 min after blocking FcγRs. ROS production was measured by Total ROS Detection Kit (ENZO, 51011), according to the manufacturer's instructions.

### q- and qRT-PCR

RNA and DNA was isolated from the right hemisphere of the infected mice as previously described (Möhle et al., 2016). Following the isolation of the nucleic acids, semi-quantitative PCR was performed to measure the parasite burden and the cytokine gene expression levels in the brain.

### Immunofluorescence

Mice were anesthetized with isoflurane and perfused intracardially with saline followed by paraformaldehyde (PFA, 4%) in phosphate buffer (pH 7.4). The brain was removed, post-fixed with PFA overnight, cryoprotected in 30% sucrose, frozen, and 20-µm-thick sections were prepared in a cryostat. The sections were fixed with ethanol, blocked with normal serum and incubated overnight at 4◦C with the primary antibody SAG-1 (D61S clone; 1:20, Invitrogen) or PECAM-1/CD31 (MEC13.3 clone; 1:500) and Ly6G (1A8 clone; 1:200). Following the primary antibody staining, lectin staining was performed (Fluorescein labeled Dolichos Biflorus Agglutinin (DBA); 1:500, Vector Laboratories) according to the manufacturer's guidelines. Then, sections were incubated for 2 h at room temperature with secondary antibodies (Alexa Fluor-488, Thermo Fischer Scientific). Controls were carried out by omission of the primary antibodies. Sections were counterstained with Sytox Red Dead Cell Stain (1:20,000, Thermo Fischer Scientific) to visualize the cell nuclei and slides were observed under a confocal laser microscope (LSM510, Carl Zeiss).

### Statistical Analysis

Data were analyzed by Mann-Whitney test for two groups or one-way ANOVA for several groups followed by Tukey's posttest with GraphPad Prism 6 (San Diego, CA). In all cases, results were presented as mean ± standard deviation (SD) and were considered significant, with p < 0.05.

## RESULTS

### Rapid Influx of Neutrophil Granulocytes into the Brain upon Cerebral Toxoplasmosis

Infection of mice with T. gondii induces egress of Ly6Chi monocytes and Ly6G<sup>+</sup> neutrophil granulocytes from the BM to the blood. Following our previous characterization of BM derived monocytes in the CNS in 4 weeks T. gondii-infected mice, we analyzed neutrophil dynamics using the same experimental model. CD11b and Ly6C expression were used to identify the inflammatory monocytes (Ly6Chi), neutrophils (Ly6Cint) and resident monocytes (Ly6Clo) in the blood (**Figures 1A,C**). The CD11b<sup>+</sup> cells were further differentiated based on their Ly6C and Ly6G expression (**Figures 1B,D**). The neutrophil specific Ab Ly6G (1A8) was applied to distinguish neutrophil granulocytes (Ly6G+) from monocytes (Ly6G−) (**Figures 1B,D**). We observed increased percentages of circulating neutrophil granulocytes in peripheral blood of infected mice (30 ± 3.06%) as compared to non-infected controls (15.7 ± 1%; **Figures 1B,D**).

Alongside the enhanced egress of immune cells in the periphery, cerebral toxoplasmosis leads to the activation of brain resident cells and infiltration of circulating immune cells to the CNS (Möhle et al., 2014; Biswas et al., 2015). Peripheral immune cell influx into the infected brain is characterized by initial invasion of neutrophils, followed by the recruitment of BM-derived monocytes and the lymphocytes. While in non-infected controls the main immune cell population was resting resident microglia (CD45loCD11b+; **Figures 1E,F**), we observed the entry of CD45hi myeloid derived population in the brains of T. gondii-infected mice (**Figure 1G**). The recruited cells included two populations, CD45hiCD11b<sup>−</sup> cells (ungated; **Figure 1G**) and CD45hiCD11b<sup>+</sup> cells (upper gate; 10.0 ± 2.1% of the parent population). The CD45hiCD11b<sup>+</sup> cells comprised of BM-derived myeloid cells, namely monocytes, neutrophil granulocytes, macrophages and DCs. Upon infection,

brain resident activated microglia cells expressed elevated CD45 levels (CD45intCD11b+; 5.0 ± 1.06% of the parent population). The anti-Ly6G (1A8) antibody was applied to distinguish Ly6G<sup>+</sup> neutrophil granulocytes (**Figure 1H**; 3.0 ± 1.02% of the CD45hiCD11b+) from monocytes. Next, the total number of mononuclear cells was quantified in the brain during chronic cerebral toxoplasmosis (**Figure 1I**). Thus, CD45hiCD11b+Ly6G<sup>+</sup> neutrophil granulocytes (6 × 10<sup>3</sup> ± 50) formed a small yet defined population, when compared to the absolute numbers of CD45intCD11b<sup>+</sup> microglia (1 × 10<sup>5</sup> ± 2,000). These data demonstrate that cerebral toxoplasmosis leads to the activation of resident microglia and the recruitment of lymphoid and myeloid immune cells including neutrophil granulocytes.

### Localization of Neutrophil Granulocytes in the CNS during the Course of Cerebral Toxoplasmosis

Neutrophil granulocytes, the earliest immune cells to arrive at the site of infection, can deploy a plethora of mechanisms to attack invading pathogens (Denkers et al., 2012; Bardoel et al., 2014; Perez-de-Puig et al., 2015). To study their localization around T. gondii in the CNS, we performed immunofluorescence analysis of mice brain slices at the acute stage and chronic stage of the infection. We observed that neutrophil granulocytes were not only in the blood vessels, marked by PECAM1, but entered the brain parenchyma (**Figure 2A**). The recruited neutrophils were in close proximity to the T. gondii tachyzoites in the cerebral cortex during the acute stage infection (**Figure 2B**). Upon encystation of the T. gondii parasites in the chronic phase, we observed the absence of neutrophils in the vicinity of bradyzoite containing cysts (**Figure 2C**). These immunofluorescence results indicate the affinity of the neutrophil granulocytes to the infective stage of T. gondii in contrast to the dormant stage during the course of the cerebral toxoplasmosis.

### Phenotypic Analysis of Infiltrating Neutrophils Upon Cerebral *T. gondii* Infection

To further characterize neutrophil granulocytes in inflamed brains upon 4 weeks of T. gondii infection, the expression profiles of various cell surface markers were determined by flow cytometry. As control, microglia were selected, because under homeostatic condition neutrophils are not present in the CNS, and immune cells from the same isolate chosen for the comparison. A large percentage of infiltrating neutrophils expressed enhanced levels of MHC I and MHC II (83 ± 1% MHC I; 95 ± 0.5% MHC II) (**Figures 3M–P**) unlike in the periphery (25 ± 4% MHC I; 3 ± 0.5% MHC II) (**Figures 3A–D**). According to their function as antigen presenting cells (APCs) of the brain, activated microglia also expressed significant levels of MHC I and MHC II (100% MHC I; 100% MHC II) (**Figures 3M–P**). We further compared the expression of the co-stimulatory molecules CD80 and CD86 and detected these on small parts of the neutrophil population (15 ± 0.5% CD80; 17 ± 0.5% CD86) and activated microglia (18 ± 1% CD80; 52 ± 1% CD86) (**Figures 3Q–T**). The phagocytosis mediating receptor CD64 (FcgR1), which was present on a large proportion of activated microglia (89 ± 1% CD64), was only detectable on a small percentage of neutrophils (11 ± 0.5%) (**Figures 3U,V**). The chemokine receptor CXCR2 was detected on more than half of the neutrophils (65 ± 0.5%). However, only a small percentage of activated microglia expressed CXCR2 (5 ± 0.5%)

(**Figures 3W,X**). The neutrophil granulocytes in the periphery did not express any significant levels of CD80, CD86 and CD64 (**Figures 3E–J**). However, the entire neutrophil population expressed CXCR2 (**Figures 3K,L**). Hence, activated microglia and neutrophils express immune modulatory molecules in infected brains indicating their contribution to the establishment of local immunity in response to T. gondii infection.

### Cytokine Production of Infiltrating Neutrophils upon Cerebral *T. gondii* Infection

Several studies have demonstrated that neutrophil granulocytes secrete plenty of immune modulatory molecules including cytokines upon infection (Sturge et al., 2013; Bardoel et al., 2014). To determine their cytokine expression profile, in experimental cerebral toxoplasmosis (4 weeks T. gondii infection), myeloid cells from the infected brain were analyzed by intracellular flow cytometry. We observed that neutrophil granulocytes hardly produced pro-inflammatory cytokines such as IL-1α and TNF (3 ± 1% and 5 ± 0.5%, respectively). On the contrary, microglia served as a considerable source of both cytokines (50 ± 3% and 80 ± 1%, respectively) (**Figures 4A,B,E,F**). Importantly, 65 ± 0.5% of neutrophil granulocytes synthesized IL-1β (**Figures 4C,D**). Interestingly, IFN-γ was also produced by the neutrophils (85 ± 0.5%) contrary to activated microglia (**Figures 4I,J**). IL-10 and iNOS were synthesized by negligible numbers of neutrophil granulocytes (0 ± 0.5% IL-10; 2 ± 0.5% iNOS) and microglial cells (0 ± 0.5% IL-10; 8 ± 1% iNOS) (**Figures 4G,H,K,L**). Besides, all neutrophils synthesized ROS unlike activated microglia (95 ± 0.5%) (**Figures 4M,N**). In summary, these data demonstrate that neutrophil granulocytes produce multiple immune modulatory molecules including IFN-γ.

### IFN-γ Production by Neutrophil Granulocytes over the Course of Cerebral Toxoplasmosis

IFN-γ plays a critical role in the host response to cerebral toxoplasmosis. Previous studies defined NK cells, T cells and microglia as sources of IFN-γ during cerebral toxoplasmosis

FIGURE 3 | Phenotypic analysis of neutrophil granulocytes and activated microglia (A–X). Expression of activation markers and chemokine receptors in mice blood and brains after 4 weeks of T. gondii infection, were analyzed by flow cytometry, respectively. The cells were gated as described and shown in the representative plots of Figure 1. Neutrophils from the infected and non-infected blood (CD11b+CD45hiLy6G+Ly6C+), activated microglia (CD11b+CD45int), and neutrophil granulocytes from the infected brain (CD11b+CD45hiLy6G+Ly6C+), were assessed for their relative expression of the indicated molecules. (A,C,E,G,I,K,M,O,Q,S,U,W) Histograms show the representative expression level of the surface maker by the cell population in comparison to the corresponding isotype control (light gray without tint or light gray tinted). Bars mark cells positively expressing particular surface markers and numbers above bars represent the percentage of cells in the respective population: neutrophils in non-infected blood (CD11b+Ly6G+) (without any tint, dotted line), neutrophils in infected blood (CD11b+Ly6G+) (tinted, dotted line), activated microglia (CD11b+CD45int) (without any tint), neutrophil granulocytes (CD11b+Ly6G+) (tinted). (B,D,F,H,J,L,N,P,R,T,V,X) Bar graphs represent the median fluorescence intensity (MFI) for the specific marker MFI ± SD (n = 4) (neutrophils in non-infected blood: white bars; neutrophils in infected blood: gray bars; activated microglia: white bars; and neutrophils: black bars). Data represent 2 independent experiments with 5 mice per experiment. Mann-Whitney test was performed for comparisons (\*p < 0.05).

(Gazzinelli et al., 1993; Hunter et al., 1994; Gavrilescu et al., 2004; Kim et al., 2012; Gazzinelli and Sher, 2014). However, the contribution of neutrophil granulocytes to IFN-γ production in the brain has not been elucidated (Suzuki, 2002; Sa et al., 2015).Therefore, we performed flow cytometry from the brains of mice during the acute (2 weeks) as well as the chronic stage (4 weeks) of T. gondii infection. In the acute phase, 83.5 ± 0.5% of neutrophil granulocytes produced IFN-γ (1.5 <sup>∗</sup> 10 <sup>∧</sup> 3 ± 200 cells). This was the case for only 47 ± 0.5% of activated microglia. At this time-point, only negligible numbers of CD11b<sup>−</sup> lymphocytes (247 ± 50 cells) were detectable in the brain to make a significant contribution (6 ± 0.5%) (**Figures 5A,B**). However, during the chronic phase of infection enhanced number of lymphocytes infiltrated the brain, which then became the major IFN-γ producers (88 ± 0.5%) (1.5 <sup>∗</sup> 10 <sup>∧</sup> 6 ± 550 cells) (**Figures 5C–F**). Neutrophils were secondary producers of IFN-γ at the chronic stage of infection (85 ± 4%) (1.4 <sup>∗</sup> 10 <sup>∧</sup> 3 ± 60 cells) (**Figures 5C–F**). Thus, our results demonstrate that Ly6G<sup>+</sup> neutrophil granulocytes are representing an important early source of early IFN-γ in the acute phase of cerebral toxoplasmosis.

### Depletion of Neutrophil Granulocytes in Experimental Cerebral Toxoplasmosis

It is still unclear whether neutrophils contribute to control of cerebral toxoplasmosis. To investigate this, we took advantage of the previously described specific depleting anti-Ly6G mAb (clone 1A8). Following our observation that Ly6G<sup>+</sup> neutrophils are early IFN-γ producers in the acute phase of cerebral toxoplasmosis, we started the ablation at day 10 post-infection when neutrophil granulocytes began entering the brain. We continued the ablation, injecting Ly6G mAb or control IgG2 mAb every alternate day, as the infection progressed from the acute to chronic stage. Twenty-four hours after the last Ab treatment,

mice were sacrificed and the successful depletion of CD11b<sup>+</sup> Ly6G<sup>+</sup> monocytes in the blood was confirmed (7.0 ± 0.13 to 0%; **Figures 6A,B**).

In agreement with this, anti-Ly6G treatment of infected mice also resulted in a significant reduction of neutrophils in the brain. Further characterization revealed a reduction of recruited myeloid cells (24.0 ± 1.5 to 19.0 ± 0.28%) in the brains of anti-Ly6G treated infected mice (**Figure 6C**, upper panel). This observation was further confirmed with a significant decrease of Ly6G<sup>+</sup> neutrophils (**Figure 6C**, middle panel; 3.0 ± 0.5% to 0.6 ± 0.3%). Further investigation revealed reduced recruitment of CD11b+Ly6Chi inflammatory monocytes (**Figure 6C**, lower panel; 48.0 ± 3.0% to 32.0 ± 1.3%) in the brains of anti-Ly6G treated mice. Moreover, absolute cell numbers of CD45hi CD11b<sup>+</sup> myeloid cells (p < 0.05), Ly6G<sup>+</sup> neutrophils (p < 0.001) and Ly6Chi monocytes (p < 0.01) were significantly reduced. On the contrary, numbers of T lymphocytes (CD45hiCD11b−CD4<sup>+</sup> and CD45hiCD11b−CD8+) and microglia (CD45intCD11b+) remained unaltered (**Figures 6D,E**).

Using q and qRT-PCR we investigated whether the ablation of Ly6G<sup>+</sup> neutrophil granulocytes affected parasite burden and cytokine gene expression levels in brains of anti-Ly6G treated mice. We found that anti-Ly6G treatment was associated with elevated parasite burden (**Figure 7A**). Correspondingly, we observed a 40 and 60% reduction of pro-inflammatory TNF and IFN-γ, respectively, as compared to IgG treated controls (**Figures 7C,E**). Furthermore, IL-10 and IL-1β mRNA levels were reduced in anti-Ly6G treated mice (**Figure 7B,D**). Together, these results demonstrate that increased parasite burden correlates inversely with the reduced levels of TNF, IFNγ, IL-1 β, and IL-10 in brains of neutrophil-depleted mice with cerebral toxoplasmosis. Hence, the recruitment of neutrophil granulocytes and their early cytokine production appear to promote inflammatory immune responses restricting pathogen load in brains of T. gondii-infected mice.

### Heterogeneity of Neutrophil Granulocyte Subsets in Cerebral Toxoplasmosis

Depending on the infection model, different subsets of neutrophil granulocytes can be discriminated based on their phenotypic and functional properties (Tsuda et al., 2004; Pillay et al., 2010; Beyrau et al., 2012). However, in the case of cerebral toxoplasmosis, it was largely unclear whether recruited neutrophils represent a heterogeneous or rather homogeneous population in the brain. Therefore, CD45+CD11b+Ly6C+Ly6G<sup>+</sup> neutrophil granulocytes from 4 weeks T. gondii-infected mice were analyzed by flow cytometry (**Figures 8A,B**). We found that this population contained cells expressing different levels of CD62-L (**Figure 8C**). Contrary to CD62-Llo neutrophils (20 ± 0.5%), a high percentage of CD62-Lhi neutrophils (50 ± 0.5%) expressed

FIGURE 5 | IFN-γ production over the course of cerebral toxoplasmosis (A–F). Cells isolated from brains of mice infected with 2 and 4 weeks of T. gondii, were re-stimulated with Toxoplasma lysate antigen in vitro and analyzed by flow cytometry. The cells were gated as shown in Figures 1E,F. (A,C) The fraction of the total cell population expressing IFN-γ was plotted in stacked bar graphs. (B,D) Bar graphs represent the MFI of the respective fluorochrome for a particular cytokine, MFI ± SD (n = 4). (E,F) The representative plots show the gating strategy of Ly6G<sup>+</sup> neutrophils and CD11b−CD45hiLy6G<sup>−</sup> cells producing IFN-γ. The quadrant was set on isotype control. Data are representative of 2 independent experiments with 5 mice per experiment. Significant differences (\*p < 0.05, \*\*p < 0.01) were determined using the Mann-Whitney test. White bars, black bars and gray bars represent CD45intCD11b<sup>+</sup> activated microglia, Ly6G<sup>+</sup> neutrophils and CD45hiCD11b<sup>−</sup> lymphocytes, respectively.

CRAMP (**Figures 8D,E**). On the contrary, only 25 ± 0.5% of CD62-Lhi neutrophils secreted IFN-γ while this was the case for 45 ± 0.5% of CD62-Llo neutrophils (**Figures 8F,G**). However, CD62-Lhi and CD62-Llo neutrophil granulocytes produced similar levels of ROS (90 ± 0.5% vs. 95 ± 0.5%) (**Figures 8H,I**). These data suggest the phenotypic heterogeneity of brain-derived neutrophil granulocytes further defining CD62-Llo neutrophils as a source of protective IFN-γ.

### DISCUSSION

Neutrophil granulocytes are the first immune cells recruited from the periphery to secrete initial effector molecules upon injury and infection. They play a critical role in infection to eliminate pathogens via multiple mechanisms (Zhou et al., 2003; Nathan, 2006; Kolaczkowska and Kubes, 2013). The response of neutrophil granulocytes to parasitic infection is well described in the periphery, but their role within the CNS in chronic infections is not thoroughly defined. One previous study concluded that neutrophil granulocytes are the limiting factor against uncontrolled tachyzoite replication in cerebral toxoplasmosis (Bliss et al., 2001), but this conclusion was based on depletion of Gr1<sup>+</sup> cells. However, later it turned out that Gr1 expression is not restricted to neutrophil granulocytes but also expressed by monocytes (Daley et al., 2008).

We defined the phenotype and function of Ly6G<sup>+</sup> neutrophil granulocytes in the course of chronic toxoplasmosis in the CNS. Initially, we observed increased percentages of Ly6G<sup>+</sup> neutrophils in the blood of T. gondii-infected mice. In the chronic stage of the infection, brain resident microglia cells displayed an activated phenotype, and peripheral immune cells including CD11b<sup>−</sup> lymphoid and CD11b<sup>+</sup> myeloid cells infiltrated the brain. In line with previous reports, characterization of brain immune cells revealed that recruited CD11b<sup>+</sup> myeloid cells contained Ly6G<sup>−</sup> inflammatory monocytes and Ly6G<sup>+</sup> neutrophil granulocytes (Möhle et al., 2014; Biswas et al., 2015). Despite low absolute numbers of neutrophils in the chronic stage, they form a substantial portion of the myeloid cell compartment in the CNS in the acute stage of the infection. Confocal microscopic analysis identified neutrophil granulocytes adjacent to T. gondii tachyzoites in the acute infection, albeit they were distant from cysts in the chronic stage. This observation suggests that the active infective form of the parasite could potentially trigger the host defense and effector functions of neutrophil granulocytes such as generation of ROS or even phagocytosis as it has been shown in acute toxoplasma infection in the periphery (Abdallah et al., 2011).

We measured MHC I and MHC II expression on both microglia and neutrophil granulocytes. This indicates that neutrophils can acquire an APC phenotype in cerebral toxoplasmosis. Neutrophil granulocytes are generally regarded as professional phagocytes responding early to tissue infection and injury. In line with our observations, increasing evidence suggests that neutrophils can also modulate adaptive immune responses by activating CD4<sup>+</sup> T cells in vitro via upregulation of MHC class molecules (Wagner and Hug, 2005; Abdallah et al., 2011). On the contrary, the co-stimulatory molecule CD86 and the phagocytosis-related receptor CD64 (FcγR1) were detected primarily on activated microglia as opposed to neutrophils. The chemokine receptor CXCR2, typically expressed by neutrophil

granulocytes, was down-regulated upon entering the brain (Liu et al., 2010). Despite being also expressed on oligodendrocytes and neurons (Liu et al., 2010), where it mediates a wide range of functions, CXCR2 is not present on the activated microglia. Similarly, activated microglia do not express CD62-L which is crucial for leukocyte rolling, transmigration and accumulation at sites of inflammation (Rainer, 2002; Biswas et al., 2015). We identify the existence of two subsets of neutrophils in the infected brain, which can be discriminated based on their differential CD62-L expression (Zenaro et al., 2015). CXCL12, the ligand of CXCR4 is constitutively expressed on endothelial cells in the CNS (McCandless et al., 2006; Wilson et al., 2010). We showed that cerebral toxoplasmosis leads to increased expression of CXCR4 on neutrophil granulocytes. In conclusion, we demonstrate that CNS infiltrating neutrophils display distinct phenotypes in cerebral T. gondii infection.

Following surface characterization, we studied the secretion of certain inflammatory mediators by Ly6G<sup>+</sup> neutrophil granulocytes and activated microglia. The proinflammatory molecules IL-1β and ROS were primarily produced by

neutrophils in line with previous studies (Bardoel et al., 2014). Most importantly, we detected that the cytokine IFN-γ, which is the main driving factor of the host immune response against T. gondii, was secreted by neutrophil granulocytes. Neutrophils can produce IFN-γ in Nocardia asteroides-infected lungs (Ellis and Beaman, 2002) and Salmonella typhimurium-infected intestines (Kirby et al., 2002; Sturge et al., 2013). Our observation are in contrast to Sa et al. (2015) who reported that microglia are the main source of myeloid cell-derived IFN-γ in a reactivation model of cerebral toxoplasmosis in RAG−/−and IFN-γ <sup>−</sup>/<sup>−</sup> mice with a Balb/c background. However, it is important to stress that our results were obtained in WT C57BL/6 mice, which may explain the disparity. Interestingly, our observation is in line with Sturge et al. (2015) who reported that the neutrophils store IFN-γ at the promyelocyte stage in the absence of inflammation. Additionally, they showed that during acute toxoplasmosis neutrophils form an important non-lymphoid source of IFNproduction (Sturge et al., 2013). Importantly, we reveal that infiltrating neutrophil granulocytes are an important source of myeloid cell-derived IFN-γ particularly in the early phase of cerebral toxoplasmosis.

The ablation of Ly6G<sup>+</sup> neutrophils in the periphery resulted in a reduced infiltration to the CNS. The depletion was also associated with a significant reduction of IFN-γ mRNA levels in infected brains. IFN-γ is crucial for survival during T. gondii infection (Gazzinelli et al., 1993; Hunter et al., 1994; Gavrilescu et al., 2004; Kim et al., 2012; Gazzinelli and Sher, 2014). Increased parasite loads after anti-Ly6G treatment may have been a direct consequence of the antibody-mediated elimination of IFN-γ-producing neutrophils. Nonetheless, we cannot exclude indirect effects such as those resulting from impaired recruitment of other immune cells. It is important to mention that the neutrophil granulocytes ablation had no effect on the recruitment of CD4<sup>+</sup> and CD8<sup>+</sup> T lymphocytes to the brain. This suggests non-significant interaction with the T lymphocytes in terms of recruitment, however T cell activation may have been affected as it not part of the investigation. One study reported that ablation of neutrophils resulted in impaired maturation of resident microglia and brain macrophages into professional APCs, altering the T cell function (Steinbach et al., 2013). In contrary, we did not detect impaired differentiation or activation status of microglia into MHC II expressing professional CD11c<sup>+</sup> APCs (data not shown).

We previously showed that Ly6Chi monocytes express TNF and IL-10 in the CNS during cerebral toxoplasmosis (Biswas et al., 2015). Thus, the reduction of TNF and IL-10 levels after neutrophil depletion may be due to the reduced recruitment of Ly6Chi monocytes. However, the reduced IFN-γ and IL-1β levels were most likely caused by the complete ablation of Ly6G<sup>+</sup> neutrophil granulocytes as there was no change in the lymphocyte compartment.

In the current study, based on CXCR4 expression, neutrophil granulocytes displayed certain heterogeneity. Ly6G<sup>+</sup> neutrophils comprised of CD62-LhiCXCR4<sup>+</sup> and CD62-LloCXCR4<sup>+</sup> subsets. On one hand the Ly6G+CD62-LhiCXCR4<sup>+</sup> subset expressed higher levels of CRAMP and produced lower levels of IFNγ, but on the other hand Ly6G+CD62-LloCXCR4<sup>+</sup> neutrophil granulocytes expressed lower levels of CRAMP and produced higher levels of IFN-γ. Differential expression of CRAMP and IFN-γ suggest functional heterogeneity of Ly6G<sup>+</sup> neutrophil granulocytes. One subset Ly6G+CD62-LhiCXCR4<sup>+</sup> with high CRAMP expression may promote monocyte recruitment. The Ly6G<sup>+</sup> neutrophil granulocytes have been described to influence the recruitment of monocytes in injury models (Zhou et al., 2003; Soehnlein et al., 2008a,b). The other subset Ly6G+CD62-LloCXCR4<sup>+</sup> with higher IFN-γ production may establish the inflammatory response against cerebral toxoplasmosis.

Here we report that cerebral toxoplasmosis leads to infiltration of Ly6G<sup>+</sup> neutrophil granulocytes to the CNS. These neutrophils contribute to IFN-γ production during the early stages of the developing neuroinflammation. Importantly, we

identified two distinct Ly6G+CXCR4<sup>+</sup> neutrophil granulocyte subsets based on their CD62-L and CRAMP expression. Moreover, we detected a neutrophil-dependent recruitment of Ly6Chi monocytes to the CNS in chronic T. gondii infection. In summary, our findings suggest that neutrophil granulocytes perform important functions to promote host defense and exhibit heterogeneity in experimental cerebral toxoplasmosis.

### AUTHOR CONTRIBUTIONS

ID designed the experiments and supervised the project. AB, TF, HD, NM, MR, AD, UB, performed the experiments. AB, AD, UB, ID, and TS interpreted the data. AB, ID, and TS contributed to manuscript preparation. ID funded the project.

### FUNDING

This study was supported by the Deutsche Forschungsgemeinschaft (SFB 854, TP25, DU1112-5-1) to ID. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

### ACKNOWLEDGMENTS

We thank Dana Zabler and Dr. Sarah Abidat Schneider for excellent technical assistance.

### 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 Biswas, French, Düsedau, Mueller, Riek-Burchardt, Dudeck, Bank, Schüler and Dunay. 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.

# Th2/1 Hybrid Cells Occurring in Murine and Human Strongyloidiasis Share Effector Functions of Th1 Cells

Cristin N. Bock <sup>1</sup> , Subash Babu2, 3, Minka Breloer <sup>4</sup> , Anuradha Rajamanickam<sup>2</sup> , Yukhti Boothra<sup>2</sup> , Marie-Luise Brunn<sup>4</sup> , Anja A. Kühl <sup>5</sup> , Roswitha Merle<sup>6</sup> , Max Löhning7, 8 , Susanne Hartmann<sup>1</sup> and Sebastian Rausch<sup>1</sup> \*

<sup>1</sup> Department of Veterinary Medicine, Institute of Immunology, Freie Universität Berlin, Berlin, Germany, <sup>2</sup> National Institutes of Health-NIRT-International Center for Excellence in Research, Chennai, India, <sup>3</sup> Laboratory of Parasitic Diseases, National Institutes of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, MD, United States, <sup>4</sup> Section for Molecular Biology and Immunology, Bernhard Nocht Institute for Tropical Medicine, Hamburg, Germany, <sup>5</sup> Medical Department, Division of Gastroenterology, Infectiology and Rheumatology/Research Center ImmunoSciences, Charité-University Medicine Berlin, Berlin, Germany, <sup>6</sup> Department of Veterinary Medicine, Institute for Veterinary Epidemiology and Biostatistics, Freie Universität Berlin, Berlin, Germany, <sup>7</sup> Experimental Immunology, Department of Rheumatology and Clinical Immunology, Charité-University Medicine Berlin, Berlin, Germany, <sup>8</sup> Pitzer Laboratory of Osteoarthritis Research, German Rheumatism Research Center (DRFZ), Leibniz Institute, Berlin, Germany

#### *Edited by:*

Chang H. Kim, Purdue University, United States

#### *Reviewed by:*

Seung Goo Kang, Kangwon National University, South Korea Jee Ho Lee, La Jolla Institute for Allergy and Immunology, United States

*\*Correspondence:* Sebastian Rausch sebastian.rausch@fu-berlin.de

*Received:* 17 February 2017 *Accepted:* 02 June 2017 *Published:* 20 June 2017

#### *Citation:*

Bock CN, Babu S, Breloer M, Rajamanickam A, Boothra Y, Brunn M-L, Kühl AA, Merle R, Löhning M, Hartmann S and Rausch S (2017) Th2/1 Hybrid Cells Occurring in Murine and Human Strongyloidiasis Share Effector Functions of Th1 Cells. Front. Cell. Infect. Microbiol. 7:261. doi: 10.3389/fcimb.2017.00261 Infections by the soil-transmitted threadworm Strongyloides stercoralis affect 30–100 million people worldwide, predominantly in tropic and sub-tropic regions. Here we assessed the T helper cell phenotypes in threadworm-infected patients and experimental murine infections with focus on CD4<sup>+</sup> T cells co-expressing markers of Th2 and Th1 differentiation. We show that mice infected with the close relative S. ratti generate strong Th2 responses characterized by the expansion of CD4<sup>+</sup> GATA-3<sup>+</sup> cells expressing IL-4/-5/-13 in blood, spleen, gut-draining lymph nodes, lung and gut tissue. In addition to conventional Th2 cells, significantly increased frequencies of GATA-3+T-bet<sup>+</sup> Th2/1-hybrid cells were detected in all organs and co-expressed Th2- and Th1-cytokines at intermediate levels. Assessing the phenotype of blood-derived CD4<sup>+</sup> T cells from South Indian patients infected with S. stercoralis and local uninfected control donors we found that GATA-3 expressing Th2 cells were significantly increased in the patient cohort, coinciding with elevated eosinophil and IgE/IgG4 levels. A fraction of IL-4+CD4<sup>+</sup> T cells simultaneously expressed IFN-γ hence displaying a Th2/1 hybrid phenotype. In accordance with murine Th2/1 cells, human Th2/1 cells expressed intermediate levels of Th2 cytokines. Contrasting their murine counterparts, human Th2/1 hybrids were marked by high levels of IFN-γ and rather low GATA-3 expression. Assessing the effector function of murine Th2/1 cells in vitro we found that Th2/1 cells were qualified for driving the classical activation of macrophages. Furthermore, Th2/1 cells shared innate, cytokine-driven effector functions with Th1 cells. Hence, the key findings of our study are that T helper cells with combined characteristics of Th2 and Th1 cells are integral to immune responses of helminth-infected mice, but also occur in helminth-infected humans and we suggest that Th2/1 cells are poised for the instruction of balanced immune responses during nematode infections.

Keywords: nematode, strongyloides, Th2, hybrid, T-bet, GATA-3, co-expression, cytokines

### INTRODUCTION

Infections by helminths affect approximately 2 billion people globally, with soil-transmitted worms being the most prevalent. Infections with the threadworm S. stercoralis are currently estimated to afflict approximately 30–100 million people worldwide and are mostly asymptomatic (Puthiyakunnon et al., 2014). However, when unrecognized, the infection bears the risk of developing into a life-threatening condition in states of immune suppression (Weatherhead and Mejia, 2014).

Infections with parasitic nematodes lead to the instruction of type 2 immune responses marked by the differentiation of naïve CD4<sup>+</sup> T cells into T helper type 2 (Th2) cells (Anthony et al., 2007). These are characterized by the expression of the lineage-specifying transcription factor GATA-3 resulting in the competence to produce the effector cytokines interleukin (IL)- 4, IL-5 and IL-13 (Zheng and Flavell, 1997; Zhu et al., 2010). Animal studies show that Th2 responses are central to the control of enteric helminth infections by orchestrating a broad spectrum of defense mechanisms, such as the production of Th2-driven antibody subclasses, specialized macrophage effector programs and physiological changes like intestinal goblet cell hyperplasia, mucus hyper-secretion and intensified intestinal smooth muscle contractions (Finkelman et al., 2004; Patel et al., 2009; Harris and Gause, 2011; Allen and Sutherland, 2014). While primary infections are often long lasting, the resulting Th2-dominated immunological environment is highly effective in restricting experimental re-infection under laboratory conditions (Dawkins and Grove, 1981; Urban et al., 1991; Finkelman et al., 1997; Anthony et al., 2007; Eschbach et al., 2010). Many species, however, manage to re-infect their host, as exemplified by hookworms (Necator, Ancylostoma), whipworms (Trichuris), and Ascaris repeatedly infecting humans by tissue migrating larvae or the ingestion of infective eggs, respectively (Turner et al., 2003, 2008; Quinnell et al., 2004; Figueiredo et al., 2010). S. stercoralis is unique as the parthenogenic larvae are able to develop further into adults in the infected host, leading to multiple and potentially lifelong circles of autoinfection (Weatherhead and Mejia, 2014).

We have previously shown the induction of a stably differentiated hybrid T helper population with combined characteristics of Th2 and Th1 cells at the single cell level, namely the co-expression of GATA-3 and Th2 cytokines together with the lineage-specifying transcription factor and signature cytokine of Th1 cells, T-bet and IFN-γ, in experimental helminth infections. These cells, while being able to support both Th2 and Th1 immune responses, display a quantitatively reduced potential for Th2- as well as Th1-associated effector functions (Peine et al., 2013). We asked whether such Th2/1 cells also occur in helminthinfected patients and hence investigated T helper cell responses in patients infected by S. stercoralis in South India. Experimental infections with the murine model S. ratti were employed to assess whether the development and proportions of Th2/1 hybrid cells differ depending on parasite burden and phase of infection and to collect more detailed information about the prevalence of Th2/1 hybrid, conventional Th2 and Th1 cells in different organs affected by the parasite during its life cycle. Furthermore, we aimed to assess if Th2/1, similar to Th1 cells present in higher numbers, may serve as a source for IFN-γ sufficient for the instruction of classical macrophage activation.

We show that Th2/1 hybrid cells co-expressing IL-4 and IFN-γ are not restricted to a considerable range of murine helminth infections, but are also detectable in S. stercoralis infected patients. In mice, the proportion of Th2/1 hybrids within Th2 cells was independent of worm burden or phase of infection, but Th2/1 cells were most prominent in spleen and small intestine. To our knowledge, we show for the first time that human Th2/1 hybrid cells are marked by high IFN-γ and low GATA-3 expression, contrasting the IFN-γ/GATA-3 intermediate phenotype of their murine counterparts. Functionally, murine Th2/1 hybrid cells shared effector aspects with Th1 cells in producing IFN-γ in response to cytokine triggers and the ability to drive classical macrophage activation.

### MATERIALS AND METHODS

### Ethics Statement and Study Population

All individuals were examined as part of a natural history study protocol approved by Institutional Review Boards of the National Institute of Allergy and Infectious Diseases, USA and the National Institute for Research in Tuberculosis, India (ClinicalTrials.gov identifiers NCT00375583, and NCT00001230), and informed written consent was obtained from all participants.

We studied a total of 74 individuals comprising 34 clinically asymptomatic, S. stercoralis-infected individuals (Inf) and 40 uninfected, healthy individuals with endemic normal status (EN) in Tamil Nadu, South India (**Tables 1**, **2**). All individuals were recruited from a rural population by screening of individuals for helminth infection by stool microscopy and serology. Inclusion criteria were age of 18 to 65 years and willingness to give blood and stool samples for examination; exclusion criteria were past anthelmintic treatment, other helminth infections, or HIV infection. Strongyloides infection was diagnosed by the presence of IgG antibodies to the recombinant Strongyloides antigen, NIE,

TABLE 1 | Demographic profile of infected and uninfected cohorts.



<sup>a</sup>Hb, hemoglobin; RBC, red blood cells; WBC, white blood cells; HCT, hematocrit; PLT, platelets.

<sup>b</sup>NS, not significant.

as described previously (Bisoffi et al., 2014; Buonfrate et al., 2015). None of the individuals had lymphatic filariasis verified with TropBio Og4C3 enzyme-linked immunosorbent assay (ELISA) (Trop Bio Pty. Ltd., Townsville, Queensland, Australia) or other intestinal helminths (based on stool microscopy). None of the tested individuals suffered from acute tuberculosis, analyzed via QuantiFERON TB Gold-in-Tube enzyme-linked immunosorbent assay (ELISA) (Cellestis).

All infected individuals were treated with single doses of ivermectin and albendazole. All uninfected individuals were anti-Strongyloides NIE negative and negative for filarial and other intestinal helminths as well as acute tuberculosis.

### Hematological Parameters

Hemograms were performed on all individuals using the Act-5 Diff hematology analyzer (Beckman Coulter, Brea, CA, USA).

### PBMC Isolation and *In vitro* Culture

Heparinized blood was centrifuged at 400 × g for 10 min, plasma separated and stored at 4◦C. Samples were refilled with RPMI-1640 and PBMC isolated by Ficoll diatrizoate gradient centrifugation (LSM; ICN Biomedicals). Cells were then washed, resuspended in freezing medium (RPMI 1640, 10% DMSO, 45% heat inactivated FCS, Harlan Bioproducts) and stored in liquid nitrogen until usage. For in vitro culture the cryopreserved cells were thawed gently, washed twice and cultured with RPMI 1640 medium supplemented with penicillin-streptomycin (100 U and 100 µg/ml, respectively), L-glutamine (2 mM), 1% NEAA MEM (all from PAN–Biotech, Aidenbach, Germany) and 5% heat inactivated AB human serum (Biochrom, Berlin, Germany). For stimulations and intracellular staining cells were counted using trypan blue and adjusted to 1 × 10<sup>7</sup> cells/ml. 2 × 10<sup>6</sup> cells/well were placed on round-bottom 96 well tissue culture plates, stimulated with phorbol 12-myristate 13-acetate (PMA) and ionomycin (both Sigma-Aldrich, MO, USA) at concentrations of 25 ng/ml and 0.5 µg/ml or media alone and incubated for 4 h at 37◦C. Brefeldin A solution (10 µg/ml, eBioscience, CA, USA) was added after 30 min. After 4 h cells were washed, stained and analyzed as given below.

### Elisa

Plasma levels of total IgE, IgG1, IgG3, and IgG4 were evaluated using Ready-Set-Go! ELISA kits (eBioscience) according to the manufacturer's instructions. All samples were run in duplicates.

### Animal Experimentation, Mice and Parasites

Animal experiments were performed in accordance with the National Animal Protection Guidelines and approved by Federal Health Authorities of the States of Hamburg (permission number 55/13). The S. ratti life cycle was maintained in Wistar rats purchased from Charles River (Sulzfeld, Germany) as described earlier (Eschbach et al., 2010). Female C57BL/6 mice were bred in house at the Bernhard Nocht Institute Hamburg, Germany. Mice were kept in individually ventilated cages under specific pathogen-free (SPF) condition, infected by 200 or 2,000 S. ratti iL3 in the hind footpad and sacrificed by isoflurane inhalation followed by cervical dislocation at day 10 or 20 post infection at 8–10 weeks of age. Parasite burdens were examined at day 6 post infection in faces via quantification of S. ratti 28S ribosomal RNA using real-time quantitative PCR as described elsewhere (Nouir et al., 2012).

### Preparation of Single Cell Suspension

Spleens and mesenteric lymph nodes (mLN) were isolated and placed in cold RPMI 1640 wash medium containing 1% FCS, 100 U/ml penicillin and 100 µg/ml streptomycin (all from PAN – Biotech, Aidenbach, Germany) and then forced through 70 µm cell strainers (BD Bioscience, San Jose, CA) to obtain single cell suspensions. Erythrocytes in spleen cell suspension were lysed using ACK buffer containing 150 mM NH4Cl, 0.1 mM KHCO<sup>3</sup> and 0.1 mM Na2EDTA, pH 7.2 for 5 min on ice followed by two washing steps. Small intestinal lamina propria (siLP) cells were isolated as described earlier (Strandmark et al., 2016).

Lungs were perfused and rinsed with 0.9% NaCl until the tissue turned white before isolation of single cells. The lung was rinsed in ice-cold RPMI 1640 containing 10% FCS. Tissue was then transferred onto petri dishes, cut in small pieces (∼ 2 mm) and subsequently added to 10 ml PBS containing 0.1265 U/ml collagenase D and DNase (0.15 mg/ml, both Sigma-Aldrich, Steinheim, Germany) followed by incubation in a tube shaker water bath (250 rpm, 37◦C) for 1 h. Digested lung tissues were forced through 40 µm cell strainers, washed twice and erythrocytes lysed using ACK buffer followed by two washing steps.

For the isolation of PBMC, whole blood was diluted 2:1 with RPMI 1640 containing 10% FCS and layered onto Pancoll (density of 1.077 g/ml, PAN-Biotech, Aidenbach, Germany), centrifuged and washed twice with PBS.

Cells of all tissues were resuspended in RPMI 1640 containing 10% FCS, counted using a CASY automated cell counter (Roche-Innovatis, Reutlingen, Germany) and were adjusted to 1 × 10<sup>7</sup> cells/ml. Cells were plated on round-bottom 96 well tissue culture plates (Costar, Corning Inc., NY, USA), stimulated with PMA and ionomycin at concentrations of 50 ng/ml and 1 µg/ml and incubated for 30 min at 37◦C, followed by addition of Brefeldin A (10 µg/ml) and further stimulation for 2.5 h.

### Flow Cytometry

Human cells were stained with fixable viability dyes (eFluor780 or eFluor506 eBioscience), fixed and permeabilized using the eBioscience Foxp3 staining kit and stained intracellularly for the following markers: CD3 (BV510, clone OKT3), CD4 (PerCP, clone SK3), CD45RO (FITC, clone UCHL1), IL-4 (PE, MP4- 25D2), IL-5 (PE, clones MP4-25D2, TRFK5), IL-13 (FITC, clone PVM13-1), IFN-γ, (eF450, clone 4S.B3), GATA-3 (eF660, clone TWAJ) and T-bet (PE-Cy7, clone eBio4B10). Murine cells were stained with fixable viability dye (eFluor780, eBioscience) and subsequently fixed and permeabilized using the Foxp3 staining kit and stained intracellularly for the following markers: CD4 (PerCP, clone RM4-5), GATA-3 (eFluor660, clone TWAJ), T-bet (PE-Cy7, clone eBio4B10), IL-4 (PE, clone 11B11), IL-13 (Alexa488, clone eBio13A), IL-5 (PE, clone TRFK5), IFN-γ (eFluor450, clone XMG1.2). All antibodies were from eBioscience, BioLegend and BD Biosciences.

### *In vitro* T Helper Subset Differentiation

Cells from spleens and peripheral lymph nodes of naïve C57BL/6 mice were stained for CD4 (BrilliantViolet 510, clone RM4-5), CD62-L (APC-eFluor780, clone MEL-14), CD44 (PE, clone IM7) and CD25 (APC, clone 61.5) (all antibodies from eBioscience and BioLegend) and naïve CD4+CD25−CD62-LhighCD44neg cells were isolated on a FACS Aria cell sorter (BD Bioscience). Cells were plated on 48 well plates (Costar) coated with anti-CD3/anti-CD28 antibodies (1 µg/ml each, clones 145-2C11 and 37.51, from BD Biosciences). Th1 differentiation was induced by addition of recombinant murine IFN-γ and IL-12 (10 ng/ml each, PeproTech) and 10 µg/ml anti-IL-4 (clone 11B11, BioLegend, San Diego, CA, USA). Th2 development was induced by addition of 30 ng/ml recombinant murine IL-4 (PeproTech) and anti-IFNγ/anti-IL12/23 p40 antibodies (clones AN18, C17.8, BioLegend, 10 µg/ml each). Th2/1 hybrids were generated by addition of IFN-γ, IL-12 and IL-4 in the concentrations given above. Recombinant human IL-2 (10 ng/ml, PeproTech) was added to all cultures and replaced in fresh medium on day 3. Cells were analyzed for expression of T-bet, GATA-3, IFN-γ IL-4 and IL-13 after 5 to 6 days by intracellular antibody staining.

### T Helper Cell Innate Function

Cells isolated from mice infected with S. ratti for 10 days were stimulated with the following recombinant murine cytokines (all from PeproTech, Hamburg) or monoclonal antibodies (from BD Biosciences) for 6 h: IL-33 (100 ng/ml) and IL-7 (10 ng/ml), IL-18 and IL-12 (both 10 ng/ml), anti-CD3 (clone 145-2C11) and anti CD28 antibodies (clone 37.51, both 2 µg/ml). Controls were left untreated. Brefeldin A (10 µg/ml) was added after 30 min. Cells were then stained for CD4, GATA-3, T-bet, IL-13 and IFN-γ.

### Macrophage/T Cell Co-cultures and Nitric Oxide Assay

Peritoneal cells were isolated from naïve C57BL/6 mice by flushing the peritoneal cavity with ice-cold PBS (20 mM EDTA, 0.2% BSA) followed by washing in RPMI. 3 × 10<sup>5</sup> cells were plated on 96 well round-bottom plates (Costar) in 200 µl in RPMI and incubated at 37◦C for 1 h. Non-adhering cells were removed by 3 washing steps with warm RPMI. 1 × 10<sup>5</sup> in vitro generated Th1, Th2 or Th2/1 cells were added and co-cultured in 200 µl for 24 h. Lipopolysaccheride (LPS, L5529, Sigma) was added for costimulation at 0.5 µg/ml to some wells. Control cultures received 1 or 10 ng/ml recombinant murine IFN-γ. Nitrite concentrations in culture media were quantified by the Griess reaction.

### Histology

Proximal small intestinal tissue samples were fixed in formalin and histochemically stained with hematoxylin and eosin (H&E) to assess histomorphology and with periodic acid Schiff (PAS) for goblet cell quantification.

Enteritis was scored as described earlier (Rausch et al., 2009). Images were acquired using the AxioImager Z1 microscope (Carl Zeiss MicroImaging, Inc., Jena, Germany) at 100× magnification. All evaluations were performed in a blinded manner.

### Statistical Analyses

All data were assessed for Gaussian distribution. Statistically significant differences between two groups were analyzed by using Student's t-test (parametric) or Mann-Whitney Utest (non-parametric) using GraphPad Prism 7 software (San Diego, CA). For multiple comparisons, the Kruskal-Wallis test with Dunn's correction was performed. P-values ≤ 0.05 were considered to indicate statistical significance.

The regression analyses were carried out using SAS 9.4. Quantitative variables were tested for normality (PROC UNIVARIATE) and when necessary transformed to logarithmic values. After graphic assessment of the assumption of linear relationships (PROC GPLOT) and assessment of correlations between variables (PROC CORR) associations between variables were investigated by multivariable linear models with the procedure PROC REG. Models with different independent factors were compared by using the adjusted R 2 and the global F-statistics. The variable "group" as well as the main variable of the respecting analyses were always kept in the model, but other factors were removed in a backward selection according to the p-value and the change of R 2 . Model diagnostics included residues assessment (normality), test for multicollinearity and variance inflation. Observations with great influence (according to leverage hat-values or Cook's D) were excluded, if the model changed significantly after removing the observation. P-values ≤ 0.05 were considered to indicate statistical significance.

## RESULTS

## *S. ratti* Infection Results in Th2 Induction

In order to assess the immune response to S. ratti, mice were infected with 200 or 2,000 infective stage 3 larvae and dissected at day 10 or 20 post infection (**Figure 1A**). Worm burdens were assessed at day 6 by quantification of Strongyloides DNA in faces. Expectedly and as shown before, mice infected with 2,000 larvae displayed significantly higher worm burdens than mice inoculated with 200 larvae (**Figure 1B**; Eschbach et al., 2010). Histological examination of proximal small intestinal tissue samples showed that the infection did not result in immunopathological changes by day 10, while some mice

displayed mild signs of cellular infiltration at day 20 post infection (**Figure 1C**). Numbers of mucus-producing goblet cells were at best mildly and transiently, but not significantly increased by day 10 post infection (**Figure 1D**).

Expectedly, infected mice displayed significantly increased frequencies of CD4<sup>+</sup> T cells expressing GATA-3 and IL-4 in spleen, gut-draining mesenteric lymph nodes (mLN) and small intestinal lamina propria (siLP) when compared to uninfected controls (**Figures 2A–D**). Th2 cells were also elevated in blood and lung tissue of infected mice (SI Figures 1A,B). T-bet<sup>+</sup> Th1 cells tended to increase in the spleen of infected mice, reaching significance at day 20 post infection, while IFN-γ expression by CD4<sup>+</sup> T cells from spleens was transiently elevated at day 10 (**Figures 2E,F**). Frequencies of T-bet expressing cells in mLN, siLP and other organs were similar in all groups (**Figures 2G,H** and SI Figures 1C,D) while IFN-γ responses in the small intestine were significantly reduced in infected mice (**Figure 2H**).

Collectively, S. ratti infection led to local and systemic Th2 responses confirming previous studies (Eschbach et al., 2010; Blankenhaus et al., 2011), while transiently elevated IFN-γ production and expansion of T-bet expressing Th1 cells was restricted to the spleen. Pathological changes and Th2-driven increases of goblet cells in the small intestine were only mild, pointing out the relatively asymptomatic nature of S. ratti infection in mice.

### Th2/1 Hybrid Cells Occur in Mice Infected with *S. ratti*

Asking whether infections with S. ratti led to the expansion of Th2/1 cells as previously shown for other helminth infections (Peine et al., 2013), we screened the organs of naïve and infected mice for CD4<sup>+</sup> T cells simultaneously expressing GATA-3 and T-bet. We detected significantly increased frequencies of GATA-3+T-bet<sup>+</sup> Th2/1 cells in all organs of infected mice (**Figures 3A–D** and SI Figures 2A,B). The proportions of Tbet co-expressing Th2/1 cells within the total GATA-3+CD4<sup>+</sup> population were highest in spleen and siLP of S. ratti infected mice (**Figure 3E**). T-bet/IFN-γ co-expressing cells were detected in similar proportions within IL-4 and IL-13 producers from spleens of mice infected for 10 days (**Figure 3F**) and the frequencies of IFN-γ <sup>+</sup> Th2/1 hybrids in Th2 cytokine producing populations were similar when comparing day 10 and 20 post infection (**Figure 3G**).

As shown previously (Peine et al., 2013), Th2/1 hybrid cells expressed lower levels of GATA-3 and IL-4 than conventional Th2 cells (**Figure 3H**) and IFN-γ production was significantly decreased compared to Th1 cells (**Figure 3I**). Notably, T-bet expression by Th2/1 cells was significantly elevated compared to infection-derived Th1 cells (**Figure 3I**), which might reflect the acute activation status of the Th2/1 cells.

Thereby, our previous demonstration of the induction of Th2/1 hybrid cells in nematode and Schistosome infections (Peine et al., 2013) can be generalized to mice infected with different doses of threadworms.

### Human Study Population Characteristics

To evaluate whether CD4<sup>+</sup> T cells with combined characteristics of Th2 and Th1 cells also occur in human helminth-infected patients we investigated blood samples of patients infected with S. stercoralis and endemic uninfected control donors. The demographic profile of the cohorts is given in **Table 1**. No differences in age range, gender and socio-economic status were observed between the two studied groups. All individuals were healthy and free of symptoms. Hematological features of both cohorts are depicted in **Table 2**. No differences concerning hemoglobin levels, red and white blood cell counts, hematocrit or platelet counts were observed between the groups. The frequencies of lymphocytes and eosinophils were significantly elevated in the infected cohort (SI Figures 3A,D), while neutrophil frequencies were significantly lower in S. stercoralis infected patients (SI Figure 3B). Blood monocyte and basophil frequencies were not different between the groups (SI Figures 3C,E).

### Human *S. stercoralis* Infection Is Associated with Increased Th2 Responses

To compare the immune status of helminth-infected individuals and uninfected control subjects, PBMC from S. stercoralisinfected donors and healthy endemic controls were stimulated with PMA/ionomycin and assessed for markers of Th1 and Th2 differentiation by flow cytometry. The S. stercoralisinfected group displayed significantly elevated frequencies of CD3+CD4+GATA-3<sup>+</sup> Th2 cells compared to the uninfected control group (**Figures 4A,B**) and a higher ratio of Th2:Th1 cells (SI Figure 4). While frequencies of IL-4 and IL-5 expressing T cells were similar in both cohorts, the S. stercoralis-infected group displayed significantly increased levels of IL-13 expressing cells (**Figure 4B**). Linear regression analysis confirmed a significant effect of the group in terms of IL-13 levels (p = 0.0407) and showed that increased GATA-3 expression was associated with increased IL-13 expression (p = 0.0279) (SI Figure 5A). Furthermore, levels of IL-4, IL-13 and IFN-γ significantly increased with age (p = 0.0030, p = 0.0062, and p = 0.0407, respectively, data not shown).

No differences were observed for Th1 cells expressing T-bet and IFN-γ between the cohorts (**Figure 4C**), but increased T-bet expression levels were associated with higher IFN-γ expression (p < 0.0001) (SI Figure 5B).

Hence, S. stercoralis infection in humans is marked by Th2 differentiation, while frequencies of Th1 cells are indistinguishable from healthy controls.

### Th2/1 Hybrid Cells Are Detectable in Blood of *S. stercoralis* Infected Patients

We next asked whether hybrid Th2/1 cells sharing characteristics of Th2 and Th1 cells were detectable in blood samples of the study cohorts. We detected CD3+CD4<sup>+</sup> T cells distinctly co-expressing IL-4 and IFN-γ in the infected as well as the healthy control group (**Figures 4A,D**). Furthermore, the S. stercoralis-infected cohort displayed significantly increased frequencies of CD3+CD4<sup>+</sup> T cells co-expressing GATA-3 and T-bet compared to healthy

FIGURE 2 | S. ratti infection leads to systemic and local Th2 response. Phenotypes of CD4<sup>+</sup> T cells were assessed in mice infected with 200 or 2,000 S. ratti stage 3 larvae. (A) Exemplary flow cytometry plots of live CD4<sup>+</sup> T cells derived from spleen, mesenteric lymph nodes (mLN) and small intestinal lamina propria (siLP) of an uninfected control and S. ratti infected mouse (200 iL3) at the depicted days post infection. Bold numbers report frequencies of GATA-3 expressing cells, italic numbers report frequencies of GATA-3+IL-4<sup>+</sup> cells. (B–D) Frequencies of GATA-3<sup>+</sup> cells (top) and IL-4<sup>+</sup> cells (bottom) within live CD4<sup>+</sup> T cells isolated from spleen (B), mLN (C) and siLP (D) as detected after 4h of PMA/ionomycin stimulation. (E) Exemplary flow cytometry plots of live CD4<sup>+</sup> T cells as described in (A). Bold numbers report frequencies of T-bet expressing cells, italic numbers report frequencies of IFN-γ <sup>+</sup> cells. (F–H) Frequencies of T-bet<sup>+</sup> cells (top) and IFN-<sup>γ</sup> <sup>+</sup> cells (bottom) within live CD4<sup>+</sup> T cells isolated from spleen (F), mLN (G) and siLP (H) detected after 4 h of PMA/ionomycin stimulation. Mean <sup>+</sup> SD of <sup>n</sup> <sup>=</sup> 5–6 (naïve ctr.) and 4–5 (infected) mice. Data from one out of two experiments with similar results are shown. \*p < 0.05, \*\*p < 0.01, \*\*\*p < 0.005, \*\*\*\*p < 0.001 comparing infected to naïve controls.

#### Bock et al. Th2/1 Hybrid Cells in Strongyloidiasis

#### FIGURE 3 | Continued

spleen (top and bottom) and small intestinal lamina propria (middle) of an uninfected control and S. ratti infected mouse at the depicted days post infection. Numbers indicate frequencies of GATA-3+T-bet<sup>+</sup> and IL-4+IFN-<sup>γ</sup> <sup>+</sup> cells, respectively. (B–D) Frequencies of GATA-3+T-bet<sup>+</sup> cells (top) and IL-4+IFN-<sup>γ</sup> <sup>+</sup> cells (bottom) within live CD4<sup>+</sup> T cells isolated from spleen (B), mLN (C) and siLP (D) as detected after 4h of PMA/ionomycin stimulation. Significance was tested comparing infected and naïve controls. (E) Exemplary plots depicting frequencies of T-bet expressing cells in CD4+GATA-3<sup>+</sup> T cells derived from peripheral blood, spleen, mLN and siLP of mice infected for 10 days (200 iL3) stained directly ex vivo. Numbers indicate mean values and SD of 4-5 mice. (F) Concatenated plots depicting IL-4 single and IL-4/T-bet co-expressing cells (top left) and IL-13 single and IL-13/T-bet co-expressing cells (bottom left) in CD4 cells isolated from spleens of 5 mice infected for 10 days with S. ratti (200 iL3). Center and right plots show detection of T-bet<sup>+</sup> IFN-<sup>γ</sup> producing cells in the IL-4 and IL-13 producing populations. (G) Pie charts depict proportions of IFN-γ <sup>+</sup> Th2/1 cells (black) within the IL-4 (white), IL-13 (light gray) and IL-5 (dark gray) producing CD4+GATA-3<sup>+</sup> T cell derived from spleens of mice infected with S. ratti (200 iL3) for 10 (left) and 20 days (right). Mean and SD of n = 5 mice is shown. (H,I) Geometric mean fluorescence intensity (MFI) of GATA-3 and IL-4 (H) and T-bet and IFN-<sup>γ</sup> signals (I) of the depicted CD4<sup>+</sup> subpopulations derived from mice infected with S. ratti (200 iL3) for 10 days. The MFI of IL-4−IFN-<sup>γ</sup> − and GATA-3−T-bet<sup>−</sup> cells is depicted by a dotted line. Mean and SD of <sup>n</sup> <sup>=</sup> 5 mice. Data originate from one out of two experiments with similar results. \*<sup>p</sup> <sup>&</sup>lt; 0.05, \*\*p < 0.01, \*\*\*p < 0.005, \*\*\*\*p < 0.001.

controls (**Figure 4D**). Expectedly, IL-4 single producing cells clustered in the GATA-3<sup>+</sup> population, while some, but not all IFN-γ <sup>+</sup> cells expressed elevated levels of T-bet (**Figure 4E**). IL-4 <sup>+</sup>IFN-γ <sup>+</sup> Th2/1 hybrid cells, however, featured a GATA-3lowTbethigh phenotype and clustered with the fraction of IFN-γ <sup>+</sup> cells expressing the highest levels of T-bet (**Figure 4E**). Investigating the proportions of IFN-γ <sup>+</sup> cells within the total IL-4<sup>+</sup> population confirmed that the proportions of Th2/1 cells were similar in uninfected controls and S. stercoralis infected patients, hence the ratios of Th2:Th2/1 cells were similar in both cohorts (**Figure 4F**). However, higher values of conventional Th2 cells (IL-4+, IL-13+ or IL-5+) correlated with increased Th2/1 hybrid cell frequencies co-expressing IFN-γ and the respective Th2 cytokines (p < 0.0001) irrespective of the Strongyloides infection status (SI Figure 6).

Assessing the expression levels of GATA-3, T-bet, IL-4 and IFN-γ by mean fluorescence intensities (MFI) confirmed that human IL-4+IFN-γ <sup>+</sup> Th2/1 hybrid cells expressed significantly lower levels of GATA-3 and IL-4 than IL-4 (or IL-13, not shown) single producers (**Figure 4G**). The mean expression levels of T-bet and IFN-γ, however, were similar for IFN-γ single and IL-4+IFN-γ <sup>+</sup> double producing cells (**Figure 4G**).

Taken together, CD4<sup>+</sup> T cells with a Th2/1 hybrid phenotype marked by the co-expression of IL-4 and IFN-γ were detectable in Strongyloides-infected patients and healthy controls. The frequencies of Th2/1 cells were positively influenced by the magnitude of the overall Th2 response. In contrast to murine Th2/1 hybrids expressing intermediate levels of GATA-3, IL-4 and IFN-γ (**Figure 3**), human Th2/1 hybrid cells were marked by high IFN-γ production and low GATA-3 expression.

### *S. stercoralis* Infection Is Associated with a Distinct Serum Antibody Profile

Next, we assessed if S. stercoralis infection was marked by changes in the serum antibody profile. The infected cohort displayed significantly increased total levels of the Th2-associated antibody classes IgE and IgG4 compared to the uninfected control group (**Figures 5A,B**). IgE levels were positively correlated with IL-13 levels (p = 0.0011, data not shown). Interestingly, also the level of IgG3 was significantly increased in the S. stercoralisinfected group, while the levels of IgG1, a subclass associated with Th1 responses in humans, were similar in both cohorts (**Figures 5C,D**).

### Th2/1 Cells Share Effector Functions with Th1 Cells

T helper cells not only produce effector cytokines in response to T cell receptor (TCR) triggering by their cognate antigen, but may also react to combinations of IL-1 cytokine family members and STAT activators (Guo et al., 2012). Th2 cells have been shown to produce IL-5 and IL-13 in response to the combined stimulation by IL-33 and IL-7 (Guo et al., 2015), while stimulation by IL-18 and IL-12 results in IFN-γ production by Th1 cells (Robinson et al., 1997; Yoshimoto et al., 1998). We thus asked if Th2/1 hybrid cells induced in helminth infections share innate effector functions with Th2 and Th1 cells. To this end, we isolated CD4<sup>+</sup> T cells from mice acutely infected with S. ratti to compare their TCR-independent and TCR-induced effector responses. In response to IL-33 and IL-7, both Th2 cells and Th2/1 cells failed to produce significant amounts of IL-13 (**Figures 6A,B**) despite the expression of the IL-33 receptor ST2 (**Figure 6C**). After stimulation with IL-18 and IL-12, we detected similar IFN-γ responses of by Th1 and Th2/1 hybrid cells (**Figures 6A,D**).

We next compared in vitro generated murine Th1 and Th2/1 cells for their efficiency in macrophage activation. As shown previously (Peine et al., 2013), naïve cells stimulated in presence of IL-4, IFN-γ and IL-12 differentiated into a homogenous population of Th2/1 hybrids co-expressing GATA-3 and T-bet, many of which co-produced IL-4 and IFNγ in response to re-stimulation (**Figures 7A,B**). Similar to Th2/1 cells induced in vivo (**Figure 3**), in vitro-generated Th2/1 hybrid cells produced significantly lower levels of IFN-γ than Th1 cells (**Figure 7C**), despite similar expression levels of Tbet (**Figure 7A**). In co-cultures with peritoneal macrophages, spontaneous IFN-γ secretion by Th1 cells triggered similar levels of nitric oxide (NO) production by macrophages as induced by IFN-γ added exogenously (**Figure 7D**). Th2/1 hybrid cells induced approximately 10-fold lower NO-responses (**Figure 7D**), consistent with their lower IFN-γ production upon stimulation (**Figure 7C**), while classic Th2 cells failed to trigger an NO response (**Figure 7D**). However, when macrophages where additionally stimulated with lipopolysaccharide (LPS), the limited amount of IFN-γ produced by Th2/1 cells was sufficient to significantly enhance NO-production over the levels produced in response to the LPS-trigger alone (**Figure 7E**).

FIGURE 4 | Detection of increased Th2 responses in S. stercoralis-infected patients and characterization of human Th2/1 hybrid cells. PBMC where stimulated with PMA/ionomycin and cytokine responses and expression of GATA-3 and T-bet were detected by intracellular staining. (A) Exemplary plots of GATA-3 and T-bet (top) and IL-4 and IFN-<sup>γ</sup> expression (bottom) by CD4<sup>+</sup> T cells from an uninfected endemic control and a S. stercoralis infected donor. (B) Frequencies of GATA-3, IL-4, IL-5, and IL-13 expressing Th2 cells as detected in the endemic normal control group (EN) and S. stercoralis-infected group (Inf). (C) Frequencies of T-bet and IFN-γ expressing Th1 cells. (D) Frequencies of GATA-3+T-bet<sup>+</sup> and IL-4+IFN-<sup>γ</sup> <sup>+</sup> Th2/1 hybrid cells. Mean, SD and individual values are shown. (E) Backgating of IFN-<sup>γ</sup> + Th1, IFN-γ <sup>+</sup>IL-4<sup>+</sup> Th2/1 and IL-4<sup>+</sup> Th2 cells (black) on GATA-3 and T-bet expression of total CD4<sup>+</sup> cells (gray). Exemplary plots of a S. stercoralis-infected donor are shown. (F) Detection of IFN-<sup>γ</sup> co-producing cells in the total IL-4+CD4<sup>+</sup> population of a control and S. stercoralis-infected patient and ratios of IL-4 single producing:IL-4+IFN-<sup>γ</sup> <sup>+</sup> cells and IL-13 single producing:IL-13+IFN-<sup>γ</sup> <sup>+</sup> cells within the cohorts. (G) Expression levels of GATA-3 and T-bet in IFN-<sup>γ</sup> <sup>+</sup> Th1, IL-4<sup>+</sup> Th2 and IL-4+IFN-<sup>γ</sup> <sup>+</sup> Th2/1 cells (top). The MFI of IL-4−IFN-<sup>γ</sup> <sup>−</sup> cells is depicted by a dotted line. Bottom: Expression levels of IL-4 in Th2 vs. Th2/1 cells and IFN-<sup>γ</sup> in Th1 vs. Th2/1 cells. Data (mean and SD) are derived from samples of 5 S. stercoralis infected donors comprising <sup>&</sup>gt;100 IL-4+IFN-<sup>γ</sup> <sup>+</sup> cells assessed in two independent experiments. \*\*p < 0.01, \*\*\*p < 0.005, \*\*\*\*p < 0.001.

In conclusion, Th2/1 hybrid cells induced in helminth infection are reactive to cytokine triggers leading to TCRindependent IFN-γ production. IFN-γ produced by Th2/1 hybrid cells is sufficient for the classical activation of macrophages under appropriate conditions.

### DISCUSSION

Here we confirm that experimental murine and natural human infections by threadworms are predominantly associated with Th2 responses (Chiuso-Minicucci et al., 2010; Anuradha et al., 2015; Breloer and Abraham, 2016). Mice infected with S. ratti displayed locally and systemically increased frequencies of Th2 cells and only low signs of intestinal immunopathology, which is in line with the mostly asymptomatic course of threadworm infections in human patients (Siddiqui and Berk, 2001; Vadlamudi et al., 2006; Montes et al., 2010). S. stercoralis infected patients displayed elevated frequencies of blood eosinophils, Th2 cells, IgE and IgG4, while frequencies of Th1 cells expressing T-bet and IFN-γ were similar to uninfected controls.

Previously we have shown that infections with two helminth species, H. polygyrus and S. mansoni, lead to the differentiation of Th2/1 hybrid cells stably co-expressing GATA-3 and T-bet as well as Th2 and Th1 effector cytokines. This phenotype was maintained after clearance of infection and progression to memory cells, arguing against a transient Th2/1 state (Peine et al., 2013). Here we show that Th2/1 cells also occur in murine experimental threadworm infections, suggesting that Th2/1 hybrid cells are an integral part of the immune response to a wide range of helminth infections.

Th2 and Th1 differentiation have long been considered as mutually exclusive, as cells differentiated under polarizing conditions express distinct patterns of the lineage defining transcription factors GATA-3 and T-bet and the respective cytokines (Zheng and Flavell, 1997; Szabo et al., 2000; Löhning et al., 2002). Furthermore, positive feedback mechanisms and reciprocal inhibition of the developmental programs reinforce and assure the efficient and mutually exclusive differentiation under the appropriate conditions (Ouyang et al., 2000; Afkarian et al., 2002; Jenner et al., 2009). However, not only mice infected with helminths leading to strong Th2 reactions, but also human patients infected with S. stercoralis displayed Th2/1 cells coexpressing Th2 and Th1 cytokines, albeit at low levels. To date, we cannot exclude that reprogramming of Th1 cells and/or Th2 cells by opposing signals (Hegazy et al., 2010; Panzer et al., 2012) contributes to Th2/1 generation during helminth infections.

Our murine experimental data show that the proportion of Th2/1 cells within the total GATA-3 expressing Th2 population differed depending on the body compartment. The highest proportions of Th2/1 cells were detected in spleen and small intestinal tissue, while their proportions were lower in gutdraining lymph nodes and blood. We hence speculate that the low frequencies of Th2/1 cells detected in blood of threadworminfected patients may not necessarily reflect the proportion of Th2/1 cells in other compartments of the human body. Whether Th2/1 hybrids and conventional Th2 cells differ in their proliferative behavior, survival, and reaction to chemokines remains to be established. Further studies are also needed to assess if the preferential location of helminth-induced Th2/1 cells in spleen and the parasite-afflicted organ is associated with specific effector functions.

Our data show that human and murine Th2/1 cells clearly differed phenotypically: while murine Th2/1 hybrids were readily detectable by the coexpression of GATA-3 and T-bet, human Th2/1 cells expressed rather low levels of GATA-3, but were detectable by coexpression of IL-4 and IFN-γ. As not all GATA-3 <sup>+</sup>T-bet<sup>+</sup> cells from helminth-infected mice coproduced Th2 cytokines and IFN-γ simultaneously when restimulated in vitro it is important to note that the detection of human hybrids based on the coexpression of cytokines may lead to an underestimation of the proportion of Th2/1 cells in humans. Human and murine Th2/1 cells also differed functionally with respect to IFN-γ production: while human Th2/1 hybrids were marked by high expression levels of IFN-γ, murine Th2/1 hybrids expressed significantly lower levels than Th1 cells. T-bet expression was similar in human Th2/1 cells and Th1 cells, and T-bet expression of murine Th2/1 hybrids even exceeded the levels detected in Th1 cells, which may be explained by differential activation states of the cells. Still, only murine Th2/1 cells were restricted in IFN-γ production, which at least in murine Th1 cells is largely driven and quantitatively controlled by T-bet expression amounts (Szabo et al., 2000; Helmstetter et al., 2015). This functional difference in human vs. murine Th2/1 hybrids might be explained by the finding that human Th2/1 hybrid cells (characterized as

Frequencies of IL-13 producers in response to cytokine and TCR-triggering. Mean and SD of 5 S. ratti-infected mice is shown. (C) ST2 expression by Th1, Th2 and Th2/1 cells from S. ratti-infected mice. Mean fluorescence intensities of the populations are given in the right top corner. Dashed line and italic number relate to CD4+T-bet−GATA-3<sup>−</sup> cells. (D) Frequencies of IFN-<sup>γ</sup> producers in response to cytokine and TCR-triggering. Data are representative for two independent experiments. \*\*\*\*p < 0.001. Diamond symbols indicate values significantly different from the respective unstimulated control samples.

IL-4/IFN-γ co-producing cells) displayed rather low levels of GATA-3 expression. It hence seems likely that the lack of GATA-3-driven counter-regulation of T-bet functions allows unabated IFN-γ production by human Th2/1 cells (Chang and Aune, 2007; Jenner et al., 2009). Yet we can only speculate on how human Th2/1 cells retain their (reduced) production of IL-4 and other Th2 cytokines in face of low GATA-3 expression. In mice, conditional deletion of GATA-3 from established Th2 cells diminishes Th2 cell maintenance and IL-5 and IL-13, but not IL-4 production (Pai et al., 2004; Zhu et al., 2004). Whether other factors such as c-Maf expression allow human hybrids to express Th2 cytokines in a setting of low GATA-3 expression remains to be investigated (Kim et al., 1999). Similarly, it remains to be established if murine and human CD4<sup>+</sup> T cell co-expressing Th2 and Th1 cytokines differ in other regulatory elements of effector functions and if their relatively high proportion in the spleen has functional relevance.

An interesting finding in our patient survey was that not only Th2-associated antibody classes (IgE, IgG4) were elevated in threadworm-infected patients, but also IgG3 was significantly increased in the infected cohort. IgG3 is a potent pro-inflammatory antibody with high affinity to the activating Fcγ receptor I (FcγRI), the latter being central for the control of bacterial infections (Ioan-Facsinay et al., 2002; Vidarsson et al., 2014). Both IgG3 class switching by B cells and FcRI expression by human myeloid cells are positively regulated by IFN-γ (Erbe et al., 1990; Snapper et al., 1992). Hence IFNγ expression by Th2/1 cells in reactive lymph nodes or the

spleen, where their proportion is relatively high, might contribute to the diversification of antibody responses and assure that antibodies primarily involved in opsonization of pathogens or their secreted components do not go short in Th2-associated infections mainly driving IgE and IgG4 production by B cells.

We found that murine Th2/1 cells not only co-express IFN-γ and Th2 cytokines in response to TCR-triggering, but also responded to the combination of the STAT activator IL-12 and IL-1 family member IL-18 with production of IFN-γ. It is tempting to speculate that the differentiation of Th2/1 cells in response to helminths inducing strong Th2 reactions, such as S. ratti, H. polygyrus and S. mansoni, may provide an advantage to the infected host in preventing an overtly and exclusively Th2-biased immune set-up. It is critical for the immune system to quickly respond to rapidly replicating microbes controlled by Th1 responses and classically activated macrophages. Similar to e.g. enhanced IFN-γ production by NK cells sensing IL-18 and IL-12 during infection with the protozoan Toxoplasma gondii (Cai et al., 2000), Th2/1 cells co-induced with conventional Th2 cells during helminth infections and coinfiltrating the infected organs may contribute to early IFNγ production in response to cytokine triggers produced upon protozoan, bacterial and viral infections. Although most likely subordinate to Th1 cells in this regard, their activated state during ongoing helminth infection may facilitate IFN-γ responses by Th2/1 cells in response to cytokine triggers and hence help in supporting adequate Th1 reactions in face of ongoing Th2 responses.

There is experimental evidence that nitric oxide is toxic for Strongyloides parasites and limits worm fecundity as well as autoinfection in mice (Ruano et al., 2012, 2015). We show that murine Th2/1 cells are, expectedly, subordinate to Th1 cells in driving NO production by macrophages, coinciding with their limited IFN-γ production. However, in conjunction with TLR triggering, the limited IFN-γ production of Th2/1 cells was sufficient to stimulate significant NO production. Hence Th2/1 cells may be important to ensure that the helminth-infected host is able to quickly and adequately react to co-infections where an unabated IFN-γ response is crucial. Furthermore, IFN-γ production by Th2/1 cells generated during helminth infection often including tissue migratory larval stages may be advantageous in preventing bacterial dissemination facilitated by tissue damage (Pesce et al., 2008), especially upon parasite re-exposure of organs densely seeded with Th2 memory cells (Steinfelder et al., 2017). We currently assess if Th2/1 cells constitute a considerable source of IFN-γ and of other factors affecting the immune response to the parasite and, possibly, coinfecting pathogens.

Taken together we provide evidence that Th2/1 cells developing during helminth infections are not a feature restricted to the murine host, but also occur in humans. Especially human Th2/1 cells seem to be able to produce IFN-γ in similar amounts as Th1 cells, posing the question whether they may efficiently counter-regulate an overt Th2 bias in helminth infections and allergic diseases. By sharing effector functions with Th1 cells they may act as a provision against hampered responses to coinfecting pathogens in face of a strongly Th2-biased immune status.

### AUTHOR CONTRIBUTIONS

CB and SR performed all the experiments. SH and SR conceptualized and designed the research. SB, AR and YB recruited and classified human subjects. CB, MLB, AK, and RM analyzed the data. CB, MB, SH, and SR wrote the manuscript. All authors approved the final version of the manuscript.

### FUNDING

This work was funded by the Deutsche Forschungsgemeinschaft (IRTG 1673, CB, SH, SR and GRK 2046, SH, SR) and the Karl-Enigk-Stiftung (CB), and, in part, by the Division of Intramural Research, NIAID, USA. MB is funded by the DFG BR 3754/2-2.

### REFERENCES


### ACKNOWLEDGMENTS

The excellent support by the technicians Y. Weber, B. Sonnenburg, M. Müller, C. Palissa, and S. Spieckermann is acknowledged gratefully.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fcimb. 2017.00261/full#supplementary-material

associated with immune hyporesponsiveness and induction of a regulatory network. Infect. Immun. 78, 3160–3167. doi: 10.1128/IAI.01228-09


during intestinal infection by nematodes. Mucosal Immunol. 10, 661–672. doi: 10.1038/mi.2016.93.


**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 Bock, Babu, Breloer, Rajamanickam, Boothra, Brunn, Kühl, Merle, Löhning, Hartmann and Rausch. 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.

# Toxoplasma Co-infection Prevents Th2 Differentiation and Leads to a Helminth-Specific Th1 Response

Norus Ahmed<sup>1</sup> , Timothy French<sup>2</sup> , Sebastian Rausch<sup>1</sup> , Anja Kühl <sup>3</sup> , Katrin Hemminger <sup>1</sup> , Ildiko R. Dunay <sup>2</sup> , Svenja Steinfelder 1 † and Susanne Hartmann<sup>1</sup> \* †

<sup>1</sup> Department of Veterinary Medicine, Institute of Immunology, Freie Universität Berlin, Berlin, Germany, <sup>2</sup> Institute of Inflammation and Neurodegeneration, Otto-von-Guericke University, Magdeburg, Germany, <sup>3</sup> Division of Gastroenterology, Medical Department, Infection and Rheumatology, Research Center ImmunoSciences, Berlin, Germany

Nematode infections, in particular gastrointestinal nematodes, are widespread and co-infections with other parasites and pathogens are frequently encountered in humans and animals. To decipher the immunological effects of a widespread protozoan infection on the anti-helminth immune response we studied a co-infection with the enteric nematode Heligmosomoides polygyrus in mice previously infected with Toxoplasma gondii. Protective immune responses against nematodes are dependent on parasite-specific Th2 responses associated with IL-4, IL-5, IL-13, IgE, and IgG1 antibodies. In contrast, Toxoplasma gondii infection elicits a strong and protective Th1 immune response characterized by IFN-γ, IL-12, and IgG2a antibodies. Co-infected animals displayed significantly higher worm fecundity although worm burden remained unchanged. In line with this, the Th2 response to H. polygyrus in co-infected animals showed a profound reduction of IL-4, IL-5, IL-13, and GATA-3 expressing T cells. Co-infection also resulted in the lack of eosinophilia and reduced expression of the Th2 effector molecule RELM-β in intestinal tissue. In contrast, the Th1 response to the protozoan parasite was not diminished and parasitemia of T. gondii was unaffected by concurrent helminth infection. Importantly, H. polygyrus specific restimulation of splenocytes revealed H. polygyrus-reactive CD4<sup>+</sup> T cells that produce a significant amount of IFN-γ in co-infected animals. This was not observed in animals infected with the nematode alone. Increased levels of H. polygyrus-specific IgG2a antibodies in co-infected mice mirrored this finding. This study suggests that polarization rather than priming of naive CD4<sup>+</sup> T cells is disturbed in mice previously infected with T. gondii. In conclusion, a previous T. gondii infection limits a helminth-specific Th2 immune response while promoting a shift toward a Th1-type immune response.

Keywords: Th2, Th1, *Toxoplasma gondii*, *Heligmosomoides polygyrus*, co-infection, helminth

### INTRODUCTION

Gastrointestinal nematode infections affect around 24% of the human population (WHO, 2017 last modified January, 2017 http://www.who.int/mediacentre/factsheets/fs366/en/; Horton, 2003). These parasites are not necessarily fatal though they cause high morbidity including malnourishment, intestinal inflammation and anemia in both acutely and chronically infected patients. In particular, school children are affected hindering their mental and physical

#### *Edited by:*

Slobodan Paessler, University of Texas Medical Branch, United States

#### *Reviewed by:*

Patricia Talamás-Rohana, Center for Research and Advanced Studies of the National Polytechnic Institute (CINVESTAV), Mexico Georgia Perona-Wright, University of Glasgow, United Kingdom

#### *\*Correspondence:*

Susanne Hartmann susanne.hartmann@fu-berlin.de

† These authors have contributed equally to this work.

> *Received:* 23 February 2017 *Accepted:* 11 July 2017 *Published:* 25 July 2017

#### *Citation:*

Ahmed N, French T, Rausch S, Kühl A, Hemminger K, Dunay IR, Steinfelder S and Hartmann S (2017) Toxoplasma Co-infection Prevents Th2 Differentiation and Leads to a Helminth-Specific Th1 Response. Front. Cell. Infect. Microbiol. 7:341. doi: 10.3389/fcimb.2017.00341 development. Moreover, individuals living in endemic areas get frequently re-infected and immunity only develops after several decades (Mosmann and Coffman, 1989). Most areas endemic for nematodes are co-endemic for various diseases, such as malaria, tuberculosis, toxoplasmosis (Bahia-Oliveira et al., 2009), leishmaniasis and salmonellosis (Hotez and Kamath, 2009). Every pathogen should be encountered by a tailored immune response engaging a certain set of effector molecules, however this is a problem for the immune system during co-infections. Seroprevalence of Toxoplasma gondii reaches up to 70% in certain areas with high chances of co-infection with different pathogens among the human population (Bahia-Oliveira et al., 2009; Pappas et al., 2009). High prevalence of T. gondii is found in tropical regions, such as Latin America, Middle East, Africa and Southeast Asia. On the other hand T. gondii infection evokes and is controlled by a very different immune response compared to helminth infections. Detailed experimental studies are required to unravel how acute or chronic infections with certain pathogens affect the immune system, when faced with new challenges by unrelated infections.

Helminth infections are typically characterized by the activation and expansion of CD4<sup>+</sup> T helper 2 (Th2) cells, which express the transcription factor GATA-3 and secrete interleukin (IL)-4, IL-5, IL-9, and IL-13, leading to IgG1 and IgE antibody production. In addition, the Th2 response leads to eosinophilia, enhanced mucus production, specific granulomaformation around larvae, as well as specific priming of innate cells, such as macrophages. This immune response can directly, or via upregulation of effector molecules, such as RELM-β produced by goblet cells, reduce worm fecundity and enhance parasite expulsion (Artis et al., 2004; Owyang et al., 2006).

Helminths evade the host immune responses due to hostparasite interactions (McSorley and Maizels, 2012) whereby the Th2 immune responses are actively regulated by the worm (Yazdanbakhsh et al., 2001). Thus, the Th2 response is accompanied by the emergence of parasite-induced regulatory cells, such as regulatory B-cells (Breg) (Hussaarts et al., 2011), regulatory T-cells (Treg) (Taylor et al., 2012) and alternatively activated macrophages (AAM) (Gordon and Martinez, 2010), which are known to limit parasite-specific and unspecific immune responses (Steinfelder et al., 2016). Heligmosomoides polygyrus is a well-studied strictly intestinal helminth of mice featuring all of the aforementioned characteristics (Reynolds et al., 2012). In addition to this, it is a widespread natural infection of wild mice (Maaz et al., 2016). The parasites are orally taken up as infective stage 3 larvae (L3), which subsequently embed themselves into the small intestinal wall to develop into L4. They then emerge as adults into the lumen, where they prevail for weeks before being expelled, depending on the mouse strain (Bansemir and Sukhdeo, 1994; Reynolds et al., 2012).

In contrast, T. gondii is an obligate intracellular protozoan parasite, which orally infects warm-blooded vertebrate hosts. After an initial intestinal phase, T. gondiispreads systemically and converts into a dormant stage in muscle and brain tissues (Dubey, 2008). Human infections with T. gondii are common and are mostly asymptomatic in immunocompetent individuals (Ho-Yen and Joss, 1992), although they may trigger basal inflammation (Parlog et al., 2015). Here, IFN-γ plays an important role in the containment of T. gondii (Ely et al., 1999). However, a previous latent infection in immunocompromised humans can reactivate and cause life threating encephalitis if left untreated (Luft et al., 1984; Montoya and Liesenfeld, 2004). During T. gondii infection a T helper 1 (Th1) immune response is elicited. This provides a strong, protective immune response and is characterized by dendritic cells (DC) producing IL-12. The production of IL-12 leads to the differentiation of CD4<sup>+</sup> T cells into Th1 cells expressing the transcription factor T-bet and the secretion of IFN-γ. Additionally, innate cells, such as neutrophils, NK-cells and innate lymphoid cells provide other early sources of IFNγ (Gazzinelli et al., 1994; Sturge et al., 2013; Klose et al., 2014).

Differentiation of Th1 and Th2 cells has been well documented in the literature. In vitro-based studies have shown that Th1 or Th2 polarizing conditions cause differentiated cells to lose their ability to completely switch phenotype after increased cell division (Murphy et al., 1996; Grogan et al., 2001). However, other studies have shown that Th subsets have the flexibility to produce non-lineage-specific cytokines (Murphy and Stockinger, 2010; O'Garra et al., 2011; Coomes et al., 2013). Furthermore, we have previously reported a subset of Th hybrid cells expressing both transcription factors T-bet and GATA-3, as well as producing IFN-γ and IL-4 at intermediate levels during helminth infections (Peine et al., 2013).

Studies on H. polygyrus and T. gondii co-infection in mice have so far focused on a previous infection with helminths and have shown that initially CD4<sup>+</sup> and CD8<sup>+</sup> T cell immunity against T. gondii is suppressed in mice. At later stages the T. gondii-specific CD4<sup>+</sup> T cell response recovers whereas the CD8<sup>+</sup> response remains disrupted (Khan et al., 2008). In line with this, other studies have shown that prior infection with H. polygyrus induced suppression of IL-12 dependent differentiation of effector CD8<sup>+</sup> T cells as well as IFN-γ production against T. gondii. Interestingly, IL-4 and IL-10 deficiency was necessary to reverse the obstructing effect of H. polygyrus infection on the CD8<sup>+</sup> T cell response toward Toxoplasma (Marple et al., 2017). The majority of co-infection studies despite being protozoan, viral or bacterial infection, have focused on infections with helminth first due to their ability to downmodulate immune responses (Rousseau et al., 1997; Liesenfeld et al., 2004; Chen et al., 2005, 2006; Graham et al., 2005; Su et al., 2005, 2014a,b; Weng et al., 2007; Khan et al., 2008; Noland et al., 2008; Miller et al., 2009; Frantz et al., 2010; Dias et al., 2011; Potian et al., 2011; Kolbaum et al., 2012; du Plessis et al., 2013; Osborne et al., 2014; Budischak et al., 2015; Coomes et al., 2015; Gondorf et al., 2015; Rafi et al., 2015; Obieglo et al., 2016). In light of the fact, that Th2 immunity against helminths is an ongoing challenge in humans and livestock, we aimed to investigate how a previous protozoan infection affects the development of Th2 responses in CD4<sup>+</sup> T cells and protection against helminths.

We observed that a previous T. gondii infection leads to an overall suppression of H. polygyrus-specific Th2 immunity and enables H. polygyrus-specific CD4<sup>+</sup> T cells to produce IFN-γ.

### MATERIALS AND METHODS

### Animals

Female NMRI and C57BL/6 mice (8 weeks old; purchased from Janvier, Saint Berthevin, France) were bred under specific pathogen-free (SPF) conditions at the Institute of Medical Microbiology, Universitätsklinikum Magdeburg, Germany or at the Institute of Immunology, Department of Veterinary Medicine, Freie Universität Berlin. The experiments performed followed the National Animal Protection Guidelines and were approved by the German Animal Ethics Committee for the protection of animals.

### Isolation of *T. gondii* Tissue Cysts and Oral *T. gondii* Infection

Female NMRI mice were infected orally (p.o.) with type II ME49 strain T. gondii cysts. After 8–10 months, tissue cysts were collected from the brains of chronically infected mice. After perfusion, brains were mechanically homogenized in 1 mL sterile PBS. Cysts were quantified using a light microscope and 8–10 weeks old female C57BL/6 mice were infected with 2-tissue cysts p.o. by oral gavage in a total volume of 200 µl/mouse (Möhle et al., 2016)

### *H. polygyrus* Infection

The parasite Heligmosomoides polygyrus was retained by serial passage in C57BL/6 mice as described previously (Rausch et al., 2008). Mice aged 8–10 weeks old were infected by oral gavage with 200 L3 larvae in drinking water. On day 14 post infection (p.i.) mice were sacrificed by isofluorane inhalation.

### Detection of *T. gondii* Parasitemia by PCR

Toxoplasma gondii burden was determined using Roche FastStart Essential DNA Green Master kit with manufacturer's protocol. TgB1 (TIBMolbiol, Berlin, Germany) was used as a target gene and Mm.ASL (TIBMolbiol, Berlin, Germany) as a reference (Heimesaat et al., 2014). Target/reference ratios were all calculated using the LightCycler <sup>R</sup> 480 Software release 1.5.0 (Roche, Germany).

### Worm Fecundity and Worm Burden

Adult worms were isolated from the small intestine and counted. Female worms were subsequently kept individually (8 per mouse) in a 96 well round-bottom plate containing RPMI, 200 U/ml penicillin and 200 µg/ml streptomycin (all from PAA, Austria) at 37◦C. After 24 h female H. polygyrus adults were removed from the wells and fecundity was determined by counting the eggs shed per female worm using a binocular microscope.

### Preparation of Parasite Antigen

Heligmosomoides polygyrus antigen (HpAg) was prepared from adults worms that were kept in culture containing 100 U/ml penicillin and 100 µg/ml streptomycin for 24 h as described before (Rausch et al., 2008).

### Cell Culture

Cells were cultured in complete RPMI 1640 medium (cRPMI) containing 10% FCS 200 U/ml penicillin, 200 µg/ml streptomycin (all from PAA, Austria) in an incubator at 37◦C and 5% CO2.

### Single Cell Suspension Preparation

Spleen and mesenteric lymph nodes (mLN) were isolated from mice, homogenized and filtered through 70 µm cell strainers (BD Bioscience, San Jose, CA) to obtain single cell suspensions. The cells were then washed and re-suspended in cRPMI. Cells were counted using a CASY automated cell counter (Roche-Innovatis, Reutlingen, Germany). Small intestinal lamina propria (siLP) and epithelium (siE) cells were isolated by the removal of the whole small intestine that was then stored on ice in cold HBSS (w/o Ca2<sup>+</sup> Mg2+) (PAA, Pasching, Austria) containing 2% FCS and 10 mM HEPES (PAA, Pasching, Austria). Small intestines were washed through with 20 ml cold buffer using a 20G needle. After washing, mesenteric fat and Peyer's patches were removed. The small intestines were then cut open longitudinally and mucus scraped off with forceps. Additionally adult H. polygyrus worms were removed and counted using forceps. Small intestines were then washed in HBSS/FCS/HEPES and cut in 1 cm pieces and stored in 20 ml HBSS/FCS/HEPES containing 0.154 mg/ml DTE (Sigma-Aldrich, St. Louis, MO). The 1 cm pieces were incubated in a tube shaker water bath (200 rpm, 37◦C) for 15 min. This step was repeated twice and then the intestinal pieces were transferred into 20 ml HBSS/FCS/HEPES containing 5 mM EDTA and agitated at room temperature for 15 min, repeated three times. Intestinal pieces were put into fresh 20 ml HBSS/FCS/HEPES and the cell suspension containing epithelium and intraepithelial lymphocytes retrieved for density gradient isolation. Intestinal pieces were washed in RPMI to remove residual EDTA and then placed in 10 ml 37◦C complete RPMI 1640 medium containing 0.1 mg/ml Liberase (Roche, Basel, Switzerland) and 0.1 mg/ml DNAse (Sigma-Aldrich, St. Louis, MO, USA). Intestinal pieces were then incubated at 37◦C, 200 rpm for 30 min. After incubation, tubes containing intestinal pieces were vortexed vigorously to disturb remaining intestinal pieces. The intestinal pieces were then forced up and down through an 18G needle. Suspensions were then filtered over a 70 µm cell strainer and washed twice with HBSS/HEPES. The cell suspensions from siLP and siE were added on a percoll gradient (GE healthcare life sciences, Sweden). Lamina propria and epithelial cells were collected from the 40%/70% interface after centrifugation. Cells were washed in cRPMI and counted using Neubauer chambers (C-Chip, Biochrom GmbH, Berlin, Germany).

### Generation and Antigen Loading of Bone Marrow-Derived Dendritic Cells

Naïve female C57BL/6 mice were used for isolation of bone marrow from the femur and tibia. Bone marrow cells were washed in RPMI medium and 1 × 10<sup>6</sup> cells/ml were cultured in 10 ml/petri dish for 6 days in cRPMI containing 10 ng/ml recombinant murine GM-CSF (PeproTech, Hamburg, Germany). On day 3, 10 ml of cRPMI with 10 ng/ml GM-CSF was added. On day 6 BmDC were counted, seeded at 1 × 10<sup>5</sup> cells/well in a 96 well plate and stimulated with 50 µg/ml of H. polygyrus antigen (HpAg) for 24 h. Cells were then washed and used for the co-culture experiment.

### Antigen-Specific Restimulation of CD4<sup>+</sup> Splenocytes

1 × 10<sup>6</sup> splenocytes were co-cultured with 1 × 10<sup>5</sup> BmDC pulsed overnight with HpAg for 5 h in the presence of 3 µg/ml Brefeldin A (eBioscience, San Diego, CA, USA) followed by intracellular staining for CD154 and cytokines.

### Cell Proliferation Assay

Spleen cells were isolated and stained with CFSE **(**Carboxyfluorescein succinimidyl ester**)**. 1 × 10<sup>6</sup> CFSE labeled cells were cultured with 20 µg/ml H. polygyrus antigen. Cells were cultured for 6 days and restimulated with 1 µg/ml PMA and 1 µg/ml Ionomycin (both Sigma-Aldrich, St. Louis, MO, USA) in the presence of 3 µg/ml Brefeldin A (eBioscience, San Diego, CA, USA) followed by intracellular staining.

### Realtime PCR

RNA was isolated from intestinal tissue sections previously stored at −80◦C via homogenization in RNA lysis buffer. The tissue supernatant was processed with a innuPREP RNA kit (Analytik Jena, Jena, Germany) following manufacturer's instructions. 2 µg of RNA was reverse transcribed to cDNA using the High Capacity RNA to cDNA kit (Applied Biosystems, Foster City, CA). The relative expression of β-actin, resistin-like molecule-beta (Relm-β), IL-12 and IFN-γ, was determined by Real Time PCR using 10 ng of cDNA and the FastStart Universal SYBR Green Master Mix (Roche, Basel, Switzerland). Primer pairs used for gene amplification were as follows: β-actin forward: GGCTGTATTCCCCTCCATCG, reverse: CCAGTTGGTAAC AATGCCATGT, Relm-β (Retnlb) forward: GGCTGTGGATCG TGGGATAT, reverse: GAGGCCCAGTCCATGACTGA. IL-12 forward: ATGGCCATGTGGGAGCTGGAGAAAG, reverse: GTGGAGCAGCAGATGTGAGTGGCT. IFN-γ forward: ATg AACgCTACACACTgCATC, reverse: CCATCCTTTTgCCAg TTCCTC. Primer pair efficiency was determined via a standard curve. The mRNA expression was normalized to the β-actin housekeeping gene and calculated by Roche Light Cycler 480 software.

### Cytokine Detection by ELISA

3.5 × 10<sup>5</sup> splenocytes were stimulated with 20 µg/ml H. polygyrus antigen or 2 µg/ml anti-CD3 and anti-CD28 (both eBioscience, San Diego, CA, USA) for 6 days. Supernatants were analyzed for IL-4, and IFN-γ using Ready-Set-Go Elisa Kits (eBioscience, San Diego, CA, USA) according to the manufacturer's instructions.

### Antibody Isotype Detection by ELISA

Heligmosomoides polygyrus-specific IgG1 and IgG2a were measured in serum. 96 well microtiter plates coated with 10 µg/ml H. polygyrus antigen were incubated with serum diluted 1:100 with 3% BSA in PBS. Bound antibody isotypes were detected using alkaline phosphatase conjugated anti-mouse IgG1 and IgG2a antibodies diluted 1:5000 each (Rockland, PA, USA) and para-nitrophenylphosphate (Sigma, Steinheim am Albuch, Germany). All samples were run in duplicates. Arbitrary units were calculated using pooled samples as reference.

### Flow Cytometry

For surface and intracellular staining, the monoclonal antibodies listed were used: CD4 (PerCP) (RM4-5); CD192 (CCR2) (Alexa 647) (SA203G11) all from BioLegend (Biozol); CD8 (53-6.7); IL-4 (PE-Cy7) (11B11); IL-5 (PE) (TRFK5); IL-13 (Alexa 488) (eBio13A); IFN-γ (eFluor 450) (XMG1.2); CD154 (PE) (MR1); Foxp3 Alexa 488 (FJK-16s); GATA3 (eFluor 660) (TWAJ); Dead Cell Exclusion Marker (DCE) (efluor 780); DCE (efluor 506); Siglec F (PE) (E50-2440); Tbet (PE) (eBio4B10); IL-13 (eFluor 660) (eBio13A); CD11b (PE) (M1/70); F4/80 (PerCP-Cy5.5) (BM8); Ly-6G (Gr-1) (PE-Cy7) (RB6-8c5); Ly-6C (eFluor 450) (HK1.4); TNF-α (Alexa488) (MP6-XT22) all from eBioscience, San Diego, CA, USA.

For intracellular staining of cytokines and transcription factors cells were fixed and permeabilized using the fix/perm buffer kit (eBioscience, San Diego, CA, USA). FACSCantoII flow cytometer and FACSAriaIII sorter (both BD Bioscience, Heidelberg, Germany) were used for cell analysis. FlowJo software 10.2 was used for final analysis (Tree star Inc., Ashland, OR, USA).

### Statistics

Experiments were performed as shown and displayed as mean ± SD or mean ± SEM as indicated. Statistical analysis was performed using GraphPad Prism software (La Jolla, CA, USA). The level of significance was determined using the Mann Whitney U-test or Kruskal-Wallis with Dunn's multiple comparison test.

### RESULTS

### Prior Infection with *T. gondii* Results in Increased Fecundity of *H. polygyrus* in Co-infection

We investigated whether infection with T. gondii affects the control of helminth parasites in the small intestine (**Figure 1A**). Mice infected with T. gondii for 14 days followed by H. polygyrus infection did not show altered parasitemia of T. gondii in the heart compared to mice infected with T. gondii alone (**Figure 1B**). Similarly, co-infected mice did not show a significant difference in H. polygyrus adult worm burden compared to H. polygyrus single infection (**Figure 1C**). However, female worms retrieved from coinfected mice showed a significantly higher fecundity compared to worms from H. polygyrus single infection (**Figure 1D**). Thus, a previous and on-going infection with the protozoan parasite T. gondii leads to a decline of anti-helminthic control in terms of fecundity leading to enhanced egg production.

### Previous Infection with *T. gondii* Selectively Restricts Th2 Polarization in Response to Helminth Infection

The increased fecundity in co-infected mice, described as number of eggs produced per female worm ex vivo, might be due to insufficient Th2 immune pressure. To test this we compared systemic and local immune responses in co-infected and single infected groups. To differentiate between the contrasting immune responses the Th2 lineage marker GATA3 (Zheng and Flavell, 1997) and the Th1-lineage marker T-bet (Szabo et al., 2000) were used. GATA3 is also present on a subset of regulatory T cells that also express Foxp3. Regulatory T cells were excluded using their expression of Foxp3 (**Figure 2A**; Wohlfert et al., 2011).

The frequency of GATA3+Foxp3−CD4<sup>+</sup> was drastically reduced in the spleen, small intestinal lamina propria (siLP) and small intestinal epithelial layer (siE) in co-infected mice compared to mice infected with H. polygyrus only. The reduction in GATA3 was similar to levels found in uninfected mice (**Figure 2B**). Furthermore, histology of the small intestine of co-infected mice showed no GATA-3 expression compared to H. polygyrus infected mice (**Figure 2E**). On the contrary, the Th1-lineage marker T-bet in CD4<sup>+</sup> T cells was expressed in co-infected mice with higher levels to T. gondii alone in all compartments (**Figure 2C**). In addition to this, T-bet<sup>+</sup> expression in CD8<sup>+</sup> T cells was higher in siLP, siEL and mLN in co-infected mice compared to mice infected with T. gondii alone (**Figure 2D**).

### Co-infection Leads to Suppression of Th2 Cytokine Responses in CD4<sup>+</sup> T Cells

The lack of GATA3 expression prompted us to investigate whether the failure to differentiate into Th2 cells (**Figure 2**) extends to the inability to secrete Th2 cytokines. CD4<sup>+</sup> T cells were restimulated with PMA/Ionomycin and cytokine expression was assessed (**Figure 3A**). The frequency of CD4<sup>+</sup> cells producing IL-4 in spleen, siLP and mLN showed a marked reduction in co-infected mice compared to the H. polygyrus single-infected group. However, only a trend in the reduction of IL-4 was observed in siEL (**Figure 3B**). Also, the frequency of IL-5 and IL-13 in spleen, siLP, siEL, and mLN showed a significant reduction in co-infected mice in comparison to H. polygyrus infection alone (**Figures 3C,D**). This observation shows that the Th2 immune responses are suppressed locally as well as systemically in mice previously infected with T. gondii. On the contrary, CD4<sup>+</sup> cells producing the Th1 cytokine IFN-γ showed an increase in the co-infected group compared to mice infected with T. gondii alone (**Figure 3E**), mirroring the enhanced T-bet expression found in the same cell population (**Figure 2C**). Similarly, T-bet and IFN-γ expression in CD8<sup>+</sup> T cells were comparable in mice single and co-infected with T. gondii (Figure S1). To further investigate this finding, splenocytes were stimulated with anti-CD3/28 to evaluate the ability of these

FIGURE 2 | Restricted Th2 responses in co-infected mice. Cells from spleen, mesenteric lymph nodes (mLN), small intestinal lamina propria (siLP) and small intestinal epithelium (siE) were isolated and stimulated with PMA and ionomycin in the presence of Brefeldin A followed by intranuclear staining for the lineage transcription factors GATA3 and T-bet. Gating strategy shown in (A), Bar graphs showing frequencies of CD4<sup>+</sup> T cells expressing GATA3 (B), T-bet (C), and T-bet expression in CD8<sup>+</sup> T cells (D). GATA3 expression in the duodeunum of the small intestine with scale bar of 100 µm (E), data shown as mean ± SEM pooled from 2 independent experiments n = 6, statistical analysis was performed using the Mann-Whitney test. (B–D) shown as mean ± SEM, pooled from two independent experiments with n = 8–10. Statistical analysis was performed using the Kruskal-Wallis with Dunn's multiple comparison test, \*P ≤ 0.05, \*\*P ≤ 0.01, and \*\*\*P ≤ 0.001.

cells to produce the Th2 cytokine IL-4 (**Figure 3F**). The coinfected group did not show any IL-4 production compared to the H. polygyrus-infected group whereas IFN-γ was produced in all groups.

Next, we evaluated effector mechanisms that are downstream of a Th2 response. As the appearance of Th2 cytokines in helminth infection is associated with eosinophilia, we tested the influence of a co-infection on the number of eosinophil granulocytes. Consistent with the lack of Th2 cytokine secretion (**Figure 3**), co-infected mice showed a significant reduction in the frequency of eosinophils in spleen (**Figure 4A**) compared to H. polygyrus single infection. Similarly, in mice infected with both T. gondii and H. polygyrus there was reduction in RELM-β expression in small intestinal tissue compared to H. polygyrus single infection (**Figure 4B**). This indicates that effector cells and molecules normally elicited by helminth infection are directly affected by the absence of Th2 cytokines in a coinfection setting. In contrast, no changes were observed in the frequency of inflammatory monocytes (F4/80+GR1+Ly6C+) in both single and co-infected mice (**Figures 4C,D**). However, a significant reduction was observed in the production of TNF-α in inflammatory monocytes of co-infected mice in comparison to T. gondii single infected mice (**Figure 4E**),. In addition to this, co-infected mice showed a drastic reduction in H. polygyrus specific IgG1 compared to H. polygyrus single infection, which showed a prominent parasite-specific IgG1 response (**Figure 4F**). In contrast, helminth-specific IgG2a is increased in co-infected mice compared to mice infected with helminths alone. Thus, this data indicates that effector cells, effector molecules and the respective antibody response normally elicited by helminth infections are drastically altered in mice previously infected with T. gondii.

### Co-infection Results in a Helminth-Specific Th1 Profile

The lack of the Th2 immune responses in co-infection may suggest that the CD4<sup>+</sup> T cells are not responding to H. polygyrus infection. Thus, we aimed to identify helminth antigen-specific CD4<sup>+</sup> T cells in co-infected mice. For this we generated bone marrow derived dendritic cells (BmDC) from naïve animals pulsed with H. polygyrus antigen (HpAg) and co-cultured them with splenocytes from single and co-infected animals. The antigen-reactive CD4<sup>+</sup> cells were subsequently detected by the activation marker CD154 (CD40L) (**Figure 5A**; Frentsch et al., 2005). We observed a reduction in IL-4 and IL-13 in activated CD154+CD4<sup>+</sup> cells in the co-infected group compared to the H. polygyrus infected group. Interestingly, despite being unable to produce Th2 cytokines, significant numbers of CD154+CD4<sup>+</sup> T cells from the co-infected group were able to produce IFN-γ in response to HpAg, which was not observed in the H. polygyrussingle infected group (**Figure 5B**). CFSE staining can be used to identify proliferating cells due to halving of CFSE in daughter cells during proliferation. Here, we assessed the proliferation of splenocytes from single and co-infected animals in response to HpAg by gating on CFSE<sup>−</sup> cells. Interestingly, while CD4<sup>+</sup> T cells from both groups proliferated based on CFSE-staining, the Th2

FIGURE 3 | Th2 but not Th1 immune responses are absent in co-infected mice. Cells from spleen, mLN, siLP, and siE of single and co-infected animals were isolated and stimulated with PMA and ionomycin in the presence of Brefeldin A followed by intracellular cytokine staining. Gating strategy for the cytokines IL-5 and IFN-<sup>γ</sup> in CD4<sup>+</sup> cells isolated from spleen (A). Bar graphs showing frequencies of CD4<sup>+</sup> T cells expressing IL-4 (B), IL-5 (C), IL-13 (D), and IFN-γ (E) in spleen, mLN, siLP, and siE. (F) IL-4 and IFN-γ production detected by ELISA in supernatant from 3 × 10<sup>5</sup> splenocytes stimulated with (black bars) and without (white bars) anti-CD3/CD28 antibodies. (B–F) Data shown as mean ± SEM, pooled from two independent experiments with n = 8–9 Statistical analysis was performed using the Kruskal-Wallis with Dunn's multiple comparison test, \*P ≤ 0.05, \*\*P ≤ 0.01, and \*\*\*P ≤ 0.001.

effector molecules and antibody isotypes. (A) Bar graph showing the frequency of Siglec-F<sup>+</sup> eosinophils in the spleen, pooled from 2 independent experiments n = 7–8. (B) Relative Gene Expression of RELM-β compared to housekeeping gene β-actin (pooled from two experiments, n = 5–8). (C,D) Gating strategy and bar graphs showing frequency of inflammatory monocytes (F4/80+GR1+Ly6C+) in spleen stimulated with LPS (pooled from two experiments, n = 8). (E) Percentage of TNFα, with the response of the T. gondii infected group represented as 100%, stimulated with LPS, pooled from two experiments, n = 9. (F) H. polygyrus specific IgG1 and IgG2a detected by ELISA, n = 7–8 pooled from two independent experiments. Data shown as mean ± SEM. Statistical analysis was performed using the Kruskal-Wallis with Dunn's multiple comparison test, \*P ≤ 0.05, \*\*P ≤ 0.01, and \*\*\*P ≤ 0.001.

Ahmed et al. Suppressed Th2 Responses during Co-infection

cytokine IL-4 was reduced in co-infected animals compared to helminth single infection. On the other hand, IFN-γ production in response to HpAg was increased in co-infected animals (**Figure 5C**). Additionally, in vitro stimulation of splenocytes with HpAg showed similar results. Here, significant amounts of IL-4 and IL-10 could be detected in supernatants from mice infected with H. polygyrus only, while IFN-γ was significantly increased in co-infected animals (**Figure 5D**). Hence, this data supports the observation that helminth-specific CD4<sup>+</sup> T cells from co-infected animals are unable to commit to the Th2 lineage but produce the Th1 cytokine IFN-γ in response to a helminth infection instead.

### DISCUSSION

In this study a previous low-dose oral infection with T. gondii prevented the establishment of local and systemic Th2 responses normally induced by infection with H. polygyrus. Neither Th2 cytokines nor the transcription factor GATA3 as well as features triggered by IL-4 (IgG1 antibodies, RELM-ß) or IL-5 (eosinophilia) could be detected in co-infected animals. The observed increase in female worm fecundity in co-infected animals is a likely consequence of a lack in Th2 responses due to the reduced antibody levels and effector molecule RELMß, as both were shown to influence H. polygyrus fecundity (Owyang et al., 2006; McCoy et al., 2008). Interestingly the Th1 response to T. gondii was not diminished in CD4<sup>+</sup> and CD8<sup>+</sup> T cells in co-infected animals, since both subsets were able to produce similar amounts of IFN-γ in contrast to other studies, where the helminth infection precedes the protozoan infection (Khan et al., 2008; Marple et al., 2017). This observation is in line with unaltered parasitemia of T. gondii during chronic phase of infection. Furthermore, inflammatory monocytes expressing Ly6C infiltrate the brain to control T. gondii via the production of pro-inflammatory mediators, such as TNF-α, IL-1-α IL-1-β and nitric oxide synthase (Dunay et al., 2008, 2010; Biswas et al., 2015). In co-infected animals the frequency of inflammatory monocytes was not altered, but the capacity to produce TNF-α in response to LPS was reduced. However, this had no effect on parasitemia of T. gondii.

The observation of an apparent reduction in Th2 responses might be due to abolished priming and polarization events that occur at various stages, such as insufficient priming of naïve CD4<sup>+</sup> helper T cells, altered function of dendritic cells (DC) leading to aberrant polarization, or the local cytokine milieu present at the time of helminth infection. Moreover, a recent study has shown that systemic T. gondii infection leads to a long-term defect in the generation and function of naive T lymphocytes (Kugler et al., 2016). However, our findings suggest that in a low-dose infection with T. gondii, this effect is not as drastic,since helminth-specific CD4<sup>+</sup> T cells were shown to proliferate and respond to antigen as seen in **Figure 5**.

The differentiation between Th1 and Th2 cells requires positive feedback loops and cross-inhibition of other lineages for uniform Th cell differentiation (Mosmann and Coffman, 1989; Paul and Seder, 1994). In addition to this the transcription factor specific for Th1, T-bet, has the ability to suppress the Th2 transcription factor GATA-3. This might provide an explanation for the cross-regulation of cytokines in Th cell differentiation (Hwang et al., 2005). During T. gondii infection T-bet can suppress IL-4 and GATA-3 expression, thus, preventing endogenous Th2 cell associated programming (Zhu et al., 2012). In our study, the observed up-regulation of T-bet was in line with the significant reduction of GATA-3 in spleen, siLP and siEL. The reduction in GATA-3 expression followed the absent Th2 cytokine production observed in co-infection with T. gondii. Our study is in line with other studies showing suppression in the Th2 responses against helminths, when another pathogen is involved prior or at the same time during infection (Liesenfeld et al., 2004; Lass et al., 2013; Nel et al., 2014). Interestingly, the frequencies of CD4<sup>+</sup> T-cells from co-infected animals producing either Tbet or IFN-γ are similar to mice infected with T. gondii alone or even higher. This suggests that CD4<sup>+</sup> T cells expressing T-bet and IFN-γ in coinfected animals consist of pre-existing T. gondiispecific T cells and H. polygyrus-specific T-cells able to produce T-bet and IFN-γ in this coinfection setting.

In general, helminths are shown to actively induce a Th2 program that requires the programming of dendritic cells (DCs) (Steinfelder et al., 2009). DCs from skin LN exposed to Nippostrongylus brasiliensis showed transcriptional changes of different DC subsets (Connor et al., 2017). Generally, DCs are pivotal in eliciting Th2 cell responses in vivo. A depletion of the CD11c<sup>+</sup> DCs subset during infection with S. mansoni led to an abolished Th2 response and an increase in the production of IFN-γ (Phythian-Adams et al., 2010). This reduced Th2 response was also observed in H. polygyrus infection when mice were depleted of CD11c<sup>+</sup> DCs (Smith et al., 2011). However, during T. gondii infection, the induced Th1 immune response is dependent on early IL-12 production by APCs (Scanga et al., 2002). DCs prime naïve T cells, but are also a target of effector cytokines produced by previously polarized effector T cells and innate cells. In our study it is most likely that DCs are affected by the previous and ongoing infection with T. gondii and the ensuing cytokine milieu. The presence of IL-12 and IFN-γ at the time point and site of infection as shown in Figure S2 might impact the ability of local DCs to be able to prime naïve T cells for Th2 differentiation. Future studies should investigate the underlying mechanisms and involvement of DCs, such as CD8α <sup>+</sup> DCs that are a source of IL-12 during T. gondii infection (Mashayekhi et al., 2011).

Dendritic cells can also be primed in response to epitheliumderived cytokines known as alarmins. These cytokines are released during epithelial tissue damage (Swamy et al., 2010). Thymic stromal lymphopoietin (TSLP) plays an important role in mounting a Th2 response in H. polygyrus infection (Massacand et al., 2009). Also IL-33 is an alarmin and it has been shown that when DCs are treated with IL-33 they polarize CD4 T cells to produce Th2 cytokines (Besnard et al., 2011; Eiwegger and Akdis, 2011). Another tissue derived cytokine; IL-25 has been demonstrated to be involved in Th2 cytokine responses and N. brasiliensis expulsion. However, this cytokine has not been described to directly act on DCs (Fallon et al., 2006;

Wang et al., 2007). Since early Th2 polarization is dependent on the release of these cytokines that act as alarmins it might be fruitful to investigate these cytokines very early after infection with H. polygyrus in previously infected T. gondii mice.

using the Mann-Whitney test, \*P ≤ 0.05, \*\*P ≤ 0.01.

Importantly, we show that while helminth-reactive CD4<sup>+</sup> T cells are unable to produce Th2 cytokines in a co-infection, they still express significant amounts of IFN-γ after restimulation with helminth antigen. We identified IFN-γ producing H. polygyrus antigen (HpAg)-reactive T cells in co-infected animals using the activation marker CD154 and a short stimulation protocol (Frentsch et al., 2005; Chattopadhyay et al., 2006). We also saw this in CFSE−CD4<sup>+</sup> T cells that expanded in response to HpAg. In line with our findings, Coomes et al. have shown that a coinfection with Plasmodium chabaudi and H. polygyrus led to a reduction in Th2 responses. Furthermore, they observed upregulation of IFN-γ when Th2 cells from H. polygyrus-infected mice were adoptively transferred into Rag1−/<sup>−</sup> mice infected with P. chabaudi. However, blocking of IL-12 and IFN-γ only partially preserved Th2 immunity in response to H. polygyrus (Coomes et al., 2015).

In summary, our data on helminth-antigen specific restimulation of CD4<sup>+</sup> T cells suggest that naïve CD4<sup>+</sup> T cells harboring a cognate TCR for helminth antigen fail to commit to the Th2 lineage and are polarized toward a Th1 phenotype in mice previously infected with Toxoplasma. This switch in cytokine expression leads to the absence of effector features downstream of the Th2 response and consequently to higher worm fecundity in co-infected animals. Recent studies emphasized the importance of bystander activation and concurrent infections on the outcome of the immune response to unrelated pathogens or vaccines (Reese et al., 2016; Tao and Reese, 2017). In regards to the differences in the development of protective Th2 immunity observed in our study and by others in both mice and humans, it is important to focus on infections not only in a "clean" host but also in the context of individual infection history as well as co-infections.

### AUTHOR CONTRIBUTIONS

NA, SS, SR, TF, and KH performed all the experiments. SH, ID, and SS conceptualized and designed the research. NA, SS, and SH wrote the manuscript. All authors approved the final version of the manuscript.

### FUNDING

This study was funded by the German Research Foundation: GRK 2046 to SH and SS. Norus Ahmed received a stipend of the GRK 2046.

### REFERENCES


### ACKNOWLEDGMENTS

The authors thank Yvonne Weber, Marion Müller, Bettina Sonnenburg, Christiane Palissa, and Beate Anders for providing their excellent technical support.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fcimb. 2017.00341/full#supplementary-material


and clearance by recruited monocytes. Acta Neuropathol. Commun. 4:25. doi: 10.1186/s40478-016-0293-8


**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 Ahmed, French, Rausch, Kühl, Hemminger, Dunay, Steinfelder and Hartmann. 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.

# Reciprocal Interactions between Nematodes and Their Microbial Environments

#### Ankur Midha, Josephine Schlosser and Susanne Hartmann\*

Department of Veterinary Medicine, Institute of Immunology, Freie Universität Berlin, Berlin, Germany

Parasitic nematode infections are widespread in nature, affecting humans as well as wild, companion, and livestock animals. Most parasitic nematodes inhabit the intestines of their hosts living in close contact with the intestinal microbiota. Many species also have tissue migratory life stages in the absence of severe systemic inflammation of the host. Despite the close coexistence of helminths with numerous microbes, little is known concerning these interactions. While the environmental niche is considerably different, the free-living nematode Caenorhabditis elegans (C. elegans) is also found amongst a diverse microbiota, albeit on decaying organic matter. As a very well characterized model organism that has been intensively studied for several decades, C. elegans interactions with bacteria are much more deeply understood than those of their parasitic counterparts. The enormous breadth of understanding achieved by the C. elegans research community continues to inform many aspects of nematode parasitology. Here, we summarize what is known regarding parasitic nematode-bacterial interactions while comparing and contrasting this with information from work in C. elegans. This review highlights findings concerning responses to bacterial stimuli, antimicrobial peptides, and the reciprocal influences between nematodes and their environmental bacteria. Furthermore, the microbiota of nematodes as well as alterations in the intestinal microbiota of mammalian hosts by helminth infections are discussed.

#### *Edited by:*

Lorenza Putignani, Bambino Gesù Ospedale Pediatrico (IRCCS), Italy

#### *Reviewed by:*

Mario M. Zaiss, Universitätsklinikum Erlangen, Germany Chang-Shi Chen, National Cheng Kung University, Taiwan

#### *\*Correspondence:*

Susanne Hartmann susanne.hartmann@fu-berlin.de

*Received:* 22 February 2017 *Accepted:* 07 April 2017 *Published:* 27 April 2017

#### *Citation:*

Midha A, Schlosser J and Hartmann S (2017) Reciprocal Interactions between Nematodes and Their Microbial Environments. Front. Cell. Infect. Microbiol. 7:144. doi: 10.3389/fcimb.2017.00144 Keywords: nematode, helminth, microbiota, antimicrobial peptides, antibiotic resistance

## INTRODUCTION

Parasitic nematodes are responsible for widespread morbidity in humans and animals. It is estimated that approximately 1.5 billion people are infected with one or more of these organisms (Hotez et al., 2008; World Health Organization, 2017) which also pose a considerable burden for animal production (Eijck and Borgsteede, 2005; Nganga et al., 2008). The extraordinary success of these parasites speaks to their ability to withstand a multitude of stresses such as host immune pressures and infectious challenges from microbes. Most parasitic nematodes inhabit the intestines of their hosts, co-existing with numerous microbial species. In studying these dynamics, researchers have focused extensively on the host-parasite relationship; more recently, the role of the microbiota as a major third party in the relationship is better appreciated due to diverse and far-reaching influences in health and disease (Donaldson et al., 2016). Much attention in these studies is given to host immune mechanisms while the interactions between nematodes and their microbial environments are largely overlooked. Due to many technical and biological challenges associated with studying parasites, these questions can be very difficult to address. The free-living nematode Caenorhabditis elegans (C. elegans) is frequently found in nature amongst a diverse microbiota on decaying organic matter (Frézal and Félix, 2015). C. elegans has been very well characterized and its interactions with bacteria studied in considerable detail. As such, these findings might be conveyed to parasitic nematodes and greatly inform our understanding of how parasites interact with the host-microbiota, as many immune-related pathways and responses may be conserved (Tarr, 2012; Rosso et al., 2013). This review will summarize current knowledge regarding how parasitic nematodes and their microbial environments may influence each other, supplemented with insights from work in C. elegans.

### NEMATODES AND THEIR MICROBIAL ENVIRONMENTS

Studies on intestinal nematodes report a range of alterations to the composition of the microbiota, ranging from increased or decreased diversity, dysbiosis, or in some cases no identifiably significant changes (summarized in **Tables 1, 2**). Responses by C. elegans to numerous microbes have also been studied. Accordingly, parasitic helminths illuminate our understanding of how microbial communities can be altered by the presence of nematodes whereas studies of C. elegans can explain how microbial populations impact nematode physiology,


→, no change; ↑, increase; ↓, decrease; n.r., not reported; SCFA, short chain fatty acids; dpi, days post-infection; wpi, weeks post-infection; Nod2, nucleotide-binding oligomerization domain-containing protein 2; WT, wild type.


#### TABLE 2 | Effects of helminth infection on host microbiota and metabolism in mice.

→, no change; ↑, increase; ↓, decrease; n.r., not reported; SCFA, short chain fatty acids; dpi, days post-infection; wpi, weeks post-infection; Nod2, nucleotide-binding oligomerization domain-containing protein 2; WT, wild type; SFB, segmented filamentous bacteria.

reproduction and growth; considered together, one can better understand how these multifactorial biological systems operate. Herein, we focus most of our discussion on selected members of the genera Trichuris and Ascaris, as well as the rodent parasite Heligmosomoides polygyrus bakeri and the free-living C. elegans. These organisms have been selected due to their importance to human and animal health along with the abundance of available data in the literature, though additional studies from other helminths are considered where appropriate.

### The Nematodes

#### Trichuris

The whipworms Trichuris trichiura, T. suis, and T. muris infect humans, pigs, and mice respectively. The life cycles of these species are comparable and infection begins with ingestion of developed eggs (Bethony et al., 2006; Klementowicz et al., 2012). Hatching occurs within hours of ingestion, liberating L1 larvae which invade the intestinal wall and undergo successive molts to the L4 stage by 3 weeks post-infection (pi) and finally develop into mature adults by 12 weeks pi (Bethony et al., 2006). Trichurids inhabit the most dense and diverse microbial environments of their hosts: the cecum and colon (Klementowicz et al., 2012). They can survive here for 1–2 years with individual females laying up to 5,000 eggs per day (Bethony et al., 2006).

#### Ascaris

The roundworms A. lumbricoides and the closely related A. suum infect humans and pigs respectively. As with Trichuriasis, Ascariasis also spreads via the fecal-oral route in humans as well as in pigs (Dold and Holland, 2011). Within 3 h, ingested eggs containing L3 larvae hatch and by 18 h pi the larvae begin their tissue migratory phase, passing through the liver after invading the cecum and proximal colon (Murrell et al., 1997). The larvae reach the lungs and pharynx by days six to eight pi (Roepstorff et al., 1997), are swallowed and can then be found in the small intestine as L4 stage larvae before further developing into mature adults. Adult Ascarids reside in the small intestine for around 1 year with individual females producing hundreds of thousands of eggs per day (Dold and Holland, 2011).

### H. p. bakeri

The trichostrongyloid H. p. bakeri is amongst the most common helminth parasites of rodents and a well characterized model for chronic intestinal nematode infection (Wu et al., 2012). Infection is initiated by ingestion of L3 larvae which migrate into the mucosa of the small intestine. By day three pi the larvae develop into L4 within the mucosa before returning to the lumen of the duodenum by day nine pi where they develop into, and remain as, egg-laying adults for approximately 12 weeks in the wood mouse (Apodemus sylvaticus) and more than twice as long in BALB/c laboratory mice (Robinson et al., 1989; Gregory et al., 1990; Behnke et al., 2009).

### C. elegans

C. elegans is a free-living, bacterivorous nematode found predominantly in humid temperate areas (Félix and Duveau, 2012; Frézal and Félix, 2015). These nematodes can be identified in feeding and reproductive stages in rotting fruit and herbaceous stems. The life cycle is characterized by freshly hatched L1 stage larvae undergoing multiple molts before reaching adulthood in as little as 3 days. Under stressful conditions including crowding, limited food, and heat stress, L1 individuals will pursue an alternate life cycle through a pre-dauer L2 (L2d) stage, followed by a non-feeding, stress-resistant alternate L3 stage called "dauer." Dauer individuals can survive for several months without food before reentering a relatively normal development cycle when conditions are more favorable. In the laboratory, the reference N2 strain is typically cultured on a diet of the rather innocuous Escherichia coli OP50. In contrast to controlled laboratory conditions, C. elegans shares its natural habitat with a variety of organisms, including bacteria, phages, fungi, isopods, arthropods, and other nematodes (Félix and Duveau, 2012). Additionally, worms can be found in states of starvation or constipation in their natural environments (Barrière and Félix, 2005).

### Parasitic Nematodes Influence the Host-Intestinal Microbiota

The mammalian intestine is home to approximately 3.8 × 10<sup>13</sup> microbes from all three domains of life, archaea, bacteria, and eukaryotes, collectively referred to as the microbiota (Sender et al., 2016). The different regions of the intestinal tract form divergent habitats, varying in bacterial type and density. In humans, an estimated 103–10<sup>4</sup> microbial cells/mL intestinal content reside in the small intestine (Sender et al., 2016). Considerably more bacterial diversity and density is found in the large intestine with as many as 10<sup>11</sup> cells/mL contents representing thousands of different species (Zoetendal et al., 2008; Sender et al., 2016). Interestingly, microbial density and diversity of intestinal regions are inversely correlated with concentrations of host-derived antimicrobials; specialized secretory Paneth cells are a key source of α-defensins, lysozymes, and C-type lectins and are particularly prominent in the proximal intestine, the intestinal site of lowest microbial richness (Bevins and Salzman, 2011; Donaldson et al., 2016). Additionally, persistence of common gut commensals is partially mediated by resistance to host defense molecules (Cullen et al., 2015). Whether these factors influence helminth niche selection remains to be determined. Various conditions such as age, diet, health status and genetic background can impact the host microbiota composition. Maturation of the gut microbiota is characterized by increasing diversification with age while age-related changes are also due to exposure to a more varied diet (David et al., 2014; Odamaki et al., 2016). Analysis of human fecal samples can detect rapid, diet-mediated changes in microbial composition. For example, consumption of animal products increases the abundance of bile-tolerant species and decreases abundance of species that metabolize plant-derived polysaccharides (David et al., 2014). Dysbiosis, a potentially pathogenic imbalance of microbial communities, can be precipitated by exposure to pharmaceutical substances such as antimicrobials as well as the presence of infectious agents (Donaldson et al., 2016). Acute antibiotic treatment can drastically perturb the gut microbiota, decreasing species diversity (Dethlefsen et al., 2008). Alarmingly, while some changes appear reversible, other alterations can be detected in fecal samples even years after a short course of antibiotic treatment (Jernberg et al., 2007). A major clinical implication of a disturbed gut microbiota is elevated risk of enteric infection, such as with Clostridium difficile, which itself is associated with decreased microbial diversity (Milani et al., 2016). Many studies have now shown that intestinal nematodes influence their microbial niches as they establish themselves as part of their wider environment within the host; therefore, the role of helminths in dysbiosis is an area of active investigation (**Figure 1**).

The gut microbiota of humans, pigs, and mice is dominated by two of the 29 known bacterial phyla: Bacteroidetes and Firmicutes, with lower abundance phyla differing between hosts and including Actinobacteria, Deferribacteres, Proteobacteria, Spirochaetes, Tenericutes, and Verrucomicrobia (Leser et al., 2002; Eckburg et al., 2005; Consortium, 2012; Nguyen et al., 2015; Weldon et al., 2015). The diversity of the microbiota is thought to reflect the health of the intestine, with far reaching implications for the overall health of the host; greater species diversity contributes to healthy metabolic and immune functioning. Dysbiosis is associated with a reduction in intestinal biodiversity, predisposing to the outgrowth of particularly harmful bacterial species (Carding et al., 2015). The literature is mixed with respect to whether or not helminths cause dysbiosis, and some reports have indicated beneficial effects in therapeutic settings. The few studies assessing the influence of helminths on intestinal microbial communities are a mix of clinical observations and animal experiments, employing different analytical tools and acquiring microbiota samples from different sources. Hence, it is difficult to draw meaningful and generalizable conclusions from

the available evidence. Still, through a careful reading of the literature, some common threads emerge.

Analysis of fecal samples from children in rural Ecuador revealed widespread helminth infection (Cooper et al., 2013). Children co-infected with both T. trichiura and A. lumbricoides appeared to have a decreased microbial diversity compared to uninfected children and children with T. trichiura single infections who did not differ from uninfected individuals. Interestingly, in a subset of children with mixed infections the authors also reported a higher abundance of Streptococcus spp., not usually dominant in healthy individuals. Taken together, these data suggest dysbiosis in the presence of A. lumbricoides, while T. trichiura single infections did not result in drastic alterations to the fecal microbiota. Another indication of dysbiosis associated with Ascaris infection comes from a study of A. suum-infected pigs which showed worm burdendependent decreased bacterial diversity compared to control animals (Paerewijck et al., 2015). One study in humans found no difference in community structure of fecal samples in healthy volunteers infected with Necator americanus (Cantacessi et al., 2014). In contrast, a study of fecal samples from helminthinfected (T. trichiura, A. lumbricoides, hookworm) individuals in rural Malaysia found a positive association between helminthcolonization and microbiota diversity (Lee et al., 2014). Burrowed in the epithelia, T. trichiura likely interacts with the mucosal microbiota, the composition of which is known to differ considerably from fecal communities (Eckburg et al., 2005). Combined with the obstacles to sampling the small intestine, the site of A. lumbricoides and N. americanus infections, it is difficult to draw firm conclusions from these human studies regarding beneficial or harmful effects of helminth infections with respect to intestinal microbial composition. Though in the case of Ascaris, the local effects seen in the porcine small intestine (Paerewijck et al., 2015) correspond with the distal effects seen in human feces (Cooper et al., 2013).

Animal studies of helminthiases can offer more depth compared to human studies, accounting for the limitations associated with sampling only the fecal microbiota. During the larval stage, T. suis appears not to disrupt the porcine colonic microbiota (Li et al., 2012); however, chronic infections in pigs demonstrate worm burden-dependent disruption (Wu et al., 2012). T. suis also appears to promote Campylobacter infection in pigs, intensifying colitis disease severity (Mansfield and Urban, 1996; Mansfield et al., 2003; Wu et al., 2012). Chronic T. muris infections in mice considerably decrease overall microbial diversity, an effect that appears reversible upon worm clearance (Holm et al., 2015; Houlden et al., 2015). In these studies, the murine microbiota also showed a shift away from Bacteroidetes in favor of Firmicutes (Holm et al., 2015; Houlden et al., 2015). A study of wild mice has also observed an increased Firmicutes/Bacteroidetes ratio in helminth-infected individuals (Kreisinger et al., 2015) whereas mice experimentally infected with the small intestinal nematode Nippostrongylus brasiliensis showed a decrease of Firmicutes while increasing Bacteroidetes (Fricke et al., 2015). Though N. brasiliensis decreases the Firmicutes/Bacteroidetes ratio, it also promotes the reduction of segmented filamentous bacteria (SFB) which are thought to prevent colonization by bacterial

pathogens, as it was shown in SFB-colonized mice with enhanced resistance to the pathogenic Citrobacter rodentium (Ivanov et al., 2009; Fricke et al., 2015). Mice with chronic T. muris infections (Holm et al., 2015; Houlden et al., 2015) as well as those with acute H. p. bakeri infections (Rausch et al., 2013) have higher abundance of Enterobacteriaceae, a family shown to overgrow during intestinal inflammation (Lupp et al., 2007) and strongly correlated with Crohn's disease (Gevers et al., 2014) and C. difficile infection (Milani et al., 2016) in humans. These data indicate a propensity for helminths to associate with a simplified microbiota; however, whether these observations are generalizable across helminth infections in diverse mammalian hosts remains to be determined; nonetheless, these animal experiments are suggestive of helminths promoting dysbiosis in their hosts.

Despite the evidence for dysbiosis, a fairly consistent finding across various helminth infections is an increased abundance of Lactobacillaceae (Walk et al., 2010; Rausch et al., 2013; Reynolds et al., 2014; Fricke et al., 2015; Holm et al., 2015; Houlden et al., 2015; Ramanan et al., 2016), a family composed primarily of Lactobacillusspp., currently under intense investigation for use as probiotics (Walter, 2008; Salvetti et al., 2012). Additionally, studies employing intestinal nematodes as therapeutic interventions to treat intestinal inflammatory conditions have shown beneficial outcomes (Broadhurst et al., 2012; Giacomin et al., 2015; Ramanan et al., 2016). Macaques with idiopathic chronic diarrhea show signs characteristic of dysbiosis such as decreased diversity, increased bacterial attachment to the intestinal mucosa, and increased abundance of Enterobacteriaceae (Broadhurst et al., 2012). Bacterial attachment and abundance of Enterobacteriaceae were effectively decreased by T. trichiura infection, while bacterial diversity was restored, indicating a protective effect of helminth infection in this setting. Similarly, mice deficient in Nod2 are susceptible to Crohn's disease and can be colonized by inflammatory Bacteroides spp., while T. muris infection protects against pathogenic colonization (Ramanan et al., 2016). In patients with celiac disease, low-level infection with N. americanus increases gluten tolerance (Croese et al., 2015), while also increasing microbial richness when combined with gluten consumption (Giacomin et al., 2015). Taken together, these findings indicate that intestinal nematodes possess great therapeutic potential and may promote microbial restoration in conditions of pre-existing dysbiosis.

From taxonomic data referring to the microbiota composition alone it is challenging to conclude whether intestinal nematodes contribute to or ameliorate dysbiosis. As such, the influence of helminths on gut metabolic profiles may offer more clues. Methodological and technological advances allow for metabolic profiling of gut microbiota, along with identification and quantification of metabolites of interest (Vernocchi et al., 2016). Such experiments can be quite informative, though they are currently far less abundant than those reporting taxonomic changes. Metabolomic analysis of the colonic microbiota of pigs infected with T. suis demonstrated reduced capacity to metabolize and utilize carbohydrates relative to control animals, an effect seen in acute and chronic infections (Li et al., 2012; Wu et al., 2012). Interestingly, T. suis-infected pigs also showed signs of altered fatty acid metabolism, including higher levels of oleic acid relative to control pigs (Li et al., 2012). Notably, oleic acid possesses antibacterial activity (Dilika et al., 2000) and may therefore influence microbial composition during Trichuris infection. Mice chronically infected with T. muris also showed signs of reduced carbohydrate and amino acid metabolism and nutrient uptake compared to naïve animals, features which likely contribute to decreased weight gain during helminth infection (Houlden et al., 2015). Observational analysis of helminth-infected humans has attributed decreased carbohydrate metabolism to Ascaris infection (Lee et al., 2014). In contrast, wild mice infected with helminths showed mixed effects with respect to metabolomic shifts (Kreisinger et al., 2015). Notably, wild mice with H. p. bakeri appeared to have an increased capacity to metabolize carbohydrates; however, metabolites were not measured and mouse weights were not reported. Together, these observations suggest helminthmediated microbiota changes tend to reduce metabolic capability in the gut, ultimately manifesting as nutritional deficiencies.

While dietary carbohydrate utilization may be impaired by intestinal nematodes, other experiments demonstrate potential benefits of helminth infection such as increased abundance of intestinal short-chain fatty acids (SCFAs), typically produced from microbial processing of indigestible oligosaccharides (Paerewijck et al., 2015; Zaiss et al., 2015; Vernocchi et al., 2016). SCFAs can serve as energy sources and anti-inflammatory compounds and are therefore under investigation for their therapeutic potential (Vernocchi et al., 2016). Increased SCFAs have been detected during infection with H. p. bakeri in mice, A. suum in pigs, and N. americanus in humans (Paerewijck et al., 2015; Zaiss et al., 2015). Furthermore, H. p. bakeri and A. suum can produce SCFAs directly (Tielens et al., 2010; Zaiss et al., 2015). Reduced weight gain associated with helminth infection may also prove to be therapeutic in different circumstances, as illustrated by Yang et al. in a study of obese mice and mice fed obesity-inducing diets (Yang et al., 2013). In agreement with other observations, N. brasiliensis infection attenuated weight gain. Importantly, helminth infection decreased adiposity and hepatic lipid storage while improving glycemic control, implying a reversal of metabolic disease. These studies indicate that microbiota alterations and metabolomic changes associated with intestinal nematodes may offer certain benefits.

The mixed observations described herein highlight the difficulties of generalizing findings from these studies, simultaneously hinting at the massive potential to apply these insights in diverse therapeutic settings. Animal studies designed to induce strong immune responses are typically a result of a high infection dose and thus high worm burdens and may not reflect average worm burdens seen in nature. Indeed, some of the aforementioned animal experiments demonstrate microbiota changes suggestive of dysbiosis (Li et al., 2012; Wu et al., 2012; Holm et al., 2015; Houlden et al., 2015). In contrast, observations from natural settings or studies inducing low-level helminth infections suggest less dramatic changes (Cantacessi et al., 2014; Kreisinger et al., 2015). Additionally, when intestinal nematodes are introduced into individuals with intestinal

inflammation or metabolic abnormalities, they appear beneficial (Broadhurst et al., 2012; Yang et al., 2013; Giacomin et al., 2015). Taken together, these data emphasize co-evolution of helminths, gut microbiota, and mammalian hosts, such that helminths can be considered less strictly as intruders but rather members of the gut macrobiota (Gause and Maizels, 2016). In order to establish themselves in the mammalian intestine, rather than being solely beneficial or causing outright harm, nematodes must shape the niche such that a new ecological balance is found. This task is aided by some microbes and challenged by others, and the insights gained from understanding these dynamics can provide opportunities to better understand and perhaps manipulate incredibly complex biological systems. It should be noted that many potential mechanisms have been identified which act in concert to shape the intestinal microbiota. Immune mechanisms such as modifying the balance between Th1/Th17 vs. Th2 responses as well as the induction of regulatory T cells have been characterized and discussed elsewhere (Reynolds et al., 2015; Gause and Maizels, 2016; Zaiss and Harris, 2016). Furthermore, the bulk of work done thus far has focused primarily on bacteria and far less is known about the archaeal and fungal components of intestinal microbial communities. In addition to metabolomic factors touched upon here, important considerations which are relatively under-studied and not frequently discussed are the direct signals between helminths and microbes. These reciprocal interactions are considered below.

### Environmental Microbes Impact *C*. *elegans* Physiology

The study of intestinal parasites is particularly amenable to understanding alterations to microbial communities of the gut, though the microbial influence on helminths is relatively difficult to assess in these systems. As a model organism with many available genetic tools, C. elegans can bridge some of the knowledge gaps encountered when studying parasitic nematodes. When recovered from compost or isopods, which worms may use as vectors to shuttle between food sources, C. elegans is frequently found in the dauer stage (Félix and Duveau, 2012). In contrast, proliferating populations of nematodes can be found in microbe-dense environments, particularly rotting fruits or vegetation. A study by Samuel et al. characterized the natural bacterial microenvironment of C. elegans (Samuel et al., 2016). The most prevalent bacterial phyla in decaying apples were Proteobacteria, Bacteroidetes, Firmicutes, and Actinobacteria. The Enterobacteriaceae family within Proteobacteria encompassed the most abundant genera detected, while Lactococcus, Lactobacillus, Acetobacter, Gluconobacter, and Gluconoacetobacter were also quite prevalent. In some samples, Escherichia spp. were also identified, though they were not particularly abundant. Considering the immensity of bacterial diversity coming from 29 extant phyla (Yarza et al., 2014), there is considerable overlap in the microbial environments of parasitic and free-living nematodes. Experiments assessing the effect of C. elegans on an experimental microbiota could be designed, whereby the nematodes are introduced into a defined microbial environment and followed over time to assess changes to bacterial constituents. Such studies would be analogous, albeit profoundly simplified, to studies assessing gut microbial changes in helminth infections and could shed light on the underlying mechanisms contributing to compositional alterations. Akin to some reports of parasitic nematodes, observations from rotting fruit suggest that growth of C. elegans is enhanced amongst a simpler microbiota (Samuel et al., 2016).

Many studies have demonstrated the susceptibility of the C. elegans intestine to a multitude of pathogens (Couillault and Ewbank, 2002), including Salmonella enterica serovar Typhimurium (S. typhimurium) (Lee et al., 2015), Enterococcus faecalis (Sim and Hibberd, 2016), Burkholderia spp. (Lee et al., 2013), Staphylococcus aureus (JebaMercy and Balamurugan, 2012), Proteus mirabilis (JebaMercy and Balamurugan, 2012), and Pseudomonas aeruginosa (Dai et al., 2015). Beyond posing a threat to the nematode intestine, bacteria may also damage and infect the nematode surface. In experimental settings, Yersinia spp. can form biofilms that block the oral cavity of the worms, resulting in starvation and death (Tan and Darby, 2004). Wild nematodes are subject to cuticular infection by Microbacterium nematophilum (O'Rourke et al., 2006) and the fungal pathogen Drechmeria coniospora (Engelmann et al., 2011). Additionally, Elizabethkingia are capable of digesting the nematode cuticle components keratin and collagen (Riffel et al., 2003; Félix and Duveau, 2012). C. elegans is also infected by intracellular pathogens such as the microsporidian parasite Nematocida parisii (Troemel et al., 2008) and the Orsay virus (Félix et al., 2011). Though studies have demonstrated the ability of bacteria to colonize parasitic nematodes, no information is available concerning which microbes might be beneficial or harmful for them. In contrast, numerous studies in C. elegans have shown beneficial and harmful effects of specific bacterial strains. Alphaproteobacteria and Lactococcus spp. tend to be beneficial and promote C. elegans proliferation while Bacteroidetes and Gammaproteobacteria seem to impair nematode physiology and induce stress/immune responses (Samuel et al., 2016). It stands to reason that similar interactions exist between bacteria and parasitic nematodes and that the composition of the host microbiota would then influence the health of parasitic nematodes in the gut. Interestingly, nematode fecundity is subject to various stresses; the host immune response can decrease helminth fecundity, exemplified by reduced egg production by H. p. bakeri during a type 2 helper T cellmediated immune response (Strandmark et al., 2016). As bacteria influence C. elegans fecundity (Szewczyk et al., 2006), it is possible that communities within the microbiota may also influence the reproductive capacity of parasites. Larger populations of proliferating C. elegans are found in more highly degraded apples containing larger microbial populations, thereby providing more nutritional sources for the nematodes. Also, apples with higher populations of worms had microbiota which were more similar to each other than apples with lower/no worms. Apples containing proliferating populations of C. elegans also had significantly fewer bacterial species and diversity than apples with non-proliferating populations. Most importantly, proliferation was associated with Alphaproteobacteria and the absence of potential pathogens from Gammaproteobacteria

and Bacteroidetes. Perhaps there is a general preference by nematodes for decreased abundance of several Bacteroidetes organisms, given the observations in parasite-host microbiota studies discussed previously. As Lactobacillaceae are more predictive of successful H. p. bakeri establishment in mice (Reynolds et al., 2014), Enterobacteriaceae and Acetobacteraceae are predictive for larger proliferating C. elegans populations in rotting apples, while Gammaproteobacteria show an inverse relationship (Samuel et al., 2016). It is not fully understood how intestinal nematodes establish in specific habitats such as the cecum and colon in the case of Trichurids, or the small intestine in the case of Ascarids and H. p. bakeri; however, some speculate the choice is based on the abundance of nutrients or the relative lack of harmful stimuli (Davey, 1964; Bansemir and Sukhdeo, 1994, 2001). While determining the nature of habitat selection in these settings may prove experimentally elusive, the available data suggest that parasitic nematodes, as with their free-living counterparts such as C. elegans, have advantageous associations with some microbial species and detrimental interactions with others.

### Colonization of Nematodes by Microbes

All animals, including nematodes, live in association with a multitude of other species. The host-microbe relationship has gained a lot of attention in recent studies, with various reciprocal interactions ranging from parasitism, to commensalism and mutualism, seen as potent forces guiding the co-evolution of species. In addition to influencing the rich and diverse microbial environments that they inhabit, nematodes inevitably also harbor their own microbiota that is likely essential for normal development and physiology, as is the case for other organisms. Though the data are limited, studies have documented the ability of bacteria to colonize Ascaris, H. p. bakeri, as well as C. elegans.

Earlier studies demonstrated nematode-associated bacterial communities in A. suum (Hsu et al., 1986; Shahkolahi and Donahue, 1993). One study aimed at identifying the source of serotonin present in nematodes, which generally lack the enzymes necessary for de novo synthesis of serotonin from tryptophan, revealed 4 × 10<sup>9</sup> bacteria per gram of nematode intestine (Hsu et al., 1986). In addition to E. coli, Enterobacter, Klebsiella, Acinetobacter, Citrobacter, Pseudomonas, Aeromonas, and Shigella were identified among gram-negatives while Staphylococcus, Streptococcus, Corynebacterium, and Bacillus were identified amongst gram-positive organisms. The species identified suggest that the nematode's intestinal microbiota may have been derived from that of the host. Another study demonstrated that antibiotic treatment of ex vivo cultured A. suum reduced, but did not eliminate, the bacterial load carried by the worms and determined that the posterior portion of the intestine harbored the highest number of culturable bacteria (Shahkolahi and Donahue, 1993). Studies that employ more modern methods while sampling site-specific host- and parasiteassociated microbial communities could greatly enhance our understanding of the selectivity and development of an Ascarisspecific microbiota.

Another study showed that A. lumbricoides nematodes isolated from cholera patients were colonized by Vibrio cholerae (Nalin and McLaughlin, 1976). Worms were retrieved from patients after being spontaneously passed or after de-worming and a majority of worms were colonized by V. cholerae of the same serotypes found in stool samples of cholera patients. Serial sectioning of the worms revealed colonization along the entire length of the intestine, from oral cavity to anus. The bacteria were recovered and viable even after worms were in culture in saline for 6 days. The authors reported retrieval of one hookworm which also revealed two colonies of bacteria which were not V. cholerae. The authors reasoned that pathogenic bacteria may survive inside parasites when parasites are passed in the stool, as happens commonly during a cholera infection, thereby contributing to environmental spread of microbial pathogens. Further, the authors discussed the possibility of pathogenic microbes reaching the reproductive tract of female nematodes during copulation and adhering to egg shells which may also serve as a vehicle for transmission. There was no indication in this study that V. cholerae are pathogenic to Ascarids and further study in this area could reveal interesting associations between different enteric infections. Taken together, these findings highlight the clinical relevance of understanding interactions between intestinal nematodes and environmental microbes.

Walk et al. provided evidence for a H. p. bakeri-associated microbiota when they sampled L3 and adult nematodes (Walk et al., 2010), finding only 28 16S sequences obtained from larvae, concluding that they are associated with few bacteria; however, the L3-associated microbiota was completely unique and consisted of 6 bacterial families, unlike the adult-associated microbiota which was very similar to the ileum of the murine host. As H. p. bakeri hatch outside of the host, they likely enter their hosts with a distinct microbiota which changes over time in the new environment. This L3-specific microbiota may also serve as a source of non-indigenous microbes for the mammalian host.

A special microbe-helminth relationship is also well documented for filarial nematodes in which the gram-negative intracellular bacterium Wolbachia is obligatory for normal larval growth and development, embryogenesis, and survival of adult worms (Taylor et al., 2005). Such endosymbiotic relationships probably emerged from ancestral infections of the host nematode by free-living bacteria, concomitant with gene losses and genome rearrangements on both sides during coevolution (Masson et al., 2016). So far, such an endosymbiotic relationship has not been detected in other parasitic nematode species, but parallels to other microbenematode partnerships may be drawn. For example, Ascaris requires and absorbs Vitamin B12 from microbial sources (Zam and Martin, 1969). Wolbachia displays cell tropism and is restricted to somatic tissues in adult male worms, whereas females also harbor bacteria in the germline (Kozek, 1977; Taylor et al., 1999). The level of infection varies substantially during filarial development, where shortly after transmission to the mammalian host a dramatic increase in the bacterial population occurs (McGarry et al., 2004). As bacterial loads within individual worms differ, Taylor et al. hypothesized that higher levels of Wolbachia infection within a worm may potentially confer selective advantages in terms of filarial development or fecundity (Taylor et al., 2013). Transcriptomic analysis of the Wolbachia genome of Onchocerca ochengi indicated that Wolbachia may have a mitochondrion-like function in the soma, generating ATP for the nematode host (Darby et al., 2012). Hence, recent trials aim at utilizing antibiotics such as tetracycline (Hoerauf et al., 1999), and other chemotherapeutics targeting Wolbachia as a novel tool for the treatment of filarial infection and disease, reviewed in Taylor et al. (2014).

C. elegans nematodes isolated from the wild are observed to harbor live bacteria in their intestine which may exist as commensals or which may proliferate and cause obstruction and pathology of the worm gut (Félix and Duveau, 2012). Worms isolated from the wild also carry a distinct and diverse microbiota when compared to other nematode species in the same environment (Dirksen et al., 2016). Dirksen et al. reported a microbiota rich in unclassified Proteobacteria from the family Enterobacteriaceae, as well as members of the genera Pseudomonas, Stenotrophomonas, Ochrobactrum, and Sphingomonas. When isolated worms were enriched on plates of E. coli OP50, their microbiota maintained considerable similarity to freshly isolated samples despite 3 weeks in culture on E. coli plates, suggestive of a C. elegans-specific microbiota and closely developed microbial-host community. This study did not mention any changes in overall microbial population sizes over time. In the same study, a subset of 14 bacterial isolates were chosen in order to cultivate nematodes on an experimental microbiota with bacterial frequencies mirroring the nematodeassociated microbiota of worms isolated from the wild. Three genotypes of C. elegans, including two natural isolates (MY316 and MY379) and the laboratory N2 strain, were cultured from hatched sterilized eggs through to adulthood. The investigators found significant influences of genotype and life stage on the microbial composition of the nematodes. Interestingly, certain bacteria including Ochrobactrum MYb71 and Stenotrophomonas MYb57 were enriched in the nematode samples relative to the agar plates. The effect was quite pronounced in the case of Ochrobactrum which was present in only trace amounts in the agar plates but represented as much as 20% of the nematodeassociated bacterial community. Ochrobactrum was also able to persist in the intestine even under starvation conditions without being used as a food source or eliminated during the ensuing stress response, which can include upregulation of antimicrobial effectors such as lysozymes (Uno et al., 2013). This bacterial strain may be a prominent symbiont for C. elegans as it seems to use the nematode as an environmental niche in the absence of apparent fitness costs to the worm. The experimental microbiota was also shown to enhance growth and nematode population size relative to worms cultured on E. coli. While detailed analysis of this sort is absent for parasitic nematodes, the ability of microbes to colonize these worms supports the idea of wormassociated microbial communities, providing benefits similar to those seen in C. elegans and other organisms, such as supplying nutrients and providing protection from pathogens (Cabreiro and Gems, 2013; Watson et al., 2014; Lee et al., 2015; Dirksen et al., 2016).

### THE INFLUENCE OF BACTERIA ON PARASITIC NEMATODES

Living in intimate association with microbes, nematodes are subject to diverse microbial influences. Beneficial effects of microbes on nematodes may include nutrition, promotion of longevity, protection from infection by other microbes, and contributions to a hospitable environment (Cabreiro and Gems, 2013; MacNeil and Walhout, 2013). From the perspective of mammalian hosts, the question arises whether a specific microbial environment might significantly influence susceptibility to helminth infection. Further, microbes might also be a source of competition, stress, and disease for nematodes.

The microbial environment can profoundly impact establishment and propagation of parasitic nematode infections by influencing egg hatching and reproductive success. T. muris eggs may require the presence of selected bacterial species to induce hatching, as demonstrated using murine cecal explants as well as E. coli, S. typhimurium, S. aureus, and P. aeruginosa (Hayes et al., 2010). Live bacteria appear to be necessary for these effects as heat inactivation prevented hatching, while bacteriostatic gentamycin treatment did not. It was proposed that physical contact between the bacteria and eggs is required as Type 1 fimbriae facilitate E. coli-induced hatching, though additional mechanisms likely exist. The authors also reported that mice pre-treated with antibiotics had significantly lower worm burdens at day 21 pi compared to untreated animals, indicating a significant role for bacterial communities in parasitic nematode establishment. Interestingly, bacteriainduced T. muris egg hatching seems to occur efficiently with members of Proteobacteria (E. coli) and Firmicutes (Enterococcus caccae, Streptococcus hyointestinalis, Lactobacillus reuteri, and Lactobacillus amylovorus). It is possible that many different bacterial species contribute via different mechanisms for optimal T. muris egg hatching (Vejzagic et al., 2015 ´ ).

Lactobacillaceae abundance in the duodenum positively correlates with susceptibility to H. p. bakeri infection (Reynolds et al., 2014). Low-level vancomycin treatment prior to nematode infection did not significantly reduce total bacteria but elevated abundance of Lactobacillaceae and Enterobacteriaceae while reducing Eubacterium/Clostridium species in the fecal microbiota. This was associated with H. p. bakeri persistence in the host. H. p. bakeri infection also elevated duodenal Lactobacillaceae and Enterobacteriaceae. Administration of Lactobacillus taiwanensis enhanced susceptibility to H. p. bakeri infection and worm fecundity, thought to be due to the induction of immunosuppressive regulatory T cells. This study demonstrated a reciprocal interaction whereby Lactobacillus spp. promote nematode establishment which then promote growth of Lactobacilli. Similar to T. muris infections discussed above, studies in germfree mice revealed higher H. p. bakeri worm burdens in conventionally raised mice, implicating the host microbiota as a key part of the parasite's environmental niche (Wescott, 1968). Bacterial populations also influence the host's immune status which can have a profound influence on the intestinal environment for parasite establishment, though specific immune variables and bacterial species are still being identified (Cattadori et al., 2016). These data suggest not only that the microbiota is essential for parasite development, but also that particular bacteria facilitate helminth establishment.

While C. elegans and other free-living nematodes are known to subsist on microbes, the food sources of intestinal nematodes are less well established. The digested remains of yeasts can be found in the intestines of C. elegans isolated from the wild, demonstrating the capability of these nematodes to use eukaryotic cells as a food source in addition to bacteria (Félix and Duveau, 2012). There is corresponding evidence suggestive of consumption of intestinal epithelia by A. suum in pigs as well as H. p. bakeri in mice (Davey, 1964; Bansemir and Sukhdeo, 1994). Freshly isolated Ascaris nematodes were found to accumulate eukaryotic cellular material in their buccal cavities, thought to be of host origin (Davey, 1964). Labeling studies sought to determine whether H. p. bakeri accumulates ingesta, host blood, or host tissue material and found an accumulation of host tissue components rather than blood or ingesta (Bansemir and Sukhdeo, 1994). These studies did not assess the intestinal contents of the worms and these data do not preclude the ability of intestinal nematodes to digest bacterial cells, especially considering bacteria are counted amongst the intestinal contents of A. suum and A. lumbricoides as discussed previously. Further insights using modern methods could be provided by studies designed to specifically assess uptake and digestion of microbes by intestinal nematodes.

Bacteria can promote or hinder nematode proliferation by various means. From serving as direct food sources or by producing essential nutrients, microbes are required for nematode growth. Furthermore, specific species are well documented to promote helminth establishment while themselves being reinforced by nematodes, as discussed. One could imagine that intentional manipulation of microbial variables could influence parasite establishment and might possibly impact how parasitic diseases are treated in the future. In the case of C. elegans, characterizing its interactions with microbes beyond E. coli OP50 serves to improve this powerful model and enhances its utility to investigators with diverse objectives. Together, these diverse systems complement each other in elucidating the varied interactions between nematodes and microbes, thereby setting a stable foundation upon which future research can build.

### SENSING THE MICROBIAL ENVIRONMENT

Nematodes are confronted with a tremendous and highly diverse microbial environment, presenting many infectious challenges. Alterations in the intestinal microbiota during helminth infection, coupled with studies demonstrating distinctly beneficial or harmful outcomes to C. elegans physiology in response to various bacteria, illustrate the ability of nematodes to sense their microbial environments and possibly discriminate between microbial species. Parasitic nematodes must have evolved particular strategies to overcome colonization by microbes, likely employing selected antibacterial defense mechanisms to coexist with the host microbiota. C. elegans has become an important model organism for the study of innate immune defense against pathogenic bacteria (Schulenburg et al., 2004; Irazoqui et al., 2010). The C. elegans immune system is seemingly of ancient origin, and there exist homologs of nearly all of its components in other organisms, including humans (Schulenburg et al., 2008). Consequently, sensing of the microbial environment might follow conserved or comparable mechanisms in parasitic nematodes as well.

C. elegans uses different protective mechanisms when confronted with potential pathogens, such as avoidance behavior (Meisel and Kim, 2014) and activation of stress and immune responses (Samuel et al., 2016) leading to the production and release of defense molecules with antimicrobial activities (Mallo et al., 2002). Identification of microbe-associated molecular patterns is followed by signal transduction via several signaling cascades, culminating in transcriptional alterations, reviewed in Rosso et al. (2013) and Kim and Ewbank (2015). While the pathogen-recognition receptors responsible for initiating immune responses are not entirely known, candidates include F-box proteins, lectins, follicle stimulating hormone receptor homolog-1, scavenger receptors, and the Toll-like receptor TOL-1 (Kim and Ewbank, 2015). Stress and immune responses illustrate an overlap between microbial and abiotic stresses; the main responses involved include: autophagy, insulin-like receptor (DAF-2), mitogen-activated protein kinases (MAPK), transforming growth factor-β-like (DBL-1), programmed cell death pathways, and unfolded protein responses (UPR) (Kim and Ewbank, 2015). Perhaps surprisingly, C. elegans lacks a homolog of NF-κB, an essential transcription factor in the innate immune response of many invertebrate and vertebrate species (Vallabhapurapu and Karin, 2009). Other proteins regulating transcription of infection and stress-modulated genes have been identified, including cyclic AMP-dependent transcription factor-7 (ATF-7), forkhead box O (FOXO) ortholog DAF-16, GATA transcription factors (ELT-2, ELT-3), helix loop helix-30 (HLH-30), heat-shock factor-1 (HSF-1), NF-E2-related factor SKN-1, signal transducer and activator of transcription STA-2, X-box binding protein-1 (XBP-1), and basic leucine zipper domain transcription factor ZIP-2 (Kim and Ewbank, 2015). As indicated, the intricacies of these responses have been reviewed elsewhere; as such, only examples of particular interest are highlighted here.

### Pathogen Recognition

Infectious risks could be mitigated and resources preserved if pathogens were simply not encountered. Avoidance behavior is facilitated by the sole C. elegans toll-like receptor, TOL-1 (Pujol et al., 2001), and is characterized by the detection of specific microbial products, such as serrawettin W2 produced by Serratia marcescens, resulting in worms migrating away from the offending agents (Pradel et al., 2007). Unlike other organisms, C. elegans may not rely heavily on toll-like signaling for immune defense, though reports on this are conflicting (Pujol et al., 2001; Couillault et al., 2004; Tenor and Aballay, 2008). Other mechanisms such as aerotaxis also contribute to pathogen avoidance, as demonstrated by C. elegans avoidance of P. aeruginosa (Reddy et al., 2009). Lectins, carbohydratebinding proteins implicated in multiple facets of immunity in a diverse range of species, may sense or neutralize microbes (Zelensky and Gready, 2005). Finally, there is evidence for an important role for scavenger receptors such as cell death abnormal-1 (CED-1) and scavenger (SCAV) proteins in pathogen recognition and resistance (Nicholas and Hodgkin, 2004; Means et al., 2009). Comparable studies of differential activation of recognition receptors in parasites by microbial communities from different intestinal regions might provide insights into how parasitic nematodes recognize threats and whether or not activation of such pathways impacts their niche selection.

### Surveillance Immunity

In addition to sensing microbes and their toxins, organisms induce immunity-related genes due to disruptions of core physiologic processes by pathogens. C. elegans employs this surveillance immunity in response to inhibited translation, cellular damage, and mitochondrial stress, all of which can result from infection and exposure to microbial toxins (Pukkila-Worley, 2016). Transcriptional responses to these stresses include activation of immune signaling and xenobiotic metabolic pathways (Pukkila-Worley, 2016).

Toxins produced by several microbes, including diphtheria toxin from Corynebacterium diphtheria (Collier, 2001), cholix toxin from Vibrio cholerae (Jørgensen et al., 2008), and exotoxin A (ToxA) from P. aeruginosa (Dunbar et al., 2012), interfere with protein translation. When non-pathogenic E. coli engineered to express ToxA are fed to C. elegans, immune pathways required for resistance to the toxin, especially MAPK pathways, are upregulated by the worms (McEwan et al., 2012). Also, the aminoglycoside antibiotic hygromycin B blocks elongation of the amino acid chain during protein translation like ToxA (McEwan et al., 2012). Taken together, C. elegans likely responds to translation inhibition by ToxA rather than to the protein itself. An RNAi screen showed that disrupting core processes, including translation, activatesbasic leucine zipper domain transcription factor ZIP-2-dependent immune signaling, similar to responses seen during P. aeruginosa infection of C. elegans and following exposure to ToxA (Dunbar et al., 2012). More recently, another transcription factor called CCAAT-enhancer-binding-protein-2 (CEBP-2) was identified to act in concert with ZIP-2 to promote defense by C. elegans against P. aeruginosa, ToxA, and in response to inhibition of translation and other core processes (Reddy et al., 2016).

In addition to surveillance of disrupted physiologic processes, C. elegans also induces immune responses after injury of its epidermis by microbial infection and sterile wounding (Pujol et al., 2008a). During infection with D. coniospora, a fungal pathogen which damages the nematode cuticle, 4 hydoxyphenyllactic acid (HPLA) acts as a damage-associated molecular pattern recognized by the G protein-coupled receptor DCAR-1 which is required for AMP expression during fungal infection (Zugasti et al., 2014). Another study showed that epidermal damage liberates the STAT-like transcription factor STA-2, triggering AMP production (Zhang et al., 2015). These findings exemplify the ability of epithelial barriers to surveil physical damage and respond by activating innate immune pathways.

Microbial infection can often lead to mitochondrial stress; a survey of C. elegans responses to bacteria isolated from the nematode's natural environment found that 101 strains of the >550 isolates tested induced the mitochondrial stress reporter promoter hsp-6::GFP (Samuel et al., 2016). Interestingly, mitochondrial disruption also induces drug-detoxification enzymes belonging to the cytochrome P450 superfamily (CYPs) and enzymes involved in glucuronidation (Liu et al., 2014) as well as infection response gene-1 (irg-1) which is also induced during P. aerugionsa infection in a ZIP-2-dependent manner (Estes et al., 2010; Dunbar et al., 2012). Another toxin of P. aeruginosa, pyoverdin, disrupts the mitochondria of C. elegans. The nematodes respond by a specialized form of autophagy called mitophagy, whereby damaged mitochondria are degraded to resist P. aeruginosa-mediated killing (Kirienko et al., 2015).

By sensing various forms of cellular stress and damage, C. elegans is able to launch an integrated response which promotes survival and longevity (Pukkila-Worley, 2016). These responses can be induced by microbes and xenobiotics, activating immune and detoxification pathways (Melo and Ruvkun, 2012; Pukkila-Worley et al., 2014; Pukkila-Worley, 2016). As helminths may employ similar strategies to deal with microbial and xenobiotic stresses, the overlap between immune and drug detoxification responses is of particular interest. Various contributing factors have been identified for the development of anthelmintic resistance, including resistance alleles in detoxification genes, parasites not exposed to treatments, and underdosing (Vercruysse et al., 2011). The concept of surveillance immunity raises the notion that elements within the host microbiota prime helminths to resist anthelmintic treatments. Whether certain microbial species of the host intestine can promote drug resistance by activating xenobiotic detoxification within intestinal nematodes has not yet been studied.

### Signal Transduction Pathways and Transcriptional Responses

As illustrated by many studies, MAPK pathways have a fundamental function in C. elegans stress and immune responses, including the p38 PMK-1 and extracellular signal-regulated kinase (ERK) pathways. Mutations at multiple levels of the NSY-1-SEK-1-PMK-1 pathway increase susceptibility to P. aeruginosa without impacting growth on E. coli OP50 (Kim et al., 2002). PMK-1 is activated by the toll-interleukin-1 receptor domain adaptor protein TIR-1 which is required for resisting killing by numerous microbial pathogens (Couillault et al., 2004). Phosphorylation of the ATF-7 transcription factor by PMK-1 activates transcription of lectins and candidate antimicrobial peptides (AMPs), conferring intestinal resistance to various species, including P. aeruginosa, S. marcescens, and E. faecalis (Shivers et al., 2010). Small molecule-mediated stimulation of the PMK-1 pathway and subsequent activation of ATF-7 can also increase lifespan in the presence of E. faecalis (Pukkila-Worley et al., 2012). Growth of C. elegans on soil-derived non-pathogenic bacteria, Bacillus megaterium and Pseudomonas mendocina also enhanced resistance to P. aeruginosa in a PMK-1 dependent fashion (Montalvo-Katz et al., 2013). Additionally, the PMK-1 pathway is active in epidermal infections as shown by induced expression of AMPs in response to D. coniospora as well in response to epidermal injury (Pujol et al., 2008a,b). Another MAPK pathway involving ERK signaling has been shown to be protective against rectal infection by M. nematophilum, (Gravato-Nobre et al., 2005). Based on these observations, one could speculate that MAPK pathways are also essential for helminth survival amongst numerous microbes.

Insulin-like signaling via DAF-2 highlights the intersection between immunity and metabolism. DAF-2 signaling inhibits DAF-16 activation by preventing its localization to the nucleus. Loss of function mutations of daf-2 confer pathogen resistance and longevity by allowing activation of the DAF-16 transcription factor (Kenyon et al., 1993; Garsin et al., 2003). DAF-16 activation also increases resistance to non-microbial insults, such as heat stress and oxidative stress (Barsyte et al., 2001). Intriguingly, C. elegans can be primed to withstand pathogens and heat stress when exposed to pathogens during development, increasing the worm's lifespan (Leroy et al., 2012). DAF-16 activation appears to increase antimicrobial gene transcription as well as stress and detoxification genes, placing this transcription factor at the nexus of stress and immune signaling (McElwee et al., 2003). Autophagy, a lysosomal degradation pathway, is also implicated in longevity and its inhibition increases intracellular replication of S. typhimurium in C. elegans intestinal epithelia (Jia et al., 2009).

As intestinal nematodes are typically much larger than C. elegans, with female Ascaris worms growing up to 35 cm in length (Dold and Holland, 2011), the surface area available for attachment and infection by microbes is extensive. Furthermore, these nematodes are much longer-lived than C. elegans. Stress responses, especially autophagy, may be of particular importance in these organisms, though this area remains virtually unexplored. An intriguing study by Voronin et al. identified autophagy as a bactericidal mode of action in filarial worms targeting the bacterium Wolbachia (Voronin et al., 2012). The activation of autophagy coincided with the onset of rapid bacterial growth and expansion, showing that in spite of their mutualistic association, the nematode may recognize Wolbachia as a stressor and respond to regulate bacterial abundance (Taylor et al., 2013). Further studies assessing antibacterial responses by intestinal nematodes are largely lacking; however, two reports have shed some light on the issue. Adult female A. suum nematodes challenged with heat-inactivated E. coli respond by increasing transcription of two AMP families, the Ascaris suum antibacterial factors (ASABF) and the Cecropins (Pillai et al., 2003, 2005). Worms were challenged by injection of heat-killed bacteria into the pseudocoelom and appeared to demonstrate tissue-specific responses. While these data suggest an inducible defense system in A. suum, it would be relevant to study tissue-specific AMP expression using live bacterial cultures introduced externally, without wounding the worms, coupled with modern transcriptomic methods. Such an experiment would preserve pathogenassociated molecular patterns without invoking pathogenindependent, epithelial-damage responses, thereby allowing identification of transcriptional changes induced specifically by infection.

### EFFECTORS OF INNATE IMMUNE SIGNALING IN NEMATODES

After sensing microbes, typical responses by nematodes include the production and release of factors possessing antimicrobial activity (**Table 3**). Among the different effectors of innate immunity, AMPs constitute the most ancient gene-encoded antimicrobial tools of eukaryotes (Zasloff, 2002). Nematodes produce different families of AMPs, many of which have been well characterized in C. elegans, while some have also been identified in parasitic nematodes. Antimicrobial activity has been experimentally demonstrated for many of these factors. AMPs can be classified based on their amino acid constituents and structural characteristics: α-helices, β–sheets, extended, and


Yes, detected; -, not detected; \*, lectin-like activity detected; gray shading, demonstrated bactericidal activity.

looped (Melo et al., 2009). Many of the best-characterized AMPs tend to be small, cationic, amphipathic molecules thought to act by membrane disruption. While many candidate immune effectors have been proposed in the literature for C. elegans, reviewed in Kim and Ewbank (2015), here we focus on effectors with corresponding data from parasitic nematodes. AMPs such as antibacterial factors and cecropins are discussed along with larger proteins involved in the immune response including lectins, nemapores, and lysozymes. We will then conclude this section by discussing antimicrobial activities of helminth-derived products.

### Antibacterial Factors

First identified in A. suum, ABFs are present in at least 25 nematode species, including seven ASABFs produced by Ascaris and six ABFs produced by C. elegans (Tarr, 2012). Sequence analysis reported by Tarr shows that other parasitic nematode species expressing ABFs include the human hookworm N. americanus, the ruminant nematode Haemonchus contortus, and the rodent model parasite N. brasiliensis (Tarr, 2012). Body fluid isolated from the pseudocoelem of A. suum demonstrated bactericidal activity, particularly against gram positive organisms (Wardlaw et al., 1994; Kato, 1995). Kato et al. isolated ASABFα (Kato and Komatsu, 1996) and subsequently demonstrated broad-spectrum antibacterial and weak antifungal activity of a recombinant form of the peptide (Zhang et al., 2000). The observed activity was rapid, killing S. aureus in under 1 min, and could be inhibited by increasing salt concentrations (Zhang et al., 2000). Andersson et al. isolated ASABF-β and γ from A. lumbricoides which also showed antimicrobial activity, especially against gram positives (Andersson et al., 2003). ABFs were the first AMPs described in C. elegans, based on sequence similarity to their Ascarid counterparts and also demonstrated antimicrobial activity (Kato et al., 2002). ABFs belong to the cysteine-stabilized-αβ (CS-αβ) group of AMPs, are cationic, and contain a cysteine-stabilized α-helix and two β-sheets (Kato and Komatsu, 1996; Tarr, 2012). Transcripts for ASABFs have been detected in all tissues of A. suum, with some members upregulated in the body wall, intestine, and ovaries after bacterial challenge (Pillai et al., 2003). Expression patterns and anatomical localization of ABFs in C. elegans are indicative of roles in defense and digestion (Kato et al., 2002). For example, ABF-1 and 2 are detected in the pharynx, and ABF-1 and 3 in the intestine of C. elegans. The expression of some ABFs is inducible by different pathogens: ABF-2 by S. typhimurium (Alegado and Tan, 2008), ABF-1 and 2 by Cryptococcus neoformans, and ABF-3 by S. aureus (Alper et al., 2007).

### Cecropins

In contrast to ABFs which seem to be widely distributed amongst nematodes, the cecropin family of AMPs has only been found in three Ascarids: A. suum, A. lumbricoides, and Toxocara canis (Pillai et al., 2005; Tarr, 2012). The first described nematode cecropin, Cecropin P1, was originally thought to be of porcine origin due to its isolation from pig intestine (Lee et al., 1989). Subsequently, its production was correctly attributed to A. suum (Andersson et al., 2003). Andersson et al. also isolated Cecropin P1 from A. lumbricoides and demonstrated antibacterial activity against gram negative organisms (Andersson et al., 2003). Pillai et al. demonstrated broad-spectrum antimicrobial activity of all four Ascaris cecropins as well as bacterial inducibility as detected for ASABFs (Pillai et al., 2005). Cecropins are typically 31–39 amino acids in length, strongly basic, cationic, α-helical peptides which are thought to disrupt microbial membranes by first laying on the membrane surface before reorientation and insertion into the membrane leading to disrupted lipid packing and subsequent membrane disintegration (Sipos et al., 1992; Gazit et al., 1995).

### Lectins

Proteins containing lectin domains are conserved across Metazoans, including nematodes (Zelensky and Gready, 2005). Harcus et al. isolated one C-type lectin (CTL) from H. p. bakeri and two from N. brasiliensis (Harcus et al., 2009). Moreover, lectin-like carbohydrate-binding activity has previously been reported in the intestine of A. suum (Cuperlovic et al., 1987 ´ ) and lectins have been identified in excreted and secreted products (ESPs) of A. suum larvae (Wang et al., 2013). In general, CTLs are capable of binding carbohydrates and function in pathogen recognition and neutralization (Zelensky and Gready, 2005). CTLs from H. p. bakeri and N. brasiliensis are predominantly expressed by gut-dwelling adults rather than larvae; while Harcus et al. made a compelling case for host-immunomodulation by their reported CTLs, a dual role in pathogen neutralization or recognition is worth testing for parasite-derived lectins (Harcus et al., 2009). Parasite-derived CTL molecules localize in the nematode cuticle but could also be found in ESPs (Page et al., 1992). The CTLs detected in parasitic nematodes were most similar to the C. elegans-CTLs clec-48, -49, and -50. Clec-50 is expressed in the C. elegans intestine and upregulated in response to the bacterium S. marcescens (Mallo et al., 2002). Though Harcus et al. did not report antimicrobial activity for the parasite-derived molecules, studies in C. elegans suggest they may have a role in defense. In addition to clec-50, clec-49 is also upregulated in response to S. marcescens (Engelmann et al., 2011) and mutants deficient in this protein and another CTL, clec-39, have increased susceptibility to, and reduced survival and fecundity during, S. marcescens infection compared to wild-type worms (Miltsch et al., 2014). Furthermore, recombinant clec-39 and clec-49 bind S. marcescens in the absence of bactericidal activity (Miltsch et al., 2014). Additionally, in response to M. nematophilum, CTLs were found to be the most upregulated protein class in C. elegans (O'Rourke et al., 2006). While nematode-derived lectins have not yet been shown to possess bactericidal activity, the mammalian lectin RegIIIγ is antibacterial and promotes commensalism by maintaining segregation between the microbiota and the host epithelium in mice (Vaishnava et al., 2011). Though many questions remain, the limited evidence available is suggestive of a role for lectins during infectious challenges and it is possible that parasitederived lectins may confer resistance to different bacteria as has been demonstrated for C. elegans.

### Lysozymes

Lysozymes are polysaccharide hydrolases, targeting bacterial peptidoglycan leading to cell lysis (Ellison and Giehl, 1991; Monzingo et al., 1996), and have been widely found in different nematode species. Lysozymes are encoded in the genomes of Ascarids, N. americanus, H. contortus, N. brasiliensis, Brugia malayi, and Wuchereria bancrofti, though missing from the Trichurids included in the sequence analysis by Tarr (2012). Lysozymes were also detected in the ESPs of A. suum (Wang et al., 2013) and H. p. bakeri (Hewitson et al., 2011), though further study is required to determine antibacterial activities of these helminth-lysozymes. 15 lysozymes were detected in C. elegans, divided into 10 protist-type (lys-1–lys-10) and five invertebratetype (ilys-1–ilys-5), so named for sequence similarities to other organisms (Schulenburg and Boehnisch, 2008). Some lysozymes are markedly upregulated by C. elegans in response to bacterial infection (Mallo et al., 2002; O'Rourke et al., 2006; Gravato-Nobre et al., 2016). Lysozymes expressed in the C. elegans intestine (Schulenburg and Boehnisch, 2008) have been proposed to act in concert with caenopores (discussed in the next section)

FIGURE 2 | Mutual influences of intestinal nematodes and host-gut bacteria. Establishment and persistence of intestinal nematodes in the host's gut are affected by bacterial communities and lead to substantial changes of the gut microbiota. Here, nematode-microbiota interactions and their impact on the host immune response and physiology are exemplified. (1) Egg hatching: Interaction of eggs with the intestinal microbiota is needed for some species, enabling larvae to hatch. (2) Mucosal immune response: Attachment to the epithelium and tissue migratory phase might lead to bacterial translocation and manipulation of immune responses. The anti-helminth immune response is predominantly a T helper type 2 response. (3) Gut physiology: Immune responses induce changes in gut physiology via induction of goblet cell hyperplasia, mucus production, and epithelial turnover, leading to changes in the host microbiota and its metabolome. Specific subsets of bacteria directly influence host physiology through their metabolic activities (e.g., short chain fatty acids-producing bacteria). (4) Microbiota composition: Intestinal nematodes modify intestinal microbial communities via different mechanisms: a) directly via secretion of antibacterial molecules and/or excretory/secretory products, b) indirectly by metabolic and physiologic shifts influencing the gut milieu. Chronic infections often lead to reduction in bacterial diversity and outgrowth of specific bacterial species beneficial for parasite survival. (5) Host metabolism: Intestinal nematodes modify host metabolism and nutrient uptake, e.g., alter amino acid, fatty acid, and carbohydrate metabolism, with subsequent influence on gut physiology, immune reactivity, and intestinal microbiota composition. (6) Nematode's microbiota: Nematode-associated bacterial communities might reflect the host microbiota and may also serve as transmission vehicles for pathogenic bacteria. Th2 cell, T helper type 2 cell; AAM, alternatively activated macrophage.

Midha et al. Nematode-Microbe Cross-Talk

in digestion and immunity (Bányai and Patthy, 1998). For example, the C. elegans invertebrate lysozyme ILYS-3 is needed for physiological pharyngeal grinder function and for defense against bacterial pathogens (Gravato-Nobre et al., 2016). It was also shown that ILYS-3 is induced by danger signals generated both by bacterial pathogens and starvation (Gravato-Nobre et al., 2016). In healthy C. elegans, intestinal ilys-3 expression undergoes a post-developmental regulatory oscillation: levels increase after L1 hatching, decline after the L2 transition, and increase again after L4 transition becoming abundantly expressed in the intestine of adult worms (Gravato-Nobre et al., 2016). Notably, coelomocytes of C. elegans also express ilys-3 (Gravato-Nobre et al., 2016) and are scavenger cells that endocytose fluid from the pseudocoelom (Fares and Greenwald, 2001). The role of coelomocytes in parasitic nematodes is not well understood and they have not been studied in immunity. Given the demonstrated importance of lysozymes for C. elegans and the wide distribution of this protein family across species, it is expected that they support helminth survival amongst the host-microbiota.

### Nemapores

Known as caenopores in C. elegans, nemapores contain a saposin domain and share similarity with protozoan amoebapores and mammalian NK-lysin and granulysin (Leippe, 1995; Roeder et al., 2010). As with ABFs, nemapores are also cysteine-rich, but contain multiple α-helices (Mysliwy et al., 2010). Nemapores are widely distributed and in addition to C. elegans, have been found in 46 nematode species including Ascarids, Trichurids, and the human filarial worms B. malayi and W. bancrofti (Tarr, 2012). Only caenopores have been studied experimentally, with caenopore-1 and 5 demonstrating antimicrobial activity (Bányai and Patthy, 1998; Roeder et al., 2010). Caenopore-5 is constitutively expressed in the intestine, while caenopore-3 is induced by starvation as well as the bacterial strains B. megaterium and Micrococcus luteus, suggesting functions in nutrition and immune defense (Roeder et al., 2010). Interestingly, the free-living soil bacteria B. megaterium and M. luteus are both frequently found with Rhabditid nematodes (Coolon et al., 2009) and may be present in the environment of C. elegans. As with lysozymes, experimental evidence of nemapore function in parasitic nematodes is currently lacking.

### Antimicrobial Activities of Helminth-Derived Products

In addition to studies demonstrating antimicrobial activities of specific helminth-derived molecules described here, activity can also be detected from the ESPs of parasites. In one study, ESPs from the filarial worm O. ochengi were analyzed revealing 36 candidate AMPs (Eberle et al., 2015). Of the 36 candidates, the investigators were able to unambiguously attribute bactericidal activity against E. coli to three peptides while the other 33 peptides had been assessed as part of peptide mixtures. Antibacterial activity has also been shown for ESPs from adult T. suis worms (Abner et al., 2001). Different bacterial strains were tested, including S. aureus, E. coli, and C. jejuni. The observed activity was shown to be due to small (<10 kDa), boiling-resistant molecules, indicative of AMPs. Interestingly, pore-forming proteins have also been isolated from T. trichiura and T. muris, indicating that antimicrobial activity of nematode ESPs is likely due to multiple components of varying sizes and mechanisms (Drake et al., 1994). As nematodes produce a variety of antimicrobial factors, innate immune responses to microbes by helminths likely offer protection using several different strategies.

### CONCLUSIONS AND PERSPECTIVES

Antibiotic resistance is a rapidly escalating problem with devastating consequences for human and animal health. Drug resistance mechanisms described for microbes and helminths foreshadow the emergence of immense clinical and economic challenges associated with previously treatable infectious diseases. Meanwhile, inflammatory and metabolic diseases also inflict great suffering and are a source of considerable strain on healthcare systems; therefore, novel solutions are sought after to combat these difficulties. As nematodes have evolved over millions of years in a diverse microbial environment, a better understanding of how nematodes interact with microbial populations may offer innovative strategies for treating human and animal diseases. Given the importance of the microbiota and the ability of helminths to influence microbial communities in the gut (**Figure 2**), helminth infections and helminth products are being studied for their role in dysbiosis. Preliminary data suggest potential benefits of helminths in inflammatory diseases and in individuals with metabolic abnormalities. An additional application of the impact of nematodes on microbes is the development of new antimicrobials. AMPs are under investigation for their therapeutic potential and an understanding of how antimicrobial effectors are used by organisms living amongst diverse microbial populations can guide development efforts. In elucidating the impacts of different bacteria on nematodes, one can anticipate the discovery of novel anthelmintic targets. Moreover, these studies also provide insights into conserved innate immune mechanisms. The divergent environments and experimental systems of parasitic and free-living nematodes offer unique tools to investigators, the combination of which enhance our comprehension of nematode biology and evolutionary ecology.

## AUTHOR CONTRIBUTIONS

All authors listed have made substantial contributions to the planning and writing of this work. All authors have approved this work for publication.

### FUNDING

The work was supported by the German Research Foundation: GRK 2046 (AM and SH) and HA2542/3-2 (SH).

### REFERENCES


not associated with alterations in the faecal microbiota. PLoS ONE 8:e76573. doi: 10.1371/journal.pone.0076573


against Salmonella and Yersinia infections. Lett. Appl. Microbiol. 61, 523–530. doi: 10.1111/lam.12478


bacterium Serratia marcescens by Caenorhabditis elegans. Proc. Natl. Acad. Sci. U.S.A. 104, 2295–2300. doi: 10.1073/pnas.0610281104


**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 Midha, Schlosser and Hartmann. 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.

# Cytokines and Chemokines in Cerebral Malaria Pathogenesis

Josefine Dunst 1, 2, 3 \*, Faustin Kamena1, 2, 3 and Kai Matuschewski 1, 2

*<sup>1</sup> Parasitology Unit, Max Planck Institute for Infection Biology, Berlin, Germany, <sup>2</sup> Institute of Chemistry and Biochemistry, Free University, Berlin, Germany, <sup>3</sup> Molecular Parasitology, Institute of Biology, Humboldt University, Berlin, Germany*

Cerebral malaria is among the major causes of malaria-associated mortality and effective adjunctive therapeutic strategies are currently lacking. Central pathophysiological processes involved in the development of cerebral malaria include an imbalance of pro- and anti-inflammatory responses to *Plasmodium* infection, endothelial cell activation, and loss of blood-brain barrier integrity. However, the sequence of events, which initiates these pathophysiological processes as well as the contribution of their complex interplay to the development of cerebral malaria remain incompletely understood. Several cytokines and chemokines have repeatedly been associated with cerebral malaria severity. Increased levels of these inflammatory mediators could account for the sequestration of leukocytes in the cerebral microvasculature present during cerebral malaria, thereby contributing to an amplification of local inflammation and promoting cerebral malaria pathogenesis. Herein, we highlight the current knowledge on the contribution of cytokines and chemokines to the pathogenesis of cerebral malaria with particular emphasis on their roles in endothelial activation and leukocyte recruitment, as well as their implication in the progression to blood-brain barrier permeability and neuroinflammation, in both human cerebral malaria and in the murine experimental cerebral malaria model. A better molecular understanding of these processes could provide the basis for evidence-based development of adjunct therapies and the definition of diagnostic markers of disease progression.

#### Edited by:

*Slobodan Paessler, University of Texas Medical Branch, United States*

#### Reviewed by:

*Tracey Lamb, University of Utah, United States Teresa Carvalho, La Trobe University, Australia*

\*Correspondence:

*Josefine Dunst dunstj@zedat.fu-berlin.de*

Received: *23 February 2017* Accepted: *03 July 2017* Published: *20 July 2017*

#### Citation:

*Dunst J, Kamena F and Matuschewski K (2017) Cytokines and Chemokines in Cerebral Malaria Pathogenesis. Front. Cell. Infect. Microbiol. 7:324. doi: 10.3389/fcimb.2017.00324* Keywords: malaria, Plasmodium, cerebral malaria, cytokines, chemokines, endothelial activation, blood-brain barrier, neuroinflammation

### INTRODUCTION

Malaria is one of the most prevalent infectious diseases worldwide and contributes considerably to the global disease burden. With ∼200 million new cases and an estimated 430,000 deaths annually (WHO, 2016), malaria remains the most important vector-borne infectious disease. The human-adapted Plasmodium species P. falciparum and P. vivax account for the majority of malaria cases and are transmitted by the bite of an infective Anopheles mosquito. Despite considerable progress in malaria eradication over the past 15 years (WHO, 2016),

**Abbreviations:** CD, cluster of differentiation; CM, cerebral malaria; CSF, cerebrospinal fluid; DAMP, danger-associated molecular patterns; DC, dendritic cell; ECM, experimental cerebral malaria; EPCR, endothelial protein C receptor; GPI, glycosylphosphatidylinositol; ICAM, intercellular adhesion molecule; IFN, interferon; IL, interleukin; IP-10, IFN-γ-inducible protein 10; KC, keratinocyte chemoattractant; LPS, lipopolysaccharide; LT, lymphotoxin; MCP-1, monocyte chemoattractant protein 1; MIG, monokine induced by IFN-γ; MIP, macrophage inflammatory protein; NK, natural killer; NLR, NOD-like receptors, PAMP, pathogen-associated molecular pattern; PF4, platelet factor 4; Pf EMP1, Plasmodium falciparum erythrocyte membrane protein 1; PRR, pattern recognition receptor; TLR, Toll-like receptor; TNF, tumor necrosis factor; VCAM, vascular cell adhesion molecule; ZO-1, zonula occludens 1.

efforts are hampered by emerging Plasmodium resistance to commonly used anti-malaria drugs (Cui et al., 2015) and limited efficacy of the currently most advanced vaccine candidate (White et al., 2015; Olotu et al., 2016).

Among infections by human host-adapted Plasmodium species, P. falciparum infections are most likely to progress to organ-related pathology and severe malaria, and thereby account for the vast majority of malaria-associated fatalities. Along with severe anemia and respiratory distress, cerebral malaria (CM) is one of the major manifestations of severe malaria (Haldar et al., 2007). CM manifests with impaired consciousness and coma in both children and adults (Idro et al., 2010), while other clinical features differ. In addition to the characteristic diffuse encephalopathy, retinal abnormalities are frequent in children and less common in adults with CM (Beare et al., 2006; Idro et al., 2010). In contrast, CM in adults is accompanied by multiorgan disorder including renal failure and pulmonary edema, which are less frequently observed in children suffering from CM (Idro et al., 2010). Although anti-malaria treatment using artesunate was reported to improve CM outcome in children and adults (Dondorp et al., 2005, 2010), the case-fatality rate of pediatric CM is approximately 20% (Haldar et al., 2007) and sustained cognitive and/or neurological impairment may occur (John et al., 2008a). Consequently, treatment strategies, which not only target the parasite but also other mechanisms underlying CM pathogenesis, need to be developed. Since the pathogenesis of CM is still incompletely understood, further investigations are an important medical research priority, especially in the context of adjunctive therapies. Since accumulating evidence indicates that an imbalance in pro- and anti-inflammatory immune responses partially contributes to CM pathogenesis, such therapeutic approaches could target cytokines and chemokines associated with CM severity. Cytokines are polypeptides, which mediate and generate inflammatory responses. Along with their contribution to disease pathogenesis in general, cytokines also exert physiological roles at lower concentrations (Clark and Vissel, 2017). Chemokines are chemotactic cytokines, which recruit lymphocytes and monocytes to the site of pathogen encounter by binding to their respective chemokine receptor (Griffith et al., 2014). Given that leukocytes were found to sequester in the microvasculature of the brain in human CM and murine ECM (Hunt and Grau, 2003), local chemokine gradients may mediate leukocyte recruitment and thus promote CM pathogenesis. Chemokines exert their function through binding to their respective G protein-coupled chemokine receptors (**Table 1**), which induces activation of phosphatidylinositol 3 kinase (PI3K) and Rho GTPase signaling pathways, thus leading to F-actin polymerization and migration (Viola and Luster, 2008).

In this review, we highlight findings from both experimental murine models and natural human infections, and assess the current knowledge on the role of host cytokine and chemokine responses in the severe malaria complication of cerebral malaria. We also emphasize the potential inflammatory cascade resulting from Plasmodium life cycle progression after sporozoite inoculation and ultimately culminating in cerebral malaria pathology (**Figure 1**).

### CURRENT CONCEPTS IN CEREBRAL MALARIA PATHOGENESIS

Two central concepts to explain CM pathogenesis have evolved and they are likely mutually dependent- the vascular occlusion hypothesis and the inflammation hypothesis (Storm and Craig, 2014).

The concept of vascular occlusion leading to CM is based on the ability of mature P. falciparum-infected erythrocytes to sequester in the microvasculature through binding of P. falciparum erythrocyte membrane protein 1 (Pf EMP1) present on the erythrocyte surface to endothelial cell surface proteins, such as intercellular adhesion molecule 1 (ICAM-1), vascular cell adhesion molecule 1 (VCAM-1), cluster of differentiation 36 (CD36), or endothelial protein C receptor (EPCR) (Pasloske and Howard, 1994; Chen et al., 2000; Rowe et al., 2009; Smith et al., 2013; Turner et al., 2013; Lennartz et al., 2017). Sequestration occurs in various organs and, along with increased rigidity of erythrocytes, is believed to cause vascular occlusion (Dondorp et al., 2004). Additionally, microvascular obstruction during P. falciparum infection may be worsened by the formation of rosettes and clumps (Chen et al., 2000; Rowe et al., 2009; Adams et al., 2014), i.e., the binding of uninfected erythrocytes by infected erythrocytes (Handunnetti et al., 1989), and aggregation of infected erythrocytes and platelets (Pain et al., 2001), respectively. These events may cause a reduction in microvascular blood flow, ischemia, and tissue hypoxia (Medana and Turner, 2006), thereby accounting for cerebral pathology. Reduced vessel perfusion and occlusion was indeed observed by fluorescein angiography of the retina in pediatric CM cases (Beare et al., 2009) and by in vivo imaging of the microcirculation in adult patients with CM or other manifestations of severe malaria (Dondorp et al., 2008), although further studies are needed to determine the underlying cause of these observations. Additionally, sequestration of P. falciparum-infected erythrocytes was observed in the cerebral microvasculature of CM patients in post mortem brain histology studies (Ponsford et al., 2012; Milner et al., 2014, 2015).

Despite accumulating evidence, the relevance of sequestration in the development of CM is still incompletely understood, since the degree of sequestration in brains of non-fatal CM cases cannot be investigated non-invasively and is, thus, unspecified (Miller et al., 2002). Occasionally, fatal CM cases present little sequestration and vessel occlusion similar to non-CM severe malaria cases (Ponsford et al., 2012). Moreover, isolated CM cases were reported in children and adults with confirmed P. vivax mono-infections in India (Kochar et al., 2009), although P. vivax is unlikely to sequester in the microvasculature since late-stage P. vivax-infected erythrocytes are present in peripheral blood (Anstey et al., 2009). Together, cerebral malaria occasionally develops in a few P. vivax infections without obvious signs of sequestration in vivo or microvascular obstruction.

Given the unresolved role of sequestration in the pathogenesis of CM, additional factors may determine disease severity. Notably, other infectious diseases that result in systemic inflammation and fever also progress to severe forms,



*<sup>a</sup>Modified from Zlotnik and Yoshie (2012) and Griffith et al. (2014).*


including neurological complications such as sepsis-associated encephalopathy (De Backer et al., 2002; Clark et al., 2004). Strikingly, systemic cytokine levels have been described to correlate with disease severity in malaria as well as sepsis (Prakash et al., 2006; Bozza et al., 2007). These findings corroborate an earlier proposal that an imbalance in pro- and anti-inflammatory immune responses triggers immune-induced pathology and might be a leading cause of CM pathogenesis, which may be further amplified by sequestration (Clark and Rockett, 1994). In addition to inflammation and sequestration, CM is associated with endothelial activation and increased blood-brain barrier permeability, and these processes might act reciprocally and have synergistic effects (van der Heyde et al., 2006). In line with this notion, certain Pf EMP1 variants, which are associated with CM, were described to compete with activated protein C in binding to EPCR (Turner et al., 2013; Bernabeu et al., 2016). Therefore, the anti-coagulant, anti-inflammatory, cytoprotective properties, which are induced upon interaction of protein C with EPCR, might be impeded by Pf EMP1-EPCR interaction and consequently, disease mechanisms may be further exacerbated (Bernabeu and Smith, 2017; Wassmer and Grau, 2017). However, the impact of the interaction between Pf EMP1 and EPCR on inflammation and coagulation remains to be demonstrated.

Insights into the mechanisms underlying CM in humans are limited and mostly based on post mortem histopathology or correlations of serum parameters with disease outcome (Hunt and Grau, 2003). Despite potential differences in human and murine CM pathogenesis (Riley et al., 2010; Craig et al., 2012), P. berghei (strain ANKA) infection reliably causes signature symptoms of CM in susceptible C57BL/6 mice (de Souza and Riley, 2002; Hunt and Grau, 2003). This host-parasite combination is a widely used and well-established murine model for CM, termed experimental cerebral malaria (ECM), which permits mechanistic studies (Craig et al., 2012). Consequently, studies highlighted in this review include reports on human cerebral malaria cases combined with mechanistic insights based on the murine ECM model and in vitro studies.

### INNATE IMMUNE ACTIVATION DURING LIVER STAGE DEVELOPMENT

Upon transmission by the bite of an infective Anopheles mosquito, Plasmodium sporozoites rapidly migrate to the liver, invade hepatocytes and develop into thousands of merozoites. Since this developmental stage of Plasmodium parasites is clinically silent, the immune response mounted by the host in order to limit parasite expansion during liver stage development remains largely unexplored (Hafalla et al., 2011). However, it seems likely that sporozoite and liver stage recognition primes the innate immune system locally and well below the pyrogenic threshold. In fact, innate immune cells have occasionally been observed to surround P. berghei-infected hepatocytes, indicating that Plasmodium does not remain undetected during liver stage development (Liehl and Mota, 2012). Additionally, a type I interferon (IFN) response is induced in livers of mice infected with P. yoelii or P. berghei (Liehl et al., 2014; Miller et al., 2014). Such an initial type I IFN response might induce chemokines, including IFN-γ-inducible protein 10 (IP-10)/CXCL10, which in turn could recruit cells expressing the corresponding chemokine receptor CXCR3, such as T, natural killer (NK), and NKT cells, to the site of infection, which might contribute to the local immune response by IFN-γ secretion (**Figure 2**; Liehl et al., 2014; Miller et al., 2014). In good agreement, an increase in IFN-γ plasma concentration prior to onset of detectable bloodstage infection was reported in controlled human P. falciparum infection (Hermsen et al., 2003). Clearly, the initial cytokine response fails to arrest liver stage development and, thus, does not curtail the proceeding to erythrocyte infection. Interestingly, P. berghei sporozoite and blood stage infections result in ECM symptoms in a similar time frame (Kordes et al., 2011), suggesting that immune responses against liver stages might not modulate CM pathogenesis. Whether this first immune response reduces

FIGURE 2 | Innate immune response to *Plasmodium* liver stage infection. *Plasmodium* infection of hepatocytes activates interferon regulatory factors (IRF), which induce transcription of type I interferons (IFN) IFNα and IFNβ. Secretion of type I IFNs activates IFNα/β receptor IFNAR in an autocrine or paracrine manner. IFNAR signaling results in transcription of IFN-stimulated genes (ISGs), which includes chemokines, such as CXCL9 and CXCL10. Upon secretion from hepatocytes, these chemokines might recruit cells expressing the corresponding chemokine receptor CXCR3, including natural killer (NK), T, and NKT cells. Upon activation by type I IFN at the site of infection, these cell types could contribute to limiting *Plasmodium* liver stage expansion by IFN-γ secretion. Based on Liehl et al. (2014) and Miller et al. (2014).

the initial parasite number released into the blood stream and thereby influences the magnitude of early blood stage-induced immune responses remains to be tested.

### BLOOD STAGE-INDUCED INNATE IMMUNE RESPONSES

Concomitant with the release of merozoites into the bloodstream and infection of erythrocytes, a Plasmodium infection progresses from the clinically silent liver stage to the symptomatic blood stage. Merozoites are only very briefly (∼60 s) exposed to the immune system before they rapidly enter new erythrocytes (Gilson and Crabb, 2009; Beeson et al., 2016), and blood stage infection is the exclusive cause of malaria symptoms, which is associated with systemic inflammation and fever. Fever is a common and effective host defense against microbial pathogens and swiftly initiated upon the first host-pathogen interaction. The febrile response is likely triggered through a universal mechanism, in which pyrogens, such as the proinflammatory cytokines interleukin 1α (IL-1α), IL-1β, IL-6, or tumor necrosis factor (TNF), are secreted by innate immune cells upon recognition of pathogen-associated molecular patterns (PAMPs) or host-derived danger-associated molecular patterns (DAMPs) by pattern recognition receptors (PRRs) (Evans et al., 2015). In Plasmodium infection, the characteristic recurrent fever coincides with synchronized rupture of infected erythrocytes in the schizont stage (Oakley et al., 2011). Release of parasite- and host-derived molecules due to erythrocyte rupture was described to induce TNF in vitro (Kwiatkowski et al., 1989; Bate and Kwiatkowski, 1994), and peaks in TNF serum concentration were found to coincide with elevated body temperature during P. vivax infection (Karunaweera et al., 1992), indicating that malaria fever is elicited by repeated release of PAMPs and DAMPs. Although an increase in core body temperature is associated with resolution of infection (Oakley et al., 2011), such a proinflammatory immune response needs to be counterbalanced by anti-inflammatory mechanisms in order to avoid a dysregulated immune response, which might lead to complications such as cerebral malaria.

The innate immune system represents the first line of defense against pathogens and mediates recognition and clearance of Plasmodium parasites (**Figure 3**). Cells of the innate immune system such as macrophages and dendritic cells (DCs) as well as non-professional immune cells such as endothelial cells and fibroblasts express PRR. These include Toll-like receptors (TLR), C-type lectin receptors (CLR), Retinoic acid-inducible gene (RIG)-I-like receptors, and NOD-like receptors (NLR), which recognize PAMPs and host-derived DAMPs (Takeuchi and Akira, 2010). Most malaria PAMPs and DAMPs known so far are apparently recognized by TLR. Activation of TLR initiates a signaling cascade including the adaptor protein MyD88 and the transcription factors NF-κB, AP-1, and interferon regulatory factor (IRF). As a consequence, expression of genes encoding pro-inflammatory cytokines such as type I IFN, IFN-γ, IL-6, IL-12, and TNF, is induced (Eriksson et al., 2013; Gazzinelli et al., 2014).

erythrocytes from the circulation by phagocytosis. In macrophages, uptake of infected erythrocytes might not lead to secretion of pro-inflammatory cytokines due to phagosomal acidification (Wu et al., 2015). Upon rupture of infected erythrocytes, pathogen-associated molecular patterns (PAMPs) and danger-associated molecular patterns (DAMPs) are released, including microvesicles, hemozoin, and glycosylphosphatidylinositols (GPI). These potential PAMPs and DAMPs might be recognized by DC through pattern recognition receptors, resulting in the secretion of interleukin 12 (IL-12), tumor necrosis factor (TNF), and IL-6 (Wu et al., 2015). DC-derived IL-12 might activate natural killer (NK) cells, which in turn secrete interferon γ (IFN-γ) and could thereby activate macrophages (Stevenson and Riley, 2004).

Several Plasmodium-derived molecules have been recognized as malaria PAMPs based on their ability to induce cytokine responses in vitro (**Figure 3**). One of the candidate malaria PAMPs are glycosylphosphatidylinositols (GPI). Although GPI are present in all eukaryotic cells where they serve as membrane anchors for certain cell surface proteins, Plasmodium GPI are structurally distinct from human GPI (Gowda, 2007). Consequently, Plasmodium GPI moieties may be recognized by the host immune system. Indeed, P. falciparum GPI were reported to induce the pro-inflammatory cytokines TNF and IL-1β in murine macrophages in vitro (Schofield and Hackett, 1993; Tachado et al., 1996). Moreover, purified GPI immobilized on gold particles elicited pronounced TNF responses from murine macrophages in vitro (Krishnegowda et al., 2005; Zhu et al., 2009, 2011), which was attributed to recognition of P. falciparum GPI through TLR2 or heterodimers of TLR2/1 and TLR2/6 (Krishnegowda et al., 2005; Zhu et al., 2011). However, given that TLR-deficiency did not impair immune responses elicited by P. berghei ANKA in vivo and did not protect mice from ECM (Togbe et al., 2007; Lepenies et al., 2008), the mechanism of GPI-induced innate immune activation in vivo remains to be determined.

Another potential malaria PAMP is hemozoin, an insoluble polymer formed inside the digestive vacuole to detoxify heme and its conjugated redox-active iron, which is released during hemoglobin proteolysis (Francis et al., 1997; Sigala and Goldberg, 2014). Hemozoin becomes accessible during erythrocyte rupture and upon phagocytosis of infected erythrocytes, and has been described to induce expression of pro-inflammatory cytokines, such as TNF and IL-1β, as well as chemokines from human monocytes, murine macrophages (Olivier et al., 2014), and human monocyte-derived DCs (Bujila et al., 2016). Hemozoin or hemozoin-bound nucleic acids are recognized by endosomal TLR9 (Coban et al., 2005; Parroche et al., 2007), cytoplasmic inflammasomes, or cytoplasmic sensors (Kalantari et al., 2014). However, it remains unresolved whether hemozoin itself or molecules bound to hemozoin, such as DNA, activate TLR9 (Liehl and Mota, 2012). In fact, the AT-rich Plasmodium genomic DNA, a feature shared by many pathogens including Schistosoma, was described to be immunomodulatory in vitro, but this response is apparently TLR9-independent (McCutchan et al., 1984; Sharma et al., 2011).

In addition to inducing pro-inflammatory responses, recognition of parasites and Plasmodium-infected erythrocytes is crucial for the phagocytic uptake and thus removal of parasites from the circulation by macrophages and DCs (McGilvray et al., 2000; Stevenson and Riley, 2004; **Figure 3**). Interestingly, although macrophages and DCs may both contribute to early pro-inflammatory cytokine responses via activation of PRRmediated signaling, a recent study suggests that macrophage responsiveness is strongly compromised upon phagocytosis of P. falciparum- or P. berghei-infected erythrocytes or free merozoites due to pronounced phagosomal acidification (Wu et al., 2015). Instead, DCs contribute to increased serum levels of pro-inflammatory cytokines early during P. berghei infection, including IL-6, IL-12p40, and TNF (Wu et al., 2015).

Apart from parasite-derived stimuli, host-derived DAMPs such as urate crystals, heme, and microvesicles released from damaged host cells may activate the innate immune system, although so far this has only been demonstrated for microvesicles (Gazzinelli et al., 2014). Plasmodium falciparuminfected erythrocyte-derived microvesicles have been reported to induce TNF and IL-10 from monocyte-derived human macrophages in vitro, potentially through phagocytic uptake of microvesicles (Mantel et al., 2013). In line with this observation, plasma microparticles derived from P. berghei -infected mice stimulated bone marrow-derived macrophages to secrete TNF in vitro, and TNF induction was reported to be TLR4-dependent (Couper et al., 2010). In humans, increased numbers of plasma microparticles have been detected during P. falciparum and P. vivax infection, and microparticle numbers were higher in severe malaria cases, including CM, than in uncomplicated P. falciparum infections (Campos et al., 2010; Nantakomol et al., 2011; Sahu et al., 2013). Furthermore, two studies point toward infection-induced alterations in the microvesicle cargo, suggesting that not only microvesicle frequency but also content are relevant in inducing pro-inflammatory responses (Couper et al., 2010; Tiberti et al., 2016). Together, these studies indicate that microvesicles might contribute to CM pathogenesis and to other manifestations of severe malaria, and the association with pro-inflammatory immune responses warrants further investigations.

Although parasite-derived stimuli have been repeatedly reported to induce pro-inflammatory responses, the real qualitative and quantitative nature of the stimuli remains inadequately understood and may point toward a complex synergistic effect of multiple stimuli. Pro-inflammatory cytokines such as TNF, IL-1α (Kwiatkowski et al., 1990; Tchinda et al., 2007), IFN-γ and IL-12p40 (Hermsen et al., 2003), as well as chemokines, including IL-8/CXCL8 (Hermsen et al., 2003), platelet factor 4 (PF4)/CXCL4, and IP-10/CXCL10 (Wilson et al., 2011), are clearly elevated during P. falciparum infection. Several cytokines and chemokines, including TNF and CXCL10, have been found to be associated with CM severity (Kwiatkowski et al., 1990; Wilson et al., 2011), while a more recent study reported that neither plasma nor cerebrospinal fluid (CSF) TNF concentration were indicative of CM-associated mortality, yet elevated levels of TNF in CSF of pediatric CM cases were associated with long-term neurologic and cognitive deficits (Shabani et al., 2017). Consequently, it remains to be conclusively determined which cytokines and/or chemokines present suitable prognostic signatures of disease progression.

While many early studies focused on the identification of single inflammatory cytokines critically involved in malaria pathology, it is likely that numerous immune players are modulated during the course of a Plasmodium infection sequentially and/or simultaneously. Accordingly, it is conceivable that a complex interplay of immune mediators contributes to the development of severe malaria in general, and to CM pathogenesis in particular. Several studies have addressed this issue by systematically analyzing pro- and anti-inflammatory markers. Since these studies included patients from different study sites of varying age and at various time points of infection, the cytokines identified to be associated with disease severity varied substantially, and the combination of selected analytes was heterogenous between studies (Prakash et al., 2006; Jain et al., 2008; Thuma et al., 2011). An early study reported two clusters of cytokines associated with mild and cerebral malaria, respectively, in P. falciparum-infected adults (Prakash et al., 2006). According to this study, IFN-γ, IL-2, IL-5, IL-6 and IL-12 were increased in mild malaria whereas TGF-β, TNF, IL-10 and IL-1β were particularly elevated in CM. In a study cohort of P. falciparuminfected children and adults, serum TNF levels did not correlate with disease severity, and instead IP-10/CXCL10, sTNF-R2, and sFas were proposed as biomarkers of CM severity and mortality (Jain et al., 2008). Additional cytokines were elevated in malaria cases compared to healthy controls and included IL-1ra, IL-10, IL-8/CXCL8, and macrophage inflammatory protein 1β (MIP-1β)/CCL4 (Jain et al., 2008). In two cohorts of P. falciparuminfected children, TNF concentration was slightly, albeit nonsignificantly, elevated in CM compared to severe anemia cases (Thuma et al., 2011; Mandala et al., 2017). In the study by Thuma et al. (2011) conducted in Zambia, IL-10, IL-1α, IL-6, and IP-10/CXCL10 plasma levels were higher in children suffering from CM than in children with severe anemia (Thuma et al., 2011). In line with these findings, Mandala et al. (2017) found higher IL-10 and IL-6 serum levels in Malawian children suffering from CM compared to those with severe anemia, while IFN-γ and IL-8/CXCL-8 were also elevated in pediatric CM cases.

In summary, cytokine profiling continues to aid in identifying distinct patterns of pro- and anti-inflammatory cytokines and chemokines in CM patients. Although a common cytokine/chemokine signature associated with CM severity has not yet been identified, which is in part due to the fact that the combination of markers investigated varies among studies, collectively, these studies point toward important roles for certain immunoregulatory molecules in modulating CM severity. In order to describe how inflammatory mediators associated with Plasmodium infection may contribute to CM pathogenesis, we will highlight their roles in endothelial activation, blood-brain barrier permeability, and neuroinflammation, by drawing on findings obtained from in vitro studies and the murine ECM model.

### INFLAMMATION AND ENDOTHELIAL ACTIVATION

Functions of healthy endothelium include anti-coagulant properties through inhibiting platelet adhesion and aggregation, regulation of blood flow by releasing nitric oxide, controlling endothelial permeability, preventing extravasation of plasma proteins to tissue, and preventing leukocyte adhesion through suppressing adhesion molecule expression and sequestering chemokines within Weibel-Palade-bodies (Pober and Sessa, 2007). Under inflammatory conditions such as systemic inflammation during Plasmodium infection, endothelial activation may seriously impair endothelial function (**Figure 4**). A hallmark of endothelial activation is the expression of adhesion molecules such as VCAM-1 and ICAM-1 on the endothelial cell surface. Systemic endothelial activation was reported in P. falciparum-infected and in sepsis patients based on plasma levels of soluble adhesion molecules (Turner et al., 1998). Moreover, immunohistochemical analysis of fatal P. falciparum -infected CM cases revealed that expression of ICAM-1 was

FIGURE 4 | Endothelial activation and chemokine secretion. A characteristic feature of *Plasmodium* infection is endothelial activation, which is likely induced by elevated serum tumor necrosis factor (TNF) levels. Binding of TNF to its receptor (TNFR1) induces transcription of adhesion molecules, including ICAM-1 and VCAM-1, as well as chemokines (Pober and Sessa, 2007). Endothelial activation might be directly induced by infected erythrocytes, possibly through activation of pattern recognition receptors (PRR), resulting in elevated expression of ICAM-1 and chemokine secretion (Viebig et al., 2005; Tripathi et al., 2006, 2009; Chakravorty et al., 2007).

most pronounced in the brain microvasculature compared to other organs and biopsies from non-malaria cases (Turner et al., 1994). In good agreement, ICAM-1 staining was described to be more pronounced on brain endothelial cells from P. berghei (strain ANKA)-infected mice during ECM than in those isolated from P. yoelii-infected (non-ECM) mice (Grau et al., 1991). Accordingly, Icam1-deficiency protected mice from P. berghei-induced ECM (Favre et al., 1999). Considering that adhesion molecules are thought to promote binding of infected erythrocytes and leukocytes to endothelial cells, and that human as well as murine CM is associated with intravascular accumulation of leukocytes in the brain (Hunt and Grau, 2003), these findings signify a critical role for endothelial activation in CM pathogenesis.

Among other factors, endothelial activation may be induced by inflammatory cytokines. TNF and lymphotoxin α (LTα) activate human endothelial cells in vitro (Cavender et al., 1989; Pober and Cotran, 1990). Similarly, IFN-γ, IL-1α, and IL-1β function in endothelial activation (Pober and Cotran, 1990; Bauer et al., 2002). These pro-inflammatory cytokines have been found to be elevated in serum or plasma of P. falciparum-infected patients and TNF as well as IL-1α and IL-1β were described to be associated with CM severity (Prakash et al., 2006; Thuma et al., 2011). Consequently, endothelial activation observed in P. falciparum infection might in part be mediated by these cytokines. In accordance with a proposed role for endothelial activation in CM, the pathology observed in murine ECM is associated with a T helper 1 (Th 1) immune response, and cytokines, such as IFN-γ, TNF, and LTα, and immune cells, e.g., CD4<sup>+</sup> and CD8<sup>+</sup> T cells together with NK cells are involved in ECM (Yanez et al., 1996; Lucas et al., 1997; Engwerda et al., 2002; Hunt and Grau, 2003; Schofield and Grau, 2005; Langhorne et al., 2008). Of note, despite an association of TNF with disease severity in CM, Tnf-deficient mice were not protected from ECM (Engwerda et al., 2002), and blocking of TNF by anti-TNF antibodies or pentoxifylline did not improve survival in human CM (Di Perri et al., 1995; van Hensbroek et al., 1996). Consequently, blocking TNF was demonstrated to be insufficient to prevent fatal CM, and, therefore, additional mechanisms are likely to critically contribute to CM pathogenesis. Moreover, these early therapeutic interventions highlight the complexity of this malaria syndrome, indicating that single serum cytokines associated with CM severity do not necessarily translate into therapeutic approaches. Notably, elevated serum TNF in particular may represent a secondary immune response, which might further amplify severity while not being critical in initiating CM pathogenesis.

Apart from the induction of endothelial activation by proinflammatory cytokines, endothelial cells are an important part of the innate immune response since they recognize PAMPs through expression of PRR such as TLR and NLR. Endothelial cells were found to secrete pro-inflammatory cytokines, including IL-1α, IL-1β, or IL-6, as well as immunomodulatory cytokines, namely IL-10 and TGF-β, and chemokines, e.g., monocyte chemoattractant protein 1 (MCP-1)/CCL2, RANTES/CCL5, and IL-8/CXCL8, upon stimulation by proinflammatory cytokines or lipopolysaccharide (LPS) in vitro (Mai et al., 2013). Notably, microvascular endothelial cells derived from subcutaneous adipose tissue of patients with uncomplicated malaria and fatal CM were demonstrated to differ in their endothelial inflammatory response to TNF stimulation ex vivo in that MCP-1/CCL2 and IL-6 were induced to a larger extent in endothelial cells derived from CM patients. Consequently, it was proposed that inter-individual differences in the endothelial response to inflammation might account for CM severity (Wassmer et al., 2011). These results support the notion that the pro-inflammatory microvascular environment during P. falciparum infection might be enhanced by endothelial cells. Interestingly, human brain microvascular endothelial cells have been described to phagocytose P. berghei merozoites in vitro (Howland et al., 2015b). Furthermore, upon co-culture with P. falciparum-infected erythrocytes, human endothelial cell lines were reported to upregulate ICAM-1 expression (Viebig et al., 2005; Tripathi et al., 2006), to increase transcription of CCL20, CXCL1, CXCL2, CXCL8, and IL6 (Tripathi et al., 2009), and to secrete MCP-1/CCL2, MIP-3α/CCL20, and IL-8/CXCL8 (Viebig et al., 2005; Chakravorty et al., 2007) (**Figure 4**). Together, these in vitro observations suggest that endothelial cells are potentially directly involved in the immune response to Plasmodium infection. Since leukocyte sequestration in the microvasculature of the brain was described in human CM and murine ECM (Hunt and Grau, 2003), local chemokine gradients originating from brain endothelial cells might orchestrate leukocyte migration and thus promote local inflammation, thereby contributing to the development of CM.

Although the extent of the contribution of endothelial cellderived chemokines in the process of leukocyte accumulation in brains of patients suffering from CM and of ECM mice remains to be established, a number of clinical studies reported that several chemokines are elevated in serum and CSF of CM patients, thereby providing a potential link between serum or CSF chemokine levels and progression from mild malaria to CM. Chemokines that were elevated in human CM cases included MCP-1/CCL2, MIP-1β/CCL4, PF4/CXCL4, IL-8/CXCL8, and IP-10/CXCL10 (Jain et al., 2008; Wilson et al., 2011). Particularly, MCP-1/CCL2, Eotaxin/CCL11, and IP-10/CXLC10 were reported to be indicative of disease severity (Armah et al., 2007; Jain et al., 2008; Thuma et al., 2011; Wilson et al., 2011). Moreover, post mortem MIP-1β/CCL4, IL-8/CXCL8, and IP-10/CXCL10 levels were significantly elevated in cerebrospinal fluid (CSF) of fatal P. falciparum-induced CM cases when compared to fatal severe malarial anemia cases and non-malaria deaths (Armah et al., 2007). In another study, IL-8/CXCL8 was elevated in CSF of non-fatal CM cases of P. falciparum-infected children (John et al., 2008b), while MCP-1/CCL2, MIP-1α/CCL3, MIP-1β/CCL4, and RANTES/CCL5 levels were comparable to malaria-free controls. The relevance of certain chemokines in cerebral malaria pathogenesis was demonstrated in the murine ECM model, in which Cxcl4-, Cxcl9-, or Cxcl10-deficiency resulted in reduced ECM-associated mortality (Campanella et al., 2008; Srivastava et al., 2008; Nie et al., 2009). Notably, PF4/CXCL4 was elevated in plasma of P. berghei ANKA-infected mice and demonstrated to induce TNF secretion from peritoneal macrophages and T cells in vitro, and Cxcl4-deficiency resulted in reduced serum TNF and IFNγ levels during P. berghei ANKA-infection (Srivastava et al., 2008), indicating that PF4/CXCL4 contributes to establishing a pro-inflammatory environment, which might amplify further immune responses and could thereby promote CM pathogenesis. Furthermore, protection from ECM in Cxcl10-deficient mice was associated with a decrease in leukocyte sequestration in brains of P. berghei ANKA-infected mice, while parasite-specific CXCR3<sup>+</sup> T cells were increased in spleens of Cxcl10-deficient compared to WT mice (Nie et al., 2009), suggesting that IP-10/CXCL10 mediated recruitment of CXCR3<sup>+</sup> T cells to the brain might contribute to the development of ECM. A similar phenomenon might account for a decrease in ECM-associated mortality described for Cxcl9-deficient mice (Campanella et al., 2008), since recruitment of CXCR3-expressing cells can also be initiated by monokine induced by IFN-γ (MIG)/CXCL9. However, the precise mechanism by which MIG/CXCL9 contributes to ECM pathogenesis is yet to be determined. Apart from studies using specific gene deletions, chemokine transcripts were found to be induced to a higher extent in brains of ECM- compared to non-ECM mice. These transcripts included Ccl2, Ccl3, Ccl4, Ccl5, Cxcl1, Cxcl9, and Cxcl10 (Miu et al., 2008; Van den Steen et al., 2008), supporting the finding that apart from NK and T cells, monocytes and neutrophils also sequester in the microvasculature (Renia et al., 2012), e.g., through recruitment by MCP-1/CCL2 and keratinocyte chemoattractant (KC)/Cxcl1, respectively. Nevertheless, studies to identify the cell types producing these chemokines in the brain are very limited. For instance, CXCL9 was demonstrated to be expressed by endothelial cells, while the source(s) of CXCL10 in the brain during P. berghei infection remain to be conclusively determined, and could include neurons, astrocytes, or endothelial cells (Campanella et al., 2008; Miu et al., 2008) as well as recruited monocytes (Ioannidis et al., 2016).

Together, activated endothelial cells likely contribute to local inflammation by secreting cytokines and chemokines, thereby recruiting leukocytes, including monocytes, macrophages, neutrophils and T cells, which accumulate in brains of mice and humans during ECM and CM, respectively (Renia et al., 2012; Storm and Craig, 2014). Since these cell types might secrete cytokines and chemokines themselves, local inflammation and endothelial activation could be further exacerbated (**Figure 5**). For instance, potential endothelial cell-induced recruitment of neutrophils and monocytes expressing IP-10/CXCL10 may promote further recruitment of CXCR3<sup>+</sup> cells such as NK and T cells to the brain (Ioannidis et al., 2016).

### ENDOTHELIAL ACTIVATION AND BLOOD-BRAIN BARRIER INTEGRITY

The blood-brain barrier is comprised of endothelial cells forming a continuous barrier through tight junctions, a basement membrane and astrocytes, which are in direct contact with neurons and microglia. This composition is critical to minimize local inflammation and neuronal damage (Obermeier et al., 2013). In the course of Plasmodium infection, endothelial

which exacerbates local inflammation in the brain.

activation may progress to vascular permeability and loss of blood-brain barrier integrity, as indicated by hemorrhages in brains of CM patients and extravasation of dyes or antibodies into the brain parenchyma in ECM (Renia et al., 2012). Although the extent of pathological events related to blood-brain barrier function in human CM is variable, dysfunction of the bloodbrain barrier appears to be associated with progression of cerebral disease (Medana and Turner, 2006). Histology of brain sections from fatal human CM cases revealed a redistribution of the tight junction proteins occludin, vinculin, and zonula occludens 1 (ZO-1), which are central to blood-brain barrier integrity (Brown et al., 1999). Moreover, immunohistochemistry of brain sections derived from pediatric fatal CM cases indicated blood-brain barrier impairment in areas containing sequestered P. falciparum-infected erythrocytes, where they were associated with focal loss of endothelial intercellular junctions (Brown et al., 2001). Additionally, in vitro studies have demonstrated a decrease in endothelial resistance upon addition of P. falciparuminfected erythrocytes to endothelial cells (Tripathi et al., 2007; Jambou et al., 2010). This model, however, only partially reflects the response at the blood-brain barrier due to lack of barrier components such as astrocytes and pericytes (Medana and Turner, 2007).

Interestingly, MCP-1/CCL2 induces redistribution of tight junction proteins and increases endothelial permeability in vitro (Stamatovic et al., 2003; Song and Pachter, 2004; Yao and Tsirka, 2011). Thus, chemokines may contribute to organ-specific inflammation by inducing signals that promote endothelial permeability (**Figure 6**). The roles of the respective chemokine receptors are less clear in this context. This is exemplified in the

FIGURE 6 | Endothelial permeability and neuroinflammation. Through continued inflammatory insults toward endothelial cells by, for instance, circulating tumor necrosis factor (TNF) and interferon γ (IFN-γ), miR-155 might be upregulated in endothelial cells. Along with uptake of miR-451a from *P. falciparum*-infected erythrocyte-derived extracellular vesicles, reorganization of tight junction proteins such as zonula occludens 1 (ZO-1) is induced and could contribute to endothelial permeability during cerebral malaria (Lopez-Ramirez et al., 2014; Mantel et al., 2016). Additionally, chemokine receptors might induce redistribution of tight junction proteins in a G protein-dependent manner (Stamatovic et al., 2003; Song and Pachter, 2004; Yao and Tsirka, 2011), while the contribution of chemokine-induced opening of tight junctions is less clear in the context of cerebral malaria. CD8<sup>+</sup> <sup>T</sup> cell-mediated cytotoxity toward endothelial cells through recognition of parasite antigen presented on MHC class I molecules on endothelial cells likely contributes substantially to blood-brain barrier permeability during cerebral malaria (Howland et al., 2015a,b). Consequently, pro-inflammatory cytokines enter the brain parenchyma and could thereby activate astrocytes and microglia, which in turn could secrete chemokines (Capuccini et al., 2016) and thus promote leukocyte recruitment and local inflammation.

MCP-1/CCL2 receptor CCR2. Ccr2-deficiency abrogated CCL2 induced endothelial permeability in vitro (Stamatovic et al., 2003), but did not protect against ECM (Belnoue et al., 2003), indicating that other chemokines and/or additional mechanisms induce blood-brain barrier permeability.

Along with the observed redistribution of endothelial tight junction proteins and loss of intercellular junctions, growing evidence from the murine ECM model suggests that CD8<sup>+</sup> T cells are primary mediators in CM disease pathogenesis by contributing considerably to the loss of blood-brain barrier integrity (Howland et al., 2015a). Indeed, the accumulation of CD8<sup>+</sup> T cells in brains of P. berghei ANKA-infected mice appears to be critical for the development of ECM (Villegas-Mendez et al., 2012), and blood-brain barrier disruption was proposed to be a consequence of CD8<sup>+</sup> T cell-mediated cytotoxicity toward endothelial cells cross-presenting parasite antigen (Howland et al., 2015b) (**Figure 6**). Recruitment of CD8<sup>+</sup> T cells to the brain of infected mice has been described to be partly mediated by expression of the chemokine receptor CXCR3 and its IFN-γinducible ligands MIG/CXCL9 as well as IP-10/CXCL10 (Hansen et al., 2007; Villegas-Mendez et al., 2012). Indeed, NK cell-derived IFN-γ has been demonstrated to be crucial for the induction of CXCR3 expression on T cells and subsequent T-cell migration to the brain of P. berghei ANKA-infected mice (Hansen et al., 2007). In line with these findings, brain transcripts of Cxcr3 were reduced in mice deficient for IFN-γR1 in comparison to wild type mice infected with P. berghei ANKA. Additionally, while expression of MIG/CXCL9 and IP-10/CXCL10 is induced in brains of P. berghei ANKA-infected wild type mice, expression levels of these chemokines in Ifngr1-deficient mice were similar to those of uninfected mice (Palomo et al., 2013). Collectively, these findings suggest that IFN-γ-induced signaling is crucial for chemokine-mediated leukocyte recruitment to the brain during ECM, while it remains to be established to which extent activated endothelium might contribute to this chemokine response.

Importantly, Cxcr3-deficient mice were less likely to succumb to ECM (Campanella et al., 2008; Miu et al., 2008), and numbers of CD4<sup>+</sup> and CD8<sup>+</sup> T cells in brains of Cxcr3-deficient mice were reported to be reduced compared to wild type mice during P. berghei ANKA infection. However, the extent of CD8<sup>+</sup> T cell recruitment was similar in Cxcr3-deficient mice which were either resistant or susceptible to ECM (Miu et al., 2008). These results indicate that the quantity of CD8<sup>+</sup> T cells in brains of P. berghei ANKA-infected mice is not critical for the development of ECM. Rather, additional aspects such as CD8<sup>+</sup> T-cell specificity toward parasite antigen presented on endothelium, expression levels of key cytotoxic effector molecules perforin and granzyme B by CD8<sup>+</sup> T cells, and localization of these CD8<sup>+</sup> T cells within the brain might play a substantial role in the development of ECM (Miu et al., 2008; Haque et al., 2011; Howland et al., 2015b; Huggins et al., 2017). Nevertheless, it remains to be determined why susceptibility to ECM among Cxcr3-deficient mice is variable. Moreover, these findings indicate that additional mechanisms other than CD8<sup>+</sup> T cell-mediated cytotoxicity might be involved in the loss of blood-brain barrier integrity.

In fact, in addition to CD8<sup>+</sup> T cells and chemokines, extracellular vesicles derived from P. falciparum-infected erythrocytes were recently implicated in blood-brain barrier permeability in vitro (Mantel et al., 2016; **Figure 6**). These extracellular vesicles contained miRNA miR-451a, which, upon endocytosis by endothelial cells, correlated with reduced endothelial caveolin-1 expression. Interestingly, MCP-1/CCL2-induced redistribution of tight junction proteins and concomitant endothelial permeability (Stamatovic et al., 2003) was also accompanied by a decrease in caveolin-1 protein (Song and Pachter, 2004), thus pointing toward a common underlying mechanism in the induction of endothelial permeability by CCL2 and extracellular vesicles derived from P. falciparum-infected erythrocytes. Since sequestration of infected erythrocytes in the microvasculature is likely promoted by expression of adhesion molecules upon endothelial activation, locally increased release of P. falciparum-infected erythrocyte-derived extracellular vesicles carrying miRNA miR-451a might be an additional mechanism involved in blood-brain barrier permeability in CM.

In addition to miR-451a, another miRNA, miR-155, was implicated in inflammation-associated blood-brain barrier permeability. Upon stimulation of human brain endothelial cells with pro-inflammatory cytokines TNF and IFN-γ in vitro, miR-155 was found to be upregulated. The tight junction protein claudin-1 was reported to be among the candidate targets of miR-155, and cytokine stimulation or miR-155 overexpression resulted in reorganization of the tight junction protein ZO-1 along with increased endothelial permeability (Lopez-Ramirez et al., 2014), indicating that cytokines such as TNF and IFN-γ might directly contribute to blood-brain barrier permeability through induction of regulatory miRNAs. Interestingly, levels of miR-155 were recently reported to be elevated in extracellular vesicles in the circulation of P. berghei ANKA-infected mice, and miR-155-deficiency resulted in preservation of blood-brain barrier integrity and reduced ECM-associated mortality (Barker et al., 2017). Notably, plasma concentrations of IL-6, IFN-γ, and MCP-1/CCL2 were significantly elevated in P. berghei ANKA-infected miR-155-deficient compared to wild type mice in this study, suggesting that miR-155 might have additional targets other than tight junction proteins. Importantly, whether miR-155-carrying extracellular vesicles are derived from activated endothelium, and their impact on tight junction reorganization in the context of ECM, remain to be determined. Nevertheless, inhibition of the function of miR-155 might present a useful target for therapeutic approaches (Barker et al., 2017). These processes of chemokine induction, microvascular sequestration of infected erythrocytes and leukocytes, and release of extracellular vesicles carrying regulatory miRNAs are most likely not exclusive to the brain. However, endothelial permeability is likely more detrimental in the brain than in other organs and, hence, cerebral malaria may be the most severe manifestation of these processes. Additionally, brain endothelial cells have been described to express comparably low levels of thrombin-binding thrombomodulin, and excess of unbound thrombin may further promote local endothelial activation by inducing further expression of adhesion molecules (Clark et al., 2006).

Together, cytokine- or Plasmodium-induced endothelial activation may lead to chemokine induction and leukocyte recruitment as well as sequestration of infected erythrocytes, which may act synergistically in promoting endothelial permeability. Nevertheless, the precise molecular mechanisms that trigger loss of blood-brain barrier integrity in CM are incompletely understood and need to be further investigated.

### NEUROINFLAMMATION

Upon disruption of the blood-brain barrier, cytokines, chemokines, and soluble parasite products might enter the brain parenchyma and, thereby, activate astrocytes and microglia, and result in symptoms of neuroinflammation in the absence of extravasation of infected erythrocytes or leukocytes into the brain parenchyma (Combes et al., 2010). Indeed, activation of microglia and astrocytes has been observed in murine and human CM (Hunt et al., 2006; Combes et al., 2010). For instance, transcriptome analysis of microglia isolated from P. berghei-infected mice revealed that several chemokines as well as transcripts related to type I IFN signaling were differentially upregulated (Capuccini et al., 2016). This finding was confirmed in vitro by stimulation of a murine microglia cell line with IFN-β, which resulted in secretion of MCP-1/CCL2, RANTES/CCL5, MIG/CXCL9, and IP-10/CXCL10. Moreover, stimulation of human primary astrocytes with a combination of IFN-γ and LTα synergistically induced IP-10/CXCL10 secretion in vitro (Bakmiwewa et al., 2016). Additionally, co-culture of P. berghei-infected erythrocytes with a mixed astrocyte-microglia culture resulted in phagocytic uptake of infected erythrocytes and of parasitized erythrocyte-derived microvesicles by microglia and astrocytes, respectively, which in turn was associated with an induction in IP-10/CXCL10 secretion (Shrivastava et al., 2017). Furthermore, astrocytes and microglia may secrete various cytokines and chemokines upon activation (Dong and Benveniste, 2001; Medana et al., 2001). Consequently, loss of blood-brain barrier integrity and subsequent activation of microglia and astrocytes might result in further chemokine-mediated recruitment of leukocytes to the brain and subsequent amplification of inflammation (**Figure 6**). Additionally, activation of microglia might induce expression of FasL, which, upon binding to Fas expressed on astrocytes, could induce astrocyte damage (Hunt et al., 2006). However, to our knowledge, this has so far not been demonstrated in the context of CM. Since astrocytes are critically involved in maintaining blood-brain barrier properties and survival of neurons (Combes et al., 2010), their functional impairment might disrupt neuronal activity and could thereby account for the neurological impairment observed in some CM cases (Hunt et al., 2006).

### SUMMARY AND PERSPECTIVES

The human and murine immune system are in part strikingly different (Stevenson and Riley, 2004). For instance, IL-8/CXCL8 was reported to be associated with cerebral malaria (Armah et al., 2007; John et al., 2008b), while this chemokine is not expressed in mice (Viola and Luster, 2008). However, KC/Cxcl1 is considered a functional homolog of IL-8/CXCL8 in mice (Hol et al., 2010), which mediates neutrophil recruitment, and Cxcl1 transcripts were reported to be elevated in brains of ECM compared to non-ECM mice (Miu et al., 2008), suggesting that KC/Cxcl1 might be similarly involved in ECM pathogenesis. Yet, the precise contribution of IL-8/CXCL8 and KC/Cxcl1 to human CM and murine ECM, respectively, needs to be further investigated. Moreover, the murine ECM model shares several features with human CM, including aspects of histopathology and inflammatory responses. Importantly, mechanistic insights can only be gained from the murine ECM model, and many observations are in remarkably good agreement with clinical data obtained from P. falciparuminduced CM. Although a comprehensive representation of the events leading to CM pathogenesis remains elusive, a working model of an inflammatory cascade leading to CM is conceivable (**Figure 1**).

Upon establishing liver stage infection, a first type I IFN response is mounted by hepatocytes, leading to a primary activation of IFN-γ-producing NK cells (Liehl et al., 2014; Miller et al., 2014; **Figure 2**) and, consequently, induction of MIG/CXCL9 and IP-10/CXCL10 secretion, potentially from endothelial cells (Campanella et al., 2008; Miu et al., 2008). Upon progression of the Plasmodium infection to the blood stages, infected erythrocytes are recognized by DCs and induce the secretion of IL-12 and TNF (Wu et al., 2015; **Figure 3**). IL-12 contributes to further activation of NK and differentiation of Th1 cells (Stevenson and Riley, 2004), while TNF and IFN-γ activate chemokine transcription and adhesion molecule expression on endothelial cells (Pober and Sessa, 2007; Miu et al., 2008; Griffith et al., 2014). As the infection progresses further, Plasmodiuminfected erythrocytes are recognized by endothelial cells and induce expression of chemokines, such as MCP-1/CCL2 and IL-8/CXCL8 (Viebig et al., 2005; Chakravorty et al., 2007; Tripathi et al., 2009; **Figure 4**). As a result, leukocytes are recruited, including monocytes, macrophages, neutrophils, as well as T cells, and initiate a local inflammatory response (Renia et al., 2012; Storm and Craig, 2014; **Figure 5**). These cell types secrete chemokines, thereby amplifying the response leading to further leukocyte recruitment and intensifying local inflammation. Additionally, endothelial cells phagocytose merozoites and parasite material released during schizont rupture and present parasite antigens to CD8<sup>+</sup> T cells (Howland et al., 2015b; **Figure 6**), which may result in targeted elimination of antigenpresenting endothelial cells and, thus, cause damages to the endothelial lining of the blood-brain barrier. This process can be further exacerbated by openings of tight junctions mediated by chemokines and extracellular vesicle-derived miRNAs (Song and Pachter, 2004; Mantel et al., 2016). As a result, small molecules can enter the brain parenchyma and potentially activate brain-resident microglia and astrocytes, further amplifying local inflammation through cytokine secretion and leukocyte recruitment and impairing neuronal functionality (Hunt et al., 2006; Combes et al., 2010).

Even though the febrile response elicited during Plasmodium blood stage infection together with the concomitant inflammatory cytokine responses limit parasite growth and mediate the resolution of infection, imbalances in pro-inflammatory and anti-inflammatory cytokines cause progression of malaria disease to manifestations of severe malaria, such as CM, and death. Adjunctive therapies that prevent adverse effects of the immune response to Plasmodium infection are therefore urgently needed. Although immunomodulation is a promising approach to alleviate immune-mediated pathology, such therapies need to be designed carefully in order to maintain efficient control of parasite growth. Notably, adjunct therapies modulating chemokine responses may have fewer side-effects compared to therapies based on neutralizing cytokines (Ioannidis et al., 2014). Indeed, antibodymediated targeting of IP-10/CXCL10 was demonstrated to result in reduced ECM-induced mortality and parasite burden in mice, which was likely mediated by retention and expansion of parasite-specific T cells in the spleen (Nie et al., 2009). Such treatments may be relevant in other contexts as well: murine Toxoplasma encephalitis has been described to be associated with constant expression of Ccl2, Ccl3, Ccl4, Ccl5, and Cxcl10 in brains of infected mice concomitant with continuous recruitment of CD4<sup>+</sup> and CD8<sup>+</sup> T cells, which was not the case for mice in which the infection developed into chronic latency (Strack et al., 2002). Furthermore, targeted neutralization of single chemokines, including MCP-1/CCL2, MIP-1α/CCL3, RANTES/CCL5, or IP-10/CXCL10, resulted in protection of mice from experimental autoimmune encephalomyelitis, a murine model for multiple sclerosis (Karin and Wildbaum, 2015). These findings from other neuroinflammatory diseases highlight that chemokines might present a valuable target for intervention strategies in several diseases. However, efforts to design chemokine-based therapies are challenged by the complexity of the chemokine system as well as properties such as redundancy, pleiotropy, and speciation (Viola and Luster, 2008). In fact, most cell populations express several different chemokine receptors and thus single chemokine or chemokine receptor blockade may not affect disease outcome in certain pathological conditions.

Together, reliable biomarkers which predict disease outcome and allow for potential prophylactic measures are yet to be identified. Several clinical studies have reported potential diagnostic and prognostic biomarkers for CM, which apart from chemokines such as IP-10/CXCL10 and PF4/CXCL4 also include P. falciparum histidine-rich protein 2 (Pf HRP2), a protein which correlates with parasite biomass, as well as regulators of endothelial activation angiopoietin-1 and -2 (summarized in Sahu et al., 2015). Still, comprehensive studies with defined clinical parameters and systematic assessment of plasma levels of multiple inflammatory mediators need to be performed to determine whether distinct clusters of markers can be associated with disease severity in order to identify patients at risk of developing CM early during infection. Such studies will inform future investigations into mechanisms underlying disease pathogenesis in order to develop novel evidence-based malaria intervention strategies.

### AUTHOR CONTRIBUTIONS

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

## ACKNOWLEDGMENTS

Work on chemokines and cytokines in murine malaria models is funded by the Deutsche Forschungsgemeinschaft (KA 3347/4-1) and partly by the Max Planck Society.

## REFERENCES


Dong, Y., and Benveniste, E. N. (2001). Immune function of astrocytes. Glia 36, 180–190. doi: 10.1002/glia.1107


fatal brain edema during experimental cerebral malaria. Infect. Immun. 85, e00985–e00916. doi: 10.1128/IAI.00985-16


malaria and protection due to CXCR3 deficiency. J. Immunol. 180, 1217–1230. doi: 10.4049/jimmunol.180.2.1217


inflammatory pathways in human cerebral endothelium. Blood 114, 4243–4252. doi: 10.1182/blood-2009-06-226415


WHO (2016). World Malaria Report 2016. Geneva: World Health Organisation.


**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 Dunst, Kamena and Matuschewski. 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.

# Increased *Plasmodium falciparum* Parasitemia in Non-splenectomized *Saimiri sciureus* Monkeys Treated with Clodronate Liposomes

Janaiara A. Cunha1†, Leonardo J. M. Carvalho1†, Cesare Bianco-Junior <sup>1</sup> , Márcia C. R. Andrade<sup>2</sup> , Lilian R. Pratt-Riccio<sup>1</sup> , Evelyn K. P. Riccio<sup>1</sup> , Marcelo Pelajo-Machado<sup>3</sup> , Igor J. da Silva<sup>3</sup> , Pierre Druilhe<sup>4</sup> and Cláudio Tadeu Daniel-Ribeiro<sup>1</sup> \*

<sup>1</sup> Laboratório de Pesquisa em Malária, Instituto Oswaldo Cruz, Fundação Oswaldo Cruz (Fiocruz), Rio de Janeiro, Brazil, 2 Instituto de Ciência e Tecnologia em Biomodelos, Fiocruz, Rio de Janeiro, Brazil, <sup>3</sup> Laboratório de Patologia, Instituto Oswaldo Cruz, Fiocruz, Rio de Janeiro, Brazil, <sup>4</sup> Vac4All Initiative, Pepinière Paris Biotech Santé, Paris, France

#### *Edited by:*

Kai Matuschewski, Humboldt-Universität zu Berlin, Germany

#### *Reviewed by:*

Ashley Vaughan, Center for Infectious Disease Research, United States Jodi L. McGill, Kansas State University, United States

#### *\*Correspondence:*

Cláudio Tadeu Daniel-Ribeiro malaria@fiocruz.br

† These authors have contributed equally to this work.

*Received:* 09 February 2017 *Accepted:* 04 September 2017 *Published:* 21 September 2017

#### *Citation:*

Cunha JA, Carvalho LJM, Bianco-Junior C, Andrade MCR, Pratt-Riccio LR, Riccio EKP, Pelajo-Machado M, da Silva IJ, Druilhe P and Daniel-Ribeiro CT (2017) Increased Plasmodium falciparum Parasitemia in Non-splenectomized Saimiri sciureus Monkeys Treated with Clodronate Liposomes. Front. Cell. Infect. Microbiol. 7:408. doi: 10.3389/fcimb.2017.00408 A major constraint in the study of Plasmodium falciparum malaria, including vaccine development, lies on the parasite's strict human host specificity and therefore the shortage of animal experimental models able to harbor human plasmodia. The best experimental models are neo-tropical primates of the genus Saimiri and Aotus, but they require splenectomy to reduce innate defenses for achieving high and consistent parasitemias, an important limitation. Clodronate-liposomes (CL) have been successfully used to deplete monocytes/macrophages in several experimental models. We investigated whether a reduction in the numbers of phagocytic cells by CL would improve the development of P. falciparum parasitemia in non-splenectomized Saimiri sciureus monkeys. Depletion of S. sciureus splenocytes after in vitro incubation with CL was quantified using anti-CD14 antibodies and flow cytometry. Non-infected and P. falciparum-infected S. sciureus were injected intravenously twice a week with either CL at either 0.5 or 1 mL (5 mg/mL) or phosphate buffered saline (PBS). Animals were monitored during infection and treated with mefloquine. After treatment and euthanasia, spleen and liver were collected for histological analysis. In vitro CL depleted S. sciureus splenic monocyte/macrophage population in a dose- and time-dependent manner. In vivo, half of P. falciparum-infected S. sciureus treated with CL 0.5 mL, and two-thirds of those treated with CL 1 mL developed high parasitemias requiring mefloquine treatment, whereas all control animals were able to self-control parasitemia without the need for antimalarial treatment. CL-treated infected S. sciureus showed a marked decrease in the degree of splenomegaly despite higher parasitemias, compared to PBS-treated animals. Histological evidence of partial monocyte/macrophage depletion, decreased hemozoin phagocytosis and decreased iron recycling was observed in both the spleen and liver of CL-treated infected S. sciureus. CL is capable of promoting higher parasitemia in P. falciparum-infected S. sciureus, associated with evidence of partial macrophage depletion in the spleen and liver. Macrophage depletion by CL is therefore a practical and viable alternative to surgical splenectomy in this experimental model.

Keywords: malaria, *Plasmodium falciparum*, *Saimiri sciureus*, clodronate liposomes, macrophages, spleen, liver

### INTRODUCTION

According to the World Health Organization (WHO) estimates, 212 million cases of malaria occurred in 2015, with 429,000 deaths (WHO, 2016). Among the five Plasmodium species capable of causing human malaria, Plasmodium falciparum is associated with more severe and lethal forms of the disease (Rowe et al., 2009; Quintero et al., 2011). Due to increasing parasite resistance to currently available antimalarial drugs (Chrubasik and Jacobson, 2010; Dondorp et al., 2010), development of an effective vaccine against malaria is urgently needed. Currently, one of the promising vaccine candidates (RTS,S/AS01) is being tested in a Phase III clinical trial in Africa in order to inform a decision regarding its deployment (Clemens and Moorthy, 2016; Olotu et al., 2016; Otieno et al., 2016). Another candidate, the MSP3 experimental vaccine, had a promising performance in African children in a double-blind follow-up of a preliminary Phase I study (Sirima et al., 2011). However, there is no assurance that these candidates will become indeed effective vaccines in the near future. Continued evaluation of potential vaccines is therefore expected and necessary.

Animal models for pre-clinical studies are an important component in vaccine development, but the limited availability of experimental models that can harbor human malaria is a critical constraint. The neo-tropical primates of the genus Saimiri and Aotus are experimental models recommended by WHO for pre-clinical testing of malaria vaccine candidates (WHO, 2004). Studies using the Saimiri sciureus model showed that immunization against P. falciparum blood stage vaccine candidate antigens such as the glutamate-rich protein (GLURP), the merozoite surface protein-3 (MSP3), or the SE36 antigen, using different adjuvants, may induce potent antibody responses and elicit partially protective immunity upon challenge (Carvalho et al., 2004, 2005; Tougan et al., 2013). Besides being susceptible to P. vivax and to P. falciparum infection (Gysin and Fandeur, 1983; Carvalho et al., 2000, 2002, 2004, 2005; Contamin et al., 2000; Herrera et al., 2002; Collins et al., 2005), these primates are abundant in nature and can be easily handled in captivity due to their relatively small size. Furthermore, they are able to reproduce clinical and pathological manifestations of human malaria such as thrombocytopenia, changes in leukocyte counts and anemia, offering an appropriate model to study the immunopathogenesis of malaria (Contamin et al., 2000; Carvalho et al., 2003). The limited availability of immunological tools to study the immune response in these animals imposes some constraints, but human reagents can be used to some extent and specific Saimiri reagents have been generated (Garraud et al., 1998; Contamin et al., 2005; Alves et al., 2010; Riccio et al., 2015). However, these models also have disadvantages. One of the major constraints is the need for splenectomy to achieve high and consistent parasitemias (Collins, 1992). The immune response against the erythrocytic forms of the parasite is largely mediated by resident cells in the spleen (Criswell et al., 1971; Achtman et al., 2003; Leisewitz et al., 2004) and this organ has critical roles in the immune response during malarial infections. Therefore, radical surgical splenectomy poses a strong limitation for testing malaria vaccines, as the very immune responses the vaccines are intended to elicit may be affected by the intervention. Therefore, alternative approaches to allow increased P. falciparum parasitemias in S. sciureus without the need for surgical splenectomy are necessary for proper evaluation of potential vaccines.

Liposomes containing a toxic chemical induce macrophage "suicide," depleteing phagocytes in specific tissues. Clodronate, a bisphosphonate drug that activates apoptosis, has been the most widely employed compound in this respect. It has been demonstrated in several animal models, including mice, dogs, and pigs, that macrophages ingest the liposome particles by phagocytosis and are destroyed or become functionally inactivated (van Rooijen and van Nieuwmegen, 1984; Mathes et al., 2006; Kim et al., 2008). The spleen is a site of important macrophage activity and destruction of blood-borne microorganisms including Plasmodium, senescent red blood cells, and clodronate-encapsulated liposomes (van Rooijen and van Nieuwmegen, 1984; van Rooijen and Kors, 1989; van Rooijen et al., 1990; van Rooijen and van Kesteren-Hendrikx, 2002; Abbas et al., 2008). Chemical "splenectomy" by clodronate liposomes is expected to induce higher P. falciparum parasitemias as it depletes monocyte/macrophages in large numbers while preserving spleen structure and functions. Indeed, in immunocompromised mice lacking B, T, and NK cells, P. falciparum growth was fully controlled by macrophages. Conversely, P. falciparum could successfully replicate at very high parasite densities for several weeks when macrophage populations were controlled by repeated administration of clodronate-encapsulated liposomes (Badell et al., 2000; Arnold et al., 2010, 2011). If the same system can be used in P. falciparum infections in S. sciureus monkeys, it can provide a much improved model for testing malaria vaccines. In such a system, the phagocytic activity of monocytes/macrophages would be decreased allowing increased parasitemia in naive animals, but it is expected that effective vaccines inducing protective adaptative immune responses, in a scenario of preserved spleen structure and function, will be able to contain parasite growth.

In this study, we investigated whether a decrease in the number of monocytes/macrophages by using clodronateencapsulated liposomes would favor the development of consistent P. falciparum parasitemias in S. sciureus monkeys without the need for surgical splenectomy, through a "chemical splenectomy." This procedure would also be more respectful of ethical considerations than surgery, because of its transient nature resulting in fast restitutio ad integrum.

### MATERIALS AND METHODS

### Animals

The animals used in this study were male and female S. sciureus, karyotype 14-7, ranging in age from 3 to 19 years and in weight from 0.548 to 0.926 kg. The monkeys, non-splenectomized and malaria-naïve, were obtained from the breeding colony of the Primatology Service (CECAL/Fiocruz), Rio de Janeiro, Brazil. In total, 34 animals were used in three independent experiments (**Table 1**). This study was carried out in accordance


Exp, experiment; NInf, non-infected animals; Inf, P. falciparum-infected animals; CL, Clodronate-encapsulated liposomes.

with the recommendations and approved by the Fiocruz Ethics Committee on Animal Use (CEUA Licences L-0062/08 and LW-9/14).

### Clodronate-Encapsulated Liposomes

Clodronate-encapsulated liposomes (CL, 5 mg/mL clodronate) were obtained from ClodronateLiposome.org (Department of Molecular Cell Biology and Immunology of the Vrije Universiteit Medisch Centrum - University Amsterdam, VUmc - Netherlands).

### Evaluation of Clodronate-Liposome on Macrophages/Monocytes *in Vitro*

Cryopreserved Saimiri splenocytes obtained during splenectomy of three naïve animals were thawed, washed twice with pure RPMI medium (Gibco, Grand Island, New York, EUA), resuspended in RPMI supplemented with 10% fetal bovine serum (FBS, Invitrogen, Grand Island, New York, EUA) at a concentration of 5 × 10<sup>5</sup> cells/ 500 µL and incubated in 5 mL culture tubes (Falcon, BD Biosciences, San Jose, CA, EUA) in RPMI medium containing 15 mM glutamine (Gibco), 10 mM Hepes (Sigma-Aldrich, St. Louis, MO, EUA), 200 U/mL penicillin (Gibco), 200 µg/mL streptomycin (Gibco), 3 mg/mL gentamicin (Sigma-Aldrich) and 2 g/L sodium bicarbonate (Sigma-Aldrich) supplemented with 10% inactivated fetal bovine serum (Invitrogen). CL suspension was added at concentrations of 1, 10, and 100 µg/mL and incubated for 4, 8, 16, or 24 h at 37◦C and 5% CO2. After culture, cells were labeled with anti-CD14- APC antibody (clone 61D3, eBioscience, Science Center Dr. San Diego, CA, EUA), a marker of macrophages and monocytes, and quantified by flow cytometry in the CyAn cell analyser (Dako Cytomation Inc., Carpinteria, CA, EUA) by counting 60,000 events. Data were analyzed using the software Summit (Dako Cytomation).

### Evaluation of Clodronate-Liposome *in Vivo*

Plasmodium falciparum FUP strain (Carvalho et al., 2004) was recovered from cryostabilates and the parasitized red blood cells (pRBCs) were passaged by intravenous inoculation in a splenectomized donor monkey. pRBCs from the passage animal were used to infect the experimental group animals intravenously with an inoculum of 10<sup>6</sup> pRBC. The animals were divided into six groups, three with infected (Inf) animals and three control, noninfected (NInf) groups: (1) NInf receiving 1 mL of phosphate buffered saline (PBS - Sigma-Aldrich); (2) NInf receiving 0.5 mL of clodronate-encapsulated liposome (CL); (3) NInf receiving 1 mL of CL; (4) Inf receiving 1 mL of PBS; (5) Inf receiving 0.5 mL of CL; (6) Inf receiving 1 mL of CL. CL injections were performed intravenously two times a week, from day 0 of infection. Three independent experiments were performed using a total of 34 animals (**Table 1**).

The follow up of infection included a daily evaluation of parasitemia by thin Giemsa-stained blood films, hematocrit (microhematocrit method) and hemoglogin levels twice a week (using the HemoCue Hb301 system, HemoCue AB, Ängelholm, Sweden) as well as daily measurement of body (rectal) temperature. Blood monocyte count was done, twice a week starting from day 0 of infection, by flow cytometry using anti-CD14-APC antibody (clone 61D3, eBioscience) in a CyAn cell analyser (Dako Cytomation) and the data were analyzed using the software Summit (Dako Cytomation). An experienced veterinarian performed daily clinical examinations. Monkeys were treated with mefloquine (15 mg/kg—Biomanguinhos, Fiocruz, Rio de Janeiro, RJ, Brazil) when parasitemia reached 20% or above or in case the hematocrit reached 20% or below or when the monkeys presented manifestations of severe disease (prostration, anorexia). Monkeys that spontaneously controlled their parasitemia also received mefloquine on day 17 to ensure complete parasite clearance. Four days after mefloquine treatment, with animals no longer presenting patent parasitemia, all animals (including the uninfected controls) received a last injection of CL (or PBS) and were euthanized 24 h later. The animals were euthanized with 7 mg/kg thiopental (Cristália, Itapira, SP, Brazil) inoculated directly into the heart muscle, after being anesthesized with ketamine (Cristália) 100 mg/kg plus midazolam (União Química, Embu-Guaçu, SP, Brazil) 10 mg/kg. After the death of the animal, liver and spleen were harvested for histological analysis.

### Spleen and Liver Analysis

Immediately after liver and spleen harvest, macroscopic evaluation was performed and the spleen was weighted. Samples of the liver and spleen were fixed using a 4% formaldehyde solution (Merk - Darmstadt, Germany) in Millonig's buffer (0.1 M sodium hydroxide, 0.13 M sodium phosphate monobasic—Sigma-Aldrich) and later processed for histology. Paraffin-embedded sections (5 µm) were stained with hematoxylin and eosin (HE), Giemsa or Prussian blue. The slides were analyzed with the aid of a microscope (AxioImager A2, Zeiss, Oberkochen, Germany) and images were captured using an Axiocam HRM (Zeiss) and the software AxionVision Release 4.8.2 (Zeiss).

For hemozoin quantitation in the spleen, the histology slices were dyed with Nuclear Fast Red and mounted with coverslips. Each slide was then completely scanned using VSlide (Metasystems, Germany), generating high definition images (0.8 NA) of all the slices. Because the original images were too big, they were divided into three parts for analysis, by using ImageJ, where the red (nuclei) and brown/black (hemozoin) colors were separated with the color threshold function. The hemozoin areas and the total tissue area in each image were measured in square pixels. These results were then used to estimate the relative hemozoin area in each slide.

### Statistical Analysis

The significance of the differences between the results or means of all variables was examined by the nonparametric Kruskal-Wallis analysis followed by Dunn test (GraphPad Prism 6, GraphPad Software Inc., San Diego, CA, USA). Results were considered to be statistically significant when p < 0.05.

### RESULTS

### Effect of Clodronate-Encapsulated Liposomes on *S. sciureus* Monocytes *in Vitro*

Even though CL has been used as a rapid inducer of apoptosis of monocytes/macrophages in vivo and in vitro, especially in mice, there are no available data on the effect of CL in S. sciureus monocytes. In vitro incubation of S. sciureus splenocytes with CL induced death of CD14+ cells (monocyte/macrophage) in a dose-dependent manner, with marked depletion at 100 µg/mL (**Figure 1**).

## Effect of Clodronate-Encapsulated Liposomes on *P. falciparum* Infection in *S. sciureus* Monkeys

#### Parasitemia

Saimiri sciureus monkeys infected with P. falciparum and receiving PBS showed variable courses of parasitemia (**Figure 2A**). The maximum parasite density varied between 4.8 and 17.7% (peak average 8.1%). Most animals kept parasitemia below 10%, and one showed extremely low parasitemia throughout the follow up. The peak parasitemia was reached between days 11 and 15. All animals were able to control parasitemia without the need for antimalarial drug treatment. The infected animals receiving 0.5 mL CL also showed variable courses of parasitemia, which reached peak levels between days 10 and 15, with parasite densities ranging between 3.9 and 26.7%

FIGURE 1 | Effect of CL on S. sciureus splenocyte viability in vitro. CL induced dose-dependent cytotoxicity in monocytes/macrophages. 24, 149, and PA51 refer to S. sciureus identification. Medium vs. 100 µg/ml: 4 h (p = 0.0003), 8 h (p = 0.007), 16 h (p = 0.011), and 24 h (p = 0.0006). %: percentage of CD14+ cells in total number of splenocytes.

FIGURE 2 | Course of parasitemia after infection of S. sciureus with 10<sup>6</sup> pRBC/mL by FUP strain of P. falciparum. Parasitemia is shown for animals that received PBS (controls) (A); clodronate liposomes (CL) 0.5 mL (B); or CL 1.0 mL (C). Data are from three separate experiments, with 6–9 animals per group. Orange arrows indicate the timepoints when PBS or CL was given to the animals. Red arrows indicate treatment with mefloquine. Dashed line represents the limit of parasitemia established for treatment (20%) with mefloquine. \*Animals that received treatment with mefloquine before day 17 for reaching 20% parasitemia.

(average 16.4%) (**Figure 2B**). Of the six animals in this group, three (50%) required treatment with mefloquine due to high parasitemia (26.7, 19.5, and 26.7%). In the group of infected animals receiving 1 mL CL the maximum parasite density varied between 4.7 and 31.2% (average 18.5%) and peak parasitemia was reached between days 11 and 15. Of the nine animals in the group, six (66.6%) required antimalarial treatment due to high parasitemia (22; 26; 31.2; 25; 20, and 21.5%). Depending on the need for treatment and treatment time, animals received a minimum of four and a maximum of seven CL injections.

#### CD14+ Cell Count in Peripheral Blood

There was no apparent effect of P. falciparum infection or CL injections on circulating monocytes, as verified by CD14+ cell counts in peripheral blood (**Figure 3**).

#### Anemia and Other Clinical Parameters

In P. falciparum-infected animals, hemoglobin concentration and hematocrit decreased reaching minimum values during or after parasite clearance (**Figure 4**). No weight loss greater than 10% was observed. Body temperature increased 1–2◦C during infection and decreased after antimalarial treatment (data not shown). Some animals showed loss of appetite and prostration at the time of peak parasitemia. Two animals that received CL 1 mL and presented high parasitemias showed hematuria at the time of peak parasitemia that reversed after mefloquine treatment. Results of experiments 1, 2, and 3 showing the data of parasitemia, CD14 counting, hematocrit, hemoglobin and rectal temperature are shown in **Supplemental Table 1**.

#### Changes in the Structure of the Spleen and Liver

All infected animals that developed parasitemia over 20% were treated with mefloquine between days 11 and 13 of infection, whereas all other infected animals were treated on day 17. All uninfected and infected-treated animals received a last injection of CL (or PBS) four days after mefloquine treatment, when parasitemia had cleared, and were euthanized 24 h later. Spleen and liver were harvested and processed for histological analysis.

#### **Spleen**

Treatment of uninfected S. sciureus with CL caused no perceivable changes in spleen weight in relation to uninfected animals that received PBS (**Figure 5**). Plasmodium falciparum infection, as expected, led to splenomegaly, with an average 5-fold increase in spleen/body weight ratios compared to uninfected animals. Treatment of P. falciparum-infected S. sciureus with CL 1 mL led to substantial reduction (average of 55%) in the degree of splenomegaly compared to infected animals that received only PBS (**Figure 5** and **Supplemental Table 1**). The decrease in spleen/body weight ratios in animals treated with CL 0.5 mL was less prominent and did not reach significance compared to PBStreated animals, although the lack of significance is likely due to the small sample size.

The spleen of uninfected, control animals that received PBS showed typical S. sciureus splenic architecture, with well-defined limits between the red and white pulps, resting T cell areas, mostly resting follicles, and in some cases with phase I and

phase II germinal centers containing a few apoptotic centers, absence of pigment and plasmacytes, and areas of monocyte accumulation in the marginal zone and the red pulp (**Figure 6A**), as previously described (Alves et al., 2015). Uninfected animals that received CL (0.5 or 1 mL) showed decreased number of clusters of monocyte/macrophages in red pulp and of apoptotic centers in the follicles compared to animals that received PBS (**Figure 6B**).

Infected animals that received PBS showed phagocitosed malarial pigment (hemozoin) throughout the red pulp (**Figure 6C**). Erythroid precursors and plasmacytes were observed, and the marginal zone was in disarray (**Figure 6C**). Infected animals that received CL 1 mL showed substantial reduction in the accumulation of malarial pigment in the red pulp (**Figure 6D**), despite the fact that these animals in general showed much higher parasitemias (average 18.5%) as compared to infected animals that received PBS (average 8.1%), although for shorter periods of time (**Figure 2**).

The analysis of spleen sections stained with Perls, which shows ferric iron (Fe3+) within macrophages and can be used not only to observe the physiology of iron recycling but also to estimate the distribution of these cells, revealed that uninfected, control animals showed clusters of iron-containing macrophages distributed throughout the red pulp, and mild staining within follicles (**Figure 7A**). Uninfected, control animals that received CL 0.5 and 1 mL showed reduced amounts of iron-containing macrophages (**Figure 7B**).

Infected animals treated with PBS showed iron staining in the red pulp, but substantially less than uninfected controls (**Figure 7C**). However, as mentioned above (**Figure 6C**), large numbers of hemozoin-containing macrophages were observed. As a rule, there was little or absent colocalization of iron

FIGURE 4 | Hemoglobin concentration in the blood (A) and hematocrit (B) of S. sciureus. NInf, non-infected control groups; Inf, P. falciparum-infected groups. Solid lines represent the mean and standard deviation.

staining and hemozoin (macrophages showed only one staining, **Figure 7C**). Infected animals treated with CL showed less iron staining as compared to animals that received PBS, and also less hemozoin staining (**Figures 7D**, **10**).

well-defined limits between the red pulp (RP) and the white pulp (WP) was observed.

#### **Liver**

With HE staining, livers of uninfected control animals that received PBS showed typical structure, with healthy hepatocytes and absence of necrosis or vacuolization and Kupffer cells without pigment (**Figure 8A** and **Supplemental Figure 1A**). Mononuclear cell infiltrates in the portal space were occasionally observed. Livers of animals that received CL 0.5 or 1 mL showed either no significant changes or, in some cases, evidence of hepatocyte vacuolization. **Figure 8B** shows one of these areas of hepatocyte vacuolization, an event that was only occasionally found but may indicate some degree of toxicity induced by CL treatment.

Infected animals that received PBS showed large numbers of mononuclear cells and also erythroid cells, neutrophils and plasmacytes within sinusoids, mononuclear cell infiltration in the portal space, and Kupffer cells containing malarial pigment (**Figure 8C** and **Supplemental Figures 1B–D**). Diffuse vacuolization of hepatocytes was observed in two thirds of the animals. Infected animals that received CL showed areas of intense hepatocyte vacuolization, and there were apparently fewer Kupffer cells with malarial pigment than in infected animals that received PBS (**Figure 8D**).

Perls staining of liver sections revealed that uninfected animals that received PBS showed strong iron staining throughout the organ, with granules within hepatocytes and Kupffer cells (**Figure 9A**). Iron staining in liver sections of uninfected animals that received CL 0.5 or 1 mL was less intense than in those that received PBS (**Figure 9B**).

Infected animals that received PBS showed large amounts of hemozoin. Iron staining was weak throughout the organ, but still present in hepatocytes and isolated Kupffer cells or adherent macrophages within sinusoids, with little or no colocalization with hemozoin (**Figure 9C** and **Supplemental Figures 2A,B**).

Infected animals that received CL 0.5 or 1 mL showed less intense hemozoin and iron staining, and Kupffer cells or macrophages with strong iron staining were rarely observed or not at all (**Figure 9D**).

### Estimating the Efficacy of CL Treatment in Depleting Macrophages through Splenic Hemozoin Quantification

Qualitative analysis of spleen and liver, as shown in **Figures 6**– **9**, and decreased spleen/body weight ratios (**Figure 5**) indicated that CL treatment led to a decrease in macrophage population in these organs. Attempts to quantify macrophages using immunostaining with anti-human CD14, CD68, and HAM-56 monoclonal antibodies were unsucessful, probably due to the prolonged period in formalin. The effect of CL treatment on splenic macrophage numbers was therefore further estimated by quantifying hemozoin accumulation in this organ, which increases with increasing parasitemia (Sullivan et al., 1996), and hemozoin-laden macrophages remain in the spleen for several weeks (Frita et al., 2012). Although, CL-treated monkeys showed about half the splenic hemozoin accumulation as compared to PBS-treated animals (**Figure 10A**), this difference did not reach statistical significance due to small sample size and to some degree of variation between animals. However, when hemozoin concentration was adjusted to the level of parasitemia (hemozoin concentration/peak parasitemia, **Figure 10B**; hemozoin concentration/area under the curve of parasitemia, **Figure 10C**), marked differences were observed between the two groups, indicating that despite sustaining much higher parasite loads CL-treated animals showed less hemozoin accumulation in the spleen. These data indicate that splenic macrophages of CL-treated animals are either markedly decreased in numbers, as also suggested by HE staining and spleen weights, or are functionally impaired, in any case resulting in decreased capacity to phagocytose parasitized red blood cells.

### DISCUSSION

Saimiri monkeys are susceptible to infection by Plasmodium falciparum and, together with Aotus, constitute unique experimental models for malaria (WHO, 2004), making these animals particularly useful in preclinical trials of potential

malaria vaccines and for studies on pathogenesis (Carvalho et al., 2000, 2004, 2005). Pre-clinical evaluation of vaccines in these experimental models can provide valuable information about the immunogenicity, efficacy and safety of a variety of formulations, facilitating selection for humans trials. Since access to this valuable resource is restricted, with limited number of animals available for research, optimization of the model is needed to allow the design of experiments with reduced sample sizes. This goal faces two important limitations regarding the use of non-splenectomized animals: (1) they usually develop low parasitemias, which makes it more difficult to detect the effect of interventions, such as vaccines, on the course of infection; and (2) there is substantial interindividual variability in the levels of parasitemias (Contamin et al., 2000). These two factors underscore the need to increase sample sizes in experiments using these animals. The alternative to overcome these limitations has been the use of splenectomized animals, which achieve higher and more consistent parasitemias, allowing smaller experimental groups to be formed. However, this strategy also faces important setbacks: (a) the spleen plays a central role in immunity against plasmodial infections and (b) non-human primates (NHP) are chosen as models for experimental studies with human plasmodia, mainly vaccines, because of their close filogenetic relationship to man. Splenectomy pushes back the model from the natural human situation. For these reasons, splenectomy constitutes a severe handicap especially in immunity and vaccine studies. The use of the "chemical (temporary) splenectomy" described here offers therefore an animal model closer to the target population condition.

In plasmodial infections, the spleen helps to control parasite burden through innate and adaptative immune responses (Yazdani et al., 2006; Portillo et al., 2012; Gazzinelli et al., 2014). Splenic macrophages are major players in this system, clearing parasitized red blood cells through phagocytic activity. Because S. sciureus are not P. falciparum's natural host, splenic clearance seems to be particularly effective, preventing higher parasitemias. We asked whether partial depletion of the macrophage population in the spleen and other sites would result in higher and more consistent parasitemias in S. sciureus while maintaining not only the structure but also the function of the spleen. Together, our results indicate that this goal was largely achieved in this study.

Saimiri sciureus monkeys infected with Plasmodium falciparum and treated with periodic intravenous injections of clodronate-encapsulated liposomes (CL) in general developed higher parasitemias requiring antimalarial treatment than

infected monkeys treated with PBS. It is not clear why this effect was not observed in three out of 9 (33.3%) infected S. sciureus treated with CL 1 mL, as they were able to self-control parasitemia, behaving closer to PBS-treated animals. Three independent experiments were performed, and these three animals were from the same experiment, suggesting that the lot of CL used may not have been optimal. Alternatively, this result would confirm that there is some degree of individual animal variation, either in the efficacy of CL treatment or in the pre-existing state of non-adaptive immunity. It is likely that the number of macrophages will vary from one animal to the other. It is advisable that each lot of CL be assayed in vitro for monocyte killing capacity prior to the conduction of in vivo experiments.

In any case, the efficacy of CL 1 mL treatment was evident, as two thirds (6 out of 9) of the treated animals developed parasitemias over 20% requiring mefloquine treatment, whereas all eight PBS-treated animals self-controled their parasitemias. Administration of a smaller amount of CL (0.5 mL) also resulted in higher parasitemias overall, with half (3 out of 6) of the animals showing parasitemias over 20%. Results are in agreement with those recorded with P. falciparum in immunocompromised mice grafted with human RBCs (Badell et al., 2000; Arnold et al., 2010). Thus, it is very likely that higher CL doses would result in improved macrophage control, and thereby in higher parasitemias. This remains to be investigated in future experiments.

From an ethical point of view CL treatment offers the paramount advantage of avoiding heavy surgery and allowing full recovery of animals. It thus constitutes a major improvement from a scientific as well as from an ethical viewpoint over splenectomy.

Plasmodium falciparum infection and/or CL treatment had little effect on peripheral blood monocyte counts, which is not a surprise as young monocytes have limited phagocytic activity and CL targets preferentially large and active macrophages, which ingest a larger number of liposomes per cell. Indeed, CL injections resulted in marked effects in the spleen and the liver. In uninfected animals, no evident effect of CL injections was observed in spleen/body weight ratios, but the number and distribution of ferric iron-containing cells in the spleen were markedly reduced. Macrophages are involved in senescent red blood cell phagocytosis and iron recycling, and therefore CL injections largely depleted this cell population in the spleen. In Plasmodium falciparuminfected animals, CL injections resulted in milder splenomegaly and this effect was associated histologically with marked

reductions in hemozoin-containing cells and further reductions in iron-containing cells. Efforts to further substantiate the histological findings by specific macrophage staining with commercially available anti-human CD14, CD68, and HAM-56 monoclonal antibodies were unsucessful, probably due to

under the curve (AUC) of parasitemia (C) of each animal. \*\*P < 0.01.

the prolonged period the biological material was kept in formalin.

Our results also indicate that P. falciparum infection in S. sciureus causes impaired iron-recycling by macrophages. Infected animals showed large numbers of hemozoin-containing macrophages but decreased numbers of ferric iron-containing cells in the spleen. An effect on iron homeostasis was also observed in the liver. It is known that macrophages are very important actors in controlling the iron upload and metabolism in hepatocytes, mostly via hepcidin. Therefore, it is possible that the reduction in iron staining in hepatocytes after CL treatment resulted from the instability of the hepcidin production due to Kupffer cell destruction (Fleming, 2005; Makui et al., 2005). However, additional studies are needed in order to clarify the mechanisms behind this phenomenon. In addition, colocalization of hemozoin and iron was a rare event, indicating that phagocytes were busy processing hemozoin from infected red blood cells and their capacity to carry out phagocytosis of uninfected red blood cells was impaired. Although the dark and gross pattern of hemozoin might actually prevent visualization of the light blue iron staining and therefore colocalization might be underestimated, some facts suggest that iron uptake by macrophages was impaired: (i) even macrophages with little, sparse amount of hemozoin showed no iron staining; (ii) colocalization was observed in some cells heavily laden with hemozoin. Since it has been shown that hemozoin may persist for weeks to months after parasite clearance (Frita et al., 2012; Alves et al., 2015), it may negatively impact the immune responses and iron recycling of the affected individuals for long periods of time after treatment has been implemented and therefore, contribute significantly to increasing malaria morbidity. Interestingly, in infected animals, Kupffer cells and sinusoidal macrophages heavily laden with iron were observed. This finding may indicate that young cells from the bone marrow repopulate the liver in an effort to restore iron recycling and other functions.

Depletion of macrophages by CL shows a number of advantages over surgical splenectomy for the study of malaria vaccines in Saimiri monkeys. However, macrophages are very important in the immune responses against P. falciparum through direct infected red blood cell phagocytosis or antibodymediated parasite opsonization or inhibition (Bouharoun-Tayoun et al., 1990; Buffet et al., 2011). Therefore, sharp decreases in macrophage numbers might affect the antiplasmodial immune responses a given pre-clinical vaccine trial may be attempting to evaluate. The same limitation can be considered for dendritic cells, which also play relevant roles in antiplasmodial immunity (Urban et al., 1999, 2001). Indeed, phagocytic marginal dendritic cells but not interdigitating dendritic cells are known to be depleted by CL treatment (Leenen et al., 1998), which can affect immune responses to particulate antigens (Delemarre et al., 1990). These limitations need to be taken into account when using CL treatment as the strategy to induce higher P. falciparum parasitemias.

It has been shown that plasmodial infections in humans (Urban et al., 2005), Saimiri (Alves et al., 2015) and mice (Achtman et al., 2003; Carvalho et al., 2007; Martins et al., 2009) induce disarray of the spleen and other lymphoid organs, with disturbance of germinal center architecture. This finding was confirmed in the present study in S. sciureus, and CL administration had no apparent effect on this event. The splenic changes observed in P. falciparum-infected, saline-treated Saimiri 5 days after mefloquine treatment were similar to those recently reported for P. falciparum-infected Saimiri at peak parasitemia before antimalarial treatment (Alves et al., 2015). These changes included the presence of large numbers of phagocytes in the red pulp heavily laden with malaria pigment, disarray of B-cell follicles, blurred limits between the red and white pulps, and penetration of RBCs into follicles. In that study, 14 days after chloroquine treatment, spleens showed better white pulp organization and persistence of hemozoin in the red pulp but with a different pattern with more compacted, less granulous pigment (Alves et al., 2015). Therefore, the pattern of splenic changes observed in infected animals 5 days after mefloquine treatment in the present study was more similar to the changes observed at peak parasitemia prior to antimalarial treatment than after an extended period (2 weeks) after antimalarial treatment. This finding suggests that 5 days following antimalarial drug treatment is insufficient to reverse the pathological changes in spleen structure induced by P. falciparum infection.

Finally, infected animals that received CL showed more intense hepatocyte vacuolization. This effect was most likely a result of a combination of CL injection with higher parasitemias in these animals, because both uninfected animals receiving CL and infected animals receiving PBS showed some degree of hepatocyte vacuolization, milder than when the events were combined.

In conclusion, these results indicate that CL depleted splenic macrophages leading to decreased parasite phagocytosis, decreased splenic congestion, and increased parasitemias. CL administration is therefore a practical and viable alternative to surgical splenectomy in this experimental model.

### AUTHOR CONTRIBUTIONS

JC participated in study design, carried out the experiments and helped LC in drafting the manuscript; CBJ participated in study

### REFERENCES


design and carried out the experiments, MA, LP-R, ER, and MP carried out the experiments and reviewed the manuscript; IdS carried out the experiments; LC, CD-R, and PD conceived the study, participated in its design and coordination, and reviewed the manuscript. All authors have read and approved the final manuscript.

### FUNDING

This study was supported by Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES), Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq, Brazil), Fundação Carlos Chagas Filho de Apoio de Amparo à Pesquisa do Estado do Rio de Janeiro (FAPERJ, Brazil), Instituto Oswaldo Cruz (FIOCRUZ, Brazil) and Programa de Apoio a Núcleos de Excelência-PRONEX (DECIT/CNPq/FAPERJ). CD-R and LC are recipients of Research Productivity fellowships from CNPq and are supported by FAPERJ as "Cientistas do Nosso Estado."

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fcimb. 2017.00408/full#supplementary-material

Supplemental Figure 1 | Liver sections, HE (bar: 100 µm). (A) Liver of an uninfected, control animal that received PBS (code PA67), in higher magnification than Figure 8A, showing Kupffer cells (arrows). (B–D) Liver of a P. falciparum-infected Saimiri that received PBS (code 141: treated at day 17, with 0.69% parasitemia, killed 5 days later), in greater detail compared to Figure 8C. Portal infiltrates (black arrow), increased cellularity in sinusoids with presence of mononuclear cells (white arrowhead), erythroid cells (thin arrow) and plasma cells (black arrowhead).

Supplemental Figure 2 | Liver sections, Perls (bar: 100 µm). (A,B) Liver of a P. falciparum-infected Saimiri that received PBS (code 216: treated at day 17, with 1.3% parasitemia, killed 5 days later), in greater detail compared to Figure 9C. Intrasinusoidal macrophages, circulating or attached to the sinusoidal wall, laden with malarial pigment (white arrows). There was little or no colocalization of hemozoin and iron staining (black arrow).

Supplemental Table 1 | Results of experiments 1, 2, and 3 showing the data of parasitemia, CD14 counting, hematocrit, hemoglobin, rectal temperature, weight and spleen/body weight ratios.


in humans: insights from splenic physiology. Blood 117, 381–392. doi: 10.1182/blood-2010-04-202911


peripheral blood mononuclear cells from squirrel monkeys (Saimiri sciureus). J. Med. Primatol. 73, 686–693. doi: 10.1111/j.1600-0684.1998.tb00074.x


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

Copyright © 2017 Cunha, Carvalho, Bianco-Junior, Andrade, Pratt-Riccio, Riccio, Pelajo-Machado, da Silva, Druilhe and Daniel-Ribeiro. 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.

# Giardia's Epithelial Cell Interaction In Vitro: Mimicking Asymptomatic Infection?

#### Martin R. Kraft 1, 2 \*, Christian Klotz <sup>1</sup> , Roland Bücker <sup>2</sup> , Jörg-Dieter Schulzke<sup>2</sup> and Toni Aebischer <sup>1</sup> \*

<sup>1</sup> Unit 16 Mycotic and Parasitic Agents and Mycobacteria, Robert Koch-Institute, Berlin, Germany, <sup>2</sup> Institute of Clinical Physiology, Charité Campus Benjamin Franklin, Berlin, Germany

The protozoan parasite Giardia duodenalis is responsible for more than 280 million cases of gastrointestinal complaints ("giardiasis") every year, worldwide. Infections are acquired orally, mostly via uptake of cysts in contaminated drinking water. After transformation into the trophozoite stage, parasites start to colonize the duodenum and upper jejunum where they attach to the intestinal epithelium and replicate vegetatively. Outcome of Giardia infections vary between individuals, from self-limiting to chronic, and asymptomatic to severely symptomatic infection, with unspecific gastrointestinal complaints. One proposed mechanism for pathogenesis is the breakdown of intestinal barrier function. This has been studied by analyzing trans-epithelial electric resistances (TEER) or by indicators of epithelial permeability using labeled sugar compounds in in vitro cell culture systems, mouse models or human biopsies and epidemiological studies. Here, we discuss the results obtained mainly with epithelial cell models to highlight contradictory findings. We relate published studies to our own findings that suggest a lack of barrier compromising activities of recent G. duodenalis isolates of assemblage A, B, and E in a Caco-2 model system. We propose that this epithelial cell model be viewed as mimicking asymptomatic infection. This view will likely lead to a more informative use of the model if emphasis is shifted from aiming to identify Giardia virulence factors to defining non-parasite factors that arguably appear to be more decisive for disease.

#### Keywords: Giardia, giardiasis, TEER, barrier function, Caco-2, transwell, permeability

## INTRODUCTION

Giardia duodenalis (also Giardia lamblia or Giardia intestinalis) is an ubiquitous protozoan parasite of the order diplomonadida and forms a species complex of eight different phylogenetic groups (assemblages) characterized by different host specificities. Assemblages A and B are infectious to humans and other mammals (Thompson and Monis, 2012). It is responsible for more than 280 million symptomatic cases of infection, i.e., cases of giardiasis associated with gastrointestinal malfunctions (see below), every year, worldwide. Infections are acquired by uptake of dormant cyst forms of the parasite mostly via contaminated drinking water, and highest prevalences are reported for countries with poor access to higher sanitation standards, though it is one of the most common reported parasitic infections in other countries as well. Since 2004, more attention is paid to this pathogen due to WHO's "Neglected Diseases Initiative" (Adam, 2001; Savioli et al., 2006; Ankarklev et al., 2010). Giardia sp. cysts transform into their trophozoite stage after passing the host's stomach

#### Edited by:

Mario Alberto Rodriguez, Center for Research and Advanced Studies of the National Polytechnic Institute (CINVESTAV), Mexico

#### Reviewed by:

Maria Del Consuelo Gomez, Instituto Politécnico Nacional, Mexico James H. McKerrow, University of California, San Diego, United States Elisa Azuara-Liceaga, Universidad Autónoma de la Ciudad de México, Mexico

#### \*Correspondence:

Martin R. Kraft kraftm@rki.de Toni Aebischer aebischera@rki.de

Received: 13 June 2017 Accepted: 12 September 2017 Published: 26 September 2017

#### Citation:

Kraft MR, Klotz C, Bücker R, Schulzke J-D and Aebischer T (2017) Giardia's Epithelial Cell Interaction In Vitro: Mimicking Asymptomatic Infection? Front. Cell. Infect. Microbiol. 7:421. doi: 10.3389/fcimb.2017.00421 and the latter forms colonize the duodenum and upper jejunum where they replicate, attach to the intestinal epithelium using their adhesive disc and feed on luminal nutrients.

### THE CHALLENGE: SYMPTOMATIC VS. ASYMPTOMATIC INFECTIONS

In the context of this article, we define a symptomatic infection, i.e., giardiasis disease, as being characterized by acute gastrointestinal complaints, like diarrhea, abdominal pain, nausea, and vomiting (Adam, 2001; Ankarklev et al., 2010) and we like to distinguish it from asymptomatic infections defined by the absence of such acute symptoms. Asymptomatic infections according to this definition would include infections that particularly if recurring—result in malabsorption/malnutrition phenotypes or may represent pure colonization without any pathology, which was recently shown by Garzon et al. (2017) on asymptomatic children, who found no correlation between Giardia-infection and barrier dysfunction, but only of barrier dysfunction and wasting and stunting. However, infections may also trigger post-infectious syndromes such as irritable bowel disease (D'Anchino et al., 2002; Wensaas et al., 2012; Hanevik et al., 2014; Litleskare et al., 2015; Halliez et al., 2016; Nakao et al., 2017).

The reasonsfor these fundamental differences of the outcomes of infections remain unclear (Adam, 2001; Troeger et al., 2007; Klotz and Aebischer, 2015; Tysnes and Robertson, 2015). A mechanism that was proposed to link to acute symptoms was the breakdown of the intestinal barrier function of the epithelium, leading to increased permeability with bacterial invasion as a possible result (Buret, 2007; Ankarklev et al., 2010). This was probably inspired by the impressive in vitro phenotypes obtained with certain bacterial, gastrointestinal pathogens and their products/toxins/proteases (Malago et al., 2003; Fajdiga et al., 2006; Rees et al., 2008; Liu et al., 2012; Anderson et al., 2013; Fiorentino et al., 2013) as well as other parasitic protozoans like Entamoeba histolytica, Cryptosporidium parvum, or Blastocystis sp. (Li et al., 1994; Leroy et al., 2000; Buret et al., 2003; Betanzos et al., 2014; Wu et al., 2014). It is fair to say that this has motivated numerous studies with the implicit goal of identifying a similar functional correlate of acutely symptomatic giardiasis in in vitro models. Here, we review in vitro investigations that aimed at finding such a robust correlate for acute clinical symptoms and conclude that, frustratingly, this goal was largely missed. We hypothesize that this is because these models reproduce the prevailing inconspicuous course of G. duodenalis infection, rather than one that is associated with acute symptoms. We propose, as others did (Bartelt and Sartor, 2015), that nonparasite factors in combination with G. duodenalis infection are most likely causing acute symptoms and adapting in vitro models to search for such factors should be attempted. We also propose that a number of epithelial functions that may give insight into non-acute but probably more relevant symptoms linked to malabsorption/malnutrition phenotypes have not yet been investigated in vitro but should become a focus of future work.

Two studies can serve to illustrate the dilemma that we face when trying to define pathological correlates of Giardia infection at the histological/cellular level. One study performed by Oberhuber et al. (1997) analyzed 567 Giardia-positive cases identified in a retrospective study of gastrointestinal biopsy samples collected consecutively over an eight-year period in the frame of a gastrointestinal pathology service in Germany. This number of Giardia positive cases represented ∼0.3% of all samples analyzed in that period. Reasons for soliciting histopathological analyses were described as follows: "most of the patients experienced unspecific gastrointestinal complaints that prompted an upper endoscopy" (Oberhuber et al., 1997). The authors conclude that "the histology of the small-bowel mucosa is inconspicuous in most subjects with giardiasis." It is reasonable to assume that such a consecutive study on biopsies sampled a random cross-sectional collection of mostly adult patients and therefore, it is justified to conclude that most Giardiainfections do not lead to acute symptoms. In contrast, a study focusing on acutely symptomatic cases reported epithelial barrier dysfunction, altered tight junction composition, and increased signs of apoptosis (Troeger et al., 2007).

### Epidemiological Findings on Epithelial Dysfunction

There have been several epidemiological studies investigating functional correlates of disease such as gut permeability in different patient collectives suffering from G. duodenalis infection. Very recently Rogawski et al. (2017) and Kosek (2017), through a multisite birth-cohort study, showed that persisting infection within the first 6 months of life was correlated with reduced weight and height for age Z scores at 2 years of age. However, this was not dependent on diarrhea, i.e., independent of acutely symptomatic infection. More likely, stunting was the result of changes to gut permeability, assessed by Lactulose (L) and mannitol (M) excretion assays according to which permeability was positively correlated with parasite detection, which agrees with some earlier findings (Dagci et al., 2002; Goto et al., 2002) but contradicts others (Serrander et al., 1984; Campbell et al., 2004; Goto et al., 2009). Even studies using the same method vary in their results. For example, Goto et al. (2002) assessed in a cross-sectional study intestinal parasite infection and permeability of 210 Nepali children aged 0–60 months who were living in poor conditions. Fourteen percent of 173 were acutely infected with Giardia. L:M ratios suggested increased permeability in Giardia-infected (0.43; n = 8) compared to uninfected children (0.25; n = 45; p = 0,014). However, standard deviations were very high and the stepwise multiple and logistic regression analyses used is known to be prone to type I errors (Burnham and Anderson, 2004). On the other hand, in a later longitudinal study by Goto et al. (2009) the examination of 298 children from Bangladesh until their second year of age and with comparable living conditions did not find a correlation of L:M ratios and Giardia-infection. Similar results were found in Gambian children (2–15 months of age) by another longitudinal L:M study by Campbell et al. (2004). In other studies, even a negative correlation of Giardia-infection to severe diarrhea could be observed (Bilenko et al., 2004; Kotloff et al., 2013; Muhsen et al., 2014).

Overall, epidemiological evidence for altered permeability as a consequence of frequent G. duodenalis infection exists. Effects are weak and appear to be inconsistent probably because data are confounded by non-acute, asymptomatic cases included in studied patient collectives. However, the evidence is rather against G. duodenalis infection alone as being a clear cause of acute symptoms such as diarrhea. In view of this data, the quest is open for improved in vitro models and respective readout parameters that may link to distinct symptoms, such as changes of permeability, molecular transport, and absorption mechanisms.

### Investigations of Epithelial Barrier Functions in Vitro

Thinking of giardiasis disease as being linked to different syndromes of altered gut barrier function, acute and non-acute, should allow derivation of distinct readout parameters as proxies for the diverse symptoms when interrogating in vitro cell culture models with Giardia parasites. Current readouts include probing electrophysiological properties of epithelial cell monolayers such as trans-epithelial electric resistance (TEER), assessing transport and permeability changes using tracer molecules, or inferring monolayer integrity from changes of abundance and localization of proteins that make up tight junctions (**Figure 1**). In addition, epithelial tissues also function as communication layers relaying signals to other organs such as the immune system, and cytoand chemokine responses were also used as readout parameters to characterize the effect of infection.

### TEER as a Surrogate of Acute Effects on Tissue Integrity

The measurement of TEER as an indicator for paracellular permeability is a well-accepted method to estimate acute pathophysiological effects on cell barrier function of intestinal epithelial cells (Srinivasan et al., 2015). An important prerequisite for a robust readout is the presence of an electro-physiologically intact monolayer. TEER measurements to investigate the effect of Giardia trophozoites on epithelial function have been used by several groups in diverse setups but were implemented with possibly relevant experimental differences (Supplementary Table 1). Our own data (**Figure 2**) suggest that exposure of Caco-2 cells in vitro to diverse G. duodenalis isolates derived from samples of symptomatic patients does not affect epithelial monolayer integrity. This is in good agreement with some previous studies (Chavez et al., 1986, 1995; Tysnes and Robertson, 2015) but is in contrast to others (Teoh et al., 2000; Humen et al., 2011; Maia-Brigagao et al., 2012), who described a TEER decrease of up to 45%. If one assumes, as also proposed by others (Bartelt and Sartor, 2015), that G. duodenalis infection alone is insufficient to cause acute disease and, therefore, that infections are mostly asymptomatic, then one would not expect an effect on TEER. Since several studies have reported the contrary, however, it seems justified to analyze possible reasons for the discrepancies as they may point toward the currently elusive additional factors that are postulated here and elsewhere (Bartelt and Sartor, 2015) to precipitate acute symptoms.

As listed in Supplementary Table 1, studies differ in experimental detail and it is of interest to analyze whether these could have affected the overall outcome of the respective TEER measurements. For example, one variable of the popular Caco-2 model is that several clonal Caco-2 populations exist and their response can vary according to a particular readout (Katelaris et al., 1995; Sambuy et al., 2005; Liévin-Le Moal, 2013; Srinivasan et al., 2015). Therefore, we tested the parental cellline and the clonal cell line Caco-2 bbe in a similar setting with the same result (Supplementary Figures 1A,B). Of note, different subpopulations—also unintentionally established by ongoing passaging of those heterogeneous cells—may explain highly different basic TEER values between setups, which range from ∼160 cm<sup>2</sup> (Teoh et al., 2000) to >1,200 cm<sup>2</sup> (Scott et al., 2002). In our experiments, Caco-2 (parental) had a basic TEER of ∼300 cm<sup>2</sup> , whereas Caco-2 bbe offered initially ∼200 cm<sup>2</sup> (Supplementary Figure 2) but increased this value in a linear manner after 37 passages to 350 cm<sup>2</sup> at passage #53 (Supplementary Figure 3), which is a known phenomenon (Sambuy et al., 2005; Srinivasan et al., 2015).

Parasite number is another factor that varies between studies, where MOIs of 0.5–8 have been described (Humen et al., 2011). In our experiments, we used MOIs of similar range (1 and 10; Supplementary Figure 1C) as well as very high parasite doses up to MOIs of 100 (**Figure 2B** or Supplementary Figures 1A,B). Thus, it seems unlikely that parasite:enterocyte ratios can explain conflicting results between studies. Specific characteristics of chosen parasite isolates cannot be excluded, but several identical isolates were tested in more than one study—including our own experiments—with different outcome (Supplementary Table 1). Thus, other differences in experimental setups must be considered.

It has been discussed that trophozoites' physical attachment alone may increase TEER and therefore obscure possible barrier defects (Chavez et al., 1995; Teoh et al., 2000; Chin et al., 2002; Tysnes and Robertson, 2015). However, exposure to the potent Giardia detachment reagent formononetin, known to remove trophozoites almost instantly without affecting their viability (Fisher et al., 2013), also did not led to normal TEERs of infected conditions (Supplementary Figure 4A). Though, formononetintreatment right after infection seem to reduce its magnitude (Supplementary Figure 4B), this contrasts to data by Humen et al. (2011), who noted a TEER decrease which was dependent on trophozoite attachment that could not be triggered by spent medium. However, spent medium and sonicates led to a decline in TEER in other studies (Teoh et al., 2000).

In many studies, infected monolayers were washed intensively with ice cold PBS to remove parasites and subsequently measured TEER (Supplementary Table 1). Our experiences with such washing procedures led to erroneously high TEERs and highly increased variances of measurements, an effect that required several hours of re-incubation to normalize (Supplementary Figure 5). In experiments that do not rely on washing, such as the use of sonicated lysates or trophozoite-conditioned DMEM, Caco-2 monolayers were shown to reduce their TEERs when

FIGURE 1 | Epithelial barrier. Scheme depicts the epithelia barrier and possible ways of its impairment. Tight junctions are located at the lateral sides in apical proximity and consist of membrane-spanning proteins like occludin and different claudins, as well as scaffold proteins like ZO-1 which connect the cytosolic ends of tight junction proteins to the actin cytoskeleton. Exchange of those proteins, leading to different tight junction compositions, affect specific permeability. Degradation of tight junctions is followed by an increase of unspecific permeability. Loss of cellular contacts lead to apoptosis and vice versa, apoptosis induces tight junction degradation. With measuring TEER, open breaches in the epithelium can be detected directly. The use of labeled molecules can also detect such leaks, however molecules can also be actively transported transcellularly e.g., via pinocytosis.

FIGURE 2 | Isolate comparison and dose dependency. We analyzed the effects of Giardia colonization of epithelial monolayers on TEER using the Caco-2 clone bbe, since it is reportedly more homogeneous than other clones or its parental line (Sambuy et al., 2005; Liévin-Le Moal, 2013; Srinivasan et al., 2015). Although, Caco-2 cells have been derived from a human colon cancer patient, a 21-day phase of confluent incubation ensures their differentiation toward a polarized small intestinal enterocyte-like cell type, considered as a model for the small intestinal epithelial barrier (ibid.). In this setup, we tested 11 different G. duodenalis isolates at MOIs of 20, including 5 reference strains WB6, NF, S2 (assemblage AI), GS (assemblage B), P15 (assemblage E) and 6 newly established clinical isolates (1 assemblage AII and 5 assemblage B). Data is relative to measurements before infection. Uninfected controls (CTRL-) were sham treated, monolayers with induced apoptosis using 1 µM staurosporine served as positive (leaky) controls (CTRL+). Each point represents the mean of 3 independent experiments with 3 monolayers per condition per experiment. Error bars indicate standard deviation. No evidence for parasite-induced decreases in TEER was found. In contrast, all tested isolates led consistently to a dose-dependent TEER-increase, without significant differences between isolates or assemblages.

exposed to, for example, NF or S2 lysates already after 24 h (Teoh et al., 2000). By destroying compartmentalization, lysates may lead to release of factors such as proteases or other enzymes that affect epithelial cell function (Buret, 2007; Cotton et al., 2011, 2014; Bhargava et al., 2015). However, in our hands WB6 lysates and heat-inactivated trophozoites, both corresponding to a MOI of 20, did not influence Caco-2 monolayers' TEER (Supplementary Figure 6A), while filtered apical supernatants from a WB6-MOI-100-72-h condition weakly recapitulated the TEER increase noted in experiments with intact parasites (Supplementary Figure 6B), indicating that products of vital Giardia trophozoites, such as products of secreted microvesicles (Evans-Osses et al., 2017) or stressed Caco-2 cells, may be responsible for the observed TEER increase.

Another variation in experimental design is the time of culture and differentiation before infection (Supplementary Table 1). Since monolayers change TEER properties over time of cultivation (Supplementary Figure 2) and less differentiated cells may be more sensitive to disturbing stimuli, findings can be confounded which can explain the disparate results in seemingly similar studies (Teoh et al., 2000; Humen et al., 2011; Maia-Brigagao et al., 2012, and for own data see Supplementary Figure 7).

The medium composition is another important parameter in Giardia co-culture experiments. Complete substitution of apical DMEM volume with Giardia growth-medium TYI-S-33 can lead to a steep initial TEER increase of monolayers, that can be further accentuated if monolayers are infected with WB6 trophozoites. However, the presence of TYI-S-33 will eventually lead to the collapse of infected and non-infected monolayers (Chavez et al., 1986; Supplementary Figure 8). Thus, takeover of TYI-S-33 components can clearly be a confounder.

Although, presence of O<sup>2</sup> for a microaerophilic pathogen is of relevance, this parameter does not seem to be decisive either. When aerobic and anaerobic conditions were compared using Caco-2 with WB6 trophozoites, anaerobic conditions showed no TEER decrease within 77 h, but indicated a higher TEER increase in both infected and uninfected Caco-2 monolayers. Prolonged incubation led to barrier failure within 139 h in all anaerobically cultivated Caco-2 monolayers, whereas TEER of aerobic cultivated monolayers remained stable and—for infected conditions—elevated (Supplementary Figure 9). Of note, viability of WB6 in our setups remained high with 57–75% of trophozoites still alive after 48 h and 32–48% after 72 h in unmodified culture conditions (data not shown).

The sodium-dependent glucose cotransporter (SGLT)-1 has been shown to inhibit enterocyte apoptosis in Caco-2 cells under high glucose conditions (Yu et al., 2005, 2008). Therefore, it is possible that depending on the cell culture media applied apoptotic effects of Giardia are masked by high glucose concentrations, e.g., in the often-used standard DMEM. However, such an effect is unlikely to have occurred in studies that used the parental Caco-2 since they do not express SGLT-1 in relevant amounts (Turner et al., 1996; Yu et al., 2008). In studies that used the Caco-2 bbe line which express this transporter (Turner et al., 1996) this is difficult to rule out, but in our experiments with the Caco-2 bbe line and low glucose conditions, comparable to Yu et al. did not affect TEER values when compared with cultures using normal DMEM (Supplementary Figure 10).

Serum is another possible component, capable of masking apoptosis induction, since it is known to contain corresponding inhibitors (Zoellner et al., 1996). In our experiments without FBS, however, no significant differences regarding Giardiainfection were found (Supplementary Figure 7). In contrast, the magnitude of effects of staurosporine which was used as a apoptosis-inducing control on TEER was clearly affected by FBS as predicted (Zoellner et al., 1996).

Another interesting confounder may be the Giardia lamblia virus (GLV). It is described as a double-stranded RNA virus of 7 kb size and belongs to the family Totiviridae (Wang and Wang, 1986; Janssen et al., 2015). GLV can infect several but not all Giardia isolates (Miller et al., 1988), depending on the expression of a specific surface receptor (Sepp et al., 1994). The related Leishmania-specific endosymbiont "Leishmania RNA virus-1" (LRV1) is known to affect the severity of leishmaniasis (Ives et al., 2011). Likewise, GLV could also influence Giardia's virulence, though no correlation regarding GLV infestation of Giardia isolates to symptomatic or asymptomatic patients were found in the past (Jonckheere and Gordts, 1987). Our experiments at least regarding TEER—also suggest no detectable differences between GLV-infected or uninfected WB6 and GS trophozoites on Caco-2 monolayers (Supplementary Figure 11).

In summary, variations in experimental design due to the lack of standardization confound the search for robust effects on TEER induced by G. duodenalis. The current data in particular of the Caco-2 in vitro co-culture model rather suggests that Giardia alone does not induce acute barrier-defects.

### Alteration in Tight Junction Composition

Intestinal epithelial barrier function is highly dependent on cytoskeleton architecture and tight junction composition (Hidalgo et al., 1989; Troeger et al., 2007). Giardia-induced changes in the F-actin cytoskeleton were observed after incubation with sonicates (Scott et al., 2002), with living trophozoites, sonicates and spent medium (Teoh et al., 2000) or only with attached trophozoites (Humen et al., 2011). All these studies found also alterations to respective TEER values (Supplementary Table 1). Tight junction proteins, ZO-1 and claudin-1 showed in some studies altered localization, that could be prevented by various treatments, e.g., by EGF- (Buret et al., 2002) or caspase-3 inhibitor treatment (Chin et al., 2002), or MLCK-inhibitor addition (Scott et al., 2002; see summary in Supplementary Table 1). Ex vivo-analysis of epithelial tissue from symptomatic patients, showed 30% reduced claudin-1 abundance (Troeger et al., 2007), which was not evident in any in vitro study. On a technical note, all in vitro studies which showed alterations of the tight junction complex, experiments were conducted on plastic (chamber slides) or glass (coverslips) support which alter Caco-2 monolayer morphologies (Supplementary Figure 12).

In contrast, but in agreement with the TEER measurements, no significant differences between infected or uninfected monolayers with respect to F-actin and tight junction proteins were observed on electrophysiologically tight filter-supported Caco-2 layers (**Figures 3**, **4**). Occasionally, effects as described in other studies, were detectable but lacked any correlation with infection (**Figure 5**). Due to heterogeneity of Caco-2 monolayers leading to a mosaic of microvilli formation (Katelaris et al., 1995; Sambuy et al., 2005; Liévin-Le Moal, 2013), the reported G. duodenalis induced microvilli depletion (Chavez et al., 1986, 1995; Buret et al., 1990, 1991) that should manifest as apical cytoskeletal changes could have been missed since changes reported were not drastic.

In conclusion, integrity of the tight junction protein complex also suggests an asymptomatic interaction of Giardia alone with the Caco-2 model. Alterations described by others may be related to growth in non-transwell systems. Of note, described disruptive effects on ZO-1 have been described as secondary to apoptosis and not directly induced by the parasite (Chin et al., 2002; Buret et al., 2003; Bojarski et al., 2004; Zehendner et al., 2011).

#### Epithelial Barrier Dysfunction Due To Apoptosis

Apoptosis is a tightly regulated process in the gut epithelium. For giardiasis, reported rates vary greatly between studies from no changes to controls (Chavez et al., 1986, 1995; Katelaris et al., 1995; Maia-Brigagao et al., 2012; Tysnes and Robertson, 2015, and own findings), to minor increases from 1 to 1.5% in symptomatic patients' mucosa (Troeger et al., 2007), to isolate specific effects of sonicates (Chin et al., 2002), to significant increases up to 41% with a rather low MOI of 3 after just 16 h using HCT-8 cells (Panaro et al., 2007) or only significant after long term co-culture (Fisher et al., 2013). HCT-8 cells were also shown to undergo increased apoptosis in mixed isolate infections (Koh et al., 2013); a finding that we could not reproduce with Caco-2 cells using NF and S2 (data not shown). Thus, effects on apoptosis also seem to be highly dependent on experimental setup and lack the robustness to allow a clear conclusion.

### Tissue Permeability as a Surrogate of Altered Molecular Fluxes

Absorption and trans-epithelial transport of biomolecules are key functions of the gut epithelia (Kiela and Ghishan, 2016) and permeation assays can be used to probe these functions.

### Labeled Compounds as Marker for Paracellular Permeability

A number of in vitro studies assessed paracellular permeability using fluorescein isothiocyanate (FITC)-conjugated dextran. Buret et al. (2002) noted a more than 40-fold increase of the trans-epithelial FITC-dextran flux in co-cultures with Giardia trophozoites. This effect was associated with alterations on tight junction protein distributions (Buret et al., 2002) and could be abolished when monolayers were pretreated with epidermal growth factor (EGF). Since bovine EGF can act on human cells, and since serum used in co-culture experiments provides several growth factors with overlapping effects, differences between serum batches must be considered as potential confounders. Similar results in a comparable setting were also reported using sonicates of various Giardia isolates (WB, PB, NF, and S2; all assemblage A; Chin et al., 2002) and on SCBN monolayers (Scott et al., 2002). Interestingly, large molecule permeability could be abolished using Myosin light-chain kinase (MLCK)- Inhibitor ML-9 (Scott et al., 2002). The mechanism proposed implicates phosphorylation of Myosin light-chain (MLC) by MLCK activity that leads to alterations in the F-actin and ZO-1 composition and eventually to increased permeability. An ex vivo study using FITC-dextran in Giardia GS infected mice compared effects on FITC-dextran permeability on day 7 (colonization phase) and 35 (post-clearance phase) after infection and found a slight increase in permeability at both time points of ∼30% that was correlated with cleavage of occludin and also increased endocytosis of bacteria (Chen et al., 2013). Our findings using FITC-dextran correlate with our respective TEER measurements (Supplementary Figure 13), showing no increased permeability of Giardia-infected monolayers, but of apoptosis-induced controls. Thus, like before the in vitro findings are as contradictory—also with respect to this parameter—as the in vivo studies mentioned before.

### Combined Assessment of Para- and Transcellular Flux

It is known, at least for in vivo studies, that individual sugars alone are not a reliable indicator of intestinal permeability and therefore disaccharide/monosaccharide ratios should be used instead (Johnston et al., 2000). Lactulose and mannitol for example are not transported via active monosaccharide transport systems but due to their difference in size can serve as a marker for transcellular permeability through channels (mannitol) and paracellular permeability, or tight junction leakage (lactulose; Andre et al., 1988). However, L:M ratios are not used very often in vitro, but other compounds had been used instead to assess and distinguish paracellular and transcellular flux. Using horse radish peroxidase (HRP) and creatinine as transand paracellular markers, respectively, infection with both a lab adapted and a field isolate, led to increased mucosal to serosal transcellular flux. Also paracellular permeability was increased by the field isolate (R-2), which is contradictory to the observed TEER increase in the same study (Tysnes and Robertson, 2015). Another study by Hardin et al. (1997) on Giardia S2-infected Mongolian gerbils could not find increased permeability to [51Cr]EDTA (permeability similar to lactulose), but macromolecular transport of BSA was induced. A study on rats also suggest Giardia's interference with active transport, but the other way around: Glucose and glycine absorption was decreased, whereas potassium which diffuses passively through the epithelium was unaffected (Anand et al., 1980). However, one should keep in mind that malabsorption could be either a result of impaired uptake due to dissipated osmotic and ion gradients originated from leaky tight junctions and apoptotic cells, or due to a reduction in the absorptive area because of villus shortening and microvilli depletion. The latter, a reduction in absorptive area, has been shown to be one of the more consistently found pathological features of Giardia (Chavez et al., 1986, 1995; Buret et al., 1990, 1991; Troeger et al., 2007) and could explain some study discrepancies.

Results with Caco-2 cells in permeability assays have to be interpreted with caution as these cells are known to actively and passively transport contents and are extensively used in

pharmaceutical drug absorption assays (Hidalgo et al., 1989; Artursson et al., 2001; Sambuy et al., 2005; Sun et al., 2008). When cultivated on plastic surfaces and not on transwellfilters, these cells tend to form large liquid-filled vacuole-like structures (Supplementary Figure 12) probably because they are not able to route those volumes through to a basally located compartment. Also, osmotically active marker substances, like

with DAPI is shown in blue. No significant changes in any of the observed proteins due to Giardia-infection are detectable.

sugars, can confound results by interfering with osmotic gradients. Enhanced active transport by Giardia, however, is understudied, but, should become more of a focus also because antigens or whole microbes of the intestinal lumen can be transported and be of pathological relevance. Indeed, Chen et al. (2013) found evidence for enhanced endocytosis of bacteria, rendering such a scenario possible.

Nuclear staining with DAPI is shown in blue. No significant changes concerning occludin due to Giardia-infection are detectable. A MOI of 10 is sufficient to completely cover the Caco-2 monolayer. Of note, MOI 100 show artificially less trophozoites attached than MOI 10 condition, possibly due to rapid nutrient consumption followed

### Analysis of Chemokine/Cytokine Profiles in Giardia Epithelial Cell Co-cultures

by starvation and subsequent detachment of the trophozoites.

In order to validate the asymptomatic outcome of our Caco-2 setup further, several chemokines/cytokines potentially or claimed to be upregulated during giardiasis were investigated. Most of the cytokine analyses that have been published were performed using cell monolayers grown on cell culture plastic. A seminal study used micro-array analysis of Giardia-infected Caco-2 cells during the first 18 h and showed an increase in CCL20, CCL2, and CXCL1/2 mRNA abundance among others (Roxström-Lindquist et al., 2005). However, increased IL-8 and TNFα mRNAs were not detected. Another study indicated no changes to IL-8, CCL2, GM-CSF, and TNFα (Jung et al., 1995). Hence, it appears that responses to Giardia infections differ from the generalized response by enterocytes of colonic origin to bacterial infections (Jung et al., 1995). This was corroborated by Fisher et al. (2013) for CCL2, GRO-isoforms (CXCL1/2), or IL-8 in the Caco-2 co-culture system. Interestingly, Caco-2 co-cultured with macrophages alone elicited GRO-isoform and IL-8 expression but this was abolished again when Giardia trophozoites were added, indicating immune-modulatory capabilities of the parasite (Fisher et al., 2013). Giardia's ability to alter immune responses has been described by others, too (Kamda and Singer, 2009; Cotton et al., 2014, 2015). Our data using the Caco-2 transwell culture system also suggest no triggering of basolateral release of CCL2/20, CXCL1/2, IL-8, TNFα, or GM-CSF by different isolates or MOIs tested (**Figure 6**). Overall, our cytokine data also suggest an asymptomatic Giardia-infection. Whether this may be due to parasites' immune-modulatory features (Kamda

and Singer, 2009; Cotton et al., 2014, 2015) requires further studies.

## CONCLUSION

As mentioned in the introduction, one of the major enigmas in giardiasis is what distinguishes acutely symptomatic from asymptomatic outcomes of Giardia-infections. In this work, we reviewed and discussed current in vitro epithelial cell culture systems that we propose do reproduce asymptomatic hostparasite interaction rather than acutely symptomatic giardiasis. We propose for further development of these systems a structured approach that aims at identifying appropriate in vitro correlates for distinct clinical symptoms: The search for correlates of acute manifestations such as diarrhea could focus on readouts such as TEER, tight junction function, and ion secretion, but should focus on testing combinations of parasite and non-parasite factors. Food factors, or the lack of certain nutrients, could determine trophozoite virulence as well as bile or pancreatic secretions. Especially bile is a requirement for axenic trophozoite cultivation (Keister, 1983; Halliday et al., 1988) and it could stimulate the parasite not only to grow, but also to increase its virulence. It is also imaginable that certain host proteases could cleave Giardia proteins in a way that an inactive virulence function is activated, similar to what has been shown as a necessity for host cell

entry of influenza virus particles (Kido et al., 2012). Of note, Giardia specifically decreases the activity of pancreatic trypsin, but not of chymotrypsin (Seow et al., 1993), which suggests that certain proteases can be hazardous for Giardia and may disturb a finely balanced silent infection to provoke pathogenic reactions. Additionally, a host's genetic background linked to hypogammaglobulinemia, IgA deficiency or cystic fibrosis or a host's phenotype comprising reduced gastric acidity, stress, co-infections, or other concomitant diseases, could contribute to the pathogenesis (DuPont, 2013). Furthermore, since the mechanisms of trans-epithelial transport are as diverse as the respective substrates (Kiela and Ghishan, 2016) and our lack of knowledge if and how Giardia affects those transport systems, the search for correlates and explanations for more protracted effects such as stunted growth that may be linked to altered fluxes of particular biomolecules, should focus on a systematic assessment of epithelial cell functions related to those transcellular transport and absorption mechanisms. Another important factor relevant for the symptomatic outcome may be the host's microbiome, as suggested by others (Singer and Nash, 2000; Chen et al., 2013; Slapeta et al., 2015). The fact that young children after the lactation period are generally more affected by Giardia infestation than older children or adults may be an epidemiological correlate of a not-yet-settled microbiome that is more susceptible to interference by the parasite. Moreover, differences in food resources and cultural habits, resulting in distinct intestinal bacterial colonization may underlay the variances noted between studies that cover different geographic regions. Additionally, very recent publications reporting experimental findings also point toward an increasing role of the intestinal microbiota during Giardia infections (Allain et al., 2017; Barash et al, 2017; Beatty et al., 2017). Finally, since giardiasis resembles symptoms of food allergies, hypersensitivity or intolerance like coeliac disease, pathogenesis could be more related to the host's individual immune reaction (Scott et al., 2004). This might explain why Giardia-infections are less severe or even protective in developing countries, where allergic diseases are usually not as prevalent. However, due to the parasite's ability to enhance sensitization toward food antigens (Di Prisco et al., 1998), the causal relationship will be difficult to investigate. Moreover, the general in vitro methodology of Giardia-host interaction studies requires better standardization with the goal to offer robust, inter-laboratory evaluated models.

### AUTHOR CONTRIBUTIONS

All authors contributed equally to design and conception of this work. MK conducted research on literature and collected experimental data. CK, RB, JS, and TA contributed to experimental design and intellectual input. RB and JS helped interpreting data and MK, CK, and TA contributed to the manuscript.

### FUNDING

This work was financially supported by grant GRK 2046 from the German Research Foundation (DFG). The funding bodies had no role in the design of the study, collection, analysis, and interpretation of data or in writing the manuscript.

### ACKNOWLEDGMENTS

We thank Ms. Wibke Krüger and Ms. Petra Gosten-Heinrich for technical assistance, as well as Prof. Staffan Svärd and Prof.

### REFERENCES


Andre Buret for intellectual input. Furthermore, we thank Prof. Andre Buret (Department of Biological Sciences, University of Calgary, Canada) for sending us NF and S2 Giardia isolates and Dr. Marco Lalle (Istituto Superiore di Sanità, Rome, Italy) for sending us "Giardia lamblia virus"-infected isolates. Finally, we thank Mr. Jonnel Jaurigue for proof-reading the manuscript.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fcimb. 2017.00421/full#supplementary-material

Buret, A. G. (2007). Mechanisms of epithelial dysfunction in giardiasis. Gut 56, 316–317. doi: 10.1136/gut.2006.107771


infiltration in an in vivo model of bacterial toxin-induced colitis and attenuates inflammation in human intestinal tissue. PLoS ONE 9:e109087. doi: 10.1371/journal.pone.0109087


exposure to Salmonella enteritidis. Cell Stress Chaperones 8, 194–203. doi: 10.1379/1466-1268(2003)008<0194:ELOHSP>2.0.CO;2


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

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

# Translational Rodent Models for Research on Parasitic Protozoa—A Review of Confounders and Possibilities

#### Totta Ehret 1, 2, Francesca Torelli <sup>1</sup> , Christian Klotz <sup>1</sup> , Amy B. Pedersen<sup>3</sup> and Frank Seeber <sup>1</sup> \*

<sup>1</sup> FG16 – Mycotic and Parasitic Agents and Mycobacteria, Robert Koch Institute, Berlin, Germany, <sup>2</sup> Department of Molecular Parasitology, Humboldt-Universität zu Berlin, Berlin, Germany, <sup>3</sup> School of Biological Sciences, University of Edinburgh, Edinburgh, United Kingdom

Rodents, in particular Mus musculus, have a long and invaluable history as models for human diseases in biomedical research, although their translational value has been challenged in a number of cases. We provide some examples in which rodents have been suboptimal as models for human biology and discuss confounders which influence experiments and may explain some of the misleading results. Infections of rodents with protozoan parasites are no exception in requiring close consideration upon model choice. We focus on the significant differences between inbred, outbred and wild animals, and the importance of factors such as microbiota, which are gaining attention as crucial variables in infection experiments. Frequently, mouse or rat models are chosen for convenience, e.g., availability in the institution rather than on an unbiased evaluation of whether they provide the answer to a given question. Apart from a general discussion on translational success or failure, we provide examples where infections with single-celled parasites in a chosen lab rodent gave contradictory or misleading results, and when possible discuss the reason for this. We present emerging alternatives to traditional rodent models, such as humanized mice and organoid primary cell cultures. So-called recombinant inbred strains such as the Collaborative Cross collection are also a potential solution for certain challenges. In addition, we emphasize the advantages of using wild rodents for certain immunological, ecological, and/or behavioral questions. The experimental challenges (e.g., availability of species-specific reagents) that come with the use of such non-model systems are also discussed. Our intention is to foster critical judgment of both traditional and newly available translational rodent models for research on parasitic protozoa that can complement the existing mouse and rat models.

Keywords: wild rodent, protozoa, parasite, model organism, mouse, rat, translational research

## INTRODUCTION

Gregor Mendel introduced the basic concept of a "model organism" when he reported his experiments on plant hybrids. He picked peas as a model because they had clear experimental advantages for addressing his question: "At the very outset special attention was devoted to the Leguminosae on account of their peculiar floral structure (...) this led to the result that the genus

#### *Edited by:*

Kenneth Fields, University of Kentucky, United States

#### *Reviewed by:*

Aaron Conrad Ericsson, University of Missouri, United States Luciana Oliveira Andrade, Universidade Federal de Minas Gerais, Brazil

*\*Correspondence:*

Frank Seeber seeberf@rki.de

*Received:* 23 February 2017 *Accepted:* 22 May 2017 *Published:* 07 June 2017

#### *Citation:*

Ehret T, Torelli F, Klotz C, Pedersen AB and Seeber F (2017) Translational Rodent Models for Research on Parasitic Protozoa—A Review of Confounders and Possibilities. Front. Cell. Infect. Microbiol. 7:238. doi: 10.3389/fcimb.2017.00238 Pisum was found to possess the necessary qualifications." (Mendel, 1866) Since then this approach for model selection for a particular purpose is widely used, meaning that a model organism should be accessible, experimentally tractable, have short generation times, be affordable to maintain and breed, possess clearly identifiable features that are to be studied, and more recently be genetically tractable (which includes access to a sequenced genome), to name a few (Rand, 2008). Given the importance of model systems in biology, the history and diversity of model organisms has been extensively reviewed (Conn, 2008; Hunter, 2008; Müller and Grossniklaus, 2010; Ankeny and Leonelli, 2011; Bolker, 2012; Alfred and Baldwin, 2015).

The available model organisms span a great taxonomic range, and for many questions single-celled organisms such as bacteria (Escherichia coli) or yeasts (Saccharomyces cerevisiae) are sufficient. However, in many biomedical studies which aim to translate findings to humans, non-mammals are not applicable as translational models (Hau, 2008). Yet, shorter phylogenetic distances and anatomical similarities are no guarantee for translational success, as research on primates has demonstrated (detailed in section The CD28 Superagonist Antibody "Disaster"). Given the high relevance of protozoa for human and animal health and our own scientific interests in these parasites, we will concentrate on such infections in rodents. While their value as translational models is not without dispute (see section Protozoan Parasites) they have been, still are, and will continue to be, invaluable for both basic biological questions in host-parasite interactions as well as pre-clinical studies. Their importance for scientific progress is demonstrated, for example by the essential role laboratory mice played in the discovery of dendritic cells (see references in Steinman, 2012) and macrophages (reviewed by Gordon, 2007). The value of rodents, and mice in particular, has also been highlighted for leading to fundamental insights in infection biology (Buer and Balling, 2003; Vidal et al., 2008; Douam et al., 2015). Given the historical and current importance of rodents, we here explore benefits and limitations of, e.g., inbred or outbred mice, lab rodents, or rodents from the wild, etc., as well as possible external confounders, such as breeding conditions in lab facilities or interactions in an ecosystem, that might have great impact on research results. With this review, we encourage experimentalists, particularly in translational medicine, to consider a broad set of potential rodent models in order to identify and use the best available system for specific studies. We aim to provide a foundation and useful references for such decisions.

### PROTOZOAN PARASITES

Infections with protozoan parasites cause substantial illness and economic loss in humans worldwide (see **Tables 1, 2** for details; Fletcher et al., 2012; Murray et al., 2012; Andrews et al., 2014; Robertson et al., 2014; Kassebaum et al., 2016). These parasites with high impact on humans mostly are the Amoeba, e.g., Entamoeba spp.; the flagellates, e.g., Trichomonas spp., Giardia spp., Leishmania spp., and Trypanosoma spp., and the large group of Apicomplexans, which contain, e.g., Plasmodium spp., Toxoplasma gondii, and Cryptosporidium spp. They are most often transmitted to their host either via ingestion of contaminated food, water, or via a vector (e.g., mosquitoes or flies; see **Table 2**). Here we provide a brief overview of some parasitic protozoa which substantially impact humans, and of which many are referred to in our examples used in following sections.

The significance of protozoan infections for global human health is here exemplified by data (see **Table 1**) where the impact of infections by the most devastating protozoan parasites is expressed as "disability-adjusted life-years" (DALYs, used by the World Health Organization (WHO) and others as a measure of disease impact). These diseases ranked second in importance across all infectious diseases, behind lower respiratory infections, and before AIDS and tuberculosis. The great majority of this impact can be attributed to malaria alone (85% caused by Plasmodium spp.). However, the collective disease burden of the other protozoa evaluated was also substantial and in the range of influenza (Murray et al., 2012). While many of these figures, including those for malaria, are fortunately on the decline, this disease was still ranked among the top 20 leading diseases as identified by the WHO worldwide in 2015 (Kassebaum et al., 2016). In addition to these human health concerns, protozoan parasites cause significant losses in many species of domestic animals (Perry and Grace, 2009; Torgerson and Macpherson, 2011; Fitzpatrick, 2013; Torgerson, 2013) and are in some cases a conservation concern for wildlife (Pedersen et al., 2007).

While many of these parasites have a very restricted host range (infecting a single host species and/or tissue), other extremes such as T. gondii, a zoonotic parasite assumed to infect all nucleated cells in all warm-blooded animals, exist. Consequently, there is a mixture of more or less natural relationships between the parasites and the rodent hosts when the latter are used as translational models for human infections (see **Table 2**). Many parasite genera contain species which naturally infect rodents (e.g., Plasmodium, Giardia, Cryptosporidium), although in most cases the very same species do not also infect humans. For other parasite species, the rodent model has been made susceptible, frequently by genetic means, to human relevant parasites (e.g., P. falciparum or C. parvum/C. hominis). In malaria research, rodents have been successfully used as models, but the suitability of the mouse to mimic severe human malaria has been questioned (Langhorne et al., 2011). In leishmaniasis research, rodents are acknowledged for contributing to a better understanding of the immune response to the parasite (Lipoldová and Demant, 2006) but other authors point out limitations and the lack of suitability of certain mouse strains to study specific parasite genotypes (Mears et al., 2015). Research on human sleeping sickness (T. brucei) has benefitted largely from mouse models (Antoine-Moussiaux et al., 2008; Giroud et al., 2009; Magez and Caljon, 2011) but criticism has been raised that more suitable animal models should be applied to address sleeping sickness in livestock (T. congolense and T. vivax; Morrison et al., 2016).



Data shown for 2010 and 2015. For causative agents of protozoan diseases see *Table 2*. <sup>a</sup>Murray et al. (2012).

<sup>b</sup>Kassebaum et al. (2016).

<sup>c</sup>Number in brackets, without influenza.

<sup>d</sup>nr, not reported.

<sup>e</sup>Number in brackets, without cryptosporidiosis and amoebiasis.

<sup>f</sup> Number in brackets, without malaria.

### WHY ARE MICE AND RATS SUCH POPULAR MODELS?

Biomedical research depends heavily on model organisms and the majority of these are rodents, particularly in infectious disease and immunological research. A few numbers illustrate this impressively. For example, in the European Union alone, 75% of all animals used for "experimental and other scientific purposes" in 2011 were house mice (Mus spp. 61%) and rats (Rattus spp. 14%; The Commission to the Council and the European Parliament, 2013). Other rodents (gerbils, hamsters, different species of mice, and other rodents) only constitute 0.47% of animals used (**Figure 1A**). In Germany, 91.5% of animals used in research on infectious diseases were rodents, with the vast majority being Mus musculus (88.6%; **Figure 1B**). Similar numbers are reported in the United Kingdom, with 82% of all research using rodents, again dominated by M. musculus (74.6%; UK Home Office, 2015). While these numbers also include animals that were used as donors, e.g., for blood or organs and thus for in vitro experimentation these data nevertheless illustrate the dominance and importance of rodents, in particular laboratory inbred mice, as model organisms.

Another informative figure shows that the number of publications where mice and rats were mentioned in the title dominates all other model organisms included (**Figure 2**), e.g., Arabidopsis, Drosophila, S. cerevisiae, Caenorhabditis elegans, Xenopus, zebrafish, Neurospora, and Dictyostelium discoideum. It is presumably no coincidence that a sharp increase in these "mouse publications" was seen in the 1990s, given that it was when embryonic stem cell manipulation met homologous recombination of the mouse genome. This resulted in the generation of defined gene knock-out mice (**Figure 2**), a finding which was later rewarded with the Nobel Prize in Physiology (Mak, 2007). The importance of this discovery for scientific progress in infection biology cannot be overestimated. However, in rats no such methods were available until relatively recently (Tong et al., 2010; van Boxtel and Cuppen, 2010), which is reflected in the drastic increase of mouse models and a relatively stable use of rats from 1990s until now. It is likely due to the highly developed genetic tools in mice, together with the more than 450 inbred mouse strains established since the first strain (DBA/2) was developed by Clarence Cook Little, that mice, and in particular the C57BL/6 strain, are the most popular animal model (Beck et al., 2000; Festing and Fisher, 2000). However, with the advent of CRISPR/Cas9 gene modifications in rats this will likely change (Hu et al., 2013; Li D. et al., 2013; Li W. et al., 2013), since this method has worked so far in almost all organisms tried and it can most likely also be applied to wild rodents. The historic establishment of tools for mice combined with the fact that 99% of genes are conserved between the human and the mouse genomes (Waterston et al., 2002) has made and will continue to make the mouse an obvious choice for translational efforts, i.e., research to understand the basics of and find treatments for human diseases. In addition, it will be exciting to see contributions from so far poorly explored model systems.

### The Mice We Use in Experiments — Who Are They and How Do They Live?

Here we will briefly cover definitions of nomenclature for referring to different types of mice, from laboratory to "wild" rodents.

**Classical inbred mice** are defined as either being "produced using at least 20 consecutive generations of sister x brother or parent x offspring matings" or "traceable to a single ancestral pair in the 20th or subsequent generation." ("Nomenclature of Inbred Mice," defined by the Mouse Genomic Nomenclature Committee). However, it can be noted that 20 generations of inbreeding does not lead to fixed alleles in the entire genome, although for most phenotypes no differences are detected after this threshold (Chia et al., 2005). Different inbred populations exist and are referred to as strains (whereas outbred populations are often referred to as stocks). Inbred strains are genetically highly homogenous, well-defined, and often with genomes and SNP data available. In addition, extensive descriptions of (mutant) strains are available in the Mouse Phenome Database (Grubb et al., 2014) or the International Mouse Phenotyping Consortium database (Koscielny et al., 2014; see Table S1 for links) and should be consulted when planning experiments.

**Wild-derived inbred** strains are "descendants of mice captured in wild populations during the mid to late 20th century and represent several different Mus species from around the world" (Lutz et al., 2012). These mice are considered suitable for, e.g., evolutionary studies and gene mapping, but notably do not represent the genetic diversity of wild animals.

**Outbred** stocks are defined as "a closed population (for at least four generations) of genetically variable animals that is bred to maintain maximum heterozygosity" (Chia et al., 2005), meaning


TABLE 2 | The diseases caused, transmission routes, and suitability of rodents as models for human disease are listed for selected protozoan parasites.

Reference is given to a single article describing the basic biology of the respective protozoan to serve as starting point for further reading. (yes), adopted to model.

that each individual is genetically different from the others. Once established, the goal is to keep the genetic variability between generations to a minimum which is achieved by using, e.g., a certain number of breeding pairs (Chia et al., 2005). We onwards refer to inbred and outbred M. musculus models as lab mice, if nothing else is specified.

**Recombinant Inbred Strains**, RIS, are a collection of mice established by inbreeding two existing inbred strains into a set of strains (often called set or panel). Each such strain is genetically homogenous, but "parallel" strains produced from the same two well-defined ancestral strains are genetically more different from each other than either of the two ancestors (Chia et al., 2005). One advantage of using a RIS set compared to pure inbred strains is that phenotypic differences (e.g., pathogen or drug susceptibility) can be fairly easily assigned to specific genotypes (Guénet et al., 2015), and obtaining high-quality quantitative data on transcripts and proteins is feasible (Chick et al., 2016). Other options for lab mice, such as genetic crosses, will be covered in the section Humanized Mice: Rodents Which Mimic the Human Immune System.

We refer to **wild rodents** (including wild mice, e.g., species of Mus) as rodents which breed without direct intervention or manipulation by humans, in their natural habitats, e.g., farmland, forests or cities (Singleton and Krebs, 2007). Such populations may in some cases be under experimental study and manipulated, for example by regular trapping, diet manipulations and drug treatment, and will here still be considered as wild populations.

### Mouse Housing Influences Experimental Outcome

Almost all animal research facilities can house rodents in specific pathogen-free, SPF, barrier facilities. This standard includes regular screening for a large set of common pathogens (in order to detect contamination), and commonly autoclaving cages, bedding, water, food and other housing related materials to assure hygienic and controlled housing, as well as controlled light/dark cycles (Hedrich and Nicklas, 2012). For details see respective lab manuals (Ayadi et al., 2011; Hedrich, 2012). Animal psychological status has been shown to influence variability in experimental studies, including examples of more reproducible results from "happier" mice, which display less anxiety or depression-associated behavior as a result of increased animal welfare (Bayne and Würbel, 2012). Although, wild mice can run several 100 m per night (Latham and Mason, 2004) including a means for physical activity (e.g., running wheels) is not standard in animal housing. Moreover, it is debated whether such so-called enrichment of housing is always required, beneficial, or adequate for the outcome of an experiment (Bayne and Würbel, 2012), given that after decades of breeding and selection lab mice in many respects show different behaviors to wild mice (Latham and Mason, 2004).

Recent publications highlight the important role of microbiota in rodents (and humans). Even though SPF animals are the most commonly used rodents in experiments (Fiebiger et al., 2016), gut microbiota are not homogenous (e.g., in composition or bacterial numbers) in such research settings and the extent of this variation has only recently emerged. Microbiota differ between vendors and mouse strains (Hufeldt et al., 2010; Ericsson et al., 2015; Hilbert et al., 2017), different shipments from the same vendor (Hoy et al., 2015), between research animal facilities (Rausch et al., 2016) and even between rooms in the same breeding facility (Rogers et al., 2014). Determining factors for gut microbiota differences under SPF housing conditions without experimental perturbations have been analyzed, and apart from vendor, the fodder and treatment thereof is important (Rausch et al., 2016). Therefore, housing conditions strongly influence mouse microbiota. Data also suggest a general difference between inbred lab mice and wild mice in that the proportion of Firmicutes vs. Bacteroides vary, with wild rodents being dominated by Firmicutes and vice versa (Weldon et al., 2015).

and Agriculture, 2011).

model organisms like T. gondii, S. cerevisiae, and E. coli are given for comparison. Green dashed line (with corresponding y-axis on the right) illustrates articles mentioning knock-out mice, with first papers appearing in the early 1990s.

### MISLEADING RESULTS DUE TO INAPPROPRIATE ANALYSIS OR AN INAPPROPRIATE MODEL

"If you have cancer and you are a mouse, we can take good care of you" (Kolata, 1998). This famous sentence from Judah Folkman (the "father" of tumor angiogenesis) makes this point: any model - animal or even mathematical - only returns the output it is capable of producing. A translational mouse model that lacks human feature X will never give a response in X, no matter how important that particular feature is in the context of a human disease. While the mouse model has been very successful for understanding the general principles of the mammalian immune system and infectious disease (e.g., Buer and Balling, 2003), it is important to be aware of, and acknowledge, the intrinsic benefits and limitations in any model chosen for a specific experiment. However, before we focus on biological confounders we want to consider that failures in the transition from preclinical studies to humans may also be due to poorly designed or performed studies (see Couzin-Frankel, 2013; Justice and Dhillon, 2016). To illustrate that problems of very different character can challenge the suitability of translational rodent models, we will first discuss two past examples from different disciplines (sepsis and immunology) that caused vibrant discussions in the scientific community and were subsequently analyzed in great detail. They can thus provide valuable insights of general importance for scientists with different research interests. In the section Non-genetic Confounders in Rodent Infections with Protozoan Parasites we will then turn to confounders in translational rodent models of infections with protozoan parasites.

### Analysis, Re-Analysis and Meta-Analysis — Three Studies and Three Conclusions

Due to its importance for human health, research on sepsis in mouse models is heavily funded, but its translational success has so far been disappointing (van der Worp et al., 2010). Few papers in biomedical research have therefore raised such an excitement and storm of replies and counter-replies as the Seok et al. (2013) study on sepsis. They analyzed transcriptomic data from various mouse models of human inflammatory diseases, and human samples in particular from septic shock, and concluded that "genomic responses in mouse models poorly mimic human inflammatory diseases" (Seok et al., 2013). However, re-analysis of the very same data subsequently concluded the opposite, and these authors simply replaced "poorly" with "greatly" in the article's title of their reply (Takao and Miyakawa, 2015). This discussion is still ongoing, with a recent paper (Weidner et al., 2016) pointing out that the data from Seok et al. per se are good enough to compare the transcriptional responses of certain (but not all) mouse models to humans, but that the analytical tools used in the two first papers were inappropriate. The authors' conclusions were that gene set enrichment analysis (GSEA) is more appropriate than gene-to-gene comparisons, which require setting an arbitrary threshold for the determination of differentially expressed genes (as opposed to identification by statistical means). Those tools were used differently by various authors for re-analyses of the original data sets, thus leading to opposite conclusions (Seok et al., 2013; Shay et al., 2013; Takao and Miyakawa, 2015; Warren et al., 2015). A further level of complexity that makes conclusions derived from transcriptomic comparisons challenging is that for many genes there is no correlation between mRNA levels and protein quantities (see section Transcriptomes Do Not Necessarily Predict Protein Levels). Although transcriptome analysis is a fairly easily accessible and promising technique, these examples illustrate that such relatively young tools require close evaluation of the entire work-flow. Understanding and considering the physics or chemistry behind the method and to critically assess appropriate analysis methods is a community task when new scientific methods are being established.

### The CD28 Superagonist Antibody "Disaster"

It is well known that substantial differences between the mouse and human immune systems exist (Mestas and Hughes, 2004; Zschaler et al., 2014; Sellers, 2017) and that they need to be considered when using mice as preclinical models of human disease (e.g., Beura et al., 2016). In 2006, a small human phase I clinical trial aimed at alleviating rheumatoid arthritis tested a humanized monoclonal antibody, TGN1412, directed against the human T cell receptor CD28. However, instead of improving the autoimmune condition, it resulted in devastating consequences (reviewed by Hunig, 2016). It was anticipated from laboratory mouse studies that injection of the antibody would result in the preferential production of regulatory T cells, followed by a downregulation of active T cells. However, all six volunteers had to be hospitalized and at least four of them suffered multiple organ dysfunctions. TGN1412 had caused an immediate "cytokine storm" in these patients due to substantial TNF-α release, followed by dramatically increased plasma concentrations of several cytokines. This "cytokine release syndrome" (CRS) was caused by a strong activation of CD4<sup>+</sup> effector memory T cells, which eventually caused severe tissue damage. But why had preclinical studies in primate models, namely cynomolgus and rhesus monkeys, not indicated any signs of problems? What was unknown in 2006 was that those primates' CD4<sup>+</sup> effector memory T-cells do not express CD28 whereas humans do (Eastwood et al., 2010). In this particular case, the monkeys were a poor model for humans, despite their phylogenetically close relationship.

And why had the human response not been seen in the numerous rodent experiments performed prior to the trial? Interestingly, later experiments have demonstrated that at least two drastically modified lab mouse models can indeed be good models for the TGN1412 experiments. The first example is linked to the fact that, as is the case for most immunological experiments, mice in the initial studies had been raised and kept under SPF conditions. Thereby, they had no exposure to microbial antigens that would elicit CD4<sup>+</sup> memory T cells. Thus, CRS was not initiated upon TGN1412 treatment as it was in humans. Consequently, when TGN1412 was later given to non-laboratory "dirty" mammals (including rodents, see section Getting the Rodent Model "Dirty") exposed to prior environmental microbial stimuli, they experienced similar syndromes as the human volunteers (Eastwood et al., 2010). The second alternative model consists of humanized mice (see section Humanized Mice: Rodents Which Mimic the Human Immune System). A recent study made use of mice which had been reconstituted with human peripheral blood mononuclear cells. Injecting TGN1412 into those animals recapitulated a number of the disastrous immunological outcomes also seen in the initial human trial (Weissmuller et al., 2016). Importantly, the transplanted human cells also included a small amount of effector memory cells. Therefore, both the use of "dirty" and humanized mice better mimicked human biology than the rodent models which were used in the pre-clinical studies (although both these models have gained interest more recently and were, if at all, very new ideas at that time).

### NON-GENETIC CONFOUNDERS IN RODENT INFECTIONS WITH PROTOZOAN PARASITES

Lab rodent models have been essential for understanding molecular, cellular, and immunological responses; however, most of the variability inherent in natural populations is not captured by them (Pedersen and Babayan, 2011; Beura et al., 2016). Even so, sources of variation which influence experimental outcomes in lab experiments have also been identified in these very models. Reports based on lab experiments have often not accounted for such variability and instead ascribed differences between experimental groups to the aspect under study.

### Microbiota as a General Confounder for Rodent Experiments

Recently, the role of microbiota as a confounder for experimental outcomes in various scientific fields has gained increasing interest (Servick, 2016), largely due to the development of next-generation sequencing and related methods. Common approaches to study microbiota include sequencing fecal content from lab mice and germ-free mice (discussed in detail in Fiebiger et al., 2016), fecal transplants, antibiotics treatment, probiotics, addition of a specific bacterium, and infectious agents. For biomedical research it is noteworthy that microbiota influences host susceptibility to drugs. One example are proteins encoded by drug processing genes, DPGs, which are responsible for uptake, distribution, metabolism, and excretion of xenobiotics such as drugs (Aplenc and Lange, 2004; Klaassen et al., 2011). DPGs in mouse liver display different expression patterns depending on the microbiota status of the animal (Fiebiger et al., 2016). Some authors have linked microbiota differences to subsequent variation in brain activity and changes in social behavior, a concept referred to as the gut-brain axis (e.g., Foster and McVey Neufeld, 2013; Mayer et al., 2015; Gacias et al., 2016) which is proposed to depend on several factors, including the immune system. Variation in microbiota is known to influence both local and systemic immune function by altering the balance of Th1/Th2 cell composition, influencing re-localization of neutrophils, or affecting macrophage polarization (Denny et al., 2016; Lopes et al., 2016). Taken together, it is therefore not surprising that differences in microbiota can have a substantial impact on protozoan parasite infections in the gut and elsewhere. It is also easy to imagine a situation in which genetically modified mice obtained from one breeder or lab and control mice from other sources leads to unintended differences in microbiota composition with resulting influence on the outcome of infection experiments.

### Protozoan Infection Experiments Are Influenced by Microbiota

One early study pointing to the importance of microbiota for the establishment of a protozoan infection used germfree mice which were infected with the intestinal parasite G. duodenalis (Torres et al., 1992). The authors demonstrated that the microbiota influences the establishment and nature of intestinal infection with regards to severity and parasite reproductive success. A later study showed that female mice with the same genetic background were either susceptible or resistant to G. duodenalis infection (Singer and Nash, 2000). Differences were due to the origin (vendor) of the animals, and the same was true for immunodeficient mice. Co-housing led to resistance in all animals, whereas treatment with antibiotics made all animals equally susceptible to intestinal infection. It was therefore concluded that the microbiota determined the outcome of infection. These studies have recently been complemented with more in-depth investigations of the microbial community, showing changes in the amount of microbiota and its composition upon G. duodenalis infection in mice (Barash et al., 2016). Hence, not only do microbiota influence infection outcome but the parasite in turn alters the gut microbiota. These studies emphasize the complexity of gastrointestinal parasite infections. Further analysis of microbiota-parasite-host cohabitation will likely reveal interactions such as competition for nutrients or synergies in metabolism.

The complexity of microbiome influences is not limited to gut microbiota. Skin microbiota has also been shown to influence the outcome of cutaneous leishmaniasis in mice. Its causative agent, L. major, differently induced skin lesions, edema, and necrosis in germ-free mice compared to SPF mice upon intradermal infection (Naik et al., 2012). Germ-free mice displayed less disease severity, but also reduced levels of IFN-γ and IL-17A from Aβ T cells in the infected skin area compared to SPF reared animals. By orally administering antibiotics, the gut microbiota, but not skin microbiota, changed without influencing cytokine production. However, introduction of a skin commensal bacterium, Staphylococcus epidermis, did rescue IL-17A production in the skin. The authors concluded that local cytokine production was specifically linked to skin microbiota. In a different study, L. major infection altered the gut microbiota of infected animals (but differently depending on mouse strain) (Lamour et al., 2015). Infection changed how gut microbiota correlated with systemic functions such as urine metabolites, plasma metabolites, and the immune system. Such findings also highlight that simple correlations between microbiota and protozoan parasites may not be adequate to elucidate the dynamic role of microbiota during infection.

Interestingly, the microbiota does not only affect the site of infection but can also influence how host and parasite interact at other sites. Recent work provided evidence that the severity of malaria infection with rodent Plasmodium spp. can also depend on vendor. Differences in disease severity correlate with differences in microbiota composition (Villarino et al., 2016) or bacterial transcription profiles (Stough et al., 2016), demonstrating systemic effects by the microbiota. A study from 2014 also reported a mechanism for such correlations, describing production of anti-Plasmodium spp. antibodies in response to gut colonization, specifically by E. coli O86:B7 but not by the reference E. coli K12 (Yilmaz et al., 2014). These studies demonstrate that infections in specific compartments which are not colonized by commensal bacteria are nevertheless influenced and such effects must be considered in planning experiments and interpreting results.

### Excursion 1: A Reductionist In vitro Approach Using Organoids

Protozoa-host interaction studies have largely been restricted to more or less suitable rodent models, cell lines (often cancerderived), and short-lived primary tissue cultures from biopsy or surgery. Recent advances in stem cell research have paved the way for the development of self-renewing and complex tissue-like culture systems, so-called organoids, which mimic organs in their main functions and structural features (Willyard, 2015). Major advantages include that host-parasite interactions can be investigated in a primary, long-lived, organ-like tissue from the organism of choice, including humans, in real time (Klotz et al., 2012). Organoids have been developed from the gastro-intestinal tract (GIT), including stomach, gut and liver, and also from kidney and brain (Clevers, 2016). Importantly, organoids lack tissue-specific immune cells and in the GIT the microbiota, and therefore complexity is low compared to in vivo settings. However, this feature allows the researcher to construct an experimental setup with exactly the desired level of complexity, adding for instance the Mouse Intestinal Bacterial Collection (Lagkouvardos et al., 2016), human microbiota from biopsies, and/or a set of cytokines or immune cells of interest. For elucidating the role of individual actors during a parasitic infection, organoids are promising alternatives to animal models, cell culture systems, and the use of human biopsy material.

### Sex and Age

Two long-known factors that influence infection success by parasites are sex and age. The most obvious differences between the sexes are hormones (Roberts et al., 2001; Klein, 2004; Bernin and Lotter, 2014) but X chromosome-linked mutations (van Lunzen and Altfeld, 2014; Garenne, 2015) and sexspecific behavior can also affect the outcome of infectious diseases. A prominent example in protozoan infections is glucose-6-phosphate dehydrogenase (G6PD) deficiency, which protects humans of both sexes to different extents from clinical outcomes of infections with P. falciparum (Shah et al., 2016). A previously developed humanized mouse model of G6PD deficiency (Rochford et al., 2013) has recently been used in screening efforts to identify malaria transmissionblocking drugs (Wickham et al., 2016). A second example involves X-linked immunodeficiency in the B-cell responses due to mutations in the Bruton's tyrosine kinase. The mutation causes a sex-specific effect which leads to X-linked agammaglobulinaemia (XLA). Human patients and mice bearing a similar mutation (CBA/N) are more prone to develop chronic giardiasis (Skea and Underdown, 1991; Van der Hilst et al., 2002).

In rodent models of, e.g., Plasmodium spp., Cryptosporidium spp., and Leishmania spp. infection age significantly influences susceptibility (Adam et al., 2003), parasite reproductive success (Rhee et al., 1999), and severity of disease (Muller et al., 2008). In the case of Cryptosporidium (Rhee et al., 1999), hamsters displayed age-dependent differences (within the first 2 months of life) in infection persistence measured by time for shedding oocysts, whereas mice did not. The results demonstrate that rodents can be used to study cryptosporidosis, but simultaneously suggest that generalizations of these results to other species are difficult, and translational success is not obvious. For leishmaniasis, recent work showed mouse agespecific differences in the induction of adaptive immunity. Animals were exposed to a vaccine candidate based on genetically modified L. donovani and aged mice (∼16 months) had a less pronounced adaptive immune response compared to young mice (∼2 months) upon L. major challenge after vaccination (Bhattacharya et al., 2016). In Babesia microti infection of lab mice between the ages of 2 and 18 months (Vannier et al., 2004), one of three strains (DBA/2 mice) mimicked patterns seen in humans in which susceptibility and an inability to clear infection increased with age. The other two strains displayed smaller differences in susceptibility and no change in infection clearance, illustrating possibilities to use rodents as models for human babesiosis, but alerting to possible issues with interpretation and translation of results.

Given these examples it seems obvious to consider age and sex aspects when planning rodent experiments. However, it is not unusual to use only male or only female mice (Flórez-Vargas et al., 2016) based on convenience, local availability, costs, legal issues (more animals required when both sexes are examined; Clayton and Collins, 2014) or research area. In particular in infectious disease research there is a strong bias toward using female mice (Flórez-Vargas et al., 2016). One reason for this is presumably that they are less aggressive and thus cheaper since they can be housed in (experimental) groups in a single cage whereas this is challenging for male mice. Likewise, younger mice are cheaper to obtain since housing cost are lower. Thus, convenience rather than scientific reasoning might influence the choice of sex or age in many studies.

### MODELS FOR ALL PURPOSES — FROM FIXED ALLELES TO COMPLEX ECOLOGY

Even before the first draft of the mouse genome was published in 2002 (Waterston et al., 2002), scientists were aware of the relative genetic homogeneity of the lab mouse compared to wild mouse populations (Guenet and Bonhomme, 2003). Inbreeding over almost a century fixed alleles in currently available lab mice, which now represent just a fraction of the genetic variability found in nature. Although a desirable feature for some questions, this variability can be of great importance in studies of host-parasite interactions. Genetic variability might be the reason for different host susceptibility together with, e.g., the confounders discussed above (see section Non-genetic Confounders in Rodent Infections with Protozoan Parasites). In many cases allelic variations of a gene involved in immune responses were identified as the cause of infection outcome.

### When Immune Responses Depend on Genetics — Selected Examples

Allelic variation at Lsh and H2 loci is involved in the opposite outcome of the acquired immune response in L. donovani infection between e.g., CBA and BALB/c mice (Loeuillet et al., 2016). Another emerging example is the role that the inflammasome has in sensing protozoan infections (reviewed in Zamboni and Lima-Junior, 2015). During T. gondii infection in rodents, sequence differences in the pathogen sensor Nlrp1 accounts for species-specific inflammasome induction - and thus outcome - in lab mice (Ewald et al., 2014) and rat strains (Cirelli et al., 2014). Phenotypic differences between model animals can also be due to polymorphisms in the inflammasome pathway effectors, e.g., IL-18 and IL-1β. A study on a wild, natural population of field voles (Microtus agrestis) found associations with polymorphisms of IL-1β, IL-2, and IL-12β and differential susceptibility to pathogen infection (Turner et al., 2011), with the impact on susceptibility being comparable to parameters like sex and body weight. The parasites considered in the study were mostly nematodes, cestodes, and B. microti, making the influence of these cytokines' polymorphisms in other protozoan parasite infections a likely scenario.

Recently, polymorphisms in immunity-related GTPases (IRGs) affecting the rodent immune responses to T. gondii were described in a study comparing the DNA sequences of several inbred and wild-derived mice (Lilue et al., 2013). Remarkably, the authors showed that while the sequences of all the examined lab mice were highly conserved, genes in the wild-derived mice were extremely diverse, comparable to the diversity of MHC genes. One of those genes, the highly polymorphic Irgb2-b1 from a wild-derived M. musculus, when expressed in C57BL/6 fibroblasts, was sufficient to confer resistance (i.e., prevent cell lysis) to so-called virulent strains of T. gondii. While 1–10 parasites of these strains can kill a lab mouse, IRG-polymorphic wild-derived mice are resistant to infection by much higher numbers of the same T. gondii strain. Apart from highlighting extensive sequence variability of wild-derived but not classical inbred laboratory strains in these gene loci, this work emphasizes that the definition of virulence is heavily dependent on the animal of choice and that its definition should always be accompanied by stating the experimental conditions. In the following section we move from traditional, genetically homogenous and inbred mice to mice manipulated to resemble aspects of the human immune system, and then to genetic crosses between inbred and wild-derived mice. Lastly, we will turn our attention to "dirty mice" and wild rodents in natural settings.

### Humanized Mice: Rodents Which Mimic the Human Immune System

Designing an immune system with human features within the mouse—generating humanized mice—has recently emerged as an approach to expand the areas where lab mice can be used to model disease. Generation of humanized mice is based on immunodeficient animals (e.g., SCID, Rag2−/−) whose innate and adaptive immune systems are severely compromised and the animals are instead characterized by increased survival of transplanted human hematopoietic cells (Kaushansky et al., 2014; Good et al., 2015), which produce a large number of different human immune cells in the mouse. Depending on further needs, these mice can be populated with, for example human red blood cells and/or CD34<sup>+</sup> hematopoietic stem cells that further give rise to T cells and antigen presenting cells (APC). Recent advances in the development of humanized mice offer the possibility to study human infectious diseases which could previously not be investigated in mice, in the mouse model (Brehm et al., 2013). Even though the use of rodent-infecting Plasmodium spp. such as P. berghei has greatly contributed to understanding the parasite's biology and general principles of protective immune mechanisms in mammals (Craig et al., 2012), it is promising that human-infecting P. falciparum now can be researched in lab mice (Kaushansky et al., 2014). The basic strategy for generating humanized mice, and adaptations of it lead to the generation of mice in which the blood stages of human malaria parasite life cycles could be established (Kaushansky et al., 2014; Good et al., 2015).

More advanced models with engrafted human hepatocytes (FRG-NOD huHep) have been further used to establish the complete development of the pre-erythrocytic liver stage of P. falciparum after mosquito bite, including formation of exoerythrocytic merozoites, subsequently infectious to human red blood cells in the same mouse (Vaughan et al., 2012). Although, the development of the mature sexual stages (gametocytes) that are necessary to complete the parasite life cycle is still inefficient, it seems possible that complete P. falciparum (and other human Plasmodium spp.) life cycles could be routinely maintained using humanized mice. Recently, such mice were also used to conduct a genetic cross between two P. falciparum strains in those animals, something that so far was only possible in non-human primates, including chimpanzees (Vaughan et al., 2015).

These examples and others from several other infectious agents (Ernst, 2016) suggest that humanized mice will continue to contribute to a new repertoire of mouse translational models. Understanding host specificity factors for a given human pathogen is crucial for the design of susceptible humanized mice, and methods to identify such factors are described in detail by Douam et al. (2015). However promising, the establishment of these mice is relatively new and already several limitations are known (described in more detail in Ernst, 2016), which limit the extent to which the models actually mimic the human immune system. For instance, several mouse cytokines differ largely in their sequences between mouse and human, and IL-13 has no effect on human cells. This might explain the low proportions of certain human immune cell types in humanized mice. In addition, signaling and adhesion molecules are different between humans and mice and, importantly, the expression of murine and not human major histocompatibility complexes impairs the function of T cells. Attempts have been made to account for some of these limitations (Ernst, 2016) but so far humanized mice are probably best considered as a promising, but yet developing, tool in translational research.

Even so, host-specificity has limited the choice of model systems for studying protozoan parasites. The Plasmodium spp. examples and initial attempts with L. major infection in humanized mice (Wege et al., 2012) hold promise for future possibilities to investigate also other human-specific protozoan parasite species in lab mice.

### Mixing the Known — Recombinant Inbred Strains and the Collaborative Cross

In order to document an influence of genetic heterogeneity on experimental results, models beyond inbred animals are required (Phifer-Rixey and Nachman, 2015; Chow, 2016). This is also true in translational research, where humans represent a genetically diverse population. When the aim is to find differences across the genome, as in genome-wide association studies (GWAS) where a given phenotype is thought to be linked to genetics, outbred stocks of mice or rats are not a solution since they are "a genetically ill-defined set of laboratory mice that are often used erroneously in toxicology, pharmacology and basic research" (Chia et al., 2005). In addition, their usefulness is limited for practical reasons, e.g., individual phenotypic variability requires larger sample sizes than necessary with inbred strains, or the study will lack statistical power to correlate experimental differences with certain genotypes (see Chia et al., 2005; Festing, 2014). In orderto address the problem of limited genetic diversity but avoid the ill-defined genetic composition of wild animals, numerous mouse collections besides the RIS (see section The Mice We Use in Experiments — Who Are They and How Do They Live) have been established.

The concept for producing RIS sets has been extended by the Complex Traits Consortium to produce more genetically variable sets (Chia et al., 2005). These are known as Collaborative Crosses (CC) and are based on a set of 8 defined and sequenced founder strains, including three wild-derived strains of Mus and five traditional inbred strains. Although the set of strains is genetically diverse, each CC strain is at least 90% homogenous and hence genetically well-defined. The CCs were designed specifically for complex trait analysis (Churchill et al., 2004; Threadgill and Churchill, 2012) and the derived Diversity Outbred (DO) population (Churchill et al., 2012) has resulted in an even more genetically diverse mouse population (see **Figure 3**). Other derivatives of CC's concept exist, like the Heterogeneous Stock mice (Valdar et al., 2006) or the Hybrid Mouse Diversity Panel (Bennett et al., 2010). Besides GWAS studies, which map determinants for non-infectious diseases, CC animals have recently been used to map susceptibility or pathogenesis determinants in bacterial and viral infection models (Durrant et al., 2011; Ferris et al., 2013; Rasmussen et al., 2014; Vered et al., 2014; Gralinski et al., 2015; Lore et al., 2015; Smith et al., 2016). However, no data for parasite infections have been reported so far, and their large potential for exploring how host genotype influences infections needs to be explored in the future. Nevertheless, CC and DO mice also have limitations (Phifer-Rixey and Nachman, 2015). They are all derived from subspecies, which limits the genetic variation and may cause partial hybrid sterility in crosses. This, in turn, might have resulted in the elimination of genetic variation at these genomic loci. However, it is expected that further crosses will improve these models on the genetic level.

Yet, these mice will never approach the genetic diversity found in the human population or wild animal populations. In addition, individuals in natural populations encounter seasonal and spatial variability in the environment, as well as differences in climate and food availability. Wild animals are also exposed to and infected with a vast array of parasites and other pathogens, harbor different microbiota, and individuals vary in their demography, behavior and genetic composition. While it is possible to add key elements of natural variation into the above described rodent model systems, there is an increasing interest in moving beyond the controlled laboratory setting to a more realistic scenario.

### Excursion 2: Metabolic Disease in Lab Mice and Humans — Is Ecological Complexity Better than SPF Facilities?

Around 25% of the world population has a metabolic syndrome, defined by the International Diabetes Federation as either diabetes / prediabetes, abdominal obesity, high cholesterol or high blood pressure. Animal studies provide important information on these conditions (Bäckhed et al., 2004; Turnbaugh et al., 2006). The role of microbiota has gained interest also in this field because of potential new treatments which can manipulate the microbiome community and function (Borody and Khoruts, 2012). Of interest here are the large numbers of studies conducted in rodents which demonstrate, for instance, alterations in body weight and insulin sensitivity which correlate with changes in microbiota upon antibiotics treatment in these model rodents (Bäckhed et al., 2004; Turnbaugh et al., 2006). Although similar data also exists from human studies and therefore support that translation of results from rodent models to humans in this case is possible, a recent contribution to this topic questions the rodent model to mimic metabolic disease and microbiota correlations in humans. In a study performed on 57 overweight and obese adult men, the systemic effects of two antibiotic treatments compared to placebo were investigated (Reijnders et al., 2016). Changes in microbiota composition (detected by 16S rRNA microarray analysis) were observed for specific antibiotics against grampositive bacteria, but no differences were seen with the broadspectrum antibiotic. On most other readouts, the authors did not see significant responses to antibiotics treatment. Hence, changes in microbiota composition did not correlate with changes in systemic functions in humans (e.g., insulin sensitivity, energy metabolism and gut permeability; Reijnders et al., 2016), which is in contrast with rodent data (Bäckhed et al., 2004; Turnbaugh et al., 2006). Reijnders et al. (2016) discuss hypotheses for these discrepancies, mentioning treatment duration and the method of antibiotic intake (capsules or in water). In addition, a possibly important difference between the described rodent studies and the human study is the fact that humans are a "wild" population. The genetic heterogeneity and environmental influences in the human population of 57 men are indeed different from the SPF bred rodents. The effects of both previous and current microbe colonization and/or infections do influence immune responses in an individual and constitute an important difference between wild and controlled laboratory populations (reviewed in Tao and Reese, 2017). Possibly, the use of a "dirty" or wild rodent population would be a more suitable choice when the aim is to investigate correlations in a highly complex biological system.

### Getting the Rodent Model "Dirty"

In the continuum of approaches that can be employed to better understand protozoan parasite infection and immunity, "dirty" animals taken from the wild or laboratory animals exposed to wild cage-mates have emerged as a promising model (Maizels and Nussey, 2013). Arguing for such translational models, recent results demonstrate that inbred mice reared in SPF conditions have the immunological phenotype of neonatal humans, lacking

effector-differentiated and mucosally distributed memory T cells (Beura et al., 2016). In contrast, "dirty" M. musculus brought in from either a pet shop or from feral barn populations had immune responses more similar to adult humans, with high levels of memory CD8<sup>+</sup> cells, likely due to diverse microbial exposure and infection. These changes in both the innate and adaptive immune cellular responses and immune gene expression could also be recapitulated by co-housing previously SPF inbred mice with pet shop mice. While about 20% of the SPF mice died due to microbial infection, the immune response of those that survived also resembled adult humans within 4–8 weeks, with effectordifferentiated and mucosal memory T cells. In addition, within that short time frame, co-housed mice responded similarly to the wild-caught pet shop mice in terms of infection, such that they were significantly more resistant, amongst other pathogens, to challenge with the cerebral malaria model P. berghei (Beura et al., 2016). Another research group has aimed to make their inbred laboratory mouse strains "dirty" by giving them sequential infection with mouse herpes virus, influenza and an intestinal helminth in order to test how this more natural pattern of exposure to pathogens may affect immune variation and expression after vaccination (Reese et al., 2016). They found that co-infected mice had different immune gene signatures, cytokine expression and antibody levels in the blood both before and after yellow fever virus vaccination compared with their SPF lab mice controls. These expression patterns resembled those of pet store-raised mice. While getting the traditional sterile laboratory mouse models dirty may pose logistical challenges, such results should encourage researchers to revisit abandoned vaccine candidates as well as to establish different routines for testing new ones.

### Benefits of Using Wild Mice

Studies on dirty mice with the benefits described above still lack other aspects of natural variation that are important (Pedersen and Babayan, 2011). Thus, there is a need for wild model organisms that permit robust studies of the individual and environmental variation inherent in natural populations (including humans). Populations of wild mice vary in many of the same ways as humans (e.g., age, sex, condition, resources, parasite exposure, infection/co-infection, genetics, etc.), yet can provide a tractable, experimental system to test the importance of natural variability on infection, immunity and disease control. There are several key epidemiological features in wild mouse populations that closely resemble human infection dynamics, such as having great variation in infection probability, burdens and disease severity across individuals. Moreover, wild mice are commonly found chronically infected with parasites, suggesting either a high frequency of re-infection, long-lasting infections, or both (Pedersen and Babayan, 2011; Knowles et al., 2013).

One approach to start a research program on wild rodents is to study the traditional laboratory mouse species (M. musculus) in the wild (Potter et al., 1986; Viney et al., 2015). Abolins et al. (2011) found that the immune function of wild-caught M. musculus was significantly greater than lab-reared C57BL/6 mice, such that after immunization with a novel antigen wildcaught mice had higher concentrations of total IgG and IgE, produced higher and more avid concentrations of antigenspecific IgG, and had greater activation of T helper cells, macrophages and dendritic cells than lab-reared mice. While wild M. musculus offer a great parallel to lab-reared mice and can serve as comparisons for protozoan infection in the laboratory, many studies do not exhaustively sample for ectoparasites and protozoans and their true infection status is not well described. Commonly, wild M. musculus are reported to be infected with mainly ectoparasites and a few nematodes (mostly Syphacia spp. pinworms; e.g., Weldon et al., 2015); however there are records of natural infections with Giardia muris, Spironucleus muris, and Encephalitozoon cuniculi (Baker, 1998), Eimeria spp. (Ball and Lewis, 1984), Cryptosporidium spp. (Backhans et al., 2013), and T. gondii (Kijlstra et al., 2008).

### Rodents beyond Wild Mice Are Natural Hosts of a Wide Variety of Protozoa

Beyond M. musculus, there are several well-studied wild rodents that are both commonly infected with protozoan parasites and also offer tractable wild model systems for both longitudinal and experimental studies of infection and immunity. In North America, much work has focused on white-footed mice and deer mice (Peromyscus leucopus and P. maniculatus), both because they are very abundant and widespread, but also because they are competent reservoirs of important emerging zoonotic pathogens (e.g., Hantavirus and Borrelia spp.; Bedford and Hoekstra, 2015). In Europe, wood mice (Apodemus sylvaticus; e.g., Knowles et al., 2013), yellow-necked mice (A. flavicollis; e.g., Ferrari et al., 2004), bank voles (Myodes glareolus; e.g., Withenshaw et al., 2016), and field voles (M. agrestis; Smith et al., 2005; Turner et al., 2014) have been commonly studied as models for wild host-pathogen interactions and are all regularly infected with protozoan parasites. For example, wild populations of A. sylvaticus in the United Kingdom have been found to be infected with C. parvum, C. muris (Chalmers et al., 1997); > five species of Eimeria (Ball and Lewis, 1984; Higgs and Nowell, 2000); Babesia sp. and Hepatozoon sp. (Turner, 1986); two species of Trypanosoma (Noyes et al., 2002), Frenkelia microti (Svobodova et al., 2004) and T. gondii (Jackson and Siim, 1986). It is very likely that this list is a far from exhaustive.

The benefits of using wild rodent-parasite models to better understand protozoan infection dynamics include the ability to: (i) conduct longitudinal field experiments which follow marked individuals throughout their lives while measuring infection status, physiological and demographic metrics (Knowles et al., 2013; Pedersen and Antonovics, 2013; Turner et al., 2014), and crucially (ii) test the efficacy of disease control interventions at the individual and population level in an ecologically relevant environment (Knowles et al., 2013; Pedersen and Antonovics, 2013). For example, in a population of wild field voles (M. agrestis) the researchers repeatedly treated one population with a standard insecticide to reduce the prevalence of fleas, and in turn, found that this reduced the prevalence of vector-transmitted Trypanosoma spp. by ∼33% (Smith et al., 2005). In addition, in experimental field studies of both P. maniculatus and P. leucopus in the US, and A. sylvaticus in the UK, anthelmintic treatment was used to reduce nematode burdens within specific, marked animals. The treatment was found to unexpectedly increase the prevalence and/or intensity of co-infecting Eimeria spp. suggesting strong antagonistic within-host interactions between a worm and a protozoon (Knowles et al., 2013; Pedersen and Antonovics, 2013).

Research on wild rodents benefit from the extensive immunological toolbox developed in lab mice (Pedersen and Babayan, 2011). In wild populations of A. sylvaticus, innate immune responsiveness, as measured by splenocyte tumor necrosis factor responses to toll-like receptor (TLR) agonists, was found to correlate positively with Eimeria spp. fecal oocysts counts, most strongly with receptors TLR7 and TLR9 (Jackson et al., 2009). More recently, the availability of genomes for wild rodents has enabled the ability to measure immunological expression in wild rodent populations. A recent investigation of wild field voles measured expression of a wide range of innate and adaptive responses by cultured and stimulated splenocytes. Importantly, repeated measures from peripheral blood samples of IFN-y, Gata3 and IL-10 expression enabled the authors to test for correlations with specific parasite infections (Jackson et al., 2014). Taken together, wild rodents reach large sample sizes, can be repeatedly recaptured using live traps, marked and followed before and after interventions, and are commonly infected with protozoan parasites. Studying the dynamics of protozoan infections in wild rodents is a valuable resource for expanding our knowledge in infection biology and might thus be a useful addition for translational research on human protozoan infections.

### A Case for Going Wild: Do *T. gondii*-Induced Behavioral Changes Exist in Natural Habitats?

How relevant are findings which suggest parasite influences on lab mouse behavior when performed in lab environments? The so-called "manipulation hypothesis" of a T. gondii infection in rodents suggests that infection leads to subsequent changes in the animal's behavior, with one consequence being that they lose their fear for feline odor (e.g., fur or urine). Cats and other felids are the only definite hosts where sexual reproduction of T. gondii can take place. Therefore, at first sight it makes sense that such "manipulated" infected rodents would experience more fatal encounters with a cat than noninfected ones, thereby increase the chance for T. gondii to sexually reproduce with another strain from a second subsequent infected prey.

The advantage or necessity of this scenario for parasite sexual reproduction in the wild has been called into question (Worth et al., 2013), but here we focus on the fact that all reported experiments were done exclusively in lab animals (Worth et al., 2014). At first sight this might not seem problematic since M. musculus and T. gondii naturally occur together. However, it is well known, but not necessarily well appreciated, that behavioral studies of rodents can be influenced by the methods used, housing conditions, genetic background and whether they are lab or wild-derived animals (Wolff, 2003; Beckers et al., 2009; Fonio et al., 2012; Chalfin et al., 2014; Newman et al., 2015). Even differences in the microbiota can have profound effects (Hsiao et al., 2013; see section Microbiota as a General Confounder for Rodent Experiments). Moreover, lab mice have been selected for decades for docile behavior, while wild mice show anxious behavior under natural conditions (Latham and Mason, 2004; Yoshiki and Moriwaki, 2006; Fonio et al., 2012; Chalfin et al., 2014).

How well does the manipulation hypothesis apply to the natural situation of predator (cat) and prey (T. gondii-infected mouse or rat)? Ecological observations might explain some of the observed discrepancies, i.e., no behavioral differences were found in one study but were indeed found in another (Worth et al., 2014). Several studies indicate that predation risk of wild or wildderived small rodents depends more on habitat characteristics (e.g., ability to hide) than on whether the rodent senses a present predator by its odor or even by its physical presence (Orrock, 2004; Powell and Banks, 2004; Verdolin, 2006). Aversive responses to predator odor can also differ dramatically between individual lab rats (Hogg and File, 1994) and, importantly, between lab mice and wild mice (Coulston et al., 1993; Hebb et al., 2004). Thus, experiments with "fearless" lab mice in nonnatural terrains may not accurately reflect the behavioral changes induced by a parasite like T. gondii under natural conditions.

### EXPERIMENTAL CHALLENGES AND AVAILABLE RESOURCES FOR NON-TRADITIONAL RODENT MODELS

Having provided a number of reasons for considering wild rodents as alternatives, we will briefly address the experimental challenges. Approaching studies on non-model rodents is demanding but the toolbox has improved significantly compared to a decade ago (Pedersen and Babayan, 2011; Zimmerman et al., 2014). Method development in the fields of genomics, transcriptomics and proteomics together with increasing affordability provide better prerequisites for research on non-model alternatives (Jackson, 2015). Next-generation sequencing has led to constantly expanding genomic data. This is demonstrated by the nearly 4,000 eukaryotic genomes available on NCBI Genome (February, 2017) as compared to around 650 in 2013 (Ellegren, 2014).

### Database Resources for Wild Rodent Genomes

Any resources available for lab mice are to varying extents useful starting-points for work on wild rodents. The Mouse Genomes Project is the biggest collection of genomic data on rodents (Table S1). Currently it consists of whole-genome assemblages and strain-specific gene annotations of 16 inbred and wildderived mouse strains. A goal of this project is the classification of sequence variations between common laboratory strains compared to the reference strain C57BL/6J (Adams et al., 2015; Doran et al., 2016). All sequence reads, variants and assemblages can be useful references for highly recombinant outbred strains (Nicod et al., 2016) or wild rodent genomes. There are also increasing numbers of genomes and/or transcriptomes available for wild rodents (e.g., A. sylvaticus, M. glareolus, and M. agrestis; see **Figure 4**), with most of them being "work-in-progress" considering assemblage status and annotations (for details see Table S1 and links therein). The quality and coverage of these genomes vary and there is for instance no clear definition of a required coverage for referring to DNA sequences as a "genome" (Ellegren, 2014). However, they do provide a good source for homology searches for a gene-of-interest, primer design for PCR applications etc. Naturally, purification of DNA, RNA or proteins and functional PCR protocols may require protocol optimization when applied to new species but otherwise follow established schemes. Some database resources and other initiatives to promote such development are discussed below and summarized in Table S1.

While reference genomes are not a prerequisite for some studies they are, for instance, indispensable as a template in quantitative gene expression studies with high-throughput sequencing of RNA (RNA-seq; Vijay et al., 2013). However, bioinformatics pipelines are often developed for established model organisms and may require considerable adjustments for non-model organisms (Ekblom and Wolf, 2014). Although such limitations may hamper the speed of omics applications to nonmodel organisms, genomes of non-model rodents will serve as excellent resources for developing species-specific tools to measure, for example, expression of immunological responses to infection. Readers that are interested in considering genome sequencing for their own non-model organism are referred to a recent step-by-step introduction of the required workflows (Ekblom and Wolf, 2014) and to the Generic Model Organism Database (GMOD) initiative, which provides software tools and data models for subsequent representation of their annotated genomes and curated data of their model organism (O'Connor et al., 2008).

### Transcriptomes Do Not Necessarily Predict Protein Levels

It should be emphasized that genomic and transcriptomic data, as valuable as they are, provide only indirect means with regards to proteomic output in response to infection. In general, the relationship between the concentration of a given transcript and its encoded protein(s) is difficult to predict just by RNA-seq or qPCR data (Liu et al., 2016). For example, Chick et al. (2016) used the currently most sensitive technology for abundance determination of both transcripts and proteins and applied them to CC and DO mice (see section Humanized Mice: Rodents Which Mimic the Human Immune System and **Figure 3**). They showed that for many genes the levels of the corresponding protein varied substantially in genetically divergent mice. Sex also influenced protein amounts within a given species (Chick et al., 2016). These recent data emphasize the importance of quantitative proteomic measures in general to complement or validate transcriptomic data, but also highlight that genetic diversity within mice influences the results.

### Antibodies, Cytokines, and Protein Quantification

Antibody-based assays are still the cornerstone for qualitative and quantitative determination of immunological parameters like chemokines or cytokines but also other proteins of interest. Numerous well-defined and evaluated reagents exist for lab mice and rats but their usefulness for wild rodents with respect to

cross-reactivity is largely unexplored and presumably quite low in many cases. The same applies to immunological effector molecules like cytokines, for which IFN-γ is a good example. IFN-γ is known to be highly species-specific, which made the production of a recombinant protein active with M. glareolus or Microtus spp. cell lines a prerequisite (Torelli et al, in preparation). Starting from genomic sequences and going to the purification of active recombinant protein from E. coli required less than half a year and will now provide the scientific community with this important cytokine.

Developing antibodies that (cross)react with wild rodents is certainly much more time and resource-consuming, but feasible. An alternative could be parallel (or selective) reaction monitoring (PRM, SRM) which are mass spectrometry-based methods that quantify unique, specific peptide sequences of a given protein (Rauniyar, 2015; Bourmaud et al., 2016). The method was recently used to quantify several cytokines and chemokines from human cells (Muqaku et al., 2015). The appealing aspect of this admittedly demanding method lies in the fact that by carefully selecting peptide sequences conserved between rodent species, they could be used across those species at relatively low cost, once established (Hüttenhain et al., 2009). Ideally, it could thus be regarded as a community effort. A database with corresponding peptides from human and mouse proteomes does exist (see Table S1) and constitutes a useful starting point (Peptide Atlas; Deutsch et al., 2008).

Given the availability of published genomic and transcriptomic data of wild rodent species (Table S2), work similar to the studies mentioned will hopefully expand the current toolbox for non-model rodents in the near future. A dedicated web site with information on such shared resources, but also on cross-reacting reagents such as polyclonal or monoclonal antibodies or commercial cytokines and other proteins tested in non-model rodents, although not yet existing, would greatly boost the interest and ease of use of non-model organisms in future studies.

## CONCLUDING REMARKS

We are convinced that rodents will continue to be important translational models for research on protozoan parasites, given that appropriate considerations are made during experimental design. By providing some examples where translation from rodent disease models to human medicine has failed, and, more importantly, by pointing at identified reasons for inconclusive or misleading data, we wish to inspire readers to consider more than the most convenient model for future experiments. Making use of rich database-resources that are available for investigating, e.g., expected phenotypes of mice, will aid in this respect. We also hope that readers are encouraged to consider and control for various confounders such as microbiota influences and housing conditions in their experimental designs. While wild models pose some challenges, we have pointed out that these rodents possess distinct advantages with regards to genetic variability and environmental exposures that can reflect immunological responses to parasites in humans more adequately than current lab models. The increasing availability of genome and transcriptome datasets as well as improved methods for quantitative proteomics already show their impact on wild infection biology.

### AUTHOR CONTRIBUTIONS

TE, FT, CK, AP, and FS contributed to the text and approved its final version.

### FUNDING

TE and FT are supported by and TE, FT, CK, and FS are members of the Research Training Group 2046 "Parasite Infections: From Experimental Models to Natural Systems," funded by the Deutsche Forschungsgemeinschaft. Own cited work from FS and CK was supported by the Robert Koch Institute. AP is supported by a Chancellors Fellowship from the University of Edinburgh, and her cited work was supported by a NERC

### REFERENCES


grant (NE/G006830/1 and NE/G007349/1) and a Wellcome Trust Centre for Infection, Immunity and Evolution Advanced Fellowship (no. 095831).

### ACKNOWLEDGMENTS

The authors thank Toni Aebischer and Ankur Midha for valuable comments and suggestions on the manuscript.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fcimb. 2017.00238/full#supplementary-material


composition of the fecal microbiota of inbred mice. PLoS ONE 10:e0116704. doi: 10.1371/journal.pone.0116704


using gene set enrichment analysis. EMBO Mol. Med. 8, 831–838. doi: 10.15252/emmm.201506025


**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 Ehret, Torelli, Klotz, Pedersen and Seeber. 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.

# MALDI-TOF MS Profiling-Advances in Species Identification of Pests, Parasites, and Vectors

Jayaseelan Murugaiyan\* and Uwe Roesler

Institute of Animal Hygiene and Environmental Health, Centre for Infectious Medicine, Freie Universität Berlin, Berlin, Germany

Invertebrate pests and parasites of humans, animals, and plants continue to cause serious diseases and remain as a high treat to agricultural productivity and storage. The rapid and accurate species identification of the pests and parasites are needed for understanding epidemiology, monitoring outbreaks, and designing control measures. Matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS) profiling has emerged as a rapid, cost effective, and high throughput technique of microbial species identification in modern diagnostic laboratories. The development of soft ionization techniques and the release of commercial pattern matching software platforms has resulted in the exponential growth of applications in higher organisms including parasitology. The present review discusses the proof-of-principle experiments and various methods of MALDI MS profiling in rapid species identification of both laboratory and field isolates of pests, parasites and vectors.

#### Edited by:

Anton Aebischer, Robert Koch-Insitute, Germany

#### Reviewed by:

Joshua D. Shrout, University of Notre Dame, USA Maria Hadjifrangiskou, Vanderbilt University Medical Center, USA

#### \*Correspondence:

Jayaseelan Murugaiyan jayaseelan.murugaiyan@fu-berlin.de

> Received: 15 February 2017 Accepted: 27 April 2017 Published: 15 May 2017

#### Citation:

Murugaiyan J and Roesler U (2017) MALDI-TOF MS Profiling-Advances in Species Identification of Pests, Parasites, and Vectors. Front. Cell. Infect. Microbiol. 7:184. doi: 10.3389/fcimb.2017.00184 Keywords: MALDI MS typing, MALDI TOF MS, intact protein profiling, pests, parasites, species, spectra reference databases

### INTRODUCTION

Invertebrate pests and parasites of plants, mammals, birds, amphibians, and reptiles account for increased losses in the agricultural sector and continue to play a considerable role in the spread of infectious diseases (Paini et al., 2016; Poulin et al., 2016). Accurate and rapid species identification of pests and parasites is extremely important for initiating species-specific treatment procedures, understanding the epidemiology, monitoring of outbreaks, and designing control measures (Gibson, 2009; Furlong, 2015). Traditionally, trained taxonomists or entomologists visually examine or observe microscopically the morphological characteristics for species determination. However, in addition to being time-consuming, misidentification possibilities, distinguishing immature or development stages, damaged samples, cryptic species, and species differing by minor morphological characteristics make identification challenging and often impossible (McKeand, 1998). Molecular methods, whichever available, are accurate and applicable to any development stages. On the other hand, these methods are also labor intensive, expensive, time consuming, and difficult to apply for species for which sequences are not available (Wong et al., 2014). Furthermore, in certain cases such as Leishmania subtyping, which is crucial for treatment, identification generally requires several weeks for performing complex and expensive analyses (Roelfsema et al., 2011). In recent years, proteome based linear matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS profiling or MALDI MS typing), which is a well-established technique for microbial species identification, has also been successfully applied to a variety of parasites and their vectors (Seng et al., 2009; Bizzini and Greub, 2010; Patel, 2013; Nomura, 2015; Karger, 2016; Singhal et al., 2016; Yssouf et al., 2016). The popularity of the method is due to its rapidness, easy to use, high throughput analysis, accuracy, reliability equal to that of the molecular methods of species identification and cost-effectiveness despite the initial cost of the machine (Dhiman et al., 2011; Neville et al., 2011; Tran et al., 2015; Ge et al., 2016). The technique involves generation of MALDI MS spectra for a given species and pattern matching with that of the spectra database of the well-defined species to deduce the species information (Welker and Moore, 2011; Nagy et al., 2012; Suarez et al., 2013). The spectra pattern matching is carried out using automated commercial software suites, such as Bruker Biotyper (Bruker Daltonics, Bremen, Germany), VITEK MS (BioMérieux, Nuertingen, Germany: earlier Axima (Shimadzu)-SARAMIS (AnagnosTec) systems), Andromas (Andromas SAS), or MicrobeLynx (Waters) (Sogawa et al., 2011; Bille et al., 2012; Patel, 2013; Cassagne et al., 2016). The commercial software tools are usually integrated with their own spectra reference database, and utilize a unique algorithm for spectra processing, pattern matching, and result interpretation.

The manufacturer-provided database is limited only in terms of reference spectra for available microbial species, and currently the reference information for pests and parasites are not included. The software generally includes the possibility to create reference spectra for any organism to be integrated within the existing database. The database extension has been utilized to create additional reference spectra to enhance the identification confidence and to include reference spectra of the missing species, including higher organisms (Bohme et al., 2012; Murugaiyan et al., 2012, 2014; Hoppenheit et al., 2013). Several reviews have been dedicated to the recent developments of MALDI MS typing of plant nematodes and organisms related to parasitology (Ahmad and Babalola, 2014; Karger, 2016; Singhal et al., 2016; Yssouf et al., 2016). Therefore, the focus of this review is on the various approaches reported for MALDI MS typing based species identification of pests, parasites, and their vectors ranging from laboratory isolates to that of field samples.

### MALDI MS BASIS OF SPECIES IDENTIFICATION

MALDI MS based species identification involves three steps:


A small portion of biological substances e.g., microbial colonies or a drop of intact/crude protein extracted using simple procedures is directly added to the target plate and allowed to air dry. Then, the sample spot is overlaid with a drop of an excess concentration of UV-absorbing small organic compounds, referred to as a matrix. There are several choices of matrix such as α-cyano-4-hydroxycinnamic acid (HCCA/CHCA), sinapinic acid (3, 5- dimethoxy-4-hydroxycinnamic acid) (SA), and 2,5-dihydroxybenzoic acid (DHB). HCCA is most frequently reported matrix; however, there is no universally recommended matrix.

In the MALDI instrument, a small region of the crystalline sample-matrix spot (usually 0.05–0.2 mm in diameter) is irradiated using a pulsed beam of a laser, generally a nitrogen beam with a wavelength of 337 nm is used in most commercial machines. The matrix absorbs the laser energy, and rapidly heats up resulting in desorption (vaporization) or structural decomposition of the proteins and protonation to form a hot dense plume of ablated gases and ions (Clark et al., 2013). Using an electric field, the ions are accelerated into a vacuum tube that terminates in an ion detector. The ions are usually of single charge and the acceleration voltage results in the same kinetic energy applied to every single charged ions, which results in separation of ions based on mass/charge (m/z) ration in the drift or vacuum tube.

The time of flight "TOF" of ions is recorded as MALDI spectra where the x-axis represents m/z ratio and the y-axis represents the intensity (or number) of same/similar ions. MALDI MS ions are singly charged, representing the non-fragmented parent ion mass, the resulting spectra is simple, and therefore, the species of unknown samples can be easily deduced after data processing and direct pattern matching with that of the spectra established from well-defined species compiled as a database. The identity of spectral peaks or protein sequence information is not important, as the species deduced by matching the protein profiles usually at a range of 2–20 kDa (2,000–20,000 m/z) (Evason et al., 2001; Sauer and Kliem, 2010; Welker and Moore, 2011; Karlsson et al., 2015; Cassagne et al., 2016).

### ADVANCES IN INVERTEBRATE PESTS, PARASITE, AND VECTOR PROFILING

In the past 16 years (**Figure 1**), MALDI MS profiling has been successfully applied for species identification of different pests, parasites and vectors such as nematodes, protozoa, and arthropods.

### Proof of Principle Experiments

The early proof of principle experiments were focused on identification of biomarker peaks, standardization of sample preparation, matrixes and measurement optimization.

### Protozoans and Unicellular Parasites

The first report of MALDI MS typing of protozoa was demonstrated using Cryptosporidium spp. associated with human infections. The species-specific spectra were reported from oocytes of C. parvum and C. muris isolated from feces of experimentally inoculated mice, lysed by freeze-thaw cycle, and spotted with HCCA as matrix (Magnuson et al., 2000). Later, it was shown that incubation of intact oocytes and purified sporozoites for 45 min with the matrix was critical for generating mass spectra with a large number of reproducible

peaks for C. parvum oocysts (Glassmeyer et al., 2007). Subsequently, direct application of whole spores, spore shells, and soluble fractions of spore-forming unicellular parasites such as microsporidia, Encephalitozoon cuniculi, Encephalitozoon hellem, Encephalitozoon intestinalis, and Brachiola algerae isolated from humans and propagated on monolayers of Vero monkey kidney (E6) cells, displayed species-specific markers in the mass range of 2,000–8,000 Da (Moura et al., 2003). Later, species-specific peaks in the range of m/z 3,000–19,000 was reported for the waterborne protozoan parasite, Giardia spp., the causative agent of giardiasis. The cysts, cyst walls, and trophozoites of G. lamblia and G. muris isolated from feces of experimentally challenged mice, were washed, mixed with an equal volume of sinnapinic acid, incubated, and spotted for MALDI MS analysis (Villegas et al., 2006).

### Insects and Pests

Although not a pest or parasite, Drosophila has been used as a model for insect profiling possibilities. Protein extraction by simple grinding of adult whole insect in water was shown to generate species-specific spectra capable of distinguishing sibling species of Drosophila sub-species. The insect sex and matrix was not found to influence the spectra (Campbell, 2005). Likewise, species-specific peaks of varying intensities in a range of 3,000–25,000 m/z have been reported for three different aphids (plant phloem sap-feeding insects), green peach aphid Myzus persicae Sulzer, cowpea aphid Aphis craccivora Koch, and bluegreen aphid (blue alfalfa aphid) Acyrthosiphon kondoi Shinji, independent of their dietary host plants (Perera et al., 2005).

### Nematodes and Developmental Stage Discrimination

Species-specific and diagnostic peaks have also been reported for simple extracts of three plant nematodes, Anguina tritici (wheat seed-gall nematode), its closely Anguina funesta (ryegrass nematode), and Meloidogyne javanica (root-knot nematode that infects horticultural and vegetable crops; Perera et al., 2005). Several years later, direct crushing of the root-knot nematode, Meloidogyne incognita was shown to be useful in rapid discrimination between the harmless and harmful J<sup>2</sup> developmental stages and adult nematode (Ahmad et al., 2012).

### Peptide Profiling

In this approach, also referred as shotgun mass mapping or SMM, whole body protein extracts were subjected to trypsin digestion without purification or fractionation steps, and the resulting peptides were utilized for generation of MALDI MS spectra for insect vector species such as Drosophila (Feltens et al., 2010) and biting midges (Uhlmann et al., 2014). Feltens et al. had applied nano-high performance liquid chromatography coupled with electrospray ionization mass spectrometry for identification of some of the MALDI MS profiles and revealed that most of the proteins were of muscles and mitochondria. However, SMM is time-consuming and handling large set of samples is challenging.

### Data Independent Analysis and Clinical Sample Survey

The data independent species discrimination or grouping of microorganism is based on the visual examination for the presence or absence of peaks. This technique is very similar to those analyses performed before the days of the implementation of software with automated pattern matching algorithms; however, a different algorithm is used for rapid determination of the presence or absence of peaks for identification or discrimination. For example, this technique has been reported for the discrimination of the Leishmania subgenus Viannia or Leishmania exclusively based on the presence of 2 pairs of peaks (Mouri et al., 2014), as well as differentiation of protozoan parasitic Entamoeba histolytica and Entamoeba dispar (Azian et al., 2006).

### Database-Based Enabled Rapid Species Identification Extended to Field Samples

MALDI MS typing based rapid species identification is usually achieved through pattern-matching of the unknown samples with that of a spectral reference library (database) created from known organisms. The main concerns at the proteome level are differences between the various developmental stages and complexity associated with various body parts. As listed in **Table 1**, vector related reference spectra were reported using commercial MALDI instrument-software suites. The proteins were extracted through homogenization and the parameters recommended for microbial species identification was followed. HCCA and SA were reported as the most utilized matrices for Bruker Biotyper and SARAMIS (Vitex MS), respectively. Despite the success of the procedures, these reference spectra remain in-house databases. Following the compilation of vector-specific reference spectra databases, the method proved to be rapid (∼2–5 min/sample) as in the case of microbial species identification.

### Parasites

In every reported case of parasite database approaches, Bruker BioTyper software tool and formic acid/acetonitrile extraction was applied. In the first such study, 56 clinical specimens belonging to 23 species of Leishmania were cultured, and promastigote pellets were utilized for database construction (Cassagne et al., 2014). Among 69 clinical isolates used for testing, only three samples were not identified. In a similar study, a database was constructed from four reference strains and the two clinical isolate were identified as L. infantum (Culha et al., 2014). In another study, using 19 enteric parasite Blastocystis isolates from 19 patients, a database for five subtypes was created and the remaining specimens were identified by matching (Martiny et al., 2014). Likewise, the differentiation of E. histolytica and E. dispar was demonstrated after establishment of the reference spectra and discriminating peaks were matched with the proteins identified through SDS-PAGE MALDI TOF MS based protein identification approach (Calderaro et al., 2015). Bruker BioTyper based database compilation were also reported for the food nematode Trichinella (Mayer-Scholl et al., 2016), Trichomonas vaginalis (Calderaro et al., 2016), Acanthamoeba spp.(Del Chierico et al., 2016) using a database for direct identification of trypanosomatids (Avila et al., 2016).

### Toward "Vector Spectra Reference Database"

Kaufmann et al., were the first to report on utilization of SARAMIS premium software to create a reference database for two laboratory-reared C. nubeculosus biting midges (Kaufmann et al., 2011), and which was then extended to a larvae-specific database, screening of field collected samples, and applied for entomological surveys in Senegal, Africa. It was further reported as a means for discrimination of cryptic Anopheles, and demonstrated that the usefulness of mosquito eggs in species identification of field collected samples (Schaffner et al., 2014; Yssouf et al., 2014a). The Bruker Biotyper database has since been utilized to create a comprehensive "Vector specific reference spectra database" that includes spectra from leg proteins of 6 tick species, 30 mosquito species, one louse, one triatomine, one bed bug, and five flea species (Yssouf et al., 2013a,b, 2014b, 2015a). In addition, the package has been used to establish an aquatic developmental stage database starting from the larva stage to pupa of 6 mosquito species and detection of host blood meal and the presence of parasites (Niare et al., 2016; Laroche et al., 2017).

### Pests

Despite the earlier proof of concept experiments for pests, the database approach was reported for only one plant pest in which Biotyper was utilized to distinguish evolutionary and morphologically close species of spider mites. Female adults of the Kanzawa (Tetranychus kanzawai), the two-spotted (T. urticae) spider mites and three other related species, namely T. phaselus Ehara (Tp), the bean red spider mite (T. ludeni Zacher) (Tl), and the tomato red spider mite (T. evansi Baker & Pritchard) (Te) were shown to be distinguishable. On the other hand, male adults and nymphs were reported to be non-distinguishable. Direct lysing of a single intact mite on target plates using double side carbon tape was also shown to generate spectra comparable to that of the extracts from 10 pooled individuals (Kajiwara et al., 2016). Recently, it was demonstrated that using MALDI MS, Tinkerbell LT, and its associated software MicroIDSys (ASTA Inc. Suwon, Korea), the larvae of the Korean apple pest, Carposina sasakii, could be effectively discriminated in about 15 min. Without such rapid identification methods, the export of these fruits might be hampered or rejected due to time-consuming pre-export inspection (Jeon et al., 2017).

### DUAL IDENTIFICATION OF VECTORS, PARASITES, AND MEAL SOURCE

MALDI MS profiling has also been shown to be an effective method for simultaneous identification of vectors and parasite species, as in case of simultaneous identification of Borrelia crocidurae/Ornithodoros sonrai and Rickettsia spp. in ticks (Fotso et al., 2014; Yssouf et al., 2015b). Direct spotting of haemolymph from a dissected tick leg allows for the simultaneous and direct species identification of ticks and associated pathogens, and useful in species identification of parasites and vector while leaving the vector remains available


(Continued)

TABLE 1 | Reference

 database based species

identification

 of parasite vectors.


UNIVERSAL DATABASE AND FIELD

status (Laroche et al., 2017).

SAMPLES

All the reported parasitology associated databases were developed as in-house databases and usually not accessable by other researchers, although several authors have agreed to supply the reference spectra upon request to other scientists possessing the same instrumentation and software tools. However, beyond a few initiations such as SpectraBank, there is no universal reference database such as protein database Swissprot/Uniprot for the purpose of species identification (Bohme et al., 2012). Recently, it was shown that the spectra generated from two different commercial MALDI MS instruments (Axima Confidence and Bruker Ultraflex III MS) could be analyzed in a single database (SARAMIS) in which a database was constructed for 20 species of phlebotomine sand flies based on measurements with Axima Confidence MS (Mathis et al., 2015). This underscores the possibilities of creation of online public reference databases that might be useful for any type of MALDI MS machines or spectral data formats. The online public database for parasitology is of great importance in terms of economy, time, and rapid analysis of samples collected from different geographical regions or hosts. The open source reference spectra database demands standardization of influencing parameters/processes affecting the sampling, such as developmental stage, specimen storage conditions, sample preparation/extraction methods, variations due to the spotting, instrumental, and post-measurement software settings.

for other laboratory investigations (Fotso et al., 2014). MALDI MS profiling has further been successfully applied to determine the feeding patterns of mosquitoes up to 24 h post-blood meal (Niare et al., 2016). A future goal for the application of MALDI MS profiling in pest/parasites will be the rapid species identification of vectors, parasites and blood meal either through individual analysis or through simultaneous monitoring processes. The recent demonstration of rapid distinguishing of the Plasmodium infection status among Anopheles stephensi mosquitoes underscores that MALDI MS typing could be useful in entomological surveys including species-specific infection

### CONCLUSION

Despite the achievements of MALDI TOF MS in microbial species identification, the application in parasitology remains limited to in-house databases integrated into commercial software. There is no commercial software tool or reference database available for parasitology, however, most of the successful reports were based on the principles, procedures, software, and databases best described for microorganisms. The crucial factors influencing parasitology MALDI MS typing includes: specimen associated parameters (age, developmental stage, sex, differing body parts, fed state), sample preparation procedures (sample storage, protein extraction solvents, and methods, sample spotting method, and matrix), and

Identification

 System, originally developed by AnagnosTec,

Potsdam-Golm,

 Germany and now acquired and redeveloped

 as Vitek-MS by bioMérieux,

 Marcy L'Etoile, France).

measurement parameters. MALDI MS typing of field specimens will be helpful in creating a distribution map and evaluation of the spread of parasites associated with disease. Few reports describe the spectral differences among the geographically different populations (Dieme et al., 2014; Dvorak et al., 2014; Hoppenheit et al., 2014). However, one should remain cautious, as the reliability of MALDI TOF MS in phylogenetic analysis is yet to be proven. The establishment of open source software and databases might be useful in future parasitological surveys and for rapid assessment of real time infection status.

### REFERENCES


### AUTHOR CONTRIBUTIONS

JM conceived and wrote the manuscript. UR reviewed the manuscript.

### ACKNOWLEDGMENTS

We would like to thank Karsten Tedin, Freie Universitaet Berlin, Germany and Vanessa Kowbel, University of Manitoba, Canada for their assistance of skillful grammatical corrections.


homologues, by MALDI-TOF mass spectrometry. J. Asia Pac. Entomol. 20, 411–415. doi: 10.1016/j.aspen.2017.02.013


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

Copyright © 2017 Murugaiyan and Roesler. 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.

# Nematode Species Identification—Current Status, Challenges and Future Perspectives for Cyathostomins

Christina M. Bredtmann<sup>1</sup> , Jürgen Krücken<sup>1</sup> , Jayaseelan Murugaiyan<sup>2</sup> , Tetiana Kuzmina<sup>3</sup> and Georg von Samson-Himmelstjerna<sup>1</sup> \*

<sup>1</sup> Department of Veterinary Medicine, Institute for Parasitology and Tropical Veterinary Medicine, Freie Universität Berlin, Berlin, Germany, <sup>2</sup> Department of Veterinary Medicine, Institute for Animal Hygiene and Environmental Health, Freie Universität Berlin, Berlin, Germany, <sup>3</sup> Department of Parasitology, I.I. Schmalhausen Institute of Zoology, Kyiv, Ukraine

Human and animal health is globally affected by a variety of parasitic helminths. The impact of co-infections and development of anthelmintic resistance requires improved diagnostic tools, especially for parasitic nematodes e.g., to identify resistant species or attribute pathological effects to individual species or particular species combinations. In horses, co-infection with cyathostomins is rather a rule than an exception with typically 5 to 15 species (out of more than 40 described) per individual host. In cyathostomins, reliable morphological species differentiation is currently limited to adults and requires highly specialized expertize while precise morphological identification of eggs and early stage larvae is impossible. The situation is further complicated by a questionable validity of some cyathostomins while others might actually represent cryptic species complexes. Several molecular methods using different target sequences were established to overcome these limitations. For adult worms, PCR followed by sequencing of mitochondrial genes or external or internal ribosomal RNA spacers is suitable to genetically confirm morphological identifications. The most commonly used method to differentiate eggs or larvae is the reverse-line-blot hybridization assay. However, both methods suffer from the fact that target sequences are not available for many species or even that GenBank® entries are unreliable regarding the cyathostomin species. Recent advances in proteomic tools for identification of metazoans including insects and nematodes of the genus Trichinella will be evaluated for suitability to diagnose cyathostomins. Future research should focus on the comparative analysis of morphological, molecular and proteomic data from the same cyathostomin specimen to optimize tools for species-specific identification.

#### Keywords: cyathostomins, nematodes, diagnostic, PCR, MALDI-TOF MS

### INTRODUCTION

Parasitic helminths globally affect human and animal health and can be of zoonotic relevance (e.g., Ascaris spp.). In equines, the most important intestinal nematodes belong to the family Strongylidae and are comprised of two subfamilies: The Strongylinae encompassing 14 species in 5 genera (Strongylus, Oesophagodontus, Triodontophorus, Bidentostomum, and Craterostomum), and the Cyathostominae encompassing currently 50 valid species in 14 genera (Caballonema,

#### Edited by:

Lilach Sheiner, Wellcome Trust Centre for Molecular Parasitology, United Kingdom

#### Reviewed by:

Elias Papadopoulos, Aristotle University of Thessaloniki, Greece Martin Krarup Nielsen, University of Kentucky, United States

#### \*Correspondence:

Georg von Samson-Himmelstjerna gvsamson@fu-berlin.de

> Received: 31 March 2017 Accepted: 12 June 2017 Published: 28 June 2017

#### Citation:

Bredtmann CM, Krücken J, Murugaiyan J, Kuzmina T and von Samson-Himmelstjerna G (2017) Nematode Species Identification—Current Status, Challenges and Future Perspectives for Cyathostomins. Front. Cell. Infect. Microbiol. 7:283. doi: 10.3389/fcimb.2017.00283 Coronocyclus, Cyathostomum, Cylicocyclus, Cylicodontophorus, Cylicostephanus, Cylindropharynx, Gyalocephalus, Hsiungia, Parapoteriostomum, Petrovinema, Poteriostomum, Scrjabinodentus, Tridentoinfundibulum) (Lichtenfels et al., 2008), in contrast to previous publications listing 51 or 52 species in 13 genera (Lichtenfels, 1975; Lichtenfels et al., 2002). In the literature, the term "small strongyles" has either been coined to include only the Cyathostominae or all equine strongylidae except the genus Strongylus, which were designated "large strongyles" (Lyons et al., 1999). Although, still widely used, it is now recommended to avoid the terms small and large strongyles (Lichtenfels et al., 2002).

Since prevalence of the highly pathogenic Strongylus species declined after introduction of the macrocyclic lactones (Herd, 1990), the cyathostomins are currently recognized as the most important equine parasites because of (i) their up to 100% prevalence in equids (Lyons et al., 1999), (ii) numerous reports of anthelmintic resistance and (iii) their pathogenicity which becomes particularly manifest in cases of sometimes fatal larval cyathostominosis (Love et al., 1999). Anthelmintic resistance against benzimidazoles is highly prevalent worldwide and pyrantel resistance is also frequently observed whereas reduced efficacy of macrocylic lactones has rarely been reported (Kaplan, 2002; Kuzmina and Kharchenko, 2008; Von Samson-Himmelstjerna, 2012; Matthews, 2014; Nielsen et al., 2014).

Cyathostomins have a direct life-cycle with adults located in the lumen of caecum and colon, shedding eggs with the feces. First larvae (L1) hatch in the feces, molt twice to infectious third larvae (L3) which are ingested by equids. In the large intestine, L3 encyst inside the intestinal wall and may also undergo hypobiosis for months, before molting to fourth larvae (L4) (Corning, 2009). Synchronous excystation of large numbers of hypobiotic larvae potentially causes larval cyathostominosis characterized by severe inflammation leading to weight-loss, diarrhea, colic, or even death (Love et al., 1999).

Although, Cyathostomins are a threat to equine welfare and scientific efforts to address this problem are frequently undertaken, research is impaired by the lack of sufficient identification methods (Lichtenfels, 2008). This perspective addresses the different methods, their advantages and limitations and gives an outlook on possible future methods for nematode identification using the cyathostomins as paradigm.

### SAMPLES AND SAMPLING

The first challenge for species identification is the availability of suitable specimens. While strongylid eggs can be easily collected from feces, they have virtually no diagnostically useful morphological features. Strongylid L3 can be obtained from eggs using different fecal cultures methods (Smyth, 1990). However, only some L3 can be identified to the genus level, and this requires a high experience level. Only for a few species in vitro culture to the L4 (Chapman et al., 1994; Brianti et al., 2009) has been described. Therefore, adult parasites must be collected from naturally infected hosts. In horses, only a few adult strongyles are occasionally shed with the feces but collection of adult nematodes from feces after anthelmintic treatment is possible (Osterman Lind et al., 2003; Kuzmina et al., 2005; Kuzmina and Kharchenko, 2008). However, the complete worm burden representing all species in the living horse will only be documented by examination of all feces over several days, which may be associated with degradation of worms leading to distorted results. A more exact and meaningful method is the collection of adult nematodes from the content of the horse intestine (Drudge and Lyons, 1977). The critical test method, which is described in detail by Drudge et al. (1963), is a combination of both, the fecal collection over a week and collection during necropsy. This method is widely used to study the effectiveness of anthelmintic compounds (e.g., Lyons et al., 2007, 2010). Due to the need of sacrificed or slaughtered horses, these methods are restricted to research. Thus, there is a great need to develop alternatives for precise nematode diagnosis for living horses. The immediate research aim is therefore the development of effective and specific non-invasive cyathostomin identification methods.

### MORPHOLOGICAL IDENTIFICATION

For more than 100 years (Molin, 1861; Loos, 1900), a large number of cyathostomin species has been morphologically described using 93 different names. In the meantime, several previously described species are considered synonyms (Lichtenfels et al., 1998) and currently 50 species are recognized as valid. Comprehensive identification keys summing up the descriptions were published (Lichtenfels, 1975; Tolliver, 2000; Lichtenfels et al., 2008).

Morphological identification of adult strongyles relies on careful examination of faint characters at the anterior end of the adult nematodes or of the reproductive system. These traits include the size and shape of buccal capsules, internal and external leaf crowns and its extra-chitinous support as illustrated in **Figure 1** to point out that differences are very faint. Fine morphological structures of posterior end such as size and shape of the bursa, genital cone, gubernaculum, and spicules in males and shape of the tail, size and proportion of different parts of the reproductive system in females are also valuable for species differentiation (Lichtenfels, 1975; Dvojnos and Kharchenko, 1994; Lichtenfels et al., 2008). However, reliable morphological identification of adult cyathostomins can only be achieved following several years of intensive training and currently only few experts are available worldwide (Lichtenfels et al., 2008).

Whereas, adult cyathostomins can be discriminated, eggs, L1 and L2 cannot be differentiated from other nematodes of the family Strongylidae. Identification of L3 is possible for some genera such as Strongylus, Triodontophorus, Gyalocephalus, or Poteriostomum while most others can only be assigned to several cyathostomin larval types (Bevilaqua et al., 1993; Santos et al., 2016). The morphological features include qualitative and quantitative traits such as the number, arrangement and shape of intestinal/midgut cells, the length of the intestine and the length of the sheath tail.

identification methods. Anterior ends and representative MALDI-TOF MS spectra of three cyathostomin species from the closely related species (A) Coronocyclus coronatus, (B) Coronocyclus labiatus, and (C) Coronocyclus labratus are shown. Scale bars represent 100 µm. The x-axes show mass charge ratios while y-axes represent arbitrary intensity units. Spectra were baseline subtracted and smoothed using default parameters in the flexAnalysis software (Bruker Daltonics). Specimen were cleared with lactophenol to improve visibility of structural features of the cuticle. External and internal leaf crown are indicated by black and white arrows, respectively.

## MOLECULAR METHODS

To overcome the limitations of morphological identification, research has focused on molecular cyathostomin identification. These methods, once target-sequences are implemented correctly, can be applied independently of the nematode life-stage.

A target locus which proved to be useful in developing genetic markers for diagnostic and phylogenetic purposes is the ribosomal DNA (rDNA) (reviewed by Gasser and Newton, 2000; Chilton, 2004). Eukaryotic nuclear rDNA is organized in clusters of sometimes several hundred repeats. Coding sequences for 18S, 5.8S, and 28S rRNAs are interrupted by the first and second internal-transcribed spacers (ITS-1 and ITS-2) (Long and Dawid, 1980). The similarity of ITS sequences is higher within than among different species (Elder and Turner, 1995). This was also shown for strongyles, where the extent of intraspecific variation was low (0–0.3%) in comparison to interspecific differences (0.6– 23.7% for the ITS-1 region, 1.3–56.3% for the ITS-2 region) (Hung et al., 1999b).

An early approach for molecular species identification based on the ITS-2 locus is the PCR-linked restriction fragment length polymorphism (PCR-RFLP) analysis, which was first used for differentiation of single eggs of the Strongylinae (Campbell et al., 1995) and was then applied to show that the morphologically very similar Cylicocyclus ashworthi and Cylicocyclus nassatus actually represent separate species (Hung et al., 1997). Another method is the PCR-linked single-strand-conformation-polymorphism technique (SSCP-PCR; Gasser and Monti, 1997), which allows the delineation of 14 strongyle species, including 9 cyathostomins, based on ITS-2 PCR products (Hung et al., 1999a). These methods rely on the DNA from individual worms or eggs and are thus associated with time-consuming procedures if it is desired to screen a representative subset of a strongyle community.

Species identification from mixed parasite DNA from fecal samples and/or copro-cultures was demonstrated after ITS-2 sequences for 28 strongyle species (including 22 cyathostomin species) were determined and species specific primers evaluated for four common species (Hung et al., 1999b). Although, this method theoretically allows species-specific research on pooled samples, it is limited to the identification of only few species.

The variability of the 26S-18S rDNA intergenic-spacer (IGS) was used for species differentiation of 16 cyathostomin species with a range of interspecies variation of 31–56% (Kaye et al., 1998). The obtained sequences were used to develop a PCR-ELISA for the identification of six common cyathostomin species (Hodgkinson et al., 2003) and a Reverse-Line-Blot-Assay (RLB) to simultaneously identify 13 strongyle species (Traversa et al., 2007). Both methods have been used to monitor the species composition before and after anthelmintic treatment (Hodgkinson et al., 2005; Cer ˇ nanská et al., 2009; ˇ Ionita et al., 2010; Traversa et al., 2010). Re-evaluation and validation of existing and new oligo-probes increased the number of species that can be identified with RLB to 18 (Cwiklinski et al., 2012). PCR-ELISA and RLB are qualitative methods detecting the presence or absence of the different species. A semi-quantitative approach applying replicates of pooled larvae was positively evaluated to enable screening of many cyathostomin populations in parallel (Kooyman et al., 2016).

Despite being recognized as "a less suitable target than ITS for quick diagnostic tests," due to its high substitution rates and high possibility of intraspecific polymorphisms, the mitochondrial Cytochrome oxidase c subunit I (COI) is used for species differentiation and could indicate cryptic species (Blouin, 2002). Twenty two COI sequence haplotypes (overall 10.8% rate of intraspecific nucleotide difference) were found within C. nassatus using specimen from different hosts and geographic origins, while only little variation (0.0–0.6% differences) was seen in the ITS-2 sequences suggesting cryptic species within C. nassatus (Traversa et al., 2008) and maybe other cyathostomin morpho-species as well. Analysis of ITS-1 and ITS-2 sequences of Cylicostephanus minutus individuals showed 3.0 and 7.4% differences also indicating the presence of a cryptic species complex (Hung et al., 1999a). The combination of markers on questionable species appears useful to investigate the occurrence of cryptic species complexes.

Whereas the objective of research on cyathostomin species identification is on the one hand to improve the available diagnostic tools, it aims on the other hand to contribute to the understanding of the phylogenetic relationships between the different taxa. Therefore, three gene loci, the ITS-2, COI and 28S rRNA were compared for their phylogenetic usefulness in strongyles. It was encountered that the high level of substitution saturation renders COI unsuitable for phylogenetic analysis. The remaining loci, ITS-2 and 28S rRNA, both showed similar groupings of cyathostomins. Combining both loci resulted in a tree with improved bootstrapping support for the internal nodes (McDonnell et al., 2000) pointing towards the importance of the simultaneous application of different molecular markers. This can also be seen in a study analyzing ITS-1 and ITS-2 sequences of 30 strongyle nematode species, including 23 cyathostomin species that questions the widely accepted separation of Strongylinae and Cyathostominae and proposes a framework to systematically analyze future datasets of strongyle nematodes (Hung et al., 2000). Findings consistent with the latter phylogenetic analysis were shown in a study focusing on the genus Cylicocyclus which proposed a separation of Cylicocyclus in two clades but statistical support for this hypothesis was relatively weak (Bu et al., 2013).

### SEROLOGICAL METHODS

Larval cyathostominosis is caused by the simultaneous reactivation and emergence of high numbers of hypobiotic larvae (Love et al., 1999) causing severe pathology. Usually, no eggs are expelled due to absence of adults making coproscopic diagnosis unfeasible (Murphy and Love, 1997). This leads to the aim of prepatent detection of cyathostomin infections using serology to be able to assess the risk of larval cyathostominosis based on the estimation of the mucosal cyathostomin worm burden.

One promising approach identified anti-larval IgG(T) serum antibody responses to two antigen complexes, only elicited by larvae, as potential markers for prepatent cyathostomin infections (Dowdall et al., 2002). Subsequent purification of native antigenic complexes from larvae resulted in higher IgG(T) signals in an ELISA and reduced the number of false positive responses (Dowdall et al., 2003). Further evidence for an immunodiagnostic potential of these markers is given by a study where the mucosal worm burden of naïve and infected horses was assessed and found to significantly correlate with the IgG(T) serum levels. Additionally, sera from horses with clinical suspicion of larval cyathostominosis had significantly increased antigen-specific IgG(T) levels (Dowdall et al., 2004). One antigenic complex could be identified as cyathostomin gut-associated larval antigen-1 (Cy-GALA-1) and allocated to the species Cyathostomum pateratum (McWilliam et al., 2010), followed by the characterization of the orthologous antigens of four additional common cyathostomin species. An ELISA was developed based on recombinant Cy-GALA proteins, which allows the detection of the immune response to cyatostomin larvae. Cross-reactivity to other parasites was not observed and is unlikely, because of the diversity of orthologous GALA sequences of non-cyathostomin species (Mitchell et al., 2016). In the absence of experimental single species infections, crossreactivity between cyathostomin species is hard to evaluate and diagnostic tests should therefore include a panel of different Cy-GALA proteins to detect most larval cyathostomin infections.

If in future available for routine diagnosis this approach could be of clinical relevance and help ruling out or confirming a larval cyathostominosis in horses with unspecific symptoms of wasting or colic. However, due to the lack of species-specificity serological methods will not help gaining detailed knowledge on the role of individual species in larval cyathostominosis.

### PROTEOMICS METHOD

The proteome based matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS profiling) species identification of microorganisms has already revolutionized diagnostic microbiology. Species identification is based on the molecular masses of proteins such as ribosomal and other abundant proteins. A small amount of microorganisms or crude extracted intact proteins is transferred to specially designed target plates and allowed to co-crystallize with an inert, UV absorbing matrix such as α-Cyano-4-hydroxycinnamic acid. A pulsed 337 nm laser beam irradiates the samples to form a dense ion plume. The resultant ions are accelerated through a vacuum tube to reach the detector and separated according to their charge/mass (m/z) ratio and the time of flight (TOF) for each is measured. The mass range m/z 2,000–20,000 is generally applied for species identification through pattern matching of the spectra peaks with that of a reference spectra database. The method is popular due to its cost-effectiveness, reliability and availability of specially designed linear MALDI machines equipped with software tools and reference databases. This method has been evaluated for a variety of microorganisms such as bacterial, fungal, and viral pathogens (Wieser et al., 2012; Clark et al., 2013). In the past two decades, this technique was established for rapid characterization of eukaryotic cell lines and for species differentiation of protozoan parasites (e.g., Leishmania, Giardia) and arthropods (e.g., mosquitoes, ticks, tsetse flies) (Hoppenheit et al., 2013; Singhal et al., 2016; Yssouf et al., 2016). Regarding nematodes, first diagnostic use of MALDI-TOF has been described to identify different races of Ditylensus dipsaci (Perera et al., 2009) and closely related species of root-knot and seed-gall nematodes (Perera et al., 2005; Ahmad et al., 2012). Despite these developments, extensive studies on the application of MALDI-TOF MS for rapid species identification for helminth have not been reported. Recently, MALDI-TOF MS was applied for rapid species identification of Trichinella spp. after adopting a simple formic-acid/acetonitrile extraction from pooled larvae and compilation of a reference database (Mayer-Scholl et al., 2016). This approach could also be extended to cyathostomin species identification. Preliminary data to evaluate the potential for MALDI-TOF MS for cyathostomins revealed distinct patterns for adult individuals of different species (**Figure 1**).

Of course, master-spectra libraries can only be generated with validated, correctly identified material. This requires that proteomic data are obtained from morphologically and molecularly identified individual specimen. For arthropods, this TABLE 1 | Comparison of methods for cyathostomin species identification.


issue can be solved by using always e.g., a wing or leg for proteomic and any other body part for molecular analysis. For nematodes, which are not segmented, this is not trivial since no defined body parts can be reproducibly cut off at exactly the same position without altering the protein spectrum. Therefore, methods need to be developed that reliably allow to conduct both methods using exactly the same starting material, despite the fact that the protein extraction usually involves conditions that damage DNA. Nevertheless, it was possible to extract DNA of sufficient quality from the acetonitrile/formic acid insoluble material to successfully amplify and sequence the ITS-2 region for the three specimen shown in **Figure 1**. These were 100, 99, and 97% identical to Genbank accession numbers JN786951.2, JN786947.2, JN786949 respectively, which confirmed the morphological identification in each case.

Possible limitations could be the different spectra elicited by different development stages, as seen in tick species identification. However, despite changes of the overall MS protein profiles, the ticks could be classified correctly according to certain specific peaks (Karger et al., 2012). Characterization of species-specific peak patterns, independent of development stages, therefore needs to be part of future research to implement MALDI-TOF MS as a possible diagnostic tool.

### CONCLUSION

Different approaches have been used over the past decades to improve cyathostomin species delineation. All methods have their advantages and limitations (**Table 1**) and none is already fully satisfying for the research questions to be answered and all are far away from applicability in routine laboratory diagnosis. Comprehensive research on different aspects improving the discrimination of individual cyathostomin species, such as inclusion of several molecular markers and additional proteomic profiles could be of great help in the future. This should include the morphological identification together with the description of the genotype (molecular) and phenotype (proteomic) data in association with the currently accepted taxonomic classification. Ideally, morphological, molecular and proteomic data from the same individual should be used to take advantage of all three approaches to identify the complete species spectrum in the Cyathostominae and delineate their phylogenetic relationship.

### AUTHOR CONTRIBUTIONS

CB performed MALDI-TOF experiments and literature surveys. CB and JK drafted and edited the manuscript. TK contributed

### REFERENCES


microphotographs. TK, JM, and GS contributed to writing of the manuscript. CB, JK, JM, and GS designed the general outline.

### FUNDING

The work received financial support from the Deutsche Forschungsgemeinschaft (DFG GRK2046).

assay for the species-specific identification of cyathostomins (Nematoda, Strongylida) from benzimidazole-treated horses in the Slovak Republic. Vet. Parasitol. 160, 171–174. doi: 10.1016/j.vetpar.2008.10.078


Hodgkinson, J. E., Lichtenfels, J. R., Mair, T. S., Cripps, P., Freeman, K. L., Ramsey, Y. H., et al. (2003). A PCR-ELISA for the identification of cyathostomin fourthstage larvae from clinical cases of larval cyathostominosis. Int. J. Parasitol. 33, 1427–1435. doi: 10.1016/S0020-7519(03)00140-1


of ticks by whole-animal matrix-assisted laser desorption/ionization mass spectrometry. Ticks Tick Borne Dis. 3, 78–89. doi: 10.1016/j.ttbdis.2011. 11.002


**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.

The reviewer MN declared a past co-authorship with one of the authors GvS to the handling Editor, who ensured that the process met the standards of a fair and objective review.

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

# The Intestinal Eukaryotic and Bacterial Biome of Spotted Hyenas: The Impact of Social Status and Age on Diversity and Composition

#### Emanuel Heitlinger 1, 2 \* † , Susana C. M. Ferreira3 †, Dagmar Thierer <sup>3</sup> , Heribert Hofer <sup>3</sup> and Marion L. East <sup>3</sup>

<sup>1</sup> Research Group Ecology and Evolution of Molecular Parasite Host Interactions, Leibniz Institute for Zoo and Wildlife Research, Berlin, Germany, <sup>2</sup> Institute for Biology, Molecular Parasitology, Humboldt University, Berlin, Germany, <sup>3</sup> Department of Evolutionary Ecology, Leibniz Institute for Zoo and Wildlife Research, Berlin, Germany

In mammals, two factors likely to affect the diversity and composition of intestinal bacteria (bacterial microbiome) and eukaryotes (eukaryome) are social status and age. In species in which social status determines access to resources, socially dominant animals maintain better immune processes and health status than subordinates. As high species diversity is an index of ecosystem health, the intestinal biome of healthier, socially dominant animals should be more diverse than those of subordinates. Gradual colonization of the juvenile intestine after birth predicts lower intestinal biome diversity in juveniles than adults. We tested these predictions on the effect of: (1) age (juvenile/adult) and (2) social status (low/high) on bacterial microbiome and eukaryome diversity and composition in the spotted hyena (Crocuta crocuta), a highly social, female-dominated carnivore in which social status determines access to resources. We comprehensively screened feces from 35 individually known adult females and 7 juveniles in the Serengeti ecosystem for bacteria and eukaryotes, using a set of 48 different amplicons (4 for bacterial 16S, 44 for eukaryote 18S) in a multi-amplicon sequencing approach. We compared sequence abundances to classical coprological egg or oocyst counts. For all parasite taxa detected in more than six samples, the number of sequence reads significantly predicted the number of eggs or oocysts counted, underscoring the value of an amplicon sequencing approach for quantitative measurements of parasite load. In line with our predictions, our results revealed a significantly less diverse microbiome in juveniles than adults and a significantly higher diversity of eukaryotes in high-ranking than low-ranking animals. We propose that free-ranging wildlife can provide an intriguing model system to assess the adaptive value of intestinal biome diversity for both bacteria and eukaryotes.

Keywords: eukaryotome, eukaryome, parasites, amplicon sequencing, spotted hyena, social status, bacterial microbiome, age classes

#### Edited by:

Susanne Hartmann, Freie Universität Berlin, Germany

#### Reviewed by:

Anastasios D. Tsaousis, University of Kent, United Kingdom Jingwen Wang, Yale University, United States

\*Correspondence:

Emanuel Heitlinger heitlinger@izw-berlin.de; emanuel.heitlinger@hu-berlin.de † Shared first author.

Received: 09 February 2017 Accepted: 02 June 2017 Published: 16 June 2017

### Citation:

Heitlinger E, Ferreira SCM, Thierer D, Hofer H and East ML (2017) The Intestinal Eukaryotic and Bacterial Biome of Spotted Hyenas: The Impact of Social Status and Age on Diversity and Composition. Front. Cell. Infect. Microbiol. 7:262. doi: 10.3389/fcimb.2017.00262

**197**

## INTRODUCTION

Mammalian hosts have a long evolutionary history with the diverse communities of prokaryotes (Dethlefsen et al., 2007; Ley et al., 2008; Douglas and Werren, 2016) and eukaryotes (Hafner and Nadler, 1988; Glendinning et al., 2014), here designated as the intestinal biome, present in their gastrointestinal tract. Although the intestinal biome was traditionally viewed as a rather inert species assembly, recent insights have revealed that its composition is influenced by many host traits and that it in turn can have a considerable impact on its host (Turnbaugh et al., 2007; Graham, 2008; Costello et al., 2012; Wegner Parfrey et al., 2014; Kreisinger et al., 2015). Recently, interest in the community of bacteria in the intestine (often called the microbiome, even though bacteria do not comprise all microscopic organisms in the intestines) has grown (Round and Mazmanian, 2009; Turnbaugh et al., 2009; Lozupone et al., 2012), whereas the communities of unicellular (Wegner Parfrey et al., 2014) and especially multicellular eukaryotes in the gastrointestinal tract (eukaryome or eukaryotome) have received less attention.

Following birth, the mammalian gastrointestinal tract is gradually colonized by prokaryotes and eukaryotes (Palmer et al., 2007; Koenig et al., 2011; Lozupone et al., 2012; Yatsunenko et al., 2012). Throughout an individual's lifespan, the composition of the intestinal biome can be altered by various factors, including diet (Turnbaugh et al., 2009; Amato et al., 2013), as exemplified by the transition from the consumption of milk during infancy (Pond, 1977) to an "adult" diet when weaned. Composition and diversity is also influenced by interactions between species within the intestinal biome (Ezenwa, 2004; Benson et al., 2010; Glendinning et al., 2014; Lee et al., 2014; Kreisinger et al., 2015), developmental changes in immune function (Dowling and Levy, 2014) and age related changes in exposure to prokaryotes and eukaryotes in the environment, food and conspecifics (Palmer et al., 2007; Koenig et al., 2011; Lozupone et al., 2012). Despite these factors that induce variation in the intestinal biome among individuals within a species, there is evidence of species-specific bacterial microbiome signatures in mammals (Ley et al., 2008; Ochman et al., 2010; Yildirim et al., 2010; Degnan et al., 2012; Menke et al., 2014).

While the majority of studies on the diversity and composition of bacterial microbiomes are based on partial sequencing of 16S ribosomal RNA genes, the application of metagenomics (the sequencing of the complete genetic repertoire of the host's intestinal microbiome) has revealed links between the bacterial microbiome of the intestine and the host's metabolism (Gill et al., 2006; The Human Microbiome Consortium, 2012). The bacterial microbiome influences host nutrition, fat storage and the metabolism of vitamins and minerals (Turnbaugh et al., 2009; Tremaroli and Bäckhed, 2012; Leone et al., 2015). Experimental studies with germ-free mice also show that the bacterial microbiome of the intestine can affect postnatal development of the hypothalamic-pituitary-adrenal axis (Sudo et al., 2004). Furthermore, bacterially derived fermentation products from the intestinal microbiome help regulate the maturation of microglia which contribute to an active immune defense of the central nervous system (Erny et al., 2015) and there is now general agreement that the bacterial microbiome influences the host immune system (Round and Mazmanian, 2009; Hooper et al., 2012; Sivan et al., 2015).

The impact of intestinal eukaryotes on their hosts has mostly been studied in relation to the pathologies caused by specific helminths and protozoans (Wegener Parfrey et al., 2011; Andersen et al., 2013; Rajilic-Stojanovi ´ c and de Vos, ´ 2014). We are not aware of any studies on intestinal biomes comprehensively screening for eukaryotes, including multicellular species. This is surprising, as it has been argued that eukaryotes may have an important ecological function in intestinal ecosystems similar to that of keystone species in terrestrial or aquatic ecosystems (Lukeš et al., 2015). By extension this should particularly apply to large multicellular eukaryotes. Currently, most research on the intestinal biomes of mammals has focused on the bacterial microbiome of humans, wild and captive non-human primates, laboratory mice and zoo animals; hence knowledge on the intestinal biomes of free-ranging wild mammals is limited.

In social mammals, social status often determines access to resources (Clutton-Brock and Huchard, 2013) thereby affecting health status (Sapolsky, 2005), wound healing (Archie et al., 2012), immune gene expression (Tung et al., 2012), immune defenses (Flies et al., 2016; Snyder-Mackler et al., 2016) and the likelihood and impact of pathogen infection (East et al., 2001, 2015; Höner et al., 2012). Social status is also likely to affect the species composition and abundance of bacteria and eukaryotes in the gastrointestinal tract.

We present, to our knowledge, the first study to simultaneously investigate both the bacterial microbiome and eukaryome of a wild mammalian species. We applied a multi-amplicon sequencing approach for bacteria and eukaryotes (metabarcoding) to fresh feces to investigate the effect of social status and age on the (distal) intestinal bacterial microbiome and eukaryome of individually known spotted hyenas (Crocuta crocuta) in the Serengeti National Park (NP) in northwestern Tanzania. This highly social carnivore lives in fission-fusion groups termed clans, in which natal females and their offspring are socially dominant over immigrant males (Kruuk, 1972). Migratory movements of ungulates cause large fluctuations in the abundance of food resources in clan territories (Hofer and East, 1993a,b), which profoundly affect the foraging behavior of clan members (Hofer and East, 1993a,b). When migratory ungulates are absent, clan members undertake regular long-distance (approximately 80–140 km) foraging trips (termed commuting trips) to areas containing abundant migratory prey (Hofer and East, 1993b). As access to food within a clan's territory is determined by social status, high-ranking females commute far less often than low-ranking females (Hofer and East, 2003) and their reduced foraging effort is reflected in their lower fecal glucocorticoid metabolite (fGCM) concentrations (Goymann et al., 2001b).

High-ranking females are more often exposed to infectious pathogens than low-ranking females because of their more frequent social interactions with clan members (East et al., 2001). Even so, high-ranking females have lower intestinal parasite burdens than low-ranking females, probably because they can allocate more resources to immune processes than low-ranking females (East et al., 2015). Hence, we expect the bacterial microbiome and the eukaryome of high-ranking animals to be more diverse than those of low-ranking animals, because high species diversity is generally considered an index of ecosystem health (Cardinale et al., 2006; Costello et al., 2012; Reich et al., 2012). In contrast, we expect the intestinal biome of low-ranking animals to be less diverse than those of high-ranking animals.

As in all mammals, the intestinal bacterial microbiome and the eukaryome of juvenile spotted hyenas develops after birth. In our study population, juveniles are more often infected with specific pathogens than adults (Goller et al., 2013; Nikolin et al., 2017), including the eukaryote Dipylidium sp. (East et al., 2013). Moreover, juveniles have lower protection from antibodies (East et al., 2001) and their diet includes maternal milk until 12–18 months of age (Hofer and East, 1995). For these reasons we expect the bacterial microbiome and eukaryome of juveniles to be less diverse than that of adults.

We tested our predictions concerning the effects of age and social status by assessing sequence read counts of taxonomically annotated ribosomal sequence variants (RSV) and hence the composition and abundance of genera in the bacterial microbiome and the eukaryome of individually known hosts. We also compared these results with those generated by the classical coprological method of parasite egg or oocyst counts as applied, for instance, by East et al. (2015) to assess whether the results of these methods were strongly correlated.

### METHODS

### Study Population

The spotted hyena (hereafter hyena) study population included three closely monitored clans that are part of an ongoing longterm research program in the center of Serengeti NP, northwest Tanzania. Individuals were recognized by their spot patterns, ear notches, scars and bald patches (Frank, 1986; Hofer and East, 1993a) and sexed using the dimorphic shape of the phallic gland (Frank et al., 1990). Age was determined from the observed date of birth or based on observations of pelage, position of the ears, level of coordination when walking and body size, with an accuracy of ±7 days as previously described (East et al., 2003). Animals were categorized as juveniles when <24 months of age, and adult when ≥24 months of age. Females were allocated a social rank within the dominance hierarchy using submissive responses in dyadic interactions (East et al., 2003). To compare rank positions across clans, individuals were assigned a standardized rank within a dominance hierarchy by distributing ranks evenly between the highest rank (standardized rank +1) and the lowest rank (standardized rank −1), with the median rank being scored as 0 (Goymann et al., 2001a). Females holding standardized ranks with 0 or above 0 were categorized as highranking, those holding a standardized rank of less than 0 as low ranking.

### Sampling

Forty-two fecal samples (35 adult females and 7 juveniles) were collected immediately after defecation from individually known animals between 2009 and 2012. Samples were thoroughly mixed and aliquots were stored in formalin (4%) for parasite egg counts and preserved in RNAlater (Sigma–Aldrich, St Louis, MO, USA) for molecular genetic analyses. Samples in RNAlater were initially stored frozen at −10◦C, transported frozen and then stored at −80◦C for genetic analysis (East et al., 2013). DNA was extracted using the Macherey-Nagel Nucleo-spin soil DNA extraction kit (Macherey-Nagel, Düren, Germany) following the manufacturer's recommendations and using the Peqlab Precellys 24 homogenisator (VWR International Group, Erlangen, Germany). We assume that fecal biomes are representative of intestinal biomes, since a strong relationship between the two was demonstrated for freshly collected (Menke et al., 2015) and properly stored samples (Menke et al., 2017).

### Parasite Egg Counts

Parasite egg or oocyst counts were conducted on aliquots of a subset of 32 fecal samples (27 adult females and 5 juveniles, 20 high-ranking and 12 low-ranking individuals) that were suitable for this procedure using a modification of the McMaster flotation technique (Gordon and Whitlock, 1939). To enhance the detectability of eggs, a solution of potassium iodide (KI) was used with a specific weight of 1.5 g ml−<sup>1</sup> (Meyer-Lucht and Sommer, 2005; Schwensow et al., 2007). Four McMaster chambers were counted for each sample with a dilution factor of 1:15. After combining the feces with the KI solution, it was vortexed for 3 min and then sieved in order to remove bigger debris. Parasite eggs or oocysts were identified according to their morphology and counted using a light microscope, 1 h after preparing the fecal suspension, with a magnification of 100x (10x eyepiece lens × 10x objective lens). Pictures were captured using the software ProgRes CapturePro version 2.5, 2007 (Jenoptik, Jena, Germany). During fecal egg or oocyst counts, eukaryote parasites were identified at the genus or family level on the basis of their morphology and size, with the exception of oocysts from the order Coccidia because they are very similar in terms of their morphological appearance. For Ancylostoma, the Taeniidae and the Coccidia two morphological types based on two size classes of eggs or oocysts were distinguished. For the nematode Ancylostoma, the two identified size classes for eggs were <80 µm and ≥80 µm, for the cestodes from Taeniidae, the two size classes were < 45 µm and ≥45 µm, and for the Coccidia oocysts, the two size classes were <20 µm and ≥20 µm, respectively. The results are expressed as fecal egg counts per g feces (FEC) or fecal oocyst counts per g feces (FOC). All egg and oocyst counts were done blind with respect to the life history stages and characteristics (age, social status) of the individual hyenas from which the fecal sample was taken and analyzed.

### Multi-Amplicon PCRs and Sequencing

The Fluidigm Access Array integrated fluidic circuit (Fluidigm, San Francisco, California, USA) was used to run 48 × 48 PCR reactions in 2,306 compartments of a microfluidics device. Target specific PCR primers (Supplementary File 1) for 18S and 16S small ribosomal subunits (SSU) were used in a "four primer" PCR approach, following the manual provided by Fluidigm. Briefly, target specific primer pairs were combined with "CS1" and "CS2" adapters at their 5′ and 3′ ends, respectively, on the microfluidics device. These target specific primer pairs were used to prime 48 target specific reactions for each of 48 samples, using default cycling parameters. After harvesting all products from the separate samples into a 96 well microliter plate, a second PCR was performed on a 10-fold dilution, by introducing Illumina sequencing oligonucleotides "PE5" at the "CS1" adapter and "PE7" at the "CS2" adapter as well as a sample identifier sequence between "CS2" and "PE7" based on the Access Array Barcode Library (Fluidigm, San Francisco, California, USA) for Illumina Sequencers (384 single direction). Samples were pooled and selected by size using Agencourt AMPure XP Reagent beads (Beckman Coulter Life Sciences, Krefeld, Germany). For further cleanup, PCR fragments between 400 and 1,000 bp were purified by PippinPrep using the 1.5% agarose DNA gel cassettes (Sage Science Inc., Beverly, Massachusetts, USA). Suitability of PCR products for sequencing (e.g., by checking for an absence of primer multi-meres) was confirmed using the Agilent 2200 Tape Station with D1000 ScreenTapes and D1000 Reagents (Agilent Technologies, Santa Clara, California, USA). Sequences were generated at the Berlin Center for Genomics in Biodiversity Research (BeGenDiv) on the Illumina MiSeq machine (Illumina, San Diego, California, USA) using version 3 chemistry and 600 cycles of (paired-end) sequencing. The sequencing data can be accessed through the accession number PRJNA386767 at NCBI Short Read Archive (SRA).

### Bioinformatic Analyses

Sorting of sequencing reads in different samples was performed using the bcl2fastq utility version 2.17.1.14 (Illumina, San Diego, California, USA) based on the sample identifier oligos. All subsequent bioinformatic, taxonomic and statistical analyses were performed in R version 3.3.2 (R Development Core Team, 2016); below we cite the R packages used for specific steps in the analysis.

Sequences were quality trimmed and screened for erroneous reads using the fastqPairedFilter function of package dada2 version 1.2.1 (Callahan et al., 2016) with parameter settings of truncLen=c(170,170), maxN=0, maxEE=2, truncQ=2. Further stratification of the full "samples by amplicon" matrix was performed using package MultiAmplicon version 0.1 (Heitlinger, 2017): primer sequences were trimmed in read pairs matching with zero mismatches starting at position one in both forward and reverse reads. Sequencing reads were sorted into an amplicon when they contained the sequences of a specific pair of primers. The MultiAmplicon package was also used as a wrapper to process identified amplicons with the dada2 workflow. Briefly, sequences were dereplicated, RSVs were inferred using the function dadaMulti (with options err=NULL, selfConsist=TRUE), forward and reverse reads were concatenated (using function mergeMulti, option justConcatenate=TRUE), a table of RSV occurrence was collated for each sample and sequences likely to be chimeric and introduced during PCR were screened and discarded.

Technical replicates for which PCRs failed were identified by hierarchical clustering of primer-stratified read numbers, with water as negative controls, and marked for exclusion. For each sample, RSV counts were obtained as the sum of the remaining technical replicates and normalized for sequencing depth using a simple scaling by the median of the per sample RSV counts. All taxonomic assignments were done blind with respect to the life history stages and characteristics (age, social status) of the individual hyenas from which the fecal sample was taken and analyzed.

### Taxonomic and Statistical Analyses

Taxonomy was inferred for RSVs using a ribosomal database project naïve Bayesian classifier (Wang et al., 2007) through dada2's "assignTaxonomy" function. As a training data set for the classifier, the SILVA\_123 database (Quast et al., 2013) was expanded with highest scoring unique BLAST hits (blastn; Altschul et al., 1997) of our RSVs in the NCBI nt database. This training dataset was thus curated with a focus on eukaryotic 18S sequences in our study system and then used in an assignment of the taxonomy levels "phylum," "class," "order," "family," and "genus" to each of our RSVs. Problems arose for annotation of putative Eimeria reads (among others) because of an incongruence between the taxonomic systems of SILVA and NCBI, so two separate sets of results will be described for this genus.

RSV counts, annotation with taxonomy information and hyena specific sample data were combined into a single R object for all amplicons using the package phyloseq version 1.18.0 (McMurdie and Holmes, 2013). Data were merged across different amplicons via their taxonomic annotation at the genus level using the function "tax\_glom" with the parameter "NArm=TRUE." This excludes genera not annotated at the genus level (or annotated with uninformative terms such as "undefined"). We also excluded genera annotated as the only genus in their respective phylum and genera with an "undefined" phylum annotation. We report estimates of RSV diversity derived from different amplicons with caution, as we recognize that many raw RSVs prior to taxonomic annotation represent different parts of the same marker genes and hence our diversity estimates are inflated for this measure. Even so, they can be used to compare the diversity between individuals, because all samples were processed using the same amplification and bioinformatics workflow.

We used the strength of association between "sequence abundance" (number of ribosomal sequence reads annotated for specific genera or higher level taxa) and FEC or FOC to screen for the most likely sequenced taxon in samples for which we had morphological evidence from the egg or oocyst counts. Because the larger size class of Ancylostoma occurred in only two samples, FEC for both size classes were added together for the correlation of egg counts with sequence abundance data. Spearman rank correlation coefficients were calculated in R base and recorded not only for "target taxa" within the taxonomic scope of morphological discrimination but in a "blinded" approach for the comprehensive set of all sequence counts at all levels of taxonomy. The strongest correlations were then screened for taxonomic agreement with FEC and FOC. The four highest positive correlations always contained the target taxa as defined by the observed morphology. In the results (below) we report p-values to test for significant association between the number of ribosomal sequence reads and FEC or FOC for the highest or second highest Spearman correlation coefficients. We also constructed linear models using (1+log10) transformed data to visualize the linearity of the association and report the regression equations as a predictive tool.

Richness—the number of taxa (RSVs or genera) present —, evenness—the evenness in the distribution of sequencing read numbers for different taxa—and diversity—an index that takes into account both richness and evenness—(see Legendre and Legendre, 2012) were calculated after the random selection of a data subset (rarefaction) from all sample counts at the sequencing depth of the library sequenced to the lowest depth. We used rarefaction instead of normalization for diversity estimates to avoid problems of overestimating presence in more deeply sequenced samples. We used the package phyloseq with its function estimate\_diversity to estimate measures of diversity and species richness with the help of package vegan version 2.4-1. We calculated the number of genera as "observed richness," the Chao1 index of diversity (Chao, 1984) as a measure of species diversity and Pielou's J as a measure of evenness (Pielou, 1975). Number of genera, Chao1 index and Pielou's J were compared between juveniles and adults or high-ranking and low-ranking hyenas using the exact version of the Mann-Whitney U test in the package coin version 1.1-2 (Hothorn et al., 2006) to obtain appropriate p-values for small sample sizes and in the presence of ties.

In order to compare the composition of the bacterial microbiome and the eukaryome between individuals, life history stages and social status categories, the diversity of the bacterial microbiomes and eukaryomes and its underlying variation across individuals was efficiently summarized by multidimensional scaling (MDS), a non-parametric ordination technique. This technique has previously been highly successful in summarizing similarly complex data sets (e.g., Burgener et al., 2009), along new dimensions called here MDS axes. MDS uses pairwise Bray-Curtis dissimilarities (Bray and Curtis, 1957) on (1+log10) transformed data of sequence abundances per genus. On the same transformed data partial least squares (PLS) models (Hastie et al., 2013), as calculated by package caret version 6.0-73, were used as a supervised machine learning technique (Wold et al., 1984) to predict age and rank category for the individual from which the sample was collected. PLS models also produce a set of axes or "directions" (Hastie et al., 2013) called here PLS axes. The optimal number of PLS axes retained in the final model was determined using leave-one-out cross-validation. Each sample received a PLS score on each PLS axis, which documents how well sample categories can be differentiated on that axis, and each taxonomic unit as a "predictor" received a PLS "loading" on each axis which documents to what extent the taxonomic unit contributed to the PLS axis. We subsequently used Fisher's exact test from R base to test for overrepresentation or underrepresentation of phyla visually identified to be abundant at the extremes of the distribution of loadings. We tested highest and lowest quartile of loadings on the single PLS axis of the model addressing rank category differences in the eukaryome. As the PLS model for differences of the microbiome in age categories contained more than one axis, the PLS loadings of the first two axes separating the samples were combined by multiplication before this procedure.

Package DESeq2 version 1.14.0 (Love et al., 2014) (function "DESeq") was used to test for differences in the abundance of individual genera between age and rank categories. In contrast to normalization or rarefaction used in other analyses the function estimated "size factors" to address differences in sequencing depth between different samples. These factors were then used as offset when a mean abundance was fitted for each taxon in generalized linear models (glm) using a negative binomial distribution and a dispersion parameter specific to that taxon. Maximum likelihood estimates for glm coefficients were obtained and likelihood ratio tests were conducted by subtracting the log-likelihood of the full model including different estimates for a focal contrast (age or rank category) from a reduced model without this difference. The resulting likelihood-ratio was compared then to a χ 2 -distribution. Resulting p-values were corrected for multiple testing using the Benjamini and Hochberg method (Benjamini and Hochberg, 1995) and expressed as false discovery rates, with the significance threshold set to 0.05.

## RESULTS

High throughput sequencing of multiple amplicons provided a comprehensive survey of the intestinal biome of hyenas, comprising 986 genera, most of which belonged to the eukaryotic biome, as described in detail below.

### Sequencing Based Assessment

We obtained a total of 3,195,831 sequencing read-pairs from amplicon sequencing for 18S small ribosomal subunits of eukaryotes and 16S ribosomal subunits of bacteria. Variant inference on a single base resolution level for these sequences revealed a total of 24,604 RSVs, which were summarized via shared taxonomic assignments at the genus level.

### Taxonomic Diversity of Hyena Intestinal Biomes

We identified 201 genera (3,725 RSVs) of bacteria as constituents of the bacterial microbiome of the hyena. We also identified 656 genera (20,879 RSVs) of eukaryotes in the same fecal samples. The number of genera annotated varied greatly between phyla of both bacteria and eukaryotes (**Figure 1A**, **Table 1**). High genus level diversity was observed in the bacterial phylum Firmicutes and in the eukaryote phyla Ascomycota, Chlorophyta and Basidiomycota. The classic phyla of intestinal parasites— Nematoda, Apicomplexa and Platyhelminthes—showed an intermediate number of genera but a high number of sequencing reads. Phylum Chordata was represented with few genera and a large number of sequencing reads (**Figure 1A**, **Table 1**).

The eukaryote sequences derived from 18S amplification did not solely originate from organisms in the eukaryome but also included organisms that were part of the hyena's diet or those that were accidentally ingested. Taxonomic annotation of eukaryote derived DNA can help to assess these potential sources. For some taxa a role as intestinal inhabitants or food items can safely be assigned. We therefore categorized eukaryote phyla in

our taxonomic annotation in categories of likely (1) eukaryome (including parasites, commensals and mutualists), (2) food items, (3) passing material, and (4) undetermined role (**Table 1**). We consider for example Nematoda, Platyhelminthes, Apicomplexa and Microsporidia as classical eukaryotic parasites and hence members of the eukaryome. Other organisms not conventionally considered to be parasites but rather commensals or with an unknown effect on their host comprised most phyla of fungi. DNA sequences from the phylum Chordata were most likely originating from hyena food items (**Table 1**).

### Sequence Based Abundance Counts Correlate with Fecal Egg or Oocyst Counts but Are More Sensitive

Eukaryote parasites were identified at a genus level for the genus Ancylostoma (detected in 25 of 32 samples), more conservatively at the family level for cestodes in families Diphyllobothriidae (detected in 26 of 32 samples) and Taeniidae (detected in 6 of 32 samples) and protists only at the level of order for Coccidia (detected in 17 of 32 samples). We also detected in one or two TABLE 1 | The diversity of genera and phyla of bacteria and eukaryotes recorded from the intestinal biome of the spotted hyena as extracted by amplicon sequencing of fecal samples.


\*The first number is for the annotated dataset analyzed for genera; the second number is the full dataset analyzed for RSVs.

\*\*Number of genera not considered correctly annotated based on phylum level abundance (see methods) or annotated as "undefined" at the phylum level.

samples species from Spirurida, Trichuris spp. and Dipylidium spp. (**Table 2**).

The amplicon sequencing based abundance estimates correlated significantly and positively with egg or oocyst counts, as detailed below, for those species for which we had more than six positive FEC or FOC samples. These correlations also helped to assign plausible taxonomic status to taxa with morphologically indistinguishable eggs or oocysts. Ancylostoma FEC correlated best with the sequence counts for the genus Ancylostoma among all reported genera (Spearman's rho, ρ = 0.54, n = 32, p = 0.002). There was a slightly better positive correlation with counts summarized for the order Rhabditida, to which Ancylostoma belongs (ρ = 0.58, n = 32, p < 0.001): seven samples with relatively high sequencing counts (range 84–4,035) and zero FECs were recovered and all samples reporting FEC in Rhabditida had at least 407 sequences counted for Ancylostoma. Within this order, annotations for 23 other genera were reported. The genus Ancylostoma contributed most (59%) sequencing reads annotated within the order. The genera Ostertagia (20%) and Haemonchus (12%) also contributed substantial numbers of sequencing reads annotated as Rhabditida. Adding together reads annotated as Ancylostoma and Haemonchus resulted in



Indicated are identified taxa, their prevalence in %, the mean and median with respective 95% confidence intervals. Table was sorted by prevalence. "spp.": an unknown number of species which may or may not belong to more than one genus.

a correlation slightly stronger (ρ = 0.59, n = 32, p < 0.001) than that observed for the whole order Rhabditida. When reads classified as Ostertagia were added to the reads annotated as Ancylostoma, the correlation with Ancylostoma spp. in the FEC became slightly weaker (ρ = 0.54, n = 32, p = 0.002) (**Figure 2A**).

Similarly, FEC for taxa identified as originating from one or several species in the family Diphyllobothriidae correlated best with annotated RSV counts for the family Diphyllobothriidae (ρ = 0.69, n = 32, p < 0.001). The genus Diphyllobothrium provided the vast majority (98%) of counts annotated in the family Diphyllobothriidae and reads for the genus Spirometra provided the remaining (2%) of counts in this family. The correlation of FEC with annotated RSV counts for Spirometra alone was slightly weaker (ρ = 0.67, n = 32, p < 0.001) than for the entire family. The FEC results for four samples contained no Diphyllobothrium eggs but produced positive Diphyllobothrium sequence results, with annotated RSV counts ranging from 9 to 39,699; all samples with FEC above zero produced annotated RSV counts in the range of 79–44,993 for Diphyllobothriidae. **Figure 2B** visualizes this relationship.

When annotated RSV counts for the genera Besnoitia, Toxoplasma, Isospora, and Eimeria were added together, their total number correlated best (ρ = 0.79, n = 32, p < 0.001) with FOC for the small size class (<20 µm) of Coccidia oocysts. This comparison produced a linear model on log transformed data with an excellent fit (r <sup>2</sup> = 0.92; **Figure 2C**). FOC were zero for three samples with sequencing counts ranging from four to five for these genera. The number of RSV counts annotated as Eimeria correlated best (ρ = 0.60, n = 32, p < 0.001) with the large size class (≥20 µm) of Coccidia oocyst counts (**Figure 2D**). This relationship did not follow the pattern reported for other correlations, since substantial numbers of FOC were reported for samples with zero abundance in terms of annotated RSV reads.

### Adult Female Hyenas Have a Bacterial Microbiome Which Is More Diverse Than and Differs in Composition from That of Juveniles

Consistent with our prediction, the bacterial microbiome of juvenile hyenas contained a significantly lower number of genera (lower richness; Mann-Whitney U-test, U = 184.5, n = 42, p = 0.012), had a lower diversity (U = 174.5, n = 42, p = 0.034) and showed a trend toward lower evenness (U = 168, n = 42, p = 0.063) than that of adults (**Figure 3A**). The bacterial microbiome of adult hyenas contained a median of 49 bacterial genera whereas those of juveniles hosted a median of 41 genera.

A more detailed analysis for each bacterial phylum revealed that the higher richness and diversity of genera in adults than juveniles predominantly occurred in the phyla Tenericutes (U = 203, n = 42, p < 0.001) and Bacteroidetes (U = 173, n = 42, p = 0.037). Microbiomes of adults had a significantly lower diversity for Actinobacteria than those of juveniles (U = 53, n = 42, p = 0.021). The differences in composition between the bacterial microbiomes of juveniles and adults were confirmed using multidimensional scaling ordination, which showed that juvenile microbiomes are less uniform in composition between individuals than those of adults (**Figure 3B**).

The distinct composition of adult and juvenile bacterial microbiomes was underlined by the results of a PLS regression which correctly assigned samples to age categories with an accuracy of 93% in leave-one-out cross evaluations. The optimal model retained three PLS axes, the first two of which are plotted with PLS scores of samples and loadings of bacteria phyla in **Figure 3C**, illustrating how well age categories were separated by their PLS scores. A more detailed analysis demonstrated that genera in the phyla Tenericutes tended to be (odds ratio = 4.83, Fisher test, p = 0.069)

specific linear model on (1+log10) transformed data, R<sup>2</sup> as a measure of goodness of fit, and a line representing the predicted relationship. The panels additionally include a representative micrograph depicting the egg or oocyst counted.

and Bacteroidetes were (odds ratio = 5.42, Fisher test, p < 0.001) characteristic for adult microbiomes, Actinobacteria for juvenile microbiomes (odds ratio = 2.93, Fisher test, p = 0.009). Testing individual genera of bacteria for differences in abundance between age categories resulted in six genera with significant false discovery rates of <0.05, with three more abundant in juveniles and three more abundant in adults (**Figure 3D**).

In contrast to the bacterial microbiome, we detected no significant differences in eukaryome richness (U = 106, n = 42, p = 0.75), diversity (U = 96, n = 42, p = 0.50), evenness (U = 128, n = 42, p = 0.68) or genus abundance between hyena age categories (false discovery rate for all single genera glms >0.05).

### A More Diverse Eukaryome in High-Ranking Than Low-Ranking Hyenas

Estimates of observed richness (U = 293, n = 42, p = 0.004) and diversity (U = 288, n = 42, p = 0.007) in terms of RSVs in the eukaryome were significantly higher in high-ranking than low-ranking hyenas (**Figure 4A**). The same was true for inferred genera (richness, U = 299, n = 42, p = 0.002; diversity,

FIGURE 3 | Bacterial genera richness, diversity and microbiome composition in different age categories. (A) Box plots depicting distributions of richness (observed counts of genera richness per phylum), diversity (Chao1 index) and evenness (Pielou's J) estimates on rarefied (see methods) genera counts for juveniles and adults. \*Significant differences (p < 0.05) based on exact Mann-Whitney U tests. (B) Non-metric multidimensional scaling (MDS) ordination based on pairwise Bray-Curtis dissimilarities partially separated juvenile from adult samples based on different compositions of taxonomic units. (C) A comparison of PLS scores (for samples) and PLS loadings (for genera) from the first two PLS axes of an optimized partial least squares model, demonstrating a clear separation of adult and juvenile samples. Genera colored by phylum can be used to assess the taxa contributing (PLS loading) to the differences underlying this distinction. (D) Log2-fold change inferred by generalized linear models testing for differences between adults and juveniles for each genus with a false discovery rate (adjusted p-value) of < 0.05. The numerical value of the false discovery rate is given below the dot for each genus color-coded for its respective phylum.

U = 299.5, n = 42, p = 0.002; **Figure 4B**). Social status had no significant effect on eukaryome evenness. A significantly higher number of genera (richness) occurred in high-ranking than lowranking individuals for the phyla Basidiomycota (U = 275, n = 42, p = 0.021), Ascomycota (U = 272.5, n = 42, p = 0.025) and Blastocladiomycota (U = 250, n = 42, p = 0.049). There was a trend toward higher diversity in high-ranking than low-ranking hyenas for Apicomplexa (U = 254.5, n = 42, p = 0.079).

It was not possible to determine compositional differences between eukaryomes of high-ranking and low-ranking animals in our dataset using unsupervised ordination techniques. PLS regression models, however, were able to assign animals to social status categories with 71% accuracy in leave-one-out cross evaluations. Most high-ranking individuals were clearly separated from low-ranking individuals by the single PLS axis retained in the model, the overlap with low-ranking individuals being confined to a minority of high-ranking individuals (**Figure 4C**). The loading of several genera belonging to the phylum Basidiomycota tended to increase the PLS scores for high-ranking individuals (odds ratio = 1.55, Fisher test, p = 0.074). There was also a trend that genera in the Apicomplexa through their loadings increased the PLS scores for low-ranking individuals (odds ratio = 2.01, Fisher test, p = 0.057). Both trends contributed to the separation of high-ranking and low-ranking individuals. Interestingly, different genera in the Apicomplexa were drivers in terms of their PLS loadings on the single PLS axis for the classification of samples as belonging to both highranking and low-ranking individuals (odds ratio = 3.17, p = 0.011; testing the combined upper and lower quartiles of loadings).

In contrast to the eukaryome, we detected no significant differences in bacterial microbiome richness (U = 184.5 n = 42, p = 0.84), diversity (U = 182.5, n = 42, p = 0.80), evenness (U = 184, n = 42, p = 0.84) or genus abundance between high-ranking and low-ranking animals (false discovery rate for all individual genera glms >0.05).

### DISCUSSION

Amplicon sequencing (metabarcoding) revealed that spotted hyenas in the Serengeti NP had a diverse intestinal biome, with at least 201 identified genera of bacteria and 656 identified genera of eukaryotes (**Table 1**, **Figure 1A**). The bacterial phylum with the highest diversity of identified genera was the Firmicutes, as were the Ascomycota, Chlorophyta and Basidiomycota among the eukaryotes (**Figure 1B**). Positive, significant correlations between amplicon sequence abundance estimates and FEC or FOC (**Figure 2**) suggest that results from amplicon sequencing are quantitatively valid and highly sensitive measures, and also can aid the differentiation of taxa with similar egg and oocyst morphologies. Consistent with our predictions, adult females had a significantly more diverse intestinal bacterial microbiome than juveniles (**Figure 3**) and the diversity of eukaryotes was significantly greater in high-ranking than low-ranking animals (**Figure 4**). However, contrary to our prediction, there was no effect of social status on the diversity of the bacterial microbiome and also no difference between the diversity of eukaryotes in adult and juvenile intestinal biomes.

Studies of the intestinal bacterial microbiome in humans report an increase in the number of taxa and taxonomic diversity of bacteria with age (Palmer et al., 2007; Tannock, 2007; Koenig et al., 2011). Our results were similar in that the distal intestinal bacterial microbiome of adult hyenas was significantly more diverse and differed in composition from that in juvenile hyenas. Probably both social and physiological factors contributed to these age related differences. During approximately the first 12 months of life, juveniles remain at the clan's communal den (Hofer and East, 1993c) and for most daylight hours they rest together in underground burrows where conditions are probably conducive to the spread of bacteria among resting juveniles (Höner et al., 2012). Dens also function as important social centers where clan members meet and interact, thereby enhancing the transmission of hyenaassociated pathogens between clan members (East et al., 2013; Olarte-Castillo et al., 2016). Greeting ceremonies (East et al., 1993), in which participants stand head-to-tail and sniff and lick each other's anogenital area, are probably important for the spread of bacteria and their inclusion in the intestinal biome by the fecal-oral route. Since juveniles frequently participate in greeting ceremonies (East et al., 1993), greetings probably aid the colonization of the juvenile intestinal biome by both bacteria and eukaryotes. In line with this idea there is growing evidence that socially mediated transmission contributes to the maintenance of a diverse bacterial microbiome in a taxonomically broad range of species, including chimpanzees (Pan troglodytes), in which social processes help maintain a diversity rich in commensals and mutualists (Moeller et al., 2016), and social insects where social transmission of bacteria provides protection against virulent pathogens (Koch and Schmid-Hempel, 2011). Throughout the long period in which they are nursed, juvenile hyenas receive bacteria from their mother, and the nutritional and immunological components of milk (Hofer et al., 2016) may also affect the composition of the intestinal bacterial microbiome of juvenile hyenas.

As high-ranking individuals are highly sought after social partners, we predicted a socially mediated spread of a more diverse bacterial microbiome in high-ranking than low-ranking individuals but our results did not show this. One possible explanation is that our resolution, to the level of RSVs and genera, was insufficient. Our approach might have failed to distinguish functional differences in the relationship (e.g., mutualistic vs. pathogenic) of variants of the same species with their hosts, as in the case of distinct variants of Escherichia coli (von Mentzer et al., 2014). Hence our taxonomic resolution even at the RSV level—was possibly insufficient for a fine-grained analysis of the diversity of the bacterial microbiome and the functional relationships of variants to their host. An alternative interpretation of our results is that high rates of contact between adult female clan members homogenize their intestinal bacterial biomes, as proposed by ecological network theory (Wilson, 1992).

In contrast to the bacterial microbiome, intestinal eukaryotes are—with the exception of some fungi—traditionally considered parasites and thus detrimental to their host. However, extending

tests. (C) A comparison of PLS scores (for samples) and PLS loadings (for genera) are visualized on the single PLS axis of an optimized partial least squares model, demonstrating a separation of the majority of samples from high-ranking individuals from samples from low-ranking animals. On the y-axis random scatter is introduced for visualization. The underlying genera are color-coded for their respective phylum.

the findings made on bacteria in the last decades it seems likely that the mammalian intestinal biome also includes diverse commensals and mutualists (Wegner Parfrey et al., 2014; Lukeš et al., 2015). In microbial communities at least, an increase in species richness increases the potential for metabolic interactions and dependencies between community members, resulting in more stable communities because they become more independent from the environment (Zelezniak et al., 2015). More stable communities are less likely to suffer perturbations in their composition (dysbiosis) as reported for many diseases. To what extent this also applies to intestinal eukaryotic communities is unclear at present. We found that high-ranking animals had a significantly more diverse eukaryome than low-ranking animals and we interpret this result to indicate a healthier intestinal ecosystem in highranking animals. Previously we have shown that female social status determines access to food resources within the clan territory and foraging effort (Hofer and East, 1993b). As a result, low-ranking females have higher foraging costs and higher fGCM concentrations (Goymann et al., 2001b) indicative of an elevated allostatic load. Low-ranking adult females more often resort to resource allocation trade-offs that reduce allocation of resources to immune processes, which therefore resulted in higher burdens of the intestinal helminth Ancylostoma spp., especially during lactation (East et al., 2015). A greater allocation of resources by high-ranking females to immune processes might contribute to maintaining mutualistic microorganisms whilst keeping pathogens in check (Hooper et al., 2012). It seems likely that the immune system of high-ranking females better limits parasitic infections than those of low-ranking females. Their higher contact rate with other clan members (East et al., 2001) and monopolization of parasite-infected social resting sites and parts of carcasses (on which they feed) can then enable highranking females to absorb, establish and maintain a more diverse eukaryome.

To alleviate primer bias and handling problems arising from primers targeting mainly food items we used a multiamplicon sequencing approach (Heitlinger, 2017). To assess to what extent these methods provide a quantitative estimate of the abundance at a particular level of resolution achieved between taxa, we correlated FEC or FOC with sequence reads. Our results indicate that sequencing based estimates were more sensitive both in the detection and the identification of parasite taxa than morphologically based FEC or FOC. We also found that the number of sequence reads was moderately to strongly positively correlated with FEC or FOC counts (**Figure 2**), which generally indicates that a degree of quantification was possible using amplicon sequencing, as reported by previous studies (Kartzinel et al., 2015; Pornon et al., 2016). For analyses of this kind, it is important to take into account the fact that taxa vary in the amount of DNA present in traditionally counted entities such as eggs, oocysts or whole individuals as this will influence the number of sequencing reads obtained (Blanckenhorn et al., 2016). The stronger correlations between the number of annotated sequence reads and coccidian FOC counts than between annotated sequence reads and helminth FEC may be due to the different life cycles of these parasites. Coccidian parasites live in the cells of the epithelial lining, thus contribute DNA via oocytes shed into the lumen of the intestines whereas adult helminths reside in the lumen of the intestines, thus potentially can contributed DNA both in the eggs they shed and from adult worms. The extent of discrepancies introduced by adult worm DNA would then depend on the relationship between egg numbers and helminth tissue in the DNA preparation. For some helminth parasites, this relationship may also depend on adult female/male ratios and other biological processes such as densitydependent effects on worm fecundity, which introduce biases in the estimate of hookworm burdens (Anderson and Schad, 1985).

Cestodes in the family Diphyllobothriidae produce eggs with a very similar morphology, thus reliable identification of Spirometra spp. and Diphyllobothrium spp. based on egg morphology is challenging (Thanchomnang et al., 2016). Our sequencing based taxonomic assignments suggest that eggs from both Spirometra spp. and Diphyllobothrium spp. were included in these egg counts. Similarly, Ancylostoma spp. have a typical strongyle egg type, hence differentiation between Ancylostoma spp. eggs and those of species such as Haemonchus spp. is difficult. Although Haemonchus spp. are parasites of ungulates, they might be detected in hyena feces by both amplicon sequencing and FEC after ingestion if hyenas fed on the viscera of an infected ungulate prey. The detection of Haemonchus spp. eggs in the feces of hyena does not indicate to us at present that these parasite species are members of the hyena's eukaryome. Our sequencing results revealed that our FOC counts for the order Coccidia (phylum Apicomplexa) likely comprised oocysts from species in three genera. We thus conclude that our amplicon sequencing approach offered an improved resolution and sensitivity over traditional egg or oocyst identification techniques and that combining the results from both approaches can provide complementary information.

Currently, little is known about the gastrointestinal biome of most wildlife species in "natural" ecosystems. Nevertheless, components of Darwinian fitness correlated with intestinal biome features can provide insight into the function of associated organisms for their host. At a more basic level, research on a broad range of wild mammals is required to permit the correct functional role (parasite, commensal or mutualist) to be assigned to individual taxa present in intestinal biomes, and in the case of predatory species to correctly differentiate passaging material such as parasites of prey from true host gastrointestinal biome constituents. For the latter, high resolution amplicon markers (e.g., COI; Hebert et al., 2003) and analysis of correlations of the composition of the apparent intestinal biome and ingested food items might offer solutions. Annotation of taxa with functional roles additionally requires databases collating such information (Poelen et al., 2014). Whereas studies on humans and laboratory animals have revealed the importance of the bacterial microbiome in host nutrition, physiology and immune processes, little is known about the impact of intestinal eukaryote diversity and composition on any host. We propose that the assessment of intestinal biomes in freeranging wildlife in the context of host fitness in terms of survival or reproductive success can help to identify beneficial and adverse community compositions for different demographic or social categories of host populations and distinguish those from dysbiosis. The hologenome concept of evolution proposes that evolution in complex organisms should, in addition to considering interactions between an individual's genome and its environment, also consider its interactions with the products and physiological processes arising from the combined genomes of the microorganisms it hosts (Zilber-Rosenberg and Rosenberg, 2008; but see Douglas and Werren, 2016). This has led some to suggest that, during the current period of rapid environmental change, the plasticity of the gastrointestinal microbiome may help some vertebrate populations adjust in an appropriate manner (Alberdi et al., 2016). We suggest that research on the roles of intestinal biomes for humans and wildlife should in addition encompass both unicellular and multicellular eukaryotes, including those traditionally thought of as parasites and the vast majority of organism so far unknown for their impact on hosts, before we can arrive at a balanced view of the benefits and costs of different community compositions of intestinal biomes.

### DATA DEPOSITION

Raw data has been deposited under accession number PRJNA386767 at NCBI Short Read Archive (SRA).

### ETHICS STATEMENT

All protocols were non-invasive and adhered to the laws and guidelines of Tanzania. Permission to conduct research in Tanzania was granted to HH, ME, and SF by the Tanzania Commission for Science and Technology. Permission to undertake research within the Serengeti National Park was granted by the Tanzanian National Parks Authority, and the research was approved by the Tanzanian Wildlife Research Institute. The research was also approved by the Committee for Ethics and Animal Welfare of the Leibniz Institute for Zoo and Wildlife Research under the approval number 2008-11-02.

### AUTHOR CONTRIBUTIONS

EH, SF, HH, and ME designed the study, EH, SF, DT, HH, and ME collected the data, EH and SF

### REFERENCES


analyzed the data, EH, SF, HH, and ME wrote the manuscript. All authors approved the final version of the manuscript.

### FUNDING

The study was funded by the Leibniz Institute for Zoo and Wildlife Research and the DFG Research Training Group 2046 "Parasite Infections: From Experimental Models to Natural Systems" (SF).

### ACKNOWLEDGMENTS

We thank Sarah Benhaiem, Nelly Boyer, Annie Francis, Katja Goller, Janine Helms, Thomas Shabani for their assistance, Luis Santos for his contribution to PCR primer selection, Ines Lesniak for help with preparing the Fluidigm Access Array PCR chips and two reviewers for their constructive comments on the manuscript.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fcimb. 2017.00262/full#supplementary-material


Yildirim, S., Yeoman, C. J., Sipos, M., Torralba, M., Wilson, B. A., Goldberg, T. L., et al. (2010). Characterization of the fecal microbiome from non-human wild primates reveals species specific microbial communities. PLoS ONE 5:e13963. doi: 10.1371/journal.pone.0013963

Zelezniak, A., Andrejev, S., Ponomarova, O., Mende, D. R., Bork, P., and Patil, K. R. (2015). Metabolic dependencies drive species co-occurrence in diverse microbial communities. Proc. Natl. Acad. Sci. U.S.A. 112, 6449–6454. doi: 10.1073/pnas.1421834112

Zilber-Rosenberg, I., and Rosenberg, E. (2008). Role of microorganisms in the evolution of animals and plants: the hologenome theory of evolution. FEMS Microbiol. Rev. 32, 723–735. doi: 10.1111/j.1574-6976.2008.00123.x

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

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

# Modeling Immune Response to *Leishmania* Species Indicates Adenosine As an Important Inhibitor of Th-Cell Activation

Henrique A. L. Ribeiro1, 2 \*, Tatiani U. Maioli <sup>2</sup> , Leandro M. de Freitas <sup>3</sup> , Paolo Tieri <sup>1</sup> and Filippo Castiglione<sup>1</sup>

<sup>1</sup> Consiglio Nazionale delle Ricerche, Istituto per le Applicazioni del Calcolo, Rome, Italy, <sup>2</sup> Departamento de Nutrição, Universidade Federal de Minas Gerais, Belo Horizonte, Brazil, <sup>3</sup> Núcleo de Biointegração, Universidade Federal da Bahia, Vitória da Conquista, Brazil

Infection by Leishmania protozoan parasites can cause a variety of disease outcomes in humans and other mammals, from single self-healing cutaneous lesions to a visceral dissemination of the parasite. The correlation between chronic lesions and ecto-nucleotidase enzymes activity on the surface of the parasite is addressed here using damage caused in epithelial cells by nitric oxide. In order to explore the role of purinergic metabolism in lesion formation and the outcome of the infection, we implemented a cellular automata/lattice gas model involving major immune characters (Th1 and Th2 cells, IFN-γ, IL-4, IL-12, adenosine −Ado−, NO) and parasite players for the dynamic analysis of the disease progress. The model were analyzed using partial ranking correlation coefficient (PRCC) to indicate the components that most influence the disease progression. Results show that low Ado inhibition rate over Th-cells is shared by L. major and L. braziliensis, while in L. amazonensis infection the Ado inhibition rate over Th-cells reaches 30%. IL-4 inhibition rate over Th-cell priming to Th1 independent of IL-12 are exclusive of L. major. The lesion size and progression showed agreement with published biological data and the model was able to simulate cutaneous leishmaniasis outcomes. The sensitivity analysis suggested that Ado inhibition rate over Th-cells followed by Leishmania survival probability were the most important characteristics of the process, with PRCC of 0.89 and 0.77 respectively. The simulations also showed a non-linear relationship between Ado inhibition rate over Th-cells and lesion size measured as number of dead epithelial cells. In conclusion, this model can be a useful tool for the quantitative understanding of the immune response in leishmaniasis.

#### Keywords: leishmaniasis, cutaneous, adenosine (Ado), model, lattice-gas, inflammation

## INTRODUCTION

Leishmaniasis is an infectious disease caused by parasites from Leishmania species. It is considered a neglected tropical disease, affecting 96 countries worldwide (Alvar et al., 2012). More than 700.000 new cases are reported by WHO every year (WHO, 2017). The disease presents in two main different clinical forms, visceral and cutaneous leishmaniasis, and their outcome ranges from self-healing cutaneous lesions to disseminated lesions or to visceral

#### *Edited by:*

Anton Aebischer, Robert Koch-Insitute, Germany

#### *Reviewed by:*

Weihuan Fang, Zhejiang University, China Antonio M. Mendes, Instituto de Medicina Molecular (IMM), Portugal

> *\*Correspondence:* Henrique A. L. Ribeiro h.deassis@iac.cnr.it

*Received:* 04 November 2016 *Accepted:* 22 June 2017 *Published:* 20 July 2017

#### *Citation:*

Ribeiro HAL, Maioli TU, Freitas LM, Tieri P and Castiglione F (2017) Modeling Immune Response to Leishmania Species Indicates Adenosine As an Important Inhibitor of Th-Cell Activation. Front. Cell. Infect. Microbiol. 7:309. doi: 10.3389/fcimb.2017.00309 dissemination of the parasite, possibly leading to death if not properly treated (Carvalho et al., 1985; Alvar et al., 2012). The clinical form and severity of leishmaniasis depend on the parasite species/strain involved and on the host immune response mounted (Sacks and Noben-Trauth, 2002; Hurdayal and Brombacher, 2014).

The immune response involved in leishmaniasis has mainly been studied in mice. The animal models most frequently used are BALB/c and C57BL/6 mice infected with L. major, which represent typical models for susceptibility and resistance respectively. The susceptibility in the BALB/c strain is related to development of (Th2)-polarized immune response, genetically characterized by a high production of interleukin (IL)-4 by CD4<sup>+</sup> T cells (Sacks and Noben-Trauth, 2002; Mougneau et al., 2011). Conversely, the C57BL/6 strain represents a model for resistance to L. major, due to (Th1) polarized immune response, characterized by the production of high amounts of interferongamma (IFN-γ) by CD4<sup>+</sup> T cells (Sacks and Noben-Trauth, 2002; Mougneau et al., 2011). The differentiation of naïve helper T (Th) cells into Th1 or Th2 cells depends on antigen presentation, timing, and cytokines produced by dendritic cells (DCs) after contact with the parasite (Vieira et al., 1994; Hurdayal and Brombacher, 2014). When DCs express nuclear factors such as STAT6 and secrete IL-4 and IL-10, they instruct naïve Th cells to differentiate into the Th2 phenotype (Dent et al., 1999; Sacks and Noben-Trauth, 2002). DCs expressing the nuclear factors STAT4 (Buxbaum et al., 2002), STAT1, and IL-12, induce the differentiation of naïve Th cells into the Th1 subtype, in turn producing IFN-γ (Stamm et al., 1999; Sacks and Noben-Trauth, 2002; Jayakumar et al., 2008).

The immune responses to other species of Leishmania such as L. braziliensis and L. amazonensis are less well studied than that to L. major, and susceptibility mechanisms are not as clearly linked to a Th2 immune response as they are in the L. major model. However, it seems to be clear that control of parasite growth is always dependent on IFN-γ and other inflammatory cytokines. In the L. braziliensis BALB/c infection model, secretion of IFN-γ and TNF-α, is observed but no Th2 immune response, is induced even in the absence of IL-12 (Souza-Neto et al., 2004; Vargas-Inchaustegui et al., 2008). Also, susceptibility to L. amazonensis is related to very low levels of IFN-γ production and low cell proliferation rate in response to its expression of serine phosphate on its membrane and to a high expression of ecto-nucleotidases (Maioli et al., 2004; Franca-Costa et al., 2012).

Parasites including the Leishmania species employ mechanisms to escape the immune response, interfering with signaling pathways of antigen-presenting cells (APCs) and with the differentiation of Th cells (Mougneau et al., 2011). An important escape mechanism employed by Leishmania may be related to the conversion of trinucleotides to adenosine (Ado). It has been reported that ATP leads mostly to pro-inflammatory signals while Ado acts by limiting the inflammation (Bours et al., 2006; Cekic and Linden, 2016). Our group and others have shown that increased ecto-nucleotidase activity on the surface of these parasites correlates with different virulence levels of the cutaneous form of the disease (Maioli et al., 2004; Marquesda-Silva et al., 2008). L. amazonensis has the highest activity of ecto-nucleotidases, which leads to a higher concentration of Ado, decreasing the capacity of DC to present antigens and induce differentiation of Th cells leading to less Th cell proliferation and cytokine production (de Souza et al., 2010; Leite et al., 2012; **Figure 1**).

The complex dynamics of the Leishmania-host interaction can be addressed with a certain degree of success by using mathematical and computational modeling. A system of ordinary differential equations (ODEs) was proposed (Nelson and Velasco-Hernandez, 2002) and later expanded (Biswas et al., 2016) to describe the dynamics of macrophages and parasites in the early phase of infection prior to the development of the adaptive immune response. Another approach—agentbased modeling (i.e., a class of discrete computational models; Castiglione, 2009)—has been implemented with L. major infection data to describe the dynamics of parasites and macrophages in the later phase of infection (Dancik et al., 2010). The authors found that the decrease in the number of macrophages following peak infection could be explained by their uptake of necrotic tissues. Paradoxically, they also found that a decrease in the parasite reproduction rate might eventually lead to more parasites. In line with these results, a different ODE model showed a negative correlation between parasite load in the initial stage of the infection and the overall number of parasites at the end of the observed time window (Länger et al., 2012). The hypothesis arising from both models is that a smaller reproduction rate elicits a weaker immune response, resulting in higher survival rates of the parasite.

Despite some progress, comprehensive models for Leishmania-host interactions and leishmaniasis progression have not been yet fully implemented: most models so far are based only on specific aspects of the disease such as the interaction between macrophages and parasites. To our knowledge, this is the first model of leishmaniasis that covers the general dynamics of the infection and takes into account the importance of purinergic (i.e., adenosine- and ATP-based) signaling. Here we propose a model of cutaneous leishmaniasis caused by different Leishmania species, L. major, L. braziliensis, and L. amazonensis, which aims to establish a minimum set of rules that can describe the development of infection for each species and to test the importance of Ado release as a virulence factor. We created a model composed of the key immune competent cell types (CD4<sup>+</sup> Th cells, macrophages, DCs and epithelial cells) and molecules (IL-4, IL-12, IFN-γ, Ado, and nitric oxide, NO) reported in the literature, and the parasite. By modeling such key aspects together, we were able to demonstrate the effect of Ado on the number of parasites and the lesion formation process, thus showing the importance of Ado in the inhibition of inflammatory Th cells in Leishmania infection, as well as in the lesion formation processes.

### MATERIALS AND METHODS

### The Computational Model

One of the first lattice-gas models (Pandey and Stauffer, 1989, 1990) of infectious diseases was constructed with the aim of simulating events occurring at infected sites and draining lymph

nodes. This type of model comprises a space-representing lattice, where the "sites" on the lattice can take a given number of different states. Evolution of the simulation, i.e., state change at the sites, is done in discrete time steps. In each time step, the state change (or not) at a given site is determined by the state of the site itself and the neighboring sites. Here, a bi-dimensional lattice with six neighbors per lattice-point and periodic boundary conditions was implemented. A representation of a portion of this lattice can be seen in **Figure 2**.

Each lattice site, representing a given volume, contains a combination of cells, cytokines and other molecules (Equations 2–10 and **Table 1**). These entities may be present in low or high concentration, represented by 0 or 1 respectively. These entities will interact and as a result, their states will change during the simulation steps (**Figure 2**).

In mathematical terms, the whole lattice L × L is represented by the vector S(t) = (S1(t), ... , SL×L(t)) where each Si(t), a bitword representing site i at time t, Si(t) = (si,1(t), ... , Si,n(t)) and si,<sup>k</sup> (t) represents the concentration (0 for low and 1 for high) of the entities k, at time t in lattice point i. The binary state of si,k (t + 1) at time (t + 1) depends on the state of the neighboring sites, including the site s itself at the previous iteration t,

$$s\_{i,k}\left(t+1\right) = HS\left(\sum\_{j \in I\_i} s\_{j,k}\left(t\right) - \theta\_k\right) \tag{1}$$

where I<sup>i</sup> is the set of neighbors of lattice point i. The value θ<sup>k</sup> is a threshold value that is zero for all entities except for epithelial cells for which it is 2, avoiding the unrealistic scenario of having islands of living cells in the middle of dead ones. The function HS is the Heaviside step function HS(x) = 1 for x > 0 and 0 for x ≤ 0.

At any time step t, this value is calculated for all entities k = 1, ... , n and lattice pointsi = 1, ... , L 2 . Then, entities in the same site i interact with each other through Boolean rules representing the reaction rules (Equations 2–10 below). The resulting value si,k (t + 1) represents the new micro-state at time t + 1. Equation (1) leads to the propagation of the entities on the lattice and together with the reaction rules represents the reaction-diffusion terms (thus including diffusion, creation and annihilation of particles) of the model. A rationale for each of Equations (2)– (9) is provided in this section and further validation from the literature is provided in **Table 1**.

The model considers n = 9 particles representing biological entities, namely: IL-12; IL-4; interferon IFN-γ (indicated IFN), activated T-helpers lymphocytes type 1 (Th1) and type 2 (Th2), Leishmania parasite (L); nitric oxide and reactive oxygen species (collectively indicated as NO); adenosine (Ado); epithelial cells (E). Besides these, three other entities are implicitly represented: macrophages, DC and naïve T helper cells. In details along the simulation, we consider that each one of these three entities is always present, as in work by Pandey and Stauffer (Pandey and Stauffer, 1989, 1990).

The reaction rules of the model are described as follows. Equations (6)–(8) use the realization of a Bernoulli event ψ(ki) that takes value 1 with probability k<sup>i</sup> ; in other words, ψ(ki) models a chemical reaction as a stochastic event occurring with rate k<sup>i</sup> . Values of k<sup>i</sup> were selected to agree with those published in the literature (**Table 3**). In the equations below, the symbol "∨"

represents an "OR" logical operator, the symbol "∧" represents an "AND," while "¬" represents a "NOT."

$$\text{mL4} = \text{Th2} \tag{2}$$

$$I\text{FN} = \text{Th1}\tag{3}$$

$$\text{IL12} = \text{L} \land \text{IL4} \tag{4}$$

Ado = NO ∧ L (5) Th1 = {- ψ k1 ∧ L ∧ (IL12 ∨ φIL12) ∧ ψ k2 ∧ IL4 ∨ - Th1 ∧ ψ k3 ∨ L } ∧ ¬(ψ k4 ∧ Ado) (6)

$$\text{Th2} = \left(\dot{\psi}\left(\mathbf{k\_1}\right) \land \dot{\mathbf{L}} \land \neg \text{(IL12} \lor \dot{\phi}\_{\text{IL12}})\right) \lor \left[\text{Th1} \land \dot{\left(\psi\left(\mathbf{k\_3}\right)\right)}\right]$$

$$\{\text{2}(L)\}\{\}\land\neg(\psi\{k\_{4}\}\land Ado)\tag{7}$$

NO = IFN (8)

$$L = \psi\left(k\_5\right) \land L \land \neg NO \tag{9}$$

$$E = E \land \neg \text{NO} \tag{10}$$

Equations (2) and (3) show the production of the two antagonistic key cytokines IL-4 and IFN-γ produced respectively by Th2 and Th1 cells (Mougneau et al., 2011; Alexander and Brombacher, 2012).

IL-12 is produced by DCs in the model in response to Leishmania (L) and IL-4, represented in Equation (4). The rule of IL-4 instigating the production of IL-12 by DC is established in the literature. For a review see the work of Hochrein et al. (2000). In the case of leishmaniasis, this process has been observed for L. major in mice (Biedermann et al., 2001).

Equation (5) describes Ado production after NO and Leishmania signals, here being modeled as the conversion of nucleotides to Ado by Leishmania ecto-nucleotidase after host cell content released upon injury caused by NO (**Figure 1**) (Marques-da-Silva et al., 2008; de Souza et al., 2010). In this model, Ado represents the ratio between adenosine and ATP (ATP and other nucleotides). We consider that there is always some kind of host cell present in a lattice point. These cells may be the epithelial cells represented explicitly in the model or may be cells from sub-epithelial tissues.

Equations (6) and (7) show the interactions leading to the differentiation and survival of Th1 and Th2 cells. These two equations are very similar and can be broken down into three parts: (1) activation/priming, (2) recruitment/survival, and (3) Ado inhibition.

The leishmania antigen (L) being presented by DC to naïve T-cells is modeled by the terms "ψ k1 ∧ L ∧ (IL12 ∨ φIL12) ∧ ¬ ψ k2 ∧ IL4 " and "ψ k1 ∧ L ∧ ¬(IL12 ∨ φIL12) ," leading to activation and priming to Th1 or Th2 cells respectively. This process is non-deterministic, with ψ(k1) the likelihood of finding a T-cell with TCR specific to leishmania antigens. As can be observed, IL-12 drives the priming of naïve T-cells to Th1 phenotype (Mougneau et al., 2011) and IL-4 acts as an inhibitor of Th1 priming (Szabo et al., 1997). The φIL<sup>12</sup> codes for the need of IL-12 for Th1 priming. If φIL<sup>12</sup> is set to 1 T-cells will be primed to Th1 in spite of IL-12 (Vargas-Inchaustegui et al., 2008). IL-4 does not drive the activation to Th2 in agreement with evidence suggesting that, in the case of leishmaniasis, the Th2 phenotype may be acquired in the absence of IL-4 (Noben-trauth et al., 1996; Mohrs et al., 2000; Stetson et al., 2002). Note that IL-4 inhibition over Th1-cells priming is a probabilistic event with probability k2.

The term "Th ∧ ψ k3 ∨ L ," present in both Equations (6) and (7), leads to activated Th-cells expansion. This expansion occurs mainly in the regions where the antigens (L) are present. In the absence of antigen Th-cells die out with a half-life determined by "ψ k3 ." This term accounts for clonal expansion and recruitment (Malherbe et al., 2000; Mougneau et al., 2011). Note that in this term antigen leads to the survival of Th cells, which is also in agreement with the literature (Reckling et al., 2008).

Ado inhibition was modeled by the third part of Equation (6) and (7) (¬(ψ k4 ∧ Ado)), letting Ado inhibit activation and survival of Th cells. This process is also probabilistic and higher probabilities represent higher Ado/ATP ratios. For reviews of the effect of Ado on the immune system see the works of Bours et al. (2006) and Cekic and Linden (2016).


\*Concepts that apply to both Th1 and Th2 cells.

Column 1 Equation number, column 2 brief description of the Equation, column 3 key topics and concepts modeled by the Equation, column 4 bibliographical references justifying these topics.

Equation (8) represents macrophage activation by IFN and subsequent NO production (Mougneau et al., 2011; Podinovskaia and Descoteaux, 2015).

Leishmania duplication in Equation (9) depends on the parameter k<sup>5</sup> representing the reproduction of the parasite inside macrophage and its clearance by NO (Mougneau et al., 2011; Podinovskaia and Descoteaux, 2015).

Equation (10) models the way in which epithelial cells will either be killed by NO or survive and multiply (Murphy et al., 2008; Mougneau et al., 2011). Epithelial cells are affected by NO and do not influence any other entity; they were included in the model so that it was possible to simulate a wound.

The lattice is initialized with all lattices containing epithelial cells and just three adjacent points containing leishmania (**Figure 2**).

### Animation

The software MATLAB R2012b was used to extract data with the function grabit and to create AVI animations (Supplementary Material).

### Sensitivity Analysis

Parameter sensitivity was performed by using Latin Hypercube Sampling (LHS a statistical method for generating a near-random sample of parameter values from a multidimensional distribution) on a grid of 243 combinations of the five parameters of the model (k<sup>1</sup> . . . k5). For each combination, the average of three independent runs was taken. Each execution consisted of 480 iterations and the area under the parasite curve was used as a reference. Partial Ranking Correlation Coefficient (PRCC) between the five parameters and the area under the curve was measured with the software R (function pcc of the package sensitivity). **Table 2** shows the range of values tested.

### RESULTS

### Predictions are Coherent with Experimental Results in Animal Models

L. major is the most studied Leishmania species, mainly in C57BL/6 and BALB/c mice. These models represent aspects of resistance and susceptibility to the disease and provide a good agreement with known facts about the immune response to this infection (Belkaid et al., 2000; Cangussú et al., 2009). Our approach was first to simulate C57BL/6 mouse L. major infection to tune the model, and then to try to adapt it to others cutaneous leishmaniasis as those involving L. braziliensis and L. amazonensis. We first started a process of extracting rules and tuning parameters by searching the literature and comparing simulated results with data from Belkaid et al. (2000) (**Figure 3**).

The set of parameter values (ψ(ki)) were adjusted to agree with the literature data (**Table 3**). Two L. major specific parameters were found: k<sup>2</sup> (50% of IL-4 inhibition rate over Th-cell priming), and φIL<sup>12</sup> (Th1 priming independent of IL-12). The parameter k<sup>4</sup> (Ado inhibition rate over Thcells) showed the same value in L. major and L. braziliensis (5%). The three parameter values k<sup>1</sup> [Th-cell activation probability (0.001%)], k<sup>3</sup> [Th-cells survival probability (18%)] and k<sup>5</sup> [Leishmania survival probability (35%)] are not specific since these values are the same in the three different species (**Table 3**). So, these shared parameters values can be reused in other Leishmania species simulations. Thus, L. major was associated with IL-4 inhibition rate over Thcell priming and Th1 priming independent of IL-12, while Th-cells activation and survival probability and Leishmania survival probability are common parameters shared by all these species.

TABLE 2 | Range of values tested in LHS-PRCC (latin hypercube sampling-partial ranking correlation coefficient in the sensitivity analysis).


### Changing Parameters Allows Simulation of *L. braziliensis* or *L. amazonensis* Infection

The first point we observed was that there is no evidence of IL-4 or IL-12 priming Th-cells during infection of C57BL/6 mice with L. amazonensis or L. braziliensis, and the absence of IL-12 does not impair the control of L. braziliensis (Maioli et al., 2004). This was simulated by setting φIL<sup>12</sup> to 1, which permits Th1 priming independent of IL-12. This change completely abrogates IL-4 production and lead to smaller lesions in agreement with observations of L. braziliensis infection (**Figures 4**, **7**). Notice in **Table 3** that for these types of leishmaniasis the inhibition of IL-4

TABLE 3 | Parameters and values used to simulate each of the three models of cutaneous leishmaniasis.


For some of Leishmania species, the inhibition of IL-4 over Th1 priming is indicated as "not applicable" (NA) since there is no IL-4 production or inhibition.

Column 1: parameter; column 2: parameter description; column 3, 4, 5: values used to simulate L. major, L. braziliensis, L. amazonensis. NA, Not Applicable.

over Th1 priming is indicated as "not applicable" (NA) since there is no IL-4 production or inhibition (Maioli et al., 2004).

One explanation for the differences between leishmaniasis coming from different species is the parasite ability to metabolize ATP by membrane ecto-nucleotidase enzymes. L. amazonensis has higher efficiency than other species in metabolizing purines by membrane ecto-nucleodidase and this increases the Ado concentration in the microenvironment, so a higher Ado/ATP ratio can inhibit immune response (Marques-da-Silva et al., 2008; de Souza et al., 2010). To simulate this, the rate of Ado inhibition (k4) over Th-cell activation and survival was increased and eventually we found an unresolved disease with thicker lesions full of live parasites in agreement with the expected findings for L. amazonensis infection in C57BL/6 mice (Marques-da-Silva et al., 2008; **Figures 4**–**6**).

Our simulated data was compared with the results of reported time-series for these three models of leishmaniasis (Ji et al., 2003; Maioli et al., 2004; Marques-da-Silva et al., 2008). These papers report time-series forlesion size measured as thickness of footpad or ear swelling in mice. Our models simulate lesion in the form of a superficial wound. However, the general dynamics of the lesion must be the same. The data used in the simulation was normalized to compare with the real lesions (**Figures 4A–G**).

The lesion size time series from the literature (Ji et al., 2003; Maioli et al., 2004; Marques-da-Silva et al., 2008) and the simulated data were normalized by the size of L. amazonensis lesion at the 6th week. This normalization allowed checking if the models could simulate lesions with similar relative sizes.

Comparison of our simulated data from models of L. braziliensis and L. amazonensis infection with footpad thickness measured by Maioli et al. (2004) showed that the models made good predictions for lesion growth and healing (**Figures 4A,B**). The relative size of the lesion caused by L. braziliensis with respect to that caused by L. amazonensis is also correctly simulated. The time-step length defined in **Figure 3** (80 steps/week) and the starting point of zero weeks post-infection was used.

An agreement between the model and the biological data was observed in the lesion formation induced by L. amazonensis and by L. major simulations; the peak and recovery time of L. braziliensis infection and the growing of lesion in L. amazonensis infection were correctly predicted (**Figures 4C,D**).

The predictions made by the model agreed with biological data in the range of the variability (shown in **Figure 4H**). These comparisons indicated that the three models are fine-tuned using correct parameters values, and can simulate the progression and outcome of the lesion size induced by infection with Leishmania species.

### Sensitivity Analysis of the Model

Sensitivity analysis was performed by testing combinations of parameters in the range shown in **Table 2**. The procedure was to vary the parameter values in the broadest range as possible given their restriction. For k1, which represents the TCR-antigen specificity probability, the value must be small, so the range tested included values from 0.00001 to 0.01%. Values of k<sup>3</sup> above 20% do not make sense, because it would simulate replication and not

FIGURE 5 | Infection resolution rate as a function of the percentage of adenosine (Ado) inhibition. Simulations were run for 3000 iterations with parameters k1 = 0.001%, k2 = 0%, k3 = 18%, k5 = 35%, φIL<sup>12</sup> =1 and Ado inhibition rate over Th cells (k4) varying from 0 to 40%. Resolution rates were calculated over 10 independent simulations. Simulated mice were reported as cured (number of parasites equal zero) or non-cured (number of parasites greater than zero) after 3,000 iterations. The figure shows that 20.5% adenosine inhibition rate over Th cell is the critical value beyond which no cures are observed.

half-life, and conversely k<sup>5</sup> cannot be below 20%; a minimum of 30% was used for biological fidelity considerations.

Sensitivity analysis reveals that the capacity of Ado to inhibit naïve Th cells activation (k4) is an important parameter (**Table 4**). This parameter has the largest influence on the number of parasites throughout the simulation, overcoming the parasite

TABLE 4 | Sensitivity analysis performed on parameters k1,...,k5 with Latin Hypercube Sampling (LHS) and Partial Ranking Correlation Coefficient (PRCC) (Gomero, 2012).


The PRCC is shown as average ± standard deviation measured with 200 bootstraps.

growth rate (related to the value k1). The correlation is positive, showing that Ado increases the susceptibility to leishmania infection. IL-4 also showed a positive correlation with parasite number, which indicates that it is a susceptibility promoter. However, Th cell activation probability and Th cell survival probability (related to Th-cell half-life) showed a negative correlation with parasite number. This makes sense since Th cell activation probability and Th cell survival probability can both be linked to decreasing the probability of parasite survival. A variation of this test was tried in which k<sup>1</sup> was fixed at 0.001%, and it had very similar results. In another variation, IL-12 production was turned off and again the results were similar to those in **Table 4** (data not shown).

### Inhibition of Th-Cell Activation by Adenosine Is the Key Factor in the Leishmaniasis Outcome

Following sensitivity analysis that points to the capacity of Ado to inhibit cell activation as the most important parameter, a deeper exploration was performed. **Figure 5** shows the percentage of resolution of the infection in respect to Ado inhibition in Th-cells. These simulations were performed without IL-12 production and with the capacity of Ado to inhibit Th-cell activation (k4) varying between 0 and 40%. Simulated mice were considered cured if after 3,000 iterations the parasite number was equal to zero. The threshold for not curing leishmaniasis is around 20.5% inhibition of Th-cell activation by Ado. The outcome experiments showed that inhibition rate agrees with the phenotype for L. amazonensis infection, but we decided to use 30% inhibition instead of 20.5%. Nevertheless, 20.5% inhibition capacity of Ado in Th-cell activation may fit data from infection with less virulent strains of L. amazonensis.

The effect of Ado on the number of parasites is intuitive, with the number of parasites growing with as Ado concentration and its capacity to inhibit Th-cells increases. It was evaluated by the size of the lesion measured in terms of the number of "dead" lattice points in respect to the Ado inhibition rate in effector Thcells (k4) (**Figure 6**). In this experiment, we let lesions evolve for 480 iterations (6 weeks of simulation). This shows that inhibition in the range of 0–20% does not have a strong effect on the maximum lesion size; inhibition from 20 to 40% inhibition leads to an increase in the lesion size and finally with Ado inhibition over effector Th-cells from 40 to 100%, the lesion size decreased. In this extreme case (>40% of inhibition) the number of parasites continuously increase but immune response is almost completely inhibited, which explains smaller lesions, while in the middle case (20–40%) there is an immune response strong enough to cause lesions but not strong enough to control the parasite growth.

These results indicate that a range of 0–15% in the inhibition in Th-cells' activation by Ado represents the peak of a lesion that eventually heals, while points greater than 20% represents the size of the lesion at the end of simulation since these lesions do not heal and grow continuously. Besides that, lesions with inhibition up to 55% represent a single globular dense wound while points above this threshold represent the sum of several diffuse lesions (Supplementary Material).

A comparison between the pattern of simulated lesions and real lesions is shown in **Figure 7**. In **Figure 7B** it is possible to see the pattern of simulated L. braziliensis lesion (**Table 3** column 4) and the comparison with real lesion (**Figure 7A**). We can see the model correctly predicts this wound as the smallest of the tree lesions simulated. **Figure 7D** shows the pattern produced by the simulation of L. major infection (**Table 3** column 3) and **Figure 7C** shows the real lesion. Similarly, the simulated lesion of L. amazonensis (**Table 3** column 5) is seen in **Figure 7F** and a real lesion in **Figure 7E**. The model correctly predicts that as the largest lesion. Assuming a lattice-point diameter of 10 µm, which is roughly the size of one host cell, the dimension of the entire lattice is about 1 cm<sup>2</sup> , which is about the size of a mouse ear. In this respect, we observed that the simulated lesions are comparable in size to real lesions in mice. The evidence above indicates that these models are fine-tuned and can simulate the infection profile with different Leishmania species, and different induced immune response and lesion outcome at an equivalent size to what is seen in mouse models of infections.

### DISCUSSION

The model we present here was created with the aim to reproduce the behavior of Leishmania infection from experimental parasite burden curve related to the immune response developed by the host in the mouse model.

In this model, the time-step length was not predefined by the parameters in Equations (2)–(10) but defined a posteriori by superimposing simulated dynamics with real data. In this sense, comparisons with experimental data (Belkaid et al., 2000) serve as a qualitative validation of the model and as a means of extracting values such as time-step length and the starting point of simulation (**Figure 3**). During the fine-tuning, it was found that the key parameter controlling step-size was Th-cell activation probability (k1). Any value of k<sup>1</sup> in the range of 0.001– 100% probability reproduced published data but with markedly differences in the step-size. With a probability of activation of 100% the time step is close to ∼80 h (∼half a week), while when such probability is 0.001% the time step is 2 h (80 steps/week). The inverse reasoning can be applied to check that it can make perfect sense. In a long period, such as ∼80 h of antigen exposure, the probability of finding a Th-cell specific for leishmania antigen will be close to 100%, while for a smaller period (e.g., 2 h) this probability will be very low at least for low dose inoculations.

Araujo et al. (2014), reproduced in accordance with the policy of the journal. In (B,D,F) iteration is the step in which the screen shot of the lesion was taken and dead cells is the number of lattice-points that contain dead epithelial cells.

The other four parameters had an important bias in the outcome of the disease but they were found not to influence the timestep length. It is interesting to notice that k<sup>2</sup> and k<sup>4</sup> are related to levels of inhibitors (IL-4 and Ado) which do not depend on time while k1, k3, and k<sup>5</sup> are related to Th-cell activation, and Th-cell survival probabilities and leishmania replication rate., All these depend on time, but only k<sup>1</sup> affects time-step length.

The parameter Th-cell half-life (k3) is sensitive to Th-cell activation (k1). With k<sup>1</sup> high (>10%) the k<sup>3</sup> value is nearly irrelevant but for k<sup>1</sup> low (<0.01) k<sup>3</sup> must have a value greater than zero for a cure to be possible. Biologically, this finding makes sense since Th-cells have half-lives of days (Murphy et al., 2008) and k<sup>1</sup> low and k<sup>3</sup> close to zero would lead to an unrealistically short half-life of hours or minutes. It can be also observed that k<sup>3</sup> correlates negatively with the number of parasites throughout the simulation (**Table 4**). Most experiments used a value of 18% for k<sup>3</sup> (**Table 3**). This value will give a Thcell half-live, in the absence of antigen, of about 11.5 steps (1 day), which is roughly close to the half-life of these cells in vivo. Throughout the simulation, the survival of Th cells will also be influenced by parasite presence and Ado production. Entities representing molecules (cytokines, Ado and NO) were modeled with a lifespan of 1 time step (2 h), in agreement with the fact that these substances have half-lives of minutes to hours (Bocchi, 1991; Loffler et al., 2007).

A broad range of values of k<sup>1</sup> could reproduce qualitatively the profile published by Belkaid (Belkaid et al., 2000): indeed Th-cells activation probability in the range of 10-100% leads to a growth in parasite number of the order of 10–10<sup>2</sup> . The same study reported that the number of parasites, following an injection of 10<sup>2</sup> early developmental stage (amastigotes) parasites, has grown up to 10<sup>5</sup> late developmental stage (promastigotes) parasites within 4.5 weeks (Belkaid et al., 2000). It is to be noted that, according to von Stebut (2007), 90% of parasites injected are killed by the complement within 3 min: this means a 103–10<sup>4</sup> -fold increase in the number of parasites. These figures are better reproduced by an activation probability k<sup>1</sup> ∼0.001%. The parameter controlling parasite growth rate (k5) also plays a role in tuning the results to better reproduce literature data. In this work, the pair of values k<sup>1</sup> = 0.001% and k<sup>5</sup> = 35% was very successful in reproducing published data.

The comparisons made in **Figure 3** also showed that the simulation starts at 2.5 weeks post-infection. The key reason why it could not simulate the first 2.5 weeks of the disease is that this model is a quadratic approximation of the real phenomenon. As an approximation, it cannot reproduce the whole dynamics but only a certain range, similar to what one would expect with a Taylor linearization over a critical point. Besides that, 3 lattice-points containing parasites can be translated to 10<sup>2</sup> parasites in a mouse which is not very far from the expected value after 2.5 weeks post-infection given that only about 10 of 100 injected parasites survive after 3 min (von Stebut, 2007). A similar approach of only simulating the late phase of Leishmania infection was used by Dancik et al. (2010); the simulations in his work start at 3.5 weeks post-infection. Nevertheless, the present model simulates part of the early silent phase of infection that according to Belkaid et al. (2000) lasts for the first 3 or 4 weeks.

It was possible to expand a L. major model to simulate other cutaneous leishmaniasis (L. braziliensis and L. amazonensis). As it can be seen in **Table 3** there is no IL-4 induced production of IL-12 for these other leishmaniasis (for a review on this see Hochrein et al., 2000). This adequately models the finding of Maioli et al. (2004) that no IL-4 producing cells could be detected during L. braziliensis or L. amazonensis C57BL/6 infection. An increase in Ado concentration to inhibit the activation of Th-cells was used to simulate L. amazonensis infection, in agreement with literature reports of incurable lesions in C57BL/6 mice infected with L. amazonensis, and with higher ecto-nucleotidase activity in the surface of this parasite (Berredo-Pinho et al., 2001; Ji et al., 2003; Maioli et al., 2004; Marques-da-Silva et al., 2008; de Souza et al., 2010; Gomes et al., 2015). This higher activity may lead to higher Ado and lower ATP concentration at the site of infection.

Notice that for simulation of L. braziliensis infection in C57BL/6 mice it was not necessary to tune the parameters of the model, but only to turn φIL<sup>12</sup> on; this change was nevertheless extracted directly from literature (Souza-Neto et al., 2004; Vargas-Inchaustegui et al., 2008). In the case of L. amazonensis only Ado inhibition over Th-cells (k4) was changed.

Comparison of the lesion size modeled in these simulations with experimental data from Marques-da-Silva et al. (2008), Maioli et al. (2004), Ji et al. (2003) (**Figures 4A–G**) showed that the model is capable of simulating different mouse models of cutaneous leishmaniasis. These comparisons not only validate the new simulations (L. braziliensis and L. amazonensis) but they also serve as a further validation for the L. major model. It is interesting to notice that the simulations were fitted to these other published data (**Figures 4A–G**) using the step-size found with Belkaid et al. (2000) comparison (80 steps/week). The appearance of the lesion takes a few weeks in mice. Thus, the starting point had to be changed from 2.5 to zero weeks post-infection. A justification for that is that in Belkaid's article (Belkaid et al., 2000), the infection was induced with a low dose of inoculum (10<sup>2</sup> parasites injected in the ear), while in the other studies in our comparison the infection was induced with high dose inoculum (105–10<sup>6</sup> parasites injected in the footpad). It is therefore reasonable to assume a different starting point for the appearance of the lesion. It is interesting that, according to the simulations, the inoculum size does not change the dynamics of late infection.

It is important to highlight that the comparisons in **Figure 4** are only meant to be qualitative, since different lesion metrics (lesion area vs. thickness) are used. The aim of these comparisons is to see if the model can capture key aspects of lesion dynamics such as when they appear and disappear and the time when they reach a peak. Data in **Figures 4A–G** were normalized by the size of L. amazonensis infection lesion in C57BL/mice at 6 weeks post-infection. This approach made it possible to compare the relative sizes of different cutaneous leishmaniasis lesions. We observed a good agreement between the model and literature with L. amazonensis lesions, i.e., the largest lesion, followed by L. major, and by L. braziliensis, which induces the smallest wounds in C57BL/6 mice.

While k<sup>1</sup> is the key value for setting the time-step length, Ado inhibition over effector cells (k4) is the most important parameter for simulating the outcome of the disease. This can be observed in the sensitivity analysis (**Table 4**) and in **Figures 5**, **6**. This shows that this parameter has the most influence over the number of parasites throughout the infection and it also controls whether the infection will be cured. These results agree with Marques-da-Silva et al. (2008) and correlate the different outcomes between the three leishmaniasis in this study with ectonucleotidase activity on the surface of parasites and a correlation of ecto-nucleotidase activity with the virulence of different strains of L. amazonensis (de Souza et al., 2010).

Lesion growth is influenced by Ado concentration as is shown in **Figures 6**, **7**. These results show that this purine has a nonlinear relationship with growth. Moderate inhibition of Thcells activation may lead to the largest lesions, in agreement with Marques-da-Silva et al. (2008), Maioli et al. (2004) and Ji et al. (2003). But high Ado concentration would have a stronger inhibition of Th-cell activation and it may lead to smaller and more diffuse lesions (Supplementary Material). These data may correlate with the different outcomes of L. amazonensis infection in humans (Silveira et al., 2009; Hombach and Clos, 2014). It may be the case that for some genetic backgrounds, the immune system responds strongly to Ado with more secretion of IL-10 and TGF-β leading to diffuse cutaneous leishmaniasis outcome. Nevertheless, such effects remain speculative and more studies are needed to fully decipher the influence of Ado in human leishmaniosis.

This model also includes the role of IL-4. This cytokine will induce the production of IL-12 (Hochrein et al., 2000) and at the same time inhibits the action of IL-12 in Th-cells (Szabo et al., 1997). It is reported in the literature that if IL-4 is presented in the earlier phase of leishmaniasis it leads to the production of IL-12 and prime Th1 cells, while if it is presented during Th-cells priming it will lead to Th2 production and will suppress the expression of the IL-12 receptor (Biedermann et al., 2001). Furthermore, Belkaid et al. (2000) showed that IL-4 is produced during the whole infection (weeks 1–22) and that it peaks during the Th-cell priming phase (weeks 4–8) without impairing the Th1 response. The apparent contradiction between these two results can most likely be explained by the dose of IL-4. Biedermann and coworkers (Biedermann et al., 2001) injected 1 µg of IL-4 into the mice, which in a conservative calculation will lead to 10<sup>4</sup> pg/ml of that cytokine, while Belkaid (Belkaid et al., 2000) measures physiological doses of 10–100 pg/ml in the lymph node. The current work simulates physiological conditions that are more in line with what Belkaid (Belkaid et al., 2000) showed. Nevertheless, sensitivity analysis showed that IL-4, despite stimulating the production of IL-12, promotes susceptibility to infection, in accordance with the literature (Biedermann et al., 2001; Cangussú et al., 2009).

It was observed in the animations of L. major infection (Supplementary Material) that there is a spatial negative correlation between L. major parasites and NO production. This finding agrees with results from (Cangussú et al., 2009).

We have shown how a robust model of cutaneous leishmaniasis has been built and validated. It consists of a minimum set of rules that can describe and differentiate L. amazonensis, L. braziliensis, and L. major infection. The simulator agrees with key experimental results published in the literature such as the antagonism between IL-4 and IFN-γ and brings new lights to the influence of Ado signaling in these infections. We could observe a non-linear relationship between this purine and lesion formation and to confirm that it is the key factor in differentiating cutaneous leishmaniasis. Higher Ado concentration can inhibit Th-cell activation leading to infection outcomes that have been seen in experimental models of leishmaniasis. This model can therefore represent a valuable tool for answering questions regarding cellular and molecular players, inflammatory processes related to these infections and finally exploration of combination therapies involving drugs such as sodium stibogluconate, topical paromomycin preparations, etc.

### AUTHOR CONTRIBUTIONS

HR, TM, LF, PT, and FC conceived of the study, HR and FC ran the simulations, all authors contributed to the analysis, wrote the manuscript and finally read and approved the final manuscript.

### ACKNOWLEDGMENTS

We thank EU FP7 project MISSION-T2D (600803) Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) and Fundação de Amparo a Pesquisa de Minas Gerais (FAPEMIG) (Brazil) for the fellowships to HR and TM and

### REFERENCES


the Consiglio Nazionale delle Ricerche (CNR—Italy) to support the work. Dr. Luis Carlos Crocco Afonso (Universidade Federal de Ouro Preto—Brazil) is kindly acknowledged for his critical contribution to the work. We also thank Priscila Guerra (Centro de Pesquisas Gonçalo Moniz/FIOCRUZ—Salvador—Brazil) for providing the L. braziliensis lesion picture in **Figure 7**. Finally, we thank Clare Sansom for carefully reading the manuscript.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fcimb. 2017.00309/full#supplementary-material


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

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

# Strain- and Dose-Dependent Reduction of *Toxoplasma gondii* Burden in Pigs Is Associated with Interferon-Gamma Production by CD8<sup>+</sup> Lymphocytes in a Heterologous Challenge Model

#### Malgorzata Jennes 1 †, Stéphane De Craeye2 †, Bert Devriendt <sup>1</sup> , Katelijne Dierick <sup>2</sup> , Pierre Dorny 3, 4 and Eric Cox <sup>1</sup> \*

<sup>1</sup> Laboratory for Immunology, Faculty of Veterinary Medicine, Ghent University, Merelbeke, Belgium, <sup>2</sup> National Reference Laboratory for Toxoplasmosis, Operational Direction Communicable and Infectious Diseases, Scientific Institute of Public Health, Security of Food Chain and Environment, Brussels, Belgium, <sup>3</sup> Department of Biomedical Sciences, Institute for Tropical Medicine, Antwerp, Belgium, <sup>4</sup> Laboratory for Parasitology, Faculty of Veterinary Medicine, Ghent University, Merelbeke, Belgium

#### *Edited by:*

Anton Aebischer, Robert Koch-Insitute, Germany

#### *Reviewed by:*

Susan M. Bueno, Pontificia Universidad Católica de Chile, Chile Anthony P. Sinai, University of Kentucky, United States Anja Joachim, Veterinärmedizinische Universität Wien, Austria

*\*Correspondence:*

Eric Cox eric.cox@ugent.be † These authors have contributed equally to this work.

> *Received:* 06 January 2017 *Accepted:* 18 May 2017 *Published:* 08 June 2017

#### *Citation:*

Jennes M, De Craeye S, Devriendt B, Dierick K, Dorny P and Cox E (2017) Strain- and Dose-Dependent Reduction of Toxoplasma gondii Burden in Pigs Is Associated with Interferon-Gamma Production by CD8<sup>+</sup> Lymphocytes in a Heterologous Challenge Model. Front. Cell. Infect. Microbiol. 7:232. doi: 10.3389/fcimb.2017.00232 Toxoplasma gondii is a worldwide prevalent parasite of humans and animals. The global infection burden exceeds yearly one million disability-adjusted life years (DALY's) in infected individuals. Therefore, effective preventive measures should be taken to decrease the risk of infection in humans. Although human toxoplasmosis is predominantly foodborne by ingestion of tissue cysts in meat from domestic animals such as pigs, the incidence risk is difficult to estimate due to the lack of screening of animals for infection and insights in location and persistence of the parasite in the tissues. Hence, experimental infections in pigs can provide more information on the risk for zoonosis based on the parasite burden in meat products intended for human consumption and on the immune responses induced by infection. In the present study, homo- and heterologous infection experiments with two distinct T. gondii strains (IPB-LR and IPB-Gangji) were performed. The humoral and cellular immune responses, the presence of viable parasites and the parasite load in edible meat samples were evaluated. In homologous infection experiments the parasite persistence was clearly strain-dependent and inversely correlated with the infection dose. The results strongly indicate a change in the amount of parasite DNA and viable cysts in porcine tissues over time. Heterologous challenge infections demonstrated that IPB-G strain could considerably reduce the parasite burden in the subsequent IPB-LR infection. A strong, however, not protective humoral response was observed against GRA7 and TLA antigens upon inoculation with both strains. The in vitro IFN-γ production by TLA-stimulated PBMCs was correlated with the infection dose and predominantly brought about by CD3+CD4−CD8α bright T-lymphocytes. The described adaptive cellular and humoral immune responses in pigs are in line with the induced or natural infections in mice and humans. Previous studies underscored the heterogeneity of T. gondii strains and the corresponding virulence factors. These findings suggest the potential of the IPB-G strain to elicit a partially protective immune response and to reduce the parasite burden upon a challenge infection. The IPB-G strain could be used as a promising tool in limiting the number of viable parasites in edible tissues and, hence, in lowering the risk for human toxoplasmosis.

Keywords: *Toxoplasma gondii*, infection, pigs, IFN-γ, T lymphocytes, immunity

### INTRODUCTION

Toxoplasmosis is a parasitic infection caused by the intracellular protozoa Toxoplasma gondii. This parasite has a complex lifecycle and affects its definitive host as well as various intermediate hosts, among which domestic and wild animals and humans (Dubey, 2010). During its distinct developmental phases, the parasite manifests itself as a tachyzoite, a bradyzoite or a sporozoite in the oocyst. The sexual multiplication proceeds only within the definitive hosts (domestic or wild members of the family Felidae), wherein five morphologically different generations of the gamonts develop in the enterocytes, leading to formation of the gametocytes. Following the fertilization of the microand the macrogamete and the rupture of the infected cell, the unsporulated oocysts are discharged into the intestinal lumen. The final hosts are responsible for the extensive shedding of oocysts in the environment (Dubey, 1995; Afonso et al., 2008). The sporulated oocysts containing eight sporozoites show a high resistance to different environmental factors and under convenient circumstances may remain infectious for 1.5 years (Dubey, 2010). On average, the final host sheds at least one million oocysts in the acute phase of the infection, resulting in a massive contamination of the environment. This explains the persistence of the parasite in wild reservoir and livestock (Black and Boothroyd, 2000; Afonso et al., 2008; Innes, 2010; Opsteegh et al., 2016). The sporulated oocysts release the sporozoites upon ingestion by the intermediate host, followed by a differentiation to tachyzoites, several cell divisions in the enterocytes, and eventually dissemination to the peripheral tissues. There, the fast-multiplying tachyzoites convert into tissue cysts with the slowly dividing bradyzoites, and remain there as the dormant stage of the infection. The predation of the intermediate host or its tissues by the Felidae leads to the new sexual reproduction cycle, in which the bradyzoites transform back to tachyzoites and merozoites (Dubey, 1995; Afonso et al., 2008).

Several infection routes have been described for the different hosts of T. gondii; the majority of the herbivorous animals acquires the infection through ingestion of water or plants contaminated with oocysts. The predation of other mammals or birds or ingestion of the placenta and/or the aborted offspring of small ruminants facilitates the transmission of toxoplasmosis to carnivores and omnivores (Black and Boothroyd, 2000; Innes, 2010). In humans, foodborne toxoplasmosis mainly results from the consumption of raw or undercooked meat from infected animals, like domestic pigs. The global prevalence of this parasite includes one third of the human population and as such represents one of the most common parasitic zoonosis worldwide (Tenter et al., 2000; Ajzenberg et al., 2002; Aspinall et al., 2002; Bosch-Driessen et al., 2002; Kijlstra and Jongert, 2008; Innes, 2010; Robert-Gangneux and Dardé, 2012; Torgerson and Mastroiacovo, 2013). Consequently, infection in human has a severe short- and long-term impact, ranging from congenital or adult toxoplasmosis in healthy individuals to T. gondii-induced encephalitis in immune-compromised patients. In particular congenital toxoplasmosis in seronegative pregnant women has a very severe clinical relevance for the fetus, since the infection may result in abortion, intracranial calcifications, mental retardation or chorioretinitis in the newborn (Peyron et al., 2017). Finally, acquired toxoplasmosis has recently been associated with an increase in suicide rates or Parkinson's disease (Israelski and Remington, 1993; Holland, 2003; Lester, 2010; Miman et al., 2010; Wang et al., 2017). Therefore, numerous preventive measures are recommended in an attempt to decrease the global infection burden in the human population. The commonly applied precautions include hygienic processing of water and meat, such as boiling of surface water and avoiding the consumption of raw or undercooked meat, as only long term freezing at −12◦C or baking above 67◦C can effectively deactivate tissue cysts. Especially, pork is often consumed undercooked and is processed in many other meat products, reaching on average 300 consumers per pig (Fehlhaber et al., 2003; Belluco et al., 2016). Additionally, direct contact with contaminated soil, plants or cat feces should be avoided by wearing gloves when gardening or emptying the litter box, and by thoroughly washing fresh vegetables and fruits. Providing clear information on these preventive measures to seronegative pregnant women, in combination with a frequent serological screening to detect the acute infection during pregnancy, has proven to be successful in decreasing the infection rate (Breugelmans et al., 2004; Peyron et al., 2017). In livestock, the preventive measures are predominantly focused on strictly indoor housing, preventing access for cats, rodent control and the appropriate carcass disposal (Jones and Dubey, 2012; Robert-Gangneux and Dardé, 2012). Whereas environmental contamination as well as the prevalence of toxoplasmosis in sheep is overall high, the situation in regard to pigs may vary per country or the farm management. Consequently, the lack of uniform validation of the variety of serological assays, and the missing gaps in the correlation between the persistence of antibodies and parasite in pork are still to be improved. Nevertheless, the estimated average prevalence in the pig population seems to be very low in European countries (2.2%)

and the USA (2.7%), presumably due to a shift from small and less strictly confined to large scale facilities, implementing allin-all-out or farrow-to-finish models (Hill et al., 2008; EFSA, 2012; Guo et al., 2015). However, the recent rise of organic or free-range farming in order to improve animal welfare seems to contribute to an increase in infection rate in pig livestock and, as such, to the incidence of foodborne human toxoplasmosis (Kijlstra et al., 2004; Dubey et al., 2012; EFSA, 2012). The risk for humans to become infected by consumption of undercooked or raw pork is also not clear. The knowledge of the parasite persistence in edible tissues of naturally infected pigs is limited, as are the role of strain or dose in the parasite survival in the host. Such information might be of pivotal importance for vaccine development.

In the last decades numerous potential vaccine candidates have been experimentally tested mainly in mice and to a lesser extent in pigs. The formulations varied between a single recombinant parasitic protein or a combination of antigens, among which surface antigens (SAGs) or excretion/secretion proteins (GRAs, ROPs, MICs), but also DNA vaccines encoding B or T cell epitopes have been evaluated. However, the degrees of success were variable and did not led to a commercial vaccine (Vercammen et al., 2000; Letscher-Bru et al., 2003; Jongert et al., 2008; Li et al., 2011; Cao et al., 2015; Wagner et al., 2015). Vaccination did move forward by the use of attenuated viable strains, which resulted with a single commercially available vaccine for sheep, but their efficiency is tested under experimental circumstances and strictly species-dependent, and cannot yet be extrapolated to other livestock species or humans (Katzer et al., 2014; Burrells et al., 2015). Nevertheless, several experimental data in pigs reported reduction in parasite burden in infected and subsequently heterologous challenged pigs, in which the choice of the strain had an important effect on the viability of the parasite (Solano Aguilar et al., 2001; Dawson et al., 2004, 2005; Kringel et al., 2004; Garcia et al., 2005; Verhelst et al., 2011, 2015). In these studies, the involvement of the innate and acquired immune system was observed, dominated by antibodies production against the parasitic antigens, and by the Th1-type of the immune response. Depending on the experimental model, high levels of anti-GRA7 alone or anti-GRA1, -GRA7 and –TLA IgG's were detected upon an inoculation with IPB-G strain or a DNA GRA1-GRA7 cocktail vaccination, followed by the RHstrain or IPB-G challenge, respectively (Jongert et al., 2008; Verhelst et al., 2011), whereas a challenge with heterologous M4 strain oocysts after an experimental inoculation with viable S48 strain tachyzoites elicited a high TLA-specific IgG production (Burrells et al., 2015). In addition to the enhanced humoral immunity, a polarized Th1-immune response was observed after inoculation with a variety of the infectious T. gondii strains (Solano Aguilar et al., 2001; Dawson et al., 2004, 2005; Jongert et al., 2008; Verhelst et al., 2015). The significantly increased IFN-γ protein concentration in serum and the supernatant from the cultured PBMCs, and IFN-γ mRNA or DNA expression in PBMCs and intestinal lymphoid tissues, appeared positively correlated with the duration of the experiments (Solano Aguilar et al., 2001; Dawson et al., 2004, 2005; Jongert et al., 2008; Verhelst et al., 2015). In parallel with IFN-γ, also other cytokines were involved in the immune response against the parasite, as shown in the infection with the VEG-strain oocysts and the increased secretion of IL-15 and TNF-α (Dawson et al., 2005). Subsequently, a Th-2 response profile with predominantly IL-10 as anti-inflammatory cytokine was observed after the early phase of the infection, dominated by IFN-γ production, as mentioned earlier (Solano Aguilar et al., 2001; Aliberti, 2005). In contrast, IL-12 (IL-12p35 and IL-12p40) mRNA expression was not detected in PBMCs shortly after inoculation (7 and 14 dpi) in another study in pigs (Dawson et al., 2005). Nonetheless, even an excessive production of parasite-specific antibodies or Th-1/Th2-response cytokines did not provide a full protection during the acute phase of the infection, preventing from the cysts formation. Despite the active role the different components of the host's immune system play in the early stage of T. gondii infection, it remains a subject of discussion and ongoing research, whether the intermediate host can clear the tissues from the cysts on long term. It is noteworthy, however, that several studies in pigs notified reduced or undetectable counts of the parasite DNA in multiple porcine tissues, and a decline in viability of the cysts, as tested by bio-assay in mice (Jongert et al., 2008; Verhelst et al., 2011, 2015; Burrells et al., 2015). Taking into account the lack of an obligatory screening of pigs or pork meat to prevent transmission to humans, knowledge on the pig as an intermediate host for T. gondii, and in particular strategies to reduce the amount of viable parasites in tissues, may contribute to diminishing the risk of zoonosis by consumption of porcine meat (EFSA, 2007, 2012; Opsteegh et al., 2016). In light of these data, the aim of this study was to confirm differences between T. gondii strains in persistence of the parasite in tissues of experimentally infected pigs and to relate the dose and strain to the immune responses in the pigs upon a single infection or a heterologous challenge.

### MATERIALS AND METHODS

### *T. gondii* Strains

Two T. gondii strains were used for the experimental infections: the IPB-Gangji (IPB-G) strain and the IPB-LR strain. The first one was isolated from the placenta of a patient with congenital toxoplasmosis and is highly virulent in mice. It produces a large number of tissue cysts and has an atypical mixed type I and type II genotype (Ajzenberg et al., 2002). The latter was isolated from pigs and belongs to genotype II, which is less pathogenic and commonly present in the European pig population (Dubey, 2009; Dubey et al., 2012). Both strains are maintained at the National Reference Laboratory for Toxoplasmosis (Scientific Institute for Public Health, Brussels, Belgium) by passage in Swiss female mice, since there is no alternative available to obtain a sufficient number of tissue cysts for the inoculation experiments than via bio-assay, as approved by the Ethical Committee (no. 20140704-01) and conform the European legislation (2010/63/EU). Tissue cysts from both strains were isolated from homogenized brain tissue, counted by phasecontrast microscopy and suspended in 10 ml of sterile phosphate buffered saline (PBS) solution at the desired concentration (700 cysts for the low dose and 6,000 for the high dose). The animals were inoculated within 8 h after cysts isolation. The inoculum for the negative control group was prepared identically from naive Swiss mice.

### Animals and Experimental Design

Two-week-old Belgian Landrace piglets were tested for the presence of anti-T. gondii serum antibodies (IgM and IgG) with an indirect immunofluorescence assay (IFA) as described previously (Verhelst et al., 2015). For the infection experiments, 3-week-old newly weaned, seronegative piglets were selected and randomly assigned to 10 groups of 3 animals (**Table 1**). These groups were housed in isolation units (Biosafety permit no, AMV/11062013/SBB219.2013/0145) at the Faculty of Veterinary Medicine, Ghent University, Belgium. All experiments were approved by the Ethical Committee of the faculties Veterinary Medicine and Bioscience Engineering at Ghent University (EC 2009/149).

In a first experiment we aimed to study the effect of a low or high infection dose of two different T. gondii strains on the humoral and cellular immune responses and tissue cyst persistence until 120 days after inoculation (**Table 1**). In a second experiment we focused on the effect of a subsequent challenge with a heterologous strain at 60 dpi and the persistence of the parasite in the tissues at 120 dpi (**Table 1**). In study 3 we compared kinetics of the IFN-γ producing porcine T cell subsets following infection with high doses of the IPB-G or the IPB-LR strain until 98 dpi (**Table 1**). In each experiment the peripheral blood monomorphonuclear cells (PBMCs) were sampled at regular intervals for the detection of cytokine mRNA by RT-qPCR, and for the quantification of the IFN-γ producing T cell subsets, respectively. At euthanasia, PBMCs and lymphocytes from the peripheral lymph nodes and spleen were isolated for further in vitro assays, whereas heart, diaphragm, skeletal muscles and brain were collected to determine the parasite load as explained further. The experimental timeline presenting the collected samples and the sampling intervals is shown in **Figure 1**.

### Humoral Immune Response

For each experiment the seroconversion was monitored during the first 2 weeks after inoculation (wpi) by daily and subsequently weekly blood collection from the vena jugularis until 120 days post infection (dpi).

### Antibody ELISA's with Recombinant GRA7 and Native TLA Antigens

As dense granule protein 7 (GRA7) is considered as a marker of an active infection, being expressed by all T. gondii stages, recombinant GRA7 is frequently used to demonstrate the immune response during acute and chronic toxoplasmosis in humans and animals (Jacobs et al., 1999). GRA7 was prepared as previously described (Jongert et al., 2007). Briefly, GRA7 was produced as a His-tagged fusion protein by Escherichia coli (E. coli) TOP 10 cells (Life Technologies, Ghent, Belgium) and purified under denaturing conditions (8 M urea, 0.1% SDS) using nickel-nitrilotriacetic acid (Ni-NTA) chelate affinity column chromatography (Ni-NTA Superflow, Qiagen, Venlo, The Netherlands). GRA7 was then eluted from the Ni-NTA column using 250 mM imidazole and further purified by sequential dialysis steps reducing the urea and SDS concentration to 0.1 M and 0.01%, respectively.

T. gondii total lysate antigen (TLA) from tachyzoites of the RH-strain was prepared as previously described (Jongert et al., 2007) in the biosafety level 2 laboratory (Biosafety permit no, 415240), as approved by the Ethical Committee (no. 20140704- 01) at the National Reference Laboratory for Toxoplasmosis (Scientific Institute for Public Health, Brussels, Belgium). TLAbased assays show a high reactivity due to a broad range of antigens in the lysate, however, differences in the production method can affect the composition of the lysate (Gamble et al., 2005; Ferra et al., 2015). Concisely, tachyzoites were diluted with PBS and then purified by differential centrifugation and filtration through a 5 µm syringe filter (MilleX <sup>R</sup> SV, Merck KGaA, Darmstadt, Germany). The tachyzoite suspension was


Groups, a single low or high dose of the IPB-G strain (Glow and Ghigh); a single low or high dose of the IPB-LR strain (LRlow and LRhigh); a high dose of the IPB-G strain, followed 60 dpi by a high dose of the IPB-LR strain (Ghigh/LRhigh); a high dose of the IPB-G strain 60 dpi (Ghigh1/2t); a high dose of the IPB-LR strain, followed 60 dpi by a high dose of the IPB-G strain (LRhigh/Ghigh).

then lysed by alternating sonication with cooling cycles using an Ultrasonic disintegrator (MSE, Leicester, United Kingdom). To evaluate the protein content of the lysate, the bicinchoninic acid (BCA) reaction (Thermo Scientific Pierce BCA protein Assay Kit, Erembodegem, Belgium) was used. Finally, the TLA was aliquoted and stored at −20◦C until further use.

Both TLA and GRA7 were used in indirect Enzyme-Linked Immunosorbent Assays (ELISA's) at 10 µg/ml to detect T. gondiispecific IgM and IgG antibodies in serum samples diluted 1/50 with the goat anti-pig IgM- and IgG-Horse Radish Peroxidase (HRP) conjugate (Bethyl Laboratories Inc., Montgomery, Texas, USA), respectively, and 2,2′ -azino-bis(3-ethylbenzothiazoline-6-sulphonic acid (ABTS) as substrate-chromogen solution (Verhelst et al., 2015). On each plate previously collected sera from one positive and three negative control animals as established by IgM and IgG immunofluorescence assay (IFA) were included and diluted 1/50 in dilution buffer (0.05% Tween-20 in PBS). The absorbance was measured at 405 nm (TECAN Spectra Fluor, Tecan Group Ltd., Männedorf, Switzerland) and the obtained data were analyzed in Microsoft Excel. Serum samples from infected animals were considered positive when exceeding the cut-off value calculated using the formula: mean OD<sup>405</sup> negative controls + 3 x its standard deviation (SD).

### Immunofluorescence Assay

The presence of IgM and IgG antibodies against T. gondii was also evaluated by IFA using slides coated with formalinefixed tachyzoites from the T. gondii RH-strain (Toxo-Spot IF, Biomérieux, Marcy-l'Etoile, France). Briefly, serum samples, diluted 1/50 in PBS, were applied to the slides for 30 min at 37◦C, followed by washing with PBS. After drying, a second incubation with fluorescein isothiocyanate (FITC)-conjugated goat anti-swine IgM(µ) or IgG (H+L) (KPL, Maryland, USA) antibody (diluted 1/25 in PBS with Evans Blue as counter dye) was performed for 30 min at 37◦C. After washing, drying and mounting with PBS-buffered glycerol, the slides were observed by fluorescence microscopy (Carl Zeiss, Germany). The cut-off read-out was established with positive and negative reference sera at a 1/50 dilution.

### Detection of the Cellular Immune Response

PBMCs were isolated from 20 ml heparinized blood (LEO Pharma, Ballerup, Denmark) by density gradient centrifugation (800 × g at 18◦C, 25 min) using LymphoprepTM (Axis-Shield, Oslo, Norway) (Sonck et al., 2010). Subsequently, the cell pellets were resuspended in leukocyte medium [RPMI-1640 (GIBCO BRL, Life Technologies, Merelbeke, Belgium), supplemented with fetal calf serum (10%) (Greiner, Bio-One, Merelbeke Belgium), non-essential amino acids (100 mM) (Gibco), Na-pyruvate (100 µg/ml), L-glutamine (292 µg/ml) (Gibco), penicillin (100 IU/ml) (Gibco), streptomycin (100 µg/ml) (Gibco), and kanamycin (100 µg/ml) (Gibco)]. The cells (10<sup>6</sup> cells/well) were cultured for 6 and 72 h upon stimulation with either TLA (10 µg/ml) as a heterologous challenge or the mitogen concanavalin A (ConA, Sigma-Aldrich, USA; 5 µg/ml) as a positive control.

### Cytokine mRNA Quantification by RT-qPCR

After 6 h of incubation with TLA, ConA or medium, the cells were lysed by adding 350 µl of RLT-buffer (Qiagen) supplemented with 1% β-mercaptoethanol (99%, Thermo Fisher Scientific, Aalst, Belgium) and stored at −80◦C until RNA isolation. Total RNA extraction and conversion into cDNA was performed using the RNeasy kit (Qiagen) and the iScript kit (Biorad, Hercules, CA, USA), respectively. The purity of the RNA was assessed by an on-column DNase digestion step as recommended by the supplier. The amount of cytokine cDNA was then tested by quantitative polymerase chain reaction (qPCR). The qPCR reaction mix consisted of 12.5 µl iQ SYBR Green Supermix (Biorad), 0.5 µl of each primer set at a concentration of 20 µM, 1.5 µl PCR grade water and 10 µl of the 1/100 diluted cDNA. Interleukin (IL)-10, IL-12A, IL-17A, and interferon-gamma (IFN-γ) cDNA was amplified with the primer

sets presented in **Table 2**. In order to normalize the cytokine expression, β-actin, glyceraldehyde phosphate dehydrogenase (GAPDH) and the ribosomal 18S gene were used as reference genes (**Table 2**). Special care was taken to choose a set of primers on different exons or spanning exon-exon junctions to exclude the amplification of genomic DNA. The qPCR amplification protocol consisted of an initial denaturation at 95◦C for 3 min, followed by 45 cycles of 95◦C for 15 s and 61◦C for 20 s. After each run, a melt curve analysis was performed to confirm the presence of the correct amplicon and to exclude false positives due to the formation of primer dimers. The cDNA was tested in duplicate for each cytokine and the three reference genes (GADPH, β-actin, r18S), showing a stable expression. The mRNA expression in PBMCs was calculated with the CFX96 ManagerTM Software v3.1 (Biorad), using a mathematical model (delta-delta Ct method). The mean value was determined for the target cytokines and normalized relative to the geometric mean of the reference genes (Verhelst et al., 2015).

### Flow Cytometric Detection of IFN-γ Production

The flow cytometric detection of IFN-γ-producing proliferating lymphocytes was performed on cultured PBMCs 72 h after heterologous stimulation with TLA (10 µg/ml). First, the cell division marker Violet Proliferation Dye 450 (VPD450, BD Biosciences, Erembodegem, Belgium) was added to the isolated mononuclear cells, showing a diminishing fluorescence after each cell division. At the end of the incubation period, a protein transport inhibitor, Golgi PlugTM, was added and the cells were fixed and permeabilized using the Cytofix/CytopermTM kit (both from BD Biosciences). Subsequently, cells were stained using murine monoclonal antibodies (Mab) against CD3 (IgG1, clone PPT3), CD4 (IgG2b, clone 72–14-4), and CD8 (IgG2a, clone 11/295/33) and anti-isotype-specific conjugates (goat anti-mouse IgG1-PerCP-Cy5.5; Santa Cruz Biotechnology, Dallas, Texas, USA), goat anti-mouse IgG2b-FITC (Southernbiotech, Birmingham, Alabama, USA) and goat anti-mouse IgG2a-Alexa Fluor <sup>R</sup> 647 (InvitrogenTM, Merelbeke,

TABLE 2 | List of primers for qPCR.


Belgium). Finally, phycoeryrthrin (PE)-conjugated mAb against porcine IFN-γ (Mouse IgG1, BD Biosciences) was added to identify the lymphocyte subsets producing IFN-γ. A minimum of 10,000 events was recorded within the proliferating cell gate (Appendix 1). The IFN-γ secretion in the different lymphocyte subsets was determined and compared with the results of the isotype-matched control (Mouse IgG1-PE, Abcam, Cambridge, UK) using a FACSAria III and FACSDIVATM software (both from BD). The gating strategy is included in the supplementary data (Appendix 1).

Animals were euthanized at 98 dpi and the splenocytes and lymphocytes from the peripheral lymph nodes (mediastinal, mesenteric and popliteal) were isolated as previously described (Verhelst et al., 2011). Subsequently, the cells were stimulated with the same antigens as the PBMCs for 6 h and 72 h, whereafter the same staining occurred for flow cytometric analysis as for the PBMCs.

### Detection of the Parasite: Bio-Assay and qPCR

In experiments 1 and 2, all animals were euthanized at 120 dpi and the parasite load was determined in brain (Br), heart (He), spleen, diaphragm (Di) and skeletal muscles (m. gastrocnemius (Mg), mm. intercostales (Ic), m. longissimus dorsi (Ld), and m. psoas major (Mp) by qPCR and a bio-assay. For this, 100 g of each tissue was homogenized in 10 ml 0.85% sodium chloride (NaCl) and digested with pepsin [0.8 g/l pepsin in 7 ml/l hydrogen chloride (HCl)] for 1 h for brain and 2 h for the other tissues, while stirring in a water bath at 37◦C. The obtained suspension was filtered and centrifuged for 15 min at 1,180 × g, the supernatant removed and the pellet resuspended in 10 ml PBS supplemented with 40 IU/ml gentamicin. For the bio-assay, 1 ml of the tissue suspension was inoculated intraperitoneally into 5 naive Swiss female mice. The mice were observed frequently on a daily base for the next 5 weeks and euthanized in respect to the human end points in case of acute toxoplasmosis associated with suffering or reduced welfare. The surviving mice were euthanized and tested serologically by immunofluorescence for the presence of T. gondii IgG antibodies or by qPCR for lungs and ascites when pre terminated due to the ethical aspects in case of acute toxoplasmosis. To determine the parasite load by qPCR DNA was extracted from the tissue suspensions with the QIAamp DNA Mini kit (Qiagen). A 10-fold serial dilution of T. gondii DNA prepared from RH-strain tachyzoite suspension containing 10<sup>6</sup> parasites per ml was used as a standard, with a detection limit of 2–4 tissue cysts per 100 g of tissue. Real-Time PCR (RT-PCR) amplifying both the T. gondii repeat element (AF146527) and the ribosomal 18S rDNA of the host cells was performed as previously described (Rosenberg et al., 2009).

### Statistics

The parasite-specific antibody and IFN-γ responses in different groups at different time points are presented as means ± SD. A one-way Analysis of Variance (ANOVA) was performed, followed by post hoc Bonferroni's and Dunnett's Multiple Comparison Tests for antibody production and cytokine response, respectively, to discriminate between infected and control groups (GraphPad Prism 5). A p <0.05 was considered statistically significant.

### RESULTS

### Parasite Burden and Immune Response after Single Inoculation with a Low or a High Infection Dose of the IPB-G or IPB-LR Strain

#### GRA-7 and TLA-Specific Antibody Response

The GRA7-specific IgM antibodies appeared approximately at the same time in the low and the high dose group (10 and 9 dpi, respectively) upon inoculation with the IPB-G strain and declined gradually from 14 and 11 dpi, respectively, until 91 dpi (**Figure 2A**). In contrast, in the IPB-LR infected group a pronounced IgM production was detected 8 dpi in the low dose group and even a stronger response at 10 dpi in the high dose group (**Figure 2B**), but both declined to control levels around 12 dpi. GRA7-specific IgG antibodies were detected shortly after IgM, irrespective of the inoculation strain, and remained detectable until the end of the experiment (**Figures 2C,D**). Nevertheless, the high dose of the IPB-LR strain induced the highest levels of GRA7-specific IgG.

TLA-specific IgM occurred earlier than GRA7-specific IgM, namely 7 to 8 dpi in the Glow and Ghigh groups. In the latter, the IgM response remained present until the end of the experiment, slightly increasing in time, irrespective from the infection dose (**Figure 3A**). On the contrary, for the LRlow and LRhigh groups the seroconversion to TLA-specific IgM was prominently present until 21 to 28 dpi, showing again the highest concentration in the high dose group (**Figure 3B**). The TLA-specific IgG antibodies appeared approximately 14 dpi in both dose groups inoculated with the IPB-G strain (**Figure 3C**), but already at 8 dpi in animals infected with the IPB-LR strain. There, the antibodies increased significantly starting from 28 dpi and remained elevated until

120 dpi (**Figure 3D**). In animals inoculated with the IPB-G strain no dose effect was neither seen for TLA-specific IgM nor for IgG production (**Figures 3A,C**), whereas the high dose induced a higher response for both IgM and IgG upon inoculation with the IPB-LR strain (**Figures 3B,D**). The IFA results confirmed the seroconversion from T. gondii - negative toward IgM positive animals and the persistence of the IgG antibodies in each infection experiment (data not shown).

### TLA-Specific IFN-γ mRNA Responses in PMBCs

PBMCs were restimulated in vitro with TLA for 6 h, where after IL-10, IL-12A, IL-17A, and IFN-γ mRNA responses were determined. No detectable IL-10, IL-12, and IL-17A mRNA production was observed (data not shown) in any infected group, irrespective of the strain or infection dose. However, a substantial increase in IFN-γ mRNA production was observed from 1 month post infection (mpi) onwards in the majority of the inoculated animals as compared to the control animals. This response was least pronounced in the animals infected with the low dose of the IPB-G (not significant, p = 0.39) (**Figure 4A**), followed by a significant (p < 0.01) and highly significant (p < 0.001) increase in the high dose of the IPB-G strain group (**Figure 4B**). In the low dose of the IPB-LR group we noticed a steady though not significant (p = 0.18) increase (**Figure 4C**). The highest IFNγ production was observed in the animals infected with the high dose of the IPB-LR strain starting from 1 mpi (p < 0.01), which stayed high throughout the experiment, becoming highly significant (p < 0.001) at 2, 3, and 4 mpi (**Figure 4D**). No detectable IFN-γ level was detected in splenocytes from IPB-G or IPB-LR infected animals (data not shown).

### Parasite Load in Tissues

At 120 dpi, the parasite load was determined in heart and striated skeletal muscles by qPCR and a bioassay, while in brain by qPCR

only. In animals infected with the IPB-LR strain, the highest parasite load was found in brain followed by heart (**Table 3**). Interestingly, the inoculation dose did not affect the distribution or load in the tissues. Besides brain and heart, also intercostal muscles and the longissimus dorsi were consistently positive in qPCR, whereas heart of all 6 animals was also positive in the bioassay.

A different pattern was seen in pigs inoculated with the IPB-G strain (**Table 3**), where a clear effect of the inoculation dose on the parasite distribution and load in the tissues was found. When inoculated with the low dose, the parasite was present in more tissues and in higher amounts than when inoculated with the high dose. However, even when inoculated with the low dose, the longissimus dorsi and psoas major remained negative in all three animals in this group. Summarizing, animals inoculated with the high dose of IPB-G showed the lowest amount of T. gondii DNA in their tissues. Brain, gastrocnemius and the longissimus dorsi were negative, whereas for diaphragm and psoas major only one sample was positive in the bioassay and qPCR, respectively. These results strongly suggest a dose-dependent decreased burden of the IPB-G strain in the examined tissues following inoculation, pointing toward an immune-mediated reduction of the parasite load.

### Parasite Tissue Load and Immune Response in a Subsequent Infection Model with Two *T. gondii* Strains

In order to assess if the low parasite load observed in some tissues after infection with the IPB-G strain was related to an immune response, in a second experiment animals were first infected with the high dose of one strain, followed 60 days later with the high dose of the other strain (**Table 1**). Since we hypothesized an effect of the inoculation with the IPB-G strain, an additional control group was included in the study inoculated with the high dose at 60 dpi and 60 days before euthanasia.

#### GRA-7 and TLA-Specific Antibody Response

As in the first experiment, inoculation with the IPB-LR strain induced higher GRA7- and TLA-specific IgM production than with the IPB-G strain, independently from the order of inoculation (**Figures 5A,B**). This was most pronounced for the TLA-specific IgM response (**Figure 5B**). The presence of a clear TABLE 3 | Parasite load (number of bradyzoites per 1E+08 cells) by qRT-PCR after inoculation with two T. gondii strains, in comparison with bio-assay (number of positive/total tested).


Groups, a single low or high dose of the IPB-G strain (Glow and Ghigh) 120 dpi; a single low or high dose of the IPB-LR strain (LRlow and LRhigh); a high dose of the IPB-G strain, followed 60 dpi by a high dose of the IPB-LR strain (Ghigh/LRhigh); a high dose of the IPB-G strain 60 dpi (Ghigh1/2t); a high dose of the LR strain, followed 60 dpi by a high dose of the IPB-G strain (LRhigh/Ghigh). Tissues: Br, brain; Ha, heart; Di, diaphragm; Ic, intercostal m.; Mg, gastrocnemius m.; Ld, long. dorsi m.; Pm, poas major m.; nt, not tested.

TLA-specific IgM response after the second inoculation with the IPB-LR strain was remarkable and suggests differences in antigen expression between both strains (higher immunogenicity, different antigens or other reasons), leading to the induction of a primary immune response against various TLA antigens. Interestingly, the increase in TLA-specific IgM levels in the IPB-G infected animals upon initial inoculation at day 60 (Ghigh1/2t) was higher in comparison with Ghigh or Glow groups (**Figure 3A**), suggesting maturation of the immune system. Similar to our findings of the first experiment (**Figures 2C,D**, **3C,D**), the GRA7- and TLA-specific IgG antibodies appeared within 2 weeks following the primary inoculation with the IPB-G or IPB-LR strain (**Figures 5C,D**). A pronounced booster response against GRA7 occurred upon the heterologous challenge at 60 dpi in both re-infected groups, as evidenced by a much faster increase in IgG levels in contrast to animals from the Ghigh1/2t group (**Figure 5C**). The TLA IgG was not boosted following the heterologous infections in both challenged groups (**Figure 5D**). However, these distinct IgG responses were more pronounced by the challenge with the IPB-LR strain than with the IPB-G strain.

### TLA-Specific IFN-γ mRNA Responses in PMBCs and Spleen

IFN-γ responses after the initial inoculation with both T. gondii strains (**Figure 6**) were comparable with those in the first experiment (**Figure 4**). In two of the three animals receiving the IPB-G strain as a first inoculation (Ghigh/LRhigh), IFN-γ mRNA expression could not be detected 1 mpi in the PBMCs recall assay with TLA. However, from 2 mpi all three animals showed a significantly (p < 0.05) to highly significant (p < 0.001) increased IFN-γ mRNA level (**Figure 6A**), similarly to the expression seen in animals receiving the IPB-LR strain as a first inoculum (LRhigh/Ghigh) (**Figure 6B**), even though the latter showed a more homogenous response from 1 mpi onwards (p < 0.01). In both groups the IFN-γ mRNA expressions remained significantly elevated (p < 0.05 to p < 0.001) during the experiment and no additional increase was seen after inoculation with the heterologous strains. Slightly lower yet significant (p < 0.05) IFN-γ production was observed in group Ghigh1/2t, inoculated for the first time at 60 dpi (**Figure 6C**).

or ++ or xx or ## or " or ◦◦: <sup>P</sup> <sup>&</sup>lt; 0.01; \*\*\* or xxx or ### or "' or ◦◦◦; <sup>P</sup> <sup>&</sup>lt; 0.001; ns, not significant.

While no detectable cytokine production was found at 120 dpi in the spleen of animals from both infection groups in the first experiment (data not shown), in the heterologous infection model significant IFN-γ transcript levels were detected 60 days after the second infection for the Ghigh/LRhigh (p < 0.05), Ghigh1/2t (p < 0.01), and LRhigh/Ghigh (p < 0.001) (**Figures 7A,B**).

#### Parasite Load in Tissues

To assess the parasite load in tissues animals were euthanized 60 days after a first (Ghigh1/2t group) or second infection (Ghigh/LRhigh and LRhigh/Ghigh groups). Interestingly, animals receiving only the high dose of the IPB-G strain (Ghigh1/2t) showed a parasite distribution and load (13 tissue samples negative in a bioassay) in between those observed at 120 dpi with the IPB-G low dose (10 tissue samples negative in a bioassay) and the high dose (17 samples negative) in the first experiment (**Table 3**). This could suggest that the reduction in parasite load induced by the IPB-G strain was already appearing at 60 dpi upon inoculation with the high dose. The tissue distribution at 120 dpi in animals, which first received the IPB-LR high dose and 60 days later the IPB-G high dose, could also be explained by this phenomenon. At 120 dpi the animals showed a different parasite load and tissue distribution (12 samples negative) than in the first experiment (1 and 2 samples negative after infection with the low and the high dose, respectively). Animals receiving first the IPB-G high dose and 60 days later the IPB-LR high dose, showed a wider parasite tissue distribution and a higher parasite tissue load (7 samples negative), more comparable to animals receiving only the IPB-LR strain.

(LRhigh/Ghigh). Piglets infected with a high dose of the IPB-G strain at day 60 served as a control (C) (Ghigh1/2t). Cells were restimulated in vitro with TLA and IFN-γ mRNA was quantified with RT-PCR. The thick lines indicate the group mean. The significance level: \*P < 0.05, \*\*P < 0.01; \*\*\*P < 0.001.

orally infected with (A) a low dose of the IPB-G strain (Glow), a high dose of the IPB-G strain (Ghigh), a high dose of the IPB-G strain followed 60 days later by a high dose of the IPB-LR strain (Ghigh/LRhigh), a high dose of IPB-G at day 60 (Ghigh1/2t), or (B) a low dose of the IPB-LR strain (LRlow), a high dose of the IPB-LR strain (LRhigh) or a high dose of the IPB-LR strain followed 60 days later by a high dose of the IPB-G strain reversed infection model (LRhigh/Ghigh). Cells were isolated at 2 mpi (group: Ghigh1/2t) or 4 mpi (groups: Glow, LRlow, Ghigh, LRhigh, Ghigh/LRhigh, LRhigh/Ghigh) and restimulated in vitro with TLA. IFN-γ mRNA was quantified with RT-PCR. The thick lines indicate the group mean. The significance level: \*P < 0.05, \*\*P < 0.01; \*\*\*P < 0.001.

### The Involvement of CD4<sup>+</sup> and CD8<sup>+</sup> T Cells in the Strain-Dependent IFN-γ Production

Results of the above experiment supported our hypothesis that the IPB-G strain reduced the parasite burden. Since many studies suggested an important role for IFN-γ responses in controlling T. gondii infections, in a next experiment we compared the kinetics of IFN-γ producing T cell subsets in blood, following infection with the high dose of both strains in an in vitro TLA recall assay. Depending on the T. gondii strain, differences in the kinetics of circulating IFN-γ producing T cell subpopulations were observed (**Figure 8**). The CD3+CD4+CD8α <sup>−</sup>and CD3+CD4+CD8α dim represent porcine T-helper cells, while CD3+CD4−CD8α bright cells are cytotoxic T cells (Gerner et al., 2015). Animals inoculated with the IPB-LR strain showed at 21 dpi a significant increase in the CD3+IFN-γ <sup>+</sup> T cell subsets (CD4+CD8α <sup>−</sup>, CD4+CD8α dim, and CD4−CD8α bright T cells), with the CD4+CD8α <sup>−</sup> T-helper cells (up to 22.2 ± 8.3% of CD3+IFN-γ producing cells) being most prevalent (**Figure 8A**). This latter population remained stable, whereas the CD4−CD8α bright population gradually increased from 9.3 ± 0.87% of the IFN-γ <sup>+</sup> producing cells at 28 dpi to > 40% (41.1 ± 17.4%) at 98 dpi (**Figure 8C**). In contrast, the percentage CD4+CD8α dimIFN-γ <sup>+</sup> cells gradually decreased from 16.4 ± 1.4% to <2.1 ± 0.8% at the end of the experiment (**Figure 8B**). In animals inoculated with the IPB-G strain, a similar increase in the percentage of CD4+CD8α dimIFN-γ + cells at 21 dpi to 10.4 ± 0.4%, and a subsequent gradual decrease to 2.2 ± 0.5% was seen (**Figure 8B**). The IFN-γ + within CD4+CD8α <sup>−</sup> and CD4−CD8α bright T cells gradually increased, reaching significantly higher levels (22.6 ± 11.9% and 18.7 ± 9.5%, respectively) at 84 dpi as compared to 0 dpi (**Figures 8A,C**). Intriguingly, the increase of the latter T cell population was clearly less pronounced in the Ghigh (p < 0.05) than in the LRhigh group (p < 0.01).

The difference in circulating CD4−CD8α bright T cell populations between both high dose groups seems to be reflected on the long term in the significantly higher (p < 0.05) percentage of CD4−CD8α bright IFN-γ <sup>+</sup> T cells in the popliteal lymph nodes (LN) in the LRhigh group than in the Ghigh group at 98 dpi (**Figure 9C**). Additionally, a similar difference in the percentage of the same population between both groups can be detected in the mesenteric LN. However, in mediastinal LN, which drain heart and diaphragm, a higher percentage of CD4−CD8α bright IFN-γ <sup>+</sup> T cells was found in the Ghigh group, although not significantly different from the LRhigh group. The distribution of the other T cell subpopulations in different lymphoid tissues shows a comparable pattern: a higher percentage of the CD4+CD8α <sup>−</sup>IFN-γ <sup>+</sup> T was found in the mesenteric (p < 0.05) and popliteal (p > 0.05) LN of the LRhigh group than in the Ghigh group (**Figure 9A**). Likewise for the CD4−CD8α bright T cells, a reverse situation was noticed in the mediastinal LN. Regarding the CD4+CD8α dimIFN-γ <sup>+</sup> cells, relatively low percentages were detected in both infected groups. The highest counts were found in the mesenteric and popliteal LN in the LRhigh group, followed by the mesenteric LN in the Ghigh group, whereas they were nearly absent in the other sampled LN (**Figure 9B**).

FIGURE 8 | The percentage of IFN-γ <sup>+</sup> T lymphocyte subsets in PBMCs after a single IPB-G or IPB-LR inoculation. IFN-γ <sup>+</sup> T lymphocyte subsets in PBMC's of pigs after oral infection with a high dose of the IPB-G (Ghigh) or the IPB-LR strain (LRhigh). Cells were restimulated in vitro with TLA and demonstrated by flow cytometry following triple staining for IFN-γ, CD3, CD4, and CD8 (A). IFN-γ <sup>+</sup> cell populations were identified as (A) CD3+CD4+CD8<sup>α</sup> <sup>−</sup>IFN-<sup>γ</sup> +, (B) CD3+CD4+CD8<sup>α</sup> dim IFN-γ <sup>+</sup>, (C) CD3+CD4−CD8brightIFN-<sup>γ</sup> + lymphocytes. The results represent mean percentages ± SD for each group; \* (IPB-G) or x (IPB-LR): P < 0.05, \*\* or xx: P < 0.01; \*\*\* or xxx: P < 0.001 in comparison with day 0.

#### FIGURE 9 | Continued

in vitro with TLA and demonstrated by flow cytometry following triple staining for IFN-γ, CD3, CD4, and CD8 (A). IFN-γ <sup>+</sup> cell populations were identified as (A) CD3+CD4+CD8<sup>α</sup> <sup>−</sup>IFN-<sup>γ</sup> <sup>+</sup>, (B) CD3+CD4+CD8<sup>α</sup> dim IFN-γ +, (C) CD3+CD4−CD8brightIFN-<sup>γ</sup> <sup>+</sup> lymphocytes. The results represent mean percentages ± SD for each group; \*P < 0.05, \*\*P < 0.01; \*\*\*P < 0.001.

### DISCUSSION

In the performed experiments, we compared the single or subsequent infection in pigs inoculated with either a high or a low dose of the IPB-G and the IPB-LR strains for the tissue specific parasite load and the accompanying immune response. The IPB-G strain has a mixed type I/II genotype, while the IPB-LR strain has a classic type II genotype (Jongert et al., 2008; Dubey et al., 2012). The antibody response against GRA7 and TLA, which are frequently used in serological assays in different species, was monitored until 120 dpi to confirm the successful inoculation and persistence of the infection (**Figures 2**, **3**). Overall, for both IgM and IgG, independent from the infection dose, the GRA7-specific antibodies were detected very soon after the initial infection, starting from 10 dpi (**Figure 2**). Similar to previous results (Verhelst et al., 2015), we detected a late TLA-specific IgG response from 28 to 35 dpi onwards (**Figure 3**). However, results of the present study demonstrated that strain and dose are important factors to consider, since primary GRA7 and TLA-specific antibody responses could be detected earlier during infection and were higher upon inoculation with a high dose of the IPB-LR strain as compared to the IPB-G strain. A low dose on the other hand resulted in a later and less prominent seroconversion. Interestingly, whereas primary antibody responses were comparable in the heterologous challenge model, a clear IgM response was seen after the challenge with the heterologous strains, indicating exposure to other antigens, presumably due to the genetic diversity of both strains (**Figure 5**). Burrells et al. (2015) described a significant TLA-specific IgG increase after challenge of pigs with the heterologous strain M4 upon inoculation with the S48 strain. Strikingly, the challenge was performed with oocysts, after an initial inoculation with tachyzoites, stressing the expression of related or identical variability antigens in correlation to the parasite stage and the strain.

Together with a robust humoral response following T. gondii infection, a strong innate and cellular immune reaction is well described in mouse and human models, involving several populations of immune cells as well as different activation pathways (Aliberti, 2005; Miller et al., 2009; Andrade et al., 2013; Gazzinelli et al., 2014; Sturge and Yarovinsky, 2014). To date, it is well known that innate immune cells (macrophages, dendritic cells (DC's) and neutrophils) are involved in the acute stage of the infection by triggering the myeloid-differentiation primary response protein 88 (MyD88) signaling pathway after uptake and intracellular recognition of the parasite by CC-receptor 5 (CCR5) or Toll-like receptor (TLR) 11 and 12 in mice; TLR7, 8, and 9 in human and TLR7 and 9 in other mammals like pigs (Miller et al., 2009; Andrade et al., 2013; Koblansky et al., 2013; Gazzinelli et al., 2014). In particular interferon regulatory factor 8 (IRF8)<sup>+</sup> dendritic cells, activated by the uptake of the parasite's protein profilin, are crucial for the induction of IL-12 secretion in mice. Human and porcine DCs and monocytes are activated by the recognition of the parasite's ssRNA and DNA via TLR7 and TLR9, respectively, prior to their pro-inflammatory cytokine response (Uenishi et al., 2012; Andrade et al., 2013).

Consequently and irrespective of the activated TLRs, the DCdriven IL-12 production leads to the activation of T-helper 1 cells and Natural Killer (NK) cells (Sturge and Yarovinsky, 2014). The latter massively produce IFN-γ, which not only continuously activates macrophages via a positive feedback mechanism, but also elicits the expression of the GTPases. The GTPases family includes four subfamilies: the very large inducible GTPases (VLIG), the Mx proteins, the immunity-related GTPases (IRGs) such as p47 or p65, and the guanylate-binding proteins (GBPs). The p47 IRG offers a robust protection against intracellular pathogens, being recruited to the parasite attachment site at the host cell (MacMicking, 2004; Taylor et al., 2004; Liesenfeld et al., 2011). Subsequently, a lethal damage to the parasitophorus vacuole (PV) is induced, leading to the rupture of the infected cell and release of the parasite into the cytosol (Gazzinelli et al., 2014). The infected cell undergoes necrosis, simultaneously with an enhanced local immune response. It is important to mention that IFN-γ-inducible IRGs are well studied in murine models, where 23 different genes have been identified to date, but the corresponding genes are not present in the human genome, which includes only one gene and one pseudogene (Bekpen et al., 2005; Zhao et al., 2009). This implies that both species deploy other intracellular defense mechanisms against T. gondii. The data on the identification of porcine GTPases are scarce, but a high similarity to the human IRGs is mentioned (MacMicking, 2004). Only two porcine GBPs have been reported until now: GBP1 and GBP2, whereas in humans 7 different GBPs have been identified (MacMicking, 2004; Li et al., 2016). More IRG's have been found using Affymetrix GeneChip <sup>R</sup> Porcine Genome Array but a detailed study in pigs is lacking (Fossum et al., 2014).

Thus, the continuous production of IFN-γ seems to be necessary in maintaining a delicate balance between the host immune system and the parasite's evasion strategies. Additionally, this cytokine plays a pivotal role in controlling both the acute and chronic phase of infection, as it facilitates stage conversion from the tachyzoite to the bradyzoite in acute toxoplasmosis and suppresses the opposite conversion during chronic infection (Denkers, 1999). Likewise, we detected a significant increased IFN-γ production by PBMC's after inoculation with two different T. gondii strains (**Figure 4**), which corroborates our previous results when inoculating pigs with the IPB-G strain (Verhelst et al., 2015). Here, we demonstrated a time- and dose-dependent increase in IFN-γ mRNA expression upon infection with the IPB-G strain. Several studies focused on experimental infection in pigs reported a time-dependent increase of IFN-γ levels in serum, supernatant from cultured PBMCs and IFN-γ mRNA expression in PBMCs and intestinal lymphoid tissues (Solano Aguilar et al., 2001; Dawson et al., 2004, 2005; Verhelst et al., 2015).

On the contrary, the inoculation with the low dose of the IPB-LR strain was almost as potent in inducing a relatively fast and strong IFN-γ production by PBMCs as the high dose of the same strain, which resulted in high IFN-γ mRNA levels at already 2 mpi, that were maintained until 120 dpi. Interestingly, the IFN-γ mRNA production in the LRhigh dose group did not show any increase over time, implying reaching the maximum capacity from 1 mpi onwards. In line with our findings, IL-12 (IL-12p35 and IL-12p40) mRNA expression was not detected in PBMCs in the acute phase of the infection (7 and 14 dpi) in an earlier study in pigs (Dawson et al., 2005).

Based on the results of the detection of IFN-γ and the parasite DNA in tissues (**Table 3**), a balance between the host defense mechanisms and the invasion of the parasite was probably established soon after inoculation with the IPB-LR strain. Namely, the high IFN-γ production during the infection study was associated with the high counts of parasite DNA in the animal tissues. On the contrary, in the IPB-G groups IFN-γ production was elevated in the late phase of the infection and resulted in a very low to undetectable parasite load in the tissues, implying that high IFN-γ levels can tip the balance in favor of the host. Intriguingly, based on our observations and unpublished data from the acute infection model with the same strains, we speculate that exposure to a high dose of the IPB-G strain is more effective in activating innate immunity than the low dose of the same strain or the inoculation with the IPB-LR strain. The strain-dependent differences in the IFN-γ production profile may result from the genetic and thus, biological features of the used strains, indicating expression of variable virulence factors toward the intermediate host. In our opinion and according to others (Hunter and Sibley, 2012), the IPB-LR as genotype II strain, activates other pathway than the IPB-G strain, which is of an atypical genotype (mixed genotype I and II). The virulence factors initiating a pathway would be GRA15 (via NF-κB) for the former, and ROP18 (via STAT3/STAT6 pathway) for the latter. Consequently, the IPB-LR induces Th1 type of protective immunity and remains persistent in the chronic phase. When looking at the IPB-LR infected groups, no substantial difference was noticed in the IFN-γ production pattern or in the amount of T. gondii DNA at the end of the experiment. In our view, the IPB-LR strain does not show an acute virulence but it is adapted to persist within the intermediate host and, as such, increase own survival. To support these speculations, we observed a much milder clinical manifestation upon inoculation with the IPB-LR strain (Jennes et al., unpublished data). We further hypothesize that resistance to the chronic infection in IPB-G model results from the high acute virulence and the subsequent fast elimination of the tachyzoites before they can successfully multiply and disseminate. Consequently, fewer parasites can survive the initial parasitemia, which eventually will lead to reduced numbers of cysts in the tissues. Importantly, we conducted the studies in a homogenous pig population in order to exclude host diversity; however, as the strains are maintained by serial passage in mice, their virulence might be altered compared to the original isolate. Therefore, it is tempting to speculate that the high IFN-γ production together with the lower parasite counts in the porcine tissues originate from a coevolution toward host tolerance and

reduced virulence, as suggested earlier by others (Gazzinelli et al., 2014). Furthermore, looking at the total IFN-γ expression following reinoculation with the heterologous T. gondii strain in the second experiment, no obvious difference between the groups could be observed (**Figure 6**). In the Ghigh/LRhigh group we detected an initial increase, which was followed by a steady decrease after the challenge. The IFN-γ production profile in the reversed infection model (LRhigh/Ghigh) supports our previous findings, showing a constant IFN-γ detection over time. Interestingly, in some animals basal or low level of cytokine mRNA were detected at 1mpi, followed by a substantial increase at the later time points, similar to the single infection experiment (**Figures 4A**, **6A**). However, the final IFN-γ concentration at the end of the experiment upon a heterologous challenge was 10 times lower than after a single high dose inoculation. We could speculate that the primary infection with a mixed genotype I/II strain, characterized by a high acute virulence and long-term STAT3 and STAT6 activation, partially modulates the immune response upon the challenge with genotype II strain. As the result, the initial impairment of the Th1 response after the challenge leads to a lower than in a single infection model IFN-γ production, and elimination of a certain fraction of the parasites. Consequently, a reduced amount of the tachyzoites disseminate to convert into bradyzoites. The latter has been shown by a lower parasite load 60 dpi challenge than in IPB-LR experiment (**Table 3** and **Figure 6**).

In regard to the involvement of immune cells in controlling the parasite's dissemination to the tissues and the chronic phase of toxoplasmosis, different populations seem to play a role. As described earlier and analogous to the acute infection stage, the production of IFN-γ is gradually taken over from the innate immune cells by T lymphocytes (Guan et al., 2007). Experimental infections in mice (Jongert et al., 2010; Suzuki et al., 2012) demonstrated the importance of CD4<sup>+</sup> and CD8<sup>+</sup> IFN-γ producing T cells in maintaining a chronic T. gondii infection, but the exact contribution of each subset remains unknown. Miller et al. (2006) describes higher production of IFN-γ by murine CD4<sup>+</sup> cells upon in vitro stimulation by infected macrophages or by TLA, but admits that the higher protective potential against dissemination of the parasite by CD8<sup>+</sup> or CD4<sup>+</sup> lymphocytes is not simply expressed by the amount of this cytokine. Indeed, it seems that IL-4 and IL-10 cytokines, produced by CD4<sup>+</sup> lymphocytes in addition to IFN-γ, might down regulate this protective capacity against the parasite. In line with that, due to their IFN-γ-independent cytolytic activity, the role of primed CD8<sup>+</sup> T cells in the host's immunity during chronic toxoplasmosis has been widely acknowledged (Wang et al., 2005; Suzuki et al., 2012; Sa et al., 2013). In pigs, only a few experiments identified CD8<sup>+</sup> and CD4+CD8<sup>+</sup> cells in the acute phase of the infection as the major source of the IFN-γ production (Solano Aguilar et al., 2001; Dawson et al., 2005). The additive or synergistic effect of CD4<sup>+</sup> T cells on the activity of the CD8<sup>+</sup> T cell population should not, however, remain neglected. In our study, regardless of the strain, the CD4−CD8α bright T cell subset contained the most IFN-γ positive cells, followed by the CD4+CD8α <sup>−</sup> subset, whereas the CD4+CD8α dim T cell subset showed very few IFN-γ positive cells (**Figure 8**). Additionally, the CD4−CD8α bright population showed a temporal increase in IFN-γ production in animals infected with IPB-LR, while the percentage of this subset was rather declining from 4 wpi onwards, when infected with the IPB-G. The IFN-γ production resulting from the induced toxoplasmosis in pigs and the involvement of the different lymphocyte populations are in line with other studies, where the in vitro cytokine profile was investigated until 14 (Dawson et al., 2005), 40 (Solano Aguilar et al., 2001), or 56 dpi (Verhelst et al., 2011). However, opposite to the pig model, in murine experiments only two T lymphocyte subsets were differentiated (CD4<sup>+</sup> and CD8+). Furthermore, the extent of the cellular response was positively correlated with the infection dose and the time-interval from the inoculation and was higher when induced by the strain with a greater tissue persistence. In perspective of future experiments in pigs, it would be desirable to focus on other immune cells, involved in the responses throughout the infection such as the immunosuppressive T regulatory (Treg) or Th17 cells. As recently shown in mice, the robust immune reaction expressed by the high IFN-γ levels in the acute phase of the infection severely reduces the activity of Tregs in a IL-2 dependent and IL-10 independent manner (Tenorio et al., 2011; Olguin et al., 2015). In a longitudinal clinical case study of human acquired cerebral toxoplasmosis a dual function of the Treg population was described, by simultaneous down regulation of CD4<sup>+</sup> and activation of pathogen-specific CD8<sup>+</sup> T lymphocytes (Rb-Silva et al., 2017). In human congenital infections not only CD4<sup>+</sup> Treg cell population seems to be involved in the immune reaction triggered by T. gondii, but also a different subset of CD4<sup>+</sup> or CD8+, namely Th17. The activity of this population is independent from IFN-γ, IL-4 and perforin activation, as their migration to the inflammation sites in initiated by certain chemokines. Interestingly, the results of the in vitro PBMCs stimulation with tachyzoites showed a higher percentage of CD4<sup>+</sup> IL17 producing cells above the CD8<sup>+</sup> (Silva et al., 2014). By investigating the fluctuations of the activity of Tregs and Th17 cells during the acute and chronic phase of the infection, we might determine whether these populations are also involved in the persistent immunity toward the parasite.

When considering the parasite load in the tissues and the viability of the cysts in the bio-assay, a clear correlation was found between the amount of detected DNA and the dose of the used strain (**Table 3**). In general, we observed a decline in the concentration of the parasite's DNA in animals when inoculated with a high dose of the IPB-G, but not the IPB-LR strain. This dose- and time-dependent decline is prominently present in different tissues in the Ghigh group in comparison with Glow and Ghigh1/2t groups, indicating that the effect of the high dose is particularly visible after a longer infection time. These results are in line with the findings of Verhelst et al. (2011), where neither parasite DNA nor viable parasites were detected in certain muscle tissues 6 months after initial infection with the IPB-G strain. In the same study, brain and heart of all animals remained infectious as determined by a bio-assay and qPCR. However, these findings are opposite to results obtained in rats and cats, showing that inoculation with increasing amounts of tachyzoites or bradyzoites resulted in a decreased survival rate or in a higher number of tissue cysts, respectively (De Champs et al., 1998; Cornelissen et al., 2014). In addition and similar to our results, others reported a reduction in parasite burden in strain-vaccinated and challenged pigs (Kringel et al., 2004; Garcia et al., 2005; Jongert et al., 2008; Burrells et al., 2015). In these experiments vaccination with oligonucelotides, antigens or infection with attenuated strains can enhance Th1 responses to elicit sufficient protection during the acute phase of the infection, resulting in a lower parasite burden in comparison with the infected control animals. Referring to that, the strains used in our study differ greatly in terms of genetic background and associated virulence, both in mice and in pigs. Therefore, we have grounded scientific reasons to believe that the observed differences in the parasite load upon infection with both strains, especially in the heterologous co-infection model, are not coincidental.

For most tissue samples this dose- and strain-dependent reduction in the number of the tissue cysts in qPCR is consistent with the bio-assay in mice. It is noteworthy that a few qPCRpositive samples were negative in the bio-assay, indicating that parasite DNA, but not viable parasites were present. Consequently, these results show a substantially higher sensitivity of the qPCR method used here above the bio-assay, since it has been optimized and successfully applied for the detection of the parasite in various human or animal tissues, with a detection limit of 2–4 tissue cysts in 100 g of sample (De Craeye et al., 2011). Conversely, both techniques gave positive results on all samples derived from animals upon single inoculation with the IPB-LR strain. Therefore, we can assume that the reduced parasite load occurred due to the earlier infection with the IPB-G strain in the group Ghigh/LRhigh. This phenomenon does not seem to be limited to type II strains (Velmurugan et al., 2009; Suzuki et al., 2012), but is also common in type I strains (Burrells et al., 2015).

Summarizing, the groups infected with the IPB-LR strain can serve as a classical model of T. gondii persistence in its intermediate host. The prominent production of parasitespecific antibodies, consistent amounts of IFN-γ and activation of cytotoxic T lymphocytes on the one hand and welldistributed DNA concentration together with isolation of the viable parasite on the other hand, clearly prove an established balance between the host immunity and the pathogen's activity. The parasite's persistence appears to be beneficial for the two, under conditions that the immunocompetent host can resist the immunomodulation by T. gondii. On the contrary, as the partial or total removal of the tissue cysts was observed in the IPB-G infected animals together with the increasing IFN-γ production profile on both the mRNA and protein level, we propose that the IPB-G strain induces a robust immune reaction in the host in the early phase of the infection. This IFN-γ-mediated response in pigs can lead to the resistance of the host to parasite invasion

### REFERENCES

Afonso, E., Lemoine, M., Poulle, M. L., Ravat, M. C., Romand, S., Thulliez, P., et al. (2008). Spatial distribution of soil contamination by Toxoplasma gondii in relation to cat defecation behaviour in an urban area. Int. J. Parasitol. 38, 1017–1023. doi: 10.1016/j.ijpara.2008.01.004

by elimination of the tissue cysts during the chronic infection. Further experiments to unravel the nature of this resistance are warranted as it could serve an important role in vaccination strategies and in the risk assessment for food safety and human health.

### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the ethical standards defined by the EU legislation on the use of laboratory animals (2010/63/EU) and in accordance with the Belgian law (the Royal Decree 29/5/2013 and the Royal Decree 30/11/2001). The protocol was approved by the Ethical Committee of the faculties Veterinary Medicine and Bioscience Engineering at Ghent University (approval number no. 176 2009/149) and by the Ethical Committee of the WIV-ISP Institute (approval no. 176 20140704-01). The Biosafety Level 2 permit for working with pathogens for Ugent, Merelbeke: AMV/11062013/SBB219.2013/0145. The Biosafety Level 2 permit for working with pathogens for WIV-IPH, Brussels: 415240.

### AUTHOR CONTRIBUTIONS

MJ designed the study, acquired and analyzed the data and drafted the manuscript. SD designed the study, acquired, and analyzed data and revised the manuscript. BD helped to design the flow cytometry study and revised the manuscript. PD and KD helped to design the study, gave valuable input and revised the manuscript. EC designed the study, analyzed the data and wrote the manuscript.

### ACKNOWLEDGMENTS

This study was granted by the Belgian Federal Public Service for Health, Food Chain Safety and Environment (grant RF 09/6213). The authors wish to thank A. Leremans for the technical assistance with the bio-assay and M. Boutry for her excellence laboratory skills with qPCR and cytokine RT-PCR. We also wish to thank R. Cooman for the animal management. B. Devriendt is supported by an FWO postdoctoral research grant. The Hercules Foundation supported the purchase of research equipment (AUGE035).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fcimb. 2017.00232/full#supplementary-material


Sources of Zoonoses, Zoonotic Agents and Food-borne Outbreaks in 2010. The EFSA Journal 10:2597. 442. doi: 10.2903/j.efsa.2012.2597


case report with longitudinal analysis of T cell subsets. BMC Infect Dis. 17:66. doi: 10.1186/s12879-016-2159-x


encephalitis in genetically resistant BALB/c mice. J. Cytokine Res. 25, 338–344. doi: 10.1089/jir.2005.25.338


**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 Jennes, De Craeye, Devriendt, Dierick, Dorny and Cox. 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.

# Long-Term Temporal Trends of *Nosema* spp. Infection Prevalence in Northeast Germany: Continuous Spread of *Nosema ceranae*, an Emerging Pathogen of Honey Bees (*Apis mellifera*), but No General Replacement of *Nosema apis*

Sebastian Gisder <sup>1</sup> , Vivian Schüler <sup>1</sup> , Lennart L. Horchler <sup>1</sup> , Detlef Groth<sup>2</sup> \* and Elke Genersch1, 3 \*

#### *Edited by:*

Anton Aebischer, Robert Koch Institut, Germany

#### *Reviewed by:*

Katherine E. Roberts, University of Exeter, United Kingdom Robert John Paxton, Martin Luther University of Halle-Wittenberg, Germany

#### *\*Correspondence:*

Detlef Groth dgroth@uni-potsdam.de Elke Genersch elke.genersch@hu-berlin.de

*Received:* 23 February 2017 *Accepted:* 20 June 2017 *Published:* 06 July 2017

#### *Citation:*

Gisder S, Schüler V, Horchler LL, Groth D and Genersch E (2017) Long-Term Temporal Trends of Nosema spp. Infection Prevalence in Northeast Germany: Continuous Spread of Nosema ceranae, an Emerging Pathogen of Honey Bees (Apis mellifera), but No General Replacement of Nosema apis. Front. Cell. Infect. Microbiol. 7:301. doi: 10.3389/fcimb.2017.00301 <sup>1</sup> Department of Molecular Microbiology and Bee Diseases, Institute for Bee Research, Hohen Neuendorf, Germany, 2 Institute of Biochemistry and Biology, University of Potsdam, Potsdam-Golm, Germany, <sup>3</sup> Institut für Mikrobiologie und Tierseuchen, Freie Universität Berlin, Fachbereich Veterinärmedizin, Berlin, Germany

The Western honey bee (Apis mellifera) is widely used as commercial pollinator in worldwide agriculture and, therefore, plays an important role in global food security. Among the parasites and pathogens threatening health and survival of honey bees are two species of microsporidia, Nosema apis and Nosema ceranae. Nosema ceranae is considered an emerging pathogen of the Western honey bee. Reports on the spread of N. ceranae suggested that this presumably highly virulent species is replacing its more benign congener N. apis in the global A. mellifera population. We here present a 12 year longitudinal cohort study on the prevalence of N. apis and N. ceranae in Northeast Germany. Between 2005 and 2016, a cohort of about 230 honey bee colonies originating from 23 apiaries was sampled twice a year (spring and autumn) resulting in a total of 5,600 bee samples which were subjected to microscopic and molecular analysis for determining the presence of infections with N. apis or/and N. ceranae. Throughout the entire study period, both N. apis- and N. ceranae-infections could be diagnosed within the cohort. Logistic regression analysis of the prevalence data demonstrated a significant increase of N. ceranae-infections over the last 12 years, both in autumn (reflecting the development during the summer) and in spring (reflecting the development over winter) samples. Cell culture experiments confirmed that N. ceranae has a higher proliferative potential than N. apis at 27◦ and 33◦C potentially explaining the increase in N. ceranae prevalence during summer. In autumn, characterized by generally low infection prevalence, this increase was accompanied by a significant decrease in N. apis-infection prevalence. In contrast, in spring, the season with a higher prevalence of infection, no significant decrease of N. apis infections despite a significant increase in N. ceranae infections could be observed. Therefore, our data do not support a general advantage of N. ceranae over N. apis and an overall replacement of N. apis by N. ceranae in the studied honey bee population.

Keywords: honey bee, *Apis mellifera*, *Nosema* spp., epidemiology, replacement

### INTRODUCTION

The Western honey bee Apis mellifera is a valuable generalist pollinator for many flowering plants in both natural and agricultural ecosystems. In agriculture, commercial pollination of crop plants, that depend on insect pollination for fruit set and seed production, is provided mostly by managed A. mellifera colonies which can, therefore, be regarded as productive livestock. The cultivation of pollinator-dependent crops is expanding all over the world; hence, there is an increasing demand for insect pollination in worldwide agriculture (Aizen et al., 2008, 2009; Aizen and Harder, 2009). Although, this demand is partially met by a globally increasing number of managed honey bee colonies (Aizen et al., 2008, 2009; Moritz and Erler, 2016), increasing problems with honey bee health resulting in severe honey bee colony losses pose a serious threat to human food security. Research of the last decade has identified a multitude of factors like pathogens, pesticides, and abiotic stressors being associated with unusually high and inexplicable losses of honey bee colonies (Genersch, 2010; Ratnieks and Carreck, 2010; Cornman et al., 2012; Pettis et al., 2013; Goulson et al., 2015). Among the pathogens studied and discussed in this context are two microsporidian parasites, Nosema apis (N. apis) and N. ceranae, (Cox-Foster et al., 2007; Higes et al., 2008; Genersch, 2010) which infect adult honey bees (Bailey, 1955).

Microsporidia are highly specialized, spore-forming fungi which are optimally adapted to an obligate intracellular parasitic life style (Keeling and Fast, 2002). Outside of host cells, microsporidia exist as metabolically inactive, infective spores. For N. apis and N. ceranae, the infection process starts with the ingestion of infective spores by an adult honey bee. The spores germinate in the midgut thereby extruding the polar tube. If the polar tube pierces a host cell, the sporoplasm is injected into the cell through the polar tube (Bigliardi and Sacchi, 2001; Franzen, 2005). Following the injection of the sporoplasm, it takes about 96 h until the first environmental spores are produced by an infected cell (Gisder et al., 2011). The spores are released into the gut lumen through cell lysis and leave the body of the infected host by defecation (Bailey, 1955; Bailey and Ball, 1991). Heavy Nosema spp.-infections of adult honey bees may result in dysentery (Bailey, 1967). Adult bees suffering from diarrhea will show abnormal defecation behavior, i.e., will defecate inside the hive, resulting in fecal spots on combs and frames. Nest mates cleaning these spots will ingest Nosema spp. spores and become infected (Bailey and Ball, 1991). Infections with Nosema spp. are widespread in honey bee populations. Most infected honey bees do not develop nosemosis and do not show any obvious symptoms like dysentery but may have an increased foraging or flight activity (Woyciechowski and Kozlowski, 1998; Dussaubat et al., 2013) despite impaired orientation and homing skills (Kralj and Fuchs, 2010; Wolf et al., 2014) and may have a suppressed immune system (Antunez et al., 2009; Chaimanee et al., 2012), as well as a reduced life span (Wang and Moeller, 1970; Malone and Giacon, 1996; Fries, 2010).

Nosema apis-infections in honey bees have been studied intensively over the last 100 years and there is little debate on the rather low impact of this parasite on A. mellifera colonies (Bailey and Ball, 1991). However, the impact of N. ceranae-infections on colony health and survival is still controversially discussed (Higes et al., 2008; Genersch et al., 2010; Gisder et al., 2010; Guzman-Novoa et al., 2011; Stevanovic et al., 2011; Fernández et al., 2012). The emerging picture is that N. ceranae might cause colony death in warmer climates like Southern Europe (Higes et al., 2007, 2008, 2009; Martin-Hernandez et al., 2007; Botías et al., 2013; Cepero et al., 2014) whereas colony losses in Northern Europe or the Americas could not be associated with N. ceranae so far (Invernizzi et al., 2009; Genersch et al., 2010; Gisder et al., 2010; Williams et al., 2010; Guzman-Novoa et al., 2011) suggesting a climatic influence on N. ceranae virulence (Gisder et al., 2010) or differences in N. ceranae susceptibility between regionally predominating A. mellifera subspecies (Fontbonne et al., 2013; Huang et al., 2015).

Initially it was thought that N. apis is specific for the Western honey bee A. mellifera (Zander, 1909), while its congener N. ceranae was described as a microsporidian parasite of the Eastern honey bee A. cerana (Fries et al., 1996), a native of South- and Southeast Asia. Although, experimental infection showed from the very beginning that N. ceranae can also successfully infect A. mellifera (Fries, 1997), it took nearly a decade until the first natural infections of A. mellifera colonies with N. ceranae were reported (Higes et al., 2006; Huang et al., 2007). It soon became evident that N. ceranae was not only much more widespread than expected in the global A. mellifera populations but that is was even the predominant species in many regions (Klee et al., 2007; Chen et al., 2008; Williams et al., 2008; Invernizzi et al., 2009; Chen and Huang, 2010; Yoshiyama and Kimura, 2011; Copley et al., 2012). Based on this epidemiological evidence it was suggested that N. ceranae is replacing N. apis in the honey bee populations worldwide. This process is thought to be driven by an asymmetric within-host competition between N. apis and N. ceranae favoring the spread of N. ceranae (Williams et al., 2014; Natsopoulou et al., 2015) although not all studies observed interspecific competition between N. apis and its congener N. ceranae (Forsgren and Fries, 2010; Milbrath et al., 2015).

However, a pan-European study on the prevalence of N. apis and N. ceranae reported that in South-European countries, such as Italy and Greece, N. ceranae had indeed practically replaced N. apis while this was not observed in Northern Europe (Ireland, Sweden, Norway, and Germany) (Klee et al., 2007). These data pointed to climatic factors differentially influencing assertiveness, establishment, spread, and, hence, prevalence of N. apis and N. ceranae. Experimental evidence exists showing that N. ceranae spores, but not N. apis spores, nearly lose their ability to germinate and, hence, their infectivity when exposed to temperatures close to or below freezing (Fenoy et al., 2009; Fries, 2010; Gisder et al., 2010). In addition, experimental infection of adult bees showed proliferation of N. ceranae—but not of N. apis—to be unaffected by temperatures above 33◦C (Martin-Hernandez et al., 2009). These data strongly argue for an advantage of N. ceranae over N. apis in warmer climates. In contrast, the cold-sensitivity of N. ceranae spores might slow down the replacement process in colder climates (Gisder et al., 2010), a hypothesis that could recently be substantiated by mathematical modeling of the replacement process when taking into account the parameters warmer and colder climate (Natsopoulou et al., 2015). However, long term epidemiological data on Nosema spp. prevalence allowing the observation of the spread of the emerging pathogen N. ceranae and evaluating the proposed process of replacement of N. apis by N. ceranae in a given honey bee population have been lacking so far. To fill this gap, we here present our results of a 12 year cohort study on the prevalence of N. apis and N. ceranae in Northeast Germany conducted on a cohort of about 230 honey bee colonies. The duration of the study, and the size of the cohort enabled us to statistically analyse the long term temporal trends in prevalence of N. apis- and N. ceranae-infections in the study area. We also show data from laboratory experiments substantiating our epidemiological data. We provide evidence that the continuous spread of N. ceranae and continuously increasing levels of N. ceranae-infection prevalence at population level not necessarily result in the replacement of N. apis.

### MATERIAL AND METHODS

### Bee Samples, Field Survey and Molecular Differentiation of *N. apis* and *N. ceranae*

The data set on Nosema spp. prevalence comprises data from spring 2005 to autumn 2016, which were collected in the course of a 5 year longitudinal cohort-study on Nosema spp. epidemiology (Gisder et al., 2010) and of the still ongoing "German Bee Monitoring Project" (Genersch et al., 2010). About 23 apiaries located in Northeast-Germany (**Figure 1**) participated in the projects with 10 colonies ("monitoring colonies") each. Monitoring colonies that collapsed during the study were replaced by colonies from the same apiary, if available by a nucleus colony made from the collapsed colony in the previous year. This procedure ensured that each apiary always contributed 10 monitoring colonies throughout the study period. Due to the long duration of the study, some fluctuation of participating apiaries could not be avoided. However, nearly half of the apiaries (11 of ∼23) participated for more than 9 years and six of them even for the entire duration of the study, i.e., 12 years; at least 20 bee keepers provided samples over a time period of consecutive 5–11 years (**Figure 1**). When an apiary dropped out, a similar apiary in terms of size, bee race, landscape, region, and history of losses and diseases was chosen

as replacement and included in the study as soon as possible. This resulted in an annual mean of 22.67 ± 1.72 (mean ± SD) apiaries participating in spring and 24.0 ± 2.83 (mean ± SD) apiaries participating in autumn. All monitoring colonies were sampled twice a year, in spring and in autumn, resulting in a total of 5,600 honeybee samples collected and analyzed from the participating apiaries over the 12 year study period (**Table 1**).

number of years for which data are available for each apiary (yellow, 12 years;

green, 9–11 years; purple, 5–8 years; blue, 1–4 years).

Sampling of bees as well as diagnosis of N. apis and N. ceranae were performed essentially as already described (Gisder et al., 2010). Briefly, from each apiary, a group of 10 bee colonies [annual mean: 10 ± 0.31 (mean ± SD) colonies in spring and 10.01 ± 0.14 (mean ± SD) colonies in autumn] was randomly selected at the beginning of the study or when the beekeeper entered the study and designated "monitoring colonies." From these colonies, bee samples were collected in spring and autumn each year and were stored at −20◦C until analysis. Spring samples collected end of March/beginning of April consisted of dead bees fallen onto the bottom board during the winter season (representing the bees that died over winter) to enable sampling of colonies that collapsed during the winter season (October to March) as well as of surviving colonies. Autumn samples collected in late September/beginning of October consisted of live in-hive bees taken from a super above the queen excluder thus ensuring that only the oldest bees (representing the most frequently infected bees) were sampled (Fries et al., 2013). Diagnosis of Nosema spp. infections was performed by microscopic examination of 20 homogenized bee abdomens according to the "Manual of Standards for Diagnostics and Vaccines" published by the Office International des Epizooties (OIE), the World Organization for Animal Health (Anonymous, 2008). The moderate sample Gisder et al. Long-Term Temporal Trends of Nosema ceranae Infections

TABLE 1 | Prevalence of colonies infected with N. apis only (N. apis) or N. ceranae only (N. ceranae) or with N. apis and N. ceranae (co-infection) from spring 2005 to autumn 2016.


Given are the total number of analyzed colonies per year and season as well as the numbers (n) and proportions (%) of colonies within each infection category.

size is adequate because the experimental unit is the colony (Doull and Cellier, 1961; Doull, 1965). Infection status of the colonies represents detectable levels of infection above 15% with 96% probability of detection (Fries et al., 1984, 2013; Pirk et al., 2013) which can be considered biologically relevant (Higes et al., 2008). For molecular species differentiation, Nosema spp.-positive homogenates were processed and analyzed via PCR-RFLP (restriction fragment length polymorphism) as previously described (Gisder et al., 2010). Results were further verified by re-analyzing randomly selected samples via a recently developed differentiation protocol (Gisder and Genersch, 2013) which is based on the detection of speciesspecific sequence differences in the highly conserved gene coding for the DNA-dependent RNA polymerase II largest subunit. Based on the diagnostic results, four infection categories were defined: Microscopic analysis resulted in the category "Nosema spp." while molecular differentiation allowed for the categories "N. apis" (single infection), "N. ceranae" (single infection), and "co-infection" (infection with both N. ceranae and N. apis) (**Table 1**).

### Purification of *Nosema* spp. Spores for *In Vitro*-Infection

Honey bee colonies of the apiary of the Institute for Bee Research were screened for Nosema spp.-infections by microscopic analysis of 20 randomly collected adult bees (see above and Anonymous, 2008). Nosema spp.-positive samples were molecularly differentiated as previously described (Gisder and Genersch, 2013) to identify samples either containing only N. ceranae or only N. apis spores. Purification of N. apis or N. cerane spores was exclusively performed with freshly sampled bees, because freezing or long-term storage affect spore viability and infection rate (Fenoy et al., 2009; Fries, 2010; Gisder et al., 2010). Midguts were carefully isolated from individual bees by using fresh forceps for each bee. Twenty midguts were pooled in 1.5 ml reaction tubes and spore purification was performed as already described (Gisder et al., 2010).

Viability of the purified spores was checked via in vitrogermination. To this end, an aliquot of freshly isolated spores was air-dried onto glass slides for 30 min at room temperature. Germination was triggered by adding 20 µl of 0.1 M sucrose solution buffer directly to the dried spores. Germination process was analyzed under an inverse microscope (VWR, Darmstadt, Germany) at 400x magnification with phase contrast. Nosema spp. spores were counted in a hemocytometer (Neubauer-improved, VWR, Darmstadt, Germany) under an inverse microscope (VWR, Darmstadt, Germany) at 100x magnification. Only those spore preparations that were able to germinate under in vitro conditions were used for cell culture experiments.

### *In vitro*-Infection of Cultured IPL-LD-65Y Cells

The insect cell line IPL-LD-65Y derived from the gypsy moth Lymantria dispar was obtained from the Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ, Braunschweig, Germany) and maintained for routine culture as given in the accompanying data sheet. For in vitro-infection of cultured IPL-LD-65Y-cells, aliquots of about 5 × 10<sup>7</sup> spores, purified as described above, were dried in 1.5 ml reaction tubes (Eppendorf, Hamburg, Germany) in a vacuum concentrator (Eppendorf, Hamburg, Germany) for 30 min at 30◦C. Subsequently, infection of IPL-LD-65Y cells with germinating spores was performed as previously described (Gisder et al., 2011). Briefly, IPL-LD-65Y cells were infected with freshly isolated N. apis or N. ceranae spores with a multiplicity of infection (MOI) of 20. Infected cells (100 µl with 2.5 × 10<sup>5</sup> cells/ml) were seeded in the cavities of six 96-well microtiter plates. N. apis- and N. ceranae-infected cells were incubated at 21◦ , 27◦ , or 33◦C. Infected cells were centrifuged on glass slides at the time points 24, 32, 48, 72, and 96 h post initial infection and were subsequently Giemsa-stained as described (Gisder et al., 2011). The number of meronts, sporonts, and mature spores of N. apis or N. ceranae was counted under an inverse microscope Eclipse Ti-E (Nikon Instruments, Düsseldorf, Germany) at 600x magnification in 10 individual cells for each time point as well as for each temperature and expressed as mean ± SD.

### Statistical Analysis

For statistically analyzing the seasonality of Nosema infections, spring vs. autumn, the Wilcoxon signed rank test was used because the proportions of infected colonies (**Table 1**) were not normally distributed. In addition, the Spearman rank correlation was determined with R (version 3.2.5, R Development Core Team, 2016) to analyse the relationship between infection categories. The Spearman correlation coefficient determined the strength of the monotonic relationship between season and infection prevalence with effect sizes between 0.10 and 0.29 representing weak correlations, coefficients between 0.30 and 0.49 representing medium correlations, and coefficients of 0.50 or above representing strong correlations.

For each time point, the expected rate of co-infections (Eco−inf) was calculated as the product of the observed rates of single infections with either N. apis (Rapis) or N. ceranae (Rceranae): Eco−inf = Rapis ∗ Rceranae. Subsequently, the differences between the observed and expected rates of co-infections were calculated for each time point. Because those differences were normally distributed, a one sample t-test was used to check if these differences were significantly different to zero.

The statistical analysis of temporal trends was performed using RStudio (version 0.99.489) based on R using version 3.2.5. For visualizing infection prevalence data, dotplots were plotted with R, separately for spring and autumn. Generalized linear models (GLM) were fitted with lme4 (Linear Mixed-Effects Models, version 1.1-12) (Bates et al., 2015) for exploring the data set and visualizing the relationship between the dependent variables (Nosema spp.) and the independent variables (year). For statistical analysis of N. apis and N. ceranae prevalence over the 12 year study period we used mixed-effect binary logistic regressions analysis defining year as fixed factor and apiary as random factor to take into account the lack of independence of data within each apiary. Even after 12 years of sampling, the amount of data is still not sufficient to define colony as random factor to fully acknowledge relative data dependence. The sampling consisted of about 230 individual colonies per season, stratified within apiaries, and the prevalence of N. apis-, N. ceranae-, or coinfections at the individual level were analyzed with defining "0" if absent or "1" if present in each colony. Odds ratios (ORs) and 95% confidence intervals [CIs] were used to assess the strength of the associations.

For statistical analysis of the counted number of different developmental stages of Nosema spp. in infected IPL-LD-65Y cells, individual student's t-tests for each time point were performed followed by Benjamini-Hochberg correction (Benjamini and Hochberg, 1995). A p < 0.05 was considered significant for the statistical tests.

### RESULTS

### Prevalence and Seasonality of *Nosema* spp.-Infections

The huge data set on Nosema spp.-infection prevalence in Northeast Germany, which was generated during the 12 year longitudinal cohort study, provided a unique opportunity for a comprehensive analysis of the spread and success of Nosema spp., and especially of N. ceranae, in a restricted honey bee population. We first analyzed the seasonality of Nosema spp.-infections based on classical microscopic diagnosis without molecular species differentiation. The data revealed a clear and expected (Bailey and Ball, 1991) seasonality of Nosema spp.-infections for the whole duration of the study period with spring values being always higher than the autumn values of the same year and autumn values being always lower than the spring values of the following year (**Figure 2A**).

Molecular species differentiation of all Nosema spp.-positive samples enabled analysing the seasonality of N. apis-, N. ceranae-, and co-infections (**Table 1**). The same seasonality as already observed for Nosema spp.-infections was also evident for N. apisinfections over the entire study duration despite for the time point "spring 2007" when less colonies where found infected with N. apis than in the preceding autumn 2006 and the following autumn 2007 (**Figure 2B**). With this exception for "spring 2007," when only 3.1% of the colonies carried detectable N. apisinfections, the proportion of N. apis-infected colonies varied between 6.4% (spring 2013), and 18.7% (spring 2016). In autumn, the prevalence of N. apis-infected colonies ranged between 1.5% (autumn 2014) and 8.0% (autumn 2005).

For N. ceranae-infections, the described seasonality with higher prevalence in spring than in the following autumn and lower prevalence in autumn than in spring next year could be observed from autumn 2006 onward until spring 2016, whereas between spring and autumn 2016 the prevalence of N. ceranaeinfections did not decrease as expected but instead further increased (from 9.1 to 10.5%; **Figure 2B**). Spring prevalence from 2007 to 2016 varied for N. ceranae-infections between 7.6% (spring 2015) and 14.9% (spring 2007), while autumn prevalence ranged between 1.2% (autumn 2015) and 8.2% (autumn 2013).

The prevalence of colonies co-infected with N. apis and N. ceranae showed the same seasonal pattern fluctuating between spring (higher prevalence) and autumn (lower to no prevalence). Values for co-infection prevalence ranged between 1.6% (spring 2010) and 10.0% (spring 2008) in spring and between 0.0% (autumn 2008, 2009, 2010, 2012, 2014, 2015) and 2.0% (autumn 2011) in autumn (**Figure 2B**).

Statistical analysis of the seasonality of Nosema spp.-, N. apis-, N. ceranae-, and co-infections using a Mann-Whitney test confirmed the above given, rather descriptive evaluation (for all infection categories, p < 0.01). Spearman correlation analysis further substantiated this finding (**Figure 3**). A strong negative correlation (coefficient values between −0.69 and −0.87) was found between season and all infection categories indicating that in each year and for all four infection categories (Nosema spp.-, N. apis-, N. ceranae-, co-infection) the infection prevalence decreased significantly from spring to autumn (for all infection categories: p < 0.01). Medium to strong positive correlations (coefficient values between 0.44 and 0.85) were found between the infection categories implying that all infection categories followed the same prevalence trend. For example, high infection prevalence for N. apis correlated with high infection prevalence for N. ceranae- or co-infections. This correlation was significant for all infection categories (p < 0.05).

An interesting question in regard to co-infections was, whether or not the observed prevalence of co-infections in spring was congruent with the expected prevalence. To answer this question, we first calculated the rate of expected co-infections for each year from the rate of observed N. apis- and N. ceranaeinfections in this season. Comparing these values with the observed frequency of co-infections revealed that over the entire study period, the observed prevalence of co-infections was always significantly {one sample t-test; M = 0.037, [0.0195, 0.0546], t(23) = 4.6389, p < 0.01} higher than expected when assuming that the occurrence of co-infections was only influenced by the prevalence of single infections (**Figure 4**).

### Temporal pattern of *Nosema* spp. Prevalence

For evaluating spread and assertiveness of the emerging honey bee pathogen N. ceranae, we analyzed the temporal patterns of N. ceranae-, N. apis-, and co-infections by plotting and statistically analysing the respective values separately for the spring (**Figure 5**) and autumn (**Figure 6**) seasons between 2005 and 2016. While the patterns for N. apis- and coinfections in spring did not show a consistent trend, the pattern for N. ceranae-infection prevalence suggested a continuously increasing trend over the years (**Figure 5A**). Generalized linear models (GLM) of the prevalence data confirmed this interpretation (**Figures 5B–D**). Logistic regression analysis (**Table 2**) demonstrated that the continuous increase in spring prevalence of N. ceranae-infections observed over the entire 12 year study period, i.e., between 2005 and 2016, was on average about 5% per year (Odd Ratio: 1.05 [1.01, 1.1]) and was significant (GLM, Likelihood Ratio test of the model, p = 0.02) (**Figure 5B**). This increase, however, was not accompanied by any significant (GLM, Likelihood Ratio test of the model, p = 0.95) change in the spring prevalence of N. apis-infections (Odd Ratio: 1.0 [0.96, 1.04]) (**Figure 5C**). Likewise, no significant trends (GLM, Likelihood Ratio test of the model, p = 0.17) were observed for co-infections in spring (Odd Ratio: 0.96 [0.9, 1.02]) (**Figure 5D**).

The dotplot of autumn prevalence of N. ceranae-, N. apis-, and co-infections (**Figure 6A**) showed a different pattern with an increasing trend for N. ceranae- being accompanied by a decreasing trend for N. apis-infections. This finding could be substantiated by GLM-analysis and Likelihood Ratio tests of the models (**Table 2**, **Figures 6B–D**). In autumn, the prevalence of

FIGURE 3 | Spearman correlation matrix values showing the pairwise correlation coefficients for the parameters season and the four infection categories (Nosema spp.-, N. apis-, N. ceranae-, and co-infections). Shape and orientation of the ellipses represent the data clouds of the respective correlation coefficients. Positive correlations are illustrated in blue and negative correlations in red. Coefficients (white digits) between |0.10| and |0.29| represent weak associations, coefficients between |0.30| and |0.49| represent medium associations and coefficients of |0.50| or above (beyond) represent strong associations.

N. ceranae-infections was significantly (GLM, Likelihood Ratio test of the model, p = 0.0003) increasing by an average of about 15 % per year (Odd Ratio: 1.15 [1.07, 1.25]) over the study period (**Figure 6B**) while at the same time the prevalence of N. apis-infections was significantly decreasing by an average of about 11% per year (Odd Ratio: 0.89 [0.84, 0.95]) (GLM, Likelihood Ratio test of the model, p = 0.0003) (**Figure 6C**). For co-infections, however, no significant (GLM, Likelihood Ratio

FIGURE 5 | Temporal patterns of prevalence of single infections with N. ceranae or N. apis and of co-infections detected in spring samples between 2005 and 2016. (A) Prevalence data for N. ceranae-, N. apis-, and co-infections in spring are plotted against year. (B) Data sets for prevalence of N. ceranae-, N. apis-, and co-infections in spring were fitted by Linear Mixed-Effects Models to visualize the relationship between the independent variables (year) and the dependent variables N. ceranae- (B), N. apis- (C), and co-infections (D). Regression lines visualizing trends for N. ceranae- (B; solid red line), N. apis- (C; solid green line), and co-infection (D; solid blue line) prevalence are shown; ribbons represent the 2nd and 3rd quartile (25–75%) of the predicted data of the model.

and 2016.

FIGURE 6 | Temporal patterns of prevalence of single infections with N. ceranae or N. apis and of co-infections detected in autumn samples between 2005 and 2016. (A) Prevalence data for N. ceranae-, N. apis-, and co-infections in autumn are plotted against year. In autumn 2011, highlighted by an asterisk, the proportions of N. apis- and N. ceranae-infections were identical, hence, only a red dot is visible. (B) Data sets for prevalence of N. ceranae-, N. apis-, and co-infections in autumn were fitted by Linear Mixed-Effects Models to visualize the relationship between the independent variables (year) and the dependent variables N. ceranae- (B), N. apis- (C), and (Continued)

#### FIGURE 6 | Continued

co-infections (D). Regression lines visualizing trends for N. ceranae- (B, solid red line), N. apis- (C; solid green line), and co-infection (D; solid blue line) prevalence are shown; ribbons represent the 2nd and 3rd quartile (25–75%) of the predicted data of the model.

TABLE 2 | Results of the binary logistic regression analysis of prevalence of N. apis-, N. ceranae- and co-infections over the 12 year study period (see also Figures 5, 6).


\*, 95% confidence interval.

test of the model, p = 0.5) change in prevalence could be demonstrated between 2005 and 2016 (Odd Ratio: 0.95 [0.81, 1.11]) (**Figure 6D**).

### *In vitro*-Infection of IPL-LD-65Y Cells

To explain the obvious success of N. ceranae over N. apis in the studied honey bee population in summer, we experimentally analyzed the proliferative capacity of both microsporidian species in infected cells at temperatures between 21◦ and 33◦C. To this end, we used an established cell culture model for N. apis and N. ceranae based on experimentally infecting cultured IPL-LD-65Y-cells. This insect cell line derived from Lymantria dispar had been shown to support replication of both microsporidian species (Gisder et al., 2011).

Intracellular proliferation of N. apis and N. ceranae at three different temperatures (21◦ , 27◦ , and 33◦C) was evaluated by determining the number of the developmental stages per cell produced during merogony (meronts) and sporogony (sporonts/spores) of Nosema spp. (**Figure 7A**). The number of both meronts and sporonts/spores increased for N. apis as well as for N. ceranae over the observation time period of 96 h at all three tested temperatures. However, the number of the different developmental stages varied between N. apis and N. ceranae infected cells depending on incubation time and incubation temperature. At 21◦C, there was no significant difference in the proliferative capacity of N. ceranae and N. apis in infected cells for both meronts and sporonts/spores at all tested time points (24, 32, 48, 72, and 96 h post-infection) (all p > 0.05) (**Figure 7B**). However, at 27◦C and even more so at 33◦C, a higher proliferation rate and a faster proliferation of N. ceranae compared to N. apis could be observed. In infected cells which were incubated at 27◦C (**Figure 7C**), the number of meronts was not significantly different between N. apis and N. ceranae after 32 and 72 h post-infection (p > 0.05) but was significantly different at time points 24, 48, and 96 h post-infection (p > 0.05). More

FIGURE 7 | In vitro-infection of IPL-LD-65Y cells with N. apis- and N. ceranae-spores. Nosema spp. proliferation was determined by counting the number of developmental stages of merogony (A; meronts, dark gray arrow heads) and sporogony (A; sporonts and intracellular spores, light gray arrow heads) in infected cells (scale bars <sup>=</sup> <sup>25</sup> <sup>µ</sup>m) incubated at 21◦<sup>C</sup> (B), 27◦<sup>C</sup> (C), and 33◦<sup>C</sup> (D). Dark gray bars represent the number of meronts per cell (mean of 10 cells <sup>±</sup> SD), light gray bars represent the number of sporonts/spores per cell (mean of 10 cells ± SD). Statistical analysis of the number of developmental stages was performed with student's t-tests for each time point and temperature. Statistical results given above the bars refer to the comparison of sporonts/spores produced by N. apis and N. ceranae (not significantly different: n.s., p ≥ 0.05; significantly different: \*, 0.05 < p < 0.01; \*\*, 0.01 < p < 0.001; \*\*\*, 0.001 > p ≥ 0.0001; \*\*\*\* p < 0.0001).

importantly, the number of counted sporonts/spores at 32, 48, 72, and 96 h post-infection was significantly higher (p < 0.05) in N. ceranae- than in N. apis-infected cells (**Figure 7C**). When the host cells were incubated at 33◦C, the number of N. apis meronts was significantly (all p < 0.01) higher than the number of N. ceranae meronts at 32, 48, 72, and 96 h post-infection (**Figure 7D**) and the numbers of sporonts/spores were significantly higher at 32, 48, 72, and 96 h post-infection (all p < 0.01) in N. ceranae-infected host cells than in cells infected with N. apis.

### DISCUSSION

### Prevalence of *N. ceranae* Infections Follows the Same Seasonality as *N. apis* Infections

Nosema ceranae is an emergent pathogen of the Western honey bee A. mellifera. Its first detection in colonies of A. mellifera dates back a decade (Higes et al., 2006; Huang et al., 2007), although, it obviously switched host from A. cerana to A. mellifera about 40 years ago (Teixeira et al., 2013) and is now endemic in the global A. mellifera population. Several differences between N. ceranae and its congener N. apis have been reported, including differences in virulence, in seasonality of infections, and in temperature dependence of spore germination and biotic potential. Most of these differences seem to work in favor for N. ceranae resulting in its continous spread in the bee population and a supersession of N. apis in many regions (Klee et al., 2007; Chen et al., 2008; Williams et al., 2008; Invernizzi et al., 2009; Chen and Huang, 2010; Stevanovic et al., 2011; Yoshiyama and Kimura, 2011; Copley et al., 2012). However, studies claiming lack of seasonality of N. ceranae or replacement of N. apis in a given honey bee population are rather short-termed studies rarely performed over more than 2 years and most often involving only a limited set of samples. Our epidemiology data based on observing a cohort 230 bee colonies sampled twice a year over 12 years now revealed a different picture at least for the study area.

Nosema apis-infections are known to follow a seasonal pattern with spring prevalence being higher than autumn prevalence. This seasonality can be explained by the pathobiology of N. apis: (i) Only the spores of N. apis are infective; (ii) older bees are more likely to be infected and carry more spores; and (iii) spores are most efficiently transmitted through the fecal-oral route (Bailey, 1967; Bailey and Ball, 1991). Therefore, N. apis transmission within the colony is favored by conditions with low or no brood rearing and forcing adult bees to stay inside the hive for longer periods and to have close in-hive contacts (Bailey, 1967; Bailey and Ball, 1991). These conditions are regularly fulfilled during the winter months in climatic zones with winter temperatures falling below 10◦C not allowing bees to fly out (Winston, 1987). Instead, in these regions honey bee colonies hibernate by longlived adult winter bees forming a winter cluster around the queen bee and not leaving the hive for weeks or months until weather conditions allow cleansing and foraging flights and restarting brood rearing to replace the old winter bees (Winston, 1987). This explains why N. apis-infection levels increase over winter but normally decrease over summer when the rather shortlived summer bees are engaged in foraging, are able to defecate outside the hive, and when newly raised bees regularly replace older more heavily infected bees (Bailey, 1967; Bailey and Ball, 1991; Retschnig et al., 2017).

In contrast, N. ceranae-infections were described to lack this characteristic seasonality (Higes et al., 2006, 2010) suggesting fundamental differences in pathobiology and preferred routes of transmission which would be interesting to investigate. To analyse this suggested lack of seasonality, we collected bee samples in spring and autumn without gap over 12 years from a cohort of around 230 honey bee colonies and analyzed all samples for the presence of Nosema spp. spores and performed molecular species differentiation in all Nosema spp.-positive samples. Surprisingly, the data clearly disproved that N. ceranaeinfections differ from N. apis-infections in regard to seasonality. Quite the contrary was true: All four infection categories, Nosema spp.-, N. apis-, N. ceranae-, and co-infections, followed the same seasonal pattern with spring prevalence of infection regularly being higher than autumn prevalence suggesting that N. ceranae and N. apis circulating in Northeast Germany are similar in regard to pathobiology and preferred transmission routes. Since reports on the lack of seasonality predominantly stem from South Europe (Higes et al., 2006, 2010), further experimental studies are necessary to analyse whether the differences in seasonality between the Northern and Southern parts of Europe are due to climatic factors or intraspecies differences in N. ceranae.

### No Evidence for a General Advantage of *N. ceranae* Over *N. apis* and for an Overall Replacement of *N. apis* by *N. Ceranae*

In many regions of the world, prevalence data collected for N. apis and N. ceranae indicated that N. ceranae has become the dominant species in the worldwide honey bee populations and it was suggested that N. ceranae has replaced or is about to replace its congener globally (Chen et al., 2012; Martin-Hernandez et al., 2012). However, in Europe, a South to North gradient was observed with N. ceranae being dominant in Southern European countries already 10 years ago while at that time N. apis was still dominant in the Northern part of Europe (Klee et al., 2007) which might reflect an already discussed climatic aspect in N. ceranae spread and assertiveness (Fenoy et al., 2009; Martin-Hernandez et al., 2009; Gisder et al., 2010; Chen et al., 2012; Natsopoulou et al., 2015).

Congruent with this South to North gradient (Klee et al., 2007), at the beginning of our epidemiology study we observed very low levels of prevalence for N. ceranae-infections in Northeast Germany compared to N. apis-infections. This starting condition, the size of the cohort, and the design and duration of the study provided a unique opportunity to follow the spread of the emerging pathogen N. ceranae and analyse the impact of this spread on its congener N. apis, well established in the observed honey bee population. Our epidemiology data show that starting from a very low level, the prevalence of N. ceranaeinfections significantly increased continuously in the observed cohort of honey bee colonies during the last 12 years. This increase was true for both time points of sampling, in spring (showing the development over winter) and autumn (showing the development over summer) clearly indicating that N. ceranae became successfully established and expanded its presence in the honey bee population of Northeast Germany.

With regard to replacement of N. apis by N. ceranae, the obtained epidemiology data showed a complex picture. For assuming a replacement process at the population level, N. apis infection prevalence should have concomitantly decreased during the study period. However, a significant decrease in N. apis-infection prevalence was only observed for autumn indicating that during the bee season in summer N. ceranae successfully competed with N. apis at the population level over the course of the study. Surprisingly and in contrast to autumn, no significant change in N. apis infection prevalence was evident in spring despite a significant increase in N. ceranae prevalence. Therefore, no replacement of N. apis by N. ceranae in the honey bee population of Northeast Germany took place over winter during the last 12 years. Instead, the increase in N. ceranae prevalence in spring came on top of the unaltered N. apis infection prevalence suggesting that the two microsporidian parasites did not compete with each other over winter at the colony and population level. In addition, the long term stability of N. apis-infection frequency in spring indicate, that whatever mechanisms are acting on N. apis during summer and causing its decrease in the population, they are compensated for and reversed during winter preventing a supersession of N. apis through N. ceranae in the observed honey bee population.

Replacement of N. apis by N. ceranae at the population level during summer but not during winter points to different mechanisms acting on or influencing the two microsporidian parasites in summer and over winter. Although the exact mechanism responsible for presence (summer) and absence (winter) of replacement at the population level still remain elusive, experimental data providing explanations at the individual bee level for the increase in N. ceranae infection prevalence over summer exists. In a recent study by Martin-Hernandez et al. (2009), infection experiments with caged bees were performed at different temperatures and the "biotic index" was calculated for both microsporidia as the total N. apis or N. ceranae spore count per day after infection. This "biotic index" was higher for N. ceranae than for N. apis at 25◦C but no significant difference could be observed at 33◦C (Martin-Hernandez et al., 2009). Although these results did not provide convincing proof for an advantage of N. ceranae over N. apis during summer, they pointed into an interesting direction. Therefore, we extended the approach and performed infection experiments in cell culture (Gisder et al., 2011), which allowed a detailed analysis of the time course of proliferation and of the proliferative potential of N. ceranae and N. apis at different temperatures. Our in vitro results revealed a significant advantage of N. ceranae over N. apis at 27◦ and 33◦C, the normal range of daily maximum temperatures in summers in Northeast Germany. At both temperatures, N. ceranae completed its replicative cycle faster and replicated more efficiently than N. apis. These results were in accordance with a recent study, suggesting a generally higher proliferation rate for N. ceranae compared to N. apis in experimentally infected, caged bees incubated at 30◦C for 20 days (Huang and Solter, 2013). Earlier and higher production of spores, which are transmitting the disease within and between colonies, may translate into higher infection prevalence at population level. These data explain an increase of N. ceranae infection levels, however, they still do not explain the observed replacement of N. apis by N. ceranae over summer.

For replacement of N. apis by N. ceranae, a simple increase in N. ceranae infection prevalence is not sufficient but a successful interspecies competition, with N. ceranae at least more often than N. apis winning the game, is necessary. Again, only experimental data at the individual bee level are available. Co-infection experiments with caged bees and simultaneous feeding of N. apis and N. ceranae spores did not provide evidence for intrahost competition between the two species (Forsgren and Fries, 2010; Milbrath et al., 2015). In contrast, sequential feeding of spores of the two species resulted in within-host competition: The first parasite significantly inhibited the growth of the second, regardless of species (Natsopoulou et al., 2015). This would have prevented the spread of N. ceranae because N. apis had been present in the bee population before N. ceranae arrived and would always have been first. However, this so-called "priority effect" proved to be asymmetric and N. ceranae exhibited a stronger inhibitory effect on N. apis than N. apis on N. ceranae (Natsopoulou et al., 2015). Mathematical modeling proposed that this priority effect will result in a successful replacement process at population level even when taking into account that the cold sensitivity of N. ceranae but not of N. apis spores (Fenoy et al., 2009; Gisder et al., 2010) provides a disadvantage for N. ceranae during cold winters (Natsopoulou et al., 2015).

However, for spring samples our epidemiology data clearly showed that although N. ceranae-infection prevalence increased over time, this increase did not result in a replacement of N. apis. Remarkably, N. apis-infection prevalence in spring remained rather stable over the 12 years study period although the autumn infection prevalence and, hence, the infection prevalence at the beginning of winter, has been declining during this period. Therefore, the two Nosema species rather not competed during winter and the mechanisms promoting the increase of N. ceranae in the studied honey bee population over winter did not influence the prevalence of N. apis.

Furthermore, we observed higher than expected co-infection rates in spring suggesting that there is no interspecies within-host competition at colony or population level during overwintering. The co-infection levels rather suggested that an infection with any one of the two microsporidia pre-existing in a colony favored an additional infection of the colony with the other microsporidium. This is in contrast to the above mentioned report (Natsopoulou et al., 2015) showing interspecies withinhost competition with a priority effect favoring the spread N. ceranae over N. apis. However, this inter-species competition was shown at the individual bee level whereas our epidemiology data concern the colony and population levels. And indeed, at the colony and population level it is hardly conceivable how an N. ceranae infection of one bee or colony might inhibit a nestmate or a neighboring colony, respectively, to become infected by N. apis - and the other way round. Actually, the concept of interspecies within-host competition of an obligate intracellular parasite due to competition for the same limited cellular energy resources cannot easily be translated to the colony or population level where limitation or shortage of resources (in this case: new hosts) is not yet the problem. However, if the prevalence of N. ceranae-infections keeps increasing like it did over the last 12 years, within-colony and between-colony competition might become an issue once all colonies are infected with either one of the microsporidia. Therefore, a continuation of this study will further our understanding of the long term epidemiology and interspecies competition at population level of these two important honey bee pathogens.

### AUTHOR CONTRIBUTIONS

EG and SG conceived and designed the study and the experiments. SG, VS, and LH carried out the experiments and the microscopic and molecular diagnosis of Nosema spp. SG, VS, and DG performed the statistical analysis. EG supervised all work, SG supervised the laboratory experiments, and DG supervised the statistical analysis. SG and EG wrote the paper. All authors revised the manuscript and approved the final version.

### FUNDING

This study was supported by the Ministries responsible for Agriculture of the German Federal States of Brandenburg (MIL) and Sachsen-Anhalt (MLU), Germany, by a grant from the German Ministry of Agriculture through the Bundesanstalt für Landwirtschaft und Ernährung (BLE, grant no. 2816SE004), and

### REFERENCES


by a grant of the German Research Foundation (DFG, GRK2046). VS received a fellowship from the German Research Foundation (DFG, GRK2046).

### ACKNOWLEDGMENTS

We thank Marion Schröder, Andrea Jäkisch, and Caspar Schöning for sampling the honey bee colonies over the last 12 years, numerous students for help with processing the bee samples, and Alice Ballard for help with the GLM analysis.


homing ability of Nosema infected honey bees. PLoS ONE 9:e103989. doi: 10.1371/journal.pone.0103989


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

# NF-κB-Like Signaling Pathway REL2 in Immune Defenses of the Malaria Vector *Anopheles gambiae*

Suzana Zakovic and Elena A. Levashina\*

Vector Biology, Max-Planck Institute for Infection Biology, Berlin, Germany

The blood feeding requirements of insects are often exploited by pathogens for their transmission. This is also the case of the protozoan parasites of genus Plasmodium, the causative agents of malaria. Every year malaria claims the lives of a half million people, making its vector, the Anopheles mosquito, the deadliest animal in the world. However, mosquitoes mount powerful immune responses that efficiently limit parasite proliferation. Among the immune signaling pathways identified in the main malaria vector Anopheles gambiae, the NF-κB-like signaling cascades REL2 and REL1 are essential for eliciting proper immune reactions, but only REL2 has been implicated in the responses against the human malaria parasite Plasmodium falciparum. Instead, constitutive activation of REL1 causes massive killing of rodent malaria parasites. In this review, we summarize our present knowledge on the REL2 pathway in Anopheles mosquitoes and its role in mosquito immune responses to diverse pathogens, with a focus on Plasmodium. Mosquito-parasite interactions are crucial for malaria transmission and, therefore, represent a potential target for malaria control strategies.

#### *Edited by:*

Susanne Hartmann, Freie Universität Berlin, Germany

#### *Reviewed by:*

Tony Nolan, Imperial College London, United Kingdom Antonio M. Mendes, Instituto de Medicina Molecular (IMM), Portugal

#### *\*Correspondence:*

Elena A. Levashina levashina@mpiib-berlin.mpg.de

> *Received:* 06 March 2017 *Accepted:* 01 June 2017 *Published:* 21 June 2017

#### *Citation:*

Zakovic S and Levashina EA (2017) NF-κB-Like Signaling Pathway REL2 in Immune Defenses of the Malaria Vector Anopheles gambiae. Front. Cell. Infect. Microbiol. 7:258. doi: 10.3389/fcimb.2017.00258 Keywords: *Anopheles gambiae, Plasmodium*, NF-κB signaling, IMD, REL2 pathway, malaria, vector biology, innate immunity

### INTRODUCTION

Mosquitoes are vectors of human infectious diseases with immense importance for public health. Malaria, caused by the Plasmodium protozoa, is the deadliest disease transmitted by Anopheles mosquitoes. Plasmodium development in the mosquito takes about 3 weeks. A series of mosquito factors affect malaria transmission, among them female longevity, nutritional fitness and efficient immune responses. The ookinete is by far the most fragile parasite stage, attracting the majority of immune responses. It swiftly develops from the sexually created zygote, and its task is to escape the dangerous gut environment by traversing the unicellular epithelium and to hide beneath the basal lamina that lines the mosquito midgut. If successful, ookinetes transform within the second day into vegetative protective oocysts that in a fortnight give rise to thousands of sporozoites that migrate and invade the salivary glands to be ready for a new transmission. Vector-parasite molecular interactions have been studied mostly in the laboratory model of infections of A. gambiae, the major malaria vector in the sub-Saharan Africa, with the rodent malaria parasite P. berghei. Although rodent and human parasites have similar invasion strategies in the insect vector, their elimination is mediated by two distinct nuclear factor-κB (NF-κB) immune pathways, REL2/Imd and REL1/Toll. NF-κB, initially discovered for its DNA-binding activity to an immunoglobulin-κ light chain enhancer in B lymphocytes, emerged as the central regulator of immune responses in animal kingdom (Sen and Baltimore, 1986). In Anopheles, experimental activation of REL2 aborts development of P. falciparum ookinetes, while constitutive induction of REL1 kills rodent parasites (Frolet et al., 2006; Garver et al., 2009). However, in both cases the molecular mechanisms of parasite recognition and killing remain unknown. In this review, we discuss immune responses of A. gambiae, with a focus on the REL2 signaling pathway, which is believed to be the major regulator of mosquito immune responses against human malaria parasites.

### IMMUNE SIGNALING AND PATHOGEN RECOGNITION

The Toll and Imd pathways in Drosophila regulate expression of hundreds of infection-inducible genes. While the Toll pathway, initially described in Drosophila embryonic development, is essential for defenses against Gram-positive bacteria and fungi (Lemaitre et al., 1996; Rutschmann et al., 2002), the Imd pathway orchestrates responses against Gram-negative bacteria and viruses (Kaneko et al., 2004; Costa et al., 2009).

Immune activation is mediated by recognition of pathogen derived molecules, such as metabolites, nucleic acids, or cell wall components that are released during pathogen growth and division (Vance et al., 2009). Among microbial and host factors that induce Imd signaling, the best characterized is the DAPtype peptidoglycan (DAP-PGN), a cell wall component of Gramnegative bacteria. DAP-PGN recognition at the cell surface is mediated by transmembrane peptidoglycan recognition proteins (PGRPs) (Choe et al., 2002; Gottar et al., 2002; **Figure 1**). In contrast, activation of the Toll pathway is initiated by binding of circulating recognition complexes to the lysine-type PGNs or glucans (Gobert et al., 2003; Leulier et al., 2003). This binding triggers downstream serine protease cascades, leading to processing and binding of the endogenous factor Spaetzle to the transmembrane receptor Toll (Morisato and Anderson, 1994; Schneider et al., 1994). Activation of both pathways culminates in the phosphorylation and release of the NF-κB-like transcription factors Dif, Dorsal and Relish from the inhibitors, and their translocation into the nucleus. The release of the Imd transcription factor Relish involves Caspar, which in its inactive state prevents cleavage of the Relish inhibitory domain (Kim et al., 2006; **Figure 1**). Phosphorylation of Cactus, the negative regulator of the Toll pathway, leads to its degradation and release of the transactivators Dif and Dorsal (Wu and Anderson, 1998).

The Imd role in Drosophila immunity has been recently extended to antiviral responses. Surprisingly, instead of protection, constitutive activation of the Imd pathway by depletion of a secreted cytokine-like molecule Diedel, enhances viral pathogenesis (Lamiable et al., 2016). Although the underlying mechanism is not entirely clear, in the absence of Diedel, viral infection triggers the pathway through an alternative, non-canonical cytoplasmic route by bypassing the function of the PGRP receptors (Lamiable et al., 2016). Similarly, non-canonical activation of this pathway was also reported in ticks that lack transmembrane PGRP-LC and death domain proteins, Imd and FADD, but feature a conserved ubiquitination module, Relish and Caspar (**Figure 1**). The tick pathway is induced by lipid components of membranes specific to PGN-deficient bacteria (Shaw et al., 2017). Importantly, the same lipids also induce the Imd pathway in Drosophila cell lines, suggesting that non-canonical activation of the Imd pathway may be evolutionarily conserved (Shaw et al., 2017).

### REL2 PATHWAY IN *A. GAMBIAE*

The vast knowledge that accumulated on the immune signaling in Drosophila has served as a blueprint for studying Anopheles mosquitoes. Sequencing of the A. gambiae genome benefited identification of the conserved components of the pathway (Christophides et al., 2002; Holt et al., 2002). However, in spite of considerable interest and potential importance in antiparasitic responses, surprisingly little is known about the targets of the Imd/REL2 pathway in Anopheles.

Genomic searches in A. gambiae identified three potential receptors: PGRP-SD, PGRP-LB and PGRP-LC. PGRP-SD has not been characterized, whereas functional analysis of PGRP-LB did not reveal its role in mosquito survival upon bacterial infections (Meister et al., 2009). Instead, the structure and function of PGRP-LC were characterized in a great detail (Meister et al., 2009). PGRP-LC encodes three splice variants (LC1-3) that differ in the organization of their extracellular PGN-binding domains. Structural modeling uncovered the potential of all three isoforms to bind both types of PGNs, highlighting a striking difference between the mosquito PGRP-LC with broad sensing capacities and the Drosophila PGRP-LC, which binds exclusively DAP-PGNs. The broad specificity of PGN binding of the Anopheles PGRP-LC was further substantiated by functional analyses that demonstrated equally critical role of the receptor in mosquito survival to Gramnegative and Gram-positive bacteria (Meister et al., 2009). Although infections with both bacteria induced expression of genes encoding antimicrobial peptides (AMPs) Cecropin1 and Defensin1, their transcriptional induction was PGRP-LC independent (Meister et al., 2009), therefore, the mechanisms underlying the PGRP-LC-mediated resistance to bacteria remain to be elucidated.

PGRP-LC plays an important role in regulating proliferation of the mosquito microbiota after blood feeding (Meister et al., 2009) and may, in big part, explain the role of the pathway in modulating development of Plasmodium parasites. Indeed, similar to the phenotype of PGRP-LC silencing, clearing mosquito microbiota by antibiotics prior to infections increases Plasmodium loads, whereas feeding mosquitoes with bacteria boosts their resistance to Plasmodium in a PGRP-LC-dependent manner (Meister et al., 2009). Therefore, it is possible that the REL2 pathway is activated after blood feeding by massive bacterial proliferation, whereas Plasmodium parasites are simple bystanders in this process and do not directly induce mosquito immune responses. Further transcriptomics studies examined mosquito responses to infections with P. falciparum and P. berghei and identified species-specific patterns of gene expression

FIGURE 1 | Schematic overview of Imd pathway in Drosophila melanogaster. DAP-type PGNs trigger the activation of Imd signaling by direct interaction with the immune cell. Of Peptido-Glycan Recognition Proteins (PGRPs) that bind these pathogen-derived molecules, transmembrane PGRP-LC is the main receptor linked to activation of Imd pathway (Choe et al., 2002; Gottar et al., 2002). Its activity is enhanced in circulation by secreted PGRP-SD and intracellularly by the cytosolic PGRP-LE (Takehana et al., 2002; Iatsenko et al., 2016). Both proteins can directly bind DAP-PGNs and promote PGRP-LC activity. Another extracellular PGN-binding protein PGRP-LB antagonizes PGRP-LC activity by scavenging PGNs in circulation (Zaidman-Rémy et al., 2006). PGN binding induces conformational changes in PGRP-LC that promotes recruitment of the death domain-containing proteins Imd, FADD and DREDD caspase from the nucleus to the plasma membrane, and it is followed by subsequent polyubiquitination of DREDD by ubiquitin E3 ligase IAP2, cleavage of Imd by DREDD and exposure of K63 site for polyubiquitination by IAP2 and E2 conjugating enzymes Bendless, Effete and Uev1a (Paquette et al., 2010; Meinander et al., 2012). The K63-polyubiquitin chains, most likely, serve as activators of TAK1 kinase via the ubiquitin-binding domain of its regulatory protein TAB2 (Paquette et al., 2010). TAK1/TAB2 complex phosphorylates IKK complex, which consists of β and γ subunits. IKKβ further phosphorylates the NF-κB-like transcription factor Relish, while a regulatory IKKγ subunit regulates DREDD-mediated cleavage of Relish (Ertürk-Hasdemir et al., 2009). Relish consists of the Rel Homology Domain (RHD) and the inhibitory ankyrin-repeat rich domain (ANK) (Dushay et al., 1996). DREDD caspase cleaves the ANK domain from RHD. RHD translocates to the nucleus and initiates transcription of target genes. Caspar acts as a negative regulator of the pathway by inhibiting DREDD-dependent cleavage of Relish (Kim et al., 2006). Immunomodulatory cytokine Diedel restrains deleterious non-canonical activation of Imd in presence and absence of viral infection (Lamiable et al., 2016). Receptor that activates the pathway to viruses is not yet known However, epistasis analyses placed Diedel function between Imd and IKKγ, as mutants for both Diedel and Imd were more prone to spontaneous pathogenesis than Diedel/IKKγ double mutants (Lamiable et al., 2016). Additionally, activation of the pathway is held in check by other factors, including CYLD, Dnr1, and Pirk. Finally, transcriptional activity of Relish is regulated at the chromatin level through interactions with a nuclear co-factor Akirin and BAP60 component of Brahma chromatin remodeling complex. Akirin recruits BAP60 complex to promoters of a subset of Relish effector genes and hence regulates their transcription (Goto et al., 2008; Bonnay et al., 2014). Positive and inhibitory interactions are depicted with → and |–, respectfully; black—well established, and red—yet unknown interactions. Color coding highlights our current knowledge on the pathway in Anopheles gambiae. Confirmed pathway components are indicated in color. Components depicted in gray represent orthologs identified by genomic searches, but whose function was not experimentally validated. Components in gray with dashed lines are absent in A. gambiae.

(Dimopoulos et al., 2002; Dong et al., 2006). Nevertheless, a significant overlap observed in the responses to bacterial and Plasmodium infections provides further support to the hypothesis that REL2 modulation of Plasmodium development may be triggered by the PGRP-LC-mediated recognition of bacteria.

Regardless of the trigger, it is expected from the Drosophila model that conformational changes of PGRP-LC will recruit the death-domain containing receptor-adaptor complex (Imd, FADD and DREDD), which will, via the TAK1/TAB2 complex, activate the IKK signalosome and inhibit the negative regulator Caspar (**Figure 1**). Relish activation is achieved by phosphorylation by the IKK signalosome and by cleavage of the inhibitory ankirin domain by the DREDD caspase (**Figure 1**). Transcriptional activity of Relish at the promoters of some genes is further regulated by its nuclear co-factor Akirin. Using RNAi silencing and its effects on P. falciparum infections, most of the pathway components were functionally confirmed in A. gambiae, except for TAK1, whose depletion did not impact Plasmodium development (Meister et al., 2005; Garver et al., 2012; Ramphul et al., 2015).

Further studies of the REL2 pathway identified some mosquito-specific particularities. In contrast to Drosophila, REL2 in mosquitoes encodes three alternatively-spliced isoforms that were first identified in another mosquito species, Aedes aegypti (Shin et al., 2002; Antonova et al., 2009). In A. gambiae, Meister et al. (2005) described two REL2 forms: a long transcript (REL2- F) coding for the full-length protein consisting of Rel-homology domain (RHD) and ankyrin-rich repeat (ANK), and a short form (REL2-S) encoding only RHD. Although authors proposed that the isoforms regulate expression of distinct sets of genes, both isoforms regulate Plasmodium development (Garver et al., 2012). Currently, the molecular mechanisms underlying activation of REL2 in Anopheles remain unresolved. It is unknown whether REL2 requires proteolytic activation by CaspL1 (Anopheles ortholog of DREDD) and how Caspar inhibits its activation. Surprisingly, functional analyses by gene silencing suggested a genetic interaction between Caspar and REL2-S, whereas regulation of REL2-F was not investigated (Garver et al., 2012). These results are inconsistent with the proposed mechanism of Relish inhibition by Caspar in Drosophila, where Caspar binding to the inhibitory ankirin domain prevents its cleavage by DREDD. However, as REL2-S lacks the ANK domain, the inhibitory mechanism of Caspar in Anopheles remains unclear. Co-silencing of Caspar with Imd, FADD, CaspL1 or IKK2, rescued the negative effect of the single Caspar knockdown on parasite development, confirming the role of these components in activation of REL2. Of particular interest is the role of Akirin, the nuclear co-factor of REL2, which regulates chromatin conformation and provides access to the promoter regions of a set of effector genes in Drosophila (Goto et al., 2008; Bonnay et al., 2014). As Akirin contributes to antiparasitic responses (Da Costa et al., 2014), better understanding of its function and of its target genes should shed light on regulation of Plasmodium killing.

So far, abortion of Plasmodium development was observed only upon experimentally-induced activation of the NF-κB pathways. Therefore, species-specific elimination of Plasmodium parasites by REL2 and REL1 seems to be independent of parasite recognition and may be explained by the activation of distinct sets of effectors. It is also plausible, that Plasmodium species differ in their susceptibility to the immune defenses triggered by these pathways. Therefore, understanding the molecular mechanisms underlying the pathway-specific parasite elimination may provide interesting insights into biology of Plasmodium species.

### EFFECTORS

Antimicrobial peptides are powerful effectors of innate immunity (Jenssen et al., 2006). They bind and directly kill a broad spectrum of pathogens by disrupting cell membrane integrity (Yeaman and Yount, 2003). Insect AMPs are synthesized by the fat body with some contribution of hemocytes, and are secreted into the hemolymph shortly after infection. In addition, some AMPs are also produced by epithelial cells in a tissue-specific manner (Tzou et al., 2000). Expression of the AMP genes Drosomycin and Diptericins in the fat body is regulated by Toll and Imd, respectively, whereas both pathways contribute to expression of other AMP genes (e.g. Defensin, Drosocin, Metchnikowin, Attacins, and Cecropins) (Ferrandon et al., 2007).

Several AMP genes have been identified in A. gambiae: Defensins (Def1-5), Cecropins (Cec1-4), Gambicin (Gamb), and Attacin (Holt et al., 2002; Mongin et al., 2004). Antimicrobial and antifungal activities of the recombinant Cec1, Gamb, and Def1 peptides were demonstrated in vitro against filamentous fungi (Cec1, Gamb, Def1), Gram-negative (Cec1, Gamb) and Gram-positive bacteria (Def,1 Cec1, Gamb) (Vizioli et al., 2000, 2001a,b). In vivo silencing of Def1 increases mosquito susceptibility to Gram-positive bacteria but does not affect development of P. berghei (Blandin et al., 2002). At the transcriptional level, however, expression of Def1 was upregulated by infections with human and rodent parasites (Tahar et al., 2002). Gambicin, the only mosquito-specific AMP, exhibited some activity against P. berghei ookinetes in vitro, whereas depletion of Gamb in vivo increased mosquito susceptibility to Gram-positive bacteria, P. berghei and, to a lower extent, to P. falciparum (Vizioli et al., 2001a; Dong et al., 2006). Interestingly, transgenic over-expression of Cec1 fused to the Shiva toxin inhibited development of P. berghei oocysts in A. gambiae (Kim et al., 2004). In spite of these results, the exact role of antimicrobial peptides in the mosquito defenses against Plasmodium is still incompletely understood.

Identification of the effectors of REL2 and REL1 is crucial for understanding the specificity of malaria killing in the mosquito. However, only a handful of immune genes in Anopheles have been assigned to either pathway. Frolet et al. (2006) did not observe any changes in the expression of AMP genes upon constitutive activation of REL1, leaving an open possibility of their regulation by REL2. In vitro studies provided some support of Cec1 and Gamb regulation by REL2 (Meister et al., 2005), but a more recent study in vivo suggested a dual regulation of AMP genes by both pathways (Garver et al., 2009).

The complement-like system emerged as a powerful arm of the mosquito immune responses to a broad spectrum of pathogens. The central component of this system, the thioestercontaining protein 1 (TEP1), is a major determinant of malaria killing (Blandin et al., 2004; Garver et al., 2009; Molina-Cruz et al., 2012; Nsango et al., 2012). TEP1 binds to the surface of invading Plasmodium ookinetes and bacteria, and promotes their killing by lysis and phagocytosis, respectively (Blandin et al., 2004). TEP1 is a highly reactive protein and requires a complex of two leucine-rich repeat proteins [leucine-rich repeat immune protein 1 (LRIM1) and Anopheles Plasmodium-responsive leucine-rich repeat 1C (APL1C)] to prevent its precocious activation and precipitation (Frolet et al., 2006; Fraiture et al., 2009). TEP1 or LRIM1 co-silencing with Cactus completely reverts the refractory phenotype of Cactus knockdown in A. gambiae infections with P. berghei (Frolet et al., 2006). Silencing of TEP1 also results in higher intensities of P. falciparum infections (Garver et al., 2009), however, levels of TEP1 protection against P. falciparum vary with the genotype and genetic complexity of Plasmodium infections (Molina-Cruz et al., 2012; Nsango et al., 2012). Similar to AMPs, expression of the complement-like genes is regulated by both pathways (Frolet et al., 2006).

Another interesting multi-member protein family with potential roles in mosquito immune responses is the family of fibrinogen related proteins (FBNs or FREPs). It comprises 59 members in A. gambiae, 37 in A. aegypti and 14 in D. melanogaster (Christophides et al., 2002; Dong and Dimopoulos, 2009). Only few FBNs have been functionally characterized. Silencing of FBN9, FBN22 and FBN39 impairs mosquito survival upon bacterial infections; depletion of FBN8, FBN9, FBN30, and FBN39 increases mosquito susceptibility to Plasmodium parasites, while silencing of FBN1 decreases parasite loads (Dong and Dimopoulos, 2009; Li et al., 2013; Simões et al., 2017). Little is known about regulation of FBN expression, except for FBN9, whose expression is regulated by REL2 (Garver et al., 2009). FBN9 binds to malaria parasites and bacteria, thereby exposing them for killing by an as yet unknown mechanism (Dong and Dimopoulos, 2009). Surprisingly, transgenic expression of FBN9 in the fat body driven by a blood feeding-inducible promoter did not enhance mosquito resistance to P. falciparum (Simões et al., 2017). This unexpected result highlights the importance of tissuespecific REL2 regulation, which has not been addressed yet.

## CELLULAR IMMUNE RESPONSES

Several independent reports implicated REL2 pathway in the immune responses of hemocytes, the mosquito blood cells. Transcripts encoding the pathway components were identified in the hemocyte-enriched transcriptome and their expression levels were further upregulated by blood meal and by P. berghei ookinetes (Baton et al., 2009). Furthermore, hemocytes synthesize proteins whose expression is regulated by the REL2 pathway (e.g., AMPs, TEP1, LRIM1, and FBN9) (Levashina et al., 2001; Baton et al., 2009). Some of these proteins (TEP1, TEP3, PGRP-LC, and LRIM1) also contribute to the efficient phagocytosis of Gram-positive and Gram-negative bacteria (Moita et al., 2005). In Drosophila, hemocytes also serve as messengers in inter-organ communication. For example, an amplitude of systemic immune responses induced by localized infections is diminished in hemocyte-depleted fruit fly mutants, revealing hemocyte contribution to the amplification of the fat-body mediated immune responses (Charroux and Royet, 2009; Wu et al., 2012). The reactive oxygen species that act as local triggers of hemocyte activation, also efficiently activate the Imd pathway (Foley and O'Farrell, 2003; Wu et al., 2012), further supporting the potential role of this pathway in hemocyte activation.

## MELANIZATION

The Anopheles REL2 pathway negatively regulates melanization, a process of melanin deposition in defense mechanisms (such as wound healing or pathogen killing), metamorphosis and tanning during development. Silencing of PGRP-LC and of both REL2 isoforms not only renders mosquitoes more susceptible to Plasmodium but also induces melanization of P. berghei ookinetes (Meister et al., 2005, 2009; Frolet et al., 2006). The reverse is observed in Drosophila, where the intracellular receptor PGRP-LE is absolutely required for melanization (Takehana et al., 2004). REL1 pathway, on the other hand, promotes melanization in both insect species (Ligoxygakis et al., 2002; Frolet et al., 2006). Surprisingly, simultaneous activation of the REL1 pathway by Cactus knockdown and inhibition of the REL2 pathway by REL2 silencing abolishes Plasmodium melanization, revealing the complexity in the regulation of this immune reaction and a potential cross-talk between the two pathways (Frolet et al., 2006).

## CONCLUSIONS

Although a significant progress has been achieved in identifying the components of the REL2 pathway in mosquitoes, many questions remain unanswered. Little is known about the role of post-translational modifications, such as ubiquitination and phosphorylation, which act as important pathway regulators in Drosophila. Moreover, completely unexplored areas are the contributions of the epigenetic modifications acting at the promoter level to fine tune immune responses. Understanding how chromatin conformations modulate expression patterns of effector genes may offer new insights into the complexity and specificity of REL2-mediated immune responses to a broad range of pathogens. However, even before considering the complexity of epigenetic modifications, it is pivotal to characterize the REL2-specific effectors, which are currently only vaguely known; and to address the questions of Plasmodium recognition and pathway activation upon infection in order to properly understand the parasite-host interaction and Plasmodium killing in the mosquito.

Recent studies in ticks and Drosophila discovered a noncanonical cytoplasmic route of pathway activation that bypasses the PGRP receptor-adaptor complex. These observations open new research avenues regarding receptor(s) and molecular mechanisms of pathway activation. The conservation of this noncanonical route in the evolutionarily distant organisms, such as ticks and Drosophila, may suggest that PGN recognition by PGRPs and further signal transduction by the deathdomain module, appeared later in evolution, after separation of arachnids from insects. Instead, the core pathway from the ubiquitination module via Caspar to Relish seems to be broadly conserved across arthropods. Better understanding of the REL2 immune pathway in the malaria mosquitoes should advance our knowledge of conserved mechanisms of innate immunity and may identify new targets for vector-mediated malaria control.

### AUTHOR CONTRIBUTIONS

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

### REFERENCES


### ACKNOWLEDGMENTS

The authors wish to thank Dr. Maiara Severo for critical reading of the manuscript. SZ is supported by a fellowship from German Research Foundation (DFG), GRK 2046.


**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 Zakovic and Levashina. 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.

# Parasite Carbohydrate Vaccines

Jonnel A. Jaurigue1, 2 and Peter H. Seeberger 1, 2 \*

<sup>1</sup> Department of Biomolecular Systems, Max Planck Institute of Colloids and Interfaces, Potsdam, Germany, <sup>2</sup> Institute for Chemistry and Biochemistry, Freie Universität Berlin, Berlin, Germany

Vaccination is an efficient means of combating infectious disease burden globally. However, routine vaccines for the world's major human parasitic diseases do not yet exist. Vaccines based on carbohydrate antigens are a viable option for parasite vaccine development, given the proven success of carbohydrate vaccines to combat bacterial infections. We will review the key components of carbohydrate vaccines that have remained largely consistent since their inception, and the success of bacterial carbohydrate vaccines. We will then explore the latest developments for both traditional and non-traditional carbohydrate vaccine approaches for three of the world's major protozoan parasitic diseases—malaria, toxoplasmosis, and leishmaniasis. The traditional prophylactic carbohydrate vaccine strategy is being explored for malaria. However, given that parasite disease biology is complex and often arises from host immune responses to parasite antigens, carbohydrate vaccines against deleterious immune responses in host-parasite interactions are also being explored. In particular, the highly abundant glycosylphosphatidylinositol molecules specific for Plasmodium, Toxoplasma, and Leishmania spp. are considered exploitable antigens for this non-traditional vaccine approach. Discussion will revolve around the application of these protozoan carbohydrate antigens for vaccines currently in preclinical development.

#### Edited by:

Brice Rotureau, Institut Pasteur, France

#### Reviewed by:

Susu M. Zughaier, Emory University, United States Charles Kelly, King's College London, United Kingdom Rashika El Ridi, Cairo University, Egypt

#### \*Correspondence:

Peter H. Seeberger peter.seeberger@mpikg.mpg.de

Received: 21 February 2017 Accepted: 26 May 2017 Published: 12 June 2017

#### Citation:

Jaurigue JA and Seeberger PH (2017) Parasite Carbohydrate Vaccines. Front. Cell. Infect. Microbiol. 7:248. doi: 10.3389/fcimb.2017.00248 Keywords: carbohydrate, vaccine, malaria, toxoplasmosis, leishmaniasis

### INTRODUCTION

The first vaccine was the smallpox inoculation introduced in 1796 by Jenner that used whole, attenuated organisms to generate a protective immune response. Our ever-increasing understanding of the underlying immune response that drives vaccination has led to modern subunit vaccines that are formulated to exacting standards. These subunit vaccines use defined protein or carbohydrate antigens implicated in virulence and disease to generate a more directed, nuanced immune response against the offending pathogen.

Over two centuries of vaccinology has seen significant progress in protection from many viral and bacterial diseases affecting humans such as polio, Haemophilus influenza type B, Streptococcus pneumoniae, and Neisseria meningitis, saving millions of lives each year. And yet, it is disheartening to consider that no vaccine exists for human parasitic infections that continue to cause suffering in many parts of the world (Nyame et al., 2004; Astronomo and Burton, 2010; Hoffman et al., 2015).

More than a million people die each year from diseases like malaria and leishmaniasis, with lifelong disability, disfigurement, and suffering for those that are living with disease (Hotez et al., 2014). As it stands, diseases caused by protozoan parasites are a leading cause of death the world over yet vaccine strategies for global parasitic diseases such as malaria (John et al., 2008; Seder et al., 2013; Tinto et al., 2015; Gosling and von Seidlein, 2016; WHO, 2016b) and toxoplasmosis (Jongert et al., 2009) have been pursued for decades. The failure in developing an effective vaccine against human parasite infection lies in part to the complexity of parasite biology compared to other microbes (Astronomo and Burton, 2010; Hoffman et al., 2015). Another failure is our relatively poor understanding of the protein and carbohydrate antigens relevant to parasitic virulence. Given that parasite disease pathology often arises from complex host immune responses to parasite antigens, ongoing research to better understand host-parasite interactions in disease is vital to parasite vaccine development (Schofield and Grau, 2005).

Carbohydrates are considered compelling, exploitable targets for vaccination to overcome the challenges that have prevented the realization of a human parasite vaccine (Nyame et al., 2004; Rodrigues et al., 2015). They are abundantly present on the surface of parasites and play a key role in host-parasite interactions (Rodrigues et al., 2015). Unique carbohydrate antigens characterize multiple developmental stages and tend to be immunoreactive for both protozoan and helminth parasites (Nyame et al., 2004). Adding to the appeal of carbohydrate antigens is the ongoing, systematic characterization of parasite glycobiology regarding their structure, function and biosynthesis (Nyame et al., 2004; Rodrigues et al., 2015). In this review, we will discuss the advantages and disadvantages of modern, carbohydrate antigen-based subunit vaccines, and reflect on the latest developments of carbohydrate vaccines for major protozoan parasites.

### CARBOHYDRATE VACCINE ANATOMY

### Carbohydrates Antigens

Carbohydrates are abundant on the surfaces of all cells and exist as poly- and oligosaccharides attached to proteins and lipids (Horlacher and Seeberger, 2008). They are involved in key biological processes such as cell adhesion, modulatory processes, and structural functions (Varki and Lowe, 2009). For pathogenic microbes, carbohydrate interactions are utilized for attachment (Kline et al., 2009) and invasion (de Groot et al., 2013). In turn, pathogenic microbial carbohydrates can be recognized by host immune systems to induce the production of carbohydrate-specific antibodies that can serve protective functions (Astronomo and Burton, 2010). Carbohydrates have been exploited for protective vaccination for decades given their crucial roles in development, growth, and disease (Varki and Lowe, 2009).

Carbohydrate biosynthesis is not directly template-driven as is the case for nucleic acid and proteins (Rodrigues et al., 2015). Instead, their biosynthesis is a complex, multi-enzymatic process (Delorenzi et al., 2002) forming linear and branched molecules with varied linkages. The result is a class of biopolymers of great complexity and diversity. Their heterogeneity means that access to pure, defined carbohydrates remains a challenge (Liu et al., 2006; Adibekian et al., 2011; Geissner and Seeberger, 2016). Microbial cell culture can be a readily available biological source of carbohydrate antigens, however carbohydrate isolation can be complex and even tedious (Geissner and Seeberger, 2016). Furthermore, isolation is not so straightforward for carbohydrates of low abundance or when microbial culture is not possible. This is especially true for parasites, which generally require more complex culturing conditions compared to bacteria. Fortunately, carbohydrate synthesis technology continues to advance and is increasingly becoming an alternative to biological isolation to source carbohydrates (Anish et al., 2014). Moreover, carbohydrate synthesis allows for pure, completely defined carbohydrate antigens as a basis for synthetic carbohydrate vaccines.

### Carrier Proteins

Virtually all vaccines rely on antibody production and subsequent immunological memory against the target antigen for their protective effect (Zinkernagel, 2003). Accordingly, the immune response against the antigen is a key consideration when it comes to vaccine design where strong, long lasting immune responses are desired. In this regard, proteins and carbohydrates are generally regarded as thymus dependent or thymus independent antigens, respectively, which characterizes the type of antibody immune response they elicit.

### Thymus Dependent and Thymus Independent Antigens

Thymus dependent (TD) antigens, such as proteins, are taken up by antigen presenting cells (APCs). The endocytosed antigen is processed through a series of catalytic steps which liberate small peptide fragments. These peptide fragments form complexes with MHCII molecules and thereby are able to be displayed on the MHCII molecules of the APC. T cells specific to the displayed MHCII-peptide complex are co-stimulated by the APC (Avci and Kasper, 2010). The activated T cells go on to "help" the antigen-specific B cells, promoting their proliferation, affinity maturation, antibody isotype switching and long lasting immunological memory (Pulendran and Ahmed, 2011).

Carbohydrate antigens are classed as thymus independent (TI) antigens and are poorly immunogenic compared to TD antigens (Weintraub, 2003). Carbohydrate antigens are recognized by APCs through pathogen recognition receptors (Blander and Sander, 2012), endocytosed, and processed into oligosaccharide epitopes. These oligosaccharides are not presented on MHCII molecules, but instead are presented directly on the APC cell surface to activate carbohydrate antigenspecific B cells. Due to the lack of T cell activation, B cells are activated in the absence of affinity maturation and isotype switching (Mond et al., 1995) which leads to the predominant production of low affinity IgM antibodies, and only low levels of IgG (Mond and Kokai-Kun, 2008; Astronomo and Burton, 2010; Berti and Adamo, 2013). The IgM antibodies bind in the micromolar range, compared to TD antigen-derived IgG antibodies that bind in the nanomolar range (Broecker et al., 2016; Geissner et al., 2016). B and T cell memory is often not achieved and the immune response is short lived (Adams et al., 2008; Mond and Kokai-Kun, 2008; Hütter and Lepenies, 2015). Moreover, children less than 2 years of age fail to mount an antibody immune response to TI antigens (Weintraub, 2003; Landers et al., 2005).

### Glycoconjugates

In the 1920s and 1930s, covalent conjugation of carbohydrate antigens to protein scaffolds was first explored to investigate the interaction between the TD antigen properties of proteins with TI carbohydrate antigens (Avery, 1931). This early work formed the basis of the first glycoconjugate carbohydrate vaccines produced in the 1980s (Schneerson et al., 1980; Beuvery et al., 1982; Wessels et al., 1993), where bacterial capsular polysaccharides (CPS) were covalently linked to so-called carrier proteins. Immunization with these CPS-protein glycoconjugates enabled T cell-mediated B cell activation against the target carbohydrate antigen, and induced long term immune memory even in infants (Stein, 1992).

The mechanism whereby TI carbohydrate antigens can activate the immune system like TD protein antigens via glycoconjugation remains an open area of research. A longstanding mechanistic explanation argues that the carbohydrateprotein conjugate is recognized by carbohydrate-specific B cells. The protein component of the conjugate is subsequently endocytosed, processed and presented on MHCII molecules of the carbohydrate specific B cell. Through the MHCII-peptide complex a peptide-specific T cell can activate the carbohydratespecific B cell, resulting in TD immune responses against the desired carbohydrate antigen (**Figure 1**; Lucas et al., 2005). A more recently proposed second mechanism argues that after uptake by carbohydrate-specific B cells the glycoconjugate is processed into glycopeptide fragments. Thus, the peptide portion of the fragment can form an MHCII-peptide complex, enabling the simultaneous presentation of the hydrophilic carbohydrate portion to carbohydrate-specific T cells. The T cell interaction activates the presenting B cell accordingly (Avci et al., 2011).

Five TD carrier proteins are currently used in licensed carbohydrate vaccines (Pichichero, 2013). Of these licensed carrier proteins, a non-toxic mutant of diphtheria toxin (CRM197) is often used in vaccine development research for parasitic diseases. Another carrier protein often used in parasite vaccine development is keyhole limpet haemocyanin (KLH). Unlike many other carrier proteins KLH is itself highly glycosylated and is a good promotor of TD and TI responses (Nyame et al., 2004), however it is not licensed for human use.

The linker conjugating the carrier protein to the carbohydrate antigen is another key consideration of carbohydrate vaccines. Synthetic carbohydrates with defined carrier protein binding sites are advantageous for generating well-defined glycoconjugates with predictable conjugation sites. To minimize the immunogenicity of the linker (Buskas et al., 2004; Gotze et al., 2015) short linkers with no functional groups are best used to ensure that immune responses against the desired antigen epitope are not detrimentally affected. The processing of the linker after uptake by APCs is also an important consideration, to allow for the release of vaccine oligosaccharide and peptide moieties for antigenic display.

### Adjuvants

Adjuvants are substances that modulate or bolster an effective immune response against the antigens in the vaccine. Formulations are typically emulsions and vesicles that can serve as a delivery vehicle for antigen vaccine components and allow for the slow release of vaccine antigen components over time. Adjuvants can assist antigen immunogenicity by increasing local inflammation and antigen uptake by APCs, and aid in their migration to lymph nodes (Petrovsky and Aguilar, 2004; Di Pasquale et al., 2015). Ideally, adjuvants reduce the amount of antigen or number of immunizations needed for vaccination (Petrovsky and Aguilar, 2004).

Frontiers in Cellular and Infection Microbiology | www.frontiersin.org

Aluminum-based mineral (Alum) salts served as the only adjuvants for human vaccines for decades (Zimmermann and Lepenies, 2015) and are used for a wide range of vaccines that aim to predominantly induce antibody-mediated immune responses (Bhowmick et al., 2014). Despite its worldwide use, research on its mechanism of action is ongoing (De Gregorio et al., 2008). Today, alum, virosomes, lipid A derived adjuvants, and squalene adjuvants are all in use for human licensed vaccines (Astronomo and Burton, 2010; Di Pasquale et al., 2015; Zimmermann and Lepenies, 2015) and are commonly used in vaccine research. Crude Freund's adjuvant (CFA) is a powerful immunogen often used as an adjuvant for candidate parasite vaccines, but it is not approved for licensed human vaccines.

Modern vaccines that use better defined, or even synthetic antigens, are generally less immunogenic than crude, whole organism vaccines. Research is ongoing in advancing the efficacy of adjuvant systems for vaccines (Petrovsky and Aguilar, 2004). Many of these new adjuvant systems that are currently being tested in bacterial vaccines are of great interest for adaption to candidate vaccines of parasite diseases. For example, an adjuvant used in animal models, α-GalCer, was covalently attached to S. pneumoniae CPS to act as both carrier and adjuvant for the carbohydrate antigen (Cavallari et al., 2014). Endosomal processing by carbohydrate-specific B cells displays α-GalCer antigens via CD1d, stimulating iNKT cells (a class of T cells) to promote B cell hypermutation, class-switching and immunological memory (Cavallari and De Libero, 2017). Zwitterionic polysaccharides (ZPSs) are another emerging class of potentially self-adjuvanting carrier isolated from commensal anaerobic bacteria. ZPSs have the special property of containing one positive and one negative charge on adjacent monosaccharides. When they are processed by APCs this crucially enables their presentation on MHCII complexes, leading to T cell activation. Thus, they are able to activate adaptive immune responses for conjugated carbohydrate antigens in the absence of carrier protein (Berti and Adamo, 2013), leading to the possibility of fully-carbohydrate vaccines (Nishat and Andreana, 2016).

### The Success of Carbohydrate Vaccines

The concept of carbohydrate antigen vaccines started in the 1920s with the first published evidence that "residue antigens" of Streptococcus pneumoniae were the CPS of the bacterium (Heidelberger and Avery, 1923, 1924), later demonstrated to be important for virulence and serotype specificity (Hütter and Lepenies, 2015). The CPS of S. pneumoniae was shown to produce CPS-specific antibodies (Tillett and Francis, 1929) that protected against the disease symptoms of S. pneumoniae (Hütter and Lepenies, 2015), leading to the first CPS antigen vaccine against S. pneumoniae in 1947 (Grabenstein and Klugman, 2012). The advent of antibiotics at around the same time somewhat stalled vaccine research, as antibiotics became the preferred method for bacterial disease prevention (Grabenstein and Klugman, 2012; Hütter and Lepenies, 2015). The rise of antibiotic resistance in the following decades (Davies and Davies, 2010) led to a resurgence in carbohydrate vaccine research (Vliegenthart, 2006).

In the 1970s and 80s, CPS-based vaccines for S. pneumoniae were licensed and approved in the USA and Europe. Increasing numbers of strain-specific CPS were added to further itinerations of the vaccine to increase its efficacy, culminating into a 23 valent CPS antigen vaccine first licensed in 1983, protecting vaccinated adults against 87% of S. pneumoniae disease in the USA (Grabenstein and Klugman, 2012; Cavallari and De Libero, 2017). Glyconjugate carbohydrate vaccines were later introduced from the 1990s onward to allow for vaccination of broader demographics, especially infants. Glycoconjugate carbohydrate vaccines against disease caused by H. influenzae type b infection lead to its virtual elimination within countries with widespread coverage (Lindberg, 1999). Similarly, the use of glycoconjugate vaccines against Neisseria meningitides has also met with effective results (Girard et al., 2006). Vaccines for S. pneumoniae are also now glycoconjugates. Several other glycoconjugate vaccines are currently under development, such as a carbohydrate vaccine for group B streptococcus (De Gregorio and Rappuoli, 2014; Lepenies, 2015). Today, glycoconjugate vaccines have substituted pure carbohydrate vaccines where possible, with the most common carrier proteins being CRM<sup>197</sup> and tetanus toxin (Cavallari and De Libero, 2017).

### EXPLOITING CARBOHYDRATE ANTIGENS FOR PROTOZOAN PARASITE VACCINES

The key components of CPS carbohydrate vaccines—antigen, carrier protein, linker, and adjuvant—have remained largely consistent since their inception. Current parasite vaccine research follows the same component model in this regard. However, as illustrated in the following sections, a recurring theme for carbohydrate parasite vaccines is the need to steer away from traditional prophylactic vaccine development strategies that were successfully applied to bacterial infections.

In addition to carbohydrate vaccines that can induce sterile protection against the parasite itself, vaccinating against deleterious immune responses in host-parasite interactions is another strategy. Much of the pathology of parasitic disease is due to the host's own immune responses against the parasite. For example, the deadly manifestation of severe malaria is a result of the host's own immune response against parasite-derived molecules, causing the toxic, hyperinflammatory response associated with severe malaria (Boutlis et al., 2002; Krishnegowda et al., 2005; Patel et al., 2007). The Plasmodium falciparumspecific glycoform of glycosylphosphatidylinositol (GPI) was identified as the putative toxin implicated in disease (Schofield and Hackett, 1993). GPI molecules are present on the surface of virtually all eukaryotic cells and serve as surface protein anchors, but parasite-specific GPIs occur at relatively high levels in parasitic protozoa (Gowda, 2002). Vaccination strategies aimed at neutralizing the effect of this malaria toxin are being pursued (Schofield, 2007). Similar carbohydrate vaccine strategies for other protozoan parasites are also employed (Buxbaum, 2013). Anti-toxin vaccines have proven successful for other microbial infections, such as the diphtheria toxoid vaccine (Playfair et al., 1990; Schofield, 2007).

We will explore the latest developments for both traditional and non-traditional carbohydrate vaccine approaches for three of the world's major protozoan parasitic diseases—malaria, toxoplasmosis, and leishmaniasis.

### Plasmodium

Malaria is an intra-erythrocytic parasitic disease caused by Plasmodium protozoan species. It is among the most devastating infectious diseases of human history and over 3.3 billion people across 97 countries are at risk of infection today (WHO, 2013b, 2014). An estimated 198 million new clinical cases of malaria occur globally with over 80% of new clinical cases occur in sub-saharan Africa alone (WHO, 2013b). Severe malaria disease develops in 5% of P. falciparum infections and accounts for 98% of all malaria-related deaths. There are an estimated 584,000 deaths per year attributed to severe disease, mostly young children under the age of 5 (WHO, 2013b). Estimates of severe disease-related deaths are as high as 1.2 million (Murray et al., 2012). Without a vaccine, at least e2.4 billion are spent annually on malaria control programs using bed nets, insecticides and drug treatments (WHO, 2013b).

A malaria vaccine should be possible considering that naturally acquired immunity to disease symptoms develops over time. Vaccines against malaria aim to reduce morbidity and mortality, and should be effective in protecting against severe malaria. In the long term, the vaccine should also protect against all clinical disease (WHO, 2013a). To this end, many stages of the parasite lifecycle are targeted by vaccines that decrease parasite load. Examples of some prophylactic vaccines include the whole P. falciparum sporozoite vaccine currently in field trials (Seder et al., 2013), and the protein antigen based RTS,S vaccine which is the most advanced example of malaria vaccine currently under development in Phase III (John et al., 2008; Tinto et al., 2015; Gosling and von Seidlein, 2016) and Phase IV trials (WHO, 2016b). At present, the RTS,S vaccine has not been licensed for use as a malaria vaccine (WHO, 2016a).

#### Anti-Toxin Vaccine

Severe malaria pathology is largely considered to arise from toxic effects of P. falciparum GPI (PfGPI), which exists either as protein free glycolipids, or as the major carbohydrate modifications for proteins essential for erythrocyte invasion (Gowda et al., 1997; Schofield and Grau, 2005). PfGPI-specific antibodies are found in adults of malaria-endemic areas, and may be inhibiting the ability of PfGPI to induce the hyper-inflammatory response associated with severe malaria. This is still up for debate, since studies that find an association between PfGPI-specific antibody titer and protection from severe malaria (Brasseur et al., 1990; Naik et al., 2000; Gowda, 2002; Keenihan et al., 2003; Perraut et al., 2005) are balanced by studies that find no such association (de Souza et al., 2002; Boutlis et al., 2005; Cissoko et al., 2006; Gomes et al., 2013; Mbengue et al., 2016). However, PfGPI-specific antibodies are reported to show relevant action in modulating immune responses by protecting immune cells against severe P. falciparum-induced inflammatory responses in-vitro (Schofield et al., 1993; de Souza et al., 2010).

To evaluate the effect of using PfGPI in an anti-toxin vaccine, a PfGPI hexasaccharide was synthesized, conjugated to KLH carrier protein and emulsified in CFA (**Figure 2**). C57BL/6 mice were immunized with the glycoconjugate, and then challenged with P. berghei ANKA in a mouse model of malaria. Immunization resulted in significant protection against severe malaria, with clearly reduced death rates at 75% survival. The immunized mice were also protected from acidosis, pulmonary oedema, cerebral syndrome and fatality characteristic of the disease model. Apparently, the induction of protective, PfGPI-specific antibodies ameliorated a hyper-inflammatory response against the PfGPI toxin, specifically shown to neutralize the parasite-induced production of TNF-α by macrophages in vitro. Furthermore, vaccination did not change parasitaemia levels, demonstrating that the effect of the carbohydrate vaccine candidate was through the neutralization of GPI's toxic effects, rather than interfering with parasite replication (Schofield et al., 2002).

Following this work, synthetic PfGPI were utilized as biomarkers to improve our understanding of the PfGPI-specific antibody response (Kamena et al., 2008; Tamborrini et al., 2010). The synthesis of PfGPI molecules continues to be explored (Gurale et al., 2016).

#### Sterile Immunity Vaccine

In healthy adults, up to 1–5% of circulating IgG and IgM are specific to the carbohydrate known as α-Gal (**Figure 3**; Macher

Ino-1,2-cyclic-phosphate that was chemically synthesized, conjugated to KLH, and used for immunization. The carrier protein was attached to the top PEtN moiety, where PEtN is an abbreviation for phosphoethanolamine. The full PfGPI (in gray) is shown for context, where R groups denote lipid moieties.

and Galili, 2008). In contrast to other mammals, humans do not express α-Gal (Galili and Swanson, 1991) allowing for immune reactivity to this carbohydrate (Galili et al., 1984). Exposure to microbiota expressing this glycan drives production of α-Galspecific antibodies (Macher and Galili, 2008) which are believed to contribute to immune resistance against α-Gal-expressing pathogenic microbes (Bishop and Gagneux, 2007; Cywes-Bentley et al., 2013). P. falciparum and other animal model Plasmodium spp. express the α-Gal carbohydrate (Galili et al., 1998) possibly bound to GPI-anchored surface proteins (Yilmaz et al., 2014). Early studies already uncovered an association between α-Galspecific IgM and protection from P. falciparum infection in humans. Thus, the effect of immunization on the production of α-Gal-specific antibodies was investigated (Yilmaz et al., 2014).

Genetically modified mice unable to express α-Gal (Yang et al., 1998) were immunized for the production of α-Gal IgG and IgM. The immunizations were either inoculation with α-Gal expressing E. coli bacteria, α-Gal rich rabbit RBCs, or synthetic α-Gal conjugated to BSA. Adjuvants used were CFA and tolllike receptor 9 agonists that enhanced the immunogenicity. The mice were then challenged with P. berghei ANKA infection by Anopheles mosquito bites. It was found that α-Gal immunization reduced the risk of parasite transmission, thereby providing sterile immunity. This sterile immunity appears to be due to the cytotoxic action of α-Gal specific IgM and subclasses of IgG against the inoculating parasites. Moreover, the antibodies were shown to inhibit hepatocyte transmigration of the parasite (Yilmaz et al., 2014).

Sterile immunity against malaria infection through the induction of α-Gal-specific antibodies is not similarly present for people living in malaria endemic regions, possibly due to low levels of naturally acquired, protective α-gal-specific antibodies. Promoting a T cell-dependent immune response against the α-Gal carbohydrate was shown to enhance the protective effect of these antibodies in the mouse model (Yilmaz et al., 2014). Moreover, naturally acquired α-galspecific antibodies may enhance the immunogenicity of antigens enriched with α-gal epitopes by augmenting the T cell response following immunization. This suggests that coupling the α-Gal carbohydrate to existing malaria vaccine candidates could enhance their immunogenicity (Benatuil et al., 2005; Yilmaz et al., 2014). The effect of α-Gal carbohydrate vaccination against other protozoan parasites expressing α-Gal, such as Trypanosoma spp. and Leishmania spp., may also be considered (Yilmaz et al., 2014).

### Toxoplasma

Toxoplasma gondii is found worldwide, capable of infecting nucleated cells of many warm-blooded animals (McLeod et al., 2009; Debierre-Grockiego and Schwarz, 2010) and is estimated to infect half of the world's population (Liu et al., 2012). The disease burden for humans has been well-documented (McLeod et al., 2009). Transmission to humans is either through consumption of food contaminated with tissue cysts and meat products from infected animals or by ingestion of oocysts released in the feces of infected cats (Kijlstra and Jongert, 2008).

Immunocompetant individuals tolerate T. gondii infection which is either asymptomatic or manifest with mild flu-like symptoms. However, the formation of parasite-containing tissue cysts prevents the clearance of the parasite after infection (Montoya and Liesenfeld, 2004). When latent carriers of T. gondii are later immunocompromized they are at risk of severe inflammatory reactions in the brain and central nervous system after liberation of the parasites from these cysts (Luft and Remington, 1992). Pregnant women encountering their first infection can also transmit the parasite to the unborn child leading to retardation or abortion (Remington et al., 2004).

Early vaccination strategies using live parasites (Cutchins and Warren, 1956) and fixed parasites (Krahenbuhl et al., 1972) protected from subsequent challenge. Today, vaccines against toxoplasmosis aim to limit acute parasitemia, protect against congenital toxoplasmosis, reduce the number of tissue cysts, or lessen parasite transmission (Kur et al., 2009). To date, sterile immunity against T. gondii has not been achieved (Jongert et al., 2009). A live vaccine for veterinary toxoplasmosis exists to give limited protection during pregnancy (Buxton and Innes, 1995) but is not approved for human use and does not fully eliminate the parasite (Liu et al., 2012).

#### GPI Vaccine

Glycolipid GPI anchors of T. gondii (TgGPI) have been considered as possible vaccine antigens due to their effect in modulating inflammatory TNF-α responses against the parasite (Debierre-Grockiego, 2010). In 2015, TgGPI were explored as possible vaccine candidates for the first time.

GPI anchors are highly abundant on T. gondii (∼10<sup>6</sup> copies per cell) (Tsai et al., 2012) and exist as a protein-attached or protein-free glycoform (**Figure 4**). Both glycoforms induce inflammatory reactions like macrophage TNF-α production (Debierre-Grockiego et al., 2003) through TLR-2 and TLR-4 signaling (Debierre-Grockiego et al., 2007), which exacerbates toxoplasmosis in mice (Hunter et al., 1996). A vaccine that

induces TgGPI-specific antibodies could ameliorate TgGPImediated inflammatory effects and lower disease burden (Debierre-Grockiego, 2010).

The two major TgGPI glycoforms were synthethized and covalently conjugated to CRM197. BALB/c mice were then immunized with either one of the glycoconjugates before challenge with virulent T. gondii RH strain. Immunization did not provide protection for mice in the lethal challenge model. Furthermore, the induced antibodies failed to exhibit any significant effect on the inflammatory response for either group (Gotze et al., 2015). Analysis of the immune response indicates that antibody induction was directed away from the desired carbohydrate side branch of the TgGPI, and more toward the linker used to attach the carbohydrate to the carrier protein. In a potential next step, the formulation has to be adjusted (Gotze et al., 2015).

Ongoing research utilizing synthetic TgGPI molecules has helped to identify biomarkers of acute and latent infection. During acute infection, high levels of TgGPI-specific IgM and IgG are present, while latent infection shows a reduced IgM response (Gotze et al., 2014).

### Leishmania

There are two million new cases of leishmaniasis every year as the disease is increasingly becoming a worldwide health burden (Desjeux, 2004). The vector borne, facultative intracellular parasite (Chappuis et al., 2007) enters mononuclear, phagocytotic cells such as macrophages. Cutaneous leishmaniasis is the most common form of disease, notable for skin ulcers that result in disability and scarring for the infected patient (Seeberger, 2007; WHO, 2017). Leishmania donovani causes visceral leishmaniasis and is the most severe form of disease characterized by fever, substantial weight loss, anemia, swelling of liver and spleen, and possible death. It is believed to be second only to malaria in terms of fatal infection (Seeberger, 2007; Aebischer, 2014). Patients are treated with antimony drugs which are costly, toxic and increasingly ineffective against resistant parasite strains. A vaccine against leishmaniasis is a desirable, economic strategy to combat this disease (Lee et al., 2012; Singh et al., 2016).

A cocktail of heat-killed Leishmania parasites is a clinically tested vaccine. However, the efficacy of the vaccine to prevent disease is not confirmed (Armijos et al., 2004; Velez et al., 2005). Other attempts involving killed or attenuated parasites for leishmanial vaccine development have not resulted in a licensed vaccine (Topuzogullari et al., 2013).

### Vaccine Efforts

Leishmania parasites express lipophosphoglycans (LPG) on their cell surface. This molecule is composed of a GPI anchor, a repeating phosphorylated disaccharide fragment, and variable cap oligosaccharides (Seeberger, 2007). The LPGs are important for survival and virulence of the parasite (Spath et al., 2000) and vaccination preparations with purified LPG is protective against cutaneous leishmaniasis (McConville et al., 1987; Russell and Alexander, 1988; Moll et al., 1989). Regarding the variable capping oligosaccharide, a unique capping tetrasaccharide (**Figure 5**) was identified as vital for parasite invasion of macrophages (Descoteaux and Turco, 2002) and was the key component of a synthetic carbohydrate vaccine for leishmaniasis (Liu et al., 2006).

The capping oligosaccharide moiety of LPG was synthesized in initial immunological studies exploiting this antigen for vaccination. The carbohydrate was loaded onto virosomes that served as an integrated carrier and adjuvant, before being used to immunize BALB/c mice. Oligosaccharide-specific IgM and IgG1 responses were produced, indicating that the immune response to the carbohydrate antigen was T cell-dependent. Furthermore, the antibodies against the synthetic carbohydrate were crossreactive with natural carbohydrate antigens of Leishmania parasites indicating its possible utility as a vaccine antigen (Liu et al., 2006). A further study immunized BALB/c mice with synthetic LPG capping oligosaccharides conjugated to CRM<sup>197</sup> and emulsified in CFA. This vaccine candidate produced IgG antibodies specific for the parasite. The oligosaccharides were used to evaluate immune responses of infected humans and dogs as the basis for a diagnostic test (Anish et al., 2013). Given the carbohydrate vaccines currently in production, animal model

challenge studies to test the protective effect of these synthetic molecules is the next logical step.

LPG vaccination to protect against cutaneous leishmaniasis employed animal challenge studies that evaluated the LPG component of whole L. amazonensis antigen (LaAg). It was known that intramuscular, systemic immunization of LaAg results in deleterious disease outcomes after challenge, however LPG depletion rendered LaAg protective against Leishmania infection with respect to lesion growth and parasite load compared to non-depleted LaAg. This indicated that LPG was the component responsible for deleterious effects of LaAg inoculation. During further investigations, mice were intranasally vaccinated with LPG alone to determine whether intranasal vaccination of the diseasepromoting component could promote protection against cutaneous leishmaniasis (Pinheiro et al., 2007). Intranasal vaccination with LPG was protective, and provided further evidence for the potential of utilizing LPG in a carbohydrate antigen vaccine that would protect against cutaneous leishmaniasis.

Recently, the role of L. mexicana protein-free GPI molecules (GIPL) in disease has been elucidated, shedding more light on immunologic pathways affecting glycolipid-specific antibody responses (**Figure 6**). After it was determined that L. mexicana infection induces GIPL-specific IgG1 responses, a monoclonal antibody against GIPL was shown to bind to the surface of

parasites, and promote IL-10 production in macrophages cocultured with parasite. Production of IL-10 is deleterious as it blocks an effective immune response that is needed to kill parasites and resolve skin lesions in cutaneous leishmaniasis. In humans, GIPL-specific antibodies are produced in response to infection with cutaneous leishmaniasis. Opsonization of parasites with these antibodies was shown to promote deleterious IL-10 production in macrophages. The role of GIPL-specific antibodies in both mouse and humans opens the possibility of generating a carbohydrate vaccine that can induce competing, non-pathogenic antibody isotypes that can protect against L. mexicana infection (Buxbaum, 2013).

### CONCLUDING REMARKS

Carbohydrate vaccines represent a promising application of glycobiology to human health. The immense successes of bacterial CPS vaccines should, in theory, be similarly achievable with parasite carbohydrate vaccines. With continued advances in parasite culture handling, and carbohydrate synthesis technologies, it is a great time to apply the development strategies employed for bacterial CPS vaccine research to parasite carbohydrate vaccine research.

### AUTHOR CONTRIBUTIONS

JJ contributed to the drafting of the article. PS and JJ contributed to all revisions of the article.

### REFERENCES


### FUNDING

The authors thank the Max-Planck Society for generous financial support. JJ is funded by DFG Research Training Group Parasite Infections (GRK 2046).

### ACKNOWLEDGMENTS

The authors gratefully acknowledge Paulina Kaplonek for critically editing the manuscript.


diverse prokaryotic and eukaryotic pathogens. Proc. Nat. Acad. Sci. U.S.A. 110, E2209–E2218. doi: 10.1073/pnas.1303573110


**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 Jaurigue and Seeberger. 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.

# Engineering of Genetically Arrested Parasites (GAPs) For a Precision Malaria Vaccine

Oriana Kreutzfeld\*, Katja Müller\* and Kai Matuschewski\*

Department of Molecular Parasitology, Institute of Biology, Humboldt University, Berlin, Germany

Continuous stage conversion and swift changes in the antigenic repertoire in response to acquired immunity are hallmarks of complex eukaryotic pathogens, including Plasmodium species, the causative agents of malaria. Efficient elimination of Plasmodium liver stages prior to blood infection is one of the most promising malaria vaccine strategies. Here, we describe different genetically arrested parasites (GAPs) that have been engineered in Plasmodium berghei, P. yoelii and P. falciparum and compare their vaccine potential. A better understanding of the immunological mechanisms of prime and boost by arrested sporozoites and experimental strategies to enhance vaccine efficacy by further engineering existing GAPs into a more immunogenic form hold promise for continuous improvements of GAP-based vaccines. A critical hurdle for vaccines that elicit long-lasting protection against malaria, such as GAPs, is safety and efficacy in vulnerable populations. Vaccine research should focus on solutions toward turning malaria into a vaccine-preventable disease, which would offer an exciting new path of malaria control.

#### *Edited by:*

Nathan W. Schmidt, University of Louisville, United States

#### *Reviewed by:*

Jason Scott Stumhofer, University of Arkansas for Medical Sciences, United States Scott E. Lindner, Pennsylvania State University, United States

#### *\*Correspondence:*

Oriana Kreutzfeld kreutzfo@hu-berlin.de Katja Müller muekatja@hu-berlin.de Kai Matuschewski kai.matuschewski@hu-berlin.de

> *Received:* 14 March 2017 *Accepted:* 04 May 2017 *Published:* 31 May 2017

#### *Citation:*

Kreutzfeld O, Müller K and Matuschewski K (2017) Engineering of Genetically Arrested Parasites (GAPs) For a Precision Malaria Vaccine. Front. Cell. Infect. Microbiol. 7:198. doi: 10.3389/fcimb.2017.00198 Keywords: malaria, *Plasmodium*, vaccine, liver stage, live attenuated parasite, sporozoite, immune memory

## INTRODUCTION

Malaria remains the most important vector-borne infectious disease and affects half of the world's population. Globally, >200 million infected individuals develop clinical symptoms, and >400,000 die because of severe malaria, primarily children in Sub-Saharan Africa (WHO, 2016a). New strategies for malaria prevention and eradication are thus urgently required. The current malaria control programs target the causative agents, Plasmodium falciparum, P. vivax, and three other human-infecting Plasmodium parasites, at different life cycle stages, which together reduces morbidity and mortality in endemic regions. Attacking the parasite at its vector stage with long-lasting insecticide treated bed nets and insecticide-indoor residual spraying combined with access to rapid diagnosis and artemisinin-based combination therapy for clinical malaria episodes are recommended by the WHO (WHO, 2016a).

**Abbreviations:** β2M, β<sup>2</sup> microglobulin; CAS, chemically arrested sporozoites; CSP, circumsporozoite protein; DC, dendritic cell; ETRAMP, early transcribed membrane protein; FASII, fatty acid synthesis II; G3PAT, glycerol-3-phosphate acyltransferase; G3PDH, glycerol-3-phosphate dehydrogenase; GAP, genetically arrested parasite; IK2, eukaryotic initiation factor-2α kinase; MHC, major histocompatibility complex; PALM, Plasmodium-specific apicoplast protein important for liver merozoite formation; PDH, pyruvate dehydrogenase; PKG, protein kinase G; PP1, eIF2α-P protein phosphatase 1; MSP1, merozoite surface protein 1; PVM, parasitophorous vacuole membrane; RAS, radiation attenuated sporozoites; SLARP/ SAP1, sporozoite and liver stage asparagine-rich protein; TRAP, thrombospondin-related anonymous protein; UIS, upregulated in infectious sporozoites.

Repeated exposure to Plasmodium transmission in malaria-endemic countries leads only to very slow acquisition of naturally acquired immunity that rapidly wanes. Development of protective immunity is likely hindered by blood infection, the exclusive cause of malaria-related morbidity and mortality. Therefore, the pre-erythrocytic phase of the Plasmodium life cycle in the mammalian host is particularly attractive as immunization agent since no clinical symptoms are associated with this first replication phase (Prudêncio et al., 2006; Silvie et al., 2008a; Matuschewski et al., 2011). It also allows priming of CD4<sup>+</sup> and CD8<sup>+</sup> T cells by presenting parasite antigens to the host immune system via MHCI and MHCII by dendritic cells and MHCI by infected hepatocytes, respectively (Hafalla et al., 2011). Prevention of disease by vaccination is an ideal addition to the portfolio of malaria intervention tools; however, it remains one of the greatest challenges in medical research.

Approaches to design a protective long-lasting malaria vaccine are wide-ranging, and some of them have already reached clinical trial phases (WHO, 2016b). Among these candidates is RTS,S/AS01, which is the first malaria vaccine that achieved licensure for administration in malaria endemic areas. RTS,S/AS01 is a hepatitis B-based subunit vaccine and contains a fragment of the circumsporozoite protein (CSP) (Stoute et al., 1997). Recent phase III clinical trial results revealed an unfavorable protection level casting doubt on the impact of RTS,S/AS01 in malaria control efforts (RTS,S Clinical Trials Partnership, 2011; Olotu et al., 2016).

A benchmark for a malaria vaccine exists for several decades already, and this experimental vaccine with proven long-term protection is a whole sporozoite vaccine. Small immunization studies in mice, non-human primates, and humans demonstrated that radiation attenuated sporozoites (RAS) elicit sterile protection against Plasmodium challenge infections (Nussenzweig et al., 1967, 1969; Clyde et al., 1973; Gwadz et al., 1979; Hoffman et al., 2002). Irradiation induces DNA breakage in the parasites, which reduces the numbers of nuclear divisions and limits liver stage expansion to early schizonts (Silvie et al., 2002). Timing and radiation dosage is critical since over-irradiated sporozoites arrest early at the unicellular stage, thus decreasing protection in experimental cohorts (Friesen and Matuschewski, 2011). Successful hepatocyte invasion and initial intra-hepatic development of live, metabolically active sporozoites are required to elicit strong immunity, and, therefore, pose a substantial hurdle toward an affordable and undemanding sporozoite vaccine. The central importance of live and metabolically active sporozoites has been corroborated with heat-killed sporozoites, which elicit only very weak and short-term, antibody-mediated protection against subsequent Plasmodium sporozoite challenge infections (Hafalla et al., 2006). On the other hand, suboptimal irradiation harbors the risk of breakthrough infections during vaccination a considerable drawback concerning safety. RAS arrest early during liver stage development and express primarily sporozoitederived antigens, including CSP and thrombospondin-related anonymous protein (TRAP), which likely contribute to priming of T cell-mediated immunity.

Liver stage developmental arrest can also be achieved by an alternative approach, termed chemical attenuation of sporozoites (CAS), which is based on simultaneous administration of normal sporozoites and anti-malarial drugs (Belnou et al., 2004; Putrianti et al., 2009; Roestenberg et al., 2009; Friesen et al., 2010; Bijker et al., 2013). While they offer interesting evaluations in small-scale exploratory clinical studies, these approaches are critically reliant on continuous clinical supervision during drug administration and currently bear no translational perspective.

The potential to induce lasting protection by live attenuated, metabolically active parasites led to the engineering of genetically attenuated parasites (GAPs) as tailored whole parasite vaccines against malaria infections (Mueller et al., 2005a). Murine malaria models employing the rodent malaria parasites P. berghei and P. yoelii enable the exploration of liver stage-specific proteins and their importance for parasite survival. In the past years, over 120 genes have been targeted by experimental genetics and analyzed for defects during the Plasmodium life cycle in both vector and mammalian host (Janse et al., 2011). Many gene knockouts resulted in normal parasite life cycle progression or refractoriness to targeted deletion, indicative of redundant functions during life cycle progression or vital roles for blood infection, respectively. Additionally, arrest of the parasite development in the mosquito midgut, prior to salivary gland colonization, or ahead of hepatocyte invasion was frequently observed. Accordingly, only very few candidate genes fulfill the criteria of potential GAP vaccine candidate lines (**Table 1**).

Herein, we review the most recent developments in GAP vaccine discovery. We assess the different GAPs that have been generated in the last 12 years and evaluate their vaccine potential. Finally, we address the challenges and obstacles in designing a GAP vaccine for vulnerable populations in malaria-endemic countries.

### FIRST GENERATION GAPs: PROOF OF PRINCIPLE STUDIES

The first preclinical studies that showed successful generation of GAP lines and their efficacy in immunization protocols targeted Plasmodium berghei genes that represent members of the early transcribed membrane protein (ETRAMP) family (Mueller et al., 2005a,b). The selected genes, i.e., upregulated in infective sporozoites gene 3 (UIS3) and UIS4, fulfilled three principal criteria: (i) stage-specific gene expression in pre-erythrocytic stages, thereby allowing recombinant knockout strains to be selected during blood stage transfection; (ii) abundant expression in pre-erythrocytic stages, indicating likely vital roles during this phase of the life cycle; and (iii) genes that are unique to Plasmodium and related haemosporidian parasites, which all share the hallmark of the first obligate population expansion phase in the host liver (Matuschewski et al., 2002). In vitro studies in cultured hepatoma cells revealed that P. berghei as well as P. yoelii 1UIS3 and 1UIS4 parasite lines arrest early in liver stage development after completion of sporozoite transformation to liver stages, but before onset of parasite


#### TABLE 1 | *Plasmodium* genes targeted for GAPs. *Plasmodium berghei* (black), *P. yoelii* (green) and *P. falciparum* (blue) are listed.

<sup>a</sup>References list only the first report in the respective Plasmodium species.

b Incomplete stage-specific gene knockout.

replication (**Figure 1**; Mueller et al., 2005a,b; Tarun et al., 2007).

Immunization studies with one prime and two booster vaccinations showed complete long-term protection with P. berghei and P. yoelii 1UIS3 and 1UIS4 GAPs in C57BL/6 and BALB/cJ mice, respectively (**Table 2**) (Mueller et al., 2005a,b; Tarun et al., 2007). The study on P. berghei 1UIS4 GAPs established frequent occurrence of breakthrough infections, i.e., a proportion of animals develop blood infections during the immunization procedure (Mueller et al., 2005b). This safety concern needs careful examination before progression of P. falciparum GAPs to human testing can take place (Matuschewski, 2013). Notably, the identification of orthologous genes for UIS3 and UIS4 in P. falciparum has so far remained elusive, although various ETRAMPs emerge as potential candidates (Spielmann et al., 2012). To date, only one P. falciparum ETRAMP has been targeted by reverse genetics (MacKellar et al., 2010). This protein, termed Pf ETRAMP 10.3, apparently performs essential functions during blood infection, since it remains refractory to targeted deletion and does not complement the P. yoelii 1UIS4 defects. Accordingly, more research is needed to identify and target P. falciparum ETRAMPs

GAPs are depicted in black and gray, respectively. Knockouts of the murine and human Plasmodium species are shown on the left and right side, respectively.

to generate this class of GAPs in P. falciparum and test their vaccine potential in small-scale human trials.

Another first generation GAP line in P. berghei and P. yoelii are parasites that lack P36p (also termed P52), a member of the Plasmodium-specific 6-Cys protein family (van Dijk et al., 2005). Parasites lacking P36p (P52) or its paralog, P36, arrest again early during liver stage development after completion of sporozoite transformation (**Figure 1**) (Ishino et al., 2005; van Dijk et al., 2005). 1P36p sporozoites transmigrate and invade hepatocytes, wherein they initiate the formation of a parasitophorous vacuole membrane (PVM). Although, they start to mature into liver stage trophozoites, 1P36p parasites suddenly abort this development, most likely because maintenance and maturation of the PVM are critically impaired (van Dijk et al., 2005).

Once more, occasional breakthrough infections in mice inoculated with 1P36p sporozoites were observed (van Dijk et al., 2005). Interestingly, in the 1P36p/P36 double knockout parasite expression of the signature merozoite surface protein 1 (MSP1) could not be detected, and the mechanism of breakthrough infections remains unsolved (Ploemen et al., 2012). Targeted gene deletion of a second member of the 6-Cys family, termed B9, led to similar early arrest and occasional breakthrough infections (Annoura et al., 2014).

Since genes of the 6-Cys protein family are remarkably conserved across Plasmodium species, P. falciparum GAP lines were generated to show proof of principle of liver stage attenuation by targeted gene deletion in human malarial parasites (van Schaijk et al., 2008; van Buskirk et al., 2009). However, in order to advance this approach to clinical testing in humans, safety is of utmost importance. Accordingly, it remains enigmatic why a Pf GAP based on P52 (P36p) and P36 was selected for a first phaseI/IIa clinical trial (Spring et al., 2013), despite the alarming

#### TABLE 2 | *P. berghei* and *P. yoelii* GAPs: long-term immunization studies.


<sup>a</sup>P. berghei and P. yoelii knockout parasite lines are displayed in black and gray, respectively.

<sup>b</sup>Only studies where i. v. challenge was performed at least 3 weeks after the last boost are listed.

c Immunizations were performed in the P. berghei-C57BL/6 and P. yoelii-BALB/cJ combinations, respectively. preclinical data. As predicted, human trials with Pf1P52/P36 had to be suspended because of breakthrough infections (Spring et al., 2013).

A potential, albeit untested, advantage of 6-Cys proteinbased GAPs is enhanced antigen presentation via MHCI on infected hepatocytes, since maintenance of the parasitophorous vacuole is impaired (van Dijk et al., 2005). Perhaps even more important, the propensity to increase apoptosis in 1P36pinfected hepatocytes (van Dijk et al., 2005) might enhance crosspriming by dendritic cells (DCs) that phagocytose apoptotic bodies (Leiriao et al., 2005), although this conjecture remains controversial (Renia et al., 2006).

Together, the first generation GAPs established that precise developmental arrests during the first clinically silent, intrahepatic Plasmodium expansion phase can be engineered by tailored removal of individual vital genes from the entire Plasmodium genome (**Table 1**). These uniform, genetically defined parasites consistently elicit lasting protection against sporozoite challenge infections in vaccination protocols with three consecutive GAP sporozoite inoculations (**Table 2**).

### GAPs TARGETING LIVER STAGE DIFFERENTIATION: SAFETY FIRST

The observations of breakthrough blood infections during the immunization protocol in a proportion of animals (Mueller et al., 2005b; van Dijk et al., 2005) initiated the search for candidate Plasmodium genes that are key developmental factors, for instance transcription factors at the nexus of sporozoite to liver stage transformation. Such a factor was identified during the analysis of sporozoite-specific (S) genes (Kaiser et al., 2004). Targeted gene deletion of this factor, termed S22 or sporozoite and liver stage asparagine-rich protein (SLARP or SAP1) (Aly et al., 2008; Silvie et al., 2008b), resulted in a complete arrest of the parasite at early liver stage development prior to nuclear division (**Figure 1**). A novel hallmark of 1SLARP parasites was the differential down-regulation of many liver stage-specific mRNAs and their corresponding proteins, including PVMresident proteins such as UIS3 and UIS4 (Aly et al., 2008; Silvie et al., 2008b), suggesting pleiotropic defects as a result of the absence of a major transcriptional regulator of liver stage differentiation.

Most importantly, 1SLARP parasites are the first and only GAPs reported to fulfill all criteria of safe arrest (Aly et al., 2008; Silvie et al., 2008b). Accordingly, after the safety failure of Pf1P52/P36, all Pf GAPs developed for clinical testing in humans include the corresponding PfSLARP knockout (Mikolajczak et al., 2014; van Schaijk et al., 2014a; Kublin et al., 2017). However, immunization studies showed a lack in long-term protection in animals immunized with P. berghei 1SLARP parasites, where only 40% of all animals were protected 3.5 months after immunization (**Table 2**) (Silvie et al., 2008b). It is conceivable that 1SLARP parasites display a smaller array of antigens, but this has to be experimentally tested employing systems immunology approaches.

Notably, targeting of other key factors important for liver stage gene expression did not result in a similar complete liver stage arrest. Deletion of the liver stage-specific transcription factor of the apetala 2 family, termed AP2-L, only resulted in a developmental delay (Iwanaga et al., 2012). Knockout of the eukaryotic initiation factor 2α (eIF2α) kinase (IK2/UIS1), which is critical for sporozoite latency, and stage-specific knockout of the corresponding eIF2α-P protein phosphatase 1 (PP1/UIS2), which is a regulator of protein translation after hepatocyte invasion, led to incomplete early arrests before and immediately after hepatocyte invasion (Zhang et al., 2010, 2016). Targeted deletion of another regulator of sporozoite latency, the RNAbinding protein PUF2, reproduced the 1IK2/UIS1 phenotype, again with an unsatisfactory safety profile (Gomes-Santos et al., 2011; Müller et al., 2011).

In conclusion, immunization data together with the demonstration of a very early, complete arrest indicate that 1SLARP parasites are comparable to RAS, with the important distinctions of a precision life cycle arrest in humans (Kublin et al., 2017) and safe handling of 1SLARP-infected Anopheles mosquitoes for vaccine production.

### LATE ARRESTING GAPs: IMPROVED IMMUNOGENICITY BUT LACK OF SAFETY

Studies employing co-administration of normal sporozoites and anti-malarial drugs have consistently shown superior immunity of late liver stage and/or early blood cycle arrest in murine malaria models (Belnou et al., 2004; Friesen et al., 2010; Friesen and Matuschewski, 2011) and small scale human trials (Roestenberg et al., 2009; Bijker et al., 2013), suggesting that a late liver stage arrest offers multiple advantages, perhaps including broader antigen presentation (Borrmann and Matuschewski, 2011). Unexpectedly, antibiotic-induced arrest at the transition from late liver stages to blood infection leads to better protection than a later arrest, after a few rounds of blood stage replication, induced by chloroquine treatment (Friesen and Matuschewski, 2011). This indicates an immune-modulatory effect by infected red blood cells and potential benefits of a complete arrest at the liver stage.

Therefore, tailored arrest toward the end of liver stage maturation was an important third step in GAP vaccine design. A CAS-based arrest using the antibiotic azithromycin showed that specific targeting of the Plasmodium apicoplast, a relict non-photosynthetic plastid organelle, resulted in late arrest after complete liver stage maturation (Friesen et al., 2010). Accordingly, two complementary approaches targeting key factors in the Plasmodium apicoplast led to generation of late-arrested GAPs and their testing in vaccine studies (**Figure 1**), namely deletion of a fatty acid biosynthesis enzyme (Butler et al., 2011) and a Plasmodium-specific protein of unknown function (Haussig et al., 2011). In both cases potent protection against reinfection was reported and superior protection correlated with extended liver stage maturation.

The fatty acid synthesis II (FASII) pathway produces saturated fatty acids in the apicoplast. It includes a cyclic reaction that catalyzes fatty acid elongation and a large pyruvate dehydrogenase (PDH) complex that forms acetyl-CoA for the elongation cycles (Yu et al., 2008). Targeted deletion of an enzyme of the cyclic reaction, namely trans-2-enoyl-ACP reductase (FabI), in P. berghei revealed a specific defect during liver stage maturation (Yu et al., 2008). However, the life cycle arrest was incomplete and sporozoite inoculations resulted in substantial breakthrough infections in C57BL/6 mice (**Table 1**) (Yu et al., 2008). Accordingly, immunization studies were not performed.

The liver stage defects of mutants in the FASII biosynthesis pathway in murine malaria parasites were confirmed in P. yoelii studies by targeted deletion of 3-oxoacyl-ACP synthase I/II (FabB/F) and β-hydroxyacyl-ACP dehydratase (FabZ) (Vaughan et al., 2009). Since breakthrough infections were once more absent in the P. yoelii model testing of 1FabB/F parasites as latearrested GAPs in vaccination protocols was possible (**Table 2**) (Butler et al., 2011). This study reported better protection and a larger CD8<sup>+</sup> T cell response in comparison to 1SAP1 and RAS parasites. Targeted gene deletion of additional P. yoelii enzymes of the FASII and the subsequent lipid biosynthesis pathway resulted in similar arrests in liver schizont maturation prior to merozoite formation (Pei et al., 2010; Lindner et al., 2014). The target genes were the E1α and E3 subunits of PDH, glycerol-3-phosphate acyltransferase (G3PAT), and glycerol-3 phosphate dehydrogenase (G3PDH) (**Figure 1**) of which the latter performed well in vaccine protocols (**Table 2**).

The previous observation of breakthrough infections in the first study of a P. berghei FASII pathway knockout (Yu et al., 2008) was subsequently confirmed (**Table 1**) (Annoura et al., 2012; Shears et al., 2017) and strictly limits the results from studies in P. yoelii. The additional, unexpected finding of aborted parasite development in mosquitos infected with the corresponding P. falciparum knockouts (Cobbold et al., 2013; van Schaijk et al., 2014b) essentially eliminated the possibility to develop GAP vaccines by targeted deletion of the FASII pathway. Together, these results also raise the important question, which prerequisites have to be fulfilled to transfer discoveries from murine models to Pf GAP vaccines.

The second strategy of targeting essential apicoplast functions to generate late liver stage-arrested parasite lines built upon bioinformatic prediction of Plasmodium-specific apicoplast targeted proteins (Haussig et al., 2011). The first target that satisfied these criteria was P. berghei Plasmodium-specific apicoplast protein important for liver merozoite formation (PALM), a protein of unknown function. Targeted deletion of PALM did not affect parasite growth or apicoplast morphology, but resulted in an even later arrest after completion of liver stage development prior to merozoite release (Haussig et al., 2011). Although, immunizations with 1PALM resulted in robust long-term protection after only two immunizations in the stringent P. berghei-C57BL/6 model (**Table 2**), consistent doseindependent breakthrough infections preclude the translation to human Plasmodium species, unless this mutant is combined with an additional knockout that causes a similar life cycle arrest. Several attempts to target other biochemical pathways within the apicoplast, including iron-sulfur cluster biogenesis or primary reactions in heme biosynthesis, did not yield 1PALM-like GAPs (Haussig et al., 2013, 2014; Rizopoulos et al., 2016).

Additional late arresting candidate genes have been identified; however, in depth analysis and immunization studies are often lacking. An interesting case is protein kinase G (PKG), which is shared across different life cycle stages and already essential during blood infection, as it plays prominent roles in merozoite egress and gametogenesis. Since generation of a 1PKG parasite is incompatible with blood infection, a stagespecific knockout by FLP/FRT recombination in sporozoites was engineered to study the role(s) in pre-erythryocytic development (Falae et al., 2010). This analysis revealed a late arrest in liver stage maturation and wild type breakthrough infections due to incomplete recombination (**Table 1**). Immunization studies were not conclusive since challenge infections were performed only one week after high dose immunizations. However, the example of PKG illustrates that a tight late arrested parasite might be achieved once a suitable liver stage-specific gene at the nexus of stage conversion is identified.

### GAPs WITH MULTIPLE GENE DELETIONS: SYNERGISTIC OR ANTAGONISTIC?

Development of parasite lines that harbor multiple gene deletions is likely to increase safety; however, whether synergistic or antagonistic effects modulate immunogenicity is less straightforward to predict and likely depends on the selected knockout combination.

Soon after the first proof-of-principle studies a GAP parasite line that harbors two consecutive gene deletions, namely 1UIS3 and 1UIS4, was engineered in P. berghei (Jobe et al., 2007). As expected, 1UIS3/UIS4 parasites displayed a complete arrest in early liver stages, indicating that vaccine strains harboring multiple independent gene deletions perform safer than single knockout GAPs. Importantly, long-term protection against a high-dose sporozoite challenge infection was complete (**Table 2**).

In marked contrast, when a double knockout was performed for the P. berghei paralogs P36p (P52) and P36, which are neighboring genes and likely arose through gene duplication, safety was not improved (Annoura et al., 2012), strongly suggesting that independent genes need to be targeted in multiple gene knockout strategies. Since 1SLARP GAPs lead to a complete termination of liver stage development (Aly et al., 2008; Silvie et al., 2008b), they constituted the obvious platform for combinations with knockouts of the 6-Cys gene family (**Figure 1**) (Mikolajczak et al., 2014; van Schaijk et al., 2014a; Kublin et al., 2017). However, it remains to be shown whether addition of B9, P36, and/or P36p (P52) knockout provides any additional benefit beyond perception of additional gene deletions. An important investigation with combinatorial knockouts will be the systematic expression profiling of liver stage-specific genes, as was previously done for 1SLARP parasites (Silvie et al., 2008b). Such an analysis will provide first insights into the expected antigenic repertoire displayed by the respective GAPs.

Instead of adding additional gene deletions, which might not add significantly to vaccine efficacy, combination of a gene deletion with transgene expression of additional factors could amplify antigen presentation during pre-erythrocytic development. Potential avenues include perforation of the intracellular niche, i.e., the parasitophorous vacuole, activation of innate immune sensing pathways, and expression of blood stage and gametocyte antigens. Such a strategy is exemplified by the expression of perfringolysin O, a cholesterol-dependent cytolysin, in P. berghei 1PDH-E1α parasites (Nagel et al., 2013). Addition of the transgene could substantially reduce, albeit not completely abolish, breakthrough infections of the single gene deletion, providing a rationale for further bioengineering efforts to achieve premature rupture of the PVM. However, a principal concern in gain-of-function mutants is that parasites containing mutations in the transgene promoter and other regulatory elements, which lead to a reduced transgene expression, will be swiftly selected. Therefore, GAPs that express additional Plasmodium antigens in order to broaden the immunogenic repertoire without further reducing parasite fitness might be a particularly rewarding research direction.

### IMMUNE MECHANISMS OF GAPs: THE CENTRAL ROLE OF EFFECTOR MEMORY CD8<sup>+</sup> T CELLS

Cytolytic, interferon gamma (IFNγ)-secreting CD8<sup>+</sup> T cells were identified early on as the key mediators of protection after radiation-attenuated sporozoite immunizations (Schofield et al., 1987; Weiss et al., 1988; Romero et al., 1989). Amongst all CD8<sup>+</sup> T cell subsets, presence of those with an effector memory phenotype, i.e., CD45RBlo, CD44high, CD62Llo, and CD122lo consistently correlated with long-lasting protection (Guebre-Xabier et al., 1999; Berenzon et al., 2003). Studies employing immunization of β<sup>2</sup> microglobulin (β2M) knockout mice and recognition of a CSP-specific T cell clone by hepatocytes with a matching MHCI haplotype revealed that liver stage antigens are presented to the CD8<sup>+</sup> T cells via MHCI on the surface of infected hepatocytes (White et al., 1996; Balam et al., 2012). Elimination of infected hepatocytes in immunized mice could be linked to perforin secreting cytotoxic CD8<sup>+</sup> T cells and IFNγ production (Mellouk et al., 1987; Schofield et al., 1987; Malik et al., 1991; Rodrigues et al., 1993; Sano et al., 2001). Employing advanced in vivo imaging techniques direct proximity to antigen-specific CD8<sup>+</sup> T cells was shown to be required for cytolytic killing of Plasmodium-infected hepatocytes (Cockburn et al., 2013; Kimura et al., 2013). Although, antigenspecific CD4<sup>+</sup> T cells are expected to be central to mount an effective T cell response, their roles in B cell help to produce antibodies that inhibit sporozoite attachment and invasion of the liver is only minor. Antibodies are not essential for vaccineinduced protection, and protective CD8<sup>+</sup> T cell responses can be mounted without CD4<sup>+</sup> help (Schofield et al., 1987; Rodrigues et al., 1993).

Based on the insights from studies with irradiated sporozoites, CD8<sup>+</sup> T cell-dependent elimination most likely is the immune effector mechanism in GAP-immunized animals. Indeed, a study employing P. berghei 1UIS3 parasites showed that immunizations of B and T cell-deficient rag1−/<sup>−</sup> and IFNγ −/− knockout mice failed to induce protection, confirming the central role of adaptive immune responses leading to IFNγ production in vaccine-induced immunity (Mueller et al., 2007). Immunizations of B cell-deficient mice and common laboratory mice after depletion or adoptive transfer of CD4<sup>+</sup> and CD8<sup>+</sup> T cells fully corroborated the central importance of CD8<sup>+</sup> T cells, but neither of antibodies nor of CD4<sup>+</sup> T cells. Of note, primaquine treatment results in efficient cure of liver stage parasites and reversed 1UIS3-mediated protection (Mueller et al., 2007). This is in perfect agreement with the requirement for parasite persistence as metabolically active, cell cycle arrested intra-hepatic stages in order to maintain long-term protection (Scheller and Azad, 1995).

Another early study demonstrated the central role of CD8<sup>+</sup> T cells, and particularly IFNγ-secreting effector memory cells, in protection induced by a P. berghei 1UIS3/UIS4 double knockout parasite line (Jobe et al., 2007). Immunization of β2m−/<sup>−</sup> mice, which are deficient in MHCI expression and CD8<sup>+</sup> T cells, abrogated vaccine efficacy. The direct comparison to RAS showed that both immunization strategies induce similar immune responses, but GAP-immunized animals displayed consistently higher levels of IFNγ-secreting effector memory cells (Jobe et al., 2007). Accordingly, it is conceivable that the insights gained from RAS immunizations can be extrapolated at least to 1UIS3/UIS4 GAP vaccines.

The findings obtained with P. berghei GAPs were fully supported by a study reporting that depletion of CD8<sup>+</sup> T cells, but neither CD4<sup>+</sup> T cells nor IgG1 antibodies, abolished protection by P. yoelii 1UIS3 or 1UIS4 parasite lines (Tarun et al., 2007). Notably, CD8<sup>+</sup> T cells from P. yoelii GAP immunized mice induce apoptosis of infected, in vitro cultured hepatocytes by contact dependent, perforinmediated cytotoxic killing, with only partial involvement of IFNγ (Trimnell et al., 2009). Short-lived CD11ahi, CD62Llo, CD44hi antigen-experienced effector CD8b<sup>+</sup> T cells, which also express CD11c, expanded swiftly after one immunization with P. yoelii 1UIS4 sporozoites (Cooney et al., 2013). Furthermore, these cells are KLRG1+CD127<sup>−</sup> terminal effector cells, which upon restimulation with infected hepatocytes secrete IFNγ, TNF, and IL-2 and express CD107a and perforin. Late-arresting P. yoelii GAPs induced larger CD8<sup>+</sup> T cell responses in comparison to RAS, and the CD8α loCD11ahi effector memory phenotype, characterized by low expression levels of CD27, CD62L, and CD127, was increased; however, this study was done in C57BL/6 mice only (Butler et al., 2011).

While the cellular mechanisms that lead to elimination of infected hepatocytes after immunization are relatively well understood, the target epitopes displayed by infected hepatocytes remain less clear (Hafalla et al., 2011). There is growing evidence that a combination of epitopes rather than a single protective antigen correlates with vaccine-induced protection (Grüner et al., 2007; Hafalla et al., 2013). The present list of pre-erythrocytic T cell epitopes in H2-K<sup>d</sup> -restricted BALB/cJ and H2-K<sup>b</sup> -restricted C57BL/6 mice remains short (Romero et al., 1989; Hafalla et al., 2013; Murphy et al., 2013; Lau et al., 2014; Müller et al., 2017). The few immunogenic epitopes identified thus far underscore the fundamental differences of immunity elicited by eukaryotic pathogens in comparison to viruses and bacteria (Hafalla et al., 2013).

The mechanisms leading to efficient priming of effector immune responses against pre-erythrocytic parasites are still incompletely understood. Efficient priming almost certainly requires cross-presentation by dendritic cells (DCs) and is most likely a combination of DC-mediated antigen presentation in different organs, including skin-draining lymph nodes, spleen, and liver (Sano et al., 2001; Chakravarty et al., 2007; Cockburn et al., 2011; Balam et al., 2012). Most studies focused on the CSP epitope, which is abundantly expressed on the sporozoite surface and only recognized in H2-K<sup>d</sup> -restricted BALB/cJ mice (Romero et al., 1989; Sano et al., 2001; Chakravarty et al., 2007; Balam et al., 2012). Whether GAPs aborting development at the trophozoite stage or later present additional antigens explaining superior CD8<sup>+</sup> T cell responses remains elusive.

### TOWARD UNIFIED GUIDELINES FOR TRANSLATIONAL GAP RESEARCH

There are currently no rules established for GAP development in P. falciparum; however, defining a set of requirements that need to be fulfilled before moving on to P. falciparum studies would significantly decrease odds of clinical trial failures. An obvious criterion from murine malaria models, i.e., P. berghei and P. yoelii, is complete liver stage arrest in a large number of animals employing both murine models simultaneously. Humanized mouse models and primary human hepatocytes can provide preliminary indications on developmental arrest in the liver and, thus, potential safety concerns of a P. falciparum GAP vaccine candidate. However, these test systems might not provide the level of stringency necessary to predict breakthrough infections, which are presently best captured in preclinical tests in large groups of mice. In order to develop potent P. falciparum GAPs beyond early arresting 1SLARP mutants, multiple independent gene deletions, ideally in unrelated physiological processes and in different cellular compartments, need to be considered.

Another desirable standard is robust long-term immunity against challenge infection. Identifying the correct parasite/host combination is important to draw valid conclusions from immunization experiments (Matuschewski, 2013). The P. berghei-C57BL/6 combination remains the most robust vaccine model to date, whereas immunizations using P. yoelii mutants in BALB/c mice are only of modest predictive value, since protection is easily achieved in this model. Other combinations, such as P. berghei and BALB/c mice, are invalid because of refractoriness of particular combinations of mouse strains and infections with certain murine malaria sporozoites (Scheller et al., 1994). This notion is illustrated in immunizations with P. berghei 1P36p sporozoites, where only one immunization dose in an inappropriate host strain, BALB/c mice, resulted in complete protection up to 3 months after immunization (van Dijk et al., 2005). Immunization with P. yoelii 1P36p/P36 parasites also induced sterile protection in BALB/c mice after a single injection (Labaied et al., 2007); however, such protection in C57BL/6 mice was only elicited after three rounds of immunization (van Dijk et al., 2005). One important immunological difference is that protection in the P. yoelii-BALB/c model largely depends on the immunedominant CD8<sup>+</sup> T cell epitope of CSP. In marked contrast, T cell responses in the P. berghei-C57BL/6 model are likely multifactorial (Hafalla et al., 2011, 2013), which closely mimics infections in human populations with a large range of MHCI haplotypes and only infrequent CSP–specific CD8<sup>+</sup> T cell responses (Offeddu et al., 2012). The recent evaluation of Grammomys dolichurus, an Afrotropical arboreal rodent, which naturally harbors rodent Plasmodium infections, as a model to study pre-erythrocytic vaccine strategies will be an important addition for preclinical evaluation of safety and immunogenicity of GAP vaccines (Conteh et al., 2017). In good agreement with the data from murine Plasmodium models, the natural host is highly susceptible to P. berghei sporozoite induced infections and multiple high immunization doses are required for robust protection (Conteh et al., 2017).

Together, comparative studies on vaccine efficacy should include the P. berghei-C57BL/6 model (Friesen and Matuschewski, 2011), to better define long-term protection against re-challenges. We propose that a GAP vaccine line should be evaluated for vaccine safety and immunogenicity in a two-step preclinical process before advancing to human trials. First, comparative evaluation of the candidate P. berghei and P. yoelii GAP lines in the respective mouse strains, C57BL/6 and BALB/c, will pre-select GAPs that are completely arrested upon high dose sporozoite inoculations and elicit long-lasting (>30 days) sterile immunity. Second, confirmation of safety and immunogenicity of the P. berghei GAP line in G. dolichurus immunization and challenge study provide an evidence-based rationale for translation to P. falciparum GAP trials.

### ROADBLOCKS TOWARD TRANSLATION OF GAPs

Intravenous injections are the preferential route of vaccine administration in mouse models; however, in real life this is inapplicable. Pediatric vaccines are exclusively administered either by intramuscular syringe injection or orally. It is evident that a malaria vaccine must adhere to the same safe routes of administration. Alternative methods that are used in drug delivery, such as intradermal, intravenous, or intraperitoneal injections, are unreasonable for a malaria vaccine to be delivered in resource-poor health infrastructures, because of the risks associated with these routes of administration, except under physician's care.

Plasmodium sporozoites apparently lack the ability to transmigrate through muscle or fat tissue. Accordingly, after intradermal, intramuscular, subcutaneous, or intraperitoneal injection only a very small proportion of sporozoites reach a blood vessel, resulting in reduced liver infection. This was confirmed in human volunteer studies, where intramuscular or intradermal syringe injections of cryopreserved sporozoites showed substantial delays in blood infections and high doses were required to consistently induce blood infections (Shekalaghe et al., 2014; Gómez-Pérez et al., 2015). As expected, protection after immunizations via these routes was negligible both in the P. berghei model and in volunteer studies with P. falciparum sporozoites (Epstein et al., 2011; Nganou-Makamdop et al., 2012). Thus, development of a GAP vaccine that can be administered in the muscle is a principal hurdle that needs to be overcome by bioengineering efforts.

The necessity for booster immunizations remains another critical limitation, which is unlikely to be removed. However, a reduction of the vaccine doses would facilitate the distribution of a malaria vaccine considerably. A better understanding of how to best amplify an initial protective CD8<sup>+</sup> T cell response toward a sustained effector-memory T cell response will be essential. It is conceivable that timing of antigen expression is important, and the temporal dynamics of expansion and contraction of antigenspecific T cells need to be analyzed to inform vaccine protocols (Hafalla et al., 2013; Murphy et al., 2013; Billman et al., 2016).

Cryopreservation of sporozoites is so far the only way of efficiently preserving the infectivity of attenuated sporozoites, posing huge logistic constraints. Thus, development of a vaccine formulation, which is stable under cooled conditions or, ideally, at ambient temperature and induces long-term protection, remains a critical bioengineering milestone. Plasmodium sporozoites are particularly sensitive to environmental conditions, and it remains entirely speculative whether an appropriate preservation process can be implemented. A GAP vaccine formulation might be further improved by addition of an adjuvant; however, no examples exist yet for live attenuated Plasmodium vaccines.

### REFERENCES


### OUTLOOK

Building on the success of live attenuated, metabolically active RAS, genetic engineering of liver stage-arrested parasites offers unprecedented opportunities to develop a precision malaria vaccine. Comparative studies with GAPs that display distinct temporal arrests during liver stage maturation provide a foundation for systems immunology approaches, which might in turn lead to a better mechanistic understanding of immune effector mechanisms that contribute to lasting protection against re-infection. Ultimately, evidence-based design of safe and effective whole sporozoite P. falciparum and P. vivax vaccines involves major research investments in preclinical murine malaria models before translation to the human parasite is warranted. Proper design of human clinical trials with predictive power for vaccine safety, negligible adverse events, and vaccine efficacy in young children living in very diverse tropical countries remains very challenging.

### AUTHOR CONTRIBUTIONS

All authors contributed equally to this work and approved the manuscript for publication.

### ACKNOWLEDGMENTS

The authors thank Diane Schad for the artwork on **Figure 1**. Work on genetically arrested parasites in murine malaria models is funded by the Deutsche Forschungsgemeinschaft through the graduate program 2046 "From experimental models to natural systems" (project B1).


malaria parasite late liver stage development. Cell. Microbiol. 11, 506–520. doi: 10.1111/j.1462-5822.2008.01270.x


**Conflict of Interest Statement:** KMa is listed as inventor on international and national patents "Live genetically attenuated malaria vaccine" and "Live genetically engineered protozoan vaccine." These patents were filed by the authors' non-profit institutions to promote the development and distribution of malaria vaccines to people in need worldwide, in accordance with a global access strategy.

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

digital media

of impactful research

article's readership