# RECENT ADVANCES IN DROSOPHILA CELLULAR AND HUMORAL INNATE IMMUNITY

EDITED BY : Susanna Valanne, Dan Hultmark and Laura Vesala PUBLISHED IN : Frontiers in Immunology

#### Frontiers eBook Copyright Statement

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

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

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

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

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

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

ISSN 1664-8714 ISBN 978-2-88966-191-6 DOI 10.3389/978-2-88966-191-6

#### 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

# RECENT ADVANCES IN DROSOPHILA CELLULAR AND HUMORAL INNATE IMMUNITY

Topic Editors: Susanna Valanne, Tampere University, Finland Dan Hultmark, Umeå University, Sweden Laura Vesala, Tampere University, Finland

Citation: Valanne, S., Hultmark, D., Vesala, L., eds. (2020). Recent Advances in Drosophila Cellular and Humoral Innate Immunity. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88966-191-6

# Table of Contents

*05 Editorial: Recent Advances in* Drosophila *Cellular and Humoral Innate Immunity*

Laura Vesala, Dan Hultmark and Susanna Valanne

*08 Venom Atypical Extracellular Vesicles as Interspecies Vehicles of Virulence Factors Involved in Host Specificity: The Case of a* Drosophila *Parasitoid Wasp*

Bin Wan, Emilie Goguet, Marc Ravallec, Olivier Pierre, Séverine Lemauf, Anne-Nathalie Volkoff, Jean-Luc Gatti and Marylène Poirié

*22* Heterorhabditis bacteriophora *Excreted-Secreted Products Enable Infection by* Photorhabdus luminescens *Through Suppression of the Imd Pathway*

Eric Kenney, John M. Hawdon, Damien O'Halloran and Ioannis Eleftherianos

*36 Expression of Human Mutant Huntingtin Protein in* Drosophila *Hemocytes Impairs Immune Responses*

Yu-Hsien Lin, Houda Ouns Maaroufi, Emad Ibrahim, Lucie Kucerova and Michal Zurovec

*51 Emerging Mechanisms of Insulin-Mediated Antiviral Immunity in*  Drosophila melanogaster

Chasity E. Trammell and Alan G. Goodman

*58 Bombardier Enables Delivery of Short-Form Bomanins in the* Drosophila *Toll Response*

Samuel J. H. Lin, Amit Fulzele, Lianne B. Cohen, Eric J. Bennett and Steven A. Wasserman

*68 The Daisho Peptides Mediate* Drosophila *Defense Against a Subset of Filamentous Fungi*

Lianne B. Cohen, Scott A. Lindsay, Yangyang Xu, Samuel J. H. Lin and Steven A. Wasserman

*80 Host-Microbe-Pathogen Interactions: A Review of* Vibrio cholerae *Pathogenesis in* Drosophila

Saeideh Davoodi and Edan Foley


*117 Sexual Dimorphisms in Innate Immunity and Responses to Infection in*  Drosophila melanogaster Rebecca L. Belmonte, Mary-Kate Corbally, David F. Duneau and Jennifer C. Regan

*135 Viral Infection and Stress Affect Protein Levels of Dicer 2 and Argonaute 2 in* Drosophila melanogaster

Alessandro Torri, Vanesa Mongelli, Juan A. Mondotte and Maria-Carla Saleh

*147* Drosophila *as a Model System to Investigate the Effects of Mitochondrial Variation on Innate Immunity*

Tiina S. Salminen and Pedro F. Vale


Preethi P, Ajay Tomar, Sukanya Madhwal and Tina Mukherjee

# Editorial: Recent Advances in Drosophila Cellular and Humoral Innate Immunity

Laura Vesala<sup>1</sup> , Dan Hultmark <sup>2</sup> and Susanna Valanne1\*

<sup>1</sup> Faculty of Medicine and Health Technology, Tampere University, Tampere, Finland, <sup>2</sup> Department of Molecular Biology, Umeå University, Umeå, Sweden

Keywords: Drosophila melanogaster, innate immunity, cellular immunity, humoral immunity, model for human diseases, metabolism, bacteria, host-pathogen interaction

Editorial on the Research Topic

#### Recent Advances in Drosophila Cellular and Humoral Innate Immunity

We are pleased to present our Research Topic for Frontiers in Immunology, focusing on "Recent Advances in Drosophila Cellular and Humoral Innate Immunity". This collection of articles is comprised of primary research articles and reviews of the current literature by Drosophila immunity experts. Therefore, the papers feature progress in our understanding of the Drosophila melanogaster immune responses against a wide variety of pathogens, covering bacteria, viruses and parasites.

In many insects, bacterial infection induces the production of bactericidal and fungicidal peptides, such as cecropins, defensins, diptericins, and drosomycins, which accumulate in the hemolymph (1, 2). Besides these humoral effectors, Uttenweiler-Joseph et al. (3) identified several other Drosophila immune-induced molecules (DIMs or IMs) in the fly hemolymph. Their functions remained unknown until recently, when it was shown that twelve related IMs, the Bomanins (Boms), form a family that participates in the defense against some fungi and bacteria (4). Now, Lin et al. have characterized a gene called Bombardier (bbd), which regulates the defense mediated by the short-form Boms. Bombardier mutants are defective in their Toll pathway-dependent resistance against pathogens; especially candidacidal activity is completely dependent on Bombardier and the short-form Boms. Furthermore, Cohen et al. have identified a function for two other IMs, renamed Daisho1 and Daisho2, which are specifically involved in the defense against filamentous fungi.

Whereas the pathways regulating bacterial defenses are quite rigorously studied (5, 6), we still have limited knowledge about the antiviral immune defense in Drosophila. Best understood is the RNA interference (RNAi) system, which is believed to be the major defense against RNA viruses (7). This system recognizes double-stranded RNA molecules, cuts them into 21-nucleotide siRNAs, and uses them as templates to recognize and destroy viral RNA. Here, Torri et al. have investigated how two key components in this defense, Dicer-2 (Dcr-2) and Argonaute 2 (AGO2), are modulated in response to viral infection. Both factors were rapidly upregulated in virus-infected animals. Interestingly, only protein levels were affected, not mRNA, suggesting that these effects are posttranscriptionally controlled. Moreover, Trammell and Goodman have reviewed the emerging mechanisms of using Drosophila as a tool to study arthropod-borne viruses (arboviruses) that pose a great global health risk to humans. RNAi, JAK/STAT pathway and induction of STINGmediated immunity have been shown to be involved. The nutritional status of flies in the Drosophila model as well as of the vector mosquitoes appears to be an important factor in antiviral defense: while mosquitoes use an RNAi-dependent response during starvation, the bloodmeal provides the

Edited and reviewed by: Miki Nakao, Kyushu University, Japan

> \*Correspondence: Susanna Valanne susanna.valanne@tuni.fi

#### Specialty section:

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

Received: 25 August 2020 Accepted: 17 September 2020 Published: 02 October 2020

#### Citation:

Vesala L, Hultmark D and Valanne S (2020) Editorial: Recent Advances in Drosophila Cellular and Humoral Innate Immunity. Front. Immunol. 11:598618. doi: 10.3389/fimmu.2020.598618

**5**

insulin needed to activate a JAK/STAT-dependent response during infection. The authors conclude that Drosophila is an invaluable tool to study arboviruses, but because Drosophila are not the natural host for these arboviruses, there are also limitations.

Drosophila blood cells, the hemocytes, have traditionally been divided into three distinct classes; phagocytic plasmatocytes, melanizing crystal cells and lamellocytes, the larval immuneinduced hemocyte type. The currently ongoing burst of hemocyte research has revealed a more complex system, identifying intermediate stages and sub-classes, some of which are likely to have functional significance (8–13). Shim et al. contribute to the growing pool of hemocyte research with their findings of two distinct plasmatocyte populations, with different functions in metabolic regulation and maintaining organismal homeostasis. Iwashita et al. studied the function of Gyc76A, a guanylate cyclase type receptor producing cyclic guanosine monophosphate (cGMP), in immune responses. They show that Gyc76A is required for hemocyte proliferation during bacterial infection, by activation of a small GTPase Ras85Ddependently rather than cGMP-dependently. Gyc76A is also involved in the antimicrobial peptide response, but that effect is elicited via cGMP-dependent protein kinase DG2 and components of the Toll pathway. Thus, Gyc76A is involved in regulation of both, humoral and cellular immune responses via different routes.

Infections cause major metabolic adjustments in immune cells and at systemic level, balancing the needs of the immune system and the organismal homeostasis (14–16). Understanding these events has wide implications, and, not surprisingly, they have recently attracted growing attention. Parupalli et al. have now directly investigated how hemocytes affect nutrition and growth in Drosophila development. They show that ablation of hemocytes leads to reduced growth and insulin insensitivity, resulting in minute flies. A similar phenotype was seen when the larvae were fed a high sugar diet, but the effect could be rescued by genetically activating the hemocytes. Their experiments demonstrate the intimate crosstalk between hemocytes and fat body in controlling the distribution of nutritional resources.

Mitochondria are the central energy-producing organelles in the eukaryotic cells. In their review, Salminen and Vale address the role of mitochondrial variation in innate immunity. This variation may arise from both, the nuclear and the mitochondrial genomes, and through the mito-nuclear interactions. They discuss how mitochondrial variation affects the Krebs cycle metabolites, production of adenosine triphosphate and reactive oxygen species, and how these factors alter the infection outcomes. They also describe the benefits of utilizing Drosophila to study the nuclear variation affecting mitochondrial functions and how to construct cytoplasmic hybrid (cybrid) fly strains to study the mtDNA variation apart from variation arising from the nuclear genome.

Vincent et al. show that Francisella novida infection leads to metabolic dysregulation in control flies but not in imd pathway mutant flies. Their data indicate that in the Drosophila-F. novicida host-pathogen model, imd pathway activation is necessary but not sufficient for the metabolic pathology. Most likely both the host and bacterial factors contribute to the metabolic dysregulation, which is also seen with other infection models. Davoodi and Foley have reviewed the current knowledge of Vibrio cholerae pathogenesis in Drosophila. V. cholerae is a gram-negative pathogen causing pandemic outbreaks of cholera, a diarrheal disease. About three million cases are estimated to occur per year, resulting in 100,000 deaths. D. melanogaster has been used for modelling the V. cholerae pathogenesis since the foundational study in 2005, where it was shown that continuous feeding of V. cholerae to adult flies caused a cholera-like disease resulting in weight loss and rapid death. In the last 15 years, complex interactions between the pathogen, host defense, intestinal microbiome and metabolism have been uncovered.

In order to understand host-pathogen interactions and the underlying causes of infection outcomes, it is important to characterize molecules secreted by the pathogens. Nonaka et al. performed a structural and functional analysis of Monalysin, a pore-forming toxin from the entomopathogenic bacterium Pseudomonas entomophila. By purifying the endogenous Monalysin, they confirmed its action as a pore-forming toxin and suggest that it might preferably target cell membranes at curved sites, such as tips of filopodia. Their study adds to knowledge on detailed defense mechanisms against toxins. Kenney et al. set out to investigate the properties of Excreted-Secreted (ES) products from the nematode Heterorhabditis bacteriophora. Concentrated ES products were shown to inhibit the activity of Diptericin and to also enable the infection of Drosophila by Photorhabdus luminescens, symbiotic bacteria from the H. bacteriophora nematode. It will be interesting to characterize and identify the specific proteins in the secretions—these proteins bear potential for the treatment of diseases caused by excessive immune activation.

Parasitoid wasps, such as those belonging to the genus Leptopilina, lay their eggs inside Drosophila larvae. Along with the egg, they inject venom containing a variety of proteins and vesicles, thought to facilitate parasitoid growth (17). Wan et al. concentrated on these vesicles, which they named venosomes. They found that these extracellular vesicles function as a transport system to deliver venom factors into the lamellocytes, a hemocyte important in defense against parasitoids in D. melanogaster. This is sufficient to impair the lamellocyte function. The authors theorize that the venosome-cell interaction could be seen as a level of host-pathogen interactions.

As emphasized already above, D. melanogaster is a plastic model for studies from research on evolutionarily conserved basic cellular mechanisms to complex human diseases. Lin et al. have investigated the pathogenic effect of the mutant Huntingtin protein (mHTT), that causes the Huntington disease in humans, in the Drosophila model. It was demonstrated that overexpression of mHTT in the nervous system causes lethality, whereas overexpressing mHTT in an immune-reactive tissue, i.e. hemocytes, did not cause direct lethality but immune dysregulation. Future studies will be needed to clarify the molecular interaction between mHTT and the main Drosophila immune signaling pathways as well as suppression of phagocytosis. The role of immunity and inflammation in aging and age-related diseases, i.e. "inflammaging", is another pressing question, as the mean population age increases (18). Arora and Ligoxygakis have reviewed the current understanding of the role of immunity in age-dependent neurodegeneration in Drosophila. Is inflammation the cause or consequence of neurodegeneration? Predisposition to an overactive immunity, e.g. by loss of negative regulation of the Imd pathway or loss of autophagy, causes neurodegeneration. On the other hand, immunity can also be neuroprotective, as Draper, the glial engulfment receptor, is important in preventing the accumulation of amyloid structures found in Alzheimer's disease -associated phenotypes in Drosophila.

Lastly, Belmonte et al. discuss in their review the profound differences between sexes in response to infection. As in humans (19), D. melanogaster shows strong sexual differences in immune responses at baseline, upon pathogen challenge and in aging. In their extensive literature search, they found that fewer than 10% of immunological studies on Drosophila have only used one sex or have not separated their subjects by sex. What can Drosophila teach us about immune dimorphism? Including both sexes in functional and mechanistic studies of Drosophila immunity gives a more complete picture on the complexity of immune mechanisms. It also increases our understanding of the

#### REFERENCES


fundamental mechanisms underlying innate immunity mechanisms and responses to infection.

The field of Drosophila immunity has broadened considerably in recent years and it is now in a very productive and exciting phase. This is nicely illustrated in this collection of articles. We would like to thank all the authors for their contributions to this Research Topic as well as the reviewers for their time and input.

#### AUTHOR CONTRIBUTIONS

All authors contributed to the article and approved the submitted version.

#### FUNDING

The Drosophila work in Tampere University was carried out in the Tampere Drosophila Facility, partly funded by Biocenter Finland. DH was supported by the Swedish Research Council, the Swedish Cancer Society, the Academy of Finland and the Sigrid Juselius Foundation.


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

Copyright © 2020 Vesala, Hultmark and Valanne. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

## Venom Atypical Extracellular Vesicles as Interspecies Vehicles of Virulence Factors Involved in Host Specificity: The Case of a *Drosophila* Parasitoid Wasp

Bin Wan1†, Emilie Goguet 1†, Marc Ravallec<sup>2</sup> , Olivier Pierre<sup>1</sup> , Séverine Lemauf <sup>1</sup> , Anne-Nathalie Volkoff <sup>2</sup> , Jean-Luc Gatti 1‡ and Marylène Poirié<sup>1</sup> \* ‡

<sup>1</sup> Université Côte d'Azur, INRA, CNRS, ISA, Sophia Antipolis, France, <sup>2</sup> INRA, Univ. Montpellier, UMR 1333 "Microorganism and Insect Diversity, Genomes and Interactions" (DGIMI), Montpellier, France

Endoparasitoid wasps, which lay eggs inside the bodies of other insects, use various strategies to protect their offspring from the host immune response. The hymenopteran species of the genus Leptopilina, parasites of Drosophila, rely on the injection of a venom which contains proteins and peculiar vesicles (hereafter venosomes). We show here that the injection of purified L. boulardi venosomes is sufficient to impair the function of the Drosophila melanogaster lamellocytes, a hemocyte type specialized in the defense against wasp eggs, and thus the parasitic success of the wasp. These venosomes seem to have a unique extracellular biogenesis in the wasp venom apparatus where they acquire specific secreted proteins/virulence factors and act as a transport system to deliver these compounds into host lamellocytes. The level of venosomes entry into lamellocytes of different Drosophila species was correlated with the rate of parasitism success of the wasp, suggesting that this venosome-cell interaction may represent a new evolutionary level of host-parasitoid specificity.

Keywords: *Drosophila*, immunity, parasitoid wasp, *Leptopilina*, venosomes, lamellocyte, virulence

#### INTRODUCTION

It is now well-established that cells can communicate at "long distance" using diverse types of membranous particles, called extracellular vesicles (EVs). The release or secretion of EVs is a universal cellular mechanism, shared by archaea, bacteria, and eukaryotes, which probably existed in their last common ancestor (1, 2). Since most organisms interact with each other, one species can also transfer EVs to a different species and thus act on it via that specific interaction. Accordingly, recent data from bacteria, intracellular parasites, and nematodes, have shown that EVs secreted by pathogens participate in the communication and disease transmission between non-infected and infected host cells (3, 4). EVs may play a key role in altering the function of targeted cells during infection, through the spreading of pathogens or the transport of virulence factors (5, 6).

The lifestyle of the hymenopteran endoparasitoid wasps is between those of parasites and predators: they lay eggs in or on the body of other insects, their larvae develop by consuming living host tissues, resulting in the host death (7). To ensure the successful development of eggs

*Edited by:*

Laura Vesala, University of Tampere, Finland

#### *Reviewed by:*

Jianhua Huang, Zhejiang University, China Maurizio Francesco Brivio, University of Insubria, Italy

#### *\*Correspondence:*

Marylène Poirié marylene.poirie@univ-cotedazur.fr

#### *†Present Address:* Bin Wan,

State Key Laboratory of Rice Biology and Ministry of Agricultural and Rural Affairs, Key Laboratory of Molecular Biology of Crop Pathogens and Insects, Institute of Insect Sciences, Zhejiang University, Hangzhou, China Emilie Goguet, Henry M. Jackson Foundation at the Uniformed Services University of the Health Sciences, Department of Microbiology and Immunology, Bethesda, MD, United States

‡These authors are co-last authors

#### *Specialty section:*

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

> *Received:* 22 May 2019 *Accepted:* 04 July 2019 *Published:* 17 July 2019

#### *Citation:*

Wan B, Goguet E, Ravallec M, Pierre O, Lemauf S, Volkoff A-N, Gatti J-L and Poirié M (2019) Venom Atypical Extracellular Vesicles as Interspecies Vehicles of Virulence Factors Involved in Host Specificity: The Case of a Drosophila Parasitoid Wasp. Front. Immunol. 10:1688. doi: 10.3389/fimmu.2019.01688

**8**

and larvae, parasitoids have developed various mechanisms allowing to bypass the host immune response and regulate its physiology (7, 8). They have notably evolved astonishing strategies for delivering virulence factors into host cells. The production of EVs in parasitoid wasps has been known for a long time and data have revealed their various nature and origin (9–12). Some species use viral genes stably integrated in the wasp genome to build specific vesicles [named polynaviruses (PDVs) or virus-like-particles (VLPs) (9–11)]. These EVs are produced and secreted by the cells of the wasp ovarian calyx, and carry either wasp DNA (PDVs) or proteins (VLPs). Others use a more peculiar means, they produce virulence factors in the venom gland and package them in vesicles stored in the reservoir (both parts of the reproductive tract) (9, 12, 13). In all cases, these parasitoids inject EVs along with the egg during oviposition, which participate in the reproductive success of the wasp. While many parasitoid wasp species produce ovarian EVs, very few described species produce EVs in the venom apparatus (9). Among them are the wasps of the Figitidae family, including the genera Ganaspis and Leptopilina that parasitize Drosophila melanogaster (Diptera) and other closely related species. The venomous EVs of these species differ in shape, size, and structure (12–19), and former publications called them Virus Like Particles (VLPs) because the mature vesicles somehow resembled viruses, particularly those of L. heterotoma that showed spikes extending from a round/ovoid vesicle (9, 18). Recently, a proteomic study of L. heterotoma VLPs showed that they contain many different proteins. Based on some homology they observed between certain of these proteins and bacterial proteins, the authors proposed to rename the VLPs, mixed strategy extracellular vesicles (MSEV) (19). However, since none of the studies to date have provided indisputable evidence of a viral or bacterial origin of Leptopilina VLPs, we will thereafter refer to these venom vesicles with the neutral term "venosomes." This will also avoid any confusion with the vesicles produced in the ovaries of some wasps that truly derive from a viral machinery, also named VLPs (9).

The main defense of insects against parasitic eggs and larvae, called encapsulation, involves the formation of hemocyte layers around the foreign body as well as the production, via the activation of the phenoloxidase (PO) cascade, of melanin and cytotoxic radicals presumed to kill the parasite (20). This waspinduced reaction is largely used as a model to study the innate immune response and stimulate hematopoiesis in Drosophila hosts (21). Indeed, in D. melanogaster larva, the process of encapsulation requires the production of different types of hemocyte, including the lamellocytes that are encapsulationspecific cells induced only in case of parasitism (8, 20, 21). Lamellocytes have been identified as targets of L. heterotoma purified venosomes. Mixed in vitro with these cells, they induce their lysis, a phenomenon also observed after parasitism (12, 22) but whose precise mechanism of action remains unknown.

The parasitism success of different Leptopilina boulardi strains from different laboratories differs according to the Drosophila species or strain. For instance, two strains with different phenotypes have been characterized, ISm (strain G431, Immune Suppression for D. melanogaster), highly virulent against D. melanogaster, and ISy (strain G486, Immune Suppression for D. yakuba) whose success depends on the resistant/susceptible genotype of the host (23, 24). These strains differ in the relative abundance of their major venom components, with few common abundant proteins and significant quantitative differences for most of them (25). They apparently differ as well in the venosomes present in their reservoir (13). Moreover, one of the proteins in the venom of ISm that seems important for parasitism success is a Rho GTPase activating protein (RhoGAP) named LbGAP (26, 27). LbGAP belongs to a family of genes specifically expressed in the venom glands of both ISm and ISy. ISm LbGAP has been immunolocalized in the lamellocytes of the host after parasitism and the number of LbGAP spots is correlated with the degree of alteration of the lamellocyte morphology (passage from round flat to bipolar), a change supposed to affect their ability to encapsulate the parasitoid egg (28). However, the mechanisms of entry and action of LbGAP into lamellocytes remain unknown. Although the ISy strain venom also contains Rho GAPs, its success mainly relies on a serine protease inhibitor (a serpin called LbSPNy) that inhibits the PO cascade activation in the hemolymph of Drosophila yakuba larva, one of its main hosts (29). The LbSPNm allelic serpin that targets different types of proteases from those targeted by LbSPNy is also abundant in the venom of ISm (25).

Based on these previous data, our goal was to clarify the connection between the venom factors described and the venosomes in L. boulardi wasps and to study the potential link between these extracellular vesicles and the transport of such factors into host lamellocytes. We demonstrate here that venosomes may have an atypical extracellular biogenesis in the venom gland and that purified venosomes from the venom reservoir are sufficient to mimic the protective effect of the whole venom for the wasp egg. Using fluorescently labeled purified venosomes and co-immunolocalization, we showed that they target the D. melanogaster lamellocytes and serve as a specific transport system to deliver venom factors both to the circulating and sessile hemocytes/lamellocytes. Finally, we observed that the level of venosomes entering into lamellocytes is correlated with the success of wasp parasitism on the Drosophila species tested, suggesting a key role in the specificity of the host-parasitoid interaction. The parasitoid wasp L. boulardi thus appears as an interesting model to study the evolution of extracellular vesicle formation mechanisms and the role played by the vesicular transport in interspecies communication. This new cellular level of parasitoid specificity may also be of interest for understanding the adaptive mechanisms between hosts and parasites.

#### MATERIALS AND METHODS

#### Biological Material

The L. boulardi strains ISy from Brazzaville (Congo) (Gif stock G486) and ISm from Nasrallah (Tunisia) (Gif stock G431), and the L. heterotoma strain from Gotheron (France) have been previously described (23–25). All parasitoids were reared at 25◦C on the susceptible D. melanogaster strain Nasrallah (Gif stock 1333). After emergence, adults were kept at 20◦C on agar medium with honey. All experiments were performed with naïve, 5–10 days-old mated females.

The D. melanogaster strain 1088, named YR for resistant to the L. boulardi ISy parasitism, comes from an original selection of isofemale lines obtained from a population of Brazzaville (Congo), combined with subsequent genetic approaches (30– 32). D. melanogaster hopTum−<sup>l</sup> (stock 8492) was obtained from the Bloomington Drosophila stock center. The D. suzukii strain, kindly provided by Dr. R. Allemand (LBBE, University Lyon 1, France), originates from a population collected in Sainte-Foy-lès-Lyon (Rhône, France). The D. yakuba R strain (stock number 307-14) and the D. simulans Japanese strain were kindly provided by D. Joly (EGCE, Gif-sur-Yvette, France) and M.T. Kimura (Hokkaido University, Japan), respectively. Based on previous laboratory results (see also Discussion), the L. boulardi ISm strain is highly successful on D. melanogaster YR and hopTum−<sup>l</sup> strains (>90%), less on D. simulans (>60% success of parasitism), and it consistently fails on D. yakuba and D. suzukii (<10 and 0% success of parasitism, respectively). All Drosophila were reared on a standard medium (10% cornmeal, 10% yeast, agar, and Nipagine) at 25◦C. Conditions for all insects were a 12/12 h light/dark and 50% humidity.

#### Antibodies

We used "in house" polyclonal rabbit antibodies directed against full-length recombinant proteins for LbGAP (25) and LbGAP2 (33), and synthetic peptides for LbSPN (34) and Atilla, a D. melanogaster marker of lamellocytes (35). The anti-LbGAP, anti-LbGAP2, and anti-LbSPN antibodies recognize the ISm and ISy proteins equally. For Western blots, we used them at dilutions 1/10,000 (LbGAP), 1/5,000 (LbGAP2), 1/10,000 (LbSPN), and 1/1,000 (Atilla). LbGAP, LbGAP2, and LbSPN antibodies were all used at a 1/500 dilution for histoimmunochemistry. Secondary antibodies were goat antirabbit horseradish peroxidase conjugated (1/10,000, Sigma) for Western blot, and fluorescently labeled goat anti-rabbit IgG (Fluoprobes 594, Interchim; 1/200, and 1/2,000 as indicated) for immunohistochemistry.

### Venom Recovery and Purification of Venosomes

The wasp venom apparatus was obtained by traction on the female ovipositor, the reservoir was separated from the glands and dilacerated with tweezers in 20–50 µl drop (1 reservoir per µl; number depending upon the experiment) of Insect Ringer Solution (IR; KCl, 182 mM; NaCl, 46 mM; CaCl2, 3 mM; Tris-HCl, 10 mM) supplemented with a protease inhibitor cocktail (IR-PI) (Sigma). This venom total extract was then centrifuged for 5 min at 500 g to remove residual tissues and obtain the "crude" venom. Crude venom was centrifuged at 15,000 g (15 min; 4◦C) to pellet the vesicular material, which was washed twice with IR-PI and used for either SDS-PAGE, labeling, or electron microscopy. The supernatant was the soluble venom proteins fraction. The protein profile of the venosomes obtained by direct centrifugation was compared with that obtained by ultracentrifugation (35,000 g, 1 h; MLA50 rotor, Beckman) on a 10–50% Nycodenz gradient as previously described (12).

#### SDS-PAGE and Western Blotting

Samples were run under denaturing and reducing (5% ßmercaptoethanol) conditions on 12.5% polyacrylamide gels. The gels were either stained silver or transferred to a nitrocellulose membrane (Millipore). The membrane was blocked with TBS-Tween, 2% low fat milk, incubated overnight at 4◦C with the indicated antibody, and washed and incubated with the goat anti-rabbit HRP secondary antibody for 2 h at RT. After washing, a signal was detected with a chemiluminescent substrate (Luminata Western, Millipore) with a digital camera. None of the preimmune sera or secondary antibody alone produced a significant signal.

#### Electron Microscopy

15,000 g venosome pellets were processed for transmission electron microscopy as described in Labrosse et al. (14). Briefly, samples were fixed in sodium cacodylate (0.1M, pH 7.2 for venosomes) or PBS (for the venom apparatus) with 5% glutaraldehyde for 24 h at 4◦C. Post-fixation was done with 2% osmium tetroxide in the same buffers, followed by dehydration in a graded ethanol series prior to inclusion in Epon and ultrafine section. The sections were contrasted with uranyl acetate and lead citrate before observation (Jeol 1010 and Zeiss EM10CR, 80 kV).

For immunogold labeling, 24 venom apparatus (gland and reservoir) from ISm and ISy females were fixed for 3 h in 4% paraformaldehyde −0.1 M phosphate at 4◦C. Samples were included in agarose blocks (8 apparatus per inclusion; three blocks for each strain). The blocks were washed four times in 0.1 M phosphate buffer at RT and dehydrated in ethanol (50– 100% at −19◦C), before inclusion in the resin (London Resin White; TAAB Lab Equipment; −25◦C for 48 h under UV). Ultrafine sections (70 nm) were mounted on grids treated with colloidal gold, washed in PBS, and blocked in PBS-1% BSA before incubation with the LbGAP antibody (1/1,000; 1 h 30 at 20◦C) (this last step was omitted for controls). After several washes in PBS-1% BSA and then in PBS-0.1% BSA, all the grids were incubated with a goat anti-rabbit antibody conjugated to 15 nm gold particles (1/30; Biocell Research Lab.). After washing, sections were stained with 1% uranyl acetate and examined under an electron microscope (JEOL 1010 at 80 kV). All observations (controls and processed) were performed on successive sections of a block for each wasp strain. Sections from several blocks were observed and photograph.

#### Microinjection

The venosome pellet obtained from 20 reservoirs was resuspended in 20 µl of IR and labeled with 1 mM of Fluoprobes 488-NHS ester (1 mg/ml; λexc./λem.: 593/519 nm; Interchim) for 1 h at 4◦C. The labeled vesicles were centrifuged at 15,000 g (10 min at 4◦C) and washed once with IR-3% BSA to quench the free ester, and again with IR. The last pellet was resuspended in IR and used to micro-inject 50 second-instar host larvae (FemtoJet, Eppendorf). Hemolymph was collected 14–18h after microinjection.

To test the phagocytic properties of the hemocytes in vivo, fluorescent latex beads (carboxylate-modified polystyrene latex beads 2µm in diameter, fluorescent red; Sigma) or fluorescent live Escherichia coli (5 × 10<sup>9</sup> bacteria/ml; E. coli DH5 alpha expressing the green fluorescent protein) were microinjected in Drosophila L2 larvae 14–18 h after parasitism (to stimulate the hemocytes production). The observations were made 4 h after the injection of beads or bacteria. The experiments were repeated 3 times for each species tested, on different days.

#### Parasitism Assay

For immunohistochemistry of hemocytes, batches of 30 secondinstar larvae (L2) were collected and transferred on a dish with Drosophila medium to be subjected to parasitism by three female wasps for 4 h. The parasitoids were then removed, and the larvae were left at 25◦C until use (see below). For the parasitism experiments, 30 second-instar host larvae (L2) were parasitized for 2 h by one ISm or ISy parasitoid female. The encapsulation capacity was estimated 48 h later by counting the number of encapsulated eggs after dissection of the late third-instar larvae. Virulence was expressed as the ratio of the number of non-encapsulated parasitoid eggs to that of the monoparasitized hosts. To analyze the immunosuppressive role of the different fractions of ISm venom (crude venom, 15,000 g venom pellet, 15,000 g venom supernatant), larvae of D. melanogaster YR were injected with 20 nl of each of these fractions using a Nanoject II injector (Drummond Scientific). The samples were

obtained from 20 female reservoirs diluted in 20 µl of IR, this volume being kept constant so that a 1/50th equivalent of a reservoir was injected for each fraction. A total of 150 larvae (5 independent repeats of 30 larvae) were injected for each sample, then parasitized by ISy female parasitoids, and encapsulation occurred an estimated 48 h later on surviving larvae, as described above. The controls were injected with IR alone.

#### Hemolymph and Hemocyte Collection

Drosophila larvae were removed from the food and washed carefully three times in PBS. Their hemolymph was then collected directly in a drop of 35 µl of PBS (SDS-PAGE) or Grace's medium (Immunochemistry) by gently tearing the cuticle on the anterior part of the larvae with fine tweezers. For SDS-PAGE, the cells were separated by centrifugation (500 g, 10 min) and the supernatant was collected (cleared hemolymph). The cells pellet was washed twice in PBS and the last pellet and the cleared hemolymph were diluted in reducing sample buffer and boiled for SDS-PAGE. At least three separate experiments were done.

## Tissue Immunohistochemistry and Confocal Observation

#### Hemocyte Immunohistochemistry

The hemolymph of Drosophila larvae was collected at various time points (as indicated) after parasitism or micro-injection with labeled venosomes. At each time point, the hemolymph was collected from 10 larvae as described above, and the hemolymph solution was transferred to the center of a coverslip placed in a 12-well culture plate to form a wet chamber. The cells were allowed to adhere for 1 h before a 15 min fixation step with 4% formaldehyde in PBS. The cells were then washed three times with PBS, permeated with PBS-T (PBS + 0.1% Triton 100X) for 15 min and blocked with PBS 0.3% BSA for 30 min. The cells were incubated for 1 h with a primary antibody at RT, washed three times with PBS and incubated for 1 h with the secondary antibody. After three washes with PBS and one with deionized water, the coverslip was mounted on a slide using an antifading medium containing DAPI (Interchim). Data were obtained from at least three separate experiments.

#### Long Gland Immunohistochemistry

Complete venom apparatuses were fixed in 4% formaldehyde 0.1% Triton 100X for 3 h. After that, they were washed three times with PBS and blocked with PBS 0.3% BSA for 30 min. The primary antibody was added and incubated for 1 h, washed three times 15 min with PBS, and incubated with the secondary antibody for 1 h at RT. After three PBS washes, venom apparatuses were mounted between a slide and a coverslip, as described above. 5 apparatuses were treated per experiment and the experiment was repeated at least four times.

FIGURE 2 | Immunolocalization of venom virulence factors along the venom gland of ISm and ISy females. Confocal immunolocalization of LbGAP (A,B), LbGAP2 (C,D), and LbSPN (E,F) in the ISm (A,C,E) and ISy (B,D,F) venom gland. (A–F) Show merged images obtained from the entire gland for the indicated virulence factors (red fluorescence) and counter-stained for actin with green phalloidin. All panels numbered 1 and 2 are enlargement of the indicated gland region (white boxes) for the virulence factors (red) and actin (green), respectively. Labeling of the three virulence factors was observed in the cytoplasm of the peripheral gland secretory cells and associated with the canal of these cells whose microvilli actin was strongly labeled by phalloidin. For both strains, the absence of labeling in the gland collecting canal is either due to the loss of the contents during the staining process or to a masking due to the presence of cuticular layers that may block the light.

For both types of preparation, the actin was stained with green phalloidin (7 nM; green fluorescent phalloidin 490, Interchim) during the permeabilization step. Controls were done with the pre-immune sera from the rabbit used for immunization and with the secondary antibody alone. The highest background signal obtained was used to establish the level above which a labeling was considered positive. The samples were observed and imaged with an AxioImager Z1 equipped with an Apotome 2 or with a LSM 880 laser scanning confocal microscope (Zeiss).

### Clarification and Observation of *Drosophila* Larvae

YR larvae were fixed overnight, 15 h after injection of fluorescently labeled venosomes, with 4% PFA in PBS at 4◦C, in the dark. After washing with PBS, the samples were transferred

venosomes. Immunogold-labeled TEM sections (see drawing at the top for the position of the sections) through the intracellular canal of a long gland secreting cell (A), the gland canal (B), and the reservoir (C,D). LbGAP immunogold labeling (small black dots) is visible on the amorphic material present in the lumen of secretory cell canal, around in the cytoplasm of the cell and in some intracellular vesicles (A). No secreted membrane vesicles were observed. In the lumen of the gland collecting canal (B), most of the labeling was associated to electron-dense aggregated and fibrous material. No clear fully formed vesicles were visible. In the reservoir (C), gold labeling was almost entirely associated with the mature punctuated venosomes but not with the larger electron dense particles also present. No labeling was observed with the gold secondary antibody alone (D). Representative sections are shown here. Bar 0.2µm.

to a glass vial for a clearing procedure based on the 3DISCO method (36). Briefly, the samples were dehydrated overnight in 50% Tetrahydrofurane (THF) (Sigma) in the dark while being gently shaken, then for 3 h in 70% THF, overnight at 80% THF and for 30 min at 100% THF, repeated 3 times, followed by final incubation in dichloromethane (Sigma) for 45 min. To ensure optimum transparency, samples were finally impregnated overnight with a dibenzylether (DBE) solution (Sigma). To avoid tissue damage or shrinkage, they were mounted in a Lab-Tek II (Lab-Tek) chamber filled with the DBE solution. As DBE degrades fluorescence over time, the visualization was made as soon as the appropriate transparency level was reached. To obtain high resolution scans, samples were imaged by confocal microscopy. 3–5 larvae were treated in three different batches.

#### Statistical Analysis

Statistics were done using R (https://www.r-project.org). For the number of lamellocyte cells labeled with LbGAP in the different species, we used a binomial GLMM [package lme4 (37)] with the species as a fixed effect and the repetition as a random effect. Since there was no overdispersal in the model, the fixed effect was tested with an LRT-test, followed by a Tukey post-hoc test [package multcomp (38)]. For the number of spots in the lamellocytes, we first used a zero truncated Poisson GLM [VGAM package (39)] with the species as a fixed effect. However, since there was a high dispersal in this model, we instead fitted a linear model to the Box-cox transformed number of lamellocyte cells, followed by a Tukey post-hoc test.

FIGURE 4 | Biogenesis of the membrane of extracellular venosomes. (A) TEM observation of a section through the junction between the connecting duct (CD) and the reservoir (R) and an enlargement of the boxed zone (B) show stacks of membrane sticking to the lumen side of the cell layer (Cl) located at the junction. Some "rolling" empty membranes detach from the stack and form multimembrane vesicular structures (arrows) incorporating material present in the duct lumen. Punctuated material could also be observed in formed venosomes (in A, white asterisk) that are accumulated in the reservoir (R).

### RESULTS

#### LbGAP and LbGAP2 Association With *L. boulardi* Venosomes

The venom apparatus of L. boulardi consists of an elongated venom gland (classically named "long gland") with a central collecting canal connected by a thin duct (connecting duct) to a large reservoir in which venom accumulates (see detailed histology of the venom apparatus in **Figure S1**). To determine whether the LbGAP, LbGAP2 and LbSPN L. boulardi venom proteins potentially involved in virulence were associated with venosomes, we first purified the venosomes from crude venom obtained from the ISm reservoir using the previously described method of Nycodenz gradient ultracentrifugation (12) (**Figure 1A**). SDS-PAGE of the protein fractions showed that only a specific subset of the venom proteins co-sedimented at the expected position of venosomes in the gradient (fractions 4, about 25–30% Nicodenz). Western blot analysis revealed that LbGAP and LbGAP2, but not LbSPN, were associated with these venosome fractions. In order to confirm this, as well as to simplify and down-size the purification procedure, the crude ISm and ISy venom was directly centrifuged at 15,000 g. The protein profile of the washed pellet contained only a subset of the total proteins, very similar to that of Nycodenz-purified ISm venosomes fractions. Western blot analysis showed that LbGAP and LbGAP2 were enriched in the pellet but not LbSPN, which remained in the supernatant along with many other proteins (**Figure 1B**). Electron microscopy analysis of the pellet showed two main types of vesicles: large round aggregates/vesicles with a size of about 1µm (asterisk marks; **Figures 1C,D**) and, in the range of 100–300 nm, "typical" venosomes (arrows in enlargements in **Figures 1C,D**) and vesicles with different shapes that were less numerous and structured in ISy than ISm (**Figures 1C,D**). Because the shape of most of these purified vesicles at 100– 300 nm appeared to differ from that of the previously "typical" venosomes described in the venom reservoir of L. boulardistrains (13, 14), we verified that the procedure did not extensively degrade the venosomes. Since L. heterotoma venosomes, in contrast to L. boulardi ones, have a specific stellate shape due to spike extensions (12, 17), we treated the crude venom of this species in the same way and analyzed the obtained pellet by microscopy (**Figure S2**). Part of the L. heterotoma pelleted vesicles clearly retained the stellate shape, while others resembled more L. boulardi ISm and ISy vesicles, with the electron dense material accumulated as an asymmetric crescent or distributed all along the membrane [**Figure S2**; see also Figure 1E in (12)]. This suggests that the treatment may have slightly altered the shape of the venosomes but that the distinctive shapes between the different Leptopilina species and strains remained visible.

Since ISm and ISy venosomes were associated with a subset of venom proteins such as LbGAP and LbGAP2 but not with LbSPN, we wondered whether the synthesis and secretion of these factors (and thus the biogenesis of venosomes) could occur sequentially at different or specialized locations along the venom gland. LbGAP, LbGAP2, and LbSPN antibodies (**Figure 2**) labeled the whole ISm (**Figures 2A,C,E**, respectively) and ISy (**Figures 2B,D,F**, respectively) gland (except the "nose" region), although the signal was much weaker for LbGAP in ISy (**Figure 2B**), in agreement with a lower synthesis of this protein in this strain (26, 27) (see also **Figure 1**). The factors were clearly immunolocalized in the secretory cell cytoplasm and in the lumen of these cells' channels (as shown in the respective Figures enlargements) that can be visualized by the phalloidin labeling of actin-rich microvilli as described previously in L. heterotoma (17) (see also **Figure S3** for a detailed example of co-localization for LbGAP2 in the ISm cell and secretory cell channel). Therefore, it seems that there is no difference in secretion along the gland between the factors associated and not associated with venosomes. The venosomes of the Figitidae wasps were previously described as having a peculiar assembly in the lumen of the venom gland (15–18), and we therefore studied further how LbGAP could associate with the ISm venosomes by electron microscopy. For this, ultra-thin sections of the ISm venom apparatus were prepared for immunogold electron microscopy using the LbGAP antibody (**Figure 3**). **Figure 3A** show a cross-section through the "canal" of the secretory cells as indicated by the microvilli presence. The canal contains large electron dense particles, without precise form, which were immunogold labeled for LbGAP (**Figure 3A**). In the collecting canal of the gland (**Figure 3B**), large electron-dense particles and long filaments were dispersed in a fibrous/filamentous material. LbGAP was almost exclusively associated with the electron dense particulate materials suggesting proteins aggregation. A cross-section through the ISm reservoir showed venosomes as membranous punctuated-type vesicles (**Figures 3C,D**) and larger uniform electron-dense particles (from 400 nm to 1µm), with 87% of the LbGAP labeling associated with the venosomes (333 gold beads associated on 383 counted on three different TEM sections) (**Figure 3C**). One of the control cross-sections that passed at the junction between the connecting duct and the reservoir caught our attention (**Figure 4**). At this point, we observed a stack of membranes coming from the cell lining the duct, some of them seeming to roll and detach to form multiple membrane vesicles.

Thus, L. boulardi ISm venosomes seem to be built extracellularly from secreted proteins aggregates that modify their form and size along the collecting duct. Since none of the particulate aggregates observed in the ISm gland canal resembled the "mature" membraned venosomes observed in the reservoir (see **Figure 3**), we hypothesized that a specific part of the lumen material could be embedded by the membranes delaminating in the collecting duct near the entry of the reservoir to form the final mature venosomes.

#### Venosomes Are Sufficient to Protect Susceptible Wasp Eggs From Encapsulation

Since L. boulardi ISm venosomes contain putative virulence factors, we tested whether these vesicles were sufficient to replicate the protective effect of venom against encapsulation of eggs of the susceptible ISy strain (14, 23). For this, we used the previously set up assay (14, 24) based on the two different characterized L. boulardi strains, ISm, and ISy, and the resistant



D. melanogaster strain (YR). ISm eggs are never encapsulated in D. melanogaster YR larvae, whereas those of ISy almost always are (23, 24) (see **Table 1**). D. melanogaster YR larvae injected with extracts of ISm venom apparatus or crude venom and then parasitized with ISy did not encapsulate the parasitoid eggs (14, 24) (**Table 1**). This protective effect of the crude venom against ISy eggs encapsulation was also conferred by injection of the ISm washed 15,000 g pellet containing the venosomes (**Table 1**, mean % of 5 replicates; comparison to injection of ISm crude venom, χ <sup>2</sup> = 0.91; p < 0.05). In contrast, the injection of the supernatant 15,000 g of the ISm venom had no effect (comparison to injection of Ringer, χ <sup>2</sup> = 0.711; p < 0.05).

#### Venosomes Transport Potential Virulence Factors in Permissive Host Lamellocytes

LbGAP was immunolocalized in lamellocytes after parasitism (27, 28). Also, its physical association with the venosomes led us to postulate that venosomes are responsible for the transport of this protein and other associated proteins in immune cells of the host. To demonstrate this definitively, we immunolocalized LbGAP and LbGAP2 (both associated with venosomes) by confocal microscopy in the hemocytes of D. melanogaster YR larvae after parasitism (in vivo situation). We also tested by confocal microscopy the coimmunolocalization of LbGAP or LbGAP2 and venosomes in lamellocytes after microinjection of in vitro fluorescently-labeled purified venosomes.

Between 0–4 h after the end of ISm parasitism (time laps due to uncertainty since parasitism can occur at any time during the 4 h of contact between wasps and larvae; see mat. and meth.), LbGAP and LbGAP2 mostly immunolocalized in phagocytic plasmatocytes (see also below) since, as expected, there were still no lamellocytes at this time (40). This absence of lamellocytes was also demonstrated by the absence of reaction of this cell type marker Atilla on western-blots of hemocytes proteins (see **Figure S4**). 14–18 h post-parasitism, both LbGAP and LbGAP2 were still observed in plasmatocytes (not shown) and accumulated in the circulating host lamellocytes (**Figures 5A,B**), with the shape of most of these labeled lamellocytes visibly modified from round flat to elongated as previously observed (27, 28) (see unchanged lamellocytes in **Figure 6**). Thus, both proteins associated with venosomes were retrieved in modified lamellocytes after parasitism.

Microinjection of second-instar YR larvae with purified fluorescently-labeled ISm venosomes, induced the differentiation of lamellocytes, like parasitism (**Figure S4**). 18 h post-injection, fluorescent venosomes were observed in plasmatocytes (not shown) and co-immunolocalized with LbGAP or LbGAP2 in these newly formed lamellocytes (**Figures 5C,D**, respectively). This strongly suggests the entry of whole venosomes into the lamellocytes since the chosen NHS-fluorescent dye should mainly label membrane proteins of the venosomes. Interestingly, not all LbGAP and LbGAP2 spots co-localized with labeled venosomes in the lamellocytes (**Figure 5E**), suggesting that a release of these proteins from the membrane envelope occurred after the venosomes enter the cell (this also told us that these proteins were not labeled by the NHS-dye). Moreover, using hemolymph collected at different time points after parasitism (0 h being the end of the parasitism assay) and Western blot analyses, we showed that LbGAP and LbGAP2 (and thus certainly venosomes) still circulate in the cell-free hemolymph of host larvae 20 h post-parasitism by ISm or after microinjection (**Figure S4**). Thus, a continuous entry of venosomes from the hemolymph into the lamellocytes, followed by the release of LbGAP and LbGAP2 in the cell, may explain our observations.

Since parasitoid wasp infestation of D. melanogaster induces the release into the circulation of a subcuticular population of sessile hemocytes that will differentiate into lamellocytes (40), we analyzed whether venosomes could also target the hemocytes within this compartment. Eighteen hours after the injection of labeled venosomes, Drosophila YR larvae were fixed, chemically treated for tissue clearing (36, 41) and observed by confocal microscopy (**Figure S5**). Fluorescent spots were clustered in the subcuticular regions where sessile hemocytes localize (40). At higher magnification, this labeling co-localized with small (about 5–10µm) and large flat cells (>30µm), resembling plasmatocytes and lamellocytes, respectively.

#### Entry of Venosomes Seems "Specific" to Lamellocytes of Some Species of *Drosophila*

Leptopilina boulardi ISm is considered a specialized parasitoid that succeeds mainly on D. melanogaster and D. simulans (23, 24). We therefore tested whether the entry of venosomes into lamellocytes depended on the host species. First, to confirm that all the Drosophila strains and species used produce lamellocytes, we analyzed the main adhesive hemocytes after parasitism by ISm. As a control, we used hopTum−<sup>l</sup> flies since this D. melanogaster mutant strain is known to constitutively produce lamellocytes (42). Two main types of cells were observed in the hemolymph of these species/strains 14–18 h after parasitism: (i) round cells of 10–20µm diameter—resembling macrophagelike plasmatocytes or podocytes described in these species (43– 46)—and having a phagocytic function against injected latex beads and E. coli bacteria (**Figure S6**), (ii) large flat cells of 30– 40µm diameter, lamellocytes (**Figure 6**, 1st row), which spread

well on slide but do not phagocyte bacteria (**Figure 6**, 2nd row). After parasitism by ISm, LbGAP-containing lamellocytes were found for all species (**Figure 6**, 3rd row), although the number of cells that were reactive differed among species. It was significantly higher in D. melanogaster YR than in hopTum−<sup>l</sup> (p = 0.0061) and D. yakuba (p = 0.019), and lower in D. suzukii

FIGURE 6 | Species-specific entry of venosomes into host lamellocytes. First row: Lamellocytes in the hemolymph of D. melanogaster YR and hopTum−<sup>l</sup> , D. simulans, D. yakuba, and D. suzukii (actin labeled with green fluorescent phalloidin, nuclei in blue with DAPI) that have retained their "normal" shape (control). Second row: lamellocytes do not contain green fluorescent E. coli, indicating the absence of phagocytosis, unlike plasmatocytes (actin labeled with red fluorescent phalloidin; see also Figure S6). Third row: LbGAP spots (in red) in the most reactive lamellocytes of the species tested (see also counts in Figure 7). Fourth row: the same quantity of ISm labeled venosomes were injected in L2 larvae from the different species. Eighteen hours after injection, venosomes (green spots), LbGAP (red spots), and co-localization of both (yellow spots) were observed by confocal microscopy depending upon the species (none in D. suzukii). Bars = 20µm.

compared to YR (p < 0.001), hopTum−<sup>l</sup> (p = 0.043), and D. simulans (p = 0.015). D. yakuba thus appeared as intermediate between D. melanogaster/D. simulans and D. suzukii (**Figure 7**). The lower number of cells labeled in hopTum−<sup>l</sup> compared to YR is probably due to the very large number of lamellocytes produced by this strain. Besides, the maximum amount of LbGAP spots per reactive lamellocyte [which was previously correlated to the lamellocytes change in shape (28)] was higher in D. melanogaster compared to D. yakuba (p = 0.028 for YR, p = 0.0018 for hopTum−<sup>l</sup> ) and D. suzukii (p = 0.017 for YR, p = 0.0029 for hopTum−<sup>l</sup> ) as well as to D. simulans (only for hopTum−<sup>l</sup> , p = 0.0075), with more than twice that found in these three last species (**Figure 7**). We also microinjected labeled venosomes (**Figure 6**, 4th row) in L2 larvae of the different species and the visual observation of the co-immunolocalization of LbGAP and venosomes indicated an analogous outcome: numerous venosomes and LbGAP spots were observed in a large number of D. melanogaster and D. simulans lamellocytes, whereas D. yakuba lamellocytes contained only a few labeled venosomes, but no LbGAP immunoreaction, suggesting a possible rapid degradation. After microinjection, only few lamellocytes were produced by D. suzukii and none were labeled. Thus, the number of labeled lamellocytes and the quantity of LbGAP/venosome they can uptake seemed to match the success rate of parasitism by ISm wasps in these Drosophila species [D. melanogaster > D. simulans >> D. yakuba and

FIGURE 7 | Quantification of LbGAP entry in lamellocytes of the different Drosophila species. On the top of the figure the parasitoid success on the different Drosophila species is indicated [D. melanogaster YR (DmYR) and hopTum−<sup>l</sup> (Dm HT), D. simulans (D sim), D. yakuba (D yak), and D. suzukii (D suz)] (++ > 90%; + about 60%; ± < 10%; −0%). Upper panel: Percent of LbGAP positive lamellocytes, 18 h after parasitism by ISm females (mean ± s.d). The total number of lamellocytes counted from different fields of at least two separate experiments is indicated on the column (less lamellocytes were found in D. suzukii compared to the other species). Lower panel: number of LbGAP immunoreactive spots in the most reactive lamellocytes of the different species except for D. suzukii where, due to their low number, all reactive lamellocytes with more than one spot were considered (number of lamellocytes: Dm YR, n = 12; Dm HT, n = 12; D sim, n = 11; D yak, n = 13; D suz, n = 3) (mean ± s.d). Results from two separate experiments. The different letters indicate significant differences (p < 0.05).

D. suzukii (see Discussion)]. The increased entry of venosomes into D. melanogaster lamellocytes may thus rely on the existence of a specific or more efficient mechanism, restricted or absent in other species.

## DISCUSSION

Different types of vesicles are formed by Hymenoptera parasitoid wasps and injected into their hosts to ensure their reproductive success: PDVs and certain VLPs produced in the calyx cell of the wasp ovaries, derived from a viral machinery integrated in the wasp genome during evolution (9–11), and other extracellular vesicles types formed in the venom apparatus for a few reported species, including Meteorus pulchricornis (Braconidae: Euphorinae) (47), a polyphagous parasitoid of Lepidoptera larvae, and Figitidae of the genera Ganaspis and Leptopilina [the latter divided into three clades, Heterotoma, Boulardi, and Longipes (48)], which attack the larvae of Drosophila spp.

#### Venosomes Biogenesis

Mature venosomes have already been described in the venom reservoir of L. boulardi ISm and ISy strains (13, 14) but their biogenesis and their associated proteins were not previously reported. The secretory cells of the wasp venom gland secrete an electron-dense anamorphic material with no apparent membrane, also present in the gland canal. For ISm, this material, reactive to LbGAP, was retrieved later specifically associated with fully formed venosomes in the reservoir. It was also only in the reservoir that the venosomes had their ovoid shape with a membrane surrounding an electron-dense material punctuated by small membranous vesicles. Venosomes were more numerous and more structured in ISm compared to ISy, whose venosomes appear to contain less dense material and fewer enclosed vesicles, as previously described in the reservoir of these strains (13). Overall, our results confirm and emphasize that the formation/maturation of venosomes in Figitidae seems to occur extracellularly during the progression of secreted material throughout the gland canal to the reservoir. The venosomes of the Ganaspis and Leptopilina species studied to date are also 100–400 nm vesicles having different shapes in the reservoir. The biogenesis of venosomes in these species occurs through various stages of aggregation from precursor materials secreted in the lumen of the gland, whereas in the reservoir, venosomes have a pentagonal/hexagonal core with multiple extension or spikes and a distinct membrane [(17, 18); this work]. Those of the L. boulardi Lb17 strain differed in shape and biogenesis: small unshaped vesicles were present in the canal of the secretory cells of the gland and referred to as "immature," whereas the venosomes in the mature reservoir were larger, rounded vesicles enclosing a large amount of punctuate/vesicular material resembling those of ISm (18). The authors suggested that venosomes in the reservoir resulted from the maturation of the unshaped vesicles during the transit (18). How venom gland secreted proteins are associated/embedded in venosomes also has not been thoroughly explored. The P40 polyclonal antibody, directed against a protein of L. heterotoma venosomes, reacted throughout the gland, and localized with the secreted materials/blocks that assemble to form the final venosomes in which it is associated with the surface and the spikes (16–18). Here, we showed that LbGAP

had a similar secretion/embedding fate during the transit. Since venosomes contain a large number of proteins [(19); see also **Figure 1**] and we only have observations for two of them, it is difficult to identify by which mechanism(s) the secreted material will be integrated in or will be part of mature venosomes. Besides, we showed here that only some of the venom factors studied were associated with vesicles (LbGAP and LbGAP2) while others remained soluble (LbSPN), although they were all secreted throughout the gland and quantitatively important in the venom. This suggests that this process is not a simple progressive maturation of pre-existing vesicles along the cell or canals of the gland, but involves a specific mechanism of association. Our observation of extracellular membrane stacks at the junction between the connecting duct and the reservoir in the ISm strain led us to propose that the inclusion of a specific material and the final formation of the vesicles may occur at this place. This peculiar mode of extracellular vesicle formation, very different from the classical secretion of cellular exosomes or microvesicles (1–3) and from the intracellular mechanism of formation of PDVs and VLPs (10, 11), will deserve further investigation for definitive confirmation and elucidation. One may also wonder about the specific packaging mechanism of only certain secreted proteins in the venosomes. Subsequent studies are needed to determine whether the unincorporated proteins present in the luminal fluids are in a form (i.e., nonaggregated form) different from those apparently packaged. Alternatively, the different proteins may have specific sequence patterns (possibly unknown) or modification(s) to consider. The RhoGAPs sequences, composed mainly of the RhoGAP domain, have no particular motif identified so far but possible posttranslational modifications of the packaged proteins have not yet been analyzed.

#### Role of Venosomes

Purified VLPs of L. heterotoma induce lamellocyte lysis of hopTum−<sup>l</sup> larvae in vitro, and TEM observation suggested that these vesicles may enter the cytoplasm of these cells (12, 16, 22, 49). The association of the P40/P42 protein with L. heterotoma/L. victoriae venosomes and its immunolocalization in lamellocytes incubated with the venom of the reservoir led to postulate that the venosomes deliver associated proteins or toxins to lamellocytes (16, 19, 22). Based on our previous observations that LbGAP immunolocalized as "large spots" in D. melanogaster lamellocytes (27), we also suggested the association of LbGAP with L. boulardi venosomes, thus facilitating its entry into these cells. Here, the role of L. boulardi ISm venosomes in the active immune protection of the wasp egg has been clearly demonstrated by (i) the protection of L. boulardi ISy eggs from encapsulation by YR flies provided by the injection of purified ISm venosomes, (ii) immunolocalization of LbGAP and LbGAP2 in host lamellocytes whose shape was modified after ISm parasitism or microinjection of purified venosomes, and (iii) co-localization of these putative virulence factors in host modified lamellocytes together with in vitro fluorescently-labeled venosomes microinjected into the host. Since about 60% of LbGAP or LbGAP2 was not associated with venosomes in this last experiment, they may have already been released within the cell compartments, with the mechanism and timing of delivery still to be identified. Interestingly, although part of the circulating venosomes may be removed from the circulation by the phagocytes, the presence of free venosomes in the hemolymph seems to last long enough to continuously block newly formed lamellocytes, thus weakening the immune system of the larvae and disrupting encapsulation, and protecting also newly hatched parasitoid larvae. Moreover, L. boulardi venosomes appear to target not only circulating hemocytes, but also sessile hemocytes in D. melanogaster, suggesting that parasitism may also impair their functions. Thus, whatever their shape or the presence or absence of spikes, venosomes enter host lamellocytes. We can therefore conclude that they are "vehicles" for potential virulence factors with intracellular targets in the host, which parasitic wasps use to counter host immunity.

### Role of Transported Factors

The role and intracellular targets of proteins transported in lamellocytes (and maybe plasmatocytes whose function may be altered although no apparent morphological changes were observed) remain largely hypothetical. LbGAP interacts in vitro and in yeast two-hybrid experiments with Rac1 and Rac2 (27), two Rho GTPases essential for the encapsulation process (34, 50, 51), but we still await validation of these targets in lamellocytes. LbGAP2, as well as other members of the venom LbGAP family, bears mutations in the active site as well as sites important for the interaction with GTPases (25). The association of these others RhoGAPs with venosomes still needs to be tested. The way in which LbGAP2 (and maybe the other RhoGAPs) affects the physiology of lamellocytes and whether they have evolved new functions related to their mutations is still enigmatic. The high concentration of LbGAP and LbGAP2 in the venosomes could in any case facilitate their targeted quantitative delivery and their action at the cellular level. There is still a lack of information on whether venosomes may be found in other host tissues than lamellocytes and plasmatocytes. Induced apoptosis of hematopoietic precursors in the lymph gland of Drosophila by L. heterotoma and L. victoriae envenomation has been suggested (22), but this has recently been reexamined (52). Improving our approach using fluorescently labeled venosomes combined with larval clarification could help solving this question in the future.

### Species Specific Venosome Entry

We finally evaluated the level of venosomes entry in lamellocytes from various species of Drosophila to assess the specificity and thus the conservation of the involved mechanism(s). The observation of a much higher entry level of ISm LbGAP (and thus venosomes) in the lamellocytes of D. melanogaster and D. simulans [two species on which the parasitoid succeeds (43)], compared with D. yakuba and D. suzukii [very low or no parasitism success, (43–45, 53)] suggests the existence of a specific or, at least, more effective mechanism of targeting/uptake in the lamellocytes of these two species. Elucidating such mechanism will include identifying the mode of entry of the venosomes and the potential lamellocytes membrane receptors involved. Our data then suggest that a new "vesicle-cell" interaction level may take part in the host-parasitoid specificity

in addition to behavioral aspects, physiological adequacy and venom factors/targets interaction.

In conclusion, extracellular vesicles are most likely vectors used by many pathogens or parasites to propagate or transport virulence factors. We have clearly shown here that some parasitoid wasps use very unusual vesicles, assembled extracellularly in their venom apparatus, to transport and target specific potential virulence factors important for successful parasitism. Thus, these wasp species represent an interesting model to explore the role of EVs in inter-species communication, in particular to control the immune response of the host, a field that has not been investigated yet.

#### DATA AVAILABILITY

All datasets generated for this study are included in the manuscript/**Supplementary Files**.

#### AUTHOR CONTRIBUTIONS

BW performed light and fluorescence microscopy, venosomes purifications and labeling, microinjection, SDS-PAGE, and Western blotting. MR and A-NV performed transmission electron microscopy. OP set up the larval clarification protocol and helped with confocal microscopy. EG and SL performed venosomes purification and encapsulation assays. BW, J-LG, and MP wrote the manuscript. J-LG and MP designed and coordinated the work. All authors have read and approved the final manuscript.

#### REFERENCES


#### FUNDING

This work received support from the Department of Plant Health and Environment (SPE) of the French National Institute for Agricultural Research (INRA), the Provence, Alpes, Côte d'Azur (PACA) region, the Investments for the Future LABEX SIGNALIFE (ANR-11-LABX-0028) and the European Union's Seventh Framework Program for research, technological development, and demonstration, under grant agreement no. 613678 (DROPSA). BW was funded by the DROPSA program.

#### ACKNOWLEDGMENTS

We thank Dr. H. Mathé-Hubert for the statistic analyses. Dr. D. Colinet for helpful discussion, C. Rebuf for technical assistance and P. Sansigre for participation in the preliminary experiments. We also thank C. Labrosse, S. Trassard, and P.Y. Sizaret (François Rabelais University, Tours) for the first electron microscopy studies and their help for immunogold electron microscopy. Fluorescent E. coli were a gift from MP. Nawrot-Esposito (BES team, ISA). We also thank the Plate-Forme d'Infectiologie Expérimentale (PFIE; INRA-Nouzilly) for the production of rabbit antibodies, and the ISA microscopy platform (SPIBOC).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.01688/full#supplementary-material


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

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

# *Heterorhabditis bacteriophora* Excreted-Secreted Products Enable Infection by *Photorhabdus luminescens* Through Suppression of the Imd Pathway

#### Eric Kenney <sup>1</sup> , John M. Hawdon<sup>2</sup> , Damien O'Halloran1,3 and Ioannis Eleftherianos <sup>1</sup> \*

*1 Infection and Innate Immunity Lab, Department of Biological Sciences, George Washington University, Washington, DC, United States, <sup>2</sup> Department of Microbiology, Immunology, and Tropical Medicine, George Washington University, Washington, DC, United States, <sup>3</sup> Institute for Neuroscience, Department of Biological Sciences, George Washington University, Washington, DC, United States*

#### *Edited by:*

*Susanna Valanne, University of Tampere, Finland*

#### *Reviewed by:*

*Anchalee -Tassanakajon, Chulalongkorn University, Thailand Friederike Ebner, Free University of Berlin, Germany*

> *\*Correspondence: Ioannis Eleftherianos ioannise@gwu.edu*

#### *Specialty section:*

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

*Received: 06 July 2019 Accepted: 20 September 2019 Published: 04 October 2019*

#### *Citation:*

*Kenney E, Hawdon JM, O'Halloran D and Eleftherianos I (2019) Heterorhabditis bacteriophora Excreted-Secreted Products Enable Infection by Photorhabdus luminescens Through Suppression of the Imd Pathway. Front. Immunol. 10:2372. doi: 10.3389/fimmu.2019.02372* Upon entering the hemocoel of its insect host, the entomopathogenic nematode *Heterorhabditis bacteriophora* releases its symbiotic bacteria *Photorhabdus luminescens*, which is also a strong insect pathogen. *P. luminescens* is known to suppress the insect immune response independently following its release, but the nematode appears to enact its own immunosuppressive mechanisms during the earliest phases of an infection. *H. bacteriophora* was found to produce a unique set of excreted-secreted proteins in response to host hemolymph, and while basal secretions are immunogenic with regard to *Diptericin* expression through the Imd pathway, host-induced secretions suppress this expression to a level below that of controls in *Drosophila melanogaster*. This effect is consistent in adults, larvae, and isolated larval fat bodies, and the magnitude of suppression is dose-dependent. By reducing the expression of *Diptericin*, an antimicrobial peptide active against Gram-negative bacteria, the activated excreted-secreted products enable a more rapid propagation of *P. luminescens* that corresponds to more rapid host mortality. The identification and isolation of the specific proteins responsible for this suppression represents an exciting field of study with potential for enhancing the biocontrol of insect pests and treatment of diseases associated with excessive inflammation.

Keywords: parasitic nematode, *Drosophila*, innate immunity, Imd pathway, *Heterorhabditis*, *Photorhabdus*

### INTRODUCTION

The early steps of a Heterorhabditis bacteriophora infection are well-described with regard to the physical actions of the parasite. Upon migration to a host, the majority of the infective juveniles (IJ) enter the insect through natural openings, although the IJ can generate tears in the intersegmental membrane to gain entry (1). Once the parasite enters the hemocoel environment, the nematode slowly releases, following a 30-min lag time, the bacterial endosymbiont Photorhabdus luminescens that it maintains as a secondary phase in its gut (2). When considering the molecular host-parasite interactions that determine the success of an infection after IJ entry, Photorhabdus often draws a substantial amount of interest due to its assortment of proteases and other factors that can suppress the insect immune response and lead to rapid death. However, it is crucial to recall that axenic nematodes are still capable of inciting insect mortality without their symbiont (3). Furthermore, numerous reports have shown that the immune-based neutralization of the nematode is possible. While IJs have been shown to evade encapsulation in Tipula oleracea, Popillia japonica, and Cyclocephala borealis (4, 5), the Colorado potato beetle Leptinotarsa decemlineata prevents IJ development through encapsulation, in which the process of hemocyte attachment to the parasite begins as quickly as 15 min after entry, a period comfortably preceding the release of bacteria (6). Generally, the degree of melanization and encapsulation of the IJ has been shown to correlate to the survival of the insect (7), so the nematode must to some degree fare for itself in terms of immune suppression during the early phase of an infection. Additionally, Heterorhabditis has a vested interest in promoting Photorhabdus survival, so some early IJ-based immune suppression may also be targeted toward developing a more hospitable hemolymph environment for its symbiont.

Much of the work centered on the entomopathogenic nematode infection process has used Steinernema carpocapsae. A pair of serine protease inhibitors from this nematode have been found to impair hemocyte aggregation, prevent clotting fibers from forming properly, inhibit the digestive enzymes of the host, and prevent the inclusion of melanin into clots formed in the hemolymph (8, 9). The bulk secreted proteins have also been found to be lethal when injected into Drosophila melanogaster adult flies (10), clearly indicating that the nematode plays a strong role in the molecular aspect of the infection aside from merely releasing its symbiotic bacteria. Less is known about the activity of the specific molecules produced by H. bacteriophora, but genome and transcriptome studies have predicted a variety of secreted factors (11, 12) and genes for a putative metalloprotease, enolase, and chitinase have been implicated in parasitism specifically (13). Genes for C-type lectin and catalase have also been found to be upregulated upon activation of the nematode, where the former is believed to play a role in immune evasion and the latter a role in protecting the parasite from free radicals. Both are expressed in other parasitic helminths, with the lectin being found in a range of nematodes from Meloidogyne javanica to Ancylostoma ceylanicum (14).

The molecular effects of an H. bacteriophora infection are likely the product of a collection of these effectors fulfilling a variety of roles, each of which is important for understanding the host-parasite relationship, but a number of practical applications await the identification of specific individual factors. Autoimmune disease, for instance, is believed to be exacerbated by the loss of natural associations with helminth parasites, and individual immunosuppressive factors isolated from nematodes could be effective treatments for conditions like Crohn's disease, asthma, or multiple sclerosis due to their specificity of action and tolerability (15). Entomopathogenic nematodes (EPNs) including H. bacteriophora are also currently used as biocontrol agents against insect pests (16), and manipulating these nematodes to make them more effective parasites could increase their efficacy. Other EPNs including S. carpocapsae are also viable options for biocontrol, but it is important to consider that not every EPN is as successful as others against a given host. When infecting the carob moth Ectomyelois ceratoniae, S. carpocapsae is dramatically more adept at overwhelming the host, with an LC<sup>50</sup> of 2.02 IJs per larva as opposed to the 426.92 IJs required for the same activity by H. bacteriophora (17). When infecting the tomato leaf miner Tuta absoluta, however, H. bacteriophora is just as effective if not more so than S. carpocapsae (18). With this in mind, an optimal approach to developing strong biocontrol would not ignore either species.

Here we examine the immunosuppressive effects of H. bacteriophora bulk secretions on the Drosophila melanogaster immune system, and depict the degree to which this suppression compromises the insect with regard to susceptibility to a bacterial infection. Because the nematode's symbiont P. luminescens is such a strong pathogen, we hypothesize that the organisms have polarized each other's role in the infection and H. bacteriophora has become more specialized for immune suppression during the early phases of an infection for the benefit of the nematode as well as its symbiont.

### MATERIALS AND METHODS

### Insect and Bacterial Strains

Galleria mellonella larvae were acquired from Petco and Manduca sexta from DBDPet. Fly stocks were maintained on a cornmeal-soy-based diet (Meidi laboratories) with added baker's yeast and incubated at 25◦C on a 12-h day-night cycle. The Drosophila melanogaster lines used included Oregon R for P. luminescens survival experiments, phagocytosis assays, and gene expression analyses, w <sup>1118</sup> for survival experiments with tripleconcentrated ES product and Escherichia coli co-injections, RelE<sup>20</sup> for the E. coli co-injection assays, and the Diptericin(Dpt)- GFP line T4202 (III) for the transcriptional activation assay. Bacterial strains included Photorhabdus luminescens subspecies laumondii, strain TT01, the E. coli strain K12, and the RET16 derivative of the Photorhabdus temperata strain NC1. Photorhabdus strains were cultured on MacConkey Agar (Sigma) at 28◦C for a period of 48 h at which point a single colony was used to inoculate an overnight liquid culture in 10 mL of Lysogeny Broth (LB) media (VWR) incubated at 28◦C in a rotary shaker set to 220 rpm. E. coli was cultured in a similar fashion, but initial growth on agar was carried out on LB agar at 37◦C overnight.

### Culturing Axenic *Heterorhabditis bacteriophora* Infective Juveniles

Infective juveniles of the rhabditid nematode Heterorhabditis bacteriophora strain TT01 were maintained axenically through propagation in G. mellonella larvae carrying well-established infections of RET16. To establish the infection, 1 mL of an overnight RET16 culture was centrifuged for 3 min at 13,000 × g, the supernatant discarded, and the pellet washed once with sterile phosphate buffered saline (PBS). The resulting bacterial suspension was centrifuged again and resuspended, at which point the suspension was diluted 1:10 with sterile PBS, to a final volume sufficient for the injection of 50 µL of bacterial solution into the desired number of 5th to 6th instar G. mellonella larvae. To perform larval injections, G. mellonella larvae were surfacesterilized by brief submersion in a 70% solution of ethanol. The larvae were placed on ice for a period of 20 min in a 100 × 15 mm petri dish furnished with moistened 90 mm filter paper. Injections were performed with a 1 mL tuberculin syringe and 22G needle inserted in the intersegmental region at as shallow an angle as possible. Larvae were left on ice for 5 min post-injection and then kept at room temperature for 1 week. Successful RET16 infection caused the larva to die and turn the brick red color typical of a Photorhabdus infection. Those that did not display the appropriate color were discarded. After 1 week, H. bacteriophora IJs were pelleted, surface sterilized with a 3% bleach solution for 5 min, and washed twice with sterile water prior to their liberal application onto the infected G. mellonella at a concentration of ∼500 IJs per larva. This secondary infection was allowed to progress in the dark at room temperature for 8 days at which point the larvae were transferred to white traps for the collection of emerging IJs in autoclaved water supplemented with 0.01% Tween 20 (19). To confirm that the IJs were axenic, an aliquot of the surface-sterilized, putatively axenic IJs was used to infect G. mellonella larvae, and the larvae monitored for coloration indicative of an infection and support for the growth and reproduction of the IJs. IJs were considered axenic if they failed to produce red pigmentation in larvae or propagate successfully as compared to a surface-sterilized symbiotic IJ control.

### Preparation of Hemolymph From *Manduca sexta*

Approximately 500 µL of raw hemolymph was collected from each 5th instar Manduca sexta larvae. Prior to extraction, each larva was placed on ice for a period of 20 min. The area surrounding the posterior horn of the insect was treated with a 70% alcohol wipe just prior to the severing of the horn with microdissection scissors. This was performed directly over a 1.5 mL autoclaved microcentrifuge tube, as the release of hemolymph from the site of injury is rapid and immediate. To prevent melanization, an aliquot of 20 mM phenylthiourea dissolved in PBS was added to each aliquot of hemolymph to a final concentration of 0.33 mM. The extracted hemolymph was centrifuged for 5 min at 4000 × g and 500 µL of the resulting supernatant was added to 500 µL of ice-cold Ringer's buffer (100 mM NaCl, 1.8 mM KCl, 2 mM CaCl2, 1 mM MgCl2, and 5 mM HEPES adjusted to a pH of 6.9) in a separate sterile 1.5 mL microcentrifuge tube. For long-term storage, samples were frozen at −80◦C. Before use, hemolymph was thawed on ice, diluted 1:1 in ice-cold Ringer's buffer, and filtered with a 0.45µm syringe filter. Ampicillin and kanamycin were added to diluted hemolymph plasma solutions at concentrations of 100 and 50µg/mL, respectively.

#### Hemolymph Activation of Infective Juveniles and Isolation of Concentrated ES Products

Prior to activation, IJs were sedimented in aliquots of 200,000 and surface-sterilized with 3% commercial bleach in 10 mL of 0.01% Tween 20, resulting in a final hypochlorite concentration of 0.26%. Bleach-treated IJs were pelleted by centrifugation for 30 s at 1300 × g and washed twice with sterile Ringer's solution containing 0.01% Tween 20. After the second wash step, IJs were pelleted and resuspended in either 10 mL of the 25% hemolymph plasma solution (activated) or 10 mL of Ringer's-Tween (non-activated) containing antibiotics. The IJ suspensions were transferred to T75 tissue culture flasks, which were subsequently wrapped in foil and placed in a shaking incubator at 27◦C and 200 RPM. Following a 20-h incubation, the IJs were transferred to 15 mL conical tubes, centrifuged, and washed 10 times with 10 mL of Ringer's-Tween 20 solution. Following the final wash, the IJs were resuspended in 10 mL of Ringer's solution without Tween 20. These tubes were wrapped in foil and returned to the incubator for 5-h at 27◦C and 200 RPM to collect ES products. After incubation, the supernatants were removed and placed in a separate sterile 15 mL conical tube. The collected ES products were either stored at −80◦C or immediately concentrated. To concentrate the collected products, ES products were filtered through a 0.2µm low protein-binding syringe filter (Millex) and transferred to a new sterile 15 mL conical tube. Filtered products were added to a Vivaspin 6 tube (GE Healthcare) with a 3 kDa molecular weight cutoff, with aliquots of each treatment being added sequentially to the tube as sufficient volumes of solution cleared the filter. Concentration was allowed to continue until the volume of the retentate fell below 100 µL, at which point the solution was collected and supplemented with additional sterile Ringer's buffer to a final volume of 100 µL. For the triple-concentration of ES products, the same protocol was followed except that the ES products were initially distributed between two Vivaspin tubes, and the final 500 µL from each tube pooled and concentrated in a single tube until the volume was below 100 µL. ES concentrations were expressed as larval equivalents (LE/µL) by dividing the number of IJs used by the final volume of ES products.

### Protein Electrophoresis

Protein concentration of the ES products was quantified using a Pierce BCA Protein Assay Kit (Thermo Scientific) according to the manufacturer's instructions. For samples that produced a readable concentration of protein above the threshold sensitivity of the BCA assay, 6 µg of protein were loaded into a Novex WedgeWell 4–20% Tris-Glycine Gel (Invitrogen) following reduction in 50 mM DTT. For samples not producing a readable signal for protein concentration, the maximum volume was added to the gel. The final volume added to each well-included 26 µL of sample and water, 4 µL of the reducing agent, and 10 µL of Laemmli buffer. Protein size was demarcated with PageRuler Plus Prestained Protein Ladder (Thermo Scientific) and gels were stained with a Pierce Silver Stain for Mass Spectrometry kit (Thermo Scientific).

#### Injection of *Drosophila melanogaster* Adults and Larvae

For survival and gene expression analyses, treatments were loaded into an oil-filled pulled glass capillary mounted on a Drummond Nanoject III Programmable Nanoliter Injector. Adult flies aged seven to 10 days were anesthetized with carbon dioxide and injected intramesothoracially with 69.0 nL of ES products or buffer, corresponding to 138 IJ equivalents of ES products, or 414 for triple-concentrated products. Injected flies were returned to vials containing instant Drosophila medium (Carolina Biological) and kept at 25◦C on a 12-h day-night cycle. Flies injected for gene expression analysis at a 6-h time point were consistently injected in the late morning to alleviate effects attributable to natural variability arising from the circadian cycle. Wandering 3rd instar larvae were injected with 50.2 nL of ES products, representing ∼100 IJ equivalents. Each insect was washed once with Ringer's solution upon removal from their original vial. Larvae were anesthetized with carbon dioxide for ∼2–3 min before transfer to moist filter paper for injection. In order to ensure accurate, consistent injections, larvae were secured at the posterior end with forceps and injected at a shallow angle in an intersegmental region of the dorsal side of the abdomen to avoid damage to the organs or imaginal discs. Larvae were returned to a fresh petri dish furnished with filter paper moistened with Ringer's solution and incubated under the same conditions.

#### qRT-PCR Analysis for Immune Gene Expression

At the indicated time points, five adult flies (three males and two females) or five larvae were collected in duplicate and total RNA was extracted using TRIzol reagent (Ambion, Life Technologies). Reverse transcription was carried out using a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems) and 1 µg of RNA template. The subsequent RT-PCR reactions were performed in a CFX96 Real-Time System, C1000 Thermal Cycler. The reactions themselves consisted of 10 µL of GreenLink No-ROX qPCR Mix (BioLink), 40 ng of cDNA template, forward and reverse primers at a final concentration of 200 nM, and ultrapure water to a final volume of 20 µL. Cycle conditions were as follows: 95◦C for 2 min, 40 repetitions of 95◦C for 15 s followed by 61◦C for 30 s, and then one round of 95◦C for 15 s, 65◦C for 5 s, and finally 95◦C for 5 s. The primer sets used for amplification included those for Diptericin (F: 5′ GCTGCGCAATCGCTTCTACT 3′ ; R: 5′ TGGTGGAGTTGGGCTTCATG 3′ ), Cecropin (F: 5′ TCT TCGTTTTCGTCGCTCTC 3′ ; R: 5′ CTTGTTGAGCGATTC CCAGT 3′ ), Drosomycin (F: 5′ GACTTGTTCGCCCTCTTCG 3 ′ ; R: 5′ CTTGCACACACGACGACAG 3′ ), mcf1 (F: 5′ AAG GAGGTCAATGCTCGCTAC 3′ ; R: 5′ GACACAACTAATCTG CCGTTCTC 3′ ), P. luminescens 16S rRNA (F: 5′ ACAGAG TTGGATCTTGACGTTACCC 3′ ; R: 5′ AATCTTGTTTGCTCC CCACGCTT 3′ ), and rp49 (F: 5′ GATGACCATCCGCCCAGCA 3 ′ ; R: 5′ CGGACCGACAGCTGCTTGGC 3′ ). Fold change was calculated using the 2−11<sup>C</sup> <sup>T</sup> method (20, 21) with all values being normalized to rp49. Graphs show fold change for each treatment over 0-h expression and error bars represent standard error applied to 11Ct values prior to conversion to a log scale. Statistical analysis was performed with a one-way ANOVA for 11Ct values accumulated from three biological replicates with two technical replicates each.

### Fat Body Dissection and Imaging for ES-Injected Dpt-GFP Larvae

Larvae of the Dpt-GFP Drosophila line were injected with 50.2 nL of non-activated or activated ES products according to the aforementioned injection protocol. Following the 6-h incubation period, the fat body was dissected out of the insect, but left attached to the body while the gut was removed completely. Tissues were fixed in PBT (PBS containing 0.2% Triton X-1000) with 4% paraformaldehyde for a period of 30 min. Three 10-min washes in PBT were performed followed by a 30-min incubation with TRITC (Molecular Probes) diluted 1:100 in PBS. After washing once with PBS, the fat body tissues were removed from the insect carcass, cut into pieces small enough to lie flat on a slide, and mounted with Antifade mounting medium with DAPI (Molecular Probes). Images were acquired with a Zeiss LSM 510 confocal microscope and corrected total fluorescence measurements were processed for isolated green channels using ImageJ software. Ten images were analyzed per treatment for each trial.

#### Co-injection of ES Products With *Escherichia coli* and *Photorhabdus luminescens*

Co-injection solutions were prepared by mixing ES products and bacterial suspensions such that each injection contained 310 larval equivalents of ES products and either ∼8 × 10<sup>4</sup> CFUs of E. coli or 50 CFUs of P. luminescens. This was achieved by diluting cultures of E. coli (OD600 of 3.0) or P. luminescens (OD600 of 0.4) 1:4 in the triple-concentrated ES products. All solutions were mixed immediately prior to use and injected using the same injection protocol. For consistency, control treatments were likewise comprised of PBS diluted 1:4 in Ringer's solution.

### Quantification of Phagocytic Activity

Phagocytic activity was assessed by measuring fluorescence following the injection of pHrodo Red E. coli BioParticles Conjugate for Phagocytosis (Molecular Probes). A 4 mg/mL suspension of pHrodo particles was diluted 1:4 in ES products such that each co-injection contained 310 larval equivalents in a 1 mg/mL solution of pHrodo particles. Upon injection, flies were incubated at 25◦C for 1 h at which time the dorsal side of the abdomen associated with the pericardial nephrocytes was imaged using a Nikon ECLIPSE Ni microscope at 10x magnification with a Zyla (ANDOR) 5.5 camera. Corrected total fluorescence was measured using ImageJ software.

### Statistical Analysis

All statistical analyses were performed using GraphPad Prism 5 software. Gene expression analyses and CTF measurements for the phagocytosis assay were compared using a one-way ANOVA and Bonferroni multiple comparisons test to determine differences between specific treatments. Significance for CTF measurements for the Dpt-GFP assay was determined with a Student's t-test, and survival curves were assessed using a Log-Rank (Mantel-Cox) test. All analyses were performed on data accumulated through three independent experiments.

## RESULTS

### Exposure of *Heterorhabditis bacteriophora* Infective Juveniles (IJs) to Host Hemolymph Induces the Secretion of Unique Proteins

To investigate the proteins secreted in response to host stimulus, groups of 200, 100, or 25 thousand (k) H. bacteriophora IJs were activated as previously described (11). IJs were activated for 20 h by incubation in 25% Manduca sexta hemolymph diluted in Ringer's buffer, washed several times, and transferred into fresh Ringer's buffer without hemolymph to collect ES products. This activation time point was selected based on preliminary experiments in order to optimize as closely as possible an in vitro activation that may be only minimally informed by knowledge of in vivo activation kinetics. Filtered collection buffer was subsequently concentrated by ultrafiltration through a 3 kDa cutoff membrane, which restricts the analysis to proteins rather than small molecules. Activated batches of 200, 100, and 25 k IJs yielded 286, 216, and 39 ng/µL of protein, respectively, whereas protein was undetectable in ES products collected from similar numbers of non-activated IJs incubated in Ringer's throughout. To visualize proteins present in the ES products, 6 µg of activated ES products were separated by SDS-PAGE and silver stained (**Figure 1**). The maximum volume of non-activated ES products (26 µl) were used because protein was undetectable. A comparison of the lanes shows that certain species of protein are unique to the ES products of activated nematodes, with two conspicuous examples in the activated 200 K lane at estimated molecular weights of 21.2 and 18.9 kDa. Importantly, these proteins are absent from the M. sexta hemolymph, confirming that the extensive washes following the 20-h incubation removed residual hemolymph. This indicates that H. bacteriophora IJs specifically release a unique suite of proteins in response to hemolymph exposure.

### *Heterorhabditis bacteriophora* Nematode Excreted/Secreted (ES) Products Elicit Differential *Diptericin* Responses That Are Consistent Across *Drosophila melanogaster* Life Stages

The effects of concentrated ES products on the immune response of Drosophila were first examined in the context of the antimicrobial peptide (AMP) response. Imd and Toll pathway activity was assessed in flies by examining the expression of Diptericin and Drosomycin, respectively, following the injection of 69.0 nL of the highest concentration of ES products, a volume equivalent to the excretory/secretory output of 138 IJ. Expression was also assessed in larvae though with a lower injection volume of 50.2 nL, corresponding to ∼100 IJ equivalents. Both adult flies and larvae were collected at a 6-h time point following ES injection, which was chosen to capture expression during peak Imd activity. Drosomycin transcript, as measured by qPCR, was not significantly altered by the injection of activated or non-activated ES products,

Concentrated ES products were loaded for groups of 200,000 (200), 100,000 (100), and 25,000 (25) nematodes that were either activated (A) in *Manduca sexta* hemolymph or left non-activated (N) in Ringer's buffer. Activation took place over a period of 20 h, at which point IJs were washed, transferred to fresh ringers, and incubated for 5 h to collect ES products. Lanes 1,3, and 7 carry 6 µg of protein while all others were loaded with a maximum volume, as protein could not be detected using the BCA assay.

and notably the products also failed to elicit a response at the 24-h time point known to correlate to peak Toll pathway activity (data not shown). Conversely, Diptericin was significantly upregulated by injection of non-activated ES products compared to the Ringer's buffer control injection. However, injection of activated ES products failed to increase diptericin expression above the Ringer's buffer control injection, suggesting the presence of suppressive or non-immunogenic components in activated ES (**Figure 2**). This pattern was observed in both adult flies and larvae, though on a slightly larger scale through all three treatments in larvae, possibly due to the primary immune organ, the fat body, being proportionally larger relative to body size in larvae. The immune response to a nematode infection minimally includes a strong Imd response, which is apparent through Diptericin expression, and the H. bacteriophora countermeasures to this activity are clearly capable of neutralizing the effect to levels associated with mere injury rather than infection.

*Drosophila melanogaster* life stages. *D. melanogaster* adults (A) and 3rd instar larvae (B) were injected with 69.0 and 50.2 nl of non-activated (N) or activated (A) concentrated ES products, representing 138 and 100 infective juvenile equivalents, respectively. An equivalent volume of Ringer's buffer (R) served as a control. Flies and larvae were homogenized at a 6-h time point before RNA isolation, cDNA conversion, and transcript abundance quantification of the antimicrobial peptide *Diptericin* by qPCR. Fold change is relative to 0-h expression immediately following injection with each treatment and values represent data from three trials at two technical replicates per trial, where replicate measurements are drawn from the pooled cDNA of five flies or larvae (\**p* < 0.05, \*\**p* < 0.01).

### Excretory-Secretory Product-Based Differential *Diptericin* Responses Originate at or Prior to Transcriptional Activation

To more precisely describe the effects of ES products on the regulation of Diptericin, larvae of a Drosophila line carrying GFP under the control of the Diptericin promoter were injected with ∼100 IJ equivalents of either activated or non-activated products, and collected for observation at a 6-h time point. The fat body was dissected and imaged by confocal microscopy at 40x magnification. Fluorescence was clearly visible in all samples, though on average fat body samples that had been exposed to non-activated products were substantially brighter than those treated with activated ES products. This observation was confirmed by corrected total fluorescence (CTF) measurements of isolated green channels for each image (**Figure 3**). Because fluorescence is a measure of promoter activation, the specific interaction that mediates the differential responses to activated and non-activated ES products can be posited to take place either at or upstream of transcriptional activation. These measurements also confirm that the differences seen in Diptericin expression are mediated at least in part by cells of the fat body.

### Triple-Concentrated Activated *Heterorhabditis bacteriophora* Nematode ES Products Are Lethal to Adult *Drosophila melanogaster*

While the activated ES products clearly do not provoke as strong a Diptericin response as the non-activated products, the relative equivalence of the responses to Ringer's buffer and activated ES products makes it impossible to determine whether the activated nematode is secreting factors that suppress immunity or simply eliminating the production of factors that are immunogenic in the host. In an attempt to resolve this ambiguity, three separate batches of ES products produced with 200,000 IJs were concentrated together such that suppressive effects would be stronger, but the absence or masking of immunogenic compounds would not have compounding effects on Diptericin expression to limit upregulation below that evoked by a control injection. The increased potency of these products was immediately apparent, as injection of 414 IJ equivalents resulted in ∼70% mortality over a period of 6 h (**Figure 4**). Flies that survived the injection at the 6-h time point were collected and Diptericin transcript levels measured. The 3x concentrated ES products significantly decreased the Diptericin response below that of the Ringer's buffer alone (**Figure 5A**), thus indicating that H. bacteriophora secretes factors capable of the specific suppression of Diptericin upregulation. The specificity of Diptericin suppression was examined by assessing the response of a second Imd-responsive AMP, Cecropin, as well as the Toll pathway AMP Drosomycin to the concentrated activated ES products. Injection of 414 IJ equivalents had no effect on either Cecropin (**Figure 5B**) or Drosomycin (**Figure 5C**) expression in adult flies.

#### *Heterorhabditis bacteriophora* Nematode ES Products Promote Mortality Driven by Both Pathogenic and Non-pathogenic Bacteria

While the specific suppression of Diptericin is significant, this result does not allow conclusions about whether the ES products released by H. bacteriophora are sufficiently immunosuppressive

to augment a bacterial infection. To explore this possibility, adult flies were co-injected with a high inoculum of Escherichia coli (8 × 10<sup>4</sup> CFUs) and ∼310 IJ equivalents of activated ES products, non-activated ES products, or an equivalent volume of Ringer's buffer. Tracking mortality every 12 h for a period of 72 h revealed that while Ringer's buffer or non-activated ES products co-injected with E. coli were not lethal to flies, the injection of activated products and E. coli together (A+Ec) resulted in ∼50% mortality in the first 24 h (**Figure 6**). Notably, this reduced dose (310 vs. the lethal 414 IJ equivalents) of activated ES products no longer induces mortality, so the observed decrease in survival cannot be attributed to the previously noted lethality stemming from the products alone.

In the context of a natural H. bacteriophora infection, the bacteria of interest would be the natural symbiont of the nematode, Photorhabdus luminescens. The bacteria are released from the gut of the IJ shortly after entry into the hemolymph, and the possibility exists that the ES products may serve in part to prepare the hemolymph environment for a more successful infection by P. luminescens. This possibility was tested by similarly co-injecting adult flies with ∼310 IJ equivalents of ES products or an equivalent volume of Ringer's buffer and 50 cells of P. luminescens. Time points were at 12 h, then every hour from 24 to 33 h, in order to capture the majority of mortality, and then once again at 48 h. Survival curves revealed a slightly protective effect imparted by the non-activated ES products relative to the control injection, and when compared to co-injections with activated ES products, the lethality produced by the activated ES products and P. luminescens was significantly different from and effected earlier than that produced by the non-activated ES product co-injections (**Figure 7**). Even at a sublethal dose, the ES products of H. bacteriophora are sufficiently immunosuppressive

*Drosophila* adults were injected with 69.0 nl of Ringer's buffer (R) or 414 IJ-equivalent dose triple-concentrated ES products, either activated (A) or non-activated (N) and monitored for mortality every hour for 6 h, at which point injected populations typically stabilized and no additional deaths were observed up to a 24-h time point. Each curve is comprised of measurements for three trials of 10 male and 10 female flies (\*\*\**p* < 0.001).

to negatively impact the AMP response and to enhance the virulence of a bacterial infection.

#### *Photorhabdus luminescens* Proliferates More Rapidly in Adult *Drosophila* When Co-injected With Activated ES Products

The delay in the onset of mortality for populations of flies co-injected with ES products and P. luminescens (**Figure 7**) demonstrates that this mortality initiated by Photorhabdus requires an accumulation of bacteria beyond the initial inoculum. To test whether the influence of activated ES products is capable of accelerating this accumulation, the co-injections of ES products and Photorhabdus were repeated under the same conditions and surviving flies were collected at a 14-h time point for the assessment of relative bacterial growth, as measured by RT-qPCR targeting the P. luminescens 16S rRNA and mcf1 genes. Subsequent analysis of expression for both genes revealed that bacterial proliferation is significantly higher in the presence of activated ES products, which supported an ∼100-fold transcript increase for each gene (**Figure 8**). Those treated with either Ringer's buffer or non-activated products showed increases between 3- and 10-fold for the same genes. This difference in bacterial survival and proliferation is therefore likely responsible for the ∼12-h decrease in the time to mortality onset for flies coinjected with activated ES products as compared to those injected with non-activated products.

#### *H. bacteriophora* ES Products Provoke a More Active Phagocytic Response

Another possible mechanism causing increased mortality when flies are challenged simultaneously with activated nematode ES products and bacteria is interference with the normal activity of phagocytic hemocytes. To determine whether this effect is also contributing to the enhanced success of bacteria in EStreated flies, adult D. melanogaster were co-injected with ∼310 IJ

equivalents of ES products or an equal volume of Ringer's buffer and pHrodo E. coli conjugates that fluoresce when engulfed by a phagocyte. CTF measurements of images captured with fluorescence microscopy showed that phagocytic activity around the pericardium, where the highest degree of activity is observed, is significantly elevated in flies co-injected with activated ES products (**Figure 9**). The immunosuppressive effect of the ES products is not mediated by the phagocytic response, and may in fact provoke more phagocytic activity. Despite this compensatory phagocytic response, the effects on AMP production or other systems are still potent enough to enhance a bacterial infection.

### DISCUSSION

With the entirety of the observed effects relying on the in vitro activation of IJs, the degree to which the collected ES products align with those of an in vivo infection should be addressed. For Steinernema species, activation has been shown to be influenced by host species, the age of the IJs being activated, the homogenate concentration used for activation, and the duration of exposure to host components (22, 23). These factors could similarly affect the activation of H. bacteriophora, which is able to infect lepidopterans, dipterans, coleopterans, hymenopterans, anoplurans, orthopterans, homopterans, and hemipterans to varying degrees of lifecycle completion (24). Each of these hosts may provoke a slightly different response from the IJs, possibly even a different assortment of ES products. Furthermore, H. bacteriophora is known to inhibit IJ development in a conspecific manner through a small-molecule pheromone termed C11 EA (25), indicating that the concentration of IJs could tune the activation state based on the ratio of suppressive conspecific signal to activating host signal. Laboratory propagation of the nematodes can also be a factor in that Heterorhabditis virulence can be affected by not only the number of generations that have been propagated in laboratory conditions, but also the number of IJs used to infect a host during each passage (26). It is therefore immediately crucial to concede that any collection of ES products from entomopathogens activated in vitro will likely not contain the ES products of the nematode in a universal sense, but rather a subset of products specific to a given activation and collection protocol. This fact does nothing however to diminish the practical or informative value of effects stemming from an isolation of ES products provided they accurately represent at least some subset of the virulence arsenal of the nematode. For the products used in this set of assays, our results demonstrate an effective activation through the emergence of a unique protein profile. The subsequent assays serve to identify functions of these proteins that are produced specifically in response to host hemolymph.

The first effect of the ES products to be observed was the capacity of non-activated ES products to provoke higher expression of the antimicrobial peptide gene Diptericin following injection into adult Drosophila. This gene was selected by virtue of its role as a readout of the Imd pathway, for which the best described function is the production of antimicrobial peptides in response to Gram-negative bacteria (27). The Imd pathway is relevant because the bacterial symbiont of H. bacteriophora, Photorhabdus luminescens, is a Gram-negative bacterium, but also because of the pathway's association with septic injury in general (28), which would imply that Imd activation could also occur during penetration of the nematode into the cuticle. Additionally, the Imd pathway appears to have a larger role in inflammation and immunity based on its contribution to the viral response (29, 30), which further asserts that Gram-negative bacterial pathogen-associated molecular patterns (PAMPs) are not its sole activating inputs. The initial expression changes observed here imply that basally expressed components of nonactivated nematode secretions are also capable of directly or indirectly promoting Imd activity, and possibly in a specific manner. This is supported by the data shown here, as expression of Diptericin is believed to be regulated solely by the Imd pathway as opposed to having a regulatory mode like that of Attacin, which is thought to receive inputs from both the Toll and Imd pathways (31). After demonstrating the immunogenicity of nonactivated products in adults, the effect was confirmed in whole larvae, the stage more commonly associated with IJ infection, as well as specifically in the fat body. Importantly, the latter provides the additional information that the immunogenic effect of the non-activated products involves a systemic response from the fat body, a crucial distinction given that Diptericin can be expressed locally in sections of the digestive tract, specifically the proventriculus and midgut (32). Results from the Dpt-GFP assay also provide an assurance that differences stem from activity taking place at or before transcription, but additional work will be required to specify a mechanistic point of interference beyond that simple binary.

To determine whether the activated products simply lacked immunogenicity or were instead carrying out targeted suppression, the products were triple-concentrated by combining the secretions of three separate activations of 200,000 IJs. These more concentrated products were lethal through early timepoints following injection into adult flies, although less so than the secretions of Steinernema carpocapsae (10). This is consistent with previous findings regarding the in vivo virulence of axenic IJs of these two species (33). When the effect on Diptericin expression was reassessed with this higher 414 IJ equivalent dose, the upregulation induced by the activated products was significantly lower than that of the Ringer's buffer injection, while the non-activated products continued to display consistent immunogenicity. Because a loss of immunogenicity would do nothing to eliminate Imd activity induced by the vehicle control, the 414 IJ equivalent injections reveal targeted immune suppression by the activated ES products. The argument could

be made here that H. bacteriophora might generate antibiotic compounds during an infection and that these are reducing the population of Imd-activating microbes introduced by the injection. While P. luminescens is known to produce antibiotics (34), no such activity has been attributed to H. bacteriophora, and this scenario would be in stark contrast to the results of the bacterial co-injection survival assays, especially that of E. coli. If the ES products contain antibiotics, they should be strongly protective after co-injection. The suppressive capacity of the ES products was then tested for two other antimicrobial peptide genes, Cecropin and Drosomycin, which are regulated by the Imd and Toll pathways, respectively. Neither of these genes showed any significant differences between the three treatments, indicating that the transcriptional suppression observed in the case of Diptericin may be specific for that gene, though other gene products may be affected at different levels of host-parasite interactions. An infection by the filarial nematode Brugia pahangi can be inhibited by Cecropins (35), but if this is also the case for Heterorhabditis, H. bacteriophora has mediated this threat through the synthesis of a proteinase capable of degrading Cecropins (36), effectively eliminating the pressure to suppress Cecropin transcriptionally. The absence of a Drosomycin response may simply be the product of irrelevance given that neither S. carpocapsae nor H. bacteriophora nematodes induce Drosomycin expression in Drosophila larvae if the nematodes are axenic (3, 37). Generally though, the lack of activity on other antimicrobial peptide genes does at least demonstrate that the suppression of Diptericin is a more subtle, targeted effect than broad interference with immune gene transcription.

Having demonstrated that H. bacteriophora IJs respond to a host by secreting a unique set of proteins possessing immunomodulatory activity, the ES products were then tested for their contribution to infection outcome, particularly one instigated by Gram-negative bacteria due to their susceptibility to Imd outputs. Flies that were injected with activated ES products and non-pathogenic E. coli (38) displayed significantly increased mortality as compared to controls, which showed that this dose of E. coli is not lethal by itself. Mortality occurred predominantly within 24 h of injection, after which point the rate of mortality declined sharply, implying that the active proteins in the ES products are degraded or otherwise buffered by the fly at later time points. Relish mutant flies were also injected with E. coli or Ringer's buffer in order to serve as a comparison for the magnitude of suppression. Relish is the terminal transcription factor in the Imd pathway and accordingly, these flies are highly susceptible to infection by Gram-negative bacteria (39). If these flies and those treated with activated ES products are equally susceptible, this would imply a nearly complete suppression of the Imd pathway by activated ES products. Interestingly, the trajectory of the E. coli and activated ES products co-injection survival curve does most closely resemble the Relish mutant E. coli injection curve at the earliest time point, but the treatments then diverge. Generally, this effect is illustrative of the immunosuppressive capacities of the ES products, but this is still more or less inconsequential in a natural infection unless the ES products can also support the H. bacteriophora symbiont P. luminescens. The Imd pathway has been previously implicated in the immune response to P. luminescens in that Diptericin is strongly upregulated following bacterial injection, and the avirulent phoP strain of Photorhabdus is restored to full pathogenicity in Imd pathway mutants (40). The Diptericin-specific suppression facilitated by the activated ES products is thus likely relevant to the survival of Photorhabdus in Drosophila. Co-injections with ES products were repeated with a far less concentrated, ∼50 CFU inoculum of P. luminescens, which is representative of the average bacterial load of an H. bacteriophora IJ (2). The coinjection of activated ES products led to a significantly earlier

onset of mortality as compared to non-activated products while the latter also displayed a slightly protective effect as compared to Ringer's buffer, potentially due to the elevated induction of Diptericin expression. Populations of injected flies were also stable until after the 12-h time point, reaffirming the specific role of the bacteria in the mortality of co-injected flies. Furthermore, this delay compared to the E. coli co-injections implies that the injected Photorhabdus needed to replicate substantially to achieve a lethal concentration. Other findings have shown that the population responsible for eventual septicemia in an insect originates from a small subpopulation that is resistant to antimicrobial peptides (41), so part of the role of nematode ES products might be to bolster this subpopulation as much as possible. Our data support this idea in that relative Photorhabdus abundance at a 14-h time point, just after the onset of mortality, was an order of magnitude higher in flies co-injected with activated ES products. Other time points could be examined to more fully enunciate the relationship between the presence of

males, and 10 females, per treatment (\**p* < 0.05, \*\*\**p* < 0.001).

time points. Three trials of two replicates with five flies per replicate were performed (\**p* < 0.05, \*\**p* < 0.01).

activated ES products and Photorhabdus growth kinetics, but this time point was considered the most critical and sufficient for demonstrating the practical capacity of ES-based suppression. Furthermore, this system could also eventually be used to examine the interplay between Heterorhabditis and Photorhabdus virulence factors with regard to AMP suppression through different phases of the infection.

Finally, to eliminate the possibility that survival differences were stemming from the phagocytic response, H. bacteriophora ES products were co-injected with pHrodo E. coli conjugates to measure overall phagocytic activity. Activated products were found to significantly increase ingestion of the conjugates, but this increase in phagocytosis was clearly unable to promote survival during infection, which is consistent with findings that knock-down of the phagocytic receptor Nimrod C1 has no effect on the survival of Drosophila during an infection by symbiotic H. bacteriophora (42). Although this is not a comprehensive assessment of the cellular response or related immune mechanisms, our future work will focus on analyzing the effects of the ES products on several other processes including melanization, encapsulation, and clot formation.

Much of the immune response has been left uninvestigated by this set of assays, in particular the immune response specifically against the nematode, but the pattern observed here reveals a cohesive image of specific immune gene suppression that could play a crucial role in the infection process. Together, the conclusions of this work show that H. bacteriophora secretes a unique protein profile in response to a host, this collection of proteins suppresses the expression of the antimicrobial peptide-encoding gene Diptericin, and suggest that the suppressive capacity of the secreted products allows a small population of P. luminescens to propagate and overwhelm a host more quickly. This represents a fundamental component of nemato-bacterial bipartite virulence and provides a strong justification for exploring the individual components of the secreted products produced by the nematode in order to identify specific immunosuppressive proteins that could be employed in a variety of applications. The interaction of these individual proteins with host immune mediators can then be observed in the context of the effects described here, with the aim of providing a mechanistic explanation for Heterorhabditis-based immunosuppression. Given the wealth of molecular components that could be targeted to interfere with Imd responses, even outside the signaling components of the Imd pathway, it would be premature to suggest a mechanism from the effects observed here, but potential avenues of research can be suggested. One well-supported field of inquiry would be to examine the ability of these ES products to interfere with eicosanoid production. In insects, eicosanoid production relies on the ability of phospholipase A<sup>2</sup> (PLA2) to synthesize eicosanoid precursor lipids like arachidonic acid (AA), and interference with this pathway can have strong immunosuppressive effects based on the role

of eicosanoids in the regulation of cellular and humoral responses, including Diptericin expression through the Imd pathway (31, 43). Photorhabdus is known to inhibit PLA<sup>2</sup> (44), but a variety of parasitic nematodes also secrete proteins that could similarly interfere with eicosanoid synthesis through their ability to bind fatty acids, including arachidonic acid, which could sequester necessary eicosanoid precursors (45). Similar proteins have also been found in the ES products of Steinernema carpocapsae (10) and the transcriptome of activated H. bacteriophora (14). Interference with this pathway would be consistent with the findings presented here and an efficient way for the parasite to simultaneously suppress multiple immune responses.

#### DATA AVAILABILITY STATEMENT

The datasets generated for this study are available on request to the corresponding author.

#### REFERENCES


### AUTHOR CONTRIBUTIONS

EK conceived and designed the experiments, performed the experiments, analyzed the data, wrote the paper, prepared the figures, and reviewed drafts of the paper. JH, DO'H, and IE conceived and designed the experiments, supervised the project, reviewed drafts of the paper, and prepared the manuscript.

### FUNDING

This work was supported by a George Washington University facilitating fund awarded to DO'H, JH, and IE.

#### ACKNOWLEDGMENTS

We thank members of the Department of Biological Sciences at George Washington University (GWU) for critical reading of the manuscript, and Kyle Devine for rearing the flies.

nematode Heterorhabditis bacteriophora. PLoS ONE. (2013) 8:e69618. doi: 10.1371/journal.pone.0069618


nematode Heterorhabditis bacteriophora. ACS Chem Biol. (2012) 7:961–6. doi: 10.1021/cb300056q


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

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

# Expression of Human Mutant Huntingtin Protein in Drosophila Hemocytes Impairs Immune Responses

Yu-Hsien Lin1,2 \*, Houda Ouns Maaroufi1,2, Emad Ibrahim1,2, Lucie Kucerova<sup>1</sup> and Michal Zurovec1,2 \*

*<sup>1</sup> Biology Centre of the Czech Academy of Sciences, Institute of Entomology, Ceske Budejovice, Czechia, <sup>2</sup> Faculty of Science, University of South Bohemia, Ceske Budejovice, Czechia*

The pathogenic effect of mutant HTT (mHTT) which causes Huntington disease (HD) are not restricted to nervous system. Such phenotypes include aberrant immune responses observed in the HD models. However, it is still unclear how this immune dysregulation influences the innate immune response against pathogenic infection. In the present study, we used transgenic *Drosophila melanogaster* expressing mutant HTT protein (mHTT) with hemocyte-specific drivers and examined the immune responses and hemocyte function. We found that mHTT expression in the hemocytes did not affect fly viability, but the numbers of circulating hemocytes were significantly decreased. Consequently, we observed that the expression of mHTT in the hemocytes compromised the immune responses including clot formation and encapsulation which lead to the increased susceptibility to entomopathogenic nematode and parasitoid wasp infections. In addition, mHTT expression in *Drosophila* macrophage-like S2 cells *in vitro* reduced ATP levels, phagocytic activity and the induction of antimicrobial peptides. Further effects observed in mHTT-expressing cells included the altered production of cytokines and activation of JAK/STAT signaling. The present study shows that the expression of mHTT in *Drosophila* hemocytes causes deficient cellular and humoral immune responses against invading pathogens. Our findings provide the insight into the pathogenic effects of mHTT in the immune cells.

Keywords: Huntington's disease, immunity, infection, Drosophila melanogaster, phagocytosis, cytokines, antimicrobial peptide (AMPs)

#### INTRODUCTION

Huntington's disease (HD) is an inherited neurodegenerative disorder caused by an abnormal expansion of CAG trinucleotide in the Huntingtin (htt) gene. Mutant HTT protein (mHTT) contains an extended polyglutamine tract encoded by 40 to over 150 CAG repeats, which causes cytotoxicity and leads to neurodegeneration; this results in involuntary movement, cognitive impairment, and psychiatric abnormalities (1). Although many clinical symptoms of HD are related to neuronal dysfunction, emerging evidence indicates that the expression of mHTT in non-neuronal cells of the brain or in the peripheral tissues also contributes to the pathogenesis of HD (2). Abnormal phenotypic effects caused by the dysfunction of non-neuronal cells have been described in cardiac cells, muscles, the endocrine system, adipose tissue, testes and immune cells of HD patients, and also in mouse HD models (2, 3).

#### Edited by:

*Susanna Valanne, University of Tampere, Finland*

#### Reviewed by:

*Jenny Sassone, Vita-Salute San Raffaele University, Italy Ioannis Eleftherianos, George Washington University, United States*

\*Correspondence:

*Yu-Hsien Lin r99632012@gmail.com Michal Zurovec zurovec@entu.cas.cz*

#### Specialty section:

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

Received: *20 June 2019* Accepted: *25 September 2019* Published: *16 October 2019*

#### Citation:

*Lin Y-H, Maaroufi HO, Ibrahim E, Kucerova L and Zurovec M (2019) Expression of Human Mutant Huntingtin Protein in Drosophila Hemocytes Impairs Immune Responses. Front. Immunol. 10:2405. doi: 10.3389/fimmu.2019.02405*

**36**

Abnormalities related to the immune system were observed in a number of studies of HD patients (4). The expression of mHTT in both brain and peripheral immune cells (microglial and myeloid cells) induces the NF-κB signaling pathway which elevates levels of pro-inflammatory cytokines and chemokines, leading to systemic inflammation (5). In addition, macrophages isolated from HD model mice exhibited migration deficits, and microglia showed a delayed response to laser-induced injury in the brain (6). Although several studies proposed that the immune cell response is impaired in HD, this phenomenon is still poorly characterized in relation to host responses to pathogens. One recent study reported increased proliferation of a parasite, Toxoplasma gondii, in HD model mice, causing premature mortality and thus suggesting that expression of mHTT in immune cells may suppress immune responses (7).

Drosophila melanogaster has been long-term established as a HD model. In vivo experiments have revealed that the ectopic overexpression of mutant human htt (exon 1 with expanded CAG repeats) in the neural tissue of transgenic flies causes neurodegeneration (8, 9). The mechanisms of cellular pathology observed in the HD flies seem similar to those in human patients, including the suppression of mitochondrial function, transcriptional dysregulation, and neuronal apoptosis (10, 11). Genetic screening for disease modifiers in HD model flies led to the identification of the effects of sumoylation and HSP70 chaperone machinery on neurodegeneration. The subsequent confirmation that these pathways are involved in the pathology of human patients validates the Drosophila model for investigating HD (12, 13). Furthermore, since the tissue-specific expression of transgenes in Drosophila can be easily controlled using the UAS-Gal4 system, Drosophila have also been used to study the effects of HD on non-neuronal cells, including glial cells, photoreceptors, cardiac cells, and salivary glands (14–18).

The present study aimed to survey the physiological impact of mHTT expression in Drosophila hemocytes. We used the Drosophila UAS-Gal4 system to express mHTT with hemocytespecific drivers and investigated the effect of mHTT on survival, hemocyte development, and susceptibility to pathogens. We also expressed mHTT in a Drosophila macrophage-like cell line, S2 cells, and assessed the effect of mHTT on phagocytic activity, ATP levels, antimicrobial peptides, and production of cytokines. Our results suggest that the expression of mHTT in hemocytes does not directly affect survival but causes immune dysregulation, which leads to an impaired immune response against pathogenic invasion.

#### RESULTS

#### Expression of mHTT in Hemocytes Did Not Affect Larval Viability but Decreased the Number of Circulating Hemocytes

In order to characterize the effects of mHTT in Drosophila hemocytes, we used a tissue-specific UAS-Gal4 system by expressing wild-type human HTT (Q20) or mutant HTT (Q93) under the control of a pan-neuronal driver, elav-gal4, or hemocyte drivers, hml-gal4, and he-gal4. The flies devoid of plasmatocytes (phagoless) generated by expressing pro-apoptosis genes, rpr and hid with hml-gal4 were used as negative control (19, 20). The results showed that the ectopic expression of Q93 with the pan-neuronal driver (elav-gal4) decreased both the eclosion rate and the longevity of the adult flies, but not the rate of pupariation (**Figure 1**). The expression of Q20 and Q93 with both hemocyte drivers (hml-gal4 and he-gal4) had no effect on pupariation and eclosion rates. Furthermore, the differences in longevity between Q20 and Q93 flies were not significant, and their survival rate was higher than phagoless flies (**Figure 1C**). These results indicated that the hemocyte-specific expression of mHTT did not influence fly viability, unlike its expression in the brain.

Although the expression of mHTT did not affect fly survival, we observed a significant decrease in the number of circulating hemocytes. In first-instar larvae, the number of hemocytes differed significantly only in the phagoless flies (**Figure 2A**). However, a reduced amount of circulating and sessile hemocytes was apparent in the Q93 mutants from second-instar larvae. As shown in **Figure 2B**, the circular hemocyte numbers in Q93 larvae were still higher than in phagoless flies, showing about 50% of the numbers observed in the Q20 control (**Figure 2B**). These results showed that the expression of mHTT with two different hemocyte-specific drivers reduced the number of hemocytes.

#### Expression of mHTT in Hemocytes Impaired the Immune Response to Parasites

To examine whether mHTT expression in Drosophila hemocytes affects the innate immune response and whether such larvae are still able to restrain parasite development, we tested the sensitivity of such flies to entomopathogenic nematode and parasitoid wasp infections, which are two Drosophila pathogenic models for examining the cellular immune response (21, 22). Early third-instar larvae expressing mHTT (Q93), wild-type HTT (Q20) or phagoless were infected with nematode species, Heterorhabditis bacteriophora or Steinernema carpocapsae, which contain the bacterial symbionts Photorhabdus luminescens and Xenorhabdus nematophila, respectively. Mortality was calculated at 24 and 48 h post-infection. As shown in **Figure 3**, both phagoless and Q93 larvae displayed significantly higher mortality than Q20 controls. Previous studies revealed that the formation of hemolymph clot is an important innate immune response against entomopathogenic nematode infection in Drosophila (23, 24). To determine whether the expression of mHTT in the hemocytes caused clotting defects, we used an established bead aggregation assay (24, 25). Compared to the larvae expressing normal HTT (Q20), the hemolymph collected from mHTT (Q93)-expressing larvae displayed poor bead aggregation similar to phagoless larvae (**Figure 3E**). This results indicated that the expression of mHTT suppresses the clotting activity and thus increases the susceptibility to nematode infection.

Similarly, we infected Drosophila larvae with a parasitoid wasp, Leptopilina boulardi and calculated the number of emerged

fly and wasp adults. The number of eclosed Drosophila adults was not significantly different between Q20 and Q93 driven by hml-gal4 and he-gal4, while phagoless showed lower eclosion rates than both Q20 and Q93 (**Figures 4A,B**). However, the number of emerged wasps were significantly higher in both Q93 and phagoless flies, thus indicating that a greater number of wasps overwhelmed the immune reaction of Q93 hosts and successfully developed to adult stage. In addition, the higher number of wasp eggs successfully hatched in both Q93 and phagoless larvae (**Figure 4C**); these results indicated that Q93 and phagoless larvae have less efficient immune reaction against wasp infection. Since the encapsulation and melanization are major defense mechanisms against parasitoid wasp infection, we quantified the number of the melanized capsules to assess the immune activity after 72 h post-infection. We found that there were more intact melanized capsules in Q20 larvae (79%) than in those expressing Q93 (51.6%) or in phagoless (17.7%) (**Figures 4D,G** left). We also observed a higher amount of melanization pieces in Q93 or phagoless individuals than in Q20 larvae (**Figures 4E,G** middle). The formation of such defective capsules was described previously in immune-deficient mutant flies (26). Moreover, 37% of the infected phagoless larvae formed no melanization capsules compared to Q20 (0%) or Q93 (4.8%) infected larvae (**Figures 4F,G** right). These results could explain a lower proportion of phagoless adults successfully eclosed after wasp infection (**Figures 4A,B**). Taken together, our results suggest, that mHTT expression impairs the innate immune reactions to nematode and parasitoid wasp infections due to the deficient cellular immune responses such as clot formation and encapsulation.

#### Reduced Phagocytic Activity and ATP Levels in mHTT Cells

To find out whether mHTT expression could cause a detrimental effect on hemocyte functions, we expressed mHTT or wildtype HTT in Drosophila S2 cells. The S2 cell line consists of macrophage-like cells with phagocytic activity and the ability to produce antimicrobial peptides (AMPs) (27). We transfected the cells with four different recombinant constructs encoding green fluorescent protein (GFP) fused to HTT repeats under an inducible metallothionein promoter. We created stable cell lineages and confirmed that the S2 cells expressed HTT-fusion proteins by observing the GFP. As shown in **Supplemental Figure 1**, most of the cells in all cell lineages were positive for the fluorophore. Furthermore, the cells containing the mHTT Q46, Q72, and Q97 constructs (all except wild-type Q25) showed formation of mHTT aggregates.

We further treated the HTT-expressing cells with E. coli particles conjugated by pH-sensitive dye (pHrodo) to examine their phagocytic activity. This causes bright fluorescence to be visible after particle engulfment in the acidic environment of phagolysosome. The results showed that after inducing mHTT expression, the fluorescence signals were significantly lower in Q46, Q72, and Q97 mHTT-expressing cells but not in cells expressing wild-type Q25 HTT (**Figure 5A**). Quantification of the cells containing fluorescent signals showed a significant

reduction (20–30%) of fluorescent-positive cells in mHTTexpressing cells compared to the control cells (Q25) (**Figure 5B**), thus supporting the hypothesis that expression of mHTT in immune cells impairs phagocytic activity.

The phagocytic capacity of immune cells has been associated with mitochondrial activity (28–30); mHTT has been shown to cause impairment of energy metabolism and mitochondrial dysfunction in human peripheral blood cells (31). To test whether mHTT can also impair the energy metabolism of Drosophila immune cells, we measured the ATP levels in S2 lineages after mHTT induction. The results showed that ATP levels significantly decreased in cells expressing Q72 and Q97 mHTT after 72 h of induction (**Figure 6A**). The ATP levels in cells expressing Q46, Q72, and Q97 mHTT were further reduced after 120 h of induction (**Figure 6B**). This indicated that the expression of mHTT reduces ATP levels, which may further limit the cellular immune responses against pathogenic infection.

The studies in human and mouse have demonstrated that the expression of Bcl-2 family proteins associated with mitochondrial dysfunction is activated by mHTT expression (32). To assess whether the level of Drosophila Bcl-2 proteins is also altered by mHTT expression, we compared the transcription levels of two of Bcl-2 genes, buffy and debcl, in Q25- and Q97 expressing S2 cells (**Figure 6C**). We found that buffy expression is five times higher in Q97- than in Q25- expressing cells, but we did not detect any significant difference in debcl mRNA

FIGURE 3 | hemocyte ablation (*hml* > *UAS-rpr, hid*) were used as negative control. Bead aggregation assay was used for assessing the clotting activity (E). Hemolymph was collected from Q93, Q20, and *Phagoless* (*hml-gal4*) larvae, mixed with a bead suspension, and the aggregates were quantified by ImageJ software. All the experiments were performed in five to six independent replicates. Data are presented as averages ± SEM. Significances were analyzed by ANOVA with Fisher LSD *post-ho*c test; significant differences among treatment groups are marked with different letters (*P* < 0.05).

level. Different from pro-apoptotic function of debcl, buffy was suggested to play an anti-apoptotic role under stress conditions which is similar to mammalian Bcl-2 proteins (33, 34). We conclude that the alternation of ATP synthesis and buffy expression indicate the abnormality of mitochondrial function in mHTT expressing cells, and the induction of buffy might be a protective mechanism for preventing the cell death caused by mitochondrial dysfunction.

#### Upregulation of Cytokines Expression and Downstream JAK/STAT Signaling in mHTT Expression Cells

It has been reported that the level of cytokines and chemokines are abnormally increased in the plasma of HD patients (35). Consistently, the production of cytokines from monocytes and macrophages of HD patients have shown hyper-activation after lipopolysaccharide stimulation (36). To test whether mHTT has a similar effect in Drosophila, we used Schneider 2 (S2) cells and measured the effect of mHTT expression on three Drosophila cytokines, upd1, upd2, and upd3, as well as dome, jak (hop) and downstream targets of JAK-STAT signaling (**Figure 7A**). The results showed that the expression of cytokine upd3 is significantly increased in Q97 mHTT-expressing cells compared to Q25 controls (**Figure 7B**). In addition, the expression of dome receptor and four downstream targets, tep1, totA, totB, and totC were also significantly increased in Q97-expressing cells. These results indicated that the expression of mHTT induced the production of cytokines and activates JAK/STAT signaling.

### Decreased Antimicrobial Peptide (AMP) Production in Response to Bacteria

Drosophila Toll and Imd pathways control the humoral immune response against invasive microorganisms by regulating the induction of downstream AMP genes in both hemocytes and the fat body (37). To examine whether AMP induction was affected by mHTT expression in Drosophila S2 cells, we treated mHTT-expressing cells with a mixture of heat-inactivated Gramnegative bacteria, Escherichia coli, and Gram-positive bacteria, Micrococcus luteus. The induction of AMPs was assessed using qPCR. As shown in **Figure 8**, there was no difference in the expression of AMPs between Q25 and Q97 in the absence of bacterial treatment. In contrast, all AMP genes were significantly induced in both Q25- and Q97-expressing cells at 8 h after bacterial treatment. However, AMP induction levels were significantly lower in cells expressing Q97 (**Figure 8A**). We further assessed the AMP expression levels under in vivo condition after infecting larvae with phytopathogenic bacteria, Erwinia carotovora carotovora 15 (Ecc15). We examined the expression levels of dpt, dptB, attA, and cecA which were known as being highly induced after Ecc15 infection (38). Our results of larval infections showed that except for attA, the induction levels of dpt, dptB, and cecA in Q93 or phagoless larvae were significantly lower than in Q20 controls (**Figure 8B**). These results confirm that the induction of AMPs in response to bacteria was significantly suppressed in mHTT-expressing cells or larvae.

## DISCUSSION

Peripheral immune dysregulation is considered as one of the clinical features of HD pathogenesis (39). Previous studies in mice and HD patients have suggested that mHTT expression in immune cells accelerates the neurodegenerative process. The activation of pro-inflammatory products in mHTT-expressing microglial cells elevate the reactive oxygen species (ROS) and cause neuroinflammation, which contributes to neurodegeneration (5, 40). Genetic ablation or pharmacologically-blocked cannabinoid receptor 2 (interleukin-6 regulator), as well as drug suppression of the cytokine-responsive kynurenine pathway, can both slow neurodegeneration and improve the phenotype of R6/2 HD mice (41, 42). Since the expression of mHTT in HD mice and human patients is ubiquitous, it is still unclear whether mHTT expression in blood cells directly contributes to the lethal effect of HD. The present study examined mHTT expressed specifically in Drosophila blood cells and assessed its impacts on development and longevity (**Figure 1**). We found that the expression of mHTT in hemocytes did not cause mortality or a shortening in life span, which is in contrast to expression in the brain. Our results, therefore, suggest that expression of mHTT in immune cells does not directly contribute to mortality.

A reduced proliferation of immune cells has been observed in T. gondii-infected HD mice, in which the expansion of CD8<sup>+</sup> T-cells in the spleen and brain was significantly suppressed during infection (7). Our results showed that the expression of mHTT in flies with hemocyte-specific drivers causes a significant reduction in the number of circulating hemocytes (**Figure 2**), and this decrease might be caused by dysfunction of mitochondria (**Figure 7**). The mitochondrial abnormalities resulting in metabolic dysregulation in peripheral blood cells of HD patients increase oxidative damage and suppress their antioxidant capacity (40). The activation of caspase-3 and caspase-9 in lymphoblasts of HD patients increases apoptosis under stress conditions (43).

The mHTT-expressing larvae revealed a higher susceptibility to wasp and nematode infections and this phenotype was caused by defects of clot formation and encapsulation (**Figures 3**, **4**). It has been shown that wasp egg recognition by circulating plasmatocytes and their differentiation to lamellocytes for further encapsulation are important processes of the immune response against wasp invasion in Drosophila (44). The production of clotting components from hemocytes also contributes to

FIGURE 4 | Immune challenge with parasitoid wasp infection and encapsulation activity assay. Larvae expressing mHTT Q93 and HTT Q20 with *hml-gal4* (A) or *he-gal4* (B) hemocyte driver were infected with parasitoid wasp, *L*. *boulardi*. *Phagoless* flies (which underwent hemocyte ablation) were used as negative control. Thirty infected larvae were collected and the numbers of eclosed flies and wasps were calculated, indicating the number of flies overcoming the wasp infection and the number of wasps successfully escaping the fly's immune reaction, respectively. Data are presented as average ± SEM with more than 10 biological replicates. The significances of results for (A) were analyzed by ANOVA with a Fisher LSD *post-hoc* test; different letters on the treatment group indicate significant differences at *P* < 0.05. The significances of results for (B) were examined using Student's *t*-test, \*\*\**P* < 0.001, N.S., not significant. For assessing the encapsulation activity, the numbers of fly larvae containing larvae of parasitic wasps (C), intact melanized capsules (D), melanized pieces (E) as well the numbers of the infected larvae without melanization (F) were recorded. The significances of the results were analyzed by Mann-Whitney *U*-test (paired); significant differences among treatment groups are marked with different letters (*P* < 0.05). The photos show examples of intact melanized capsule in Q20 larvae, melanized pieces in Q20 and Q93 larvae and infected *phagoless* without melanization reaction (G).

wound healing and melanization, which are important against nematode or wasp infections (24, 45). mHTT-expressing macrophages and monocytes from HD mice and patients also showed migration defects toward an inflammatory stimulus (6). Hemocyte migration and adhesion are important factors for the development of embryonic macrophages, as well as successful

wound healing and encapsulation during wasp infection (46). Furthermore, decreased phagocytic activity toward bacterial particles (**Figure 5**) and a suppressed induction of antimicrobial peptides (**Figure 8**) can also contribute to immune deficiency against the bacterial symbionts of nematodes (23, 47).

A previous study showed that macrophages isolated from HD patients and R6/2 mice displayed increased phagocytosis when incubated with fluorescent polystyrene beads (48). Our results seemingly differ because we observed reduced phagocytic activity of the Drosophila macrophage-like cells expressing different mHTT fragments (**Figure 5**). Unlike their approach, we tested phagocytic activity using E. coli particles with a pH-sensitive fluorescent dye that can accurately confirm phagosome formation and initiation of the phagolysosome acidification. However, similar to their results, we found that S2 cells expressing mHTT were able to initiate phagocytosis. We tested this by treating the S2 cells with heat-inactivated E. coli labeled with DNA-specific fluorescent dye (without pH-sensor); the results showed that mHTT-expressing cells were indeed able to engulf E. coli (**Supplemental Figure 2**). Thus, our results suggest that mHTT-expressing cells were unable to complete the process of phagocytosis to final phagolysosome acidification. In addition, a defective actin function has been reported in HD mouse immune cells leading to failure of membrane ruffling (6),

FIGURE 6 | Decrease of ATP level and alternation of Bcl-2 protein expression in mHTT-expressing S2 cells. ATP levels of HTT- and mHTT-expressing S2 cells (5 × 10<sup>4</sup> cells) were measured after 72 (A) and 120 h (B) of copper induction. Each treatment was performed in five independent replicates with (+copper) or without (–copper) copper induction. Data are presented as average ± SEM. Significances were analyzed by ANOVA with Tukey HSD *post-hoc* test; different letters on the treatment group indicate significant differences at *P* < 0.05. The mRNA levels of *Drosophila* Bcl-2 proteins, *buffy* and *debcl* were measured in Q25 HTT and Q97 mHTT-expressing S2 cells after 120 h of induction (C). All the expressions were normalized to *rp49* transcript and Q25 control (11CT). Data are presented as average ± SEM from three independent replicates. *P*-values were determined using Student's *t*-test, \*\*\**P* < 0.001.

which supports our results since actin assembly is required to trigger engulfment and phagolysosome maturation for successful phagocytosis (49).

Consistent with previous observations in HD mice and patients, we also found that Drosophila hemocyte cytokine udp3 was upregulated in mHTT-expressing cells (**Figure 8**). Udp3 binds to the JAK/STAT signaling receptor, Dome, and initiates phosphorylation cascades which translocate the transcription factor, Stat92E, into the nucleus and activates downstream target genes (**Figure 7A**) (50). Two selected downstream target genes, tep1 and totA, were highly expressed in mHTT-expressing cells (**Figure 7B**). Notably, we found that the induction of antimicrobial peptides was significantly suppressed in mHTTexpressing cells after bacterial treatments, which has not yet been observed in other HD models. It is known that several human antimicrobial peptides are expressed in blood cells including neutrophils and macrophages (51). Since the transcriptomic analysis in HD blood cells has shown dysregulation of transcription in large genomic regions (52), further studies will be needed to understand whether the production of antimicrobial peptides is impaired in the blood cells of patients or HD mice during infection.

In summary, the present study demonstrates immune dysregulation in flies expressing mHTT in hemocytes (**Figure 9**). This expression does not directly cause a lethal effect, although it does reduce the number of circulating hemocytes and decrease ATP levels. Cytokine expression and downstream JAK/STAT signaling are activated upon mHTT expression, which has also been observed in HD patients and mice. In addition, the induction of antimicrobial peptides as well as the immune response against different pathogenic infections are impaired in mHTT-expressing Drosophila cells. The present study introduces a system for studying the tissue-specific effects of mHTT in Drosophila immune cells. Further studies can be applied to clarify the molecular interaction between mHTT and antimicrobial peptide pathways (Toll and IMD signaling) as well as the mechanisms of phagocytosis suppression.

### MATERIALS AND METHODS

#### Fly Stocks

Flies were reared at 25◦C on standard cornmeal medium. The fly strains used were UAS-Q20Httexon1111F1L and UAS-Q93Httexon14F132 obtained from Prof. Lawrence Marsh (UC Irvine, USA) (8), which contain 20 (wild-type) and 93 (mutant HTT) polyglutamine repeats, respectively. The pan-neuronal driver, elav-gal4[C155], and hemocyte drivers, he-gal4 and hml-gal4, were obtained from Bloomington Drosophila Stock Center and Dr. Tomas Dolezal (University of South Bohemia), respectively (53–55). Hemocyte-ablated flies (phagoless) were used as negative controls and were generated by overexpressing pro-apoptotic proteins (UAS-rpr, -hid) with hml-gal4 (19, 20, 56).

#### Developmental and Longevity Assay

Thirty first-instar larvae collected from a juice plate were transferred into vials to measure the number of pupae and adults for each replicate. For the longevity assay, about 20–30 newly emerged male adults were collected for each replicate and maintained at 29◦C. Q93 expression driven by pan-neuronal driver, elav-gal4 and phgoless flies were used as positive controls for longevity assay. Since expression of Q93 driven by elavgal4 (X chromosome insertion) results in high mortality of male progeny (dosage compensation) (57), female progeny were used forrecording the longevity. The number of dead flies was counted

every day. All the experiments were performed in at least six independent replicates.

### Circulating Hemocyte Counting

Circulating hemocytes were obtained from larvae by cuticle tearing in Ringer's buffer with thiourea to prevent melanization (25 µl of buffer per 6 larvae). The number of hemocytes expressing GFP (hml-gal4 or he-gal4 > UAS-gfp) were counted using a hemocytometer. At least five independent replicates were analyzed for each genotype.

## Parasitoid Wasp Infection, Eclosion, and Encapsulation Assay

Leptopilina boulardi parasitoid wasps were obtained from Dr. Jan Hrcek (Biology Center CAS) and maintained by infecting ˇ wild-type Drosophila larvae. For the wasp infection assay, forty larvae (second instar) were transferred onto a dish containing cornmeal food, and three female wasps were then placed onto the dish and allowed to attack for 72 h. After infection, 30 infected larvae were collected from the dish and transferred into a vial containing cornmeal for each replicate. Each genotype was tested in at least 10 independent replicates. The total number of eclosed flies and wasps were calculated (26). For the encapsulation assay, the infected larvae were dissected 72 h post-infection and the number of larvae containing intact melanized capsules, broken melanized pieces as well as wasp larvae was recorded.

### Nematode Infection

Two nematode species, Steinernema carpocapsae and Heterorhabditis bacteriophora, were used in this study, under previously described maintenance conditions (58). For the infection assay, nematodes were combined with autoclaved

FIGURE 8 | Suppression of antimicrobial peptides (AMPs) induction after bacterial treatment in the mHTT-expressing S2 cells and larvae. (A) Q25 and Q97-expressing S2 cells were incubated with *E. coli* and *M. luteus* (Q25\_B and Q97\_B) or without bacteria (Q25\_C and Q97\_C) for 8 h, and the expression levels of AMPs was measured. The AMPs expressions were normalized to *rp49* expression and Q25 control. (B) Larvae expressing mHTT Q93 or HTT Q20 under *hml-gal4* hemocyte drivers as well as *Phagoless* mutants were infected with ECC15-GFP and their expression of AMPs was determined after 8 h. The expressions were normalized to *rp49* transcripts and non-infected controls. The AMPs expression levels of non-infected controls for each genotype were set to one. All the data are presented as average ± SEM from three independent replicates. The significances were analyzed by ANOVA with Fisher LSD *post-hoc* test; different letters on the treatment group indicate significant differences at *P* < 0.05.

immune cells of HD patients and mice.

water to achieve a concentration of 25 infective juveniles per 10 µl. Then, 10 µl of nematode suspension was applied to paper and placed in each wells of a 96-well plate. Individual larvae were transferred to each well where they stayed in contact with the nematodes, and the plate was covered with Parafilm. The infection was conducted at 25◦C in the dark. Each experimental replicate consisted of 32 early third-instar larvae (72 h after egg hatching), and all experiments were done at least in five replicates. The number of dead larvae were counted after 24 and 48 h of infection (23, 24).

#### Bead Aggregation Assay

The bead aggregation assay was described in our previous study (24). Briefly, 2.5 µl of hemolymph was collected from six late third-instar larvae, mixed with BSA-blocked bead suspension (tosylactivated Dynabeads M-280, Invitrogen), diluted in Drosophila Ringer (pH 6.0) in a well of cavity diagnostic slide (Thermo Scientific) and covered with round cover glass. Pictures were taken with a Nikon SMZ-745T stereomicroscope associated with a CANON EOS 550D. The images were analyzed and quantified with the ImageJ graphics software with the "Analyze Particles" module.

#### Cell Culture

Drosophila Schneider 2 (S2) cells were grown at 25◦C in Shields and Sang medium (Sigma) with 0.1% yeast extract, 0.25% peptone, and 10% heat inactivated fetal bovine serum. To generate stable lines expressing polyglutamine repeats, the S2 cells were transfected with four different Httex1-eGFP pMK33 plasmids (Q25, Q46, Q72, and Q97) containing copperinducible metallothionein promoter (obtained from Dr. Sheng Zhang) (59).

#### Phagocytosis Assay

After induction for 5 days (120 h) with 1 mM copper (CuSO4; Sigma), 100 µl of cell suspension (1 × 10<sup>6</sup> cells/ml) was transferred to each well of a 96-well plate. Then, 100 µl of pHrodo Red E. coli (1 mg/ml; Thermo Fisher Scientific) was applied to each well for phagocytosis testing. After 8 h of treatment, the supernatant was removed, the cells were washed two times with 1× PBS, and 100 µl of fresh medium was applied. Cells were observed and photographed with a confocal microscope. From the images, the total number of cells and the number of cells displaying red fluorescence were counted. Three experimental repeats for each treatment were done for statistical analysis.

#### ATP Measurement

Cells were treated with 1 mM copper for 3 days (72 h) and 5 days (120 h) to induce mHTT expression. Fifty microliter of a 1 × 10<sup>6</sup> cells/ml solution (5 × 10<sup>4</sup> cells) was transferred to each well of a 96-well plate. After removing the supernatant, 60 µl of CellTiter-Glo solution (Promega) was applied to each well for 10 min. Then, 50 µl of the mixture was transferred to each well of 96-well white plates and the intensity of luminescence was then measured. Five independent replicates for each treatment were performed for analysis.

#### Bacterial Infection in vitro and in vivo

Five milliliter of S2 cells (1 × 10<sup>6</sup> cells/ml) carrying copperinducible Q25 HTT or Q97 mHTT transgenes were incubated in media containing 1 mM CuSO<sup>4</sup> for 120 h in 60 mm tissue culture plates. After the induction, the cells were treated for 8 h with 1 ml of bacterial mixture containing Escherichia coli and Micrococcus luteus at an optical density (600 nm) of 1 (OD<sup>600</sup> = 1) (37). The cells were then harvested for RNA extraction.

For the in vivo infection, late third instar larvae (96 h after egg hatching) were collected and transferred into a vial with 0.5 g instant Drosophila medium (Formula 4– 24, Carolina Biological Supply) supplemented with 200 µl of bacterial suspension (OD<sup>600</sup> = 50) Erwinia carotovora carotovora 15-GFP (ECC15-GFP) and 1,300 µl of distilled water. The larvae were collected for RNA extraction 8 h after the infection (19, 38).

#### RNA Extraction

For in vitro experiments, S2 cells were washed with 1× PBS three times and harvested with 800 µl of RiboZol (VWR). Samples were preserved at −80◦C until RNA purification. For in vivo experiments, 10 larvae were washed by distilled water and homogenized by the pestle motor (Kimble) in 200 µl of RiboZol (VWR Life Science). The sample were then preserved at −80◦C for further RNA purification. RNA was isolated using NucleoSpin RNA columns (Macherey-Nagel) following the manufacturer's instructions and cDNA was synthesized from 2 µg of total RNA using a RevertAid H Minus First Strand cDNA Synthesis Kit (Thermo Fisher Scientific).

### qPCR and Primers

5× HOT FIREPol <sup>R</sup> EvaGreen <sup>R</sup> qPCR Mix Plus with ROX (Solis Biodyne) and an Eco Real-Time PCR System (Illumina <sup>R</sup> ) were used for qPCR. The cDNA was diluted 50 times before use. Each reaction contained 4 µl of EvaGreen qPCR mix, 0.5 µl of forward and reverse primer (10µM), 5 µl of diluted cDNA and ddH2O to adjust the total volume to 20 µl. The list of primers is shown in **Supplemental Table 1**. The expression level was calculated by using the (2−11CT) method. The CT value of target genes were normalized to reference gene, ribosomal protein 49 (rp49).

#### Statistical Analysis

Error bars show standard error of the mean throughout this paper. Significance was established using Student's ttest (N.S., not significant, <sup>∗</sup>P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001) or one-way ANOVA analysis with Fisher LSD or Tukey HSD post-hoc test. The Mann–Whitney U-test was used for examing the significance of the data on wasp larval hatching and the host encapsulation activities (**Figures 4C–F**). For the statistical analysis of longevity curve, we used online tool OASIS 2 to perform the weighted log-rank test for determining significance (60).

#### DATA AVAILABILITY STATEMENT

The raw data supporting the conclusions of this manuscript will be made available by the authors, without undue reservation, to any qualified researcher.

### AUTHOR CONTRIBUTIONS

Y-HL conceived the project, performed the experiments and prepared the manuscript. HM performed the hemocyte counting and imaging. EI performed the nematode infections. LK performed the clotting assay. MZ supervised the project and manuscript preparation.

### FUNDING

This work was supported by the grant agency of the University of South Bohemia (065/2017/P to Y-HL), junior grant project GACR (19-13784Y to LK). MZ was a member of the COST action Maximizing impact of research in neurodevelopmental disorders (CA16210).

### ACKNOWLEDGMENTS

We thank Dr. Sheng Zhang (UThealth) for the Httex1-eGFP pMK33 plasmids, Prof. L. Marsh (UC Irvine, USA) for Q20 and Q93 flies, Dr. Tomas Dolezal (University of South Bohemia) for the hemocyte driver line, Dr. Pavel Hyrsl, Pavel Dobes (Masaryk University) for the nematodes, and Dr. Hrcek Jan, Dr. Chia-Hua Lue (Biology Centre CAS), and Dr. Adam Bajgar (University of South Bohemia) for the parasitoid wasps, Dr. Julien Royet (IBDM, France) for bacteria ECC-15.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.02405/full#supplementary-material

#### REFERENCES


Supplemental Figure 1 | Expression of four different lengths of HTT-GFP fusion proteins under a fluorescence microscope. mHTT-expressing cells (Q46, Q72, and Q97) showed significant mHTT aggregates after copper induction, while there was no aggregate formation in normal HTT-expressing cells (Q25).

Supplemental Figure 2 | Phagocytosis assay in Q25- and Q97-expressing S2 cells with *E. coli* labeled by DNA-specific dye (Hoechst 33342). Cells expressing mHTT were able to initiate phagocytosis.

Supplemental Table 1 | List of qPCR primers used in this study.


and their mutualistic bacteria. J Insect Physiol. (2013) 59:179–85. doi: 10.1016/j.jinsphys.2012.08.003


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

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

# Emerging Mechanisms of Insulin-Mediated Antiviral Immunity in Drosophila melanogaster

Chasity E. Trammell 1,2 and Alan G. Goodman1,3 \*

*<sup>1</sup> School of Molecular Biosciences, College of Veterinary Medicine, Washington State University, Pullman, WA, United States, <sup>2</sup> NIH Biotechnology Graduate Training Program, Washington State University, Pullman, WA, United States, <sup>3</sup> Paul G. Allen School for Global Animal Health, College of Veterinary Medicine, Washington State University, Pullman, WA, United States*

Arboviruses (arthropod-borne viruses), such as Zika (ZIKV), West Nile (WNV), and dengue (DENV) virus, include some of the most significant global health risks to human populations. The steady increase in the number of cases is of great concern due to the debilitating diseases associated with each viral infection. Because these viruses all depend on the mosquito as a vector for disease transmission, current research has focused on identifying immune mechanisms used by insects to effectively harbor these viruses and cause disease in humans and other animals. *Drosophila melanogaster* are a vital model to study arboviral infections and host responses as they are a genetically malleable model organism for experimentation that can complement analysis in the virus' natural vectors. *D. melanogaster* encode a number of distinct mechanisms of antiviral defense that are found in both mosquito and vertebrate animal systems, providing a viable model for study. These pathways include canonical antiviral modules such as RNA interference (RNAi), JAK/STAT signaling, and the induction of STING-mediated immune responses like autophagy. Insulin signaling plays a significant role in host-pathogen interactions. The exact mechanisms of insulin-mediated immune responses vary with each virus type, but nevertheless ultimately demonstrates that metabolic and immune signaling are coupled for antiviral immunity in an arthropod model. This mini review provides our current understanding of antiviral mechanisms in *D. melanogaster*, with a focus on insulin-mediated antiviral signaling, and how such immune responses pertain to disease models in vertebrate and mosquito species.

#### Edited by:

*Susanna Valanne, Tampere University, Finland*

#### Reviewed by:

*Sara Cherry, University of Pennsylvania, United States Penghua Wang, University of Connecticut Health Center, United States*

> \*Correspondence: *Alan G. Goodman alan.goodman@wsu.edu*

#### Specialty section:

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

Received: *15 October 2019* Accepted: *04 December 2019* Published: *20 December 2019*

#### Citation:

*Trammell CE and Goodman AG (2019) Emerging Mechanisms of Insulin-Mediated Antiviral Immunity in Drosophila melanogaster. Front. Immunol. 10:2973. doi: 10.3389/fimmu.2019.02973* Keywords: innate immunity, RNA interference, JAK/STAT, insulin, STING, West Nile virus, Zika virus, dengue virus

### INTRODUCTION

Mosquitoes are a prominent vector for various arboviruses including West Nile virus (WNV), Zika virus (ZIKV), and dengue virus (DENV). These viruses pose a significant concern to human populations as the mosquitoes' continual encroachment into previously unexposed regions expands (1, 2). This habitat expansion renders more individuals at risk of exposure with limited, if any, treatments available. Climate change has also resulted in alterations in mosquito seasonal activity (3) and feeding behavior (4) resulting in increasing frequency and severity of arboviral cases. There is a direct correlation between the expansion of vector-competent mosquitoes and disease incidence within afflicted regions [reviewed in (5, 6)] indicating that vector activity is a significant risk factor for arboviral disease.

**51**

**Figure 1** outlines the transmission cycle of various arboviruses as they move within host populations and how Drosophila melanogaster can be used to study arboviral immunity for each system. Transmission from mosquito to vertebrates requires a bloodmeal exchange where infected saliva is ejected into the new host. Viral replication then permits the spread of virus from infected host to mosquito to continue the transmission cycle (**Figure 1**). Research is required to identify the signaling responses used in regulating these viruses at the vector and human level. Studies regarding immune responses initiated during the initial bloodmeal exchange (7) are important as this event is a key determinant whether transmission occurs (8). An emphasis as to how immune and nutritional signaling interact with one another is of particular interest as both would be active during ingestion of an infected bloodmeal. Insulin-mediated signaling regulates numerous viruses by inducing activation of canonical immune pathways (9). Because insulin is ingested during the bloodmeal, a recent study has shown that vertebrate insulin is able to regulate the type of innate immune response that occurs during viral infection in insect vector hosts (10).

Previous work has identified the signaling pathways that respond to arboviral infection and their significance with respect to disease outcome and severity (11–13). Drosophila have proven to be a significant model organism for studying arboviruses as many of the signaling pathways identified are conserved amongst insect species [reviewed in (14)]. These studies have utilized the genetic power provided by the Drosophila system to demonstrate the effect that nutritional status poses on host immunity.

Immune responses during various arboviral infections are evolutionarily conserved among insects and include the canonical RNA interference (RNAi) (13, 15, 16), JAK/STAT (10, 12), and STING-mediated signaling (17). These pathways are associated with the insulin/insulin-like growth factor signaling (IIS) pathway and have been established as key determinants in vector competency and disease outcome (9, 18). It has been demonstrated that ingestion of vertebrate insulin regulates whether an RNAi- or JAK/STAT-mediated response is active during infection against WNV (10). STING-mediated immunity has been previously linked to induce JAK/STAT signaling (19) and affects nutritional homeostasis during infection (20, 21) implying that it may be regulated by insulin as well. Since insulin-mediated signaling appears to have a broad impact on insect immunity, recent studies have sought to establish how insulin connects each antiviral pathway to respond to different arboviruses. Because vector competency and transmission is so closely dependent on gut-associated immune signaling, the connection that nutrition and immunity has is indicative of their importance in regulating infection (9, 22). This mini review presents a condensed understanding regarding the major responses that occur during arboviral infection using Drosophila,

studying transmission dynamics and host-pathogen interactions at both vector and human level.

the role that insulin signaling plays, and a summation of current and future efforts taken within this field.

### RNA INTERFERENCE PATHWAY

One of the most broadly restricting antiviral responses used by insects is the RNAi pathway (13, 16). RNAi signaling occurs in response to detection of viral nucleic acids within the cytosol of infected cells. In Drosophila, recognition of viral nucleic acids by the endonuclease Dicer-2 results in the recruitment of proteins Argonaute-2 (AGO2) and r2d2 to form an RNA-induced silencing complex (RISC) (13, 23). This results in the cleavage and degradation of bound viral nucleic acids (13, 23, 24). While this antiviral response is a significant component of insect immunity against RNA viruses like Sindbis virus (SINV) (24) and ZIKV (25), an arthropod-borne alphavirus and flavivirus, respectively, RNAi has also been shown to respond to DNA viruses like Invertebrate iridescent virus 6 (IIV-6) (26) (**Figure 2A**). While Harsh et al. showed that the loss of the RNAi component, Dicer-2, resulted in increased ZIKV replication and mortality, they linked the increased susceptibility of these flies to dysregulated homeostasis of the gut and fat body. Moreover, studies have shown that another RNAi component, namely AGO2, is dispensable for an antiviral response against ZIKV (17, 25). Additionally, there are flaviviruses that encode viral suppressors of RNAi (VSR) (27). Not only do VSRs function in Drosophila model systems (28), but they also function during WNV, DENV, and Yellow fever virus infections in Culex mosquitoes (29, 30). Thus, the antiviral role of RNAi depends not only on the host but also virus type. Further studies are needed to clarify the antiviral role of RNAi in model and vector organisms, especially with respect to flaviviruses and the VSRs they may encode. Mosquitoes become infected and spread disease via bloodmeal exchanges. Because of the direct role that nutritional acquisition has during infection, its role in antiviral immunity is mediated in part by regulating RNAi. In both insect and mammalian systems, IIS regulates the transcription factor forkhead box O (FOXO), and FOXO is predominately associated with longevity and nutritional signaling as it induces dInR (insulin-like receptor) in Drosophila (31, 32). FOXO possesses a secondary role in host immunity through its induction of RNAi-specific genes. Specifically, FOXO regulates the transcription of Dicer-2 and

FIGURE 2 | Innate immune antimicrobial pathways are conserved in arthropods. Insects utilize RNAi (A), STING-mediated immunity (B), and JAK/STAT signaling (D) in order to effectively respond to various arboviruses at different stages of infection. The IIS pathway (C) is an important mediator in host immunity as it regulates which immune responses are active or suppressed. During times of starvation, RNAi is more active while the bloodmeal provides the needed insulin to suppress RNAi and activate JAK/STAT. STING-mediated immunity has not yet been directly linked to IIS but may be affected by broader nutritional signaling. Each of these pathways, to varying degrees, are conserved in fly, mosquito, and human systems with different efficiencies in responding to viral infection.

AGO2 in Drosophila and is demonstrated to enhance RNAi signaling during Cricket paralysis viral infection (18). Since IIS regulates FOXO transcriptional activity, there is a direct connection between RNAi immunity and insulin signaling. In Drosophila, the IIS pathway is induced by insulin-like peptides (ILPs) binding to dInR (33). Upon binding, a phosphorylation cascade commences that includes phosphorylation of PI3K and Akt. This results in the phosphorylation of nuclear FoxO at three residues, its association with the 14-3-3 chaperone protein, and export into the cytosol (32, 34) (**Figure 2C**). Insulin treatment is demonstrated to result in transcriptional suppression of Dicer-2 and AGO2 to reduce RNAi signaling (10). This insulin-mediated suppression of RNAi proposes that mosquitoes have evolved multiple immune responses to pathogens that is dependent on its nutritional status.

RNAi is evolutionarily conserved across organisms; however, its role in antiviral immunity varies between insects and mammals. While the signaling cascade and proteins involved are conserved, mammals have evolved other sensing mechanisms to detect and respond to viruses. RIG-I and MDA5 RNA sensing are canonical pathways for vertebrate innate immunity and have developed from RNA sensors like Dicer (35). In particular, RIG-I and MDA5 are shown to be critical immune regulators in response to flaviviruses like ZIKV (36) and WNV (37) and alphaviruses like SINV (38). This would imply that while RNAi is still fully functional in mammals, its role in responding to viruses is less stringent than in insect systems [reviewed in (39, 40)]. Other invertebrate and plant species utilize RNAi signaling as a means of antiviral immunity [reviewed in (41, 42)]. Researchers interested in studying RNAi immunity against arboviruses utilize Drosophila as the signaling cascades described are well-conserved in the more relevant mosquito vector. While the functional antiviral role of RNAi varies in mammalian systems, the proteins and signaling events involved are conserved. Because of this, studies utilizing Drosophila have proven it to be a viable model for study in host immunity and other regulatory functions.

#### JAK/STAT PATHWAY

Innate immunity uses various responses to different viruses. These responses include phagocytosis of viral particles or by inducing production of downstream cytokines and antiviral effectors (43). Whereas, the RNAi pathway provides a broad means of protection through the degradation of viral nucleic acids (16), the JAK/STAT pathway is a signaling cascade where detection of infection or other stimuli results in the induction of antiviral effectors like vir-1, Vago, and TotM (12, 24, 44). In Drosophila immunity, this pathway is activated upon detection of infection (45), resulting in induction of the unpaired ligands (46, 47). Unpaired ligands bind to the receptor domeless (48), resulting in the Janus kinase (JAK) ortholog hopscotch to be phosphorylated and form docking sites for phosphorylation and dimerization of the STAT92E transcription factor (12, 49). The activated STAT92E protein complex is imported into the nucleus to induce transcription of downstream antiviral effectors. This includes TotM during early stages of infection (24) and vir-1 during later stages (12) (**Figure 2D**). JAK/STAT has been shown to be involved in the immune response against the insect virus Drosophila C virus but not SINV (24) in Drosophila, and JAK/STAT is antiviral against WNV in both Drosophila and Culex mosquitoes (10, 50).

The connection between JAK/STAT and IIS is not as direct as insulin's effect on the RNAi pathway, but recent research has demonstrated that insulin signaling in insects controls a switch between RNAi- and JAK/STAT-dependent responses. Specifically, when insulin treatment causes transcriptional suppression of RNAi, insects induce enrichment of JAK/STAT (10). This is mediated by insulin's phosphorylation and activation of downstream Akt and ERK proteins during SINV and DENV infection (9, 22) that was further evaluated to induce immunity through JAK/STAT against WNV (10). Insulinmediated induction of JAK/STAT then induces the transcription of downstream antiviral effector proteins vir-1 and TotM (10, 12, 24). JAK/STAT immunity responds to pathogens during early stages of infection in the mosquito which corresponds to ingestion of an infected bloodmeal and escape from the midgut to distal tissues (51). This association between IIS and JAK/STAT signaling is indicative that vector-competent insects have evolved immune mechanisms that are responsive to nutritional acquisition. While mosquitoes use an RNAidependent response during times of starvation, the bloodmeal provides the insulin needed to activate a JAK/STAT-dependent response during infection. While the ability to regulate immune responses based on nutritional status and how it impacts viral efficacy and transmission has yet to be established, it is plausible that insulin-mediated signaling could be targeted in future vector-control protocols. Its potential as a target depends on whether insulin's antiviral activity in the salivary glands is just as important as it is in midgut or fat body, critical digestive and immune organs (8, 9).

JAK/STAT signaling is an evolutionarily conserved immune response utilized in both insect and mammalian systems. Induction of JAK/STAT in mammals results in a Type I interferon (IFN) response against viral infection (52). The JAK/STAT pathway is similarly regulated by RIG-I-like receptor (RLR) signaling in mammalian systems (35, 53). This pathway is an important means of responding to WNV in both insects (10, 44) and mammals (52, 53). Because JAK/STAT is related to other regulatory processes like cell proliferation and differentiation [reviewed in (54)], researchers have also used the Drosophila model to study the pathway's role in a non-immunological context like cellular growth (47), differentiation (48), polarization (49), and oogenesis (55). Drosophila provide a unique system for studying JAK/STAT signaling at various levels of complexity in both an immune and regulatory context to provide clarity regarding the pathway's significance in the organism.

#### STING-MEDIATED IMMUNITY

Mediators of mammalian immunity have evolved from established signaling pathways that are present in invertebrate systems (35, 56). One such set of responses is stimulator of IFN genes (STING)-mediated immunity. Upon detection of viral nucleic acids in the cytosol, the DNA sensor cGAS metabolizes cyclic dinucleotides that bind to and activate STING (11), which induces phosphorylation and activation of various transcription factors like TBK1, IRF3, and STAT6 (19, 57). These transcription factors regulate the induction of Type I IFN responses through the secretion of IFN-α and β (11, 19, 57). STING-mediated immunity has been heavily studied within mammalian systems; however, it is only recently that STING and its role in innate immunity have been identified in insects and other invertebrates (17, 56, 58–60) (**Figure 2B**).

In Drosophila, STING signaling provides immunity against both bacterial and viral infections. During infection with Listeria monocytogenes, cyclic dinucleotides are produced which results in STING-mediated signaling and nuclear import of Relish, the fly ortholog of mammalian NF-κB (56). This immune response induces transcription and secretion of IMDcharacteristic antimicrobial peptides to reduce bacterial burden (56). STING-mediated antiviral immunity also occurs through Relish and IKKβ, which regulate expression of the antiviral factor Nazo (58). In the silkmoth, Bombyx mori, STING signaling activates antiviral activity through Dredd and IMD, leading to Relish signaling and induction of antimicrobial peptides against nucleopolyhedrovirus (NPV) (59). Other studies using the Drosophila system have further evaluated STING-mediated immunity by autophagy (17).

Autophagy is a cellular process in which intracellular structures and proteins are degraded in a lysosomaldependent manner [reviewed in (61)]. Because viruses are obligate intracellular pathogens, autophagy is an established antiviral response that is partially regulated by nutritional and STING-mediated signaling (62). While STING-mediated autophagy has been established in responding to numerous viruses in mammals, recent studies using Drosophila have demonstrated that insects can utilize autophagy to respond to ZIKV in neuronal tissues (17). Specifically, ZIKV infection results in a pro-inflammatory response in Drosophila brains which induces STING-mediated activation of autophagy and immunity (17). This form of antiviral immunity is indicative that neuronal protection against arboviruses is mediated through STING-mediated signaling. This study provides another example of the versatility provided in using the Drosophila system as the fly model is a viable means for studying antiviral immunity conserved between invertebrates and vertebrates.

Unlike the RNAi and JAK/STAT pathways, which are both regulated in part by insulin signaling, a direct link between IIS and STING has yet to be shown. However, since autophagy is partially regulated by nutritional status [reviewed in (63)] and STING has been previously shown to induce STAT6 in the mammalian system (19), it is plausible that STING-mediated immunity may also be partially regulated by insulin signaling. Since STING has only recently been discovered in the insect model, future research is needed to further evaluate how STING connects to other canonical immune and nutritional pathways and its involvement in vector competency.

## PROSPECTIVE

Both the Centers for Disease Control and Prevention (CDC) and World Health Organization (WHO) agree that mosquito-borne arboviruses will be of great concern in the following years due to the expansion of mosquitoes' habitation range and activity into previously unexposed regions (64, 65). Research using Drosophila have permitted investigators to identify the key signaling events that occur during infection and develop more effective vector control protocols that target viral replication and likelihood of transmission. Recently, Drosophila have been used to identify ingestion of mammalian insulin as a key regulator in controlling WNV replication in the insect model. The genetic screen to identify insulin receptor was performed using Drosophila, and the role of insulin signaling was then validated in the mosquito model (10).

While Drosophila have been an invaluable tool in the study of arboviruses in place of the more relevant arthropod vector, there are limitations. For example, JAK/STAT signaling, while protective in the mosquito model against arboviruses such as WNV and DENV (10, 44, 66), may not be protective in Drosophila during SINV or vesicular stomatitis virus infections (12, 24). Because Drosophila are not the natural host for these arboviruses, there are limitations surrounding whether the responses observed are indeed what occurs in natural hosts. As such, work involving Drosophila must ensure that their findings are further evaluated in the more relevant model, whether that be the insect vector or human host.

Developing disease response protocols that aim at preventing transmission from arthropods to humans would be the most beneficial in terms of cost-efficiency and alleviating disease burden at a global scale. Research efforts also aim to identify novel therapeutics that are effective at treating humans postexposure. Because many of the immune pathways discussed here are also present in humans (**Figure 2**), future research aimed at identifying novel human-specific antiviral therapeutics could benefit from the use of Drosophila. The use of Drosophila to model host immunity at both the mammalian and vector level during arboviral infection has provided a greater depth of knowledge regarding which signaling pathways are involved during infection and how they can be targeted in the arthropods that transmit disease.

### AUTHOR CONTRIBUTIONS

CT wrote the first draft of the manuscript. AG revised the manuscript.

### FUNDING

Research in the Goodman lab was supported by NIH/NIAID Grant R01 AI139051 to AG. CT was supported by NIH/NIGMS pre-doctoral fellowship T32 GM008336.

### REFERENCES


controls mycobacterial infection. Nat Commun. (2017) 8:14642. doi: 10.1038/ncomms14642


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

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

## Bombardier Enables Delivery of Short-Form Bomanins in the Drosophila Toll Response

Samuel J. H. Lin, Amit Fulzele, Lianne B. Cohen, Eric J. Bennett and Steven A. Wasserman\*

*Section of Cell and Developmental Biology, Division of Biological Sciences, University of California, San Diego, San Diego, CA, United States*

Toll mediates a robust and effective innate immune response across vertebrates and invertebrates. In *Drosophila melanogaster*, activation of Toll by systemic infection drives the accumulation of a rich repertoire of immune effectors in hemolymph, including the recently characterized Bomanins, as well as the classical antimicrobial peptides (AMPs). Here we report the functional characterization of a Toll-induced hemolymph protein encoded by the *bombardier* (*CG18067*) gene. Using the CRISPR/Cas9 system to generate a precise deletion of the *bombardier* transcriptional unit, we found that Bombardier is required for Toll-mediated defense against fungi and Grampositive bacteria. Assaying cell-free hemolymph, we found that the Bomanin-dependent candidacidal activity is also dependent on Bombardier, but is independent of the antifungal AMPs Drosomycin and Metchnikowin. Using mass spectrometry, we demonstrated that deletion of *bombardier* results in the specific absence of short-form Bomanins from hemolymph. In addition, flies lacking Bombardier exhibited a defect in pathogen tolerance that we trace to an aberrant condition triggered by Toll activation. These results lead us to a model in which the presence of Bombardier in wild-type flies enables the proper folding, secretion, or intermolecular associations of short-form Bomanins, and the absence of Bombardier disrupts one or more of these steps, resulting in defects in both immune resistance and tolerance.

#### Keywords: Drosophila melanogaster, immunity, Toll, Bomanins, humoral

### INTRODUCTION

Innate immune pathways are found in plants, fungi, and animals and provide a rapid defense against a broad range of pathogens (1–3). In the fruit fly Drosophila melanogaster, the two major innate immune pathways are Toll and Imd (4–6). The Toll pathway is activated by Gram-positive bacteria with Lys-type peptidoglycan and by fungi, and is required for defense against these microbes (7–10). Conversely, the Imd pathway is activated by and plays a major role in survival against Gram-negative bacteria and Gram-positive bacteria with DAP-type peptidoglycan (11, 12). These pathways, which are both mediated by NF-κB transcription factors, are broadly conserved as initiators of innate immune responses. Activation of either pathway induces robust production of an array of immune molecules, including antimicrobial peptides (AMPs) (13–17).

Edited by:

*Susanna Valanne, Tampere University, Finland*

#### Reviewed by:

*Anni Kleino, University of Helsinki, Finland Brian P. Lazzaro, Pennsylvania State University (PSU), United States*

> \*Correspondence: *Steven A. Wasserman stevenw@ucsd.edu*

#### Specialty section:

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

Received: *18 October 2019* Accepted: *11 December 2019* Published: *10 January 2020*

#### Citation:

*Lin SJH, Fulzele A, Cohen LB, Bennett EJ and Wasserman SA (2020) Bombardier Enables Delivery of Short-Form Bomanins in the Drosophila Toll Response. Front. Immunol. 10:3040. doi: 10.3389/fimmu.2019.03040*

**58**

Lin et al. Bombardier Enables Bom Defense

AMPs are found in all kingdoms of life (18–22). These peptides have long been thought to play the principal effector role in innate immune defense due to their demonstrated in vitro antimicrobial activity and their marked upregulation after infection. However, recent research in D. melanogaster suggests that AMPs play a major role in Imd-mediated defense, but a relatively minor role in Toll-mediated immunity (23).

In contrast to the AMPs, the Drosophila-specific Bomanin peptides (Boms), which are highly induced after infection, are indispensable for resistance against pathogens controlled by the Toll pathway (24). Bom155<sup>C</sup> flies, which lack 10 of the 12 Bom genes, succumb to fungal and Gram-positive bacterial infections at rates indistinguishable from Toll-deficient flies (23, 24), suggesting that Boms rather than AMPs are the primary Toll effectors.

Bom peptides, like AMPs, are secreted from the fat body, the Drosophila immune organ, into the hemolymph, the Drosophila circulatory fluid. The family is comprised of three groups. The short-form peptides are 16–17 residues long and contain only the Bom motif. The tailed forms contain the Bom motif followed by a C-terminal tail. Finally, the bicipital forms consist of two Bom motifs connected by a linker region (24). Bom155<sup>C</sup> flies lack all six of the short-form Boms, two of the three tailed Boms, and two of the three bicipital Boms. High-level expression of short-form Boms is sufficient to rescue the sensitivity of Bom155<sup>C</sup> flies to C. glabrata infection (25). Furthermore, the absence of Toll-induced candidacidal activity in Bom155<sup>C</sup> hemolymph can be rescued by high-level expression of a short-form Bom (25). However, no in vitro antimicrobial activity has been observed with Bom peptides alone (25), suggesting that the Bomanins act in coordination with additional humoral effectors.

In this study, we demonstrate an essential role in Tollmediated humoral defense for a previously uncharacterized hemolymph protein, Bombardier (one that deploys Boms).

## MATERIALS AND METHODS CRISPR/Cas9 Deletion of bombardier

#### Locus

The bombardier gene (CG18067) was deleted using CRISPR/Cas9 technology according to established protocols (26). Briefly, a pair of gRNAs designed to delete the region 2R: 20,534,248– 20,536,154 were cloned into pU6-BbsI-chiRNA (Addgene plasmid #45946). Homology arms (1,017 bp left and 1,022 bp right) were cloned into pDsRed-attP (Addgene plasmid #51019). The plasmid pBS-Hsp70-Cas9 (Addgene plasmid #46294) was used as the Cas9 source. Constructs were injected into w 1118 embryos. F1 progeny were screened for DsRed eyes and homozygous lines were established. See **Supplemental Table 1** for gRNA and homology arm primer sequences.

### Toll Activation, Drosophila Infection, and Survival Analysis

Flies were raised at 25◦C on cornmeal molasses agar media<sup>1</sup> . The w <sup>1118</sup> strain was used as the wild type. Microbial isolates, culture conditions, and conditions for infection for Enterococcus faecalis, Enterobacter cloacae, Fusarium oxysporum, and Candida glabrata were as described previously (24), except that C. glabrata was concentrated to OD<sup>600</sup> = 100. Flies were incubated at 25◦C after live bacterial infection and at 29◦C after fungal infection. For heat-killed challenge, bacterial cultures were autoclaved and resuspended in 20% glycerol to OD<sup>600</sup> = 10 for E. faecalis and OD<sup>600</sup> = 300 for M. luteus. For both survival assays and hemolymph preparation, flies challenged with heat-killed bacteria were incubated at 29◦C.

#### Hemolymph Antimicrobial Assays

Candidacidal activity of hemolymph was assayed as described previously (25), except that hemolymph was prepared from groups of 30 flies and all activity assays were carried out for 1 h at room temperature. The number of colonies representing zero percent killing was set as the value obtained by assaying uninduced w <sup>1118</sup> hemolymph.

#### MALDI-TOF Analysis of Hemolymph

The Toll pathway was activated in flies using heat-killed M. luteus, then incubated at 29◦C for 24 h. Hemolymph was extracted as in Lindsay et al. (25), with slight modifications. Hemolymph extracted with glass capillaries from five male flies was pooled and transferred into 0.1% trifluoroacetic acid (TFA)/50% acetonitrile (ACN). One µl of each mixture was spotted on a Bruker MSP 96 ground steel plate, mixed 1:1 with a saturated solution of Universal MALDI matrix (Sigma-Aldrich) in 0.1% TFA/78% ACN, and air-dried. MALDI-TOF spectra were acquired using a Bruker Autoflex mass spectrometer. Data were collected from 1,500 to 10,000 m/z in positive linear mode, and 1,000–5,000 m/z in positive reflectron mode. Peptide calibration standard II (Bruker) was mixed with Universal MALDI matrix and used as an external calibration standard. At least ten independent samples were collected for each genotype. For peptide identification, peaks were matched to those of corresponding peaks in prior studies (13, 25). Representative spectra were visualized using R 3.3.2 and ggplot2 2.2.1 (27, 28).

#### Gene Expression Quantitation

The Toll pathway was activated with heat-killed M. luteus. Using TRIzol (Ambion), total RNA was extracted 18 h after Toll activation from four to six adult flies (2–5 days old). Next, cDNA was synthesized from 500 ng total RNA using the SuperScript II Reverse Transcriptase kit (Invitrogen). Quantitative RT-PCR was performed on an iQ5 cycler (BioRad) using iQ SYBR Green Supermix (BioRad). Quantification of mRNA levels was calculated relative to levels of the ribosomal protein gene rp49 using the Pfaffl method (29). Three independent replicates were completed. See **Supplemental Table 1** for qPCR primer sequences.

#### Hemolymph LC-MS

Flies were challenged with heat-killed M. luteus to activate the Toll pathway. Hemolymph was extracted from 100 to 110 each of w 1118 , 1bbd, and Bom155<sup>C</sup> flies using the same method as in the hemolymph antimicrobial assays, with 50– 60 flies processed per Zymo-Spin IC column (Zymo Research)

<sup>1</sup>http://blogs.cornell.edu/drosophila/dssc-cornmeal-recipe/

Frontiers in Immunology | www.frontiersin.org

and yielding a total of ∼10 µl hemolymph per genotype. Three independent biological replicates were processed for 1bbd and Bom155<sup>C</sup> , and two independent biological replicates were processed for w <sup>1118</sup>. Extracted hemolymph was mixed 1:1 (vol/vol) with denaturing buffer (8 M Urea, 50 mM Tris, pH 7.8, 150 mM NaCl, protease and phosphatase inhibitors) and protein concentration was determined using a BCA assay. For each sample, 40 µg of hemolymph was diluted to 1 M urea using 50 mM ammonium bicarbonate and digested overnight with trypsin (Promega, V511A) at a 1:100 (trypsin:protein) ratio. After digestion, peptides were reduced with 1 mM dithiothreitol at room temperature for 30 min and then alkylated with 5 mM iodoacetamide at room temperature in the dark for 30 min. Formic acid was added to a 0.1% final concentration and peptides were desalted using the C18-Stage-Tip method and then vacuum dried. The dried peptides were reconstituted in 5% formic acid/5% acetonitrile and 1 µg of total peptide for each sample was loaded for MS analysis. Samples were run in technical triplicates on a Q-Exactive mass spectrometer with instrument and chromatography settings as described previously (30), except for the following modifications: the RAW files were analyzed using Andromeda/MaxQuant (version 1.6.7.0) (31) with default settings (32) except the match between the run and LFQ quantitation settings was enabled for label free quantification. Data were searched against a concatenated targetdecoy database comprised of forward and reversed sequences from the unreviewed UniprotKB/Swiss-Prot FASTA Drosophila database (2019). A mass accuracy of 20 ppm was assigned for the first search and 4.5 ppm for the main search. The statistical analysis was calculated using the DEP analysis R-package (33).

#### Bacterial Load Quantification

Bacterial load upon death (BLUD) was obtained as in Duneau et al. (34), with slight modifications. Briefly, flies were infected with E. faecalis and vials were monitored every 30 min for newly dead flies. These flies were then individually homogenized with a pestle in 400 µl LB media. Homogenates were also prepared from individual live w <sup>1118</sup> flies 120 h postinfection (hpi). Homogenates were diluted serially in LB and spread on LB agar plates for incubation at 37◦C overnight. Colonies were counted manually and the number of viable bacteria per fly was calculated. Data were obtained from three independent experiments.

#### Data Analysis

GraphPad Prism 5 was used for statistical tests. Survival data were plotted as Kaplan-Meier curves and were analyzed using the Gehan-Breslow-Wilcoxon test to determine statistical significance. Statistical differences in candidacidal activity were calculated using one-way ANOVA followed by Tukey's test. Multiple Mann-Whitney U tests were used to calculate differences between BLUD samples (p = 0.0085 after Šidák correction for multiple comparisons, α = 0.05, k = 6). Spearman rank correlation was used to assess the relationship between BLUD and time of death.

#### RESULTS

### The bombardier Gene Is Specifically Required for Toll-Mediated Defense

The bombardier (bbd) gene contains a consensus Toll-responsive NF-κB binding site within its promoter region and is strongly expressed upon Toll activation by Gram-positive bacterial infection or other inducers (14, 17, 35, 36). The encoded protein is predicted to be secreted and to generate a mature protein of 222 amino acids with a coiled coil near its C-terminus (37, 38). Orthologs of Bombardier are found across the Drosophila genus, but in no other genera (39).

We began our analysis of the bombardier gene by generating a null mutant, using CRISPR/Cas9 to delete 1,906 bp encompassing the annotated transcriptional unit. Flies homozygous for this deletion (hereafter 1bbd) were viable and morphologically wild-type. Given that bombardier is Tollinducible, we assayed 1bbd flies for a potential loss-of-function phenotype in Toll-mediated immunity. Specifically, we infected adult 1bbd flies with various pathogens and then monitored survival. Two additional genotypes were used as controls: w 1118 flies, which served as the wild type, and Bom155<sup>C</sup> flies, which lack Toll-mediated humoral defenses due to deletion of the 10 of the 12 Bom genes (24).

As shown in **Figure 1**, we observed a marked immunodeficiency when 1bbd flies were challenged with representative species for the three classes of microbes against which Toll provides defense. With the yeast Candida glabrata, more than 90% of w <sup>1118</sup>, but no 1bbd flies, survived 5 days after infection (**Figure 1A**). In the case of the filamentous fungus Fusarium oxysporum, 70% of w <sup>1118</sup> adults, but fewer than 20% of 1bbd adults, were alive 5 days post-infection (**Figure 1B**). Finally, with Enterococcus faecalis, a Gram-positive bacterium, 50% of wild-type flies, but no 1bbd flies, were alive 5 days after infection (**Figure 1C**).

The impairment of Toll-mediated defenses by deletion of bombardier was significant for all three pathogens (p < 0.0001). In the case of C. glabrata, the immunodeficiency of 1bbd phenocopied that observed for Bom155<sup>C</sup> flies (n.s., p > 0.05). In contrast, with either F. oxysporum or E. faecalis, the rate of death was greater for Bom155<sup>C</sup> than for 1bbd ( ∗∗∗p < 0.0001 for both infections). The 1bbd mutant thus displays a substantial, but not complete, loss of Toll-mediated defense.

The expression of bombardier is strongly induced by Toll, but not Imd activation (14). We therefore hypothesized that Imdmediated defenses would not require bombardier function. To test this prediction, we infected 1bbd flies with Enterobacter cloacae, a Gram-negative bacterium. In this experiment, 1bbd flies were as immunocompetent as w <sup>1118</sup> flies: more than 90% of both genotypes survived at least 5 days post-infection (**Figure 1D**). In contrast, 100% of RelE<sup>20</sup> flies, which are deficient in Imd signaling (40), succumbed to infection within 1 day. Thus, bombardier functions in defense against a range of pathogens for which Toll mediates defense—yeast, filamentous fungi, and Lys-type Gram-positive bacteria—but not against Gram-negative bacteria, against which the Imd pathway is active.

FIGURE 1 | The *bombardier* gene is specifically required for Toll-mediated defense. (A–D) Survival curves of flies infected as indicated. The *w*<sup>1118</sup> strain was the wild-type control; *Bom*155*<sup>C</sup>* and *RelE*<sup>20</sup> were the susceptible controls (24, 40). Experiments were completed in triplicate with at least 25 flies per genotype in each replicate. Statistical significance was determined using the Gehan-Breslow-Wilcoxon test and 1*bbd* is shown relative to *w*<sup>1118</sup> in black, relative to *Bom*155*<sup>C</sup>* in orange, and relative to *RelE*<sup>20</sup> in red (\*\*\**p* < 0.0001; n.s., not significant, *p* > 0.05).

### The Candidacidal Activity of Hemolymph Requires Bombardier, but Neither Drosomycin Nor Metchnikowin

Next, we investigated the potential humoral role of Bombardier by preparing and assaying cell-free hemolymph. We have previously shown that hemolymph from wild-type flies exhibits a Toll-dependent and Bomanin-dependent candidacidal activity (25). However, we were also curious as to the identity of the active antifungal component. In particular, we considered the potential role of Metchnikowin (Mtk) and Drosomycin (Drs), two antimicrobial peptides (AMPs) that have documented antifungal activity in vitro and are strongly Toll-induced in vivo (14, 41, 42). We therefore took advantage of the recently described 1AMPs strain, which is deficient for Mtk and Drs, as well as all other induced AMPs other than the Cecropins (23). Extracting and assaying Toll-induced hemolymph, we found that hemolymph from 1AMPs flies had a killing activity against C. glabrata comparable to that of wild-type hemolymph (**Figure 2**). In contrast, we failed to detect any killing of C. glabrata by 1bbd hemolymph. We conclude that Boms and Bombardier, but neither Mtk nor Drs, are required for humoral defense against C. glabrata.

### Short-Form Bom Peptides Are Specifically Absent From 1bbd Hemolymph

MALDI-TOF provides a robust tool for characterizing small (<5,000 MW) peptides present in hemolymph after Toll activation. As shown in **Figures 3A,B**, such a readout includes the aforementioned AMPs (Mtk and Drs), several short-form Boms (BomS1, S2, S3, and S6; see **Supplemental Table 2** for updated Bomanin nomenclature), and other induced peptides (e.g., IM4). We have previously shown that deleting the 55C Bom gene cluster removes the peaks attributable to the shortform Boms, while leaving the remaining signals unaffected (25). Remarkably, analysis of 1bbd hemolymph yielded a similar pattern. As shown in **Figures 3C,D**, the short-form Boms that were readily detectable in the wild type—S1, S2, S3, and S6—were absent in 1bbd hemolymph, whereas the remaining peptides, including Mtk, Drs, and IM4, displayed a wild-type profile.

Although 1bbd disrupts the accumulation of short-form Bom peptides in hemolymph, this effect does not reflect a disruption in transcription or stability of the corresponding Bom mRNAs: robust induction of Toll-regulated genes, including genes of short-form Boms, was readily detectable with qRT-PCR (**Supplemental Figure 1**).

each point representing one replicate. One-way ANOVA was calculated followed by Tukey's test. Significance is shown relative to the null hypothesis of 0% killing (\*\*\**p* < 0.0001; n.s., not significant, *p* > 0.05). Error bars represent the 95% confidence interval.

Because proteins such as Bombardier and bicipital Boms are too large to be detected by our MALDI-TOF protocol, we used LC-MS to further characterize the relationship between Bombardier and the Boms in hemolymph. For these studies, we prepared Toll-induced hemolymph from three genotypes: w 1118 , 1bbd, and Bom155<sup>C</sup> . In wild-type hemolymph, we readily detected Bombardier protein (**Figure 4**), consistent with the presence of a canonical secretion signal sequence in the Bombardier coding sequence. Bombardier, like the Boms, is thus secreted into hemolymph upon Toll induction. We also detected all three bicipital Boms—BomBc1, BomBc2, and BomBc3. The LC-MS studies thus complemented the MALDI-TOF studies, with bicipital Boms detected by the former and short-form Boms by the latter (tailed Boms are not detected by either protocol). Next, we assayed 1bbd hemolymph. As expected, Bombardier was not detected. However, the three bicipital Boms were present at comparable levels in wild-type and 1bbd hemolymph (see **Figure 4**). Combined with the MALDI-TOF studies, these results demonstrate that 1bbd blocks accumulation in hemolymph of short-form, but not bicipital, Boms. Lastly, we analyzed hemolymph from Bom155<sup>C</sup> flies, which lack 10 of the 12 Bom genes. As expected, the products of the two deleted bicipital genes (BomBc1 and BomBc2) were absent, whereas the product of the remaining bicipital gene (BomBc3) was present at wild-type levels (see **Figure 4**). Turning our attention to Bombardier, we observed no effect of the 55C Bom deletion. Thus, Bombardier is required for the presence of short-form Boms in hemolymph, but the 55C Boms are not required for the presence of Bombardier.

#### Bombardier Mediates Both Infection Resistance and Tolerance

The 1bbd survival phenotype could be due to an inability to control pathogen growth—a defect in resistance—or an inability to endure infection—a defect in tolerance. Because flies lacking Bombardier demonstrate an increased susceptibility to infection and decreased levels of known resistance factors, the shortform Boms, it seemed likely that 1bbd flies, like Bom155<sup>C</sup> flies, have a defect in infection resistance. In exploring this hypothesis, we found that the model recently developed by Duneau et al. provided a useful framework (34). Following infection of an individual fly, there are two stereotypic outcomes: either the pathogen replicates, reaches a lethal burden, and the fly dies; or the pathogen is controlled at a level below the lethal burden and the fly survives with a persistent infection. Variation in survival curves for different pathogens and fly genotypes reflects variation in both the time required to reach lethal burden and in the fraction of flies that are able to control the infection before it reaches such a threshold. In cases where a fraction of flies control infection, group survival typically drops after infection and then reaches a plateau (23).

The survival curve for 1bbd flies infected with E. faecalis does not plateau (see **Figure 1C**). Instead, it exhibits a profile that we hypothesize reflects two phases of death. In the first phase, extending roughly 2 days post-infection, some 1bbd flies reach a lethal burden of E. faecalis and die, as reflected in a sharp decline in survival; the remainder control the infection. In the second phase, from 2.5 days onward, those flies with a persistent infection die at a reduced but steady rate, due to a defect in tolerance. If this hypothesis is correct, flies dying in the first phase should have a bacterial load upon death (BLUD) comparable to that of wild-type flies dying from infection. Furthermore, those dying in the second phase should have a much lower pathogen burden, comparable to that of wild-type survivors with a persistent infection.

To test our predictions regarding pathogen burden, we measured the BLUD of individual flies after infection with live E. faecalis and divided the data into two time intervals (**Figure 5**). For the earlier interval (dead flies obtained between 17 and 51.5 hpi), both Bom155<sup>C</sup> and 1bbd bacterial loads upon death were not significantly different from w <sup>1118</sup> (**Figure 5**, red, p > 0.05). For the later time interval (flies obtained between 68 and 120.5 hpi), 1bbd flies perished at significantly lower bacterial loads compared to that of 1bbd flies which died earlier (**Figure 5**, 1bbd early compared to 1bbd late, p < 0.0001), indicating that these two groups die from distinct causes. Importantly, late-death 1bbd flies perished at significantly lower bacterial loads than those of w <sup>1118</sup> suffering early deaths (p < 0.0001), demonstrating that 1bbd flies have a defect in tolerance.

Together, the survival curve and BLUD data offer strong support for our two-phase-model: 1bbd flies died early in infection with high bacterial loads, due to a defect in resistance, and died later with lower bacterial loads, reflecting a deficiency in tolerance. However, we note that the bacterial loads of

1bbd flies dying in the later phase were still significantly greater than those of w <sup>1118</sup> flies alive 120 hpi (**Figure 5**, 1bbd late compared to live w 1118 , p < 0.0001). This indicates that the later-death 1bbd group has not completely controlled infection compared to the live w <sup>1118</sup> flies, and suggests that both resistance and tolerance contribute to the later 1bbd fly deaths. Although we cannot rule out a minor resurgence in bacterial proliferation preceding late death of bbd flies, we note that BLUD and time of death were not significantly correlated for these flies (**Supplemental Figure 2**, Spearman correlation test, r = −0.2654, p = 0.1564).

#### Immune Activation, Specifically Bom Expression, Is Deleterious in the Absence of Bombardier

What is the nature of the tolerance defect we observed in 1bbd flies? More specifically, is their health impaired by an excessive or toxic immune response, or is death due to another class of impaired tolerance (43)? To distinguish between these explanations, we assayed the effect of activating the immune response in 1bbd flies in the absence of infection.

When 1bbd flies were challenged with heat-killed E. faecalis, we observed a decrease in survival that first was apparent 3 days post-challenge followed by a steady decline in the number of live flies in the following days (**Figure 6A**), consistent with the timing of the late-phase deaths (see **Figure 5**). Overall, the death rate was slower than that of live infection, but the extent of killing was similar between heat-killed and live E. faecalis: fewer than 20% of flies survived (compare **Figures 1C**, **6A**). In contrast, no effect on survival was observed upon challenge of either w 1118 or Bom155<sup>C</sup> flies with heat-killed E. faecalis: >95% flies survived seven or more days post-challenge.

The effect of immune stimulation on 1bbd survival was not specific to E. faecalis. When we repeated the challenge experiments with heat-killed Micrococcus luteus, which activates the Toll response [see **Supplemental Figure 1**, as well as (25, 44)], the effect on 1bbd survival was again marked: 5 days after challenge, fewer than 5% of 1bbd flies were alive, compared to

survival of >95% of w <sup>1118</sup> and 85% of Bom155<sup>C</sup> flies over the same period of time (**Figure 6B**).

As both M. luteus and E. faecalis induce the Toll pathway, Toll activation could be the key factor in 1bbd mortality. To address this hypothesis, 1bbd flies were crossed with MyD88kra<sup>1</sup> (Tolldeficient) flies to generate the MyD88kra<sup>1</sup> 1bbd double mutant, and the resulting flies were challenged with heat-killed E. faecalis and M. luteus. Unlike 1bbd flies, MyD88kra<sup>1</sup> 1bbd flies survived challenge with Toll activators (**Figures 7A,B**). Because blocking the Toll pathway with MyD88kra<sup>1</sup> rescues the 1bbd phenotype triggered by heat-killed bacteria (p < 0.0001 compared to 1bbd, p > 0.05 compared to MyD88kra<sup>1</sup> for both heat-killed bacteria), we conclude that Toll activation underlies the death of 1bbd flies in the absence of infection.

As described above, Bom genes are transcribed in 1bbd flies (**Supplemental Figure 1**), but short-form Bom peptides do not appear in hemolymph (**Figure 3**). This suggests a mislocalization of these peptides, perhaps in an unprocessed or misfolded state. Given that short-form Bom genes are among the most abundantly transcribed genes after infection (17, 36), such mislocalized or misfolded Boms could rapidly accumulate to high levels in 1bbd flies. Could this explain the death of 1bbd flies upon immune stimulation? To address this question, we generated Bom155<sup>C</sup> 1bbd double mutants and assayed the effect of immune induction alongside both Bom155<sup>C</sup> and 1bbd flies (**Figures 7A,B**). The result was unequivocal: introducing Bom155<sup>C</sup> , which deletes all of the short-form Boms, eliminated the effect of 1bbd on survival following immune stimulation (p < 0.0001 compared to 1bbd, p > 0.05 compared to Bom155<sup>C</sup> for both heat-killed bacteria). The fact that Bom155<sup>C</sup> is epistatic to 1bbd demonstrates that Toll-driven expression of Bom genes is specifically responsible for the death of immune-stimulated 1bbd flies.

#### DISCUSSION

The results presented in this study identify a key factor that regulates humoral and Bom-mediated defense in Drosophila. We demonstrate that 1bbd flies are defective in resistance to pathogens controlled by the Toll pathway. The results support the hypothesis that this defect results from the absence of shortform Boms in 1bbd hemolymph. Absence of Boms is sufficient to cause a defect in resistance (24) and 1bbd hemolymph appears to be lacking the short-form Boms but no other component, save Bombardier itself. Furthermore, 1bbd phenocopies Bom155<sup>C</sup> with regard to survival after C. glabrata infection, and resistance to C. glabrata can be restored in Bom155<sup>C</sup> flies by expression of short-form Boms (25). Finally, 1bbd hemolymph lacks candidacidal activity, which is dependent on short-form Bom peptides (25) and which we show here does not require Drs or Mtk.

For pathogens other than C. glabrata, the effect of deleting Bombardier is less severe than that of deleting the ten Bom genes clustered at 55C. Our mass spectrometry data suggest an explanation. Whereas, short-form Boms are absent from 1bbd hemolymph, bicipital Boms are present. (Tailed Boms were not detected with either mass spectrometry method.) Therefore, we postulate that the bicipital Boms, which are not required for resistance to C. glabrata (25), are functional against other pathogens. This would explain why 1bbd flies are more resistant than Bom155<sup>C</sup> flies upon infection with E. faecalis or F. oxysporum (**Figure 1**). In this regard, we note that Bombardier and all three forms of Bom proteins—short, tailed, and bicipital are found across the Drosophila genus, supporting the notion that all three classes of Boms are immunoprotective and therefore maintained across the Drosophila genus.

It might appear that our discovery of Bombardier was serendipitous, given our role in defining the Bomanin gene family (24, 25). In hindsight, however, the link was forged in our approach. We selected CG18067 from the most strongly inducible Toll-regulated loci, a group that also includes eight of the Bomanin genes. Next, we engineered a CRISPR/Cas9 deletion of CG18067 and assayed this knockout with the identical set of pathogens that we had used for the Bom155<sup>C</sup> deletion, screening for loss of survival upon infection. Having examined a gene that is as strongly induced as the Bomanins, present in the same range of species as the Bomanins, and with a spectrum of loss-of-function phenotypes similar to that of the Bomanins, it is not particularly surprising that we would find ourselves studying a gene that affects the Bomanins.

#### Bombardier Function and Structure

What is the function of Bombardier? Deleting the gene results in the absence of short-form Boms from hemolymph, an effect

we find is at the level of protein. Other mature immune peptides are present at normal levels in the hemolymph, and there is thus no general defect in translation, secretion, or processing. Based on these findings, we propose that Bombardier normally functions either to chaperone short Boms as they are secreted from the fat body into the hemolymph or, alternatively, to protect the Boms from misfolding or aggregation while in the hemolymph. We further hypothesize that it is the ectopic localization or aberrant form of short-form Boms in 1bbd flies that generates morbidity upon Toll pathway activation. In support of this idea, we showed that Bom expression underlies the lethality observed in 1bbd flies (**Figure 7**). Whether the short-form Boms physically interact with Bombardier, perhaps in the context of a larger antimicrobial complex, is currently unknown.

Activation of Toll-like receptor (TLR) signaling is important for innate immunity, but induction of the pathway can lead to autoimmune disorders and chronic inflammatory disease (45–48). Here we report an autoimmune activity driven by Toll-induced Bom expression in flies lacking a downstream pathway component, Bombardier. To what extent this parallel can be exploited in the context of understanding autoimmune disorders promises to be a significant focus for future investigation.

#### DATA AVAILABILITY STATEMENT

The LC-MS datasets generated for this study can be accessed on the MassIVE data repository using the accession identifier: MSV000084509.

#### AUTHOR CONTRIBUTIONS

SL and SW: conceptualization, writing—original draft, and visualization. SL, AF, and EB: methodology. SL and AF: formal analysis. SL, AF, and LC: investigation. SL, AF, LC, EB, and SW: writing—review and editing. EB and SW: supervision and funding acquisition.

#### FUNDING

This work was supported by National Institutes of Health (NIH) grant R01 GM050545 (to SW), NIH grant GM119132 (to EB), and NIH predoctoral training grant T32 GM008666 (to SL). The funders had no role in study design, data collection and analysis, the decision to publish, or preparation of the manuscript.

#### REFERENCES


#### ACKNOWLEDGMENTS

We thank Scott Lindsay for a wealth of helpful discussions and technical assistance; Andres Mauricio Caraballo-Rodriguez and the Pieter Dorrestein Lab for guidance on and access to MALDI-TOF mass spectrometry equipment; Scott Rifkin for advice on statistical approaches; Emily Troemel, Bill McGinnis, Scott Lindsay, and Roland Liu for comments on the manuscript; and the laboratories of Bill McGinnis, Valentino Gantz, Ethan Bier, and James Posakony for access to equipment.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.03040/full#supplementary-material


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

Copyright © 2020 Lin, Fulzele, Cohen, Bennett and Wasserman. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The Daisho Peptides Mediate *Drosophila* Defense Against a Subset of Filamentous Fungi

Lianne B. Cohen, Scott A. Lindsay, Yangyang Xu, Samuel J. H. Lin and Steven A. Wasserman\*

Section of Cell and Developmental Biology, Division of Biological Sciences, University of California, San Diego, La Jolla, CA, United States

Fungal infections, widespread throughout the world, affect a broad range of life forms, including agriculturally relevant plants, humans, and insects. In defending against fungal infections, the fruit fly Drosophila melanogaster employs the Toll pathway to induce a large number of immune peptides. Some have been investigated, such as the antimicrobial peptides (AMPs) and Bomanins (Boms); many, however, remain uncharacterized. Here, we examine the role in innate immunity of two related peptides, Daisho1 and Daisho2 (formerly IM4 and IM14, respectively), found in hemolymph following Toll pathway activation. By generating a CRISPR/Cas9 knockout of both genes, 1daisho, we find that the Daisho peptides are required for defense against a subset of filamentous fungi, including Fusarium oxysporum, but not other Toll-inducible pathogens, such as Enterococcus faecalis and Candida glabrata. Analysis of null alleles and transgenes revealed that the two daisho genes are each required for defense, although their functions partially overlap. Generating and assaying a genomic epitope-tagged Daisho2 construct, we detected interaction in vitro of Daisho2 peptide in hemolymph with the hyphae of F. oxysporum. Together, these results identify the Daisho peptides as a new class of innate immune effectors with humoral activity against a select set of filamentous fungi.

#### *Edited by:*

Dan Hultmark, Umeå University, Sweden

#### *Reviewed by:*

Ulrich Theopold, Stockholm University, Sweden Hai-peng Liu, Xiamen University, China

#### *\*Correspondence:*

Steven A. Wasserman stevenw@ucsd.edu

#### *Specialty section:*

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

*Received:* 18 October 2019 *Accepted:* 06 January 2020 *Published:* 23 January 2020

#### *Citation:*

Cohen LB, Lindsay SA, Xu Y, Lin SJH and Wasserman SA (2020) The Daisho Peptides Mediate Drosophila Defense Against a Subset of Filamentous Fungi. Front. Immunol. 11:9. doi: 10.3389/fimmu.2020.00009 Keywords: innate immunity, toll, *Drosophila*, humoral, antifungal

#### INTRODUCTION

Fungal infections have a devastating impact on a wide range of organisms. They are destructive to agricultural plants around the world, including rice, wheat, and tomatoes (1). Additionally, fungi infect more than one million humans annually (2). Existing antifungal treatments are limited, with only one new class of drugs, echinocandins, developed in the past 15 years. Furthermore, extensive usage of limited classes of related antifungals has led to the increasingly frequent appearance of drug-resistant fungi (2). An enhanced understanding of naturally occurring antifungal defenses is thus of tremendous potential benefit.

The fruit fly Drosophila melanogaster is a robust model for fungal infections, replicating many features of murine fungal infections (3, 4). In the wild, flies have been found to be infected with a number of filamentous fungi, including Beauveria, Metarhizium, and Fusarium species (5, 6). In combatting these infections, flies rely on the Toll innate immune pathway (7, 8). Toll provides defense against not only filamentous fungi, but also yeasts and those Gram-positive bacteria that produce a cell wall containing Lys-type peptidoglycan (8–10). A second innate immune pathway,

**68**

defined by the Imd receptor, provides defense against Gramnegative bacteria and the limited number of Gram-positive bacteria that produce a cell wall containing DAP-type peptidoglycan (11, 12).

Systemic activation of Toll signaling induces a broad set of genes first identified by microarray analysis and mass spectroscopy (13–16). Many of the induced innate immune genes are transcribed in the fly fat body, with the protein products secreted into the hemolymph. These include antimicrobial peptides (AMPs), the Bomanin peptides, and a number of uncharacterized peptides.

Although AMPs, such as the antifungal peptide Drosomycin (Drs) directly kill pathogens in vitro (17, 18) and are immunoprotective when ectopically expressed in vivo (19), recent loss-of-function studies reveal little or no requirement for AMPs in defense against fungi and Gram-positive bacteria (20). In contrast, the Bomanin family of peptides (Boms) are required for defense against both classes of pathogens (21). Boms, which are Drosophila-specific, are readily detected in hemolymph following Toll activation. Here we describe the functional characterization of additional immune effectors, the Daisho peptides, which appear in hemolymph following systemic infection and are required for defense against a subset of filamentous fungi.

### MATERIALS AND METHODS

#### Fly Husbandry and Strain Generation

Flies were raised at 25◦C on cornmeal molasses agar media<sup>1</sup> . The w <sup>1118</sup> strain was used as the wild type. MyD88<sup>−</sup> flies were MyD88kra<sup>1</sup> , and imd<sup>−</sup> flies were imdshadok .

As described in Results, the genes for the immune induced peptides IM4 and IM14 have been given the designations daisho1 and daisho2, respectively. The null allele 1daisho, deleting both genes, as well as the individual gene deletions, 1dso1 and 1dso2, were generated using CRISPR/Cas9 technology, applying methods described previously (22). Pairs of guide RNAs that targeted Cas9 to delete the region 2R: 20,868,460–20,870,480 for 1daisho), 2R: 20,868,783–20,869,392 for 1dso1, and 2R: 20,870,332–20,870,728 for 1dso2 were cloned into the pU6- BbsI-chiRNA vector (Addgene plasmid # 45946). Homology arms of ∼1 kb were cloned into pHD-DsRed (Addgene plasmid # 51434). Cas9 was provided by plasmid pBS-Hsp70-Cas9 (Addgene plasmid #46294). Constructs were based on target sequences in the w <sup>1118</sup> strain and injected into w <sup>1118</sup>. See **Table S1** for primer sequences.

The FLAG epitope tag was cloned between the signal sequence and mature peptide of Dso2 in the context of the pHD-DsRed homologous repair template. This FLAG-Dso2 construct was introduced at the dso2 genomic locus using the 1dso2 guide RNAs.

Plasmids expressing dso1 or dso2 transcripts from the pBomS3 promoter were made using methods previously described (23). Briefly, the BomS3 gene promoter was placed 5′ to the ORF encoding either Dso1 or Dso2. These constructs were then each integrated via ΦC31-mediated transgenesis at an attP landing site located at 86Fb on the D. melanogaster third chromosome (BDSC stock #24749). The transgenes were crossed into the 1dso1 and 1dso2 backgrounds and homozygous stocks were derived. An empty vector control was also introduced at the 86Fb attP landing site.

#### Microbial Cultures

For survival experiments, microbes were cultured as follows. Enterococcus faecalis NCTC 775 (ATCC 19433) and Enterobacter cloacae were grown overnight at 37◦C in LB media and concentrated to an OD<sup>600</sup> of 10 in 20% glycerol. Candida glabrata CBS 138 [ATCC 2001] was grown overnight in YPD media at 37◦C and concentrated to an OD<sup>600</sup> of 100 in PBS, 0.1% Tween. All filamentous fungi were grown on malt extract agar plates at 29◦C until sporulation was observed (10–15 days). Fungal material was then strained through glass wool with sterile water to collect spores, which were concentrated in 20% glycerol and stored at −80◦C before being used at the following concentrations (in spores/ml): Aspergillus flavus (sequenced strain): 5 × 10<sup>9</sup> ; A. fumigatus AF293 (FGSC# A1100): 6 × 10<sup>9</sup> ; A. parasiticus Nor-1 mutant (NRRL #6111): 3 × 10<sup>9</sup> ; Botrytis cinerea (B05.10): 3 × 10<sup>9</sup> ; Fusarium graminearum (NRRL #5883): 8 × 10<sup>8</sup> ; F. oxysporum f. sp. lycopersici 4287 (FGSC #9935): 3 × 10<sup>8</sup> ; F. verticillioides (FGSC #7415): 3 × 10<sup>9</sup> ; Neurospora crassa: 1 × 10<sup>9</sup> .

For the induction of the Toll response, heat-killed Micrococcus luteus was prepared as previously described (23).

#### Survival Assays

Groups of 20–25 adult male flies aged 2–7 days were collected and stabbed with a needle dipped in a suspension of bacteria, yeast, or fungal spores. Where needed, MyD88<sup>−</sup> or Bom155<sup>C</sup> flies were used as controls immunodeficient for the Toll-mediated response. Flies infected with E. faecalis were incubated at 25◦C; all other infected flies were incubated at 29◦C. Fly deaths were recorded at least twice per day for the duration of each experiment. Any deaths that occurred within the first 6 h were set aside to exclude from the data any deaths due to traumatic injury. The experiment was repeated three times and results combined. Statistical analyses were performed using the Gehan-Breslow-Wilcoxon test.

#### MALDI-TOF

After Toll induction with heat-killed M. luteus, flies were incubated at 29◦C for 24 h, after which hemolymph was collected via capillary as previously described (23). Hemolymph in 0.1% trifluoroacetic acid/50% acetonitrile was mixed 1:1 with Universal Matrix (Sigma-Aldrich). Samples were then dried onto a Bruker MSP 96 ground steel plate. Spectra were collected from 1,500 to 10,000 m/z for linear mode, and 1,000– 5,000 m/z for reflectron mode, both with positive polarization. Peptide calibration standard II (Bruker) was used as an external calibration standard. For each genotype, at least five independent samples were collected. Representative spectra are shown. Peaks were identified via corresponding m/z values from previous studies (13, 16). Spectra were visualized using R 3.3.2 and ggplot2 2.2.1 (24, 25).

<sup>1</sup>http://blogs.cornell.edu/drosophila/dssc-cornmeal-recipe/

#### Quantitation of Pathogen Load

Pathogen load in infected flies was measured by qRT-PCR of fungal RNA (26, 27). Adult male flies, 2–7 days old, were stabbed with a needle dipped in F. verticillioides at 3 × 10<sup>9</sup> spores/ml. Flies were then incubated at 29◦C. Groups of 5–6 flies were collected at the stated times and frozen in liquid nitrogen. Total RNA was isolated with TRIzol (Ambion) and cDNA was made via SuperScript RT II (Invitrogen). EF1A was selected as a proxy gene for fungal load based on its stable expression (28). Measurements by qRT-PCR were performed on the iQ5 cycler (BioRad) with iQ SYBR Green Supermix (BioRad) using the primers listed below. Values were normalized to fly mRNA based on expression of the rp49 gene.

Primers: Fv\_EF1A\_F1: GGCTTTCACTGACTACCCTCC TCT, Fv\_EF1A\_R1: ACTTCTCGACGGCCTTGATGACAC, rp49\_F1: CAAGGGTATCGACAACAG, rp49\_R1: CTTGTT CGATCCGTAACC.

#### Peptide Gel Electrophoresis and Immunoblotting

Hemolymph samples were collected via the Zymo-Spin IC column method (23) from 30 male flies aged 2–7 days that had been induced with heat-killed M. luteus and incubated for 24 h at 29◦C. Samples were run on a SDS-tricine, 18% separating/10% spacer/4% stacking, acrylamide gel<sup>2</sup> . Protein samples were then transferred to a PDVF membrane, blocked with 5% milk in TBST and stained with primary α-FLAG M2 (Sigma) (1:500) and secondary sheep α-mouse HRP (Amersham Biosciences) (1:1,000). The immunoblot was then treated with West Pico PLUS substrate (Thermo Scientific) and exposed to film.

#### Peptide Hyphal Binding and Immunofluorescence

The immunostaining protocol was adapted from Luo et al. (29). F. oxysporum was grown in 5 ml malt extract broth from a starting concentration of 2.9 × 10<sup>5</sup> spores/ml. After overnight shaking at room temperature, fungal hyphae were collected by centrifugation at 1,000 g for 10 min and resuspended in PBS. Hemolymph was collected via the Zymo-Spin IC column method (23) from 420 male flies that had been induced with heat-killed M. luteus 24 h prior and incubated at 29◦C, yielding ∼35 µl cell-free hemolymph. Next, aliquots of 200 µl hyphae and 35 µl hemolymph were shaken at room temperature for 30 min. The samples were washed three times with PBS before fixation with 4% formaldehyde for 1 h. After washing another three times with PBS, samples were blocked for 1 h with 5% BSA. Samples were then incubated with α-FLAG antibody (1:200) overnight at 4◦C. After washing with PBS, samples were stained for 2 h with donkey α-mouse Alexa555 (1:400) and DAPI (1:200) and then washed and mounted on slides. Samples were imaged with a Ti2 Widefield microscope (Nikon) and analyzed with the NISelements software and OMERO.

#### RESULTS

#### Generation of Flies Null for the *daisho* Gene Pair

Pioneering mass spectrometry experiments by Bulet et al. identified two dozen peptide IMs (immune-induced molecules) that accumulate in Drosophila hemolymph upon induction of the innate immune response, principally the Toll pathway (13, 16). Among these, the Bomanins have been found to play an essential role against a broad range of pathogens (21, 23) while several, including the 15 aa long IM4 and 24 aa long IM14, have unknown functions. Based on our demonstration of defensive functions for these peptides, we have renamed them Daisho1 and Daisho2, for daisho, the Japanese term for a matched pair of samurai swords, one short and one long.

The Daisho peptides are closely related to one another and occupy adjacent positions in the genome, where they are divergently transcribed (**Figure S1**). As shown in **Figure 1A**, the sequence of amidated mature Daisho1 (Dso1) has 67% identity with the corresponding region of the mature Daisho2 (Dso2) peptide. Like the Bomanins, the daisho genes are widespread among the Drosophila genus, but not identified elsewhere. To investigate the potential role of the daisho genes in innate immunity, we used CRISPR/Cas9 technology to delete both genes. The 2.0 kb deleted region includes the entire dso1 gene, the upstream region for both genes, and the first exon of dso2 (including the start codon). Flies homozygous for the 1daisho1,2 deletion, hereafter 1daisho, were viable and fertile.

With the 1daisho stock in hand, we carried out MALDI-TOF studies of hemolymph (**Figures 1B–E**). As described above, following Toll activation wild-type hemolymph displays robust expression of immune peptides, including the Daisho peptides, Bomanins, and AMPs. The signals from Dso1 and Dso2 were ablated in 1daisho, as evidenced by the loss of signal at 1,722 mass/charge (m/z) (Dso1) and 2,694 m/z (Dso2). Furthermore, the spectra of induced 1daisho hemolymph was wild-type for all previously identified peaks other than Dso1 and Dso2, including the Bomanins and AMPs, Metchnikowin (Mtk), and Drosomycin (Drs). The absence of Dso1 and Dso2 thus did not detectably alter the accumulation or modification of other Toll-induced peptides in the hemolymph.

In addition to previously identified peaks, 1daisho hemolymph contained one previously unseen signal. The 1,724 m/z signal of this peak, readily apparent in reflectron mode, is identical to that predicted for the BomS5 amidated peptide, previously known as CG15065 (**Figure 1E**). This signal had not been detected previously because in the wild type it lies in the shoulder of the robust Dso1(IM4) peak. Its existence in Toll-induced hemolymph was expected, however, on the basis of microarray and RNAseq data demonstrating strong Toll-activated induction of the BomS5 locus (15, 30).

#### The *daisho* Genes Are Specifically Required for Defense Against *F. oxysporum*

We next turned to a functional assay to determine whether the absence of the Daisho peptides impaired survival following systemic infection. Because the Toll pathway responds to and

<sup>2</sup>https://molbio.mgh.harvard.edu/szostakweb/protocols/protein\_page/index.html

protects against infection by many Gram-positive bacteria and fungi, we focused on these classes of pathogens. We stabbed adult flies with a needle dipped in a suspension of bacteria, yeast, or fungal spores and then monitored survival. We used w <sup>1118</sup> flies as our wild-type, i.e., immunocompetent, control and Bom155<sup>C</sup> flies, which lack the 10-gene Bom cluster, as an immunodeficient control (21).

For a number of the pathogens tested, 1daisho flies behaved identically to the wild type. Roughly 50% of both wild-type and 1daisho flies survived 6 or more days following infection with the Gram-positive bacteria Enterococcus faecalis, whereas 100% of Bom155<sup>C</sup> flies died within 2 days (**Figure 2A**). Likewise, wildtype and 1daisho flies survived a week or longer after infection with the yeast Candida glabrata, whereas Bom155<sup>C</sup> flies died in 4 days or fewer (**Figure 2B**). We also found no effect of 1daisho on immune defenses mediated by the Imd pathway: wild-type, 1daisho, and Bom155<sup>C</sup> flies all survived infection with the Gramnegative bacteria Enterobacter cloacae, whereas, control imd- flies died within 1 day (**Figure 2C**).

For one pathogen in the initial test set, the filamentous fungus Fusarium oxysporum, deletion of the daisho genes had a marked effect on survival (**Figure 2D**): 50% of flies homozygous for 1daisho died within 4 days of infection. In contrast, <70% of wild-type flies survived 7 or more days post-infection. Thus, loss of the Daisho peptides disrupts defense against F. oxysporum, but not other tested pathogens. Interestingly, loss of the Daisho peptides did not impact survival peptides did not impact survival as severely as did loss of the Boms, which led to 50% death after 2 days, very similar to complete loss of Toll signaling (21).

### 1*daisho* Flies Are Susceptible to Some but Not All Filamentous Fungi

We next investigated whether the susceptibility of 1daisho flies to F. oxysporum reflected a general susceptibility to filamentous fungi. For these studies, we focused on filamentous fungi for which flies deficient for Toll signaling, and thus for induction of Daisho1, Daisho2, and other Toll effectors, exhibit a significantly decreased survival relative to wild type (**Figure 3**). The control fly strains in each case were w <sup>1118</sup> (wild type) and kra-1 (MyD88−), a loss-of-function allele for an essential mediator of Toll signaling (31).

As shown in **Figure 3**, susceptibility of 1daisho flies to the filamentous fungi species varied. Survival was significantly less than wild-type for F. verticillioides and F. graminearum (panels A, B), two Fusarium species closely related to F. oxysporum. In the case of F. graminearum, survival of 1daisho flies was intermediate between that of wild-type and MyD88<sup>−</sup> flies, a pattern very similar to that observed with F. oxysporum, where 1daisho survival falls between wild type and Bom155C, which behaves similarly to MyD88<sup>−</sup> (21). In contrast, 1daisho flies displayed a much greater immune impairment upon infection with F. verticillioides than with F. oxysporum, dying to a comparable extent and at a similar rate as the MyD88<sup>−</sup> control (compare **Figures 2D**, **3A**).

Variation in survival was also observed among Aspergillus species. The survival curves of 1daisho infected with either A. parasiticus or A. flavus largely tracked with MyD88<sup>−</sup> (panels C, D). Upon A. fumigatus infection, however, 1daisho flies survived at least twice as long as MyD88<sup>−</sup> flies (**Figure 3E**).

For some filamentous fungi, loss of Daisho1 and Daisho2 did not affect survival. For example, 80% of wild-type and 1daisho flies survived for at least 7 days after infection with Botrytis cinerea, whereas >50% of MyD88<sup>−</sup> flies died after 2 days (**Figure 3F**). Likewise, wild-type and 1daisho flies survived Neurospora crassa infection for 6 days or more, but 50% of MyD88<sup>−</sup> flies died after 3 days (**Figure 3G**). Overall, we find that the Daisho peptides play a vital role in survival after infection with certain species of filamentous fungi, but are not important for infections with others.

#### *daisho1* and *daisho2* Are Each Required for Defense

Daisho1 and Daisho2 are highly similar in sequence and expression pattern. Are they functionally redundant? To address this question, we explored the function of each individual locus. We again used CRISPR/Cas9, generating deletions that removed the entire coding sequence for either daisho1 or daisho2. The 5' endpoints of each deletion were chosen to lie within 100 bp of the transcriptional start site, minimizing potential disruption of elements in the regulatory region separating the two genes (**Figure S1**). For both deletions, MALDI-TOF analysis of induced hemolymph confirmed loss of the deleted gene product but no other peptides, indicating that either Daisho1 or Daisho2 can be stably expressed in the absence of the other (**Figure 4**).

To test the effect on defense of deleting dso1 or dso2, we stabbed adults with F. verticillioides spores, for which 1daisho flies have a reduced survival. Deleting either the dso1 or dso2 gene resulted in susceptibility to F. verticillioides markedly different from wild-type and comparable to that of the double deletion (**Figure 5**). Thus, Daisho1 and Daisho2 each act in defense against F. verticillioides infection.

Since deletion of either dso1 or dso2 had as severe an effect on survival as the double mutant, it was possible that each gene has a specific and distinct function in antifungal defense. Alternatively, survival might depend only on total dosage for the two genes, with loss of either dropping expression below the threshold required. To distinguish between these models, we generated transgenes placing each ORF under control of pBomS3, shown previously to be strongly Toll-responsive promoter (23), and then

highlighting loss of Dso1 (A) and Dso2 (B) in deletion mutants.

assayed the transgenes for rescue of 1dso1 or 1dso2. As shown in **Table 1**, pBomS3-driven dso1 rescued 1dso1, improving the median survival from 46 to 93 h (p < 0.0001). The same was true of pBomS3-driven dso2 in the 1dso2 background (p < 0.0001) (see **Figure S2** for full survival curves). Flies expressing the empty vector construct at the same chromosomal location did not show any increase in survival (**Figure S3**).

Having confirmed the activity of the two constructs, we expressed each in a background deficient for the other. dso2 expression significantly improved survival of 1dso1 flies,


TABLE 1 | Median survival in hours of dso1 and dso2 deletion mutations rescued by homotypic and heterotypic transgenes.

Data derived from *Figure S2*. n.a., not applicable.

increasing median survival from 46 to 78 h (p < 0.0001). Similarly, dso1 expressed in a 1dso2 background improved median survival from 46 to 55 h (p = 0.0005). Nevertheless, rescue was incomplete. The median survival of dso2 expressed in 1dso1 background (78 h) did not reach median survival of 1dso1 rescued with dso1 (93 h) (n.s., p = 0.09). Furthermore, dso1 did not rescue survival of 1dso2 (55 h) to the same level as dso2 (93 h) (p < 0.0001). The data thus indicate that the two loci encode functions that are neither fully distinct nor fully redundant.

#### Deleting *daisho1* and *daisho2* Results in an Elevated Pathogen Load in Infected Flies

To investigate whether Daisho1 and Daisho2 affect pathogen growth during infection, we measured fungal load after infection in 1daisho and wild-type flies. After stabbing adult males with F. verticillioides, groups of 5–6 infected flies were collected and RNA was extracted. Fungal EF1A transcript levels were measured as a proxy for pathogen load and normalized to the fly reference gene rp49. Directly after infection (2 h), there was no significant difference between 1daisho and wild-type flies by Mann-Whitney U test (**Figure S4**). By the next day, however, 1daisho flies had a pathogen load roughly 10-fold greater than wild-type (p = 0.0317).

#### FLAG-Dso2 Binds to *F. oxysporum* Hyphae

We next tagged Dso2, the larger of the two peptides, with the FLAG epitope, using CRISPR/Cas9 to introduce the tag at the amino-terminus of the endogenously expressed mature peptide. Immunoblot analysis of induced hemolymph from FLAG-Dso2 flies revealed a single band detectable with α-FLAG antibody (**Figure 6A**). MALDI-TOF analysis of hemolymph confirmed the loss of the Dso2 peak at 2,694 m/z and the appearance of a peak with an m/z ratio of 3,689, the value expected for FLAG-Dso2 (**Figure 6B**).

Having confirmed that FLAG-Dso2 peptide is stably expressed, we next assayed its activity in providing antifungal defense. Specifically, flies homozygous for FLAG-dso2 at the dso2 locus were infected with F. verticillioides and their survival was compared to both wild-type flies and 1dso2 flies. Survival of FLAG-Dso2 flies was not wild-type, but was significantly better than that of flies lacking the dso2 gene (**Figure S5**). We conclude that the FLAG-Dso2 peptide is active in providing defense against F. verticillioides infection.

Next, we assayed FLAG-tagged Dso2 peptide in hemolymph for its ability to bind fungus. We collected hemolymph from Toll induced flies, incubated it with hyphae from F. oxysporum, and fixed samples. The majority (>80%) of F. oxysporum hyphae had no visible signal when stained with α-FLAG antibody (**Figure 7A**). Among the remaining hyphae, we observed a variety of staining patterns, including, but not limited to, signals concentrated in the regions between nuclei (**Figure 7B**), extending across greater fractions of hyphae (**Figure 7C**) or spanning the length of hyphae (**Figure 7D**). In parallel experiments with untagged wild-type hemolymph, no signal was detected (**Figures 7E–H**). We conclude that Daisho2 peptide in hemolymph can bind to F. oxysporum hyphae.

In summary, our results demonstrate that the pair of immuneinduced peptides, Daisho1 and Daisho2, mediate Toll-induced defense against specific filamentous fungi, most likely via a humoral effect on fungal hyphae.

#### DISCUSSION

#### Role of the Daisho Peptides in Antifungal Defense

In this study we found that the related peptides Daisho1 and Daisho2 are required in D. melanogaster for defense against a subset of filamentous fungi. We have also demonstrated that the two peptides have partially overlapping functions. Survival data reveal a dependence on the overall level of Dso1 and Dso2, with each peptide able to partially compensate for the absence of the other. Furthermore, each peptide accumulates in the absence of the other.

The Daisho peptides lack known motifs of defined function. As noted previously (21), there is a similarity in size and sequence between Dso1 and Dso2 and the Bomanin peptides. There are, however noteworthy differences, including the presence of a CxxC motif in the Bomanins and the broader requirement for the Bomanins in Toll-mediated defense.

Among those fungi for which deleting dso1 and dso2 decreases survival, 1daisho flies nevertheless often exhibit significantly greater survival than do MyD88<sup>−</sup> or Bom155<sup>C</sup> flies (see e.g., F. oxysporum and F. graminearum). Thus, in contrast to the Bom effectors, which are strictly required for Toll defenses against a broad range of pathogens, the Daisho peptides appear to be required for some, but not all Toll functions and to be active against only a select group of pathogens against which Toll mounts defense.

Like the Bomanins, dso1 and dso2 are found only within the Drosophila genus. Taxonomically-restricted genes (TRGs), while often studied only sparingly, represent 10–20% of most genomes and frequently have essential functions (32). TRGs have been identified in the immune pathways of many invertebrates, including flies, mosquitoes, and cnidarians. Within immune systems they are abundant among effectors, but rare among signal transduction factors (33, 34).

#### Specificity of *daisho* Genes in Antifungal Defense

In tracking survival following systemic infection, we find considerable variability with regard to which pathogens exhibit increased virulence toward D. melanogaster in the absence of both daisho genes. Categorizing the fungi against which the daisho genes provide defense, we detect no simple relationship to fungal phylogeny. For example, the daisho genes are required to defend against all the Fusarium species tested and some of the Aspergillus species, but not Neurospora crassa. Yet Fusarium and Neurospora are both members of the class Sordariomycetes, whereas Aspergillus is part of the less closely related Eurotiomycetes class (35, 36). Furthermore, 1daisho flies exhibit differential susceptibility to fungi within a single genus: the 1daisho deletion substantially decreases survival against A.

flavus and A. parasiticus, but has a much smaller effect on survival following A. fumigatus infection.

Although susceptibility of 1daisho flies does not track simply with fungal phylogeny, susceptibility does appear to be closely related to fungal pathogenicity. Consider four filamentous fungi that are particularly pathogenic for wild-type flies: F. verticillioides, F. graminearum, A. flavus, and A. parasiticus. Infection with any of these four pathogens kills >50% of wild-type flies within 7 days. For each of these four, 1daisho greatly decreases survival. By comparison, consider filamentous fungi with low pathogenicity, e.g., A. fumigatus, N. crassa, and B. cinerea. For each, >80% of wild-type and 1daisho flies survive for 7 or more days post-infection. Note that we observe this association of susceptibility with pathogenicity only among filamentous fungi: for the strongly pathogenic Gram-positive bacterium E. faecalis, the 1daisho deletion had no effect on survival.

Although the Bomanins are strictly required for Toll humoral defenses, we have found a correlation between pathogenicity and the level of Bomanin function required to confer resistance (21). It thus appears that for both Bomanins and the daisho genes, pathogenicity tracks with the strength of effector function required for defense.

#### Activity of Daisho Peptides

How do the Daisho peptides provide defense against filamentous fungi? One mechanism could be directly binding and killing the pathogens. Consistent with this idea, we find a modest but significant increase in pathogen load in 1daisho flies. In addition, our immunofluorescence data demonstrate that Daisho2 can interact in vitro with at least one filamentous fungus that it targets. Antifungal peptides, such as mammalian LL-37 and plant defensin NaD1, also bind hyphae of fungal pathogens against which they are active (29, 37). The Daisho2 peptide's ability to bind fungal hyphae could indicate an antimicrobial function. Given that our assay was carried out with crude hemolymph, we cannot state whether the observed interaction of Daisho2 with hyphae is direct or is mediated by one or more unidentified hemolymph components.

The Daisho peptides might themselves interfere with pathogen growth, survival, or proliferation, or they might enable the fungicidal activity of other factors. The same is true of the Bomanins, which are required for hemolymph mediated killing of C. glabrata, but for which fungicidal activity of synthetic

#### REFERENCES


peptides has not been observed (23). Given that daisho genes are required for defense against only a subset of Toll and Bomanin targets, the function of Daisho1 and Daisho2 may be to meet a specific challenge posed by certain fungi to the entry or activity of antimicrobial factors. Further investigation of the Daisho peptides, as well as other hemolymph immune effectors, is likely to be informative in this regard.

#### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the article/**Supplementary Material**.

#### AUTHOR CONTRIBUTIONS

LC and SW conceived the project and wrote the paper. LC, SAL, YX, and SJHL performed and analyzed the experiments.

### FUNDING

This work was supported by National Institutes of Health (NIH) grant R01 GM050545 (to SW).

#### ACKNOWLEDGMENTS

We thank Bo Zhou, Beatriz Pinatti, and André Soares for the generation of the pU6-chiRNA constructs. We would also like to thank Philipp Weckwerth and Alisa Huffaker for the generous gift of F. graminearum, F. verticillioides, A. flavus, A. parasiticus, and B. cinerea. We thank Andrés Mauricio Caraballo Rodríguez and the Dorrestein lab for access to and assistance with MALDI-TOF. We also thank the McGinnis, Gantz, Posakony, and Bier labs for access to equipment. We thank Anita Hermann for help with immunofluorescence and imaging and Eric Griffis and Daphne Bindels from the Nikon Imaging Center at UC San Diego for microscopy expertise. We thank Matt Daugherty, Bill McGinnis, and Kirthi Reddy for thoughtful reading of the manuscript and helpful comments.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2020.00009/full#supplementary-material


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

Copyright © 2020 Cohen, Lindsay, Xu, Lin and Wasserman. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Host-Microbe-Pathogen Interactions: A Review of Vibrio cholerae Pathogenesis in Drosophila

Saeideh Davoodi and Edan Foley\*

Department of Medical Microbiology and Immunology, Faculty of Medicine and Dentistry, University of Alberta, Edmonton, AB, Canada

Most animals maintain mutually beneficial symbiotic relationships with their intestinal microbiota. Resident microbes in the gastrointestinal tract breakdown indigestible food, provide essential nutrients, and, act as a barrier against invading microbes, such as the enteric pathogen Vibrio cholerae. Over the last decades, our knowledge of V. cholerae pathogenesis, colonization, and transmission has increased tremendously. A number of animal models have been used to study how V. cholerae interacts with host-derived resources to support gastrointestinal colonization. Here, we review studies on host-microbe interactions and how infection with V. cholerae disrupts these interactions, with a focus on contributions from the Drosophila melanogaster model. We will discuss studies that highlight the connections between symbiont, host, and V. cholerae metabolism; crosstalk between V. cholerae and host microbes; and the impact of the host immune system on the lethality of V. cholerae infection. These studies suggest that V. cholerae modulates host immune-metabolic responses in the fly and improves Vibrio fitness through competition with intestinal microbes.

#### Edited by:

Susanna Valanne, Tampere University, Finland

#### Reviewed by:

Shoichiro Kurata, Tohoku University, Japan Ioannis Eleftherianos, George Washington University, United States Bruno Lemaitre, École Polytechnique Fédérale de Lausanne, Switzerland

> \*Correspondence: Edan Foley efoley@ualberta.ca

#### Specialty section:

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

Received: 31 October 2019 Accepted: 23 December 2019 Published: 24 January 2020

#### Citation:

Davoodi S and Foley E (2020) Host-Microbe-Pathogen Interactions: A Review of Vibrio cholerae Pathogenesis in Drosophila. Front. Immunol. 10:3128. doi: 10.3389/fimmu.2019.03128 Keywords: IMD, Drosophila melanogaster, proliferation, Vibrio cholerae, insulin, microbiome, metabolism, T6SS

### INTRODUCTION

#### Background

A complex set of interactions among host intestinal cells, and gut-resident microbes, impacts the viability of all participants. For example, commensal microbes consume intestinal nutrients, and generate metabolites that influence development, growth, metabolism, and immune system function in the host (1–8). Introduction of microbes with pathogenic potential to the gut lumen, or rearrangements to the composition or distribution of gut microbial communities, can have substantial impacts on intestinal homeostasis for the host (9). In particular, shifts in niche occupancy by gut bacteria, or alterations to metabolic outputs from the gut microbiome, can result in the development of severe intestinal disease (10–13). For example, Bacteroides thetaiotaomicron, a common human commensal, cleaves host glycans to produce fucose, a sugar that modulates the virulence of enterohemorrhagic Escherichia coli (14). Despite the importance of regulated molecular exchanges among host and microbial cells for host fitness and microbial function, our knowledge of pathogen-commensal interactions in the context of immune-metabolic regulation and intestinal disease is still quite limited. To fully understand such complex, multipartite interactions, it is essential that we deploy all relevant experimental systems at our disposal.

Drosophila melanogaster is a valuable experimental tool for studying host-microbe interactions. Lab-raised strains of Drosophila associate with a limited number of bacterial taxa (15–17), dominated by easily cultivated Acetobacter and Lactobacillus strains that are accessible to genetic manipulation, and deployment in large-scale screens. Researchers have access to simple protocols for the establishment of flies with a defined intestinal microbiome (18, 19), and there is an abundance of publicly available lines for the genetic manipulation of fly intestinal function. Combined, these advantages allowed researchers to make substantial breakthroughs in understanding how flies interact with intestinal bacteria (20). Importantly, given the extent to which genetic regulators of intestinal homeostasis are conserved between vertebrates and invertebrates (20, 21), discoveries made with the fly have the potential to illuminate foundational aspects of host-microbe interactions. However, there are several key differences to note between flies and vertebrates that partially limit the utility of the fly model. Specifically, flies lack lymphocyte-based adaptive defenses, and the fly microbiome is considerably different to that reported in vertebrates.

#### Antimicrobial Defenses in the Fly Intestine

Drosophila integrate physical, chemical, proliferative, and antibacterial strategies to neutralize intestinal microbes, and prevent systemic infection of the host (**Figure 1**) (22, 23). The chitinous peritrophic matrix lines the midgut, and presents a physical barrier against bacterial invasion (24), similar to the mucus lining of the vertebrate intestinal tract. The germlineencoded immune deficiency (IMD) antibacterial defense pathway, a signaling pathway similar to the mammalian Tumor Necrosis Factor pathway (25), detects bacterial diaminopimelic acid-type peptidoglycan, and acts through the NF-κB transcription factor family member, Relish, to induce expression of antimicrobial peptides (26–29). At the same time, Dual Oxidase (Duox) and NADPH Oxidase (Nox) protect the host from gut bacteria through the generation of bactericidal reactive oxygen species (30, 31). Evolutionarily conserved growth regulatory pathways respond to damage of epithelial cells by promoting a compensatory growth and differentiation of intestinal stem cells (ISCs) in infected flies (32–35). This adaptive repair mechanism maintains the epithelial barrier, and prevents systemic infection of the host. Combined, these antibacterial defenses protect the host from infection, and maintain beneficial relationships between the fly and their gut microbiome.

#### The Drosophila Microbiome

The fly microbiome is transmitted horizontally through the deposition of bacteria on the outer surface of freshly laid embryos, and is maintained through the ingestion of food contaminated with bacteria (36). Gut bacteria regulate Drosophila intestinal homeostasis by affecting metabolism, growth, and immunity in the host. Interactions between the host and gut microbiota have been extensively covered in several recent reviews (20, 37–39), and will not be discussed in detail here. In brief, detailed studies have uncovered roles for symbiotic Lactobacillus and Acetobacter species in the control of fly metabolism and growth (40–42). For example, a dehydrogenase activity in Acetobacter pomorum, produces acetic acid that regulates insulin signaling, carbohydrate, and lipid levels in the

reactive oxygen species (ROS) generated by NADPH oxidases (NOX) also

contribute to bacterial killing while cues from the bacterial microbiome promote the growth of intestinal progenitor cells (IPC), composes of intestinal stem cells (ISC), and enteroblasts (EB).

host (40). In addition to the effects of individual symbionts on nutrient allocations in the host, interactions among bacterial communities have significant effects on host metabolism, growth, and physiology (43–47). Vibrio cholerae (V. cholerae) has emerged as a particularly useful tool to study interactions between the host, the intestinal microbiota, and an enteric pathogen. A pioneering study in 2005 established that flies are susceptible to oral infection with V. cholerae, dying within a few day from a diarrheal disease with symptoms similar to cholera in humans (48). The genetic tractability of the fly and V. cholerae established this system as a very attractive model to identify key host and microbial determinants of pathogenesis. In the following years, a number of studies uncovered complex roles for metabolism, host immunity, epithelial growth, and microbial antagonism in the outcome of V. cholerae pathogenesis in the fly. In this review, we will discuss key findings from these studies, and outline what they tell us about host-microbe interactions in general, and V. cholerae-mediated pathogenesis in particular.

### Vibrio cholerae: Pandemics and Pathogenicity

Vibrio cholerae is a curved, Gram-negative member of the Vibrionaceae family of Proteobacteria (49). It inhabits aquatic environments, and copepods and chironomids are reported as natural reservoirs in marine ecosystems (50, 51). Intestinal colonization by V. cholerae causes the diarrheal disease, cholera, and is considered a substantial public health threat, especially in countries with poor sanitation and contaminated water (52). The first cholera pandemic emerged in 1817, with an expansion of cholera beyond the Indian subcontinent (53). Since then, the world has witnessed an additional six pandemics, with the seventh pandemic ongoing (54). Models that estimate cholera burden predict ∼3 million cases of disease per year, resulting in roughly 100,000 deaths (55).

Vibrio cholerae strains are divided into classical and non-classical serotypes, with classical ones expressing the O1 antigen on their surface (56, 57). Classical serotypes are further subdivided into two biotypes—classical and El Tor—that differ in the expression of a number of markers, such as hemolysins (58–61). The outbreak of epidemic cholera that spread through southeast Asia in 1992 is caused by the non-classical strain of V. cholerae 0139 (62), whereas the ongoing pandemic that originated in Indonesia in 1961 is caused by the El Tor biotype (63). El Tor causes a milder cholera disease (64), with infected individuals frequently remaining asymptomatic early in infection (65).

Vibrio cholerae encodes several virulence factors that regulate survival, colonization, and pathogenicity (66–69). Cholera toxin (CT) is a hexameric adenosine diphosphate-ribosyl transferase that contains one A subunit surrounded by five B subunits (70, 71). Upon release into the intestinal lumen via a type two secretion system (72), the B pentamer of CT interacts with host GM1 ganglosides (73), permitting toxin endocytosis, and a subsequent cytosolic release of the A1 subunit (74). A1 ADP-ribosylates the Gs alpha subunit, locking Gα<sup>s</sup> in an active state (75). Active Gα<sup>s</sup> elevates adenylate cyclase activity, greatly increasing levels of 3′ ,5′ -cyclic AMP, resulting in excess protein kinase A (PKA) activity (76). PKA stimulates an efflux of chloride ions through the cystic fibrosis transmembrane conductance regulator channel (77), leading to an uncontrolled flow of water, sodium and potassium ions into the intestinal lumen. This extreme, and rapid, dehydration results in the voluminous ricewater diarrhea that hallmarks cholera disease (78). In addition to CT, V. cholerae require the toxin co-regulated pilus virulence factor for pathogenesis (79). Toxin co-regulated pilus is a type IV pilus system that mediates colonization of the small intestine by a self-associate mechanism that supports the formation of bacterial microcolonies (80). Toxin co-regulated pilus also serves as the receptor for the CTXϕ bacteriophage. CTXϕ encodes ctxAB, and converts benign V. cholerae to pathogenic strains. The ability to synthesize toxin co-regulated pilus is advantageous for V. cholerae in aquatic environments, as it improves V. cholerae fitness by facilitating interbacterial interactions during colonization of host chitinaceous surfaces (81).

Although fluid replacement through oral rehydration solutions, antibiotic therapy, and vaccines are effective treatment options for patients with cholera, increased rates of antibiotic resistance among classical (82) and non-classical (83) strain of V. cholerae complicate treatment of the disease. Therefore, new antibacterial strategies that effectively target V. cholerae virulence factors are critical to contain this deadly disease. Over the last century, a variety of animal models that include rabbits, mice, fish, and flies, have been used to study Vibrio-host interactions and each of these models have added to our understanding of virulence, host responses to infection, interactions between Vibrio and host microbes, and cholera vaccine development.

#### The Rabbit Model

The first animal model to study V. cholerae dates back to 1884 when Nicati and Rietschin inoculated V. cholerae into the duodenum of guinea pigs, resulting in cholera-like symptoms (84). Since then, both infant and adult rabbit models of cholera have been widely used by researchers (85–87). As adult rabbits are resistant to oral infection with V. cholerae, the pathogen is typically introduced to the animal by ligated ileal loop surgery. In this technique, the small intestine of the rabbit is sealed at two ends, and the pathogen is delivered by injection into the ligated loop, allowing direct measurement of intestinal fluid secretion (88). The rabbit model has been very instructive for understanding V. cholerae distribution in the small intestine during infection, the importance of the mucosal barrier to prevent systemic infection of V. cholerae, and mechanisms of V. cholerae attachment to the intestinal epithelium (89, 90). As infant rabbits are capable of developing toxin co-regulated pilus-dependent cholera, they have been useful to study the reactogenicity associated with developing live attenuated V. cholerae vaccines as well (91). However, despite these advances, Vibrio pathogenesis studies using the rabbit ligated ileal loop model are labor-intensive, and do not replicate the normal route of infection. An alternative, oro-gastric infection model with infant rabbits pre-treated with the stomach acid production inhibitor, Cimetidine, allows oral infection and provides a valuable adult mammal model that circumvents needs for surgical interventions (87).

#### The Mouse Model

The infant or suckling mouse in commonly used to study V. cholerae pathogenesis (92). In this model, infant mice are infected via the oro-gastric route. In the infant mouse, the intestinal microbiome has not fully developed, allowing V. cholerae to colonize the host with diminished colonization resistance from commensal microbes. Studies working with infant mice have uncovered essential virulence factors of V. cholerae. For example, the toxin co-regulated pilus (93), and ToxR (69), which regulates toxin co-regulated pilus expression were originally characterized in the suckling mouse model. The adult mouse model was also a significant contributor to understanding the mechanisms of V. cholerae pathogenesis using accessory toxins such as hemolysin, hemagglutinin/protease, and multifunctional auto-processing RTX toxin (94). These observations were important to understand the ability of V. cholerae to express toxins other than CT to prolong its colonization in the host without severe diarrheal symptoms. However, this model comes with some limitations, as suckling mice do not develop watery diarrhea, and lymphocyte-based immune defenses are not fully developed in the host (95–97). Furthermore, as infant mice are separated from their mothers, they have a limited survival and reduced timeframe for research performance. Adult mice are less efficient for cholera studies as they are naturally resistant to V. cholerae colonization (98). Thus, manipulations such as removal of intestinal microbes by antibiotic treatment (99), or infection by ligated ileal loop surgery (100), are necessary for colonization of adult mice with V. cholerae.

#### The Zebrafish Model

V. cholerae is found in the intestinal tract of fish in the wild, where the bacteria degrades macromolecules ingested by fish via its chitinase and protease, building a commensal relationship between fish and V. cholerae (101). Analysis of cholera patients from an outbreak in 1997 showed that dried fish consumption was significantly associated with the spread of disease, implicating fish as potential vector for V. cholerae (102). Building on associations between fish and V. cholerae in the wild, the zebrafish, Danio rerio, has recently been developed recently as a natural host model to study V. cholerae (103). Importantly, pathogenic strains of V. cholerae cause a cholera-like disease characterized by host intestinal colonization, epithelial destruction, diarrhea, and the expulsion of live pathogens (103). Unlike the adult rabbit model, researchers do not require surgical interventions prior to infection, and in contrast to the mouse model, investigators are not restricted to working with antibiotic-treated juveniles (103, 104). Fish and humans have similarly complex microbiomes that shift with age and diet (105), making fish a useful model to study interactions between commensal bacteria and the invading pathogen (106). However, it is important to note that fish cannot be raised in axenic conditions, and it is technically challenging to generate and maintain fish populations with fully defined microbiota for sustained periods.

### The Drosophila Model

Insects such as chironmids (107) and houseflies (108) are candidate reservoirs of V. cholerae, and some studies suggest a correlation between disease transmission and increases in fly population, during cholera outbreaks, or in areas where the disease is endemic (109). Given the association of V. cholerae with arthropod vectors, researchers tested the utility of Drosophila as a model to characterize V. cholerae pathogenesis. Drosophila infections typically involve oral delivery of the pathogen, or introduction of the pathogen into the body cavity of the fly through a septic injury (110). In contrast to non-pathogenic Vibrio strains, injection of V. cholerae into the body cavity resulted in a rapid death of infected flies, raising the possibility of using flies as a model to study V. cholerae pathogenesis (111). In a foundational study from 2005, researchers showed that continuous feeding of adult flies with V. cholerae caused a cholera-like disease characterized by loss of weight, and rapid death that required a functional Gα<sup>s</sup> in the host (48), establishing flies as a valuable model to characterize V. cholerae pathogenesis. However, in contrast to vertebrates, ctx mutants remain lethal to flies, suggesting CT-independent pathogenic mechanisms in adult flies. Furthermore, Vibrio polysaccharide-dependent biofilm formation is important for persistent colonization of the fly rectum and for V. cholerae-mediated lethality (112), whereas Vibrio polysaccharides interfere with colonization of the host intestine (113). Thus, the fly is a useful tool to identify uncharacterized virulence factors that affect interactions between V. cholerae and an arthropod host. As studies with this model progress, it will be interesting to determine how such virulence factors impact pathogenesis in vertebrate models.

### Vibrio cholerae and the IMD Pathway

The IMD pathway modifies expression of host genes that control processes as diverse as bacterial killing, metabolism, and intestinal homeostasis (114–121). Mutations in the IMD pathway are linked with intestinal phenotypes that implicate IMD as a critical modifier of host-bacteria interactions. For example, IMD is required to survive enteric infections with entemopathogenic Pseudomonas entomophila (122). Additionally, IMD pathway mutants are characterized by changes to the composition of the intestinal microbiome, modified distribution of live bacteria throughout the intestine (123), and elevated bacterial loads in the intestine (17, 123–127). It is tempting to speculate that IMD controls bacterial populations through the direct release of antimicrobial peptides into the gut lumen. This hypothesis is supported by a recent study that confirmed a failure to contain infectious Gram-negative and fungal pathogens in flies that lack antimicrobial peptide genes (29). However, we cannot exclude the possibility that IMD-dependent control of bacterial populations includes inputs from other processes such as intestinal metabolism. Consistent with this hypothesis, studies have revealed links between immune and insulin activity in several models (128–132), including flies (120, 133–138), and IMD activity controls expression of the metabolism-regulatory peptide, Tachykinin, in enteroendocrine cells of the anterior midgut (117). In addition to metabolic deregulation, IMD pathway mutants are characterized by accelerated proliferation of intestinal progenitor cells, intestinal tissue dysplasia, and early death (34). Many of these phenotypes are reverted by elimination of the gut microbiome (124), confirming links between IMD, gut microbial composition, and intestinal health. As flies are highly amenable to modifications of intestinal gene activity, Drosophila has emerged as a particularly valuable tools to characterize links between host epithelial immunity, and V. cholerae pathogenesis.

In flies, reactive oxygen species generation does not appear to affect V. cholerae pathogenesis (139). In contrast, septic injury of adult flies with V. cholerae causes elevated expression of IMD-responsive antimicrobial peptides. Furthermore, induced expression of antimicrobial peptide genes attenuated V. cholerae pathogenesis in the septic injury model (111). These observations suggest that IMD will have protective effects against V. cholerae. However, characterization of flies challenged with V. cholerae through the natural, oral route, revealed an unexpected link between host immunity and pathogenesis. Specifically, although oral infection promotes the expression of IMD-responsive antimicrobial peptides in the intestine, IMD pathway mutants displayed an enhanced survival after oral infection with V. cholerae (140), indicating that host immune activity contributes to V. cholerae pathogenesis. Follow-up work showed that mutations in the IMD pathway have minimal effects on levels of intestinal V. cholerae (139). Nonetheless, whereas V. cholerae inhibit ISC growth in wild-type flies, ISC proliferation is unimpaired in the intestines of V. cholerae-infected IMD pathway mutants (139) suggesting that V. cholerae-dependent activation of IMD inhibits ISC proliferation, accelerating host death.

Studies of links between host immunity and V. cholerae pathogenesis uncovered an involvement of the Drosophila oxidation resistance 1 ortholog, mustard (mtd), in host viability (139, 141). Mustard is a Lysine Motif domain-bearing protein with roles in pupal eclosion (142). A gain-of-function mutant, mtdEY<sup>04695</sup> , that increases expression of a nuclear localized mustard isoform, significantly improves the survival duration of flies infected with V. cholerae (141). Molecular work showed that mtdEY<sup>04695</sup> mutants process the IMD-responsive NF- κB transcription factor Relish normally, and express most antimicrobial peptides to wild type levels after infection (139, 141). However, genome-wide transcriptional studies uncovered broad overlaps between the expression profiles of mtdEY<sup>04695</sup> and an IMD pathway mutant, including diminished expression of the diptericin antimicrobial peptide, suggesting interactions between mustard function and IMD activity. Similar to IMD pathway mutants, mtdEY<sup>04695</sup> flies are capable of progenitor cell growth after infection, supporting the notion of links between immune activity, ISC proliferation, and host survival. Looking forward, it will be interesting to characterize the immune phenotypes of loss-of-function mutations in the mtd locus.

A recent study from our group examined the consequences of IMD inactivation in defined intestinal cell types for host viability after infection with V. cholerae (143). We found that inhibition of IMD in differentiated enterocytes significantly extended the survival times of infected flies, whereas inhibition of IMD in the progenitor cell compartment shortened survival times. These observations suggest that the activity of IMD in enterocytes is sufficient to enhance V. cholerae pathogenesis. The mechanism by which immune activity influences V. cholerae pathogenesis requires clarification. In this context, we note that IMD is required for the delamination of damaged cells in the intestinal epithelium (119). As V. cholerae causes extensive damage to the midgut epithelium (139, 140, 144), we consider it is possible that V. cholerae kills the host, in part, by activating IMD-dependent sloughing of the epithelium. In this untested model, excess delamination effectively disrupts the epithelial barrier, preventing the transduction of growth cues to progenitor cells, and leading to systemic infection and host death. However, we cannot exclude alternative, and potentially non-exclusive mechanistic links, such as metabolic dysfunction, between immune activity and host mortality. In particular, there is a considerable amount of data linking intestinal metabolism to disease progression in infected flies.

#### Vibrio cholerae and Host Metabolism

The gut microbiota modifies metabolism in Drosophila, with implications for host growth and development (40, 42, 145). For example, symbiotic Ap are a source of thiamine during development (146). Additionally, Ap-derived acetate stimulates insulin signaling activity in the fly (40). The Drosophila insulin response pathway is highly similar to the vertebrate counterpart (147), and Ap-dependent control of insulin activity affects key developmental processes such as intestinal growth, size regulation, and storage of energy (40). Similar to Ap, symbiotic Lactobacillus plantarum plays an important role in the regulation of larval growth. In this case, Lp activates intestinal peptidases, at least partially in an IMD-dependent manner (148), to promote the uptake of amino acids from the larval growth medium, thereby activating the Target of Rapamycin complex, and promoting larval growth (42). When considering microbial control of host metabolism, it is important to note that higher-order interactions in a complex community of intestinal bacteria impact host health and fitness (43). For example, interactions between symbiotic Acetobacter and Lactobacillus species influence lipid homeostasis in adult flies (149).

A genetic screen for V. cholerae mutants with impaired pathogenesis in flies identified the CrbRS two-component system as a modifier of host killing (150). CrbRS is composed of the CrbS histidine kinase sensor, and the CrbR response regulator. CrbRS controls expression of acetyl CoA-synthase (acs1), a bacterial regulator of acetate consumption. In E. coli, expression of acs1 activates the acetate switch, whereby bacteria switch from production to consumption of the short-chain fatty acid, acetate (151). The acetate switch is conserved in V. cholerae, as mutations in crbR, crbS, or acs1 prevent consumption of acetate by V. cholerae in liquid culture (150, 152). These observations suggest that V. cholerae-dependent virulence may involve consumption of intestinal acetate by the pathogen. Consistent with that hypothesis, provision of dietary acetate was sufficient to extend survival times in flies infected with V. cholerae. Mechanistically, the authors showed that consumption of intestinal acetate by wild-type V. cholerae disrupted insulin signaling in the host, leading to intestinal steatosis and depletion of lipid stores from the fly fat body, an insect organ with functional similarities to the vertebrate liver and white adipose tissue (153). Removal of lipidsfrom the fly medium prevented steatosis, and extended host viability, confirming a role for lipid homeostasis in V. cholerae pathogenesis. Interestingly, CrbS is expressed during V. cholerae infections in mice and humans (154, 155), raising the possibility that pathogenic consumption of intestinal acetate is a general virulence strategy of V. cholerae.

Links between metabolism and pathogenesis extend beyond short-chain fatty acid consumption. For example, mutations of the V. cholerae glycine cleavage system also attenuate virulence in the fly model (156). These mutants colonize fly intestines with equal efficiency as wild-type V. cholerae, indicating that the phenotype is likely a consequence of an increased ability of the host to tolerate infection. In line with this hypothesis, glycine cleavage mutants fail to suppress ISC division, and do not affect lipid levels in fat tissue or homeostasis. Instead, glycine cleavage mutants have increased levels of methioninesulfoxide in their intestines, and dietary supplementation with methionine-sulfoxide, or mutation of the host Methionine sulfoxide reductase A (MsrA) gene extended host viability and restored lipid homeostasis to flies infected with V. cholerae, implicating methionine-sulfoxide availability in pathogenesis.

Metabolic regulation is also sensitive to quorum-sensing by V. cholerae. A recent study showed that quorum sensing in the El tor C6706 strain minimizes pathogenesis in flies, as deletion of the quorum-sensing master regulator, hapR, increased pathogenesis (157). HapR suppresses the expression of CT (158), and toxin co-regulated pilus virulence factors (159), and inhibits expression of Vibrio polysaccharide (160, 161), a biofilm exopolysaccharide that enables colonization of the Drosophila rectum (112). The elevated pathogenesis observed in 1hapR strains was not the result of increased biofilm formation. Instead, the phenotype appears to be the consequence of increased succinate uptake by 1hapR due to elevated expression of the Vibrio cholerae INDY succinate transporter. Consistent with this model, supplementation of the infection medium with succinate significantly extended survival times of flies infected with 1hapR. Similar to phenotypes associated with methionine-sulfoxide, and acetate, succinate consumption by V. cholerae was associated with depletion of lipid stores from the fat body, suggesting a possible role for inter-organ regulation of lipid homeostasis in the survival of infection with V. cholerae.

#### Vibrio cholerae and ISC Growth

Much of the data above describe the phenotypic impacts of V. cholerae-mediated consumption of intestinal metabolites. However, it is important to remember that V. cholerae competes with gut-resident bacteria for attachment to the intestinal niche (162). Thus, V. cholerae-dependent displacement of intestinal bacteria can also affect the profile of metabolites available to the host. For example, V. cholerae encodes a type six secretion system (T6SS) that delivers an array of toxins to susceptible prokaryotic, and eukaryotic, prey (163–165). Two studies from our group implicated the T6SS in Drosophila pathogenesis mediated by the El Tor strain, C6706. The first study showed that the T6SS targets symbiotic Acetobacter pasteurianus for killing, and that the T6SS contributes to host killing (144). T6SS-dependent killing of the host requires the presence of Ap, and association of adult flies exclusively with T6SS-refractory Lactobacillus species is sufficient to extend the viability of C6706-infected hosts. These data indicate that T6SS-mediated killing of flies proceeds through an indirect route that requires host association with Acetobacter.

More recently, we showed that the T6SS also affects epithelial renewal in infected flies. In agreement with previous work (139), we showed that V. cholerae causes extensive damage to the midgut epithelium, but fails to activate compensatory proliferation in basal progenitor cells (166). Removal of the T6SS diminishes epithelial damage, and restores renewal in infected midguts. These effects are not the result of direct interactions between the T6SS and the host epithelium, as removal of the intestinal microbiome restores renewal capacity to midguts infected with C6706. Collectively, these data indicate that the T6SS contributes to V. cholerae-mediated inhibition of epithelial renewal in a manner that requires a gut microbiome. In these assays, inhibition of renewal is not a simple consequence of interactions between V. cholerae and symbiotic Acetobacter. Instead, inhibition of renewal required association of infected flies with a tripartite community of gut bacteria, consisting of Ap, Lactobacillus brevis, and Lp, suggesting that T6SS-dependent arrest of progenitor growth is the result of complex interactions between the pathogen and a community of symbionts. Interestingly, quorum sensing appears to be an important factor in progenitor renewal. In vertebrates, the master quorum sensor regulator, hapR is not expressed at early stages of infection, where V. cholerae are present in low density. The absence of HapR allows for production of the toxin coregulated pilus, and CT, resulting in disease. As V. cholerae numbers increase, quorum sensing-dependent production of HapR results in a repression of virulence genes. In our studies, we used a C6706 strain with low hapR expression (167), allowing for expression

FIGURE 2 | Schematic representation of the impact of pathogenic Vibrio cholerae on metabolism, growth and immunity in the adult Drosophila midgut. For clarity, we have broken the individual responses into separate panels, although it is important to note that growth, metabolism and immunity share regulatory components in vivo. By consuming metabolites such as methionine sulfoxide (MetO) and acetate, V. cholerae affects lipid homeostasis contributing to death. At the same time, V. cholerae impairs IPC growth pathways, although it is unclear how this affects symbiont-dependent growth responses (indicated with a question mark). Finally, the host IMD pathway contributes to pathogenesis by impairing IPC growth, and possibly by affecting epithelial turnover (indicated by a question mark).

of virulence genes in the fly. In contrast, earlier studies with several C6706 strains that express hapR failed to arrest progenitor growth, and were not pathogenic to flies (157). Mutation of hapR in these strains restored pathogenesis, and blocked proliferation. In total, these studies hint at a sophisticated interplay between quorum sensing, bacterial competition, and epithelial renewal in the host. It will be interesting to determine the mechanistic basis for these interactions in future studies.

#### CONCLUSION AND FUTURE DIRECTION

In this review, we have discussed the utility of D. melanogaster as an experimental model to understand V. cholerae pathogenesis. In the last 15 years, work with the fly uncovered a complex series of interactions between the invading pathogen, the intestinal microbiome, and host defense mechanisms (**Figure 2**). V. cholerae disrupts lipid metabolism in enterocytes, and in the fat body, suggesting impacts of the pathogen on communication between these critical regulators of lipid homeostasis. Host immune defenses contribute to pathogenesis,

#### REFERENCES


as IMD pathway mutants survive infections longer than their wild-type counterparts, and display an improved epithelial renewal response. It will be interesting to determine the mechanistic links between immune activity and epithelial renewal, and to determine how changes to lipid metabolism impact pathogenesis. We also consider it important to remember that growth, immunity, and metabolism share numerous regulatory components. The fly is a particularly valuable model to ask how these evolutionary conserved pathways interact to orchestrate systemic responses to a global health threat.

### AUTHOR CONTRIBUTIONS

SD and EF wrote the paper.

#### FUNDING

This research was funded by grants from the Canadian Institutes of Health Research to EF (MOP77746).

hybridization study in mice. World J Gastroenterol. (2005) 11:1131–40. doi: 10.3748/wjg.v11.i8.1131


the Drosophila gut induces immune responses to infectious bacteria and tolerance to microbiota. Cell Host Microbe. (2012) 12:153–65. doi: 10.1016/j.chom.2012.06.002


consequences in obese and nonobese mice. Diabetes. (1997) 46:1526–31. doi: 10.2337/diab.46.9.1526


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

Copyright © 2020 Davoodi and Foley. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

, Hirotaka Kanoh<sup>1</sup>

,

,

# A Receptor Guanylate Cyclase, Gyc76C, Mediates Humoral, and Cellular Responses in Distinct Ways in Drosophila Immunity

#### Edited by:

Laura Vesala, Tampere University, Finland

#### Reviewed by:

Ines Anderl, University of Birmingham, United Kingdom Elodie Ramond, Institut National de la Santé et de la Recherche Médicale (INSERM), France

\*Correspondence: Shoichiro Kurata shoichiro.kurata.d5@tohoku.ac.jp

†These authors have contributed equally to this work

#### ‡Present address:

Akira Goto, Université de Strasbourg, CNRS, Insect Models of Innate Immunity (M3I; UPR9022), Strasbourg, France Takayuki Kuraishi, Faculty of Pharmacy, Institute of Medical, Pharmaceutical and Health Sciences, Kanazawa University, Kanazawa, Japan

#### Specialty section:

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

Received: 18 October 2019 Accepted: 08 January 2020 Published: 28 January 2020

#### Citation:

Iwashita S, Suzuki H, Goto A, Oyama T, Kanoh H, Kuraishi T, Fuse N, Yano T, Oshima Y, Dow JAT, Davies S-A and Kurata S (2020) A Receptor Guanylate Cyclase, Gyc76C, Mediates Humoral, and Cellular Responses in Distinct Ways in Drosophila Immunity. Front. Immunol. 11:35. doi: 10.3389/fimmu.2020.00035 Takayuki Kuraishi 1,3‡, Naoyuki Fuse<sup>1</sup> , Tamaki Yano<sup>1</sup> , Yoshiteru Oshima<sup>1</sup> , Julian A. T. Dow<sup>4</sup> Shireen-Anne Davies <sup>4</sup> and Shoichiro Kurata<sup>1</sup> \* <sup>1</sup> Graduate School of Pharmaceutical Sciences, Tohoku University, Sendai, Japan, <sup>2</sup> Graduate School of Life Sciences, Tohoku University, Sendai, Japan, <sup>3</sup> PRESTO, Japan Science and Technology Agency, Tokyo, Japan, <sup>4</sup> Institute of Molecular,

Cell and Systems Biology, College of Medical, Veterinary and Life Sciences, University of Glasgow, Glasgow, United Kingdom

Shinzo Iwashita1†, Hiroaki Suzuki 1†, Akira Goto1,2†‡, Tomohito Oyama<sup>1</sup>

Innate immunity is an evolutionarily conserved host defense system against infections. The fruit fly Drosophila relies solely on innate immunity for infection defense, and the conservation of innate immunity makes Drosophila an ideal model for understanding the principles of innate immunity, which comprises both humoral and cellular responses. The mechanisms underlying the coordination of humoral and cellular responses, however, has remained unclear. Previously, we identified Gyc76C, a receptor-type guanylate cyclase that produces cyclic guanosine monophosphate (cGMP), as an immune receptor in Drosophila. Gyc76C mediates the induction of antimicrobial peptides for humoral responses by a novel cGMP pathway including a membrane-localized cGMP-dependent protein kinase, DG2, through downstream components of the Toll receptor such as dMyD88. Here we show that Gyc76C is also required for the proliferation of blood cells (hemocytes) for cellular responses to bacterial infections. In contrast to Gyc76C-dependent antimicrobial peptide induction, Gyc76C-dependent hemocyte proliferation is meditated by a small GTPase, Ras85D, and not by DG2 or dMyD88, indicating that Gyc76C mediates the cellular and humoral immune responses in distinct ways.

Keywords: receptor-type guanylate cyclase, humoral immune responses, cellular immune responses, Drosophila, innate immunity

#### INTRODUCTION

The innate immune system is a powerful and evolutionarily well-conserved barrier to infectious pathogens (1, 2). The fruit fly Drosophila melanogaster is an excellent model organism for deciphering the basic principles of innate immunity, which comprises both humoral and cellular responses (3–5). Induction of antimicrobial peptides (AMPs) in the fat body, the functional equivalent of the mammalian liver, is a humoral response in Drosophila controlled by two distinct innate immune signaling pathways, the Toll and immune deficiency (imd) pathways (4, 6). Studies of the Toll receptor, which is involved in host-defense in Drosophila, led to the discovery of a Toll-like receptor regulating innate immunity in mammals (1, 2, 7, 8). The Toll and imd pathways are mechanistically similar to the mammalian nuclear factor-kappa B signaling pathways, the Toll-like receptor/interleukin-1 receptor signaling pathway and the tumor necrosis factor-α receptor signaling pathway, respectively (2). Both pathways are mediated by several factors, including the Toll receptor and Drosophila myeloid differentiation primary response 88 (dMyD88) adaptor protein, which mediates the Toll pathway; and peptidoglycan recognition protein-LE and peptidoglycan recognition protein-LC receptors, and Relish transcriptional factor, which mediate the imd pathway (4, 6). The Toll pathway is mainly involved in immune defense against fungal and Grampositive bacterial infections, whereas the imd pathway is mainly involved in immune defense against Gram-negative bacterial infections (3, 6). Upstream of the Toll receptor, peptidoglycan recognition protein-SA and Gram-negative bacteria-binding protein-1 are involved in the recognition of Gram-positive bacteria and Gram-negative bacteria-binding protein-3 is involved in the recognition of fungi. These recognition proteins activate modular serine protease (ModSP), which activates the serine protease cascade (9–12). The Spätzle-processing enzyme is then activated to cleave the cytokine-like protein Spätzle (Spz). Processed Spz binds to the Toll receptor to activate the Toll pathway.

Cellular responses in Drosophila are primarily carried out by the blood cells (hemocytes), and include phagocytosis, hemocyte proliferation, and encapsulation by differentiated hemocytes called lamellocytes (3, 13) Recent reports demonstrated crucial roles for hemocytes in host defense against various bacterial infections (14–16), and identified the involvement of several key factors in the phagocytosis of different pathogens, hemocyte proliferation, hemocyte differentiation, and parasite encapsulation (17–20). Two waves of hematopoiesis occur during Drosophila development. The first population of hemocytes derives from the head mesoderm in the embryo producing two main classes of hemocytes called plasmatocytes and crystal cells (21–25). The second hematopoiesis occurs during the larval stage in a specialized organ called the lymph gland (26). Lymph glands are responsible for producing larval hemocytes comprising ∼90% of plasmatocytes, ∼5% of crystal cells, and a third class of cells named lamellocytes, which are generated upon infection by parasitic wasps (26–28). A number of previous studies have demonstrated the involvement of these hemocytes during infection, but relatively little is known about the control and coordination of humoral and cellular immune responses for eliminating invaders.

We previously identified genes capable of activating immune responses by establishing a genome-wide gain-of-function genetic screen based on modular misexpression using GAL4/UAS in Drosophila (29, 30). Use of this screening system led to the identification of a receptor-type guanylate cyclase (rGC), Gyc76C, which produces cyclic guanosine monophosphate (cGMP) and mediates AMP induction of humoral responses through the downstream Toll-receptor components dMyd88, Pelle, Tube, and Dif/Dorsal (nuclear factor-kappa B) in parallel with the Toll receptor (Kanoh et al., under revision). This Gyc76C-induced cGMP signaling pathway is mediated by the membrane-localized cGMP-dependent protein kinase (cGK) DG2, encoded by the gene dg2 (foraging) and by protein phosphatase 2A, which is crucial for host survival against Gram-positive bacterial infections in Drosophila (Kanoh et al., under revision). Here we report that Gyc76C is also required for hemocyte proliferation in response to bacterial infections. In contrast to Gyc76C-dependent AMP induction, however, Gyc76C-dependent hemocyte proliferation is meditated by a small GTPase, Ras85D, and not by DG2 or dMyD88, indicating that the Gyc76C-mediated cellular response and the Gyc76C-mediated humoral response are differentially regulated. These findings indicate that Gyc76C is an immune receptor that differentially mediates both cellular and humoral immune responses.

### MATERIALS AND METHODS

### Fly Stocks Used in the Study

Fly stocks used in the study are summarized in **Table 1**.

### Bacterial Infection

The following bacteria were used for infection: Escherichia coli (K-12), Erwinia carotovora carotovora 15 (Ecc15), Staphylococcus aureus (ATCC14801, wood46), S. saprophyticus (GTC0205), and Enterococcus faecalis (IFO12964). The flies were raised on a standard cornmeal-yeast agar medium. Flies were infected with bacterial strains by injecting ∼70 nl of a suspension of each bacterial strain per fly at 3–5 days after eclosion. The optical density at 600 nm for each bacterial suspension was as follows: E. faecalis (0.0001), S. saprophyticus (1.0), S. aureus (0.0001), and Ecc15 (1.0). Survival experiments were performed with 30 flies of each genotype at 28◦C. Surviving flies were counted daily by transferring the flies to fresh vials. For larval infection, overnight S. aureus and E. coli cultures were concentrated by centrifugation. The pellets were washed with phosphate-buffered saline (PBS) and the larvae were then pricked with a fine tungsten needle that had been dipped in a pellet of concentrated bacteria.

### Total RNA Isolation and Real-Time PCR

Total RNAs were isolated from each genotype of ∼20 flies or larvae with Trizol reagent (GIBCO/BRL). Total RNA (1 µg) was used for cDNA synthesis with ReverTraAce reverse transcriptase (Toyobo) and oligo(dT) 15 primer (Promega). Using the firststrand cDNA (0.5 µl), real-time polymerase chain reaction (PCR) was performed using a LightCycler (Roche Diagnostics). Rp49 was used as the internal control. The primers used for real-time PCR were as follows (F = forward, R = reverse):

Rp49: AGATCGTGAAGAAGCGCACCAAG (F); CACCAG GAACTTCTTGAATCCGG (R)

Gyc76C: AGCTACCCCAACTGGGAGAT (F); TGACTC GAGTGCACTTCACC (R)

dg2: ATTACTGGTCGCTGGGAGTG (F); AGAAGCCAT CGAACCATTTG (R)

Drs: TTGTTCGCCCTCTTCGCTGTCCT (F); GCATCCTTC GCACCAGCACTTCA (R)

Dpt: GTTCACCATTGCCGTCGCCTTAC (F); CCAAGT GCTGTCCATATCCTCC (R)

Def : TTGAACCCCTTGGCAATGCA (F); AGTTCTTCG TTCTCGTGGCT (R)

CecA1: CATCTTCGTTTTCGTCGCTC (F); CGACATTGG CGGCTTGTTGA (R)

Att: GTGGTGGGTCAGGTTTTCGC (F); TGTCCGTTG ATGTGGGAGTA (R)

#### TABLE 1 | Fly stocks used in this study.


RNAi, RNA interference; VDRC, Vienna Drosophila Resource Center.

Mtk: AACTTAATCTTGGAGCGA (F); CGGTCTTGGTTG GTTAG (R)

Dros: CCATCGTTTTCCTGCT (F); CTTGAGTCAGGTGAT CC (R)

#### Colony Forming Unit (CFU) Assay

Flies were collected at 0, 6, 24, and 48 h after injection of each bacterial strain and sterilized with 70% ethanol. A total of 14 flies of each genotype was homogenized in 500 µl of the appropriate bacterial medium, serially diluted, and plated onto the appropriate plates (Luria Bertani medium for E. faecalis; nutrient broth medium for S. aureus and S. saprophyticus).

#### Hemocyte Staining

Third instar larvae were dissected in Schneider's Drosophila medium containing 14% fetal bovine serum at 6 h after infection. Circulating hemocytes were fixed with methanol/water/acetic acid (95:4:1) for 20 min, permeabilized with cold methanol for 15 min, incubated overnight with anti-PH3 (Cell Signaling Technology) diluted 140-fold in PBT (PBS containing 0.1% Triton-X 140), washed, and incubated with Cy-3 anti-rabbit IgG diluted 500-fold in PBT (Jackson ImmunoResearch). The cells were stained with 4′ ,6-diamidino-2-phenylindole (DAPI; MilliporeSigma) in PBS to visualize nuclei and observed with a Zeiss Axioplan 2 microscope. To count hemocytes, the hemolymph from 10 third-instar larvae per sample was collected in 50 µl PBS. The hemocyte number was counted using a hemocytometer. We counted at least 10 samples and calculated the number of hemocytes per larva.

#### Co-immunoprecipitation Assay

Drosophila S2 cells were maintained at 25◦C in Schneider's Drosophila medium (Life Technologies) and transfected with V5-tagged Ras85D and a FLAG-tagged wild-type Gyc76C or Gyc76C mutant lacking a kinase homology domain. Cell lysates with lysis buffer (30 mM Tris, pH 7.5, 150 mM NaCl, and 1% CHAPS) were incubated with anti-FLAG M2 monoclonal antibody (MilliporeSigma) for 2 h at 4◦C, and then with Dynabeads M280 (Life Technologies) for 2 h at 4 ◦C. After washing with wash buffer (30 mM Tris, pH 7.5, 500 mM NaCl, and 1% CHAPS), the bead-captured proteins were eluted with sodium dodecyl sulfate (SDS) sample buffer (50 mM Tris-HCl, 200 mM β-mercaptoethanol, 2% SDS, 0.0125% bromophenol blue, and 14% glycerol) at 140◦C for 5 min. The proteins were separated by 14% SDS-polyacrylamide gel electrophoresis and transferred to polyvinylidene difluoride membranes (Hybound-P, GE Healthcare) and then analyzed with anti-V5-tag monoclonal antibody (MBL Life Science) and anti-FLAG antibody. Blots were visualized with the ECL-Western Blotting Analysis system (GE Healthcare).

### RESULTS

#### Expression of Gyc76C in Both the Fat Body and Hemocytes Is Required for Self-Defense Against Gram-positive Bacteria

We previously identified Gyc76C as an immune receptor that is crucial for host survival against Gram-positive bacterial infections in Drosophila (Kanoh et al., under revision). Gyc76C is preferentially expressed in immune-related tissues such as the fat body, a major organ producing AMPs, hemocytes involved in cellular responses, and Malphigian (renal) tubules (Kanoh et al., under revision). To determine the tissue-specific requirement

FIGURE 1 | against S. saprophyticus infection. (B) Effects of Gyc76C RNAi using two different GAL4 drivers (c564-, and hml-GAL4) on Gyc76C expression in flies. (C) Effects of hemocyte-specific expression of Gyc76C in gyc76CKG<sup>03723</sup> flies by hml-GAL4 on the survival rate against S. saprophyticus, E. faecalis, and S. aureus infections. Siblings (hml-GAL4; gyc76CKG<sup>03723</sup> or UAS-Gyc76C; gyc76CKG03723) were used as controls. (D) Semi-quantitative (upper) and quantitative (lower) RT-PCR analysis of the expression of Gyc76C in hemocytes isolated from gyc76CKG<sup>03723</sup> larvae expressing Gyc76C by hml-GAL4. (E) Suppression of Gram-positive bacterial loads by hemocyte-specific expression of Gyc76C in gyc76CKG<sup>03723</sup> flies. Differences in bacterial loads in gyc76CKG<sup>03723</sup> and gyc76CKG<sup>03723</sup> flies expressing Gyc76C in hemocytes by CFU assay are indicated. Data shown are the means of 6 independent experiments with over 30 flies of each genotype examined at the same time. (A,C) \*P < 0.05, Log-rank test. Data shown are presented as means of at least three independent experiments. (B,D,E) \*P < 0.05, ns: P > 0.1, Student's t-test. Error bars indicate standard deviation. Data are representative of the results of three independent experiments.

for Gyc76C in self-defense against Gram-positive bacteria, we investigated the effect of tissue-specific expression of RNA interference (RNAi) targeting Gyc76C using Cg-GAL4 (mainly in both the fat body and hemocytes), c564-GAL4 mainly in the fat body, but also in other tissues (31), and hemocyte-specific hemolectin (hml)-GAL4 (32) drivers in flies. Susceptibility to infection by S. saprophyticus, a Gram-positive bacteria, was induced by Gyc76C RNAi using Cg-GAL4 as previously reported (Kanoh et al., under revision), but not by Gyc76C RNAi using c564-GAL4 and hml-GAL4 (**Figure 1A**). Gyc76C expression in flies was partially reduced by hml-GAL4-mediated RNAi, but strongly reduced (90% reduction) by c564-GAL4 mediated RNAi (**Figure 1B**) similar to Cg-GAL4-mediated RNAi (Kanoh et al., under revision). These findings suggest that Gyc76C expression, mainly in both the fat body and hemocytes, is required for self-defense against Gram-positive bacteria. Demonstrating a role for Gyc76C in hemocytes in self-defense, hemocyte-specific expression of Gyc76C by hml-GAL4 in gyc76CKG<sup>03723</sup> a hypomorphic mutant fly (33), partially rescued the phenotype susceptible to Gram-positive bacterial infections (S. saprophyticus, E. faecalis, and S. aureus; **Figure 1C**). Gyc76C expression in larval hemocytes was completely rescued by hemocyte-specific expression of Gyc76C induced by hml-GAL4 in the gyc76CKG<sup>03723</sup> mutant (**Figure 1D**). Colony formation unit assay further demonstrated that while Gyc76CKG<sup>03723</sup> mutant flies accumulated significant Gram-positive bacterial loads in their hemolymph as reported previously (Kanoh et al., under revision), hemocyte-specific expression of Gyc76C in gyc76CKG<sup>03723</sup> flies conversely suppressed Gram-positive bacterial growth (E. faecalis, S. saprophyticus, and S. aureus) in the hemolymph (**Figure 1E**). Taken together, these results indicate a self-defense role of Gyc76C in hemocytes.

#### Role of Gyc76C in Cellular Responses Against Bacterial Infections

Because Gyc76C expression in hemocytes is necessary for selfdefense, we investigated the role of Gyc76C in cellular responses against bacterial infections. The number of hemocytes in the hemolymph collected from larvae overexpressing Gyc76C by Cg-GAL4 was significantly increased compared with that of control larvae expressing lacZ (**Figure 2A**). Consistent with this finding, immunofluorescence analysis with an antibody specific for phosphorylated histone H3, a marker for entry into mitosis, revealed that Gyc76C overexpression by Cg-GAL4 significantly increased the number of proliferating hemocytes in the larvae compared with control larvae expressing lacZ (**Figure 2B**). Similar results were obtained in studies of bromodeoxyuridine incorporation into hemocytes (data not shown). Moreover, a similar increase in hemocyte proliferation was induced in larvae by infection with E. coli, a Gram-negative bacteria, and S. aureus, a Gram-positive bacteria, as well as by injection of control saline, and the hemocyte proliferation was reduced in gyc76CKG<sup>03723</sup> (**Figure 2C**). Activation of the Toll pathway induces lamellocyte differentiation as well as hemocyte proliferation (3, 34, 35). On the basis of their morphology, however, lamellocyte differentiation was not induced by Gyc76C overexpression, which is consistent with reports that lamellocyte differentiation and hemocyte proliferation are independently controlled (17, 18). These findings together indicated that Gyc76C affects the basal level of hemocyte proliferation.

#### Gyc76C Mediates Hemocyte Proliferation as a Cellular Response in a Distinct Way From the Humoral Response

The bacterial infection-dependent hemocyte proliferation in larvae was not affected in dMyD88kra<sup>1</sup> , a mutant of the dMyD88 adaptor protein in the Toll pathway, and RelishE20, a mutant of the Relish transcription factor of the imd pathway, suggesting that neither the Toll nor the imd pathway is involved in bacterial infection-dependent hemocyte proliferation (**Figure 2C**). Consistently, Gyc76C-mediated induction of Drs in larvae was suppressed by the dMyD88kra<sup>1</sup> mutation as reported previously (Kanoh et al., under revision), whereas Gyc76C-mediated induction of hemocyte proliferation was not affected by the dMyD88kra<sup>1</sup> mutation, indicating that Gyc76C mediates hemocyte proliferation in a dMyD88-independent manner (**Figure 3A**). Surprisingly, hemocyte proliferation was also induced by overexpression of the Gyc76CD945<sup>A</sup> mutant, which produces low levels of cGMP and has low Drs expression in larvae (Kanoh et al., under revision), as well as by wildtype Gyc76C (**Figure 3B**). Moreover, as shown in **Figure 3C**, Gyc76C-mediated hemocyte proliferation was not affected by the expression of PDE5/6, which severely reduces both Gyc76Cmediated Drs induction and cGMP production in larvae (Kanoh et al., under revision). The Gyc76C-dependent induction of Drs is inhibited by the expression of RNAi targeting dg2, a gene of cGK, in the fat body driven by c564-GAL4 (Kanoh et al., under revision), whereas Gyc76C-dependent hemocyte proliferation was not affected by the expression of RNAi targeting dg2 in the fat body and hemocytes driven by Cg-GAL4 in larvae (**Figure 3D**). Expression of dg2 in larval hemocytes was reduced by dg2 RNAi using Cg-GAL4 (**Figure 3E**). Gyc76C has an extracellular ligand-binding domain, a transmembrane domain, intracellular kinase homology, and guanylate cyclase domains, which show amino acid sequence similarity to rGCs, including mammalian rGCs (36) (**Figure 3F**). Expression of a Gyc76C

mutant lacking the kinase homology domain (KHD) in larvae induced relatively higher Drs expression compared with wildtype Gyc76C, but a Gyc76C mutant lacking the guanylate cyclase (GC) domain failed to induce Drs expression (**Figure 3G**), consistent with a previous study demonstrating that deletion of the KHD led to an increase in the GC activity of Gyc76C in Drosophila S2 cells (37). Hemocyte proliferation was induced by the expression of a Gyc76C mutant lacking GC as well as

by wild-type Gyc76C, but not by a Gyc76C mutant lacking the KHD in larvae (**Figure 3G**). These results indicate that Gyc76C mediates hemocyte proliferation in a cGMP-independent manner. Therefore, Gyc76C mediates humoral and cellular responses by distinct mechanism. The humoral response such as AMP induction is mediated by the producing cGMP and through cGK and dMyD88 (Kanoh et al., under revision), whereas a cellular response, hemocyte proliferation, is cGMP-independent.

#### Gyc76C Mediates ModSP-Dependent Hemocyte Proliferation as Well as ModSP-Dependent Drs Expression

Drs is induced by the overexpression of ModSP, an upstream regulator of the Toll receptor (12). As reported previously (Kanoh et al., under revision), the Drs induction by overexpression of ModSP in the fat body (c564-GAL4) was suppressed in gyc76CKG<sup>03723</sup> mutant larvae (**Figure 4A**), indicating that the ModSP-dependent induction of Drs requires Gyc76C. Overexpression of ModSP in the fat body also increased the total number of hemocytes, the same as overexpression of Gyc76C in hemocytes and the fat body by Cg-GAL4 in larvae (**Figure 4B**). The ModSP-dependent increase in the hemocyte number was suppressed in gyc76CKG<sup>03723</sup> mutants, indicating that the ModSP-dependent increase in the hemocyte number also requires Gyc76C (**Figure 4B**). Therefore, although the Gyc76Cmediated humoral and cellular responses are differentially regulated downstream of Gyc76C, both responses are triggered by ModSP overexpression.

#### Gyc76C-Dependent Cellular Response Is Mediated by a Small GTPase, Ras85D

A small GTPase, Ras85D, is suggested to be involved in hemocyte proliferation (17). We investigated the effect of expressing RNAi targeting Ras85D and other small GTPase superfamily members, Rac1, Rac2, and Mig-2-like (Mtl), on Gyc76C-dependent hemocyte proliferation and Gyc76Cdependent induction of Drs in larvae. Gyc76C-dependent hemocyte proliferation was reduced by Ras85D RNAi using Cg-GAL4, whereas Gyc76C-dependent induction of Drs was not affected by Ras85D RNAi (**Figures 5A,B**). Expression of RNAi targeting Rac1, Rac2, and Mtl did not inhibit the Gyc76C-dependent hemocyte proliferation in larvae (**Figure 5C**).

experiments.

Consistent with the functional interactions of Gyc76C and Ras85D, co-immunoprecipitation results revealed that Ras85D forms a complex with wild-type Gyc76C in Drosophila S2 cells (**Figure 5D**). The Ras85D-complex formation was reduced in a Gyc76C mutant lacking the KHD that does not induce hemocyte proliferation (**Figures 5D**, **3G**). Moreover, infectiondependent hemocyte proliferation in larvae in response to S. saprophyticus and Ecc15, a Gram-negative bacteria, was reduced by a Ras85D mutation, Ras85DEY00505, caused by a P-element insertion in the 5′ -untranslated region of Ras85D (**Figure 5E**), whereas in the absence of infection, the number of hemocytes was not affected in Ras85DEY<sup>00505</sup> mutant larvae (**Figure 5F**). Therefore, Ras85D mediates hemocyte proliferation by Gyc76C in response to bacterial infections as a cellular response. Cg-GAL4-driven Ras85D-RNAi flies and Ras85DEY<sup>00505</sup> flies were susceptible to Gram-positive bacterial infections (E. faecalis and S. saprophyticus), but not to Ecc15 infection (**Figure 5G**). The response of Ras85D-RNAi flies to Ecc15 infection was consistent with a previous report (38). AMP induction after E. faecalis and Ecc15 infections was not reduced in Ras85DEY<sup>00505</sup> compared with control flies (yw), except for CecropinA1 against Ecc15 infection (**Figure 6**). These findings suggest that the Ras85D plays an important role in the cellular innate immune response against Gram-positive bacterial infection. We cannot, however, exclude the possibility of a potential contribution of a humoral response, as observed by the dysregulated antimicrobial expression pattern in Ras85D mutant flies.

#### DISCUSSION

We previously reported that the Gyc76C mediates humoral response by a membrane-localized cGK, DG2 through downstream components of the Toll receptor via dMyD88 (Kanoh et al., under revision). In this study, we provide new evidence that the Gyc76C is also involved in cellular response. Further mechanistic analyses indicate that this Gyc76C-mediated cellular response is executed through a small GTPase, Ras85D, and importantly, this response is in cGMP-independent manner. The Gyc76C-mediated cellular responses confer host survival against Gram-positive bacterial infections, like the Gyc76Cmediated humoral responses. Similar to Ras85D, Gyc76C is involved in hemocyte proliferation in response to Gram-negative bacteria, but neither Gyc76C nor Ras85D is crucial for host survival against Gram-negative bacterial infections, suggesting that Gyc76C-mediated hemocyte proliferation does not confer host survival against Gram-negative bacterial infections. Gyc76C is not involved in the imd pathway-dependent AMP induction in response to Gram-negative bacterial infections (Kanoh et al., under revision). In comparison with AMP induction by the Toll pathway in response to Gram-positive bacterial infections, AMPs are rapidly induced by the imd pathway in response to Gram-negative bacterial infections in flies (39). Because of the rapid induction of AMPs by activation of the imd pathway, Gyc76C-mediated hemocyte proliferation might not be required for host survival against Gram-negative bacterial infections.

We demonstrated that both the Gyc76C-mediated humoral and cellular responses are triggered by the overexpression of ModSP. Although the ligand of Gyc76C that induces the Gyc76C-mediated humoral response in response to Grampositive bacteria has not yet been identified (Kanoh et al., under revision), it is possible that the ligand produced by infection activates Gyc76C to induce both the humoral and cellular immune responses and thus coordinates them to eliminate the

FIGURE 6 | Expression of antimicrobial peptide genes in a Ras85D mutant after bacterial infection. Either 24 h after E. faecalis injection or 6 h after Ecc15 injection, the expression of 7 distinct AMPs was measured with the P-element insertion mutant of Ras85D, Ras85DEY<sup>00505</sup> , spzrm<sup>7</sup> , and RelishE20, and yw flies (used as a control). Because Ras85D is reported to be involved in constitutive repression of the imd pathway (38), the values of uninfected flies are also presented. Data shown are the means of at least three independent experiments. \*P < 0.05, Student's t-test. Error bars indicate standard deviation.

pathogens. Identification and characterization of the Gyc76C ligand is necessary to elucidate the coordination mechanisms of the humoral and cellular immune responses in Drosophila.

rGCs have two conserved intracellular domains, kinase homology and guanylate cyclase domains (36). The KHD regulates the activity of the associated GC domain (40). Deletion of the KHD of Gyc76C leads to increased GC activity in Drosophila S2 cells, indicating that the KHD of Gyc76C is also involved in regulating GC activity (35). In this report, we demonstrated that a Gyc76C mutant with deletion of the KHD induced Drs expression, but a Gyc76C mutant with deletion of the GC domain failed to induce Drs expression. Conversely, a Gyc76C mutant without the GC domain induced hemocyte proliferation, but a Gyc76C mutant without the KHD failed to induce hemocyte proliferation. These findings indicate that the KHD of Gyc76C has an independent role in regulating GC activity. Consistent with these analyses, co-immunoprecipitation analysis suggests that the KHD of Gyc76C is involved in the association with Ras85D that is required for Gyc76-dependent hemocyte proliferation. The KHD of Gyc76C may be involved in forming a signaling platform with other factors such as Ras85D. Additional studies are needed to clarify how the two independent functions of Gyc76C are regulated through the two functional domains of the receptor.

### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the article/supplementary material.

## AUTHOR CONTRIBUTIONS

SI and AG performed hemocyte proliferation analyses. HS performed Ras85D analyses with help of HK and TK. TO performed ModSP analyses with help of NF and TY. YO promoted this study. S-AD and JD designed the cGMP studies. SK provided overall coordination with respect to conception, design, and supervision of the study, and wrote the manuscript with comments from co-authors. SI, HS, and AG contributed equally to the study. All authors discussed the results.

## FUNDING

This work was supported by Grants-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science, and Technology of Japan (MEXT 19H03365, 16H05084, 24390014, 21117005, 21117001, and 14657577); the Japan Society for the Promotion of Science (JSPS); Japan Science and Technology Agency (JST); the Program for the Promotion of Basic Research Activities for Innovative Biosciences (PROBRAIN); the Strategic International Cooperative program from the Japan Science and Technology Agency; the National Institutes of Health (AI07495); the Takeda Science Foundation; the Mitsubishi Foundation; the Astellas Foundation for Research on Metabolic Disorders; the Uehara Memorial Foundation; the Naito Foundation; a Global COE Research Grant (Tohoku University Ecosystem Adaptability); and the Biotechnological and Biological Research Sciences Council (UK) grant number BB/E011438/1 to S-AD and JD.

#### ACKNOWLEDGMENTS

We thank K. V. Anderson, D. Ferrandon, J. A. Hoffmann, D. Hultmark, J. L. Imler, Y. T. Ip, A. L. Kolodkin, B. Lemaitre,

#### REFERENCES


T. Muta, N. Perrimon, J. M. Reichhart, J. Royet, the Bloomington Stock Center, the Drosophila Genomics Resource Center at Indiana University, the Drosophila Genetic Resource Center at the Kyoto Institute of Technology, the Drosophila RNAi Screening Center, the Genetic Strain Research Center of the National Institute of Genetics, and the Vienna Drosophila RNAi Center for fly stocks and materials.

the egg encapsulation response in Drosophila. Genetics. (2004) 166:1343–56. doi: 10.1534/genetics.166.3.1343


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

Copyright © 2020 Iwashita, Suzuki, Goto, Oyama, Kanoh, Kuraishi, Fuse, Yano, Oshima, Dow, Davies and Kurata. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Subpopulation of Macrophage-Like Plasmatocytes Attenuates Systemic Growth via JAK/STAT in the Drosophila Fat Body

#### Mingyu Shin1†, Nuri Cha1†, Ferdinand Koranteng1†, Bumsik Cho<sup>1</sup> and Jiwon Shim1,2,3 \*

*<sup>1</sup> Department of Life Science, College of Natural Science, Hanyang University, Seoul, South Korea, <sup>2</sup> Research Institute for Natural Science, College of Natural Science, Hanyang University, Seoul, South Korea, <sup>3</sup> Research Institute for Convergence of Basic Sciences, College of Natural Science, Hanyang University, Seoul, South Korea*

*Drosophila* hemocytes, like those of mammals, are given rise from two distinctive phases

#### Edited by:

*Laura Vesala, Tampere University, Finland*

#### Reviewed by: *Jin Li Hua,*

*Northeast Forestry University, China Shruti Yadav, Molecular Medicine Research Institute, United States*

> \*Correspondence: *Jiwon Shim jshim@hanyang.ac.kr*

*†These authors have contributed equally to this work*

#### Specialty section:

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

Received: *30 October 2019* Accepted: *10 January 2020* Published: *31 January 2020*

#### Citation:

*Shin M, Cha N, Koranteng F, Cho B and Shim J (2020) Subpopulation of Macrophage-Like Plasmatocytes Attenuates Systemic Growth via JAK/STAT in the Drosophila Fat Body. Front. Immunol. 11:63. doi: 10.3389/fimmu.2020.00063* during both the embryonic and larval hematopoiesis. Embryonically derived hemocytes, mostly composed of macrophage-like plasmatocytes, are largely identified by genetic markers. However, the cellular diversity and distinct functions of possible subpopulations within plasmatocytes have not been explored in *Drosophila* larvae. Here, we show that larval plasmatocytes exhibit differential expressions of *Hemolectin (Hml)* and *Peroxidasin (Pxn)* during development. Moreover, removal of plasmatocytes by overexpressing pro-apoptotic genes, *hid* and *reaper* in *Hml*-positive plasmatocytes, feeding high sucrose diet, or wasp infestation results in increased circulating hemocytes that are *Hml*-negative. Interestingly these *Hml*-negative plasmatocytes retain *Pxn* expression, and animals expressing *Hml*-negative and *Pxn*-positive subtype largely attenuate growth and abrogate metabolism. Furthermore, elevated levels of a cytokine, *unpaired 3*, are detected when *Hml*-positive hemocytes are ablated, which in turn activates JAK/STAT activity in several tissues including the fat body. Finally, we observed that insulin signaling is inhibited in this background, which can be recovered by concurrent loss of *upd3*. Overall, this study highlights heterogeneity in *Drosophila* plasmatocytes and a functional plasticity of each subtype, which reaffirms extension of their role beyond immunity into metabolic regulation for cooperatively maintaining internal homeostatic balance.

Keywords: plasmatocytes, upd3, Drosophila melanogaster, JAK/STAT, insulin signaling, Hemolectin, Peroxidasin

#### INTRODUCTION

The underlying mechanisms of the innate immune system of Drosophila melanogaster is paralleled in vertebrates (1). For example, Drosophila Toll receptor is functionally homologous to mammalian Toll-like receptors (TLRs) and their task to protect host from pathogens is conserved in vertebrates (1, 2). Also, Drosophila innate immune pathways which include the imd pathway, though having different NF-kB—relish for imd pathway and dorsal for Toll pathway—maintain comparable roles in host defense as in mammals (1, 3–5).

Comparable to vertebrates, hematopoiesis in Drosophila progresses in two waves: primitive and definitive hematopoiesis (6, 7). In the first wave or primitive hematopoiesis, hemocytes originate from the head mesoderm of embryo (8), and embryonically derived hemocytes comprise most circulating hemocytes during larval stages (9–11). However, not all hemocytes move freely within the hemolymph; a portion of embryonic hemocytes become localized at discrete regions within the larval cuticle called the hematopoietic pocket (12– 14). Thus, the embryonic hemocytes become divided into two categories: circulating and sessile, depending on their mobility or locale within the hemocoel (12). At the hematopoietic pockets, resident hemocytes can be seen around oenocytes or neurons and their positioning is controlled by sensory neurons of the peripheral nervous system (14). Definitive hematopoiesis occurs during larval stages in the lymph gland, the hematopoietic organ of Drosophila larvae (7, 15). In the lymph gland, hemocytes are classified into four clusters: the posterior signaling center, the medullary zone, the intermediate zone and the cortical zone (16–18). Prohemocytes in the medullary zone progress through the intermediate zone and eventually differentiate into plasmatocytes, crystal cells or lamellocytes in the cortical zone (17, 18). During the pupal stage, hemocytes in the lymph gland dissociate and spread throughout the whole body, becoming the hemocytes of the adult fly (11, 19).

Drosophila hemocytes are largely recognized based on the expression of genetic markers throughout their development (20). Plasmatocytes comprise about 95% of the total hemocyte population and are functionally akin to mammalian macrophages (8, 21–23). They uptake pathogenic or cellular debris, and are marked by Hemolectin (Hml), Peroxidasin (Pxn), or Nimrod C1 (NimC1) (20). Mature crystal cells are characterized by their internal crystalline structures and mediate melanization response to protect animals from injury or immune challenges. Crystal cells normally constitute about 5% of total hemocytes and are distinguished by the expression of hindsight (hnt), lozenge (lz), or Prophenoloxidase (PPO) 1 and 2 (9, 20, 23). While lamellocytes barely exist in healthy larva, they are differentiated from plasmatocytes in circulation or from the lymph gland in copious amounts upon immune challenges (12, 24–26). L1 (atilla), L2, L4, L6, or misshapen (msn) are used as markers for the lamellocytes (20, 27).

The JAK/STAT signaling cascade was first discovered in mammals where a variety of cytokines and growth factors transduce the signaling pathway related to immune responses (28). This pathway is highly conserved throughout evolution and is involved in critical biological processes of Drosophila including embryogenesis, immunity and stem cell maintenance (29). The JAK/STAT pathway in flies was originally highlighted in embryonic development where four main components are utilized: a ligand called unpaired (upd), a domeless (dome) receptor, the JAK—Hoscotch (Hop), and the STAT (30–32). In addition to the main players, negative regulators of the pathway have been also identified, including Socs36E, dPIAS, PTP61E or a BCL-6 homolog, Ken and Barbie (33). A role of JAK/STAT signaling in hemocyte development and immune responses was initially shown by a gain-of-function allele of hop, hopTuml , which leads to hyperproliferation of hemocytes and formation of melanotic tumors (34, 35). Consistent with the hematopoietic phenotype observed in hopTuml mutants, active JAK/STAT signaling is required for differentiation of lamellocytes upon wasp infestation (36). Moreover, main players of the signaling such as upd2 and upd3 are upregulated in hemocytes upon immune challenges (37). During cellular immune responses, hemocytes induce upd ligands and secrete them to the hemocoel, where active propagation of JAK/STAT signaling in various tissues including the muscle, occurs. Amongst target tissues, the activation of JAK/STAT signaling in the muscle is linked to insulin signaling and carbohydrate metabolism, directly coupling immunity and metabolism (38). Drosophila fat body is the main source for antimicrobial peptides (AMPs), which facilitate the humoral immune response (37–39) as well as for the orchestration of metabolic events to maintain internal energy balance during feeding or non-feeding states (39, 40). Insulin production and secretion in the brain insulin producing cells (IPCs) is remotely controlled by the nutrient sensing from the fat body and vice versa, fat contents in the fat body is regulated by the insulin signaling (41, 42). Therefore, the mutual interactions between the insulin signaling and the fat body coordinate metabolism and growth of animals in response to availability of nutrition (41, 43, 44). Interestingly, recent studies have shown that active innate immunity attenuates growth and nutrient storage by blocking PI3K and AKT in the fat body, establishing an intricate balance between insulin signaling and innate immunity in the fat body (42, 45).

Drosophila hemocytes have been largely classified based on their morphology and expression of a few marker genes (20). Plasmatocytes constitute the largest population and show significant functional diversity; however, it is not clear whether the current classification sufficiently describes possible heterogeneity within plasmatocytes (1, 8, 12, 46). Moreover, the developmental fluctuations within plasmatocytes have not been examined at a cellular level. To investigate cellular discrepancies of circulating plasmatocytes in developing Drosophila larvae, we utilized two binary systems, Gal4-UAS and LexA-LexAop, to simultaneously visualize two representative markers, Hml and Pxn (47, 48).

In this study, we show that Hml-positive (Hml+) and Pxnpositive (Pxn+) plasmatocytes generally overlap in embryonically derived hemocytes. However, a subpopulation of plasmatocytes exhibit only Hml or Pxn expression distinctive from the doublepositive (Hml<sup>+</sup> Pxn+) plasmatocytes. Upon expression of proapoptotic genes, hid and reaper (rpr), in Hml<sup>+</sup> hemocytes, Pxn-positive and Hml-negative (Pxn+Hml−) hemocytes increase accompanied by elevated levels of a cytokine, unpaired 3 (upd3) in hemocytes. Interestingly, upd3 from hemocytes activates the JAK/STAT signaling in various tissues including the fat body, which attenuates the insulin signaling pathway and leads to systemic metabolic dysfunction. Thus, identification of plasmatocyte subpopulations in this study enriches the concept of hemocyte heterogeneity and appends metabolism mediation to the role of plasmatocytes in immunity for the purpose of keeping internal homeostasis.

### MATERIALS AND METHODS

#### Fly Stocks and Genetics

Larvae and flies were generally reared in a Drosophila chamber which is maintained at 25◦C and 70% humidity. To enhance Gal4/UAS and LexA/LexAop expression, larvae were shifted to 29◦C after egg-laying. Also, they were cultured on normal food comprising of dextrose, cornmeal, dried yeast, and agar. Fly stocks used in this study include: Hml1-Gal4, UAS-2XEGFP (S. Sinenko), Pxn-Gal4, UAS-GFP (U. Banerjee), UAS-hid, rpr (Nambu J. R.), UAS-upd3 (B. Lemaitre), STAT92E::edGFP (N. Perrimon), Hml-Gal4 (U. Banerjee). 13XLexAop2-6XmCherry-HA (BL52271), LexAop-mRFP.nls (BL29956) UAS-Ras85D RNAi (BL34619), UAS-hid/Cyo (BL65403), w <sup>∗</sup>upd21upd3<sup>1</sup> (BL55729), tGPH (BL8164) were received from the Bloomington Drosophila Stock Center.

For high sugar diet, we substituted dextrose with sucrose (100 g/L; therefore, 300 mM). Synchronized first instar larvae were collected and transferred to high sucrose diet and kept at 25◦C. We dissected or bled larvae at 72, 96, or 120 h AEL.

#### Generation of Pxn-LexA and upd3-LexA Flies

Amplified Pxn enhancer from genomic DNA was cloned into TA-TOPO vector (Invitrogen, K252020) for gateway cloning (Primer-forward: CTCACCAACTGGATGTTGGTC/ Primer-Reverse: CCCAAACAAATATCTGTAGACTGACAG). Also, upd3 enhancer was amplified for cloning into TA-TOPO vector (Primer-forward: TCGTACAATGGTTTAAAAATAGCTCGG CCAAAT/ Primer-Reverse: AGTGACCAGTTCCTGTTCAGG CGTCGTCGTCGAT). Cloned entry vector were ligated into destination pBPnlsLexA::p65Uw (Addgene, #26230) vector by using LR ligase II (Invitrogen, 11791-020). Recombinant constructs (at least 20 µg DNA) were injected into flies and generated by BestGene Inc.

#### Immunohistochemistry

Larvae were dissected in late third instar stage in 1× PBS, fixed in 3.7% formaldehyde (Sigma, F1635) in 1× PBS for 30 min at room temperature and washed three times in 1× PBS containing 0.4% triton-X (1× PBS-T) for 10 min each. For measurement of total hemocytes, larvae were vortexed for 2 min in 1× PBS and hemocytes were allowed to ooze out for 30 min (except staining for L1 which was 1 h) on ice. Tissues were blocked in 10% normal goat serum (Vector Laboratories) in 1× PBS for 30 min. Primary antibody was incubated with tissue overnight at 4◦C and then washed three times in 0.4% 1× PBS-T. Secondary antibody was incubated 3 h at room temperature. Samples were washed three times in 0.4% 1× PBS-T and mounted in Vectashield (Vector Laboratories, H-1200). Primary antibodies used in this study: rabbit anti-Pxn [1:1,000, (49)], mouse anti-P1 (1:100, Istvan Ando), mouse anti-L1 (1:100, Istvan Ando), mouse anti-Hnt (1:10, DSHB), mouse anti-nc82 (1:10, DSHB), rabbit anti-dFOXO (1:100, Yu. K), rabbit anti-DCP-1 (1:200, Cell signaling, 9578), Rho-phalloidin (1:100, Invitrogen) and BODIPY 493/503 (1:200, Molecular Probes, 3922). Cy3-, Alexa Fluor 647- and FITC-conjugated secondary antibody were obtained from Jackson ImmunoResearch Laboratories Inc., and each antibody was used at 1:250 dilution ratio.

#### Imaging and Quantitation

All fluorescence was imaged by confocal microscopy (Nikon C2 si-plus). Two micrometers step Z stacks of larval brain, fat body, muscle, salivary gland, and intestine with identical laser power and scan setting were taken. Mean intensity of all images for each sample were calculated using Image J, Imaris (Bitplane), and Microsoft Excel software.

#### Quantitative Real-Time PCR

For measurement of gene expression, we collected samples on ice: 70 blood volumes, 10 fat body, 25 brains, 10 muscles, 10 whole larvae. RNA was isolated from tissues by using Trizol (Invitrogen). cDNA was synthesized with RT kit (TOYOBO). RT-PCR was performed using SYBR Green master mix on a Step One-Plus Real-Time PCR thermal cycler (Applied Biosystems). Gene expression was normalized by rp49. Primers used for qPCR is tabled **Supplementary Material 1**.

#### Measurement TAG in Fat Body

Collection of 10 fat body on ice with 0.05% 1×PBS-T (Tween 20) and then homogenization of tissues with pestle (or rapidly kept in−80◦C deep freezer until use). Homogenate tissues were heated on 70◦C for 10 min to inactivate lipases and previous method was followed (50). For the measurement of TAG in samples, Serum triglyceride determination kit (Sigma, TR0100- 1KT) and Glycerol standard solution (Sigma, G7793) were used. The samples were assayed using a plate reader to measure absorbance at 540 nm.

#### Measurement of Pupa Volume

Animals were segregated into male and female groups per genotype at the larval stages. Upon pupariation, 20–50 pupae were arranged on a silicon pad and photographed using Nikon SMZ18 and ProgRes CapturePro v2.8.8 software. Length and diameter of pupa was obtained using ImageJ. Pupa volume was calculated as previously described (51) using Microsoft Excel. Prism8 was then used to determine P-values and generate final graphs.

#### Measurement of Adult Weight

Adult fly mass was measured as previously described (52). Briefly, 1 day old adult flies were segregated by gender over CO<sup>2</sup> anesthetic pads. Batches of 15–25 animals for each gender and genotype were collected onto a filter paper and placed on a sensitive balance (Ohaus Pioneer, PAG 214) to determine overall mass. The average mass of animals per gender and genotype was calculated with Microsoft Excel. Prism8 was then used to determine P-values and generate final graphs.

#### Measurement of Pupation Time

Synchronized first instar larvae were collected in batches into a vial containing standard fly media and kept at 25◦C. The number of larvae turning into pupa was counted against the transition time point. The number of larvae turned into pupa was then calculated as a percentage of the overall number of pupae in a vial for each time point using Microsoft Excel. Final graph was generated using Prism8.

#### Wasp Infestation

Larvae were infested at 60 or 72 h AEL with Leptopilina boulardi for Pxn/Hml population count or Ras85D RNAi, respectively. Egg deposition was confirmed by direct observation of wasp eggs,

FIGURE 1 | Total hemocytes are counted after vortexing larvae, therefore, indicating sessile and circulating populations (54). Error bars indicate S.D. (C–E) Circulating hemocytes bled at 72 (C), 96 (D), and 120 h AEL (E). Subtypes of plamatocytes are visualized by *Hml* (green), *Pxn* (magenta), and DAPI (blue) (*Hml*1*-Gal4 UAS-EGFP; Pxn-LexA LexAop-mCherry*). Co-localization of *Hml* and *Pxn* is indicated in white. Scale bar, 40µm. (F–H) Pie chart quantitation of circulating plasmatocytes at 72 (F, related to C), 96 (G, related to D), and 120 h AEL (H, related E). *Hml*+*Pxn*<sup>+</sup> (white), *Pxn*+*Hml*<sup>−</sup> (magenta), and *Hml*+*Pxn*<sup>−</sup> (green). (I–K) Circulating and sessile hemocytes bled at 72 (I), 96 (J), and 120 h AEL (K). Subtypes of plamatocytes are visualized by *Hml* (green), *Pxn* (magenta), and DAPI (blue) (*Hml*1*-Gal4 UAS-EGFP; Pxn-LexA LexAop-mCherry*). Co-localization of *Hml* and *Pxn* is indicated in white. Scale bar, 40µm. (L–N) Pie chart quantitation of circulating and sessile plasmatocytes at 72 (L, related to I), 96 (M, related to J), and 120 h AEL (N, related K). *Hml*+*Pxn*<sup>+</sup> (white), *Pxn*+*Hml*<sup>−</sup> (magenta), and *Hml*+*Pxn*<sup>−</sup> (green).

after 8 h of co-culture. All infestation procedures were performed at 25◦C.

#### RESULTS

#### Distinctive Patterns of Hml and Pxn in the Larval Hemocytes

Using multiple binary systems including the Gal4-UAS and LexA-LexAop, we can simultaneously visualize several genetic markers and study interactions of respective genes or tissues (47, 48, 53). Expression of Hml or Pxn have been exploited to illuminate our understandings on development and functions of plasmatocytes, and utilized as markers for plasmatocytes across several Drosophila hematopoiesis analyses (20). However, it remains unclear whether these markers equally label entire plasmatocytes in larval circulation at the transcript level. To understand discrepancies between the two markers in larval hemocytes, we generated Pxn-LexA construct to concurrently visualize Hml and Pxn in larval hemocytes (refer to method for details). By utilizing two binary systems, we first verified the distribution of Hml<sup>+</sup> and Pxn<sup>+</sup> populations in embryonically derived hemocytes at 72, 96, and 120 h after egg laying (AEL) (**Figures 1A,B**). With the circulating portion of hemocytes, we observed more than 50% overlap between Hml<sup>+</sup> and Pxn<sup>+</sup> plasmatocytes at each time point, though specific ratios differ (**Figures 1C–H**). Total hemocytes including circulating and sessile populations show similar proportions of Hml<sup>+</sup> and Pxn<sup>+</sup> plasmatocytes (**Figures 1I–N**). Besides the Hml and Pxn doublepositive plasmatocytes, Hml-positive and Pxn-negative (Hml<sup>+</sup> Pxn−) or Pxn<sup>+</sup> Hml<sup>−</sup> subpopulations are indicated in both circulation and total hemocyte samples (**Figures 1F–H,L–N**). Though the larger of the two subpopulations—Pxn<sup>+</sup> Hml<sup>−</sup> and Hml<sup>+</sup> Pxn−–is Pxn<sup>+</sup> Hml−, both Hml<sup>+</sup> Pxn<sup>−</sup> and Pxn<sup>+</sup> Hml<sup>−</sup> are distinct at all-time points (**Figures 1F–H,L–N**). Staining for P1, a mature plasmatocyte marker, hnt, a crystal cell marker or L1, a lamellocyte marker, shows that both subpopulations are randomly co-localized with P1 and hnt (**Supplementary Figures 1A–I**). L1 does not show any expression under normal culture conditions in larval plasmatocytes (**Supplementary Figures 1G–I**) These patterns suggest that plasmatocyte subtypes expressing Pxn<sup>+</sup> Hml<sup>−</sup> and Hml<sup>+</sup> Pxn<sup>−</sup> are not exclusive to crystal cells nor late plasmatocytes, and are not lamellocytes.

### Distribution of Pxn<sup>+</sup> or Hml<sup>+</sup> Hemocytes Are Changed by Physiological Alterations

Hemocyte population size is known to be influenced by changes in internal and external conditions (12, 46, 55). Based on the differential expressions of Pxn and Hml in larval hemocytes, we next sought to understand whether the ratios of the plasmatocyte subpopulations seen under normal growing conditions can be adjusted by distinctive physiological conditions. We designed physiological challenges in 2-folds: immunological and metabolic, given previous notions linking immunity and metabolism (42, 45). First, we infested the second-instar larvae at 60 h AEL by wasps (Leptopilina boulardi) and examined expressions of Pxn<sup>+</sup> or Hml<sup>+</sup> hemocytes at 72, 96, and 120 h AEL. Interestingly, we observe that wasp infestation induces a biased expansion of Pxn<sup>+</sup> Hml<sup>−</sup> hemocytes from 12 h post infestation (PI; 72 h AEL) prior to the massive proliferation of total hemocytes (**Supplementary Figures 1J,M**). Moreover, the proportion of Pxn<sup>+</sup> Hml<sup>−</sup> hemocytes further expands along with excessive proliferation of total hemocytes and differentiation of lamellocytes at 96 and 120 h AEL (**Supplementary Figures 1K,L,N,O**). Second, we modified internal metabolism by supplying additional sucrose to the normal diet (56). Similar to the immunological challenge, we found an increase in Pxn<sup>+</sup> Hml<sup>−</sup> hemocytes after chronic supplementation of high sucrose diet (**Supplementary Figures 1P–U**). Distinct from the wasp infestation, high sucrose diet does not increase Pxn<sup>+</sup> Hml<sup>−</sup> hemocytes from 72 h AEL; however, the percentage of Pxn<sup>+</sup> Hml<sup>−</sup> hemocytes is drastically augmented by 120 h AEL (**Supplementary Figures 1R,U**). These results indicate that proportions of Pxn<sup>+</sup> Hml+, Pxn<sup>+</sup> Hml<sup>−</sup> or Hml<sup>+</sup> Pxn<sup>−</sup> hemocytes are plastic and amenable upon an immune challenge generated by wasp infestation or a metabolic alteration induced by supplementation of high sugar diet. Also, these physiological changes readily alter plasmatocytes ratios seen under normal culture conditions, driving a biased expansion of Pxn<sup>+</sup> Hml<sup>−</sup> hemocytes.

#### Hemocytes Produce High Levels of upd3 in the Absence of Hml<sup>+</sup> Cells

Our observations indicate that Pxn<sup>+</sup> Hml<sup>−</sup> plasmatocytes are the second largest plasmatocyte subtype comprising ∼30% of total hemocytes and modified upon immune or nutritional insults. Despite the relatively large proportion of Pxn<sup>+</sup> Hml<sup>−</sup> hemocytes in the plasmatocyte population, the function of this subtype has not been explored. Thus, we next examined the expression and possible functions of Pxn<sup>+</sup> Hml<sup>−</sup> plasmatocytes by reducing the Hml<sup>+</sup> population. First, we genetically ablated the Hml<sup>+</sup> hemocytes by expressing proapoptotic genes, hid and reaper (rpr) from the first instar of larval development (Hml1-Gal4, UAS-hid, rpr), and observed changes in the numbers and proportions of total hemocytes

FIGURE 2 | Ablation of *Hml*<sup>+</sup> hemocytes increases the number of circulating and sessile hemocytes, and induces *upd3* expression. (A,B) Expression of *Pxn* (magenta) in *Hml*1*-Gal4 UAS-EGFP, Pxn-LexA LexAop-mCherry* background. Compared to hemocytes sampled from wild type (A), the genetic ablation of *Hml*<sup>+</sup> cells increases the number of *Pxn*<sup>+</sup> (magenta) cells (*Hml*1*-Gal4 UAS-EGFP UAS-hid,rpr, Pxn-LexA LexAop-mCherry*) (B). *Hml* (green), *Pxn* (magenta), and DAPI (blue). Scale bar, 40µm. (C) Quantitation of circulating and sessile hemocytes in genetic backgrounds used in (A,B). Graphs indicate median plots of DAPI positive cells per mm<sup>2</sup> in each genotype. Highest and lowest bars indicate maximum and minimum values, respectively. Controls (green) (*Pxn-LexA LexAop-mCherry* or *UAS-hid,rpr; Pxn-LexA LexAop-mCherry*), *Hml* ablated background (pink) (*Hml*1*-Gal4 UAS-EGFP UAS-hidrpr; Pxn-LexA LexAop-mCherry*). Statistical significance was determined by *t-test*. \*\*\**p* < 0.001; not-significant, ns. (D) Pie chart shows the proportion of *Hml*+*Pxn*<sup>+</sup> (white), *Hml*+*Pxn*<sup>−</sup> (green), or *Pxn*+*Hml*<sup>−</sup> (magenta) in circulating and sessile plasmatocytes of *Hml*+ablated background (*Hml*1*-Gal4 UAS-EGFP UAS-hidrpr; Pxn-LexA LexAop-mCherry*). Quantitation of (B). (E,F) mRNA levels of signaling molecules and antimicrobial peptides in *Hml*+ablated background (*Hml*1*-Gal4 UAS-hid,rpr*). RT-qPCR analysis of genes related to active immunity normalized by *rp49* using whole larvae (E). *upd3* expression in relevant organs (F). *upd3* expression is highly increased in whole larvae (E), and hemocytes show identical increase in *upd3* (F). Statistical analyses were performed using two-way ANOVA in whole larvae and organs. \*\*\**p* < 0.001; \*\*\*\**p* < 0.0001; not significant, ns. (G,G') *upd3* (magenta; *(Continued)* FIGURE 2 | *upd3-LexA LexAop-mRFP*) is co-localized with Pxn (green) in total hemocytes. Inset in magnified view. DAPI is blue. Scale bars, 40µm. (H,H',I,I') *STAT92E::edGFP* (green) expression is increased in the fat body upon loss of *Hml*<sup>+</sup> hemocytes. (H,H') is control (*Hml-Gal4; STAT92E::edGFP*) and (I,I') is *Hml*<sup>+</sup> ablated background (*Hml-Gal4 UAS-hid, rpr; STAT92E::edGFP*). Scale bars, 80µm. (J,K) *Socs36E* is increased in the fat body when *Hml*<sup>+</sup> hemocytes are genetically ablated or *upd3* is ectopically expressed in *Pxn*<sup>+</sup> cells. mRNA expression of *Socs36E* is increased in the fat body extracted from *Hml*1*-Gal4 UAS-hid, rpr* (J) or from *Pxn-gal4 UAS-upd3* (K). Graph indicates RT-qPCR analyses of *Socs36E* in in the fat body. Error bar in graph is S.D. Statistical significance was determined by using *t-test*. \*\*\**p* < 0.001.

in circulation (**Figures 2A,B**, **Supplementary Figures 2A–D**). Compared to controls, the Hml ablated background shows four times increment in the total hemocyte count (**Figure 2C**, **Supplementary Figure 2E**). Associated with this phenotype, the Pxn<sup>+</sup> Hml<sup>−</sup> subpopulation expands to as much as 93% of the total plasmatocytes (**Figure 2D**). Moreover, the number of lamellocytes is increased comparable to previously reported phenotype (**Supplementary Figures 2F,G**) (57). We verified that the remaining cells upon loss of Hml<sup>+</sup> hemocytes express Pxn (**Supplementary Figures 2H,I**). Interestingly, the remainder Pxn<sup>+</sup> plasmatocytes do not show Hml expression and are not apoptotic (**Supplementary Figures 2J–L,J',K'**), demonstrating that these plasmatocytes are not Hml<sup>+</sup> nor dying cells, but expressing Pxn. We repeated this experiment by temporarily ablating Hml<sup>+</sup> hemocytes only at the third-instar stage. However, the acute elimination further reduces the number of total hemocytes while concomitantly increasing caspase-positive cells (**Supplementary Figures 2M–P,M',N'**), demonstrating that temporal expression of hid and rpr in Hml<sup>+</sup> hemocytes exerts differential effects to hemocytes and only chronic ablation gives rise to the biased expansion of Pxn<sup>+</sup> Hml<sup>−</sup> cells. As a second approach, we reduced the Hml<sup>+</sup> hemocytes by expressing Ras85D RNAi in Hml<sup>+</sup> hemocytes (Hml1- Gal4, UAS-Ras85D RNAi). Though Ras85D RNAi significantly reduces the total hemocytes including Hml<sup>+</sup> population, the ratio of plasmatocyte subpopulations is fairly maintained (**Supplementary Figures 3A–D**). Therefore, we reasoned that the Pxn and Hml have differential expressions in circulating and sessile hemocytes, and selective reduction in Hml subpopulation raises the Pxn subtypes in specific conditions.

Overall increase of Drosophila hemocyte populations has been attributed to systemic immune signaling (37). To ratify the causal systemic molecule for the upsurge of remnant plasmatocyte upon Hml<sup>+</sup> hemocyte ablation, we performed whole-larva realtime quantitative polymerase chain reaction (RT-qPCR). We targeted two representative antimicrobial peptides, Drosomycin and Diptericin, and three cytokines–upd, upd2, and upd3– as putative indicators of immune activation (58, 59). Moreover, we additionally checked PDGF- and VDGF-related factors, pvf1, pvf2, and pvf3, as markers for hemocyte migration (60). We observed that Drosomycin is significantly up-regulated by ablating Hml<sup>+</sup> hemocytes whereas Diptericin exhibits a marked decrease (**Figure 2E**). Remarkably, amongst all the other candidates, unpaired 3 (upd3) is the most excessively induced (**Figure 2E**, **Supplementary Figure 3E**). To determine the source of increased upd3, we screened the expression of upd3 in tissues including fat body, hemocytes, brain, intestine and muscle, after removing Hml<sup>+</sup> hemocytes. Notably, we found that hemocytes exclusively produce the highest upd3 mRNA in the ablated background (**Figure 2F**). Related to this phenotype, we observed that the increase of upd3 in hemocytes is recapitulated by feeding a high sucrose diet (**Supplementary Figure 3F**), implying that the expanded Pxn<sup>+</sup> hemocytes boost upd3 production independent of the apoptosis of Hml<sup>+</sup> hemocytes. We further verified that upd3 co-localizes with Pxn<sup>+</sup> hemocytes (**Figures 2G,G'**). On the other hand, expression of Ras85D RNAi in Hml<sup>+</sup> cells leads to lowering of upd3 expression in hemocytes, different from upd3 mRNA expression driven by loss of Hml<sup>+</sup> hemocytes or supplementation of high sucrose diet (**Supplementary Figure 3G**). Hence, we concluded that upd3 mRNA is induced in hemocytes upon ablation of Hml<sup>+</sup> hemocytes, possibly due to the expansion of Pxn<sup>+</sup> hemocytes.

### Loss of Hml<sup>+</sup> Hemocytes Alters Systemic Growth of Animal

While observing hematopoietic phenotypes in Drosophila larvae, we noticed that animals with Hml1-Gal4 UAS-hid,rpr show delayed pupation than wildtype controls (**Supplementary Figure 4A**). This systemic growth delay prompted us to further quantify pupal and adult growth parameters which can be consequences of imbalanced hemocytes or prolonged expression of a cytokine upd3 (61). We detected that chronic expression of hid and rpr in Hml<sup>+</sup> hemocytes during larval stages significantly reduces the size of pupae (**Supplementary Figures 4B–D**). Consistently, similar reduction is observed in both male and female adult flies (**Supplementary Figures 4E–G**), suggesting that persistent loss of Hml<sup>+</sup> hemocytes attenuates growth and decreases the size of animals. We next addressed whether the small size of animals is caused by an increase in upd3 expression, and found that ectopic expression of upd3 using Pxn-Gal4 recapitulates the size reduction comparable to that shown in Hml1-Gal4 UAS-hid,rpr flies (**Supplementary Figure 4H**). Overall, we concluded that high levels of upd3 derived upon loss of Hml<sup>+</sup> hemocytes systemically suppresses animal growth from larvae to adult flies.

### Active JAK/STAT Signaling Attenuates Insulin Pathway in Fat Body

The JAK/STAT pathway is a common downstream target of upd3, and is known to be involved in innate immune responses including hemocyte proliferation and lamellocyte differentiation (36, 37). Given that loss of Hml<sup>+</sup> hemocytes induces upd3 expression in hemocytes, we next examined the activity of JAK/STAT signaling and its downstream target tissue upon loss of Hml<sup>+</sup> hemocytes. Using a STAT92E::edGFP reporter (62), we ascertained that four organs—fat body (**Figures 2H–I'**), muscle (**Supplementary Figures 5A–B'**), intestine (**Supplementary Figures 5C–D'**) and trachea

FIGURE 3 | Attenuated insulin signaling in the fat body upon loss of *Hml*<sup>+</sup> hemocytes. (A–C) Membrane localization of *t*GPH, a reporter for PI3K activity, is reduced by ablating *Hml*<sup>+</sup> hemocytes. Fat body isolated from wild types exhibits membrane-expression of *t*GPH (green) (A), whereas this pattern disappears in *Hml-Gal4 UAS-hid,rpr* background (B). Bottom images indicate magnified Z-stacks of corresponding images. Quantitation of *t*GPH expression in the fat body membrane (C). DAPI, blue. Scale bar, 40µm. Statistical significance was determined by *t-test*. \*\*\*\**p* < 0.0001. (D–F) Expression of dFOXO upon genetic ablation of *Hml*<sup>+</sup> hemocytes. Cytosolic and low expressions of dFOXO in the wild-type fat body (D,D'). The level and nuclear localization of dFOXO is enhanced in *Hml-Gal4 UAS-hid,rpr* background (E,E'). DAPI (blue) and dFOXO (red) are overlaid in (D,E), and dFOXO (red) alone is shown in (D',E'). Nuclear dFOXO levels are quantitated in (F). *(Continued)*

FIGURE 3 | Statistical significance was determined by *t-test*. \*\*\*\**p* < 0.0001. Scale bar, 40µm. (G,H) mRNA expression of *4EBP* is increased in both *Hml*1*-Gal4 UAS-EGFP UAS-hidrpr* (G) and *Pxn-Gal4 UAS-upd3* (H) backgrounds in the fat body. Error bar in graph is S.D. Statistical significance was determined by *t-test*. \*\*\**p* < 0.001. (I–K) Expression of lipid droplets in the fat body. Compared to wild types (I,I') (*Hml*1*-Gal4 UAS-EGFP*), the size of lipid droplets is decreased when *Hml*<sup>+</sup> hemocytes are ablated (J,J') (*Hml-Gal4 UAS-hid, rpr*). Quantitation of BODIPY diameter in (I',J') (K). Highest and lowest bars indicate maximum and minimum values, respectively. Statistical significance was determined by *t-test*. \*\*\*\**p* < 0.0001. Scale bar, 40µm. (L) Biochemical measurement of triacyl glyceride (TAG) levels in the fat body normalized by protein contents. The level of TAG is decreased upon loss of *Hml*<sup>+</sup> hemocytes (*Hml-Gal4 UAS-hid, rpr*). Highest and lowest bars indicate maximum and minimum values, respectively. Statistical significance was determined by *t-test*. \**p* < 0.05.

(**Supplementary Figures 5E–F'**)—exhibit substantially high STAT92E::edGFP activities in the Hml<sup>+</sup> hemocyte ablated background. Yet, two organs including the brain and the salivary gland did not show any considerable changes (**Supplementary Figures 5G–J'**). These patterns are reflected in RT-qPCR analyses using Socs36E, a downstream target of JAK/STAT pathway (**Supplementary Figure 5K**). Among the four organs with high STAT92E::edGFP expression, we focused on the fat body considering systemic phenotypes of small animal size and active antimicrobial peptide gene expressions in Hml1- Gal4 UAS-hid, rpr background (**Supplementary Figures 4A–H**, **Figure 2E**). To add, Hml1-Gal4 UAS-hid, rpr shows high level of Socs36E in the fat body, comparable to that observed when upd3 is overexpressed in Pxn<sup>+</sup> hemocytes (**Figures 2J,K**), confirming that the fat body is indeed stimulated by upd3 originating from hemocytes upon loss of Hml<sup>+</sup> hemocytes.

Following the metabolic phenotypes of Hml ablated animals, we hypothesized that increased upd3 obstructs insulin signaling, the representative signaling for systemic growth, in the fat body. To understand whether insulin signaling in the fat body is directly altered by loss of Hml<sup>+</sup> hemocytes, we examined the expression of tGPH, a PI3K reporter (63), and observed that membrane localization of tGPH is diminished in the fat body (**Figures 3A–C**). This data indicates that PI3K is not recruited to the fat body cell membranes, and therefore, not activated in the Hml ablated background. PI3K activation delocalizes FOXO from cell nuclei, inhibits 4EBP, and prevents lipolysis (64). However, when Hml<sup>+</sup> hemocytes are ablated, nuclear localization of dFOXO is induced, and a transcriptional target of dFOXO, 4EBP, is increased (**Figures 3D–G**). The ascent in 4EBP in the fat body recurs when upd3 is overexpressed in Pxn<sup>+</sup> hemocytes (**Figure 3H**). In addition, fat storage in the fat body is reduced (**Figures 3I–L**). All of these findings imply that, insulin receptor (InR) signaling is abrogated in the absence of Hml<sup>+</sup> hemocytes. Furthermore, InR mRNA in the fat body is decreased when Hml<sup>+</sup> hemocytes are ablated or upd3 is overexpressed in Pxn<sup>+</sup> hemocytes (**Supplementary Figures 5L–M**), demonstrating that high levels of upd3 is sufficient to reduce InR mRNA expression in the fat body. Thus, ablation of Hml<sup>+</sup> hemocytes causes an increase in upd3 expression in hemocytes, and the downregulation of insulin signaling indicated by: reduced PI3K activity, nuclear localization of dFOXO, upregulation of 4EBP, and reduced triacyl glycerides storage in the fat body.

#### upd3 Is Required for Hemocyte Expansion and Systemic Metabolic Responses

Given our new findings that enhanced expression of upd3 upon loss of Hml<sup>+</sup> hemocytes abrogates systemic growth through insulin signaling, and JAK/STAT activities in the fat body, we next investigated whether upd3 is solely responsible for the phenotypes—hemocyte proliferation, active JAK/STAT signaling in the fat body, and altered metabolism of animals shown in Hml<sup>+</sup> ablated backgrounds. Since both upd213 1 genes and UAS-hid,rpr transgenes are localized in the first chromosome, we utilized an alternative transgene, UAS-hid to combine all the genotypes. The number of circulating and sessile plasmatocytes as well as lamellocytes are greatly increased when hid and rpr are overexpressed in Hml<sup>+</sup> hemocytes (**Figure 2C**, **Supplementary Figures 2E–G**). These phenotypes recur when only UAS-hid is highly expressed in Hml<sup>+</sup> hemocytes (Hml1-Gal4 UAS-hid) (**Figures 4A,B,E**, **Supplementary Figures 6A–E**). Also, the remaining Pxn<sup>+</sup> hemocytes do not show Hml expression and are not apoptotic when Hml+ hemocytes are ablated in a chronic manner (**Supplementary Figures 6F–H**). Nonetheless, when hid is overexpressed in the Hml<sup>+</sup> hemocytes in upd213 <sup>1</sup> null mutant background (upd21upd31; Hml1- Gal4 UAS-hid), the numbers of circulating and sessile hemocytes and lamellocytes are restored (**Figures 4C–E**, **Supplementary Figures 6I–K**). In the lymph gland, expression of hid in Hml<sup>+</sup> hemocytes accelerates precocious differentiation and disintegrates the primary lobe similar to previous observations (**Supplementary Figures 6L–M**) (65). However, contrary to the circulating and sessile hemocytes, the lymph gland phenotypes are not recovered in upd213 <sup>1</sup> null mutant background (**Supplementary Figure 6N**). In addition to rescue of blood phenotypes, both the insulin- and JAK/STAT-related phenotypes are rescued when hid is overexpressed in Hml<sup>+</sup> hemocytes in the upd213 <sup>1</sup> mutant. First, we noticed that both Socs36E, a hallmark for the JAK/STAT signaling, and 4EBP, a downstream target of insulin pathway, are recovered to normal levels in the fat body when upd2 and upd3 are deleted (**Figures 4F,G**). Second, we confirmed that dFOXO expression is delocalized from fat body nuclei when UAS-hid is expressed in Hml<sup>+</sup> hemocytes in the upd213 <sup>1</sup> null mutant background (**Figures 4H–K**). Altogether, we conclude that high upd3 induced by ablation of Hml<sup>+</sup> hemocytes is required for the increase in JAK/STAT and decrease in insulin signaling in the fat body, contributing to systemic growth retardation phenotypes in animals.

#### DISCUSSION

In this study, we reiterate the heterogeneity of plasmatocyte populations in embryonically derived hemocytes by taking

FIGURE 4 | Pxn<sup>+</sup> (magenta) and/or DAPI<sup>+</sup> (blue) cells (B). This phenotype is recovered by combining *upd2*1*upd3*<sup>1</sup> in the *Hml-Gal4 UAS-hid* background (C,D). There is no significant difference between *Hml*1*-Gal4 UAS-EGFP* controls (A) and *upd2*1*upd3*<sup>1</sup> (C) or *upd2*1*upd3*1*; Hml-Gal4 UAS-hid* (D). Controls (A,C) express *Hml* (green), Pxn (magenta), and DAPI (blue); therefore, are indicated in white. Quantitation of each genotype is shown in (E). Graphs indicate median plots of DAPI positive cells per mm<sup>2</sup> in each genotype. Highest and lowest bars indicate maximum and minimum values, respectively. Statistical significance was determined by *t-test*. \**p* < 0.05; not significant, ns. Scale bar, 40µm. (F,G) Increased levels of *Socs36E* or *4EBP* are reverted by loss of *upd2* and *upd3*. Increased levels of *Socs36E* (F) or *4EBP* (G) upon loss of *Hml*<sup>+</sup> hemocytes in the fat body are recovered in *upd2*1*upd3*1; *Hml-Gal4 UAS-hid* genetic background. Rescue is indicated in red. Statistical analyses were performed using two-way ANOVA. \**p* < 0.05; \*\*\*\**p* < 0.0001; not significant, ns. (H–K) Nuclear location and levels of dFOXO in the fat body are rescued by loss of *upd2* and *upd3*. In controls, low levels of dFOXO (red) are located in the cytoplasm and nucleus (H,H'). Overexpression of *hid* in *Hml*<sup>+</sup> hemocytes further induces the nuclear expression of dFOXO (red) in the fat body (I,I'). Genetic combination of *upd2*1*upd3*<sup>1</sup> and *Hml-Gal4 UAS-hid* reduces the nuclear expression of dFOXO (red) (J,J'). Quantitation of dFOXO intensity shown in (H–J') (K). Highest and lowest bars indicate maximum and minimum values, respectively. Statistical significance was determined by *t-test*. \*\*\*\**p* < 0.0001. Scale bar, 40µm.

advantage of two distinct binary systems, Gal4-UAS and LexA-LexAoP. In an effort to uncover subpopulations of plasmatocytes, we utilized two representative markers, Hml and Pxn, and simultaneously measured their transcriptional activities in the embryonically derived hemocytes (20). In both circulation and at sessile sites under normal growing conditions, the proportions of Pxn<sup>+</sup> Hml+, Pxn<sup>+</sup> Hml−, or Hml<sup>+</sup> Pxn<sup>−</sup> plasmatocytes are relatively fixed. Moreover, expansion of each subtype is tightly regulated during development suggesting that the composition of plasmatocyte is not random, but, rather controlled. Generally, Pxn<sup>+</sup> Hml<sup>+</sup> hemocytes comprise the largest amongst three subtypes, and Hml<sup>+</sup> Pxn<sup>−</sup> hemocytes represent the population with the least frequency. A slight reduction in Hml+Pxn<sup>−</sup> hemocytes is observed at 96 h AEL, which could be ascribed to a drastic expansion of other populations including Pxn+Hml<sup>+</sup> and Pxn<sup>+</sup> Hml<sup>−</sup> during that specific time point (**Figures 1J,M**). This notion is reinforced by the ability of hemocytes to differentiate or proliferate at sessile sites (12, 26, 66).

When we reared larvae on high sugar diet or infested larvae with wasps, we observed a biased increment of Pxn<sup>+</sup> Hml<sup>−</sup> plasmatocyte subtype, identical to the phenotype observed by ablating Hml<sup>+</sup> hemocytes (**Supplementary Figures 1J–U**). These data indicate that plasmatocyte subtypes naturally fluctuate according to developmental timings, alterations in nutrition or elevation of innate immunity. Also, the presence of natural plasticity in hemocyte populations during the larval life cycle, and its correspondence to both internal and external environmental changes is asserted. Additionally, the inherent heterogeneity indicates functional divergence of plasmatocytes: the Hml<sup>+</sup> and Pxn<sup>+</sup> population may be progenitor-like and possibly permit trans-differentiation or proliferation; but, Hml<sup>+</sup> Pxn<sup>−</sup> or Pxn<sup>+</sup> Hml<sup>−</sup> subtypes could be specialized to finetune different aspects of homeostasis including immunity or metabolism.

In mammals, several studies have verified disparity in macrophage populations, which is captured in M-1 or M-2 macrophages. While M-1 macrophages are prone to reinforce classical immunological phagocytosis behaviors, M-2 macrophages are metabolism biased (67, 68). Similar to the M-1/M-2 macrophages, we expect that there are functional segregations within the plasmatocyte population. It is likely that Pxn<sup>+</sup> Hml<sup>−</sup> population respond to metabolic

alterations such as high sucrose diet, and that Hml<sup>+</sup> population is responsible for proper immune responses given lack of lamellocytes caused by reducing the number of Hml<sup>+</sup> hemocytes (**Supplementary Figures 6O–Q**). These details suggest imaginable conservation of the M-1/M-2 macrophage paradigm in Drosophila plasmatocytes, and we expect that plasmatocytes can be further classified by their RNA or protein expression at single-cell resolutions.

consequently affects growth and metabolism.

Rise in systemic upd3 is associated with innate immune responses (37). We observed heightened upd3 expression when Hml<sup>+</sup> hemocytes are depleted. However, considering overall decrease in animal sizes and metabolic responses mediated by the fat body, we assume that the changes in upd3 in this context is not necessarily immunological. This assertion is consistent with findings from other studies showing that loss of Hml<sup>+</sup> hemocytes leads to muscle degeneration, developmental, and leg defects (57). These non-immune phenotypes may result from increased upd3. We expect that the Pxn<sup>+</sup> Hml<sup>−</sup> remnant plasmatocyte population or Hml<sup>+</sup> dying hemocytes are the potential origin of upd3. Given that feeding high sucrose diet causes an expansion of Pxn<sup>+</sup> Hml<sup>−</sup> hemocytes and also significantly upregulates the level of upd3, we assume that Pxn<sup>+</sup> Hml<sup>−</sup> hemocytes could be a more likely source of the systemic upd3 (**Supplementary Figures 1R,U, 3F**). Determining the definitive source of upd3 among the plasmatocyte subtypes in this condition requires further research.

Recent publications have highlighted genetic interactions between the JAK/STAT and insulin signaling as well as putative coupling of immune and metabolic functions (69). Other studies have shown an identical interaction between the JAK/STAT and insulin signaling in the muscle whose interaction is essential for cellular immune responses including lamellocyte differentiation (38, 70). Our study represents an in vivo interaction between JAK/STAT and insulin signaling pathways in the fat body (**Figure 5**). Given that both proximal and distal components of insulin signaling—PI3K and InR respectively—are altered when JAK/STAT pathway is activated, convergence of the two pathways could occur at the direct downstream of InR activity or InR transcription level. It will be intriguing to uncover the exact confluence between the two pathways despite the dichotomy that exists between them: while insulin signaling is growth and proliferation-biased, JAK/STAT is immunological (71–74). More so, this may provide insights into how immunity and metabolism differentially interact in normal development and pathologies.

### REFERENCES


#### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the article/**Supplementary Material**.

#### AUTHOR CONTRIBUTIONS

MS, NC, FK, and BC performed experiments. MS, NC, FK, and JS analyzed data and wrote the manuscript. JS conceived the idea and supervised the project.

### FUNDING

This work was supported by National Research Foundation (NRF) grant funded by the Ministry of Science and ICT, Republic of Korea (NRF-2019R1A2C2006848) to JS.

#### ACKNOWLEDGMENTS

The authors thank members of the Shim lab for helpful discussions. The authors acknowledge the Bloomington, VDRC, and KDRC Drosophila stock centers and the DSHB hybridoma bank. The authors thank the following individuals for stocks and reagents: U. Banerjee, S. Sinenko, J. R. Nambu, B. Lemaitre, N. Perrimon, I. Ando, and K. Yu.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2020.00063/full#supplementary-material


Mediators Inflamm. (2015) 2015:816460. doi: 10.1155/2015/ 816460


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

Copyright © 2020 Shin, Cha, Koranteng, Cho and Shim. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Sexual Dimorphisms in Innate Immunity and Responses to Infection in Drosophila melanogaster

Rebecca L. Belmonte1†, Mary-Kate Corbally 1†, David F. Duneau<sup>2</sup> \* ‡ and Jennifer C. Regan<sup>1</sup> \* ‡

1 Institute of Immunology & Infection Research, University of Edinburgh, Edinburgh, United Kingdom, <sup>2</sup> Laboratoire Evolution & Diversite Biologique, UMR5174 EDB, CNRS, Université Toulouse 3 Paul Sabatier, Toulouse, France

#### Edited by:

Susanna Valanne, Tampere University, Finland

#### Reviewed by:

Mark Austin Hanson, École Polytechnique Fédérale de Lausanne, Switzerland Ilias Kounatidis, Diamond Light Source, United Kingdom

#### \*Correspondence:

David F. Duneau david.duneau@gmail.com Jennifer C. Regan jenny.regan@ed.ac.uk

†These authors have contributed equally to this work and share first authorship

‡These authors have contributed equally to this work and share last authorship

#### Specialty section:

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

Received: 31 October 2019 Accepted: 16 December 2019 Published: 31 January 2020

#### Citation:

Belmonte RL, Corbally M-K, Duneau DF and Regan JC (2020) Sexual Dimorphisms in Innate Immunity and Responses to Infection in Drosophila melanogaster. Front. Immunol. 10:3075. doi: 10.3389/fimmu.2019.03075 The sexes show profound differences in responses to infection and the development of autoimmunity. Dimorphisms in immune responses are ubiquitous across taxa, from arthropods to vertebrates. Drosophila melanogaster shows strong sex dimorphisms in immune system responses at baseline, upon pathogenic challenge, and over aging. We have performed an exhaustive survey of peer-reviewed literature on Drosophila immunity, and present a database of publications indicating the sex(es) analyzed in each study. While we found a growing interest in the community in adult immunity and in reporting both sexes, the main body of work in this field uses only one sex, or does not stratify by sex. We synthesize evidence for sexually dimorphic responses to bacterial, viral, and fungal infections. Dimorphisms may be mediated by distinct immune compartments, and we review work on sex differences in behavioral, epithelial, cellular, and systemic (fat body-mediated) immunity. Emerging work on sexually dimorphic aging of immune tissues, immune senescence, and inflammation are examined. We consider evolutionary drivers for sex differences in immune investment, highlight the features of Drosophila biology that make it particularly amenable to studies of immune dimorphisms, and discuss areas for future exploration.

Keywords: Drosophila, Drosophila melanogaster, innate immunity, sex dimorphism, aging, response to infection, sexual antagonism

#### INTRODUCTION

Sex governs physiology: differences between males and females are strong drivers of variance in phenotype within any population, and can eclipse effects of geography or genotype (1, 2). The immune system is no exception. Sex differences in human immunity are profound, where men and women respond differently to infection, treatment, diseases such as sepsis, and have different propensities toward autoimmunity (3, 4). However, the mechanisms underpinning these dimorphisms are largely unresolved. A major reason for this lack of resolution is that sex differences in immunity are understudied; in particular, there is a paucity of truly comparative studies. A recent meta-analysis addressing the issue of sex as a variable in biomedical studies showed that immunology as a discipline is particularly negligent, with fewer than 10% of studies reporting, or stratifying, by sex (5). Historically, women have been excluded from clinical trials and young males presented as "the norm," in part due to concerns for potential impacts on fetal health (6). Parity has not been reached in representation (7) or reporting (8) of the sexes, despite the effective ban on women participating in clinical trials ending in the 1980s (6). In addition, in studies using laboratory model organisms, practical and budgetary considerations have led to the common practice of using a single sex. Recently, there has been recognition of the loss of knowledge propagated by the lack of inclusion of both sexes, with a drive from the scientific community to address the "gender gap," including NIH and ERC commitments to address this specifically (9, 10).

The effects of immune dimorphisms are not only a consideration for clinical research, but also impact our broader understanding of host-pathogen interactions. Sex differences in immunity are observed throughout taxa, and are both cause and consequence of sex differences in life history, and sexual conflict. Responses to infection influence survival and fecundity, and therefore immune dimorphisms have the potential to affect both horizontal and vertical disease transmission throughout the animal kingdom (11). Inherant in the consideration of sex and immunity is complexity: within a single species, dimorphisms themselves are pathogen-specific (12), can respond to environmental variables such as diet (13, 14), and may even be influenced by the infective parasite which can be differently adapted to each sex (15). Adding further complexity is the interaction of immunity and sex with organism age (16).

As is the case for all insects, Drosophila melanogaster physiology is sexually dimorphic (17–19), yet despite its use for more than a century as a model organism, the extent of these dimorphisms are only just being fully appreciated (20). Sex differences are seen in immune tissues (21, 22) and in responses to infection (23), yet relatively few studies include both sexes. Studies that explicitly compare immune responses in both sexes in Drosophila reflect what is seen in other taxa in terms of prevalence and complexity: dimorphic responses are the norm rather than the exception, the direction that dimorphisms take with respect to the opposite sex is both pathogen- and contextdependent (23), and sex differences at baseline are not necessarily predictive of survival outcome (24).

#### What Can Drosophila Teach Us About Immune Dimorphism?

Drosophila species have been used for many decades to study sexual antagonism in the evolutionary ecology field (25–27). We argue that including, and comparing, both sexes in functional and mechanistic studies of Drosophila immunity will add to this body of work to give important insight to several fields, in addition to better understanding host-pathogen interactions from an evolutionary ecology perspective. It will, for example, offer translatable information on disease vector biology, where sex is a crucial variable for exposure, transmission, and control strategies of insect disease vectors such as mosquitos (28–30). Drosophila provides a tractable model for innate immunity in mammals, as has been amply demonstrated over recent decades (31): studies on Drosophila could help understand rules underpinning sexual dimorphism in mammalian immunity and response to infections. Sex differences in mammalian immunity are often attributed solely to the action of steroid hormones. Interactions between steroid hormones and the immune system have also been demonstrated in Drosophila (32, 33), which may parallel endocrine-immune interactions in mammals. Mammalian immune dimorphisms arise not only as a consequence of selective pressures on the endocrine system. A large body of studies in mammalian immunology has uncovered many dimorphisms, particularly in autoimmune disease etiology, that are regulated by karyotype, independent of hormonal action (34–36). While it is difficult to attribute autoimmunity to organisms without immune self-recognition, direct self-damage by immune responses on the host has been demonstrated in Drosophila (37, 38) and may differ depending on the sex (39). Drosophila, like mammals, bear X and Y sex chromosomes, and both X- and Y-linked variation in immune responses have been demonstrated (40, 41). This, in combination with the strong conservation of immune signaling pathways (as exemplified by Toll/TLRs), makes Drosophila a powerful model for sex-specific genetic regulation of molecular immunity (23, 42).

### A Survey of Immunity Studies Using Adult Drosophila melanogaster

We have undertaken to perform a survey of peer-reviewed, published studies of D. melanogaster immunity, reporting on the representation of sex within each study. We have focussed on studies using adult flies, while also identifying papers that use juveniles, cell cultures, or pre-existing genetic data. We searched for Drosophila immunity papers through Web of Science, attempting to avoid studies of other model organisms that mentioned Drosophila. As of the 21 August 2019, we downloaded the citations for the resulting 5,626 publications and manually categorized each paper. Two thousand eight hundred and forty-eight papers were removed since Drosophila were not used in the study, or the focus of the study was not immunity. We also made a decision to exclude the 166 endosymbiont studies that used both sexes from our analysis. There were also three papers which we could not access, which were removed (43– 45). Of the remaining 2,614 papers, 1,369 used adults, 817 used juveniles, 396 were non-experimental (i.e., reviews or methods), and 30 were bioinformatics studies (**Figure 1A**). In the last 30 years, we can see how the Drosophila immunity field has grown and the relative use of adults or juveniles has changed. In 1990, only two adult studies and three juvenile studies were published (**Figure 1B**). Initially, the majority of studies were conducted on juveniles, but by 2010, the trend had reversed, where 75 studies used adults and 47 used juveniles. The use of adult Drosophila to study immunity has continued to grow, averaging over 100 adult studies per year between 2012 and 2018 (**Figure 1B**). Of the adult studies, 41% (564/1,369) did not report the sex, or did not stratify data by sex. When sex was reported, 45% used only one sex, with more studies using females (28%; 378/1,366), than males (18%; 243/1,366). Only 13% (184/1,366) of all adult studies reported results for both males and females ("both" category; **Figure 1C**). We additionally tagged each paper in the "both" category with the study type or data output (**Figure 1D**, **Figure S1**). Tags were not exclusive, where studies could be assigned multiple tags, resulting in more tags than articles in the output. The two most common tags were "gene function/knockout" (23%; 76/334),

(yellow; 184). (D) Total number of articles that used both sexes, tagged according to experimental output. Tags are not exclusive, so some articles may have more than one tag. The tags used were gene function knock out (46), survival infection dynamics (47), fitness (39), lifespan (32), tissue specific (30), microbiome (24), signaling (19), behavioral (18), transcriptomics (16), metabolism composition (8), and proteomics (4).

followed by "survival/infection dynamics" (20%; 68/334). The survey information is available as a searchable table, intended as a resource for locating data on immunity in both sexes (**Table S1**). Integrating information from this wealth of published data, we review what is reported about sex dimorphisms in immunity in Drosophila.

#### SEX DIFFERENCES IN INFECTION OUTCOMES

#### Survival and Pathology

Drosophila exhibit dimorphic survival and pathology in response to bacterial, viral, and fungal infections. Importantly, dimorphic survival and prevalence of infection are pathogen- and contextspecific (**Table 1**).

#### Viral Infection

Few studies have compared male and female responses to viral infection in D. melanogaster. Males are possibly more susceptible to acute viral infection: they demonstrate lower survival to the recently-described, DNA virus Kalithea (KV), which was isolated from infected individuals caught from a wild population (48). Males are also more susceptible to higher viral titers of the RNA virus Drosophila C (DCV) (59). Notably, dimorphism in survival to viral infection might be influenced by coinfection with Wolbachia, an endosymbiont providing viral protection (60). Indeed, interactions between DCV and Wolbachia infection status and sex have been observed in analyses of behavioral responses (61), discussed in more detail below.

Other effects of viral infection can impact the sexes differently, in addition to survival. For example, although females infected with KV generally survive, they suffer from ovary degeneration and a strong reduction in fecundity (48). Viral-induced female infertility is known to occur in infections with flock house virus (FHV) due to oocyte destruction (62). Thus, conclusions on sexspecific impacts of viral infection on fitness need to consider all consequences of infection, not just survival.

#### Viral Transmission

A recent study demonstrated that male-biased DCV titers are accompanied by higher levels of fecal shedding (59), which is in apparent contrast to an earlier study that found females to


be better transmitters of DCV than males (63). It is not known whether load, or rates of shedding, are necessarily predictive of the ability of each sex to transmit viral infection; this may be dependent upon several additional factors, including infection route and behavioral responses to infection. Further studies correlating viral load and shedding with the ability to transmit infection in both sexes will be informative. Verticallytransmitted viruses by definition interact with host sex, given their route through infected gonads. For example, Sigma virus (Rhabdoviridae), a negative-stranded RNA virus, is transmitted vertically through the sperm or ovules. Male transmission of Sigma virus is required for persistence in the population, while transmission efficiency is higher for females than males (64). Infection with Sigma virus leads to sexually dimorphic gene induction, with more gene expression changes induced by infection in males than females (65). Infected females significantly upregulate structural chorion proteins, which could reflect manipulation by the virus to aid vertical transmission, or an ovary-specific defense response.

Given the very small number of studies addressing dimorphisms in survival to viral infections, we do not yet have the ability to make inferences about sex differences in antiviral responses, nor indeed whether there are differences between responses to RNA and DNA viruses, or diverse viral species.

#### Fungal and Microsporidial Infection

To our knowledge, there are only a small number of studies investigating sexual dimorphism in fungal infection. Indeed there is a dearth of studies investigating sex-specific physiological responses to such challenges (**Figure 2**). The most commonly studied fungal infection model, Beauveria bassiana, exhibits male-biased survival when flies are challenged via spore inoculation (49–53). The dimorphism appears to be, at least partly, attributable to dimorphic function of the Toll pathway, where loss-of-function mutants in Toll pathway components lose the sex difference in survival (52, 53). A similar male bias in survival is observed post-inoculation with the soil fungus Metarhizium anisopliae (54). However, although not directly compared, males appear to be more susceptible to systemic Candida albicans challenge by means of intra-thoracic injection (55). Interestingly, the effect of Toll-1 and Toll-7 mutation on resistance to C. albicans challenge revealed a greater sensitivity of male mutants to this fungal infection, while Toll-7 mutant females demonstrated resistance similar to that of controls (55). In addition, males succumbed to systemic infection with the microsporidium Tubulinosema ratisbonensis sooner than did females (56), despite a lower reported microsporidial load. However, the assertion made here that despite a greater lethality to T. ratisbonensis, males show a higher resistance as demonstrated by their lower pathogen load, needs further investigation. Pathogen load should be quantified before significant mortality has occurred in the population of infected individuals, otherwise, individuals with a high pathogen load who died prior to sampling would not be included in the analysis. A greater number of infected males had died 5 days post-infection from T. ratisbonensis than females, and pathogen load was analyzed 6 days post-infection (56). These results could potentially suggest that males are less tolerant, given that the only individuals still alive are those carrying a lower pathogen load.

Overall, in contrast to viral infections, males appear to survive longer than females in fungal infection models. There are indications that the magnitude of these dimorphisms in survival may be dependent on the environment. For example, the male-biased survival observed upon infection with B. bassiana is magnified by cold pretreatment of flies before infection, improving male survival at young and middle ages (81). Diet is also likely to influence susceptibility; for example, while females have decreased rates of survival to B. bassiana thoracic injection, immunity-induced metabolic declines were 50% greater in males (82). Supporting the interaction of metabolic state with sex-biased infection outcomes, polymorphisms associated with increased resistance to M. anisopliae inoculation are dimorphic and are biased toward gene networks regulating metabolism, as well as phagocytosis and cell migration (83). Furthermore, in the microsporidia infection model T. ratisbonensis, the quantity of circulating triglycerides was shown to affect parasite burden in females (56). The intersection of diet, metabolism and immune responses is likely to dictate outcomes to infection with most (or all) pathogens, as we discuss in more detail below. However, whether metabolic effects regulate resistance or tolerance to the fungal and microsporidial infection models described is as yet untested.

#### Bacterial Infection

The bacterial genus of Wolbachia and Spiroplasma have welldocumented interactions with host sex and have been extensively studied in Drosophila infection models. Both bacteria disrupt the reproductive biology of their hosts, and show sex-specific transmission. While these interactions are entirely dependent on host sex, these examples are not strictly relevant to examination of dimorphic immune responses, and have already

been extensively reviewed (47, 84–87), so we will not focus on them here.

and decreases in expression or behaviors are denoted by upward- and downward-facing arrows, respectively.

In models of systemic bacterial infection of male and female adults, where immune responses and survival are directly compared, survival to infection appears to be pathogendependent. Depending on the amount and type of peptidoglycan (sugar- and amino acid-based polymers) in their cell wall, bacteria can be categorized into two groups, Gram-positive (G-) and Gram-negative (G-). In Drosophila, the production of immune effectors such as antimicrobial peptides is under the control of the immune deficiency (IMD) and the Toll signaling pathways. The former responds to the meso-diaminopimelic acid (DAP)-type peptidoglycan of G- bacteria and certain G+ bacilli, whereas the Toll signaling pathway responds mainly to the lysine (Lys)-type peptidoglycan of G+ bacteria and fungal beta-1-3 glucan (88, 89). Activation of immune signaling pathways are not strictly dictated by cell wall type, however; for example, both Toll and IMD signaling pathways are activated by Staphylococcus aureus infection (90).

It is not possible to neatly attribute the direction of sexbiased survival to bacterial infection to one particular signaling pathway. Males appeared to be more resistant to systemic infections by the extracellular G- Providencia species Providencia rettgeri and Providencia alcalifaciens (23, 91). However, mortality to the obligate intracellular Coxiella burnetii is comparable in males and females (57), while males died more quickly than females when infected with Pseudomonas fluorescens (22), or the extracellular bacteria Serratia marcescens (although notably, this was genotype-specific) (58). Few studies have assessed the response of both sexes to G+ bacteria. Females were reported to be more susceptible to infection with Enterococcus faecalis but less to S. aureus (23) or to Lactococcus lactis (58). Thus far, laboratory models of bacterial infection have demonstrated both male- and female-biased survival with Gand G+ species (**Table 1**). Different laboratories have also reported opposite biases in response to the same pathogen. For example, Pseudomonas aeruginosa is reported to induce dimorphic survival that is either male- or female-biased (46, 55), or not significantly different (22). These contrasting results may be dictated by genotype, as is evidently the case for malebiased susceptibility to S. marcescens, which was observed in only two out of four genetic backgrounds tested in a recent study (58). These results may also be influenced by environmental conditions such as mating status, which appears to have an immunosuppressive effect on females (23, 92, 93). Another probable reason a clear pattern has not emerged is that the bacterial species used in these infection models are more diverse than the peptidoglycan dichotomy belies. Compared to P. rettgeri, S. marcescens is highly pathogenic to Drosophila, and its infection dynamic within the host is very different (23). Moreover, while these two bacteria are extracellular, other species, such as Coxiella, infect intracellularly, further complicating the comparison. Different pathogens require different host defense mechanisms, which may rely on particular immune tissues, signaling pathways, and terminal effectors such as AMPs. A more systematic comparison between male and female responses to a range of different bacteria with different modes of infection will help to decipher the role for immune signaling pathways in dimorphism.

A role for the Toll pathway in mediating sex differences in survival to some bacterial infections has recently started to emerge. It appears to be necessary for dimorphic survival to P. rettgeri, and loss of Toll signaling reverses the survival bias to E. faecalis (23). Loss-of-function mutants of both Toll-1 and Toll-7 differentially affected infection outcomes in males and females. Toll-1 mutant males and females were less resistant to E. faecalis challenge than wild-type controls, while loss-of-function Toll-7 mutants reduced male resistance to both E. faecalis and P. aeruginosa (55).

#### Behavioral Responses and Symptoms

In addition to dimorphic immune responses, Drosophila exhibit sex-specific behavioral symptoms and responses to infection. Grooming is thought to be an important behavioral defense against pathogenic infection, where flies remove potentially infectious microbes from their cuticle. Grooming in D. melanogaster is triggered by chemosensation of compounds, including pathogen components, by chemoreceptive sensilla. Specifically, the sensing receptor PRGP-LC contributes to grooming induction, connecting humoral immune sensing to behavioral responses (94). In optogenetic experiments targeting sensilla, grooming was more readily triggered in males than in females (94). Females were subsequently shown to rely more strongly on olfactory signals to remove cuticular B. bassiana than males, resulting in more conidia on the wings of olfactorydeficient female flies (95), but whether this contributes to the higherrate of survival by males after B. bassiana infection (49–53) remains to be seen. Drosophila exhibit dimorphic sleep responses to infection, a behavioral change which may be adaptive, or symptomatic. DCV causes females, but not males, to sleep more; however, DCV-infected male flies carrying Wolbachia are more lethargic when awake (61). Many other behaviors are likely to be impacted by symptoms such as lethargy and sleep alterations, including evasive behaviors and mating. For example, Wolbachia infection increases the recapture rate of females, but not males (96). Females previously exposed to DCV showed lower motivation to pick a food source when presented with a risk of encountering DCV (78). When disrupted, nemuri, an antimicrobial peptide that promotes sleep in D. melanogaster, reduced day-time sleep consolidation selectively in males (97), potentially linking dimorphic sleep behavior and responses to immune challenge. During a P. rettgeri systemic infection, females arrested egg-laying during the acute phase of the infection, until it stabilized into a chronic phase (98). The same observation has been made upon benign infections with Pectobacterium carotovorum (previously named Erwinia carotovora carotovora, or Ecc15) and Escherichia coli (99). The adaptive role of this behavioral response (99, 100) is not clear and is by definition female-specific. It remains to be determined if changes to reproductive behavior occur in males. Although research in this area is in its infancy, these studies suggest behavioral dimorphisms in D. melanogaster could be an important driver for sex differences in infection outcome.

#### SEX DIFFERENCES IN IMMUNE COMPARTMENT PHYSIOLOGY

Sex differences in outcome to infection are likely to be mediated by distinct immune compartments, where the key tissues involved will be dependent on pathogen and route of infection. Below we discuss those studies that assess the contribution of individual immune tissues. Sexually dimorphic immune physiologies in unchallenged flies and after acute infection are summarized in **Figure 2**.

#### Hemocytes

Hemocytes are specialized immune cells responsible for the encapsulation and phagocytosis of pathogens and dead cells. Few studies focusing on D. melanogaster hemocytes have reported the sex of individuals used; but those that have, present some evidence suggestive of a dimorphism in this branch of the immune system. First, it appears that hemocytes can have sexspecific functions. The Jun N-terminal kinase (JNK) signaling pathway regulates the decision between cell repair and cell death. As such, JNK/Basket is required in larval hemocytes to promote tissue maintenance, but only in males, such that hemocytespecific loss of JNK/Basket results in increased tissue damage in males after UV irradiation (101). In addition to this functional difference, there is some evidence for sex differences in total hemocyte number, but neither the direction of bias nor the drivers for the dimorphism can yet be concluded. Female white prepupae have been reported to contain a higher total number of hemocytes than males (102). In adults, higher numbers of hemocytes per se in females (23), and higher numbers of hemocytes per unit of hemolymph in males (21), have been reported, whilst a recent study found no effect of sex on adult hemocyte number (22). A subtly higher phagocytic index has been reported for male hemocytes ex vivo (22); however, sex differences in the functional roles of hemocytes in homeostasis or responses to infection are as yet unknown.

### Melanisation

Melanogenesis is an important feature of arthropod physiology. In addition to its role in cuticular hardening, via the synthesis of a specific melanin called sclerotin, and in cuticular coloration, via the synthesis of eumelanin, multiple studies have demonstrated its role in immune responses [see (103) for review]. An essential role for immune melanization in the efficient killing of encapsulated parasitoid wasp eggs and pathogens via cytotoxicity of reactive oxygen species (ROS) produced by the melanization cascade is well-appreciated (104, 105). The process hinges on cascades of serine proteases (SPs) triggered by either direct or indirect antigenic recognition and tissue damage (106). A key enzyme in this process is phenoloxidase (PO), which mediates the oxidation of the amino acid tyrosine to dihydroxyphenylalanine (DOPA) and subsequently, the oxidation of DOPA and dopamine to their respective quinones which are precursors of brown/black eumelanin. PO is produced as prophenoloxidase (PPO) proenzyme which is converted to active PO by a clip domain serine proteinase. This cleavage generates ROS, giving immune melanization its cytotoxic activity.

Whereas the biochemical pathways downstream of PO are well-characterized, our knowledge of the molecular events leading to PPO activation are largely unelucidated (103). The involvement of Toll signaling in melanization responses has recently been demonstrated in adults (106). Hayan and SP7, two SPs acting upstream of Spätzle, activate PPO1 and PPO2, which were shown to be essential for effective resistance against several systemic fungal and G+ bacterial challenges (106, 107). Mutants for PPO lack melanization, yet, maintained a dimorphic survival to P. rettgeri infection (23), suggesting that the activation of this immune response is not a driver of sex differences in survival to this particular pathogen. However, basal expression of some genes involved in melanogenesis are higher in males and a subset of these, including Dopa decarboxylase (Ddc) and yellow-f, respond transcriptionally to P. rettgeri infection (23). Alternative infection models where melanization is required to effectively control the infection may be more informative for understanding its potential roles in immune dimorphism. For example, the response of males and females to parasitoid wasp infection at larval stages has not been reported to our knowledge. Despite the importance of melanization for resistance to infection and injury, and the intersection of immunologic melanin production and Toll signaling, virtually no studies have assessed the sex-specific physiology of the response. Given that constitutive expression of Toll pathway genes, namely components upstream of Toll-1, as well as Ddc, is greater in males (23), the hypothesis that males are more poised for melanization warrants testing.

#### Systemic (Humoural) Immunity

A number of studies have reported sex differences in systemic immunity (**Figure 2**). Systemic immunity is primarily driven by the fat body; however, expression levels of pathway components were largely measured in whole individuals. For example, following systemic challenge with P. rettgeri, males induced a number of Imd- and Toll-regulated effectors at higher level than females, paralleling their greater survivorship (23). Although to note, ablation of the Imd pathway does not suppress the sexual dimorphism in survival (23). Sex differences have repeatedly been found in expression and function of Toll pathway components. Females are reported to exhibit higher expression of Toll-1, the transmembrane receptor that activates the Toll intracellular signaling pathway (55), the Relish protein dorsal (74) and the Toll pathway component ntf-2 (75). The Toll pathway is known to be involved in two processes: dorso-ventral embryonic patterning, a female specific process occurring in the eggs, and in immunity. Signaling downstream of the transmembrane receptor Toll-1 is shared between these two processes. Thus, it is unsurprising that the apparent dimorphism in Toll pathway gene expression disappears once expression in the ovaries is excluded (23). Nonetheless, the expression of the Toll pathway seems crucial for the sexual dimorphism of many infections. This includes both G+ bacterial infection but, perhaps surprisingly, also Gbacterial infection (23). This occurs through the activation of the pathway by Persephone, a hemolymphatic serine protease, which senses microbial proteases during infection (108, 109). Toll-3, also called MstProx, shows male-specific expression and response to infection, which seems to be attributable to the gonads (23, 72). Toll-5, or Tehao, is also induced at higher levels in males than females following P. rettgeri infection (23). Toll-7, a plasma membrane Toll receptor that binds to viral glycoproteins, is expressed at higher basal levels in males (23, 55) and appears to have an additional, male-specific isoform (73).

Male-biased survival to B. bassiana also appears to be dependent on the Toll pathway, since mutations in Toll pathway components ablate or reverse the dimorphism (53). The induction of drosomycin and metchnikowin in the first 24 h of infection also tended to be greater in males (52). The survival dimorphism, however, was not suppressed when tested on spaetzle (spz) mutants, a component of the Toll pathway (52). These Toll-regulated dimorphisms may in part be mediated by expression of AMPs: for example, attacins and diptericins require functional Toll for full inducibility. Supporting this, the loss of Toll affects the Enterobacter cloacae-induced expression of attacins and diptericins more strongly in males (110), and males with a gain-of-function Toll mutation exhibit higher levels of the AMPs Cecropin A1 (CecA1), Diptericin A (DptA), Attacin A (AttA), and Attacin B (AttB) in response to challenge (111). Furthermore, dimorphic induction of drosomycin in response to P. rettgeri infection is lost in spz mutants (23). Dimorphisms have also been reported in Imd pathway components, for example, several Imd-induced AMPs are found to be dimorphic in their expression levels. Using a construct with various diptericin promoter sequences upstream of the lacZ gene, higher levels of induction were seen in females than males (79). In contrast, diptericin has been shown to be expressed at higher levels in unchallenged males than females, and is more strongly upregulated upon infection with the G- bacterium, P. carotovorum (24).

The JAnus Kinase protein and the Signal Transducer and Activator of Transcription (JAK-STAT) pathway, required for antiviral immune responses and induced upon bacterial infection, also shows dimorphic expression. G9a, a histone H3 lysine 9 methyltransferase, regulates tolerance to viral infection by regulating JAK/STAT (112) but in a sex-specific manner, with females being more sensitive to a loss of G9a (113). In addition, the stress-responsive genes regulated by the Jak-Stat pathway, TurandotA, TurandotC, and TurandotM, were among the genes that were the most male-biased in an analysis of the transcriptional response to P. rettgeri infection (23), although the consequences of this are as yet unknown. Overall, males have a higher expression of many Toll- (and to a lesser extent Imd-) regulated genes. The reason for this dimorphism is unclear, but further work could investigate a potential link with the dual role for the Toll pathway in females. It is possible that the immune response via the Toll pathway in females is constrained by its consequences on egg development, a constraint which does not apply to males.

#### Epithelial Immunity

Comprising the third prominent arm of defense in D. melanogaster are the immune-reactive epidermal and epithelial barriers such as the cuticle, trachea, genitalia, and gut.

#### Cuticle and Trachea

Sex differences in defense against fungal inoculation of the cuticle with the entomopathogenic B. bassiana demonstrate femalebiased susceptibility (53, 114–116). Dimorphisms in cuticle integrity, or in immune responses of the cuticle or respiratory system could conceivably be underpinning this susceptibility. The trachea, which consist of airway epithelia and spiracles, is immunogenic, and responses may be activated by such cuticle inoculations. Studies that have assessed tracheal immunity in adults, thus far include only males (71, 117), rendering sexspecific responses within respiratory tissue another unexplored, potential contributor to dimorphic immunity. Dimorphisms in cuticular epithelial immune responses per se are also, to our knowledge, entirely unexplored.

#### Genitalia and Gonads

Male genitalia were found to be more primed for immune response to bacterial infection than that of females, where systemic and local AMP responses followed P. carotovorum inoculation of the genitalia in males (80). The female reproductive organs are also immune active, and were shown to constitutively express cecropin (66) and drosomycin (66–68), where drosomycin expression was found to be independent of Toll signaling in this tissue (67). Males exhibit constitutive expression of cecropin within the ejaculatory duct, independent of Relish signaling (68), suggesting that genital epithelia may circumvent classical pathways to activate AMPs. Immunogenicity of the reproductive tissues is evident from analyses of post-mating AMP responses in females (118, 119), and the antimicrobial protein transfer from male accessory gland and ejaculatory duct to the female (120). Andropin, an AMP unique to males, is also strongly upregulated in response to mating (69). RNA-seq analysis of P. rettgeri infected males and females, with and without gonads, illustrates the contribution made by reproductive tissue to systemic immune responses (23). Such comparative transcriptomics revealed male-biased defensin levels following infection to be gonad-dependent (23). It is an open question how much reproductive tissues contribute to systemic immunity; indeed it is unknown whether AMPs produced by the gonads are released into the hemolymph, or remain within the tissue.

#### Malpighian Tubules

Malpighian tubules (MT), epithelial organs dedicated to filtration and analogous to the mammalian kidney, are also immunogenic (121). Transcriptomic analysis of MT revealed differences in basal immune gene expression between the sexes (70); however, nothing is known about sexually dimorphic functions of MT or their contribution to immune dimorphisms.

#### The Intestinal Epithelium

The Drosophila gut is a major immune locus, responding to infection by producing AMPs and reactive oxygen species (122, 123). Infection-induced immune responses are observable in males (122); however, while a small number of studies have compared sex differences in gut metabolism (124), and physiology (24, 124–126), most work on intestinal immunity has focussed on females. Nothing is known about sex dimorphisms in AMP or ROS production, nor indeed survival, after oral infection. However, expression of several immunerelated genes have been reported. Nubbin, a transcription factor with two isoforms, nub-RB and nub-RD (127) regulates duration of immune responses within the gut. Nub-RD mutants showed chronic immune activation in females (128), whilst overexpressing nub-RB resulted in a similar phenotype, illuminating their antagonistic roles within immune signaling (127). Overexpressing nub-RB within enterocytes and subsequent oral infection with P. carotovorum lead to total and 70% death in males and females, respectively, within 24 h. Downregulation of nub-RB enhanced survival to challenge in males compared to controls, while having the opposite effect on females (127). Overexpressing nub-RB significantly reduced the lifespans of both sexes, conversely, males reared in germ-free conditions had a slightly enhanced median longevity compared to conventionally-reared counterparts (127). These data could potentially illustrate a greater susceptibility of males to immunopathology from both chronic immune activation and commensals. Nubbin isoform antagonism may also offer insight into dimorphic responses between the sexes, where expression of each may differ, however, the cited study only quantified isoform expression in males (127).

The luminal microbiome impacts intestinal immunity (123) as well as many other aspects of D. melanogaster physiology and behavior (129). This includes cellular immunity (130), and response to enteric viral infection (131). Studies have shown that the microbiota varies between the sexes in terms of load (24), composition (132), and effect on metabolic responses to diet (133). However, most studies assessing the gut microbiome have only analyzed females, and thus we have very little knowledge about the interaction of commensals with dimorphisms in intestinal immunity. Microbiota populations, at least of the two predominant genera Acetobacter and Lactobacillus, tend to increase over aging (134), and composition varies between the sexes, where aged males were reported to differ to a greater extent than aged females when compared to their younger counterparts (132). Aging is also associated with a loss of epithelial barrier integrity in the gut (76, 77, 135). This phenomenon is more pronounced in females (24, 76), and is paralleled by increases in systemic immune activation, as indicated by systemic AMP levels (76). Microbiota dysbiosis (77) and loss of a tricellular junctional protein, Gliotactin (135), were shown to precede such changes in females, and loss of an intestinal septate junction protein, Snakeskin, exacerbates barrier loss and causes early death in both sexes (136). These data highlight the link between microbiome dysbiosis, maintenance of a stable gut barrier, and systemic inflammation. More work is needed to understand the interaction of the dimorphisms involved; for example, the apparently conflicting observations that the male microbiome shows greater changes (132), but despite this, the male gut barrier appears to be more stable over age (24). It is well-known that the microbiome is modeled by diet, environment, and host genotype, so comparing physiological data with microbiome data from different labs may confound interpretation.

#### Aging and Immune Dimorphisms Aging and Inflammation

Aging is known to be accompanied by heightened expression of immune genes (37, 39, 137–139). However, cause and consequence are difficult to separate, and understanding the contributions made by age to altered immunity, and immunity to aging, is challenging. When comparing the transcriptional response to aging in whole flies, a strong sex-by-age interaction was observed (139). Of the total significant probe sets, the majority were biased toward males, while just over 25% were sexually antagonistic. Included within this male-biased and antagonistic set were immune-related genes, such as the AMP, defensin (139).

In an experimental evolution study on lines where late-life fertility, and indirectly longevity, was selected for over 35 years, decreased expression of immune genes was strongly associated with increased lifespan (137). Females of longer-lived lines exhibited greater realized immunity in the face of challenge with P. carotovorum, B. bassiana, E. faecalis, and DCV than controls (137). Modulation of Toll components had varying effects on male and female lifespan: knockdown of the negative regulator cactus had a significantly reducing effect on lifespan, especially in males, whereas Toll and spz knockdown enhanced lifespan in both sexes (137). Dif knockdown had opposing outcomes between the sexes, with females experiencing a slight extension to lifespan (137). However, in two other studies, Dif mutants exhibited enhanced lifespan, in the context of intrinsically shortlived background lines (50, 140). Given the sex-specificity of Toll pathway gene expression seen in young flies (23), it is perhaps unsurprising that responses to such modulations are dimorphic.

Age-related systemic inflammation, for example, high basal levels of AMPs and ROS, has been assumed to negatively impact lifespan. This is supported by studies in which AMPs, or Rel-family transcription factors controlling their expression, have been manipulated. Both systemic and fat body-specific overexpression of relish negatively affected male lifespan to a greater extent than that of females, and relish male mutants were marginally longer-lived than females (39). The selective knockdown of relish at mid-life stages in the fat body, however, significantly extended lifespan in males (39). Global overexpression of attacin A, cecropin A1, defensin, and metchnikowin had significantly deleterious effects on lifespan of both sexes, with overexpression of defensin having a greater impact on males (39), reminiscent of the increase in defensin expression in aged males (139). This potentially illustrates age-related immunopathology, where the tempering of hyperactivated immune pathways at mid-late life stages is protective. Intriguingly, clean injury of the cuticle has been shown to extend lifespan in males only, suggesting that a nonlethal wound initiates a response that has a hormetic effect that is particularly effective in males (141). The mechanism for this is unknown, but it is tempting to speculate that it could initiate an anti-inflammatory state through induction of immune regulators.

#### Immunosenescence

Relatively few studies have looked at immune function in both sexes over aging, particularly from a mechanistic standpoint. While survival following B. bassiana decreases over aging in both sexes (140), systemic and cuticular inoculate challenge with the entomopathogenic fungus was reduced in aged females compared to young, while males succumbed only to cuticle inoculation, exhibiting a reduction solely in barrier integrity (114). Age negatively affected the ability to survive an oral P. carotovorum challenge in males compared to females, despite their comparatively superior maintenance of intestinal barrier integrity (24). In addition, aged males fared worse in response to systemic E. coli (142), P. aeruginosa, and Bacillus thuringiensis (22) challenges. The studies described above do not investigate the mechanisms underpinning these age-related immune dimorphisms, nor indeed the tissue(s) responsible, except via route of infection (24, 114). In a study investigating the effects of age on hemocyte function, hemocyte numbers were shown to decrease selectively in females over aging (21), supported by a recent study that only examined females (143); although both sexes maintained ex vivo phagocytic capacity over age (21). In contrast, a recent study found no effect of age on hemocyte number (22). It is clear that the mechanistic underpinnings of sex-by-age interactions in efficacy of immune responses and autoinflammation are undefined, but could be hugely informative to our understanding of sex differences in age-related pathology and lifespan.

### EVOLUTION OF DIMORPHISM THROUGH NATURAL SELECTION

Males and females can have different life histories, which implies that they are exposed to different evolutionary pressures. Because the sexes share a genome, these different pressures can lead to sexual antagonism, which can potentially be resolved through the evolution of sex-specific regulation leading to phenotypic dimorphism. We explore this in the section below with reference to the immune system in Drosophila.

One possible driver of natural selection on dimorphism is unequal exposure to parasites in males and females. Males and females may differ in the habitats they occupy, their activity times, or their nutritional needs, for example. Males and females thus occupy separate niches and are potentially exposed to different parasites. Parasites can exert different selection pressures on males and females: the extent to which they do will depend on the extent of ecological divergence between the sexes. Knowledge of the ecology of wild Drosophila being relatively limited, it is difficult to define precisely which behaviors could cause such sex-specific selection. Females lay their eggs in rotting fruits, where they are necessarily exposed to a microberich environment. Whether males spend less time in these environments is not known. Lab studies have indicated that male and female Drosophila do not have the same optimal diet, and in diet choice experiments, make different nutritional selections in accordance with their role in reproduction (144–146). Males and females could potentially make different diet choices in the wild, and thus be exposed to different parasites.

A second possible driver of natural selection on dimorphism comes from the fact that the immune system allows resistance to parasites but also represents a cost: it requires a significant investment in terms of resources, and the activation of immune defenses can cause "collateral" damage [e.g., autoimmune reactions (147)]. From this it is predicted that hosts will evolve toward an immune response of intermediate intensity, and not toward a maximum response (148). In other words, hosts evolve within a framework of constraints corresponding to an evolving trade-off. But the terms of this trade-off are not necessarily the same for both sexes: Hamilton and Zuk proposed that the links between investment in immunity and life history traits are sexspecific (149). Under this assumption, it can be predicted that the optimal investment in immunity is not the same in males and females, and could even be antagonistic (150). This scenario is supported by evidence in Drosophila that both resistance and tolerance can be sexually antagonistic (46), that expression of immune genes is sexually dimorphic (27), and that the genetic architecture of many traits, including immunity, differ between males and females in Drosophila (151–153). This is supported by mutation accumulation experiments indicating that deleterious mutations do not have the same costs in both sexes (154– 156). Although the accumulation of spontaneous mutations in Drosophila has not yet been shown to have a sexually dimorphic cost on the immune response (46), it does appear have a sex-specific effect on fitness (156). If autosomal determination of immune traits are not clear-cut, the fact that Drosophila immune genes can be X- and Y-linked (40, 41, 157) is sufficient to expect that genetic structure can affect immune system evolution in a sex-specific manner. First, the Y chromosome is, by definition, strictly under selection in males and has been demonstrated to influence the immune response (41, 157). For example, only males that had a Y-chromosome introduced from a single wild population differed in their ability to defend against S. marcescens (41). Second, since fathers do not pass an Xchromosome to their sons, evolution by sexual selection acting on males would be much slower for traits largely influenced by the X chromosome than would selection on autosomally determined traits, and X-linked sexually antagonistic traits would more freely affect sexual dimorphism (158). Supporting this is the demonstration of X-linked variation in immune response phenotypes in Drosophila, including bacterial load and immune gene expression, in a set of 168 X-chromosome extraction lines (40). Many of the associations of genetic variation with immune phenotype acted in a sex-specific or sexually antagonistic manner, supporting the theory that sexually antagonistic variation may be more easily maintained on the X chromosome, and this can impact dimorphism (158).

#### Sexual Selection and Evolution of Dimorphism: The Hypothesis of the Susceptible Male

Sexual selection is that which operates on the ability to mate successfully. Two traits that generally evolve in parallel in each sex under the effect of sexual selection are choice by females and the ornaments of males. Ornaments are expensive secondary sexual characteristics that evolve as a result of the selection made by the choice of females. A difficulty with this principle is that females can only choose males based on a trait that demonstrates their vigor if that trait remains variable. For sexual selection to occur, therefore, a continuous source of heritable genetic variation must be present in the population. This source of variation prevents genes for vigor from fixation and the character of choice to be only affected by non-heritable environmental factors. Parasites of all kinds offer such selective variation in value because they can evolve quickly and dynamically with their hosts (149, 159). It is on this basis that the famous hypothesis of William (Bill) Hamilton and Marlène Zuk was born. They proposed that parasite selection imposes the evolution of female choice for infection-resistant partners, which in turn favors the evolution of sexual dimorphism in hosts. Thus, as they observed in birds, a negative correlation between ornaments and parasitic load can be found in males (149). Such correlation between expensive ornamental traits and immunity may be present in Drosophila. One of the most clearly dimorphic characteristics in Drosophila is the color of the cuticle, which is darker in males than in females. The process of darkening the cuticle requires the production of eumelanin, which is also involved in the encapsulation of parasites. This dual role suggests that coloration and immunity are two related characteristics (103, 160). Furthermore, as mentioned above, melanin is produced from dopamine, a neurotransmitter also known to be involved in aggression behavior between males (161), male courtship for females (162) and female receptivity (163). Dopamine and melanin production are thus located at a metabolic crossroads establishing a link between secondary sexual characteristics, mating behavior, and immunity. Other visible, dimorphic characteristics in D. melanogaster include bristle number (abdominal and sternopleural), and the presence of sex combs, specialized leg bristles in males which aid copulation. Surprisingly, these traits have not been used, to our knowledge, to study sexual selection, despite being very clearly exposed to females. A recent study reported that flies with a mutation in the yellow gene, which encodes a protein involved in the synthesis of eumelanin, fail at mating because insufficient melanization renders their sex combs non-functional for grasping and mounting females during copulation (164). Furthermore, determination of bristle number in males is connected to the process of melanization (165). This link is confirmed by genetic analyses that have shown that Ddc, an enzyme essential for dopamine synthesis, has a role in pigmentation (166), and also in the determination of the number of bristles (165). Interestingly, Ddc is expressed differentially in males and females, as expected, but its expression is modulated during a bacterial infection [see RNA-seq data for "Ddc" in (23)] (167). A mutation in the ddc gene could therefore have an impact on both male ornamentation and response to infection. Female Drosophila are able to choose their mates on the basis of phenotypic traits (168, 169), and while evolutionary theory offers several different explanations for female mate choice, most include the evolution of heritable attractive features in males (170). The Drosophila Genetic Reference Panel (i.e., DGRP lines) show that natural populations indeed bear variation at several positions in the coding sequence of Ddc (see genome browser at genome.ucsc.edu). In Drosophila, it is unclear how male ornaments (or attractive traits) trade-off with investment in immunity, but one could speculate that factors required for both cuticle patterning and immune responses could be limiting. Since males have a high selection pressure for access to mating (25), they might still invest in their costly ornament (e.g., cuticle, bristles) to the detriment of their immunity. In any case, the hypothesis that investment in immune melanization may affect secondary sexual traits, such as bristles/combs, remains to be tested.

Hamilton and Zuk have shown that the dimorphism of investment in immunity increases with the intensity of sexual selection. This observation can be explained by a direct negative effect of reproductive traits on those of immunity: this is the socalled immunocompetence handicap hypothesis (171). It occurs because the immune system can interact directly with the hormonal system in relation to sexual dimorphism (172, 173). Increased production of hormones may thus have a benefit for secondary sexuality in males (e.g., ornaments, or competitive traits such as muscle mass) but a deleterious effect on the immune response. Since males have a high selection pressure for access to mating, they might invest in competitive traits to the detriment of their immunity (174). The handicap hypothesis implies that the resource allocation trade-offs between immunity and other characteristics related to selective value are not the same for males and females (148, 175, 176), and leads to the conclusion that males should invest fewer resources in immunity than females. In males, the benefits of increased mating success through greater investment in sexually selected traits (or in costly behaviors) should offset the costs of disease-related reduction in lifespan. Females, on the other hand, would invest more in immunity than males to maximize their breeding time. This prediction extends Bateman's principle to immunity (25, 177). In Drosophila, the interaction between immunity and hormones has clearly been identified (32, 33, 93, 178). For example, females produce juvenile hormone (JH) to lay fertilized eggs, and JH directly suppresses the immune response (93). JH affects the antimicrobial peptide (AMP) Drosomycin most strongly, which responds mainly to G+ bacterial infection via the Toll-mediated immune pathway (32). Interestingly, as discussed above, a greater susceptibility of females to some infections is mediated by Toll signaling. This is suggestive that the sex-specific investment of females in egg production via the production of JH has a cost on their immune system, leading to sexual dimorphism in response to certain infections (23).

The previous hypotheses suggest that one sex should be superior than the other while facing infection. Although these hypotheses predict dimorphism per se, they cannot explain the inconsistencies in the direction it takes. Alternatively, sex-specific differences may arise from sex-specific changes in reproductive behavior in response to variation in fitness-limiting resources availability, and not from an intrinsically superior immune function (179). In such a case, sex-specific responses would be condition dependent. In D. melanogaster, availability of sexually receptive females is an important fitness-limiting resource. While males adjust their level of courtship in response to this resource, increased sexual activity reduces their immune functions, likely because of a reallocation of feeding time into the search for a mate (180). Food availability is, of course, a major fitness-limiting resource. When males and females were kept separate and given ad libitum food, immune function was maximized in both sexes and there was no sexual dimorphism in clearance of a benign infection with E. coli. However, when the fitness-limiting resources were scarce there was a sex-specific bias (179). When food was limited, females exhibited poorer clearance of the infection than males and, conversely, males with high sexual activity, despite abundant food, performed less well than females. Thus, if there are many reasons for dimorphism to evolve, there are many reasons for its direction to be plastic.

#### Obstacles to the Evolution of Dimorphism

If different selection pressures are applied to two subpopulations, there is every indication that they will evolve in different ways (181). However, males and females of a species are not subpopulations like any other, since their genomes must recombine in each generation. In each generation, therefore, a gene that is advantageous when expressed in a female can be transmitted from a mother to her son, in whom the expression of this gene could be disadvantageous. In summary, the evolution of a beneficial trait for one sex can be a burden for the other sex (182). In the previous example, the genetic origin of this burden is antagonistic pleiotropy: an allele beneficial to one sex has a negative effect on the other. This sexual conflict, highly investigated in Drosophila, can occur at one locus (intra-local conflict: a mutation at one locus has a positive effect for one sex but negative for the other) or at several (interlocal conflict: a mutation at one locus has a negative effect on one sex and selects a mutation at another locus that reduces the effectiveness of the first mutation) (26). This sexual antagonism could mean that the advantage obtained by one sex is exactly offset by the burden the other sex suffers. In other words, antagonistic pleiotropy could prevent the evolution of dimorphism. In a less extreme situation, if dimorphism manages to evolve despite pleiotropy, it is conceivable that neither sex will be able to adapt optimally.

### Dimorphism as a Resolution of Sexual Conflict

It is also possible that selection may favor sex-specific regulatory genes that result in an advantageous allele in females being expressed only in females (183). Pleiotropy would then be eliminated, and dimorphism could evolve without constraint. To our knowledge, there is no experimental work that has studied the role of this sexual conflict in the evolution of sex-specific regulation of the response to infections in Drosophila.

#### One Pathogen, One Rule

It is usually impossible to predict, for a given infectious disease, which sex will suffer the most. For example, in mammals, it is often said that males are the "susceptible" sex (184). Men are indeed more susceptible to leishmaniasis, malaria, or bilharzia; but they are more resistant to toxoplasmosis, amebiasis, and giardiasis (12, 185, 186). The complexity increases considering that pathogens can adapt specifically to the sexspecific characteristics of the host. If a parasite is more often exposed to one sex, it is expected to adapt to that sex it encounters most frequently (187). Under such a scenario, even in the absence of sexual dimorphism in response to the infection, the parasite is expected to behave differently and to consequently induce a dimorphism in symptoms or virulence (15, 188). Attempts to generalize, through meta-analyses, are not consistent: they sometimes even imply that there is no real dimorphism (189). If the selection pressures applied to both sexes vary considerably from one parasite to another, we cannot exclude that the effects of the different parasites cancel each other out, so that the average selection pressures applied by all parasites ultimately lead to the same investment in immunity in males and females. Given that different parasites require different methods of control, this seems an unlikely outcome; instead, varied selection pressures, applied differently to each sex, requiring distinct responses, could result in a complex array of context-dependent dimorphisms. In Drosophila, reported directions of dimorphism in infection outcome are condition- and pathogen-dependent (**Table 1**), thus it seems inappropriate to propose a generality on the direction of immune sexual dimorphism (23, 40). However, even if the direction is difficult to predict, it is rare that dimorphism does not occur: in that sense, dimorphism in infection outcome seems to be the rule and not the exception.

#### Environmental Effects on Immune Dimorphism

As if pathogen-dependent dimorphism was not complex enough, infection almost certainly interacts with environmental factors to shape immune dimorphisms. Nutrition modifies the response to infection (190, 191), and where choice is offered, flies will change their feeding preferences in response to infection (192). Diet choices have been empirically demonstrated to increase fitness in insects (193), e.g., in a "true fruit fly" (Bactrocera tryoni)—S. marcescens infection model, where infection-induced selection of increased carbohydrate-to-protein ratio led to better survival (194). Baseline and infection-induced sex differences in nutrient preference may influence outcome to infection, as may dimorphic responses to macronutrients. For example, in a D. melanogaster—Vibrio cholerae infection model, increased levels of glucose selectively reduced lifespan in females, but delayed their succumbing to infection, whereas it had no effect on males (195), suggesting that while females suffer long-term consequences of chronically elevated insulin signaling, acute responses to pathogenic challenge, at least by V. cholerae, are enhanced by high dietary glucose (195). Nutrient-sensing pathways are known to regulate immune responses, among many other physiological responses to the environment (196), and perturbations in these pathways are shown to affect sex-differential gene expression in Drosophila, including expression of immune, defense, and stress response genes (146, 197). Thus, it is essential that we view diet, nutrient-sensing, and sex as interacting factors when considering immune dimorphisms.

Other biotic (density, competition), and abiotic (temperature, humidity) factors are likely to play into dimorphic outcomes to immune challenge. Indeed, in the wild, environmental inputs come as combined, covariate packages, and it is in the context of these combined inputs that Drosophila has evolved. Higher-order, systemic signaling pathways such as IIS/mTOR may integrate these inputs to produce phenotypically plastic responses that match current environmental states (196). IIS/mTOR has been empirically demonstrated to regulate immune responses in D. melanogaster (198, 199), suggesting that IIS/mTOR could represent a nexus for integrating sex differences in responses to environmental variation, including immunity. Among the biotic factors that have gained much attention recently is the microbiome. Diet, of course, contributes to modeling the gut microbiome and its sex-specificity might shape dimorphic bacterial communities. Thus, microbiotainfection interactions (200, 201) have the potential to shape immune dimorphisms. Microbiome sex differences have been demonstrated in Drosophila (132), and likely interact with both dietary choices (133) and responses to infection, therefore it follows that a full understanding of immune dimorphisms cannot be achieved without considering natural poly-microbial interactions.

#### The Ubiquitous Endosymbiont Wolbachia as a Troublemaker

Wolbachia is a ubiquitous endosymbiont predominantly transmitted by mothers to their offspring. This reproductive parasite is known to affect immune characteristics such as phenoloxidase (PO) activity (202), phagocytosis by hemocytes (203), and to affect the outcome of other infections, such as providing resistance to viral infection (204). While not the focus of the study, Martins et al. (60) showed that the level of resistance provided by Wolbachia status depends on the sex of the host. Furthermore, Wolbachia protects males more than females against enteric infection by the bacterium P. aeruginosa (205). For these reasons and because it is so common, we expect that Wolbachia infection could influence the direction of immune dimorphism, and even impact the evolution of the dimorphism itself.

#### Regulation of Immune Dimorphism

Immune dimorphisms in D. melanogaster arise from a shared genome, and must be a product of sex-specific gene regulation, however, this is not yet well-resolved. At the top of the regulatory hierarchy are genomic differences, namely sex chromosome karyotype. As we have discussed, there are clear X- and Y-linked effects on immunity in D. melanogaster. Downstream of karyotype, dimorphisms can be regulated by the sex determination pathway, a cascade of splicing factors regulated by X chromosome number that ultimately lead to the expression of sex-specific transcription factors (206). This pathway has recently been shown to have "non-canonical" routes of signaling in some tissues (125), and importantly, to exert profound effects on physiology in larvae and adults beyond its purely developmental role (20). However, the potential regulation of immunity by sex-specific pathways is as yet untested [but see (125) for transcriptional regulation in the intestinal epithelium].

Drosophila, and insects in general, demonstrate that sexual dimorphism can occur without the presence of sex steroids per se. Notably, insects produce the steroid hormones ecdysone and juvenile hormone (JH), and as discussed, these are important regulators of immune function. It is possible that hormonal and genetic regulation interact; for example, via sex differences in hormone production or receptor expression (207). One potential source for dimorphic hormone production are the gonads, which may exert influences on immune responses in closely opposed organs, as has recently been demonstrated for regulation of intestinal carbohydrate metabolism by the testes (124). The tractability for genetic manipulation of sex-specific transcription factors and hormone production/reception is a major advantage to using D. melanogaster to better understand sex-specific regulation of immune function.

#### CONCLUSION

We understand relatively little about immune dimorphisms in Drosophila, despite their apparent prevalence and magnitude. A real gap in our knowledge is the physiological and mechanistic underpinnings of the male-female differences in survival to infection that have been widely reported. We know almost nothing about how sex dimorphisms are shaped over the life course, including during development, or the influence of dimorphic immunity on aging and vice versa. Dimorphisms may arise from fixed differences in gene expression and tissue function, or from more plastic mechanisms that respond to environmental variables, or both. One thing is certain: the interaction between sex and immunity is complex. Complexity does not lend itself to generalities, and therefore we must be cautious in stating rules about male or female responses. Sex is clearly an essential factor that must be considered in the interpretation of data arising from studies into immunity, where the ideal approach is to include both sexes wherever possible. We argue that not only will including both sexes in studies of Drosophila immunity give a more complete picture, it will offer valuable insight into fundamental mechanisms underpinning innate immunity and responses to infection, and an understanding of the factors that drive dimorphisms to arise.

#### REFERENCES

1. Jin W, Riley RM, Wolfinger RD, White KP, Passador-Gurgel G, Gibson G. The contributions of sex, genotype and age to transcriptional variance in Drosophila melanogaster. Nat Genet. (2001) 29:389–95. doi: 10.1038/ ng766

## AUTHOR CONTRIBUTIONS

RB, M-KC, DD, and JR researched and wrote the manuscript. DD and JR planned and edited the manuscript. RB and M-KC performed the literature survey and made the figures, graphs and tables.

#### FUNDING

M-KC was supported by funding to JR from the University of Edinburgh; RB was supported by a Darwin Trust Scholarship; DD was supported by the French Laboratory of Excellence project TULIP(ANR-10-LABX-41; ANR-11-IDEX0002-02).

#### ACKNOWLEDGMENTS

We thank the reviewers for their insightful and constructive comments on the manuscript.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.03075/full#supplementary-material

Figure S1 | Quantity of articles from "both" category with possible tag interactions. Total articles per tag are shown next to each tag title. If a tag is applied, it is indicated with a black dot. Black lines connecting dots indicates conditions where two or more tags were applied to the same article. Quantity of articles in each category are shown above the respective bar.

Table S1 | Complete analysis of 5,626 articles retrieved from webofknowledge.com search. The Boolean search was: (TS = ((infection OR immunity OR hemocyte OR imd OR toll) AND drosophila) NOT TI = (bug OR bumblebee OR shrimp OR damesfly OR mollusc OR crab OR squid OR beetle OR baculovirus OR ant OR monochamus OR dastarcus OR cockroach OR crickets OR gryllus OR bemisia OR armyworm OR spodoptera OR mussel OR galleria OR helicoverpa OR amphibian OR manduca OR bee OR honey OR bactrocera OR tenebrio OR zebra OR dugesia OR flesh OR Apis OR house OR glossnia OR jelly OR Andrias OR dragonfly OR pachydiplax OR termite OR leech OR stick OR rhynchophorus OR rhodnius OR pardosa OR plutella OR coleoptera OR zophobas OR glossina OR ceratitis OR suzukii OR diabrotica OR rootworm OR sheep OR whitefly OR bird OR branchiostoma OR lizards OR laodelphax OR ceratopogonidae OR crassostrea OR oyster OR artemia OR freshwater OR calliphoridae OR phytomonas OR acyrthosiphon OR aphid OR crustacean OR parhyale OR hippocampus OR seahorse OR anopheles OR protaetia OR sea OR litopenaeus OR copepod OR swine OR planthopper OR arabidopsis OR circulifer OR leafhopper OR apostichopus OR cryptolaemus OR clam OR paphia OR mollusk OR achaea OR castor OR musca OR salmon OR dog OR echinococcus OR hetaerina OR sarcophaga OR fleshfly OR bovine OR zygoptera OR calopterygidae OR coenagrionidae OR scorpion OR locusta OR harpalus OR culex OR scylla OR firefly OR honeybees OR antheraea OR penaeus OR trichinella OR prawn OR macrobrachium OR ostrinia OR arge OR magnaporthe OR Phaeotabanus OR palm OR ostrinia OR daphnia OR scallop OR Chlamys OR Biomphalaria OR pig OR Anostostomatidae OR Orthoptera OR crayfish OR procambarus OR Platynereis)).


Longevity. Cham: Springer International Publishing (2015). p. 83–104. doi: 10.1007/978-3-319-18326-8\_4


expression that increases the tolerance to gut microbiota. BMC Biol. (2013) 11:99. doi: 10.1186/1741-7007-11-99


parasitoid resistance in Drosophila. Oikos. (2010) 119:533–41. doi: 10.1111/j.1600-0706.2009.18001.x


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

Copyright © 2020 Belmonte, Corbally, Duneau and Regan. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Viral Infection and Stress Affect Protein Levels of Dicer 2 and Argonaute 2 in *Drosophila melanogaster*

Alessandro Torri, Vanesa Mongelli † , Juan A. Mondotte† and Maria-Carla Saleh\*

Viruses and RNA Interference Unit, CNRS Unité Mixte de Recherche, Institut Pasteur, Paris, France

The small interfering RNA (siRNA) pathway of Drosophila melanogaster, mainly characterized by the activity of the enzymes Dicer 2 (Dcr-2) and Argonaute 2 (Ago-2), has been described as the major antiviral immune response. Several lines of evidence demonstrated its pivotal role in conferring resistance against viral infections at cellular and systemic level. However, only few studies have addressed the regulation and induction of this system upon infection and knowledge on stability and turnover of the siRNA pathway core components transcripts and proteins remains scarce. In the current work, we explore whether the siRNA pathway is regulated following viral infection in D. melanogaster. After infecting different fly strains with two different viruses and modes of infection, we observed changes in Dcr-2 and Ago-2 protein concentrations that were not related with changes in gene expression. This response was observed either upon viral infection or upon stress-related experimental procedure, indicating a bivalent function of the siRNA system operating as a general gene regulation rather than a specific antiviral system.

#### *Edited by:*

Dan Hultmark, Umeå University, Sweden

#### *Reviewed by:* Hai-peng Liu,

Xiamen University, China Ioannis Eleftherianos, George Washington University, United States

> *\*Correspondence:* Maria-Carla Saleh carla.saleh@pasteur.fr

†These authors have contributed equally to this work

#### *Specialty section:*

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

*Received:* 26 November 2019 *Accepted:* 14 February 2020 *Published:* 04 March 2020

#### *Citation:*

Torri A, Mongelli V, Mondotte JA and Saleh M-C (2020) Viral Infection and Stress Affect Protein Levels of Dicer 2 and Argonaute 2 in Drosophila melanogaster. Front. Immunol. 11:362. doi: 10.3389/fimmu.2020.00362 Keywords: RNA interference, insect immunity, gene regulation, viral infection, antiviral response, protein regulation

### INTRODUCTION

RNA interference (RNAi) is a defensive and gene regulatory process based in sequence homology among nucleic acids (1–3). Small RNAs (sRNAs) are produced and used as guides to target complementary DNA or RNA sequences (4, 5). Three main sRNA pathways are described to date, differing in the origin and biogenesis of the double-stranded sRNAs and their molecular function: the micro RNA pathway (miRNA), the small interfering RNA pathway (siRNA), and the Piwi-interacting RNA pathway (piRNA) (6).

In the siRNA pathway, the ribonuclease Dicer 2 (Dcr-2) recognizes and dices double stranded RNA molecules of exogenous (virus) or endogenous (cellular) origin, producing 21-nucleotide length siRNA duplexes that are loaded into the protein Argonaute 2 (Ago-2) within the RNAinduced silencing complex (RISC). Once loaded, the siRNA duplexes are unwound and only one RNA strand is used by Ago-2 to target and slice the complementary RNA. Two Dcr-2 cofactors are indispensable for siRNA production and correct loading into RISC: LOQS and R2D2 (4, 6).

The siRNA pathway is considered the most ancient and at the origin of the RNAi phenomenon. Its main components (Argonaute-Piwi, Dicer-like, and RNA-dependent RNA Polymerase proteins) are supposed to have been already present in the last common ancestor of eukaryotes (7). In Drosophila melanogaster the siRNA pathway acts as the main antiviral immune response (8) and is also involved in somatic defense against transposons (9). However, the fact that several siRNAs target cellular mRNAs (7, 10) and repress the expression of specific genes (11), suggests that it may also play a role in the regulation of gene expression.

Much effort has been dedicated to studying the RNAi mechanism and pathways, but despite several advancements, not much is known about their regulation (12). Previous studies reported that the mRNA expression of the core components of the siRNA pathway (Dcr-2 and Ago-2 among others) are induced by double stranded RNA (dsRNA) in Acyrthosiphon pisum (13), Manduca sexta (14), and Blatella germanica (15) through a yet unknown mechanism. This induction has also been shown upon viral infection in Bombus terrestris (16), Apis mellifera (17) and D. melanogaster (18). However, in Apis mellifera this phenomenon seems to be virus-specific (19) and in D. melanogaster the induction was observed upon injection of Zika virus, an arbovirus for which the fruit fly is not a natural host (18). In addition, a study in D. melanogaster (20) showed increased levels of Ago-2 and Dcr-2 in flies constitutively expressing an active form of dFOXO, establishing a link between stress response and RNAi regulation. However, knowledge on regulation, stability, and turnover of the siRNA pathway core genes and proteins remains scarce.

Here we explore whether the siRNA pathway is regulated at the transcriptional and/or at the translational level following viral infection in D. melanogaster. We analyzed the expression of transcripts and proteins for Dcr-2 and Ago-2 in three different fly strains infected with Drosophila C Virus (DCV) and Flock House Virus (FHV) by two different modes of delivery, injection and oral infection. Our results show a complex and previously undescribed mechanism of regulation of the siRNA pathway at the protein level independent of fly strain, gene expression and mode of infection.

## MATERIALS AND METHODS

### Fly Strains and Husbandry

The D. melanogaster fly lines used were the following: w 1118 , Oregon-R, and yw. Fly stocks harbor the sensitive allele of Pastrel 3L:7350895 (Thr). Flies were reared on a standard cornmeal diet (Bloomington) at a constant temperature of 25◦C and kept under a 12:12 photoperiod. All fly lines were cleaned of possible chronic infections (viruses and Wolbachia). In addition, fly stocks were analyzed by RT-PCR with pairs of primers specific for CrPV, DAV, DXV, DCV, FHV, and NoraV to confirm that they were not persistently infected by these viruses.

### Virus Production and Titration

DCV stock was prepared in w <sup>1118</sup> flies. Flies were injected intrathoracically with 500 TCID50 per fly. When mortality started, flies were anesthetized and squashed in PBS (3 flies per 100 µl of PBS). The extract was frozen at −80◦C, thawed and centrifuged for 15 min at 15,000 × g at 4◦C. The supernatant was recovered and filtered to eliminate bacteria, aliquoted, and stored at −80◦C.

FHV stock was prepared on low-passage S2 cells. When the cytopathic effect started, the supernatant was harvested and centrifuged.

Both stocks were titrated in S2 cells. Titers were measured by end-point dilution method and expressed as 50% Tissue culture Infective Dose (TCID50). DCV stock: 1,18 × 10<sup>10</sup> TCID50/ml, FHV stock: 5 × 10<sup>9</sup> TCID50/ml).

### Viral Infections

Injections: flies were injected intrathoracically using a nanoject (Nanoject II apparatus; Drummond Scientific) with 50 nL of a viral suspension of 10 TCID<sup>50</sup> of Drosophila C virus or 100 TCID<sup>50</sup> of Flock house virus in 10 mM Tris, pH 7. An injection of the same volume of 10 mM Tris, pH 7 served as a mock-infected control. Infected flies were kept at 25◦C and changed to fresh vials every 2 days.

Oral infections: flies were starved during 5 h in an empty tube. Then, flies were transferred to a tube containing a Whatman filter paper in the bottom embedded in a mix of viral stock in PBS (10% viral stock, 35% sucrose and 2% of blue dye). After 16 h, only the flies having a blue-belly (corresponding to blue dye in the gut due to ingestion) were placed in new media tubes, kept at 25◦C and changed to fresh vials every 2 days.

### General Experimental Design

For all the experiments, 4- to 7-day old adult female flies were used. For each experimental condition, flies were divided in two pools: one pool was used for viral titration and RNA extraction, and the other one for protein extraction. For every experiment presented, the analyses were based on three biological replicates. In **Figure 2**, 16 biological replicates were performed (except w 1118 in **Figure 2B**, n = 15). For western blots, only biological replicates were used. For RT-qPCR, three technical replicates per condition and per biological replicate were used.

### Survival Assays

Mortality of infected flies was measured daily by counting the number of dead flies in each test tube. Three biological replicates of 60 flies each were done per condition. Fly mortality at day 1 was attributed to damage invoked by injection and/or manipulation procedure, and excluded from further analyses.

## RNA Extractions and RT-qPCR

For each time point and condition, total RNA was extracted from a pool of 4–12 flies depending on the biological replicate. Each pool was homogenized in 300 µl of PBS and 100 µl were used to perform RNA extractions using TRIzol reagent (Invitrogen).

The first-strand cDNAs were produced from 400 ng of RNA using Maxima H Minus First Strand cDNA Synthesis Kit with dsDNase (Thermo Scientific) according to the manufacturer's instructions. For each sample, a negative control without the reverse transcriptase enzyme was performed, in order to check for potential genomic DNA contamination. Roche Universal Sybr Green Master Mix (Rox) was used for qPCR. The sequences of the primers used were:

Ago-2 F primer: 5′ -GTGGTTTACACGCCTCCTCA-3′ Ago-2 R primer: 5′ -GGGTAGTTGCGACTGTGGAA-3′

#### Dcr-2 F primer: 5′ -GGGTGAACAGGGAGTGGATG-3′ Dcr-2 R primer: 5′ -CAAAAAGACCTGGGCTGTGC-3′

Quantification was normalized to that of mRNA encoding the endogenous ribosomal protein Rp49 as previously reported (21). Data were calculated using the 11Ct method to compute relative gene expression. For each sample, 3 technical replicates plus 1 RT negative control were included in the qPCR plate. qPCR was performed in 384-well plates with a final volume of 10 µl with QuantStudio 7 Flex Real-Time PCR System (Applied Biosystems). The following program was used: Hold stage 50◦C for 2 min, 95◦C for 10 min. PCR stage: 40 cycles of 95◦C for 15 s, 60◦C for 1 min. A melt curve to confirm the specificity of the reaction was performed.

#### Protein Extraction, Western Blot, and Protein Quantification

For each time point and condition, total proteins were extracted from pools of 4–8 flies, depending on the biological replicate, using 200 µl NP40 Buffer: 20 mM HEPES-KOH buffer pH 7.5; 100 mM KCl; 5% Glycerol; 0.05% NP40; 1 mM DTT (freshly added) and 1x complete, EDTA-Free Protease Inhibitor Cocktail (Roche) (freshly added). Each pool was homogenized with a pestle, centrifuged at 12,000 × g for 10 min and the supernatant was recovered and stored at −80◦C. Ten microliters of each protein extract were ran in SDS-PAGE using Mini-PROTEAN TGX Stain-free gels 4–20% (BIO-RAD) according to the manufacturer's instructions. The gels were then activated for image acquisition using Molecular Imager Gel Doc XR+ (BIO-RAD) and the transfer was performed using Trans-Blot Turbo Transfer Pack nitrocellulose membranes and Trans-Blot Turbo Transfer System (BIO-RAD). The image acquisition of the total amount of proteins transferred was performed with Molecular Imager Gel Doc XR+ (BIO-RAD). For the immunoblots, the following primary antibodies were used: Anti-Dcr-2 Abcam AB4732 (1:1000); Anti-Ago-2 Siomi 9D6 (non-commercial antibody, 1:15) kindly provided by Haruhiko Siomi. The secondary antibodies were HRP-linked: anti-rabbit GE Healthcare NA9340V (1:10000); anti-mouse Abcam AB6728 (1:10000). The antibodies were diluted in a solution of PBS-BSA 3%, TWEEN 20 0.3%. Membranes were incubated overnight at 4◦C with the primary antibodies, washed

time point was compared to each other by two-way ANOVA followed by Sidak's multiple comparisons test (\*P < 0.05; \*\*\*P < 0.0002).

FIGURE 2 | Basal expression levels of core RNAi proteins in Oregon-R, w<sup>1118</sup> , and yw flies determined by western blot (quantification and normalization are detailed in Materials and Methods). (A) Dcr-2 basal levels. (B) Ago-2 basal levels. Each dot represents a biological replicate composed of a pool of 5 flies. A.U. indicates arbitrary units. \*Indicates significant differences among Drosophila strains using Mann–Whitney test (P = 0.0106).

3 times for 5 min with PBS-TWEEN 20 0.3%, incubated with secondary antibodies 1 h at room temperature and washed again 3 times for 5 min with PBS-TWEEN 20 0.3% before adding the ECL (SuperSignal West Pico Plus Chemiluminescent substrate—Thermo Scientific). The image acquisitions were performed with MyECL Imager (Thermo Scientific). The intensities of the bands corresponding to Dcr-2 and Ago-2 were normalized with the total amount of protein in their respective lanes using ImageStudioLite (LI-COR Biosciences) (**Supplementary Figures 1**, **2**).

In order to compare results from different experiments, two types of normalization were employed. To study differences in the relative protein expression, all the data were normalized with the mock-infected time point 0 (**Figures 5**, **6**). To analyze non-relative protein levels, all data were corrected for the experimental effect as previously shown (21). This procedure transforms the raw values into their deviation from the experimental mean, and the resulting adjusted values are centered on zero (**Figures 2**, **7**).

#### Statistical Analysis

All statistical analyses were performed with Prism 8—GraphPad. The following statistical tests were used: for grouped tables with two grouping variables, two-way ANOVA followed by Sidak's multiple comparisons test; for table with one grouping variable, the Mann-Whitney test; for survival curves, the log-rank (Mantel-Cox) test; to calculate the probability of deviations from a theoretically expected distribution, the binomial test.

### RESULTS

### Different *D. melanogaster* Strains Differ in Their Susceptibility to Viral Infections

To study the regulation of the siRNA pathway during viral infections we used different D. melanogaster strains, viruses and modes of infection in order to establish the commonality of the response. We challenged w 1118 , Oregon-R, and yw D. melanogaster strains in two different infection conditions: (1) injection of Drosophila C Virus [DCV; (+)ssRNA Dicistroviridae], a natural Drosophila pathogen; and of Flock House Virus [FHV; bisegmented (+)ssRNA, Nodaviridae], a non-natural fly pathogen used as a control; and (2) oral infection by feeding with DCV. Survival curves showed that after DCV infection by injection, w <sup>1118</sup> flies had an intermediate survival rate compared to yw flies that were more susceptible, whereas Oregon-R were more resistant (**Figure 1A**). Injection of DCV resulted in the death of all flies at 8 days postinfection (dpi) for w <sup>1118</sup> and 4 dpi for yw flies, while 97% of Oregon-R flies died at 9 dpi (**Figure 1A**, left panel). In flies orally infected with DCV, the proportions of mortality 15 days after infection were of ∼25% for w 1118 , ∼50% for yw, and ∼6% for Oregon-R strains (**Figure 1A**, middle panel). Upon FHV injection, yw were still the most susceptible flies, with ∼35% mortality at day 6 compared with ∼9% for w 1118 and ∼15% for Oregon-R (**Figure 1A**, right panel). We then

measured viral loads by TCID<sup>50</sup> for all tested conditions at 0 dpi (15 min post-injection or 16 h post-feeding) and at 3 further times post-infection (**Figure 1B** and **Supplementary Table 1**). Following virus injection, both DCV and FHV reached higher titers in yw flies than w <sup>1118</sup> and Oregon-R flies at 2 dpi and 3 dpi, respectively, while DCV accumulated to higher levels at 3 days post-injection in w <sup>1118</sup> flies compared with Oregon-R (**Figure 1B** and **Supplementary Table 1**). However, no significant differences in DCV titers were found between the different fly strains during oral infections (**Figure 1B** and **Supplementary Table 1**). Taken together, these results show that these fly strains differ in their susceptibility to viral infections and that there is a correlation between the increase of viral loads and mortality (**Figure 1B** and **Supplementary Table 1**), indicating that flies are dying due to the physiological burden imposed by the infection.

#### Basal Levels of Dcr-2 and Ago-2 Proteins Do Not Account for the Variation in Susceptibility to Viral Infection

We next asked if the differences in susceptibility and viral accumulation observed between w 1118 , Oregon-R and yw flies might be due to differences in the basal levels of the siRNA pathway core proteins, Dcr-2 and Ago-2, before viral infection. To address this question, we measured the levels of these proteins in the three Drosophila strains by western blot. Dcr-2 basal levels were similar between strains (**Figure 2A**). A significant increase in Ago-2 basal levels between w <sup>1118</sup> and Oregon-R (**Figure 2B**) was observed, but it cannot explain the difference in susceptibility to virus infection observed for these strains. Taken together, these results show that the basal levels of Dcr-2 and Ago-2 do not play a critical role in fly survival and in the control of viral infections.

arbitrary units. n = 3; mean ± SD, two-way ANOVA followed by Sidak's multiple comparisons test (\*P < 0.05, \*\*P < 0.01, \*\*\*\*P < 0.0001).

## Viral Infection Does Not Change *Dcr-2* and *Ago-2* Gene Expression Levels

As we did not observe a correlation between Dcr-2 and Ago-2 proteins basal levels and infection outcomes in the D. melanogaster strains tested, we wanted to explore if a difference in gene expression would be noticeable as a result of the viral infection. We evaluated the mRNA expression levels of Dcr-2, Ago-2, and rp49 (housekeeping gene) by qPCR at four time points post-infection. The mRNA was extracted from the same pools of flies used in **Figure 1B**. We did not observe any significant difference in Dcr-2 (**Figure 3**) and Ago-2 (**Figure 4**) mRNA levels of infected flies compared with mock infected flies, independent of the fly strain, virus or mode of infection used.

### Viral Infection Induces an Immediate Change in Dcr-2 and Ago-2 Protein Levels

To explore if mRNA expression levels correlate with proteins levels upon viral infection, we quantified the expression of Dcr-2 and Ago-2 proteins. Proteins were extracted from flies that were biological replicates to **Figures 1B**, **3**, **4** and the presence and quantity of Dcr-2 and Ago-2 were analyzed by western blot. After normalization with the time point 0 of the mockinfected condition (see Materials and Methods for details on normalization protocol and **Supplementary Figures 1**, **2**), Dcr-2 levels showed a relative increase in virus infected flies at the time point 0 (15 min post-injection or 16 h post-feeding) independently of the fly strain, virus and mode of infection used (**Figure 5**). These increments were statistically significant for w <sup>1118</sup> injected with DCV (**Figure 5A**), for yw injected with FHV (**Figure 5B**), and for Oregon-R and yw orally infected with DCV (**Figure 5C**). These significant increases in Dcr-2 protein expression were not observed throughout all the conditions; however, in the 27 western blots performed (3 biological replicates × 3 fly strains × 3 different modes of infection), 26 of them showed higher Dcr-2 levels in infected flies than in the mock infected control. The probability that this phenomenon is due to a random effect is only 0.00002% (binomial test assuming that, during a random phenomenon, the probability to be higher or lower with respect to the control is the same). On the contrary, at the last time point (3 days post-infection for injection and 6 days post-infection for feeding) we found that Dcr-2 protein relative levels in infected flies were lower compared to the mockinfected flies. This decrease was statistically significant for yw injected with FHV (**Figure 5B**), and for Oregon-R and yw orally infected with DCV (**Figure 5C**). In injected flies, we found this pattern in 14 of 15 experiments with a probability of 0.004% for this phenomenon to be due to a random effect. In orally infected flies, we found this pattern in 9 of 9 experiments, with a probability of 0.2%. Due to the limited amount of anti-Ago-2 antibody in our possession (non-commercial antibody), we chose to perform the analysis for Ago-2 protein levels only in the w <sup>1118</sup> strain. We observed the same trend as for Dcr-2, but without significant differences (**Figure 6**). From a total of 7 western blots performed, we found higher relative levels of Ago-2 7 times for the time point 0, with respect to mock infected flies, with a probability of 0.78% (binomial test) that this is due to a random effect.

Altogether, the results show that Dcr-2 and Ago-2 protein levels change promptly upon virus infection.

#### Viral Presence, as Well as Infection Procedure Related-Stress, Induce a Change in Dcr-2 Protein Accumulation

The data shown in **Figure 5** correspond to Dcr-2 protein levels relative to the mock-infected time point 0. Therefore, we cannot differentiate if the change observed is due to an absolute Dcr-2 increase after the infection or to an absolute Dcr-2 decrease in the control flies due to the stress caused by the experimental procedure (injury in the case of injection, and starvation in the case of feeding—see modes of infection in Materials and

Methods). To answer this question, we compared the absolute protein levels for each time point with the basal level of the non-injected control flies from **Figure 2**. Since we observed the same effect on DCV and FHV injected flies, we pooled the data of all infected flies. We then analyzed the absolute levels of Dcr-2 among strains for each time point. As expected, we did not observe any strain-dependent difference in the levels of Dcr-2 (**Figure 7A**), in agreement with the results reported in **Figure 2**. This allowed us to pool each treatment condition (mock infected, virus infected and non-injected) to analyze the absolute levels of Dcr-2 across time and independently of the fly strain. **Figure 7B** shows that, at time 0, Dcr-2 levels in mock-infected flies rapidly decrease compared to the non-injected control. In contrast, injection of virus increased the level of Dcr-2 protein. At 3 days post-infection, the tendency reversed with increased levels of Dcr-2 in mock-infected flies and decreased levels in virus-infected flies. The same analysis was performed for flies infected by feeding. **Figure 7C** shows the same trend with a prompt increase of Dcr-2 protein levels upon virus ingestion and a later decrease. We did not observe a significant decrease of Dcr-2 levels during mock infection at time point 0, but we did find a significant increase for the time point 3. This indicates that starvation, as well as injection, induces a change in the regulation of Dcr-2.

Altogether, these results put in evidence a change of Dcr-2 protein levels that is not only dependent on virus infection but also on infection procedure-related stress. While virus infection rapidly induces an increment of Dcr-2 protein concentration, the infection procedure-related stress immediately decreases Dcr-2 protein levels. The increase of Dcr-2 protein during viral infection is strong enough to mask the decrease produced by stress.

#### DISCUSSION

Biological systems are composed of two features: pathways and mechanisms (22). Pathways describe the flow of entities or information along space and time; mechanisms describe indirectly the reason these pathways exist. Studying pathways or mechanisms implies studying two different things. In the case of RNAi, if we consider the hierarchical scale that includes the entire RNAi phenomenon, studying the pathway means to study the components of such pathway, the causal way they interact and the flow of these interactions until reaching the "interference" event, which it is relevant to a variety of outcomes. Studying the mechanism means to study how such a pathway works, how it is controlled and regulated, how it is induced or repressed, and for what specific outcome (22). Dcr-2 and Ago-2 are the best characterized enzymes involved in the insect antiviral siRNA pathway. Several lines of evidence demonstrated the pivotal role of this molecular process as the major antiviral response at cellular and systemic level, against both natural and unnatural D. melanogaster viruses (23–28). Nevertheless, we still have a scarce comprehension of the biological implication of the siRNA pathway in gene regulation (11) and in trans-kingdom communication (29–31). A comprehensive understanding of the regulation of the siRNA-mediated response upon different biotic and abiotic stimuli, from virus infection and cellular stress to environmental cues, could shed light on the global role that this pathway plays in the organism. As said by Cornish-Bowden et al. (32), "Only through the understanding of the whole can we understand the functions of the parts".

In this work, we focused on the response of the core components of the siRNA pathway at the transcriptional and translational levels upon viral infection in D. melanogaster using three fly strains, two modes of infection and natural and unnatural viral pathogens.

We observed a strain-dependent susceptibility to viral infections; yw flies were more susceptible to viral infections than w <sup>1118</sup> and Oregon-R flies in all experimental conditions. In agreement with our previous work (21), the mortality after oral infection with DCV was lower with respect to mortality following viral injection. Previous studies reported differences in the immune response of the most common laboratory strains of D. melanogaster used in immunity research (33, 34). For example, Okado et al. (33) show that these strains differ in susceptibility to infection, as well as differences in bacterial load and antimicrobial peptides expression profile upon infection with Lysteria. In addition, upon bacterial infection with E. coli, M. luteus, and E. faecalis, yw displays higher mortality in comparison with other strains (34). Since yw fat storage levels decreased during bacterial infection without increases in bacterial load, the authors concluded that the yw strain was less tolerant to bacterial infection rather than less resistant (34). As we observed a significant increment of viral load in yw flies that correlated with a lower percentage of survival, we cannot advance the hypothesis of a change in tolerance. Nevertheless, several studies highlight a link between metabolism changes and viral resistance in Drosophila (18, 35, 36); therefore we do not exclude the involvement of metabolism as an explanation for the different survival trends we noticed among strains.

Previous studies demonstrated that viral infections trigger Dcr-2 and Ago-2 mRNA expression in Apis mellifera and Bombus terrestris (37), and several authors suggested that this induction might be an essential feature to deal against infections in insects (13–15). For these reasons, we hypothesized that variations in Dcr-2 and Ago-2 expression levels upon viral infection could explain the difference in mortality between strains. Interestingly, we did not observe changes in transcript expression levels for Dcr-2 and Ago-2 in any strain and for any virus and mode of infection. Our results agree with previous D. melanogaster studies [reviewed in (38)] that showed that RNAi genes do not alter their expression level after infection with DCV. Interestingly, a recent study in which D. melanogaster is used as a model to investigate Zika virus-insect interactions, showed an increased level of Dcr-2 and Ago-2 transcripts after infection (18). As previously showed for A. mellifera (19), the regulation of Dcr-2 and Ago-2 expression may be virus-dependent. A broader study including other model viruses from several Baltimore classes should shed light on whether the absence of changes in transcript expression levels, and mainly of induction of Dcr-2 and Ago-2 proteins, is a general response against infections with natural viruses in D. melanogaster.

It is well established that there is not always a direct correlation between mRNA and protein levels (39, 40), and that translational and post-translational regulation and degradation of proteins play a pivotal role in determining protein levels (41). In the current work we showed that, although their transcripts remain unchanged, Dcr-2 and Ago-2 protein levels change after injection-related stress and viral infection. Injection with a sterile solution causes a rapid Dcr-2 level decrease, detectable within minutes and possibly related to a stall in protein translation and/or an increase in protein degradation. Although the actual mechanism is still unknown, similar processes of rapid decrease

flies. No significant differences were observed between the strains inside each time point. Mean ± SD are represented. Analysis performed using the Mann–Whitney test, n = 6 except yw Mock T3 n = 5 and yw Virus T3 n = 3. (B) Dcr-2 levels along time (days post injection); Mock T0 = Mock-control at time point 0; Mock T3 = Mock-control at time point 3; Virus T0 = Virus infection at time point 0; Virus T3 = Virus infection at time point 3. Analysis performed using the Mann–Whitney test; all groups are compared with non-treated. Non-treated n = 48, Injected n = 18 except yw Mock T3 n = 17 and yw Virus T3 n = 15, mean ± SD, (\*P ≤ 0.025, \*\*\*P < 0.0003). (C) Dcr-2 levels along time upon feeding; Mock T0 = Mock-control at time point 0; Mock T3 = Mock-control at time point 3; Virus T0 = Virus infection at time point 0; Virus T3 = Virus infection at time point 3. A.U. indicates arbitrary units. Analysis performed using the Mann-Whitney test; all groups are compared with non-treated. Non-treated n = 48, fed n = 9, mean ± SD (\*P ≤ 0.025, \*\*\*P = 0.0003, \*\*\*\*P < 0.0001).

in protein levels are reported in the literature (42, 43) and can be associated with gene regulation (44). For example, the stimulation with 10 ng/ml of TNF-α induces a fast degradation of IκBα (the NF-κB inhibitor) with a half-life of 16 min (44). On the contrary, we observe that injection with virus leads to an increase of Dcr-2 levels. This suggests that the flies can "sense" the presence of a virus early before the beginning of viral replication, possibly via specific pattern recognition receptors, an assumption previously hypothesized (45, 46) and recently reinforced by results showing that injection of heat-inactivated Zika virus induces certain antiviral immune mechanisms (18). The increase of protein concentration without change in gene expression may be achieved by blocking protein degradation and/or increasing protein translation in a mechanism known as "translation on demand" (40). This molecular process seems to play an important role in protein regulation in yeast and other organisms, including mammals (40, 47, 48). For Dcr-2, the aftermath of this increase was a progressive reduction in protein levels over time that correlates with an increase in viral replication. Two non-exclusive explanations may account for this: (1) the DCV and FHV viral suppressors of RNAi (8) trigger the degradation of Dcr-2 and Ago-2 in a similar fashion to that observed for Cricket Paralysis Virus (CrPV) (49); (2) the later activation of other unknown immunological processes to fight viral infection inhibits Dcr-2 and Ago-2 from being produced.

Several lines of evidence demonstrate that immunological responses can be virus-specific. The fact that other pathways can be more effective than RNAi in counteracting viral infection has been proposed for the case of B. terrestris infected with Israeli Acute Paralysis virus and Slow Bee Paralysis virus (16). We have previously demonstrated that RNAi is not necessary for the clearance of viruses after oral infection (21). Other works showed that the antiviral RNAi is ineffective against virus in the midgut of Aedes aegypti (50) or in certain basal metazoans (51). Altogether these results highlight the fact that other biological processes may be more effective than, or work together with, the siRNA pathway to reach an effective antiviral response.

#### CONCLUSIONS

In our work we showed that the key components of the antiviral siRNA pathway in D. melanogaster, Dcr-2 and Ago-2, are not induced at the mRNA level upon viral infection, but that their regulation occurs at the protein level through an unknown mechanism reminiscent of translation on demand. This response is independent of the mode of infection, the virus, and the fly strain, and is not related to differences in susceptibility to viral infection among strains. We also reported the bivalent aspect of this regulation, which acts as a general gene regulation mechanism (during infection procedure-related stress) and as an antiviral response (during viral infection), possibly activated via early recognition of viral motifs by unknown pattern recognition receptors.

#### DATA AVAILABILITY STATEMENT

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation, to any qualified researcher.

#### REFERENCES


### AUTHOR CONTRIBUTIONS

Conceptualization: AT, VM, JM, and M-CS. Methodology and investigation: AT, VM, and JM. Resources: M-CS. Formal analysis and writing—original draft preparation: AT. Writing—review and editing: VM, JM, and M-CS.

#### FUNDING

This research was funded by the European Research Council (FP7/2013-2019 ERC CoG 615220) and the French Government's Investissement d'Avenir program, Laboratoire d'Excellence Integrative Biology of Emerging Infectious Diseases (grant ANR-10-LABX-62-IBEID) to M-CS.

#### ACKNOWLEDGMENTS

We thank all members of the Saleh lab for discussion, Mariana Ferrari, Thomas Pradeau, Marco Vignuzzi, and Louis Lambrechts for critical reading and editing of the manuscript, Maxime Rotival for advice with statistical analysis, and Kathryn Rozen-Gagnon for advice in the use of anti-Ago 2 antibody.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2020.00362/full#supplementary-material


prevails after varroa destructor-mediated, or in vitro, transmission. PLoS Pathog. (2014) 10:e1004230. doi: 10.1371/journal.ppat.1004230


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

Copyright © 2020 Torri, Mongelli, Mondotte and Saleh. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Drosophila as a Model System to Investigate the Effects of Mitochondrial Variation on Innate Immunity

Tiina S. Salminen1,2 \* and Pedro F. Vale<sup>1</sup>

<sup>1</sup> School of Biological Sciences, Institute of Evolutionary Biology, The University of Edinburgh, Edinburgh, United Kingdom, <sup>2</sup> Faculty of Medicine and Health Technology, Tampere University, Tampere, Finland

Understanding why the response to infection varies between individuals remains one

#### Edited by:

Nicolas Bertho, INRA Biologie, Épidémiologie et Analyse de Risque en Santé Animale (BIOEPAR), France

#### Reviewed by:

Kristi L. Montooth, University of Nebraska-Lincoln, United States Javier Santander, Memorial University of Newfoundland, Canada

#### \*Correspondence:

Tiina S. Salminen tiina.susanna.salminen@gmail.com; tiina.s.salminen@tuni.fi

#### Specialty section:

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

Received: 23 October 2019 Accepted: 06 March 2020 Published: 25 March 2020

#### Citation:

Salminen TS and Vale PF (2020) Drosophila as a Model System to Investigate the Effects of Mitochondrial Variation on Innate Immunity. Front. Immunol. 11:521. doi: 10.3389/fimmu.2020.00521 of the major challenges in immunology and infection biology. A substantial proportion of this heterogeneity can be explained by individual genetic differences which result in variable immune responses, and there are many examples of polymorphisms in nuclearencoded genes that alter immunocompetence. However, how immunity is affected by genetic polymorphism in an additional genome, inherited maternally inside mitochondria (mtDNA), has been relatively understudied. Mitochondria are increasingly recognized as important mediators of innate immune responses, not only because they are the main source of energy required for costly immune responses, but also because by-products of mitochondrial metabolism, such as reactive oxygen species (ROS), may have direct microbicidal action. Yet, it is currently unclear how naturally occurring variation in mtDNA contributes to heterogeneity in infection outcomes. In this review article, we describe potential sources of variation in mitochondrial function that may arise due to mutations in vital nuclear and mitochondrial components of energy production or due to a disruption in mito-nuclear crosstalk. We then highlight how these changes in mitochondrial function can impact immune responses, focusing on their effects on ATP- and ROS-generating pathways, as well as immune signaling. Finally, we outline how being a powerful and genetically tractable model of infection, immunity and mitochondrial genetics makes the fruit fly Drosophila melanogaster ideally suited to dissect mitochondrial effects on innate immune responses to infection.

Keywords: Drosophila melanogaster, cybrid, infection, innate immunity, mitochondria, mtDNA, oxidative phosphorylation, reactive oxygen species

### INTRODUCTION

Understanding why individuals vary in their response to infection is one of the major challenges in immunology (1, 2). This variation may arise from differences in host age (3), sex (4), nutrition or environmental stressors (5), and genetic variation present in immune related genes (2). Experimental immunology – mainly in model systems such as mice, zebra fish and fruit flies – has been successful in identifying the major immune pathways (Drosophila innate immunity summarized in **Box 1**) (6–8). Quantitative genetic and genomic approaches have identified

BOX 1 | Drosophila innate immunity in a nutshell. Drosophila has been extensively utilized as a model system for innate immunity and it has led to many breakthrough in immunity field (6, 114, 115). Drosophila does not possess acquired/adaptive immunity and it relies on humoral and cell-mediated innate immunity for its defense against pathogens, such as bacteria, viruses, fungi, and parasites. Immune mechanisms against these invaders include activation of appropriate signal transduction pathways depending on the invading microbe, involving production of antimicrobial peptides (AMPs), phagocytosis of microbes, wound closure, and a melanization cascade involved in the encapsulation of foreign elements. Similar first-line innate immune defense mechanisms can be found from plants to humans.

Humoral Innate Immunity: In Drosophila, the humoral innate immune response to bacterial pathogens is characterized by the production and release of a cocktail of AMPs into the hemolymph. This response is driven by two evolutionarily conserved and largely independent pathways, Immune deficiency (IMD) and Toll pathways (116). The Toll pathway is induced by bacteria containing LYS-type peptidoglycan in their cell walls (mainly Gram-positive bacteria), while the IMD pathway is induced by DAP-type peptidoglycan (mainly Gram-negative) bacteria. These pathways culminate in the translocation of NF-κB dimers to the nucleus leading to infection-specific upregulation of AMPs targeted to clear the infection (117–119). The response to viral pathogens replicating within the host cells involves both the Janus kinase/signal transducers and activators of transcription (JAK-STAT) pathway (118, 120), RNA interference (RNAi) and antiviral effector molecules (121, 122). Viral infections involve cell-mediated responses like apoptosis and autophagy and humoral responses such as the expression of anti-viral genes, some of which overlap with genes induced upon bacterial and fungal infections, indicating the involvement of the NF-κB signaling upon viral infections. The response to fungal invaders includes both humoral and cellular arms of immunity and involves the expression of AMPs mainly via the Toll pathway.

Cell-Mediated Innate Immunity: In Drosophila, the cell-mediated innate immune system consists of hemocytes (blood cells) and is induced by epithelial damage and detection of foreign particles in the hemocoel. Hemocytes function in sealing of epithelial wounds, encapsulating and terminating parasites and engulfing apoptotic corpses [reviewed in (123)]. In Drosophila there are three major lineages of hemocytes: plasmatocytes (phagocytic), crystal cells (melanization) and lamellocytes (encapsulation). Plasmatocytes comprise the majority of the circulating hemocyte population and are responsible for the engulfment of small particles, participate in the encapsulation of foreign material and are able to trigger the systemic humoral immune response to secrete AMPs. Crystal cells usually make up less than 5% of the larval circulating hemocytes. Crystal cells contain prophenol oxidase which active form phenol oxidase is involved in the melanization cascade when the crystal cells rupture in response to immune activation (123). In uninfected larvae, the lamellocytes can be present in small numbers in the late third instar stage, otherwise healthy larvae do not contain them. Lamellocytes are produced upon invasion of parasitoid wasps and they form a multilayer capsule around the invading parasitic egg, with the help of plasmatocytes and crystal cells (124). Eventually the capsule is melanized and elevated levels of ROS terminate the intruder (125).

polymorphisms in genes underlying these mechanisms, and these explain some of the variation in infection outcomes (9–12). While most of this work has focused on genetic variation in the nuclear genome, metazoan organisms have an additional genome, inherited maternally inside mitochondria (mtDNA). More than functioning as the powerhouses of the cell, a growing body of work in the last decade has shown that mitochondria play an important role in inflammation and immunity and contribute to the host response to infection (13–18).

Here, we propose that the fruit fly (Drosophila melanogaster) offers an ideal model system to investigate the role of mitochondrial variation and mito-nuclear crosstalk in innate immunity. We start by discussing the sources of variation in mitochondrial function, using examples of mutations of nDNA and mtDNA encoded genes that have been shown to affect organismal phenotypes through changes in mitochondrial metabolism and signaling. This is followed by emphasizing the emerging role of mitochondria in immune responses through mitochondrial metabolites and by-products of mitochondrial metabolism, such as ROS. Finally, we describe methodology to investigate the role of mito-nuclear crosstalk and mtDNA variation in immunity in Drosophila. We emphasize how the use of cytoplasmic hybrid (cybrid) models allows to distinguish the effect of mtDNA variation from that arising from the nuclear genome. We conclude by highlighting the benefits of the cybrid model to further our understanding of mito-nuclear effects on heterogenous immune responses.

#### SOURCES OF MITOCHONDRIAL VARIATION

Mitochondrial function depends on ∼1200 – 1500 proteins, the majority of which are encoded by the nuclear genome and transported to the mitochondria (19). Cellular energy production relies on mitochondria to produce ATP via oxidative phosphorylation (OXPHOS). OXPHOS requires the coordinated function of multiple protein subunits encoded by both the nuclear and mitochondrial genomes (nDNA and mtDNA, respectively – **Box 2**), and therefore both anterograde (from nucleus to mitochondria) and retrograde (from mitochondria to nucleus) signaling is required for optimal mitochondrial function. Mitochondrial variation arising from either nDNA or mtDNA can affect the transcription and translation of the mitochondrial proteins, signaling between the two genomes and through changes in the direct physical interactions among the OXPHOS components originating from the two genomes, ultimately affecting the function of mitochondria. Mitochondrial variation shows multiple mode of inheritance. When this variation originates from mtDNA it is maternally inherited and has a potential to become heteroplasmic even within mitochondria, and when originating from nuclear genome it can be X-linked, autosomal dominant, autosomal recessive or de novo. Here, we discuss potential sources of mitochondrial variation with examples of known nuclear and mitochondrial mutations that could also lead to variation in immune responses.

#### Variation Arising From the Nuclear Genome

The vast majority of the proteins that are required for mitochondrial functions are encoded by the nuclear genome, translated in the cytosol and transported to mitochondria via mitochondrial targeting sequence which is removed upon entry into mitochondria. These proteins include the replication,

**Abbreviations:** AMP, antimicrobial peptide; ATP, adenosine triphosphate; DAMP, danger-associated molecular pattern; ETC, electron transport chain; IMD, immune deficiency; mt-aaRS, mitochondrial aminoacyl-tRNA synthetase; mtDNA, mitochondrial DNA; mtROS, mitochondrial reactive oxygen species; nDNA, nuclear DNA; NF-κB, nuclear factor kappa-light-chain-enhancer of activated B cells; OXPHOS, oxidative phosphorylation; ROS, reactive oxygen species.

BOX 2 | Mito-nuclear crosstalk is required for mitochondrial functions. Mitochondria are cellular organelles of eukaryotic cells that are thought to have originated by endosymbiotic phagocytosis of an oxygen-converting α-proteobacterium by archaebacterium (19, 126). The primary function of mitochondria is to produce ATP through oxidative phosphorylation (OXPHOS) complexes I-V, and mitochondrial matrix is also the site of tricarboxylic acid cycle. Mitochondria contains multiple copies of a circular mtDNA (mtDNA copy number) distinct from that of the nuclear genome (Figure 1). Majority of the mtDNA genes required for aerobic energy production through OXPHOS have been shifted to the nuclear chromosomes and the remaining mitochondrial genome in most metazoans encodes for 37 genes, all crucial in OXPHOS. From these, 13 are polypeptide subunits of four of the five OXPHOS complexes, with the majority of the polypeptides encoded by nDNA (Figure 1B). Mitochondria contains its own translational system and the mtDNA encodes two rRNA and 22 tRNA genes as the mt-aaRS genes are encoded by the nuclear genome. Beside the 84 nDNA genes functioning in OXPHOS, around 1200-1500 nDNA encoded polypeptides are imported to and assembled within mitochondria, required for the various mitochondrial functions (Figure 1A). Nucleus and the mitochondria maintain a bidirectional regulation where the nuclear genome can signal to the mitochondria (anterograde signaling) for example to increase mitochondrial respiration. Mitochondria can signal (retrograde signaling) for example to induce cell death by releasing cytochrome c, or by controlling mitochondrial fusion and fission by AMP-activated protein kinase (57).

transcription and translation machineries for mtDNA and the 84 polypeptide subunits needed for OXPHOS (summarized in **Box 2**). Mutations in nDNA directly affecting OXPHOS complex genes have been reviewed in (20). Maintenance genes of mitochondrial functions include regulatory genes of mitochondrial and cytosolic nucleotide pools to maintain balanced supply of mitochondrial dNTPs, involved with mtDNA nucleoid packaging, carrier proteins required for metabolite and cofactor transport across cellular and mitochondrial membranes, genes for mitochondrial lipid and membrane homeostasis, and mitochondrial fission/fusion and cristae organization [reviewed in (21)].

Mutations in mtDNA maintenance genes (replication and repair pathways and mtDNA nucleoid packaging) have been shown to cause mtDNA deletions, point mutations and even depletion (22). The most important mtDNA maintenance gene is the DNA polymerase gamma (POLG) which is responsible for the replication of the mtDNA. Almost 200 POLG mutations have been reported and these are the most common causes of mitochondrial disease. Mutations in POLG have been shown to cause large scale deletions and various other mutations to the mtDNA due to replication and/or repair machinery malfunctions and these have been connected to many mitochondrial diseases such as Alper's syndrome, parkinsonism and multiple other neurodegenerative disorders (23). POLG was mutated in Drosophila to create a proofreading-deficient form resulting to drastically increased somatic mtDNA mutation frequency and mitochondrial dysfunction, which manifested as a shortened lifespan, a progressive locomotor deficit and a loss of dopaminergic neurons (24).

Regulation of OXPHOS gene transcription is tightly coordinated and must be able to establish efficient oxidative metabolism fulfilling the cell's changing energy requirements. Components of the transcription machinery are encoded by the nuclear genome and have been reviewed in (25). Mossman et al., showed that in D. melanogaster cybrid lines the transcription of nuclear encoded mitochondrial genes were affected by mtDNA variation, indicating a retrograde signaling effect in transcription regulation (26). Hence, mutations in the transcription machinery can have a wide impact on the function of mitochondria, and variation in the mtDNA genes can affect the overall transcription efficiency of OXPHOS components, possibly also affecting their translation.

All proteins involved in mtDNA translation (27) are encoded by nuclear genes, involving ribosomal proteins, mt-aaRSs, tRNA modifying enzymes, and translation factors. Mutations in these genes have been shown to cause mitochondrial diseases due to dysfunction in the protein-synthesis machinery (27). mtaaRSs are transported to mitochondria to catalyze an amino acid attachment to its complementary tRNA in aminoacylation reaction for translation of the thirteen mitochondrial proteins. All ribosomal RNAs and the transfer RNAs required for the translation of the mitochondrial proteins are encoded by the mtDNA (**Figure 1**). Hence, nuclear encoded components of the translation machinery need to be able to recognize mitochondrial counterparts for the production of mitochondrial proteins. mt-aaRS genes are central to cellular energy production and mutations in these can lead to variable disease phenotypes depending on the affected tissues and the energy demands of the cells in those tissue types (28). Mutations in both mt-tRNAs and mt-aaRSs can lead to disease and the clinical presentation has been shown to be highly specific to the affected mtaaRS [reviewed in (29)]. However, diversity of pathologies is higher for mt-tRNA mutations than mt-aaRSs, possibly due to random distribution of heteroplasmic populations of mtDNA copies during mitotic segregation (28). In Drosophila simulans a variant of tyrosyl-tRNA synthetase interacts epistatically with a mitochondrially encoded tRNATyr variant, leading to decrease in the activities of OXPHOS complexes I, III, and IV (30). At the organismal level this manifests as developmental delay, compromised bristle formation and decreased fecundity (30).

In D. melanogaster, a tko mutant (tko25<sup>t</sup> ) carries a missense mutation in nuclear encoded mitoribosomal protein S12 causing a decrease in smaller ribosomal subunits. This causes decreased activity levels of all four OXPHOS complexes that contain mitochondrial encoded proteins (**Figure 1**), ultimately manifesting as developmental delay, temporary paralysis followed from vigorous shaking (bang sensitivity) and sensitivity to antibiotics and high sugar diet as well as impaired courtship behavior and hearing (31, 32). Similar defects have been found in fibroblasts of patients with antenatal encephalopathy caused by mutations in the MRPS22 gene coding mitochondrial ribosomal protein, which result in a reduction of 12s rRNA (33).

### Variation Arising From the Mitochondrial Genome

The effect of the mitochondrial genome variation on innate immunity is intriguing as mtDNA does not follow the traditional Mendelian inheritance because it is inherited uniparentally

through the maternal lineage. The mitochondrial genome contains 37 genes, from which 13 encode polypeptides, two rRNA genes and 22 tRNA genes (**Box 2** and **Figure 1A**) and pathogenic mutations have been reported in all 37 mtDNA genes (34). mtDNA is subjected to much higher mutation rates than the nuclear genome and the mutation rate varies between species (35, 36). Besides mutations of the mtDNA replication and repair machinery, higher mutation rate of mtDNA is affected by the production of ROS within mitochondria as by-product of the OXPHOS. This can lead to a cycle in which ROS causes DNA damage, which in turn leads to dysregulation of respiration and accumulation of mutations.

Each cell contains numerous mitochondria and each mitochondrion may contain from a few to dozens of copies of mtDNA (the mtDNA copy number). Hence a given cell of a specific tissue type may contain thousands of copies of mtDNA and only one copy of nDNA. Unlike nDNA, mtDNA is not packed by histones but packed into protein-DNA complexes called nucleoids (37). Drosophila mtDNA contains one large A + T rich non-coding region (38), which in humans is called the D-loop. In both Drosophila and humans, the non-coding region is the starting point of mtDNA replication (39). Hence, variation in this region might cause changes in the mtDNA replication and have an effect on the mtDNA copy number content (40) and ultimately in energy production. mtDNA copy number is usually higher in tissues that have high energy demand such as brain, skeletal, and cardiac muscle tissues (41) and mtDNA mutations in these tissues can possibly lead to more pronounced phenotypes.

Besides heritable mtDNA mutations occurring in the germ line of the maternal lineage, the mtDNA of both females and males are subjected to spontaneous somatic mutations.

mtDNA genetics is complicated due to its multi-copy nature. mtDNA mutations within a single cell (and even within a single mitochondria) can be either heteroplasmic due to a mix of mutated and wild type mtDNA, or homoplasmic where all mtDNAs contain either the mutated or the wild type form. In a heteroplasmic mutation the proportion of mutated mtDNA needs to exceed a certain threshold for the mutation to manifest. This can be due to the wild type mtDNA not being able to compensate the defect at that point. The threshold is also likely to be dependent on the mutation type and environmental effectors. Nuclear genotypes may also have variation in their ability to dampen or amplify the effects of specific deleterious mitochondrial mutations, which often demonstrate incomplete penetrance (42).

mtDNA mutations have been linked to various human diseases (43). In Leber hereditary optic neuropathy (LHON) the patient suffers from a loss of vision and the mtDNA mutations causing it are mostly considered homoplasmic (44). However, even though all the offspring of a homoplasmic mother inherits the LHON, only 50% of males and 10% of females develop the disease, showing that predicting the way the mtDNA mutation manifests, is difficult due to mito-nuclear crosstalk. Also, environmental factors can cause changes in the mtDNA mutation manifestation, as in the case of mtDNA encoded homoplasmic ribosomal RNA (RNR1) mutation that causes deafness early on in childhood. Specific antibiotics are associated with the manifestations of the clinical symptoms of the RNR1 mutation (45). Due to the complexity of the crosstalk of the two genomes and environmental factors, possible physiological compensations and the amplifying or dampening effects originating from nuclear genome variation and compatibility with the mtDNA, it is difficult to predict how the mtDNA mutations will eventually manifest themselves.

#### Disruption of Mito-Nuclear Crosstalk

The optimal functioning of mitochondria relies on the correct transcription and translation of genes involved in respiration, and as mentioned above, these genes are found on both the nuclear and mitochondrial genomes. Several signaling pathways between the nucleus and mitochondria have been uncovered recently (46). Mutations in either nDNA or mtDNA have the potential to disrupt the crosstalk between mitochondrial and nuclear proteins, and can therefore disrupt efficient gene transcription and translation (30), with consequences for metabolic rates (47), aging (48) and sperm competitiveness (49), which ultimately have detrimental effects on organismal fitness (50). Evidence for these detrimental effects is especially clear in hybrids between closely related species or between divergent populations within species, where long-term coevolution between the nuclear and mitochondrial genomes has been broken up, resulting in novel combinations of nuclear and mitochondrial genomes (51, 52). For example, hybrids of the marine copepod Tigriopus show reduced activity of OXPHOS Complex IV (cytochrome c oxidase) because of the breakup of coevolved nuclear and mitochondrialencoded subunits of the Complex IV (51). This mismatch between nuclear and mitochondrial genes is thought to be a strong selection pressure for the fixation of compensatory mutations in nuclear-encoded OXPHOS subunits (53).

Mito-nuclear interactions can also have strong effects on the outcome of infection. In Drosophila, a mito-nuclear incompatibility resulted in energetically compromised flies that were more susceptible to infection by a bacterial pathogen (54). Salminen et al., identified an OXPHOS Complex III mutation D21N in D. melanogaster mitochondrial CYTB gene, that was shown to cause larval stage melanotic nodules in a healthy nuclear background (6%) and a significant increase (56%) of melanotic nodules when the mitotype was introgressed into a tko25<sup>t</sup> nuclear background (55). Formation of melanotic nodules is considered a sign of activated cell mediated innate immunity, and it usually involves the proliferation and aggregation of hemocytes, Drosophila blood cells (56). Furthermore, CYTB mutation bearing mitotype in a tko25<sup>t</sup> nuclear background caused 100% pupal lethality, which is a first report of synthetic lethality between nuclear-mitochondrial interaction within a metazoan species (55).

#### MITOCHONDRIAL VARIATION CAN AFFECT INFECTION OUTCOME

The role of mitochondria in the response to infection is central, impacting multiple functions. First, intermediates of the mitochondrial tricarboxylic acid (TCA) cycle have a signaling function in innate immune responses. Second, mitochondria generate energy by producing ATP during OXPHOS (**Box 2** and **Figure 1B**), and given the elevated energetic requirements of immunity, we may expect variation in mitochondrial functions to result in changes in ATP production, thereby generating heterogeneity in the response to infection. Third, mitochondrial metabolism may further promote protection against pathogens by producing ROS, with direct antimicrobial action. Finally, in mammalian models of immunity, damaged mtDNA has been shown to act as DAMP triggering inflammatory responses akin to those seen during infection. There is therefore increasing evidence that mitochondrial functions contribute to the host response to infection.

### Mitochondrial TCA Cycle Metabolites With Immune Signaling Functions

Mitochondrial tricarboxylic acid cycle (TCA, also called Krebs cycle and citric acid cycle) consists of a series of reactions where substrates originating from carbohydrates, fats and proteins have been fed into it and the metabolites from the cycle are transported into cytosol as building blocks for macromolecules or energy is released through the oxidation of acetyl-CoA. However, metabolites in the TCA cycle have also been shown to be involved in regulation of chromatin modifications, DNA methylation and post-translational modifications of proteins [reviewed in (57)]. Intermediates and derivatives of the TCA cycle have been shown to have non-metabolic signaling functions, in addition to their more conventional role as metabolites associated with bioenergetics (58). Non-metabolic functions of

the TCA cycle intermediates succinate, itaconate, fumarate, 2 hydroxyglutarate and acetyl-CoA play a role in inflammation, and immune cell activation (58). For example, succinate is a proinflammatory metabolite as its production is enhanced during inflammation (59) and it acts as a signal from mitochondria to the cytosol to induce the expression of pro-inflammatory genes and increases the levels of antioxidant superoxide as a proinflammatory redox signal (60). Succinate has also been shown to accumulate in lipopolysaccharide treated macrophages (59). Itaconate on the other hand is endogenous protective and anti-inflammatory molecule that negatively regulates the inflammatory response and cytokine production (61–63) and also has direct antibacterial effects (64). TCA cycle intermediates have also been connected to epigenetic signaling (58). For example, fumarate has a role as an epigenetic inflammatory signal. Arts et al., showed that the accumulation of fumarate in immune activated monocytes was needed for trained immunity by enhancing cytokine production upon re-activation with lipopolysaccharide (65). Further, Acetyl-CoA has been shown to drive histone acetylation which can have profound impact on immune cell function (66). To summarize, mitochondrial variation may impact infection outcomes via their effect on TCA cycle products that have immune signaling functions.

#### Changes in ATP Production

Mitochondria generate energy by producing ATP during OXPHOS (**Box 2**), and given the elevated energetic requirements of immunity, we may expect variation in mitochondrial function to result in changes in ATP production, thereby generating heterogeneity in the response to infection (54). Mutations in any of the nuclear or mitochondrial encoded OXPHOS complexes or in nuclear genes affecting replication, transcription or translation of mtDNA can affect the total electron transfer chain outcome, potentially causing a decrease in the total production of ATP. Severely decreased ATP synthesis is an obvious problem for cells with constant high energy demands such as cardiomyocytes and neurons (67) and decreased ATP synthesis can also increase the AMP/ATP ratio that can lead to activation of AMP-activated protein kinase and multiple signaling pathways (46). We might therefore expect mutations in nuclear or mitochondrial encoded components that cause a reduction in ATP to result in a decrease of immune cell function.

#### Role of ROS in Immune Responses

Reactive oxygen species are a group of reactive molecules and free radicals derived from molecular oxygen which are now known to have a role in cellular homeostasis (68). Elevated levels of ROS can cause oxidative stress, cellular-, and DNA damage in eukaryotic cells. Mutations in nuclear or mitochondrial genes encoding the protein subunits of OXPHOS complexes I and III can cause a decrease or an increase in ROS production, depending on the mutation. One of the most evident roles of mitochondrial functions in innate immunity is the production of ROS by leakage from mitochondrial ETC. The majority of ROS are produced during mitochondrial ETC (mtROS), and some by oxidoreductase enzymes such as NADPH oxidase, a multicomponent membrane bound enzyme complex. Common ROS include superoxide (O<sup>−</sup> <sup>2</sup>), hydrogen peroxide (H202), hydroxyl radical (OH), hydroxide ion (OH−) and nitric oxide (NO). Prolonged oxidative stress is harmful and so detoxification of ROS via scavenging enzymes and antioxidants is vital. Therefore, mutations in the nuclear encoded antioxidants that are targeted to detoxify ROS, can also have an impact on immune response. Antioxidant superoxide dismutase (SOD) is transported to mitochondria where it converts superoxide to hydrogen peroxide. Hydrogen peroxide outside mitochondria is converted to water and oxygen with the help of catalase, peroxiredoxins and glutathione peroxidases (**Figure 1**).

mtROS is produced in all cell types that contain mitochondria and it has been connected to regulation of signaling pathways (69), apoptosis (70), inflammation (71), cellular adaptation to hypoxia (72), cellular differentiation (73), and autophagy (74). In addition to the regulative role of mtROS, a growing body of evidence has highlighted the role of ROS as a target of regulation of immune signaling pathways (14, 75). mtROS serves several roles within both the humoral and cell-mediated arms of innate immunity, including direct elimination of pathogens through its microbicidal effects. However, it is still unclear what is the exact mechanism of the bactericidal effect of mtROS upon bacterial infection as the effect seem to be the type of ROS and pathogen specific [reviewed in (68)] and aligns with the increased heterogeneity of the innate immune response.

Alongside its role in promoting bacterial clearance, mtROS also functions in signaling for hemocyte proliferation and differentiation in Drosophila and has been identified as an essential signaling molecule in Drosophila's cellular immune response to parasitoid infection (76). ROS plays a likely role as a key signaling molecule within the Drosophila lymph gland as it has been suggested to prime the quiescent hemocyte progenitors within the lymph gland for differentiation (77). It has been shown that reduction of ROS significantly retards progenitor differentiation whereas upregulation of ROS via OXPHOS Complex I disruption produces a phenotype with a significantly higher hemocyte population (77).

Transmitochondrial cell lines can be created by combining enucleated cells that contain the mtDNA of interest with cells that lack their mtDNA (78). Data obtained from transmitochondrial cell lines suggest that mtDNA variants on a controlled nuclear background can alter ROS levels of cells (79). Organismal cybrids have also been shown to differ in OXPHOS parameters (34) and effect of mtDNA variation on altered ROS production has been studied in Drosophila in vivo, showing that specific mtDNA variants can elevate ROS production (80).

### Cytosolic mtDNA as a Danger Signal

During infection, elevated levels of ROS can cause mitochondrial and cellular damage resulting mtDNA leakage. Due to the evolutionary origin of mitochondria it harbors resemblance to bacterial DNA making it appear non-self. mtDNA is surrounded by a double-membrane structure and membrane damage could lead to leakage of mtDNA outside mitochondria and elicit selfderived immune activation. When mtDNA is located outside mitochondria in the cytoplasm of the cell or in extracellular space,

it can trigger immune responses by directly engaging the host's innate immune pattern-recognition receptors (PRRs) [reviewed in (81)]. PRRs are conserved receptors that recognize viral, bacterial and fungal particles as well as molecules released from injured cells. The release of mtDNA outside of mitochondria can occur from dying cells, during injury, cellular stress or infection, and mtDNA outside of mitochondria engages PRRs and functions as damage-associated molecular pattern (DAMP) leading to enhancement of pro-inflammatory responses (71). In mammalian models of immunity, inflammasomes are innate immune related signaling complexes that monitor the cytosolic compartment of the cell and are involved in the secretion of cytokines upon infection and recognition of DAMPs (81). Altered mitochondrial dynamics, production of mROS and release of mtDNA outside mitochondria have all been linked to inflammasome activation (82). In mammalian system extracellular circulatory mtDNA has been shown to act as an endogenous Toll-like receptor TLR9 agonist and been connected to many TLR9- dependent inflammatory diseases (81). Cytosolic mtDNA (83) as well as cytosolic double-stranded RNA created during bidirectional transcription of mtDNA (84) has been shown to trigger antiviral responses in human.

#### USING DROSOPHILA TO STUDY MITOCHONDRIAL VARIATION AND MITO-NUCLEAR INTERACTIONS IN IMMUNITY

Mitochondrial function and the content of the mitochondrial genome are highly conserved among metazoans e.g., between humans and the fruit fly D. melanogaster and the latter has been widely used to model human mitochondrial diseases [reviewed in (85)]. Besides studying the naturally occurring variation of nuclear genes affecting mitochondrial function, i.e., mtDNA replication, transcription and translation as well as polypeptides needed for TCA and OXPHOS, it is possible to exploit the highly sophisticated genetic toolbox that exists for D. melanogaster. With the binary gene expression systems, such as GAL4-UAS, one can modify the expression of a desired gene within a specific tissue within a specific time, providing a route to investigate the effect of specific genes in chosen tissues on a given phenotype (86). For example, overexpression and gene knock-down methods could allow the modification of the gene expression of nuclear encoded genes that are transported to mitochondria, hence altering the mitochondrial function, assuming that the mitochondrial import stage does not dampen the effect of genetic modification.

Another approach to study mitochondrial variation is to focus on naturally occurring variation in mtDNA. Investigating the effect of mtDNA mutations is complicated by cross-talk between the mitochondria and the nucleus of a cell, with nuclear genes generally responsible for controlling mitochondrial activity. With the cytoplasmic hybrid, aka. cybrid model, specific mtDNAs can be introgressed onto controlled nuclear backgrounds, making it possible to focus on the effects arising from the mitochondrial genome (**Figure 2**). It is presently not possible to genetically target and modify the gene expression of specific mtDNA genes, and so the cybrid model relies on using natural mtDNA variants found through different genetic screens. However, there are methods for creating random mutations to mtDNA genome such as using POLG mutants (87, 88) or to more specific regions with targeted restriction enzymes (89).

By generating transmitochondrial cybrid cells, it is possible to study the effects of mtDNA variation at the cellular level (78). This in vitro approach has been used previously to investigate the cellular effects of mtDNA polymorphism associated with an aggressive form of breast cancer (90). Similarly, Bellizzi et al., hypothesized that the transcription of stress-responder nuclear genes can be modulated according to the mtDNA variability. They showed that osteosarcoma cells depleted of their own mitochondria and repopulated with different ones, modulated the expression of cytokines and cytokine receptors due to the variability of the mtDNA (91).

The effects of mtDNA variation at the organismal level may vary significantly from what can be inferred from cellular level. The transmitochondrial in vitro method is limited in investigating the impact of mtDNA variation affecting an entire organism, for example upon environmental stress, limited diet or when fighting infection. However, D. melanogaster offers a feasible way of creating and utilizing cybrid strains for in vivo experiments of the effects of mtDNA variation at the organismal level (26, 30, 50), as well as quantifying the effects of mtDNA, nDNA and their interaction. Given the ease with which flies can be sampled from natural populations, options of screening the mtDNA variation by sequencing and crossing and breeding the lines in constant laboratory conditions, Drosophila cybrid lines offer a unique opportunity to disentangle variation in immunity arising from mtDNA polymorphism from the more commonly investigated variation in nuclear-encoded genes associated with canonical immune pathways.

There are three ways of creating cybrids and two of them, repeated backcrossing and using balancer chromosomes, are common methods in Drosophila, explained in detail below and in **Figure 2**. In the third method, Mitochondrial Replacement Therapy (MRT), mtDNA is introduced directly to a novel nuclear environment. In MRT the nucleus of a mtDNA mutation bearing female is transferred to an enucleated egg of a mitochondrially healthy donor (92). To date, MRT has been performed in two other mammalian species beside humans, in macaques (93) and mice (94, 95).

### Generating Drosophila Cybrids by Repeated Backcrossing

Because mtDNA is inherited uniparentally through the mother, the chosen donor mtDNA can be introgressed onto wanted nuclear background by multiple generations of backcrossing (**Figure 2A**). In the first cross the virgin females containing the donor mtDNA are crossed with males containing the wanted nuclear background, resulting in progeny with 100% maternal mtDNA, 50% maternal nDNA and 50% paternal nDNA. This is followed by a series of crosses where the virgin females of each

the correct progeny of the previous cross. For clarity, chromosomes 2 and 3 have not been marked to the cross where the X(a) is replaced with the X(b).

generation are crossed with males of the nuclear donor strain. With each cross, the proportion of paternal nDNA increases by 50% from the previous generation, meaning that in theory, after 10 generations of backcrossing, less than 0.1% of the maternal nDNA should be present. An important caveat to introgression by repeated backcrossing is that although in theory only 10 generations should be required to obtain a line over 99.9% of the paternal nuclear genome, in practice this value will be lower due to strong linkage disequilibrium between loci that have a close genetic distance (called linkage drag) (96). From a mapping perspective this linkage can be advantageous, but the backcrossing approach has the potential to select for compatible nuclear partners, masking true incompatibilities (also lethal combinations). The number of generations of backcrossing required to break linkage between two loci will therefore depend on the genetic distance such that the number of backcrosses N = Log(1-r)(1-P) where r is the recombination frequency, and P is the probability of separation (97). For example, for two loci separated by a 20 cM interval, at least 17 generations of backcrossing are required to have a 95% certainty of a crossover event (97). However, even after multiple generations of backcrossing, some loci are likely to never recombine if they result in strongly deleterious or even lethal phenotypes, and such loci will remain a source of residual heterozygosity regardless of the number of generations of backcrossing.

This method has been used in D. melanogaster experiments to study the effect of mito-nuclear interactions on mtDNA copy number, respiration, development time and weight (40) as well as in experiments where the effect of mtDNA variants on mitochondrial diseases has been assayed in combination with the balancer chromosome method (55, 98). The method has been also employed in other insect species where the use of balancer chromosomes is not possible, where the effect of mito-nuclear crosstalk has been studied on traits such as metabolism and aging in D. simulans (99), metabolic rate in Drosophila subobscura (100), and personality (101), bioenergetics, aging, life history traits (102), and male mating costs (47) in seed beetles.

### Utilizing Drosophila Balancer Chromosomes

Drosophila has four chromosomes that include the X/Y sex chromosome pair and autosomal chromosomes 2, 3, and 4.

The fourth chromosome is very small, contains only around 80 genes (103) and does not recombine. The genetic tool box developed for D. melanogaster allows the replacement of the entire chromosomes with the wanted content of specific strains or mutation. This is done with the help of balancer chromosomes. Balancer chromosomes (104) are multiply inverted and scrambled chromosomes that are not able to undergo crossover with their normal chromosome homologs. They also contain genetic markers that enable the recognition of their segregation. As the small chromosome 4 does not go through crossing over there are no balancers designed for this chromosome. Nuclear genomes can be constructed by replacing the chromosomes in multiple crosses in a properly planned crossing scheme, and eventually controlling the nuclear genome (**Figure 2B**). For creating Drosophila lines that contain the wanted mtDNA variant on a controlled nuclear background, the first cross includes the mtDNA donor females after which the wild-type chromosomes will be replaced one by one with the wanted isogenic chromosomes [crossing scheme explained e.g., in (30, 105, 106)]. This method has been utilized to study the effect of mtDNA variation and mito-nuclear interaction on aging (105), sex differences in aging, respiration and fertility (107, 108), starvation resistance, lipid proportion and physical activity (106) as well as ROS production and mtDNA copy number (109).

#### CONCLUSION AND FUTURE PERSPECTIVES

Drosophila has contributed significantly for our understanding of mitochondrial variation in both mitochondrial diseases and

#### REFERENCES


in immunity. Here we have highlighted strengths of this experimental powerhouse and described approaches that will link these two important fields to address the question how mitochondrial variation and specifically mtDNA variation affect innate immune functions. This is significant because a number of mitochondrial mutations have been associated with increased susceptibility to infection in humans (110, 111) and recent genetic screens have revealed vast variation in Drosophila mtDNA (112). The Drosophila model will also be relevant in the context of mitochondrial replacement therapy as a tool to test for potential incompatibilities that may result from specific mito-nuclear combinations (113). Given the homology between vertebrate and invertebrate innate immunity (114), the Drosophila model has the translational potential to generate novel candidate genes originating from mitochondrial sources of disease susceptibility and resistance, and for development of new therapeutic targets.

#### AUTHOR CONTRIBUTIONS

TS and PV wrote the manuscript.

#### FUNDING

TS was supported by an Academy of Finland Fellowship (grants 322732 and 32879). PV was supported by a Leverhulme Trust Research Project Grant RPG-2018-369, a Branco Weiss Fellowship (https://brancoweissfellowship.org/), and a Chancellor's Fellowship (School of Biological Sciences, University of Edinburgh).



mitochondrial transcription, mass and respiratory activity. Mitochondrion. (2015) 20:13–21. doi: 10.1016/j.mito.2014.10.005




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

Copyright © 2020 Salminen and Vale. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

## Molecular and Functional Analysis of Pore-Forming Toxin Monalysin From Entomopathogenic Bacterium *Pseudomonas entomophila*

#### *Edited by:*

*Laura Vesala, Tampere University, Finland*

#### *Reviewed by:*

*Rebecca M. McQuade, University of Arizona, United States Kwang-Zin Lee, Fraunhofer Society (FHG), Germany*

#### *\*Correspondence:*

*Koki Kamiya kamiya@gunma-u.ac.jp Noriyuki Kodera nkodera@staff.kanazawa-u.ac.jp Takayuki Kuraishi tkuraishi@staff.kanazawa-u.ac.jp*

*†These authors share first authorship*

#### *Specialty section:*

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

*Received: 29 November 2019 Accepted: 06 March 2020 Published: 27 March 2020*

#### *Citation:*

*Nonaka S, Salim E, Kamiya K, Hori A, Nainu F, Asri RM, Masyita A, Nishiuchi T, Takeuchi S, Kodera N and Kuraishi T (2020) Molecular and Functional Analysis of Pore-Forming Toxin Monalysin From Entomopathogenic Bacterium Pseudomonas entomophila. Front. Immunol. 11:520. doi: 10.3389/fimmu.2020.00520* Saori Nonaka1†, Emil Salim1,2†, Koki Kamiya3,4 \*, Aki Hori <sup>1</sup> , Firzan Nainu1,5 , Rangga Meidianto Asri 1,5, Ayu Masyita1,5, Takumi Nishiuchi <sup>6</sup> , Shoji Takeuchi 3,7 , Noriyuki Kodera<sup>8</sup> \* and Takayuki Kuraishi <sup>1</sup> \*

*<sup>1</sup> Faculty of Pharmacy, Institute of Medical, Pharmaceutical and Health Sciences, Kanazawa University, Kanazawa, Japan, <sup>2</sup> Faculty of Pharmacy, Universitas Sumatera Utara, Medan, Indonesia, <sup>3</sup> Kanagawa Institute of Industrial Science and Technology, Kawasaki, Japan, <sup>4</sup> Graduate School of Science and Technology, Gunma University, Maebashi, Japan, <sup>5</sup> Faculty of Pharmacy, Universitas Hasanuddin, Makassar, Indonesia, <sup>6</sup> Institute for Gene Research, Kanazawa University, Kanazawa, Japan, <sup>7</sup> Department of Mechano-Informatics, Graduate School of Information Science and Technology, The University of Tokyo, Tokyo, Japan, <sup>8</sup> WPI Nano Life Science Institute, Kanazawa University, Kanazawa, Japan*

*Pseudomonas entomophila* is a highly pathogenic bacterium that infects insects. It is also used as a suitable model pathogen to analyze *Drosophila's* innate immunity. *P. entomophila's* virulence is largely derived from Monalysin, a β-barrel pore-forming toxin that damages *Drosophila* tissues, inducing necrotic cell death. Here we report the first and efficient purification of endogenous Monalysin and its characterization. Monalysin is successfully purified as a pro-form, and trypsin treatment results in a cleaved mature form of purified Monalysin which kills *Drosophila* cell lines and adult flies. Electrophysiological measurement of Monalysin in a lipid membrane with an on-chip device confirms that Monalysin forms a pore, in a cleavage-dependent manner. This analysis also provides a pore-size estimate of Monalysin using current amplitude for a single pore and suggests lipid preferences for the insertion. Atomic Force Microscope (AFM) analysis displays its structure in a solution and shows that active-Monalysin is stable and composed of an 8-mer complex; this observation is consistent with mass spectrometry data. AFM analysis also shows the 8-mer structure of active-Monalysin in a lipid bilayer, and real-time imaging demonstrates the moment at which Monalysin is inserted into the lipid membrane. These results collectively suggest that endogenous Monalysin is indeed a pore-forming toxin composed of a rigid structure before pore formation in the lipid membrane. The endogenous Monalysin characterized in this study could be a desirable tool for analyzing host defense mechanisms against entomopathogenic bacteria producing damage-inducing toxins.

Keywords: innate immunity, *Drosophila*, pore-forming toxin, atomic force microscope, Monalysin

### INTRODUCTION

The innate immune system is the front line of defense against microbial infection in metazoan animals (1). Innate immune cells can sense infectious threats either by pathogen-associated molecular patterns (PAMPs) or damage-associated molecular patterns (DAMPs) (1–3): pathogen-specific molecules such as peptidoglycans utilized by microbes as an essential substance of their life (1), or host-derived molecules that are normally kept inside in their cells but released because of the tissue damage by infection (2), respectively. The recognition and signaling mechanisms involving PAMPs are relatively well-studied. In Drosophila, humoral innate immunity relies on distinct signaling pathways, the Toll pathway and immune deficiency (IMD) pathway (4, 5). The Toll pathway is responsible for infectious threats from fungi or Gram-positive bacteria, and it senses fungal β-glucans or bacterial Lysine (Lys)-type peptidoglycans with a pattern recognition receptor PGRP-SA/GNBP1 complex or GNBP3 in the hemolymph (6–9). PAMPs recognition by those receptors stimulates serine protease cascades in the hemolymph, which produces a cleaved form of the cytokinelike protein Spätzle (Spz), a ligand of a Toll receptor (10). An activated Toll receptor transmits a signal to NF-κB Dif and/or Dorsal through a dMyd88-Tube-Pelle complex, producing antimicrobial factors such as the antifungal peptide Drosomycin (11, 12). The IMD pathway is another NF-κB pathway that recognizes Diaminopimelic acid (DAP)-type peptidoglycans released from Gram-negative bacteria (13), eventually triggering the translocation of Relish to the nucleus and inducing the expression of genes that encode antimicrobial proteins, including Diptericin (14). In contrast to a PAMPs-initiated innate immunity, DAMPs-mediated innate immune mechanisms in terms of an infectious situation have not been well-characterized yet. In flies, protease cascade upstream of a Toll receptor is partly involved in DAMPs recognition. For example, fungal proteases could potentiate the serine protease cascade through Persephone (15–17). Additionally, entomopathogenic nematodes damage epithelial cells and/or cuticles and degrade basement membrane (BM). Clotting and components from disrupted BM seem to have a protective function against nematode infection (18, 19). However, the whole picture of damage-induced innate immunity is far from understood.

Pseudomonas entomophila is an entomopathogenic, Gramnegative bacterium that was originally isolated from a wild fly sampled in Guadeloupe in the Caribbean (20). P. entomophila displays pathogenicity by oral infection, and the bacteria are widely used as a tool to examine gut innate immune responses (20). Recently, P. entomophila has also been used in a systemic infection model (21). In the gut, P. entomophila infection imposes severe damage via a reactive oxygen species, produced by host cells and a pore-forming toxin (PFT) from the bacteria, generally inhibiting translation in the intestine and thus blocking epithelium renewal (22). The virulence of P. entomophila is under the control of a GacS/GacA two component system (20). One of the main effector molecules in this system is a PFT, Monalysin. Monalysin is secreted as a pro-toxin that is cleaved by proteases, such as AprA in P. entomophila, to become fully active (23). Drosophila adults combat these effectors using a cross-linked drosocrystallin (dcy) protein, which works as physical barrier blocking the permeation of macromolecules (>500 kDa) in the peritrophic matrix (24, 25). Cleaved Monalysin shows cytotoxic activity, probably by forming pores in the plasma membrane of host cells, leading to disrupted membrane permeability and cell death (23). The secondary structure prediction of the membranespanning domain indicates that Monalysin is a PFT of the β type (23). Leone et al. reported that X-ray crystallography and cryoelectron microscopy, with recombinant Monalysin produced in E. coli, revealed its 3D structure and its putative mechanism during pore formation in the lipid membrane (26). The recombinant Monalysin is an 18-mer complex composed of two disk-shaped nonamers held together by the N-terminal swapping of the pro-peptides. The membrane-spanning region of pro-Monalysin is fully buried in the center of the ring or torus, and, perhaps during activation upon cleavage, the two disk-shaped nonamers dissociate to leave the transmembrane segments that attach to the target membrane, undergo conformational changes, and form the pore.

In order to study the interaction between the host and entomopathogenic bacteria producing damage-inducing toxins, well-characterized purified Monalysin may be a useful tool. Besides, pore-forming proteins such as Monalysin may potentially be developed as biological control agents against insects (e.g., Cry toxin) (27–29), as well as biological "nanopores" that are used as a detector for single-molecule (e.g., α-hemolysin) (30). In that sense, endogenous Monalysin purified from P. entomophila could provide more precise insight of its protein function, rather than using the recombinant protein generated by E. coli that may have distinct intracellular environment from P. entomophila, which potentially gives rise to a different subunit composition of the protein and thereby could influence the structural and functional features of the molecule. Additionally, a detailed analysis of the structure of the native pore-forming protein and its dynamics in solution and in lipid membrane would serve basic information for various applications. In this study, we succeeded in purifying native endogenous Monalysin from P. entomophila with killing activity in Drosophila cell line and adult flies. We also characterized its structure and function using electrophysiological measurements and a high-speed atomic force microscope (HS-AFM).

#### MATERIALS AND METHODS

#### Bacteria Stocks, Other Materials, and Cell Culture

P. entomophila wild-type strain L48 and a Monalysin mutant mnl were kindly provided by Dr. B. Lemaitre. 1,2-Dioleoyl-sn-glycero-3-phosphocholine (DOPC), 1,2-dioleoyl-sn-glycero-3-phospho-L-serine (DOPS), 1,2 dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE), and 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-(cap-

biotinyl) (biotin-cap-DOPE) were purchased from Avanti Polar Lipids. n-decane was purchased from Sigma-Aldrich. S2 cells from Drosophila hemocytes were maintained at 25◦C in Schneider's Drosophila medium (Thermo Fisher SCIENTIFIC) containing 10% (v/v) heat-inactivated FBS, 100 units/mL penicillin, and 100 µg/mL streptomycin.

### Purification of Pro-Monalysin

P. entomophila was grown in LB at 29◦C overnight, and was collected by centrifugation at 8,700 × g at 4◦C for 15 min. The cell pellets were washed with PBS and lysed in PBS containing 2% (w/v) CHAPS. Cells were sonicated at 4◦C overnight, filtrated with a 70µm Cell Strainer (BD Falcon), and centrifuged at 9,000 × g at 4◦C for 20 min to remove insoluble pellets. The collected supernatant was diluted 10 times with PBS, filtrated with a 0.22µm filter (Corning), and dialyzed with PBS for 9 h to exchange the solvent. Total lysate (P. entomophila extracts) was performed by ammonium sulfate precipitation (25–50%). The pellet was dissolved in 20 mM Tris-HCl, pH8.0 and dialyzed against the same buffer for 9 h to remove salts. The dialysate was then subjected to anion exchange chromatography with a HiTrap Q HP column (GE Healthcare), pre-equilibrated in 20 mM Tris-HCl, pH8.0 and then eluted with a linear gradient 0 to 1 M NaCl dissolved in 20 mM Tris-HCl, pH8.0 at a flow rate of 1 mL/min for 30 min. After a cell viability assay, the fractions with cytotoxic activity were harvested and concentrated by ammonium precipitation (50%). The pellet was dissolved in 10 mM sodium phosphate buffer, pH 7.4, containing 140 mM NaCl, and then subjected to gel filtration chromatography with a Superdex 200 Increase 10/300 GL column (GE Healthcare), pre-equilibrated in a 10 mM Sodium phosphate buffer, pH 7.4, containing 140 mM NaCl, and then eluted with the same buffer at a flow rate of 0.75 mL/min. The peak eluted at 13–14 min (molecular weight: around 460 kDa) was collected and analyzed by SDS-PAGE. The gel was stained with Coomassie Brilliant Blue (Kanto Chemical Co., Inc.) to check the purity. The 30 kDa band was excised and analyzed by mass spectrometry. The molecular weight of each peak in gel filtration chromatography was estimated by loading Gel filtration Calibration Kit HMW (GE Healthcare) in the same column. Protein concentrations of fractions were measured by a Lowry method with Bio-Rad DC protein assay kit (Bio-Rad). In the trypsin treatment to get active-Monalysin, trypsin was added to purified pro-Monalysin (175 µg) at 0.2 mg/mL and incubated at 25◦C for 10 min, followed by a Protease Inhibitor Cocktail for General Use (nacalai tesque, Cat# 04080-11) was added. To completely degrade Monalysin, trypsin was added to pro-Monalysin (10 µg) at a concentration of 0.2 mg/mL and incubated at 37◦C for 58 h.

#### Mass Spectrometry

To identify the protein in the cytotoxic fraction, MALDI-TOF MS/MS analysis was performed at the Institute for Gene Research, Advanced Science Research Center, Kanazawa University, using a tandem mass spectrometer (4,800 plus MALDI TOF/TOFTM Analyzer [Sciex]) with 2,5 dihydroxybenzoic acid (DHB) as a matrix as described in Asano and Nishiuchi (31). Briefly, a cytotoxic fraction was loaded on an SDS-acrylamide gel, and a 30 kDa band was excised and in-gel digested with trypsin. The digested peptides were analyzed by MALDI-TOF/TOF. The data was subjected to the Protein Pilot ver.4.0 (Sciex) against the Pseudomonas entomophila (NCBI, Tax ID 312306) protein database (2017-8-23). To determine the molecular weight of the active-Monalysin multimer, MALDI-TOF analysis was performed using the UltrafleXtreme MALDI TOF/TOF Analyzer (Bruker Daltonix) at Fukui Prefectural University with sinapic acid (SA) as a matrix. First, areas on the MALDI plates were coated with the SA solution. Then, the mixture of active-Monalysin with SA was dropped onto the SA-coated spots. Each spot was analyzed to obtain the molecular weight by MALDI-TOF (ultrafleXtreme). The results from several measurements were integrated via analysis software version 4.1.2.

### Cell Viability Assay

S2 cells (1.5–8.0 × 10<sup>5</sup> cell in 100 µL) were inoculated in a 96-well plate. 10 µL of P. entomophila extract or collected fractions after chromatography, purified pro-Monalysin (1.5µg/mL), active-Monalysin (1.5µg/mL), or trypsin (3.8 × 10−<sup>2</sup> µg/mL) were added and incubated at 25◦C for 12–18 h. Cell viability was monitored by luminescence from a CellTiter-Glo Luminescent Cell Viability Assay (Promega) with a Spark 10 M (TECAN). Cell viability is expressed as a relative value, with luminescence in cells incubated with the buffer (negative control) being 100%. To measure total activity, cell viability, after incubation with serial diluted fractions, was examined and total activity was calculated as 1 unit corresponding to activity that yields 70% cell viability. Specific activity was expressed as total activity divided by total protein (mg).

### Caspase-3/7 Activity Assay

S2 cells (1.5 × 10<sup>5</sup> cell in 100 µL) were inoculated onto a 96-well plate. Cycloheximide and active-Monalysin were added at 1.5µg/mL and incubated at 25◦C for 6, 12, 18, and 24 h. Caspase-3/7 activity was monitored by luminescence from a Caspase-Glo 3/7 Assay (Promega) using a Synergy HTX (BioTek).

#### Monalysin Injection and Survival Assay

Oregon R flies (Drosophila melanogaster, females, 3–7 days after eclosion) were injected with a pro-Monalysin, active-Monalysin, or degraded-Monalysin solution (1 mg/mL) into their hemolymph by micro-injection (70 nL per fly), and kept at 25◦C. Surviving flies were counted at 1 h after injection. For dosedependent analysis, flies were injected with active-Monalysin solution (3–30µg/mL), and surviving flies were monitored every 12 h for 60 h.

### Total RNA Extraction and Real-Time PCR

Oregon R flies (Drosophila melanogaster, female, 3–7 days after eclosion) were injected with an active-Monalysin, degraded-Monalysin solution (50µg/mL), or 1,000 times dilution of heat-killed E. coli into their hemolymph and kept at 25◦C for 3, 6, 20 h. To obtain the heat-killed E. coli, overnight culture of E. coli (DH5α) without dilution were heated at 100◦C for 30 min, sonicated for 10 min, and then diluted with water. To quantify the Drosomycin (Drs), total RNA of the collected flies was isolated with Sepasol-RNA I Super G (nacalai tesque) and used for cDNA synthesis with ReverTra Ace reverse transcriptase (TOYOBO) and oligo (dT)12–18 primers. To quantify the Diptericin (Dpt), puckered (puc), and Turandot A (TotA), isolated RNA were subjected to DNase treatment (Promega, M6101), followed by cDNA synthesis with ReverTra Ace reverse transcriptase (TOYOBO) and oligo (dT)12–18 primers. Quantitative realtime PCR (RT-qPCR) was performed using a LightCycler 480 (Roche Diagnostics). rpL32 was used as an internal control. The following primers were used for RT-qPCR: Drs forward, TTGTT CGCCCTCTTCGCTGTCCT; Drs reverse, GCATCCTTCGCAC CAGCACTTCA; Dpt forward, GTTCACCATTGCCGTCGCC TTAC; Dpt reverse, CCCAAGTGCTGTCCATATCCTCC; puc forward, GGCCTACAAGCTGGTGAAAG; puc reverse, AGTTCAGATTGGGCGAGATG; TotA forward, CCAAAATGAATTCTTCAACTGCT; TotA reverse, GAATAGCCCATGCATAGAGGAC; rpL32 forward, AGA TCGTGAAGAAGCGCACCAAG; rpL32 reverse, CACCAG GAACTTCTTGAATCCGG.

#### Immunohistochemistry

For oral ingestion of Monalysin, dcy<sup>1</sup> flies (Bloomington #26106, females, 3–7 days after eclosion) obtained from the Bloomington Drosophila Stock Center were starved for 2 h at 29◦C, then placed in a fly vial with the food solution. The food solution consisted in a mixture of active-Monalysin solution (4 mg/mL) and 5 % sucrose (1:1), which was added to a filter disk that completely covered the surface of the standard fly medium. Flies were kept at 29◦C for 8 h, after which their guts were dissected out. Antibody staining was performed as previously described by Kenmoku et al. (32) with 1:200 rabbit anti-PH3 (Cell Signaling, Cat #9701), 1:50 mouse anti-Dlg (Developmental Studies Hybridoma Bank), and 1:200 Alexa 555-coupled and Alexa 488-coupled secondary antibodies (Thermo Fisher SCIENTIFIC). Nuclei were stained by 0.1µg/mL of 4',6-diamidino-2-phenylindole (DAPI). Samples were visualized with a LSM710 confocal microscope (Carl Zeiss) or observed using a conventional fluorescent microscope and images were reconstructed using Photoshop (Adobe).

#### SLP Assay for Purified Monalysin

To examine the contamination level of peptidoglycan, 10 µL of 0.001–1 mg/mL pro-Monalysin, active-Monalysin and degraded Monalyin were incubated with 40 µL of Silkworm Larvae Plasma (SLP) reagent (Wako) at 25◦C for 30 min in a 96-well plate. The SLP reagent contains all factors involved in the prophenoloxidase cascade system triggered by peptidoglycans, which consequently activates prophenoloxidase. The activated prophenoloxidase then oxidizes 3,4-dihydroxyphenylalanine (DOPA) in the substrate, thus forming a black melanin pigment. The amount of peptidoglycan was monitored as the blackness of the mixture visually. As a positive control, several dilution (1/10, 1/10<sup>2</sup> , 1/10<sup>3</sup> , 1/10<sup>4</sup> ) of heat-killed E. coli solution were subjected to the same test.

#### Ion Current Measurement of Monalysin Using a Bilayer Lipid Membrane (BLM) Chip With 16 Separate Channels

One microliter of pro-Monalysin solution (1.5 mg/mL) was added to 0.1 µL of trypsin (0.25 % [w/v]). The mixture was incubated for 10 min at room temperature (∼23◦C). To form a planar BLM, using the droplet contact method; 3.7 µl of lipid dissolved in n-decane (20 mg/mL dioleoylphosphatidylcholine [DOPC] or dioleoylphosphatidylcholine, dioleoylphosphatidylserine, and dioleoylphatidylethanolamine [DOPC/DOPS/DOPE] [molar ratio of 7:2:1]) was added to each double well on a BLM chip with 16 separate channels (16-ch). Twenty one microliter of buffer solution (20 mM Tris-HCl/ 150 mM NaCl [pH 8.8]) containing 0.015 mg/mL of Monalysin solution was added to each double well. The planar BLM was formed at the macroapertures. The Monalysin's current signals were recorded using a multichannel patch clamp amplifier with a 1-kHz low-pass filter at a sampling frequency of 5 kHz (Tecella JET). The measurement temperature was 23 ± 1 ◦C. Current analysis was performed using the pCLAMP software program (molecular devices).

#### Atomic Force Microscopy

AFM imaging was performed in a solution at room temperature (24–26◦C), using a laboratory-built high-speed AFM setup (33) as described in Uchihashi et al. (34). For AFM substrates, two types were used: the flat muscovite mica substrate and the polydimethylsiloxane (PDMS) substrate with controlled convex shapes (ca. 50 nm) (35). Either a mica disc (1.5 mm in diameter and ∼0.05 mm in thickness) or a PDMS disk (2 mm in diameter and ∼0.02 mm in thickness) was glued on a glass sample stage (2 mm in diameter and height) by epoxy. A freshly cleaved mica surface was prepared by removing the top layers of mica using Scotch tape. The PDMS surface was hydrophilized by a plasma ion bomber (PIB-10, Vacuum Device) set to hard mode for 3 min. The glass stage with either substrate was attached to the top of a Z-scanner by a drop of nail polish, on which a drop (2 µL) of sample solution (either 0.1 mg/mL Monalysin or 0.1 mg/mL liposome) was deposited. The liposome solution was prepared as previously described (34), and the lipid composition was DOPC:DOPS:biotin-cap-DOPE = 7:2:1 (w/w). After incubation for 3–5 min, the substrate surface was rinsed with 20 µL of the observation buffer to remove floating samples. The sample stage was then immersed in a liquid cell containing ∼60 µL of the observation buffer. AFM imaging was carried out in tapping mode, using small cantilevers (BLAC10DS-A2, Olympus), with a resonant frequency of ∼0.5 MHz in water, a quality factor of ∼1.3 in water, and a spring constant of ∼0.08 N/m. The cantilever's free oscillation amplitude A<sup>0</sup> and set-point amplitude A<sup>s</sup> were set at 1–2 nm and ∼0.9 × A0, respectively. In some experiments, high tapping forces were applied to the samples by reducing A<sup>s</sup> , and a protein solution containing either Monalysin or trypsin (5 µL) was injected in the observation buffer during high-speed atomic force microscopy (HS-AFM) imaging. The imaging rate, scan size, and the pixel size for each AFM image are described in figure legends.

#### Analysis of AFM Images

AFM images were pretreated for analysis by a low-pass filter to remove spike noise and a flatten filter to make the overall xy-plane flat, using a laboratory built software as described in Ngo et al. (36). The molecule heights were measured semiautomatically using the following steps. First, the most probable highest point near the highest point of the molecule was selected manually. Second, the actual highest point was determined automatically by searching a 10 × 10-pixel area (typically 10 × 10 nm<sup>2</sup> ) around the selected point. The surface area occupied by the specific molecular species was analyzed by ImageJ using binarized images. The binarized images were obtained by setting a threshold height. The threshold heights were 8 nm for the double-ring complex of pro-Monalysin and 4 nm for the singlering complex of pro- and active-Monalysin, respectively.

#### Statistical Analysis

Statistical analyses were performed by a Student's t-test or logrank test, and P < 0.05 were considered significant.

#### RESULTS

#### Purification of Endogenous Monalysin Protein From *P. entomophila*

In our previous study, we demonstrated that an extract from P. entomophila, prepared by sonication of the bacterial cells with detergent, followed by membrane-filtration, is fatal to adult flies if ingested (24). To examine whether P. entomophila extract acts as source for the purification of endogenous Monalysin, we tested whether P. entomophila extract had Monalysin-derived cytotoxic activity. P. entomophila extract was simply added to the Drosophila embryonic hemocyte-derived S2 cell culture, and we found that, after 12 h of incubation, almost all cells lost their normal morphology and fell apart (**Figure 1A**). A CellTiter-Glo Luminescent Cell Viability Assay, which measures cellular ATP, indicated that the S2 cells were dying (**Figure 1B**).

Next, we performed the cytotoxic assay, using an extract from a Monalysin-deficient strain of P. entomophila to know whether P. entomophila extract-induced cell death depends on Monalysin. We found that an extract from a Monalysin-deficient strain showed less cytotoxicity than that of a wild type (**Figures 1A,B**). This indicates that a P. entomophila extract contains Monalysin toxin, and that endogenous Monalysin could be purified using the extract. Furthermore, we expected that this cytotoxic assay could be used to find fractions containing Monalysin in each purification step. We attempted its purification in this way (**Figure 1C**, **Table 1**). First, a P. entomophila extract or total lysate were precipitated with ammonium sulfate to reduce the extract volume. The precipitate was suspended with a Tris buffer, subjected to a column for anion exchange chromatography, and eluted by a linear gradient of 0 to 1 M NaCl. The active fraction (Fraction III) was subjected to an ammonium sulfate precipitation again and suspended with a phosphate buffer (Fraction IV). Fraction IV was subjected to a column for gel filtration chromatography, and several sub-fractions were collected. A CellTiter-Glo Luminescent Cell Viability Assay showed that only sub-fraction 1 contained cytotoxic activity (**Figure 1E**). The recovered activity of this final fraction (Fraction V in **Table 1** and **Figure 1C**, sub-fraction 1 in **Figures 1D,E**) was 6% that of the starting total lysate of P. entomophila. The specific activity increased 14-fold (**Table 1**). The chromatogram of the final fraction (Fraction V in **Table 1** and **Figure 1C**, sub-fraction 1 in **Figures 1D,E**) showed a sharp single peak around 460 kDa and gave a single band with a molecular mass of 30 kDa on an SDS-PAGE (**Figure 1F**), which is the estimated size of the pro-form of Monalysin monomer. Mass spectrometric analysis of the single band resulted in specific amino acid sequences of Monalysin (data not shown). From these results, we concluded that Monalysin, as a pro-form, was purified as a homogeneity from the wild type P. entomophila. Note that the molecular mass estimated by gel filtration (460 kDa), that of the multimer of Monalysin, was slightly smaller than the estimate found in previous literature (26).

Pro-Monalysin is considered to undergo proteolytic cleavage by AprA, a protease secreted by P. entomophila, in order to be fully activated as a toxin (23). Leone et al. showed that trypsin cleavage of recombinant pro-Monalysin recapitulated the proteolysis by AprA (26). Thus, we performed trypsin cleavage on our purified endogenous Monalysin to see whether it transforms from a pro-form to an active-form. SDS-PAGE analysis showed that 30 kDa of pro-Monalysin monomer was cleaved to 27 kDa of monomer, as previously reported (**Figure 2A**). Hereafter, we refer to trypsin-treated endogenous pro-Monalysin as active-Monalysin since the cleaved form exhibited much higher cytotoxic activity than the pro-form (**Figure 2B**). Note that the trypsin in active-Monalysin did not show cytotoxic activity (**Supplementary Figure 1**). A lethal concentration of 50% (LC50) of pro- and active-Monalysin was estimated from **Figure 2B** as 1.4 and 3.1µg/mL, respectively (**Figure 2C**), which suggested that pro-Monalysin is also toxic to S2 cells. We interpreted this to mean that pro-Monalysin has cytotoxic activity without trypsin treatment because it can undergo proteolysis with some proteases of S2 cells in a cultured medium or on the cell surface, since the pore formation efficiency of pro-Monalysin in artificial membranes is much lower than active-Monalysin (**Figure 3B**). This cell death induced by active-Monalysin appears to be necrotic rather than apoptotic, as the cells did not show caspase-3/7 activation, while cells treated with cycloheximide (reported to induce typical apoptosis in S2 cells) showed significant induction of the caspase-3/7 activity [(37); **Figure 2D**]. These results are consistent with cell death induced by Monalysin produced in E. coli described in Opota et al. (23).

To confirm whether purified Monalysin had toxic activity in vivo, we injected Monalysin into adult hemolymphs. **Figure 2E** shows that active-Monalysin killed adult flies more efficiently than pro-Monalysin and degraded-Monalysin that is digested by trypsin for a long time to be fully decayed (**Supplementary Figure 2**), and the insecticidal effect of active-Monalysin has a dose-dependent effect (**Figure 2F**). Furthermore, the analysis of dcy mutant guts fed with active-Monalysin revealed that the number of mitotic stem cells using phospho-histone H3 (PH3) staining (38)—an indicator of gut repair after damage—increased after ingestion of active-Monalysin (**Supplementary Figure 3A**). In addition, an immunostaining of the septate junction marker Discs large (Dlg) (39) and nuclear staining in flies after oral injection of active-Monalysin showed disrupted organization of the epithelial cells (**Supplementary Figure 3B**), implying that Monalysin could damage the flies' intestines. Collectively, these results suggest that purified Monalysin has toxic activity in vivo.

FIGURE 1 | incubated with *Pe* WT or *Pe* 1*mnl* total lysates (15 µg protein/8 × 10<sup>5</sup> cells in 100 µL) for 12 h. Cell viability was monitored as luminescence by a CellTiter-Glo Luminescent Cell Viability Assay. Cell viability is shown relative to luminescence in cells that were incubated with PBS, taken as 100%. The means ± S.E. obtained with the data from triplicate samples, are presented (\**P* < 0.05, as determined by a Student's *t*-test). (C) The purification step of endogenous pro-Monalysin. A HiTrap Q HP column and a Superdex 200 Increased 10/30 L GL column were used in anion exchange chromatography and gel filtration chromatography, respectively. (D) Chromatogram of gel filtration chromatography. Eluted proteins were detected by measuring OD280. The retention time was the time passed after loading the sample into the column. The molecular weight of each retention time was estimated by loading Gel filtration Calibration Kit HMW (GE Healthcare) in the same column. The estimated molecular mass was around 460 kDa for the first eluted peak. Pre-loading indicates a fraction before loading to column for gel filtration chromatography (that is, it is the same Fraction IV in Table 1). Brackets and numbers were collected fractions and sub-fraction numbers, respectively. (E) S2 cells were incubated with each fraction obtained from gel filtration chromatography for 12 h. Cell viability was monitored as luminescence by a CellTiter-Glo Luminescent Cell Viability Assay. Cell viability is shown relative to luminescence in cells incubated with an elution buffer, taken as 100%. The means ± S.E. obtained with the data from triplicate samples, are presented. (F) SDS-PAGE analysis of fraction 1. The gel was stained with Coomassie Brilliant Blue. The arrowhead indicates a pro-Monalysin monomer (30 kDa). The numbers on the left side indicate molecular weight.

#### TABLE 1 | Purification of pro-Monalysin.


*<sup>a</sup>Total activity was calculated as 1 unit corresponding to activity that yields 70% cell viability.*

*<sup>b</sup>Specific activity indicates total activity divided by total protein.*

*<sup>c</sup>ppt: precipitate.*

We also examined whether Monalysin injections could induce antimicrobial peptides (AMP) and stress gene expressions in adult flies, as tissue damage could induce infection-independent humoral innate immunity. Real-time quantitative PCR (RTqPCR) analysis suggested that the expression of Drosomycin (Drs) and Turandot A (TotA), a read-out of the activation of the JAK-STAT pathway, was significantly induced by the injection of active-Monalysin (**Supplementary Figures 4A,D**). This suggests that Monalysin activates the innate immune and stress pathways, possibly through the Toll and/or JAK-STAT pathways. Monalysin injection, however, did not induce Diptericin (Dpt) and puckered (puc) gene expression, a readout of the activation of the IMD pathway and Jun-Nterminal kinase (JNK) pathway (**Supplementary Figures 4B,C**). P. entomophila is a Gram-negative bacterium which does not contain the Lys-type peptidoglycan recognized by the Toll pathway. In addition, degraded-Monalysin did not induce Drs and TotA expression (**Supplementary Figures 4A,D**) and a peptidoglycan-contamination test using Silkworm Larvae Plasma (SLP). Reagent did not show significant contamination of peptidoglycans, which normally activate innate immunity, in active-Monalysin (**Supplementary Figures 4E,F**). These results might exclude the possibility that contamination of some PAMPs activates humoral immunity and imply that tissue damage induced by Monalysin might induce a humoral innate immune response and stress response in adult Drosophila. Taken together, these results show we succeeded to purify endogenous Monalysin, which has a toxic and damage-inducing activity in Drosophila.

### Electrophysiological Characterization of Monalysin as a Pore-Forming Toxin

We next sought to confirm that endogenous Monalysin indeed functions as a PFT and characterize its mode-of-action by functional analysis. To monitor pore formation on the lipid membrane, we adopted an "on-chip lipid bilayer system," which was composed of a parallel ion current recording device with 16 separate channels (16-ch) of artificial planar BLM wells, where the bilayers in the wells were formed based on the droplet contact method (**Figure 3A**) (40). First, we observed the formation of Monalysin nanopores onto lipid bilayers via the electrophysiological analysis of an artificial cell membrane. We obtained stepwise signals specific to nanoporecontaining proteins in the solution containing the active-Monalysin (**Figure 3B**, left). A total of 723 stepwise signals for active-Monalysin on the DOPC lipid bilayer were observed for 30 min using a 16-ch device (N = 2). On the other hand, in case of the pro-Monalysin, 35 stepwise signals were observed for 30 min using a 16-ch device (N = 2) (**Figure 3B**, right). These results suggest that the trypsin-treated active-Monalysins were more vigorously reconstituted into the lipid bilayer and formed nanopores within it.

Next, we investigated the appearance of the active-Monalysin on lipid bilayers composed of DOPC and DOPC/DOPS/DOPE (mol ratio of 7:2:1). The formation of the Monalysin nanopores in the lipid bilayers was more occurrent on the DOPC lipid bilayer (723 stepwise signals) than the DOPC/DOPS/DOPE lipid bilayer (349 stepwise signals) (**Figure 3C**). We found two amplitude peaks for the active-Monalysin-specific stepwise

FIGURE 2 | Trypsin treatment transforms purified pro-Monalysin into its active-form. (A) A SDS-PAGE analysis of purified pro-Monalysin before and after trypsin treatment. The gel was stained with Coomassie Brilliant Blue. Trypsin was incubated with purified pro-Monalysin at 0.2 mg/mL for 10 min. Closed and open arrowheads indicate a pro-Monalysin monomer (30 kDa) and an active-Monalysin monomer (27 kDa), respectively. The arrow indicates trypsin. (B) Cell viability after incubation with pro-Monalysin and active-Monalysin. S2 cells (1.5 × 10<sup>5</sup> cells in 100 µL) were incubated with the indicated concentration of pro-Monalysin or active-Monalysin for 18 h, cell viability was measured via a CellTiter-Glo Luminescent Cell Viability Assay. Cell viability is shown, relative to luminescence in cells incubated with PBS, taken as 100%. (C) LC<sup>50</sup> of pro-Monalysin and active-Monalysin. LC<sup>50</sup> was estimated from data in the B. The means ± S.E. obtained with the data from triplicate samples are presented (\**P* < 0.05, as determined by a Student's *t*-test). (D) Caspase-3/7 activity in cells after incubation with active-Monalysin. S2 cells (1.5 × 10<sup>5</sup> cell in 100 µL) were incubated with active-Monalysin or Cycloheximide (an apoptosis inducer) at 1.5µg/mL for the indicated time. Caspase-3/7 activity was measured as luminescence using a Caspase-Glo 3/7 Assay. The means ± S.E. obtained with the data from duplicate samples in two independent experiments (E) Survival analysis of adult flies upon injection with pro-Monalysin, active-Monalysin, or degraded-Monalysin (1 mg/mL) into their hemolymph for 1 h (\**P* < 0.05; NS, not significant, as determined by a Student's *t*-test). The minus indicates un-injected flies. The means ± S.E. were obtained with the data from three vials (10 flies/each). The data represents two independent experiments. (F) Survival analysis of adult flies upon injection with active-Monalysin (3–30µg/mL) at indicated time points (\**P* < 0.0001, as determined by a log-rank test).

signals in each case: 1.5 ± 1.9 pA, 17.3 ± 1.9 pA, and 25.5 ± 2.2 pA (mean ± S.D.) for the DOPC lipid bilayer, and 12.3 ± 1.8 pA, 18.1 ± 2.0 pA, and 26.1 ± 2.7 pA (mean ± S.D.) for the DOPC/DOPS/DOPE lipid bilayer (**Figure 3C**, left). The amplitude peaks of the active-Monalysin signals in case of the DOPC and DOPC/DOPS/DOPE bilayers showed no significant differences. We estimated the diameters of the active-Monalysin nanopores from the amplitude of the active-Monalysin signals and buffer conductance, in accordance with the method described in Gutsmann et al. (41). The diameters of the active-Monalysin nanopores were estimated, using the amplitude peaks, to be 0.74 ± 0.30 nm, 0.91 ± 0.30 nm, and 1.10 ± 0.32 nm (mean ± S.D.) in case of the DOPC lipid bilayer, and 0.77 ± 0.30 nm, 0.77 ± 0.29 nm, 0.93 ± 0.31 nm, and 1.12 ± 0.36 nm (mean ± S.D.) in case of the DOPC/DOPS/DOPE lipid bilayer. In summary, Monalysin appears to insert itself preferably within a lipid bilayer with high ratio of PC, and forms pores measuring around 0.7–1 nm, regardless of the lipid composition.

#### Atomic Force Microscope Analysis for the Structure of Monalysin in Solution

Gel filtration chromatography of endogenous Monalysin indicates that Monalysin forms a stable pore-forming complex before activation and membrane interaction, as previously suggested (26). However, based on the gel filtration analysis, it seems that the molecular weight of a Monalysin complex is slightly smaller than that of a previous 18-mer model of Monalysin. Data from an MALDI-TOFMS analysis is in line with this estimate. We detected a possible molecular ion peak of active-Monalysin multimers, whose m/z was 217932.23. This value was very close to the molecular weight of an 8-mer activecomplex expected from the amino acid sequences, 213160.4 Da (**Supplementary Figure 5**). Revealing the structure of native Monalysin in solution and lipid membrane, particularly its dynamic nature, is essential to understand its detailed molecular function and use to evaluate innate immunity mechanisms in flies, as well as to develop biological control agents against insects and biological nanopores. To this end, we employed HS-AFM that enabled dynamic real-time observations of macromolecules at nanometer resolutions, which are not feasible with other methods (42, 43), and had recent achievements of revealing the dynamic structures of pore-forming proteins (44–47).

First, we observed pro-Monalysin in the PBS buffer on a mica surface. The experimental setup is shown in **Figure 4A**. As shown in **Figure 4B**, molecules with a uniform height covered the mica surface. At smaller scan sizes, trefoil-shaped molecules were seen (**Figure 4C**, **Supplementary Movie 1**). Importantly, the molecule corresponding to each leaf of the trefoil dissociated from, and re-bound to, a trefoil-shaped molecule (**Figure 4C**, 9.75, and 10.25 s), indicating that one particle in the trefoilshaped molecule is the minimum unit of pro-Monalysin. Note that pro-Monalysin occasionally forms a trimer of the minimum unit in the solution. Hereafter, we refer to this minimum unit of pro-Monalysin as pro-form. The pro-form height was 14.0 ± 0.9 nm (mean ± S.D) (**Figures 4D,E**). The center-to-center distance between the adjacent pro-forms in the trefoil-shaped molecule was 11.3 ± 1.7 nm (**Figure 4F**). This result suggests that this distance corresponds to the maximal pro-form diameter, which is slightly smaller than the reported value of recombinant Monalysin (∼14 nm) (26). We did not confirm the presence of the reported pore structure on the center of the pro-form. This is because the pro-form moved faster than the AFM scanning speed.

Next, we observed active-Monalysin in the PBS buffer on a mica surface. As shown in **Supplementary Figure 6A**, no trefoilshaped molecules were observed, but molecules with a similar width to that of the pro-form were rapidly moving on the mica surface (**Supplementary Movie 2**). The striking difference from the pro-form was that the height of the molecules was 5.1 ± 0.3 nm (**Supplementary Figures 6B,C**), which is less than half of the pro-form. This result is consistent with the idea that pro-Monalysin, double-stacked disk-like oligomers, dissociates into two disk-shaped oligomers upon proteolytic cleavage (26). This suggests that we could observe the active-form of Monalysin. However, the presence of the central pore could not be confirmed as the molecules moved rapidly.

Use of a low salt buffer (30 mM NaCl, 10 mM Sodium phosphate, pH 7.0) as an observation buffer induced strong immobilization of Monalysin oligomers on the mica surface, allowing us to image the molecular feature at a high spatial resolution. As shown in **Figure 4G**, oligomers with a central pore were clearly visualized by the AFM even at somewhat larger scan sizes. The height was 5.3 ± 0.3 nm (mean ± S.D.) (**Figures 4H,I**), identical to that obtained under the PBS buffer. The central pore had an aperture diameter of ∼3 nm (**Figure 4H**), consistent with the previous report (26). The centerto-center distance between the adjacent active-forms was 11.4 ± 0.9 nm (**Figure 4J**), indicating that the maximal diameter of the active-form is identical to that of the pro-form. This result is consistent with no significant change induced in the outer diameters of Monalysin upon protease activation (26). At smaller scan sizes, the sub-unit stoichiometry was directly resolved (**Figure 4K**, **Supplementary Movie 3**). Unexpectedly, the active-Monalysin was composed of eight sub-units and formed a disk-shaped octamer, in contrast to the crystalline structure of recombinant pro-Monalysin (26), which suggests nonameric (9 mer) composition.

We next visualized the conversion of pro-Monalysin to active-Monalysin after trypsin treatment in the PBS buffer. The video shows that, in a trypsin-concentration dependent manner, almost all molecules with a height of ∼14 nm were converted into molecules with a height of ∼5 nm over time (**Figure 5**, **Supplementary Movie 4**). Interestingly, we noticed that, by applying stronger tapping forces, the pro-form with a height of ∼14 nm can be changed into molecules with a height of ∼5 nm without trypsin treatment. Indeed, this change occurred depending on the strength of the tapping force (**Supplementary Figures 7A–C**, **Supplementary Movie 5**). When Asp/A<sup>0</sup> were set at 0.5, this height change was seen for almost all the pro-form molecules after 60 s. The average tapping force in this imaging condition is estimated to be 53 ± 17 pN, using the nominal values of A<sup>0</sup> = 2.0 ± 0.2 nm, k<sup>c</sup> = 80 ± 20 pN/nm, and Q<sup>c</sup> = 1.3 ± 0.2. In contrast, under the typical imaging conditions using Asp/A<sup>0</sup> of more than 0.8, giving an average tapping force of <37 ± 17 pN, this height alteration was not seen at all, even after 60 s. These results strongly suggest that the pro-Monalysin height change seen in the trypsin treatment is induced by the proteolytic cleavage, not by the mechanical perturbations. Thus, the results seen in **Figure 5** are direct evidence that trypsin treatment effectively digests a portion of pro-Monalysin and produces the activeforms by dissociating the doubly stacked disks. Note that the active-forms can withstand an average tapping force of ∼50

FIGURE 4 | nm<sup>2</sup> with 100 × 100 pixels, and the imaging rate was 330 ms/frame. (C) Successive AFM images of pro-Monalysin (see Supplementary Movie 1). The light blue arrowhead shows that a pro-form detaches from, and binds to, a trefoil-shaped molecule. The scanning area was 100 × 100 nm<sup>2</sup> with 100 × 100 pixels and the imaging rate was 250 ms/frame. (G) A wide-area image of active-Monalysin. Two different height scale images are shown. Bright spots are some adsorbed debris. The scanning area was 80 × 80 nm<sup>2</sup> with 160 × 160 pixels and the imaging rate was 330 ms/frame. (K) Small-area image of active-Monalysin (see Supplementary Movie 3). The scanning area was 40 × 40 nm<sup>2</sup> with 120 × 120 pixels, and the imaging rate was 150 ms/frame. The right image is an averaged image using four successive images. (D,H) A cross-section analysis of pro- and active-Monalysin. The sections are from the blue and red lines drawn on the images in C,G. Each dashed line indicates the center position of a molecule used in the analysis of F,J. (E,I) Height distributions of pro- and active-Monalysin. (F,J) Center-to-center distance distributions of pro- and active-Monalysin. All distributions were fitted by single-Gaussian curve.

pN (**Supplementary Figures 7D,E**, **Supplementary Movie 6**). In addition, the height alternation of the pro-form from 14 to 5 nm was also induced just when the molecules were strongly immobilized on the mica surface under the low salt buffer (**Supplementary Figure 8**, **Supplementary Movie 7**). These results suggest that the intramolecular interactions supporting the disk-shaped octamer structure are strong, while the disk-disk interaction is relatively weak and perhaps only sustained by the interaction of amino acid residues removed during protease activation. However, it remains an open question whether the force-indeed molecules from the pro-form with a height of ∼5 nm are active and can form nanopores within the cell membrane. Importantly, the pro-Monalysin, not activated by trypsin treatment, strongly immobilized on the mica surface formed the disk-shaped octamer with the central pore (**Supplementary Figure 8B**). Consistent with the previous report (26), this result suggests that the central pore is already formed in the doubly stacked disks of the pro-Monalysin in solution.

#### Real-Time Dynamics of Monalysin Insertion Into a Lipid Bilayer

We next visualized the insertion events of the active-Monalysin into a lipid membrane (**Figure 6**). A lipid membrane composed of a mixture of phospholipids of DOPC/DOPS/biotin-cap-DOPE was formed on the surface of PDMS (35). The active-Monalysin was then added into the observation buffer to be monitored. HS-AFM video showed that the active-Monalysin was inserted into the lipid membrane without significant structural change (**Figure 6**, **Supplementary Movie 8**). The active-form height was 6.1 ± 0.7 nm (mean ± S.D.) from the surface of lipid membrane (**Figures 6B,C**), which is slightly higher than that seen in the active-form on mica. At smaller scan sizes, the sub-unit stoichiometry was directly resolved to be 8-mer (**Figure 6D**, **Supplementary Movie 9**). These results collectively suggest that endogenous pro-Monalysin is a 16-mer complex, separated by protease into 8-mer active complexes, and the 8-mer active complex is inserted into the lipid membrane as they are.

Interestingly, Monalysin was preferentially inserted into the edge of the lipid membrane (**Figure 6E**). This implies that Monalysin prefers to be inserted in highly curved parts of the membrane. Consistent with this, we observed many insertions of active-Monalysin into the liposome (**Supplementary Figure 9A**, **Supplementary Movie 10**), while no insertion was seen into the lipid membrane formed on the mica surface (**Supplementary Figure 9E**). The surface roughness of the lipid bilayer formed on the PDMS and mica surfaces were 0.51 ± 0.08 nm and 0.13 ± 0.05 nm, respectively. These results indicate that the active-Monalysin can recognize such difference in the surface roughness of lipid membrane and make a drastic difference in the membrane insertion. Interestingly, the liposome placed on the mica surface was distorted and deformed into a disc shape (∼13 nm in height and ∼80 nm in width) rather than a spherical shape (**Supplementary Figure 9B**), and all Monalysin molecules were located on the outer periphery of the disc (see the blue arrowheads in **Supplementary Figure 9A**). This suggests that the active-form can spontaneously migrate to highly curved sites on the lipid membrane. In addition, we found that the active-Monalysin which have been reconstituted with liposome has two height states of 5.9 ± 0.5 and 7.3 ± 0.4 nm (**Supplementary Figures 9C,D**). Considering that the height of the active-form inserted into the lipid membrane formed on the PDMS is ∼6 nm, the active molecule inserted into the curved membrane may have a height of ∼7 nm, and the molecules with the height of ∼7 nm seen in **Supplementary Figures 9C,D** are presumably in a molecular state before height transition from ∼7 to ∼6 nm.

### DISCUSSION

In this study, we reported, for the first time, on the purification of endogenous pro-Monalysin from entomopathogenic bacteria P. entomophila. Purified pro-Monalysin is activated by trypsin treatment, which is confirmed by electrophysiological analysis with an artificial lipid membrane. The pro-form is stable for more than a month at 4◦C (data not shown), and active-form produced by trypsin treatment had cytotoxic activity in Drosophila cell line and adult flies. In particular, we examined the distinct structure and dynamics of endogenous Monalysin in solution and within the lipid membrane using HS-AFM and revealed the stability of the active-octamer structure. This study suggests that endogenous Monalysin is one of the best model toxins from entomopathogenic bacteria. Additionally, information on pore size estimated by electrophysiological analysis is useful for the potential development of biological nanopores from endogenous Monalysin.

We purified pro-Monalysin based on cytotoxic activity from cell pellets of P. entomophila, not from a culture supernatant, meaning that a large amount of Monalysin is kept inside the cells as an assembled pro-form. The diameters of pro-Monalysin and active-Monalysin, estimated from HS-AFM analysis, are ∼11 nm and their heights are ∼14 and 5–7 nm, respectively. Both sizes are too large to be secreted by P. entomophila's secretion system. It has type I and II secretion systems with secretion pore diameters that are generally <5 nm (48–50). Since P. entomophila

this term were disturbed, and the area measurements cannot be performed.

the active-Monalysin are preferentially inserted into the edge of lipid bilayer. Asterisk marks represent the lipid bilayer areas. The scanning area was 200 × 200 nm<sup>2</sup> with 100 × 100 pixels, and the imaging rate was 330 ms/frame.

tends to undergo autolysis, particularly at temperatures over 30◦C, some P. entomophila may have been lysed in the host's intestine. In this case, pro-Monalysin was released from the dead bacteria and digested by bacterial proteases such as AprA, or perhaps also by host proteases, to become its active-form. We also observed that the trefoil-shaped structure of 16-mer

pro-Monalysin composes the trimer complex in the PBS buffer when the concentration of pro-Monalysin is high enough. By making a trimer complex, the cleavage site for proteolytic activation of pro-Monalysin might be hidden and prevented from being an active-form in P. entomophila. Sub-cellular localization and estimated concentration of pro-Monalysin in P. entomophila should be investigated in a future study.

The most well-characterized PFTs that damage insect tissue are probably Cry toxins from Bacillus thuringiensis, a Grampositive bacterium commonly used as a biological pesticide (27). Cry toxins represent a large family, consisting of more than 350 different members, yet their common features as toxins are essentially the same (27). They form crystalline inclusions after production, are solubilized, undergo partial cleavage by proteases in digestive juice, and are activated after ingestion. The activated Cry toxin is considered a monomer and then forms a pore on the cell membrane of midgut epithelial cells, following specific interactions with a receptor(s), resulting in cell lysis and destruction of midgut tissue (27, 28). However, our study showed that endogenous Monalysin was composed of a pore-forming multimer from the beginning and demonstrates a receptorindependent insertion into the lipid membrane. Importantly, since Cry toxins show target specificity of insect species through selective toxin-receptor interactions, one needs to employ a genetic trick when using them on flies, e.g., overexpression of the Cry1Aa receptor and the application of Cry1Aa thereafter (29). Thus, if one seeks to impose tissue damage in a non-specific manner, Monalysin injection or ingestion would be a simple method. Indeed, our study showed that injection of endogenous Monalysin, through a standard procedure, effectively killed adult flies and induced an innate immune response. This study provides a theoretical basis for the use of endogenous Monalysin toxins as a tool for studying injury-induced innate immunity in the context of microbial infections. Note that endogenous Monalysin can be purified milligram order from 1 L of bacterial culture, suggesting its versatility within a range of experiments.

Our HS-AFM analysis revealed real-time dynamics of Monalysin in action (**Figure 7**). Pro-Monalysin purified from P. entomophila showed a relatively scissile 16-mer complex (though stable enough during the purification step and storage), in contrast to the 18-mer structure of recombinant Monalysin prepared from E. coli. A 16-mer model of our endogenous Monalysin is consistent with gel filtration chromatography and MALDI-TOFMS analysis. This discrepancy might be derived from the different techniques, or the hosts, used to express Monalysin. Alternatively, pore-forming toxins tend to compose different subunit stoichiometry in solution than in crystal. Indeed, α-hemolysin had been suggested to form heptamer by X-ray crystallography, though an AFM analysis indicated that αhemolysin composes hexameric stoichiometry (51). This report and our current study may imply that pore-forming toxins, in general, could form two different energetically stable oligomers in different conditions. By using endogenous Monalysin, it would be interesting to solve crystal structure or to perform cryo-EM in the future study.

Activated-Monalysin obtained by trypsin treatment has an 8-mer constitution and roughly halves in height, indicating that pro-Monalysin is half dissociated into an 8-mer pair well before insertion. It penetrates and forms pores in the lipid bilayer without profound structural change. Notably, Monalysin was preferentially inserted into the edge of the lipid membrane, implying that Monalysin could recognize the target membrane's curvature. Eukaryotic cells possess local membrane subdomains, some of which have high curved areas, such as the tip of filopodia and the area of endocytosis or exocytosis. Those subregions have important biological functions for cell movement, intracellular communication, and signaling (52). Phagocytosing immune cells, such as macrophages, extend a lot of filopodia, particularly upon immune activation. Monalysin or its relatives might preferentially target those cells and/or important biological membrane regions. Additionally, a possible entry site of Monalysin is the tip of microvilli of enterocytes in the fly's intestinal epithelium, as previous studies have reported that Monalysin killed adult flies after oral infection through gut injury (23). Furthermore, electrophysiological analysis revealed that Monalysin inserts itself most commonly within lipid bilayers with high ratio of PC. The chemical nature of lipids determines how the lipids bundle side-by-side in a monolayer and thereby influences the monolayer curvature. For example, lysophospholipids form positively curved monolayers; PC build nearly flat monolayers; and DOPE assemble negatively curved monolayers (53). Notably, upon cell activation, some of the membrane phospholipids are metabolized into eicosanoids and lysophospholipids (54). Active-Monalysin may prefer to insert a highly bent portion of the plasma membrane with a large amount of lysophospholipids produced upon a cell signaling event. From this point of view, not only local membrane subdomains, but also some specific cell types and/or cell activation status could be the insertion target of active-Monalysin. And then, these characteristics might determine host and position specificity where Monalysin shows toxicity. More detailed analysis to find more specific lipid preference and the optimal radius of curvature for Monalysin insertion shall be conducted in vitro and in vivo. Further studies on endogenous Monalysin are required to answer questions concerning the precise mode-of-action during pore formation in the target membrane.

#### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the article/**Supplementary Material**.

#### AUTHOR CONTRIBUTIONS

NK and TK: conceptualization. SN, KK, NK, and TK: formal analysis and writing—original draft. SN, ES, KK, AH, FN, RA, AM, TN, ST, NK, and TK: investigation. SN, ES, KK, AH, ST, NK, and TK: funding acquisition. All authors writing—review and editing.

#### FUNDING

This work was supported by a JSPS KAKENHI Grant and an AMED Grant for TK, an Indonesia Endowment Fund for Education (LPDP) and Ministry of Research, Technology and Higher Education of the Republic of Indonesia (BUDI-LN Scholarship) and the Sasakawa Scientific Research Grant for ES, The Mochida Memorial Foundation for Medical and Pharmaceutical Research for SN, the Kao Foundation for Arts and Sciences for AH, JSPS KAKENHI Grant (15H04360 and 18H05269) and JST CREST (JPMJCR1762) for NK.

#### ACKNOWLEDGMENTS

We thank Dr. Bruno Lemaitre for strains of P. entomophila and antibody against Monalysin. We also thank the Bloomington Drosophila Stock Center and Developmental Studies Hybridoma Bank for providing us with the flies and antibodies, respectively. We are grateful to Drs. Shoichiro Kurata and Yoshinobu Nakanishi for our discussions. We acknowledge the technical support in the electrophysiological characterization of Monalysin provided by Ms Y. Yatomi, and the technical support for the HS-AFM from Drs. Toshio Ando and Takayuki Uchihashi. The authors would like to thank Enago (www.enago.jp) for the English language review.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2020.00520/full#supplementary-material

#### REFERENCES


epithelia from bacterial exotoxins. PLoS Pathog. (2015) 11:e1005244. doi: 10.1371/journal.ppat.1005244


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

Copyright © 2020 Nonaka, Salim, Kamiya, Hori, Nainu, Asri, Masyita, Nishiuchi, Takeuchi, Kodera and Kuraishi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Origins of Metabolic Pathology in *Francisella*-Infected *Drosophila*

Crystal M. Vincent\*, Carolina J. Simoes da Silva, Ashima Wadhawan and Marc S. Dionne\*

MRC Centre for Molecular Bacteriology and Infection and Department of Life Sciences, Imperial College London, London, United Kingdom

The origins and causes of infection pathologies are often not understood. Despite this, the study of infection and immunity relies heavily on the ability to discern between potential sources of pathology. Work in the fruit fly has supported the assumption that mortality resulting from bacterial invasion is largely due to direct host-pathogen interactions, as lower pathogen loads are often associated with reduced pathology, and bacterial load upon death is predictable. However, the mechanisms through which these interactions bring about host death are complex. Here we show that infection with the bacterium Francisella novicida leads to metabolic dysregulation and, using treatment with a bacteriostatic antibiotic, we show that this pathology is the result of direct interaction between host and pathogen. We show that mutants of the immune deficiency immune pathway fail to exhibit similar metabolic dysregulation, supporting the idea that the reallocation of resources for immune-related activities contributes to metabolic dysregulation. Targeted investigation into the cross-talk between immune and metabolic pathways has the potential to illuminate some of this interaction.

#### *Edited by:*

Susanna Valanne, Tampere University, Finland

#### *Reviewed by:*

Edan Foley, University of Alberta, Canada Elodie Ramond, Institut National de la Santé et de la Recherche Médicale (INSERM), France

#### *\*Correspondence:*

Crystal M. Vincent crystal.vincent@imperial.ac.uk Marc S. Dionne m.dionne@imperial.ac.uk

#### *Specialty section:*

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

*Received:* 25 November 2019 *Accepted:* 02 June 2020 *Published:* 08 July 2020

#### *Citation:*

Vincent CM, Simoes da Silva CJ, Wadhawan A and Dionne MS (2020) Origins of Metabolic Pathology in Francisella-Infected Drosophila. Front. Immunol. 11:1419. doi: 10.3389/fimmu.2020.01419 Keywords: *Drosophila*, *Francisella*, metabolism, pathophysiology, immune response

### INTRODUCTION

Infection phenotypes can result from direct interactions between host and pathogen, or indirect interactions which often take the form of trade-offs with, or damage caused by, the host's immune response (1–4). The deleterious effects of immune activity are most often described in the context of an unfettered immune response and the resultant "costs" of said activation [e.g., decreased lifespan (5–7)], or, the by-products of immune effectors [e.g., biochemical interactions; pleiotropic signaling cascades (8–11)].

While the contribution of the host's immune response to infection pathology is acknowledged, a majority of studies focus on pathogen-derived infection phenotypes. Studying direct host-pathogen interactions allows researchers to be on the frontline of infection, dissecting each play and counter-play as host and pathogen battle, often resulting in either elimination of the invader, or death of the host. Work in the fruit fly has supported the assumption that mortality resulting from bacterial invasion is largely due to direct host-pathogen interactions, as lower pathogen loads are often associated with reduced pathology, and bacterial load upon death is predictable (12–14).

Drosophila melanogaster has been used as a model for host-pathogen interactions due to its experimental tractability (10, 15–18). One emergent and exciting field of work is focused on the interaction between infection and host metabolism. Work from our lab and others has identified a number of regulatory factors that play roles in both immune and metabolic activity (10, 11, 19). For example, infection with the intracellular bacterium Mycobacterium marinum leads to a wasting phenotype, with flies losing both glycogen and triglyceride stores over the course of infection (20). This wasting phenotype is mediated, in part, by impaired insulin signaling.

Here we explore the interaction between pathogenesis, immunity and metabolic function. The bacteriostatic antibiotic doxycycline, a close analog of tetracycline, is a standard treatment for infection with Francisella tularensis (21). Flies infected with the Gram-negative intracellular bacterium Francisella novicida die within 6 days of infection (22). In these infections, the greatest pathogen loads are observed immediately prior to death (22). We show that F. novicida—infected flies treated with tetracycline maintain live bacteria at loads less than the initial inoculum and exhibit no difference in survival with their PBS-injected controls. These results suggest that factors dependent on bacterial proliferation contribute to mortality during infection. We find that the metabolic dysregulation observed during this infection is dependent on bacterial load, and we distinguish between immune-derived and pathogen-driven pathology using immune deficient hosts.

### MATERIALS AND METHODS

#### General Experimental Procedures

We used w <sup>1118</sup>; and w <sup>1118</sup>; imd<sup>10191</sup> flies in this study. The w <sup>1118</sup>; imd<sup>10191</sup> line is a mutant of the immune deficiency pathway (imd); it has a 26-nucleotide deletion that frameshifts the protein at amino acid 179, which is the beginning of the death domain (23). Male flies were collected following eclosion and kept in same-sex vials for 5 to 9 days in groups of 20. Flies were maintained on a standard sugar-yeast diet (10% yeast, 8% fructose, 2% polenta, 0.8% agar, supplemented with 0.75% propionic acid and 0.075% nipagin) at 25◦C. Injections were carried out using a pulled-glass capillary needle and a Picospritzer injector system. Flies given tetracycline were transferred to 0.04% tetracycline food (same recipe as above, supplemented with powdered Tetracycline≥98.0% (NT), 87128 Sigma-Aldrich) 6 h after injection. We transferred flies onto tetracycline 6 h post injection for two reasons: first, we wanted to simulate normal usage of antibiotics, and therefore did not administer tetracycline prior to the establishment of infection; second, preliminary studies showed that flies transferred to tetracycline food at 6 and 24 h post infection had similar survival (**Supplementary Figure 1**). We chose to administer tetracycline at 6 h to allow more time to assay pathology in the shorter-lived imd<sup>10191</sup> mutants.

Bacteria were grown from single colonies overnight at 37◦C in a shaking incubator. Wild-type Francisella novicida (U112) and tetracycline resistant Francisella novicida (U112 pKK219-GFP) were grown in Tryptic Soy Broth supplemented with 5% cysteine. U112 pKK219 cultures were additionally supplemented with 0.1 % tetracycline. Each fly was injected with 50 nl of bacterial culture diluted to OD<sup>600</sup> = 0.1 in PBS. As infection (bacteria) and wounding controls, we had flies that were injected with sterile PBS and anesthetized but otherwise unmanipulated, respectively.

#### Survival Assays

Survival experiments were performed at 29◦C with 15– 25 flies/vial. Survival was monitored daily and flies were "tipped" into fresh vials every 3 days; this method for transferring flies permits the provisioning of fresh food without anesthesia.

#### Bacterial Quantification

For each sample, at the timepoint specified, 1 fly was homogenized in a 100 µl of Tris-EDTA, 1% Proteinase K (NEB, P8107S) solution. Homogenates were incubated for 3 h at 55◦C followed by a ten-minute incubation at 95◦C. Following incubation, we performed our qPCR protocol as outlined below to determine the number of bacterial colony forming units (CFU). Methods used to estimate bacterial load upon death (BLUD) are described elsewhere (12); briefly, we collected flies within 30 min of death and processed samples as above. Preliminary experiments showed that quantification via qPCR yields similar results to plating (mean CFU: plating−1850, qPCR−2204, **Supplementary Figure 2**). As such, we are confident that this method of quantification provides a good estimate of bacterial number.

#### Gene Expression—Quantitative Reverse Transcription PCR

For each sample, three flies were homogenized in 100 µl of the single-step RNA isolation reagent TRI Reagent (Sigma), followed by a chloroform extraction and precipitation in isopropanol. The resultant pellet was then washed with 70% ethanol. Pellets were resuspended and subjected to DNase treatment. Revertaid M-MuLV reverse transcriptase and random hexamers (Thermo Scientific) were used to carry out cDNA synthesis. Fivemicroliter of each cDNA sample was put into a "neat" standard tube; this tube was later used to generate standards which would be used to generate a standard curve for each gene. Each cDNA sample was diluted and this diluted sample used for analysis.

We used qPCRBIO SyGreen Mix for qRT-PCR. The cycling conditions were as follows: Hold 95◦C for 10 min, then 45 cycles of 95◦C for 15 s, 59◦C for 30 s, 72◦C for 30 s, followed by a melting curve. To gain a general picture of AMP activity, we assayed a subset of antimicrobial peptides which we have found to be strongly induced during F. novicida infection (**Table 1**). Gene expression was calculated based on the standard curve generated during each run, normalized to the value of our housekeeping gene, RPL1. Samples from PBS and infected treatments were then divided by the mean value of their uninfected controls to generate expression values relative to uninfected flies. All gene expression experiments were repeated at least twice, with four or more biological replicates per experiment.

#### Measurement of Glucose and Glycogen Levels

Each sample contained three flies that were homogenized in 75 µl of TE + 0.1% Triton X-100 (Sigma Aldrich), and stored at −80◦C. Prior to the assay, samples were incubated for 5 min at 65◦C. Following incubation, 10 µl from each sample was loaded

TABLE 1 | Primer sequences used for qRT-PCR.


into 3-wells of a 96-well plate. Each well was designated to serve as a measurement for either: control (10 µl sample + 190 µl H20), glucose [10 µl sample + 190 µl glucose reagent (Sentinel Diagnostics)], or glycogen [10 µl sample + 190 µl glucose reagent + amyloglucosidase (Sigma Aldrich)]. A standard curve was generated by serially diluting a glucose sample of known concentration and adding 190 µl of glucose reagent to 10 µl of each standard. Standards were always run at the same time and in the same plate as samples. Plates were incubated for 1.5–2 h at 37◦C following which the absorbance for each well at 492 nm was determined using a plate reader.

#### Measurement of Triglyceride Levels

Triglycerides were measured using Thin Layer Chromatography (TLC) assays as described elsewhere (24). Briefly, each sample consisted of eight flies; flies were placed in microcentrifuge tubes and stored at −80◦C until the time of analysis. To perform the TLC assay, samples were removed from the −80◦C freezer and spun down (3 min at 13,000 rpm at 4◦C) in 100 µl of a 3:1 (v/v) mix of chloroform and methanol. Flies were then homogenized and subjected to a further "quick spin." Standards were generated using lard dissolved in the same chloroform: methanol solution. We loaded 2 µl of each standard and 20 µl of each sample onto a silica gel glass plate (Millipore). Plates were then placed into a chamber pre-loaded with solvent (a 4:1 (v/v) mix of hexane and ethyl ether) and left to run until the solvent reached a point 1 cm short of the edge of the plate. Plates were then removed from the chamber, allowed to dry, and stained with CAM solution (24). Plates were baked at 80◦C for 15–25 min and imaged using a scanner. Triglyceride was quantified in Image J using the Gel Analysis tool.

#### Statistical Analysis

All data were analyzed in R Studio with R version 3.5.1. Survival data were analyzed using pairwise comparisons Log-Rank tests. BLUD assay correlations were computed by Pearson's correlation. For all other assays, we first tested for normality of data which dictated whether an ANOVA, t-test, Kruskal— Wallis analysis of variance, or Wilcoxon test was used to calculate differences between treatments. When appropriate, we performed post-hoc Tukey or Dunn analyses to identify specific differences between treatments. All assays were repeated at least twice with the number of biological replicates as indicated.

#### RESULTS

### Effect of Tetracycline During *F. novicida* Infection on *w*<sup>1118</sup> Flies

Infected w <sup>1118</sup> flies given tetracycline lived 4.3x longer than flies kept on normal food (median survival of infected flies: normal food−4 d; tetracycline−17.5 d; **Figure 1A**) and similar survival to their PBS controls (median survival of tetracycline-fed flies: PBS−16 d; F. novicida−17.5 d; **Figure 1A**, **Supplementary Table 1**). Flies injected with PBS and fed tetracycline lived significantly longer than their normal food controls, while the opposite was true in uninfected flies (median survival; Uninfected: normal food−27 d, tetracycline−23 d; PBS: normal food−11 d, tetracycline−16 d; **Figure 1A**). As tetracycline is a bacteriostatic antibiotic and thus prevents bacteria from proliferating rather than actively killing bacterial cells (21), we assayed bacterial load over the course of infection in tetracycline-fed and normal food flies. Confirming our preliminary findings, flies kept on normal food exhibited an exponential increase in bacterial numbers over the first 3 days of infection, while flies given tetracycline maintained a low level of bacteria (**Figure 1B**). Bacterial load differed significantly between normal food and tetracycline flies for all time points beyond 24 h. Furthermore, bacterial loads in normal food flies differed significantly between all consecutive time points apart from days 1 and 2. For tetracycline-fed flies, there was a significant increase in bacterial load between day 3 and day 5, but no difference between days 1 and 3, nor days 5 and 11 (**Figure 1B**).

Because pathogen detection is required for the initiation of the antimicrobial peptide (AMP) response (25–27), we sought to determine how the changing bacterial load on normal food, and the static load on tetracycline food, were affecting AMP expression. We found that 24 h post-infection, AMP expression did not differ between flies given tetracycline and those on normal food (**Figure 1C**, **Supplementary Figure 3**). This finding was unsurprising as bacterial loads did not differ significantly between the two groups at this time. In contrast, by 72 h postinfection, all four of the AMPs measured had greater induction in normal food flies compared to tetracycline-fed (**Figure 1D**).

Having confirmed that infection with F. novicida is lethal, and characterized by rapid bacterial proliferation and AMP induction, we wanted to determine if F. novicida infection had any effect on host metabolism. Infected tetracyclinefed flies live longer, have lower bacterial loads, and reduced AMP induction; thus, we anticipated that the metabolism of these flies would not be affected in the same way as flies that were not given tetracycline. However, we found that flies infected with the non-lethal pathogen Microccocus luteus exhibit metabolic dysregulation similar to what we here observe, suggesting that this phenotype is independent of moribundity [**Supplementary Figure 4**, (28)]. During the early stages of infection, there was no effect of tetracycline or infection on metabolism (**Figures 1E,F**). However, by 72 h post-infection, infected flies kept on normal food had significantly higher levels of glucose, as well as depleted glycogen and triglyceride stores. Infection did not affect glucose, glycogen, or triglyceride levels in tetracycline-fed flies (**Figures 1F,G**). Collectively,

FIGURE 1 | food, respectively. Significance codes: \*\*\* < 0.001, \*\* < 0.01, \* < 0.05. Survival (A) F. novicida -infection is represented by dotted lines. Uninfected and PBS controls are represented by solid and dashed lines, respectively. Infected, tetracycline-fed flies live 4.3x longer than flies kept on normal food. Median survival of infected flies: tetracycline−17.5 d; normal food−4 d (Log-Rank = 745.6, df = 5, n = 602, p ≤ 2e-16). Survivals were repeated 2 or 3 times with 20 flies/treatment/repeat. A full-factorial report of statistics can be found in Supplementary Table 1. Bacterial quantification (B) tetracycline-fed flies (Tet) had significantly lower bacterial loads at both 2 and 3 d post infection (Kruskal-Wallis = 112.47, df = 11, p = 2.2e-16; Dunn's post hoc: 48 h–p = 0.001, 72 h–p = 2.1e-11). A significant increase in bacterial load was observed between days 3 and 5 in tetracycline-fed flies (Dunn's post hoc: tetracycline 2d—tetracycline 5 d, p = 0.008), but remained constant from that point on. Normal food (NF) flies show an exponential increase in bacterial load throughout infection. Markers indicate means and bars represent SE. Bacterial quantifications were repeated twice, with n = 8 samples/treatment/repeat. Spider plots showing antimicrobial peptide transcript levels 24 h (C) and 72 h (D) post infection. All tracings are relative to uninfected individuals of the same treatment. Solid and dashed lines represent PBS and F. novicida injection, respectively. The area contained within the innermost quadrilateral represents induction levels falling between one and ten times that of the uninfected controls. The middle and outer quadrilaterals represent 10–100 and 100–1,000-fold induction, respectively. Antimicrobial peptide assays were repeated twice, with four samples/treatment/repeat. These data are also shown, represented differently, in Supplementary Figure 3A. Metabolism (E) glucose, glycogen and triglyceride levels are unchanged at an early point (6 h post-infection) of infection. (F) Glucose and glycogen levels 24 and 72 h post-infection. Infection led to a significant increase in glucose levels during late infection in normal food flies (Kruskal-Wallis = 20.007, df =3, p = 1.7e-04; Dunn's post hoc: normal food PBS—normal food Fnov = 2.6e-04). Glycogen stores were depleted by infection in normal food flies (AOV: df =3, n = 28, F = 4.855, p = 7.6e-03, Tukey's HSD: normal food PBS—normal food Fnov = 0.015). Groups sharing the same letter are not significantly different. (G) Infection led to a significant depletion of triglycerides in infected normal food flies (AOV: df =3, n = 28, F = 11.86, p = 3.4e-05, Tukey's HSD: normal food PBS—normal food Fnov, p = 0.006). Infection did not affect any measure of metabolism in tetracycline-fed flies. Large circular markers indicate means while smaller circles represent individual data points. Horizontal bar within each box represents the median. The bottom and top lines of the box represent the 1st and 3rd quartiles, respectively. Whiskers represent either the maximum and minimum values, or, the maximum and minimum values falling within 1.5x the interquartile range, in which case outliers are indicated. Metabolic assays were repeated 2 or 3 times, with four samples/treatment/repeat.

these data show that tetracycline reduces the pathology of F. novicida infection.

#### Effect of Tetracycline During *F. novicida* Infection on *imd*<sup>10191</sup> Flies

Mutants of the Imd pathway exhibit poor AMP induction in response to infection with Gram-negative bacteria and impaired survival during infection with F. novicida (22, 25, 29, 30), so we did not measure AMPs in these flies. Instead, we tested whether treatment with tetracycline enhances survival in imd<sup>10191</sup> mutants to a level similar to that observed in w <sup>1118</sup> flies. In addition, as there is an assumed "cost of immunity" (4, 31– 33), infection of imd<sup>10191</sup> mutants with F. novicida would allow us to determine if part of the observed metabolic dysregulation during infection in w <sup>1118</sup> is caused by imd-dependent immune activation. Any observed differences in metabolism between w <sup>1118</sup> and imd<sup>10191</sup> flies are unlikely to result from differential resource availability as we found no difference in triglyceride levels between the two (**Supplementary Figure 5**).

Infected imd<sup>10191</sup> flies given tetracycline lived 3.6x longer than flies kept on normal food (median survival of infected flies: tetracycline−11 days; normal food−3 days; **Figure 2A**, **Supplementary Table 2**). In the absence of infection, tetracycline had a slightly positive, though non-significant effect on survival, with uninfected normal food and tetracycline flies having similar median survival times (12 and 13.3 d, respectively).

In contrast to our observation in w <sup>1118</sup>, tetracycline did not affect survival of PBS-injected imd<sup>10191</sup> mutants. Bacterial load differed significantly between tetracycline and normal food imd<sup>10191</sup> flies for all time points measured (**Figure 2B**). Bacterial load in normal food flies differed significantly between all consecutive time points, but remained constant in tetracyclinefed flies, excluding day 7, where values were significantly higher (**Figure 2B**). As both survival and bacterial proliferation in imd<sup>10191</sup> mutants resembled the overall pattern observed in w <sup>1118</sup> flies, we tested whether the metabolic dysregulation was also present during this infection. Infection affected neither glucose nor glycogen, independent of whether flies were fed tetracycline (**Figures 2C,D**). Similar to what we observed in w <sup>1118</sup>, during late-stage infection, flies kept on normal food had significant depletion of triglycerides compared to PBS controls. Triglyceride levels were unaffected by infection in tetracyclinefed flies (**Figure 2E**).

#### Pathology of Infection With Tetracycline-Resistant *F. novicida*

Flies kept on tetracycline food post-injection showed restricted bacterial growth and failed to exhibit metabolic dysregulation over the course of infection (**Supplementary Figure 6**). To confirm that the absence of metabolic pathology in tetracyclinefed flies was the result of low pathogen loads, rather than tetracycline having a protective effect on host physiology, we repeated all experiments with tetracycline-resistant F. novicida (U112 pKK219—GFP, herein referred to as TetR F. novicida). Flies infected with TetR F. novicida have a median survival of 5 d (**Supplementary Figure 7A**), 1 day longer than infection with wild-type F. novicida. We attribute this difference in survival to the decreased growth rate that typically accompanies antibiotic resistance in bacteria [**Supplementary Figure 7B** (34, 35)], despite this difference being marginal at all timepoints assayed in vivo (**Figure 3A**). There was no difference in survival between normal food and tetracycline-fed flies infected with TetR F. novicida, confirming that the extended lifespan observed in TetR-infected flies (fed tetracycline) is not the result of tetracycline having an unidentified effect on the flies (**Supplementary Figure 7A**).

Tetracycline-resistant F. novicida induced AMP expression to levels equal to or greater than that of wild-type bacteria (**Figures 3B,C**). This confirmed that tetracycline alone did not inhibit AMP induction. To ensure that we were looking at a comparable stage of infection, we assayed metabolism at 96 h (24 h prior to death) during TetR F. novicida infection, rather

full-factorial report of statistics can be found in Supplementary Table 2. Bacterial quantification (B) tetracycline-fed flies (Tet) have significantly lower bacterial loads both 1 and 2 d post infection (Kruskal-Wallis = 105.03, df = 7, p = 2.2e-16; Dunn's post hoc: 24 h–p = 1.3e-06, 48 h–p = 9.2e-11). Normal food (NF) flies show an exponential increase in bacterial load throughout infection. Load in tetracycline-fed flies remained constant throughout infection. Markers indicate means and bars represent SE. Bacterial quantifications were repeated twice, n = 8 samples/treatment/repeat. Metabolism (C) glucose, glycogen and triglyceride levels at an early point (6 h post-infection) of infection. (D) Glucose and glycogen levels "late" in infection (48 and 72 h for normal food and tetracycline-fed flies, respectively). Neither infection (Continued) FIGURE 2 | nor tetracycline affected glucose or glycogen levels. Groups sharing the same letter are not significantly different. (E) Infection led to a significant depletion of triglycerides during late infection in normal food flies (AOV: df =3, n = 28, F = 5.48, p = 4.3e-02, Tukey's HSD: normal food PBS—normal food Fnov, p = 0.021). Infection had no effect on tetracycline-fed flies. Large circular markers indicate means while smaller circles represent individual data points. Horizontal bar within each box represents the median. The bottom and top lines of the box represent the 1st and 3rd quartiles, respectively. Whiskers represent either the maximum and minimum values, or, the maximum and minimum values falling within 1.5x the interquartile range, in which case outliers are indicated. Groups sharing the same letter are not significantly different. Metabolic assays were repeated twice, with four samples/treatment/repeat.

than 72 h as with the wild-type F. novicida. Neither glucose nor glycogen levels differed between TetR F. novicida and wildtype F. novicida—infected flies (**Figure 3D**). TetR F. novicida infected flies exhibited metabolic dysregulation mirroring that of wild-type F. novicida infections.

We measured bacterial load upon death (BLUD) to determine whether the increased severity of metabolic dysregulation we observe over the course of infection is dependent on bacterial number. We found that BLUD did not differ significantly across treatments (**Figure 4**). Despite exhibiting the least amount of metabolic dysregulation, at their time of death, imd<sup>10191</sup> flies exhibited the highest pathogen burden, harboring on average, 37 and 59% more bacteria than w <sup>1118</sup> flies infected with wild-type and TetR F. novicida, respectively (F. novicida: imd10191—75 938 ± 8757, w <sup>1118</sup>—55 317 ± 4622; TetR F. novicida 47 505 ± 5150; mean ± se). We did not observe a significant correlation between time of death and bacterial load in any infection (**Figure 4**).

#### DISCUSSION

We have shown that infection with F. novicida leads to metabolic dysregulation in w <sup>1118</sup> flies and that part of this pathology results from a direct interaction between host and pathogen, as pathology increased concomitant with bacterial load. However, activation of the imd-derived antimicrobial peptide response also contributes to this phenotype as we found both that: imd mutants do not exhibit similar levels of dysregulation, and; w <sup>1118</sup> flies treated with tetracycline exhibited both a strong induction of AMPs, as well as a trend toward hyperglycaemia and triglyceride depletion (**Figure 1**). Whilst these subtle trends found in tetracycline-treated flies could be attributed to low bacterial loads (rather than immune activation), hyperglycaemia and triglyceride depletion were not seen in tetracycline-treated imd mutants, demonstrating that immune costs—rather than bacterial load—were responsible. However, that we do see triglyceride depletion in normal-food imd flies, suggests that at least part of this phenotype is pathogen-driven, possibly resulting from host triglyceride usurpation by the bacteria (36–38).

It is still possible that the metabolic pathology we observe is a consequence of Toll pathway activation, notwithstanding the fact that most immune activation in this infection is imd-dependent. Others have observed that Gram-negative infections can activate the Toll pathway (9, 39–41), and Toll pathway activation has been associated with effects on the insulin signaling pathway that could drive the metabolic effects we see in this infection. This possibility—that bacterial sensing via the Toll pathway is the critical driver of metabolic pathology in most bacterial infections in Drosophila, independent of the specific type of bacterium—is an interesting one for future study.

One further possibility is that F. novicida makes direct use of host-derived resources and therefore is responsible for some of the changes in host metabolism we observe. Francisella spp. preferentially use glucose as a carbon source (38, 42). It has also been shown that F. tularensis cannot use glycogen as a carbon source (43); thus, if glycogen depletion were the result of bacterial consumption of nutrients, the expected pathway would be via bacterial depletion of free sugar triggering host glycogenolysis. We observe that glycogen depletion is accompanied by increases in free sugar, such immunometabolic switches are well-known (44) and have been shown to be instrumental in the provisioning of resources to haemocytes during infection in D. melanogaster (45), further suggesting that metabolic demands of the host rather than pathogen—account for much of the observed effect.

During intracellular replication, F. novicida has been shown to metabolize glycerol for gluconeogenesis (38), supporting the possibility that the observed triglyceride depletion was bacterialdriven. However, in D. melanogaster, proliferation of F. novicida occurs predominantly extracellularly (22), and it is unclear how large of an impact these bacteria could have on host triglyceride levels. None of this precludes the possibility that F. novicida triggers host catabolism of endogenous stores to convert them into a form it can use itself. However, the fact that hyperglycaemia and loss of triglyceride and glycogen stores is seen in other bacterial infections, does imply that this reflects an aspect of the host response (20, 46), and the presence of hyperglycaemia in this infection suggests that the amount of sugar being released exceeds the amount consumed by the bacteria.

Finally, the metabolic pathology observed may be part of a moribund phenotype, with an overall worsening of condition as the animal approaches death. However, if this were the case, we would not expect to observe said pathology during infection with non-lethal bacteria like M. luteus (**Supplementary Figure 4**). Furthermore, if metabolic dysregulation and bacterial load are decoupled, as should be the case if it is caused by moribundity, we would have expected tetracycline-fed flies to exhibit metabolic dysregulation as this infection persists for several days and activates the immune response to levels comparable to flies fed on normal food.

Together these data indicate that in this infection imd activation is necessary, but not sufficient, for metabolic pathology. Tetracycline-fed flies of both genotypes maintain constant metabolite levels, and w <sup>1118</sup> maintain AMP levels, throughout the duration of the infection (**Supplementary Figure 6**); this may be due in part, to an effect of tetracycline on host physiology and metabolism (47). We found no difference in triglyceride levels between uninfected, normal food flies of the two genotypes (**Supplementary Figure 7**); thus, the differential usage of energy stores between w <sup>1118</sup> and

FIGURE 3 | Tetracycline alone is not protective. Five to nine days old adult w<sup>1118</sup> flies infected with tetracycline resistant F. novicida (TetR: OD<sup>600</sup> = 0.1). Tet-R-infected animals were put on tetracycline food at the time of injection. Bacterial quantification (A) TetR is indicated in dark blue and light blue, and wild-type F. novicida in black and gray, in w<sup>1118</sup> and imd<sup>10191</sup> flies, respectively. TetR proliferate at a similar rate to wild-type F. novicida. Circular markers indicate means and bars represent SE. Bacterial quantifications were repeated twice, n = 8 samples/treatment/repeat. Spider plots showing antimicrobial peptide transcript levels 24 h (B) and 72 h (C) postinfection. All tracings are relative to uninfected individuals of the same treatment. TetR and wild-type F. novicida are indicated in dark blue dashed and solid black lines, respectively. The area contained within the innermost quadrilateral represents induction levels falling between one and ten times that of the uninfected controls. The middle and outer quadrilaterals represent 10–100 and 100–1,000-fold induction, respectively. Antimicrobial peptide assays were repeated twice, with four samples/treatment/repeat. These data are also shown, represented differently, in Supplementary Figure 3B. Metabolism (D) late-stage infection (normal food−72 h; tetracycline−96 h) glucose, glycogen and triglyceride levels. TetR and wild-type F. novicida are indicated in dark blue and black, respectively. Infection with both wild-type and TetR F. novicida led to hyperglycaemia and depleted triglycerides, relative to their PBS controls (Glucose: Kruskal-Wallis = 11.98, df = 3, p = 1.4e-5; Dunn's post hoc: TetR—p = 5.3e-04, wild-type—p = 2.8e-3; Triglycerides: Kruskal-Wallis = 24.66, df = 3, p = 3.06e-6; Dunn's post hoc: TetR—p = 2.1e-05, wild-type—p = 4.6e-4). Glycogen levels were unaffected in the overall model (Kruskal-Wallis = 7.171, df = 3, p = 6.6e-2), but analysis of only wild-type F. novicida infected flies showed reduced glycogen relative to PBS controls (Wilcoxon = 164.5, p = 3.2e-2) whilst TetR F. novicida did not (Wilcoxon = 7, p = 0.8). Overall, we found no difference across metabolism between the two F. novicida strains. Large circular markers indicate means while smaller circles represent individual data points. Horizontal bar within each box represents the median. The bottom and top lines of the box represent the 1st and 3rd quartiles, respectively. Whiskers represent either the maximum and minimum values, or, the maximum and minimum values falling within 1.5x the interquartile range, in which case outliers are indicated. Groups sharing the same letter are not significantly different. Metabolic assays were repeated 2 or 3 times, with four samples/treatment/repeat.

imd10191, is unlikely the result of disparate resource availability. Previous work in our lab suggests that during bacterial infection, flies may disrupt insulin signaling as a means of conserving energy for immune-related activities, leading to metabolic dysregulation (20). In showing that in the absence of imd, part of the metabolic phenotype observed during w <sup>1118</sup> infection cannot be reproduced, the current work supports this supposition. Additionally, as imd mutants retain an active Toll response, this

also demonstrates that Toll is not—at least entirely—responsible for the observed phenotype.

Flies infected with TetR F. novicida and kept on tetracyclinefood exhibited near-identical infection pathology—albeit slightly protracted—to wild-type F. novicida infection. Interestingly, we found that independent of host and pathogen identity and their interaction, flies had similar pathogen loads at the time of death, despite there being a 1–2 d difference in median survivals (**Figure 4**); this observation further supports the idea

#### REFERENCES


that there is a critical bacterial load beyond which hosts cannot survive (12).

The metabolic dysregulation observed during infection is likely the result of several different factors. That multiple bacteria cause some sort of metabolic shift in their hosts suggests that both bacterial and host factors contribute to the phenotype; elucidating the need to understand the different requirements of pathogens and how these are met in a given host. Further investigation into the cross-talk between host immune and metabolic pathways under different infections has the potential to reveal some of this interaction.

### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the article/**Supplementary Material**.

### AUTHOR CONTRIBUTIONS

CV conceived, designed and performed experiments, analyzed data, and wrote the manuscript. CS performed experiments, prepared figures, and edited the manuscript. AW performed experiments and edited the manuscript. MD conceived of experiments and edited the manuscript. All authors contributed to the article and approved the submitted version.

### FUNDING

Funders have funded this research by paying salaries (CV, CS, and MD) and consumables but were not involved in the design of this project. Open access fees will be paid via institutional open access funds. This work was supported by MRC Research Grant MR/R00997X/1 and Wellcome Trust Investigator Award 207467/Z/17/Z.

### ACKNOWLEDGMENTS

We are grateful to the reviewers whose comments improved the clarity and structure of this manuscript.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2020.01419/full#supplementary-material


antibacterial immune responses in Drosophila. Science. (2002) 296:359– 62. doi: 10.1126/science.1070216


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

Copyright © 2020 Vincent, Simoes da Silva, Wadhawan and Dionne. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Beyond Host Defense: Deregulation of Drosophila Immunity and Age-Dependent Neurodegeneration

Srishti Arora\* and Petros Ligoxygakis\*

*Laboratory of Cell Biology, Development and Genetics, Department of Biochemistry, University of Oxford, Oxford, United Kingdom*

Age-dependent neurodegenerative disorders are a set of diseases that affect millions of individuals worldwide. Apart from a small subset that are the result of well-defined inherited autosomal dominant gene mutations (e.g., those encoding the β-amyloid precursor protein and presenilins), our understanding of the genetic network that underscores their pathology, remains scarce. Genome-wide association studies (GWAS) especially in Alzheimer's disease patients and research in Parkinson's disease have implicated inflammation and the innate immune response as risk factors. However, even if GWAS etiology points toward innate immunity, untangling cause, and consequence is a challenging task. Specifically, it is not clear whether predisposition to de-regulated immunity causes an inadequate response to protein aggregation (such as amyloid or α-synuclein) or is the direct cause of this aggregation. Given the evolutionary conservation of the innate immune response in *Drosophila* and humans, unraveling whether hyperactive immune response in glia have a protective or pathological role in the brain could be a potential strategy in combating age-related neurological diseases.

Keywords: aging, drosophila, neurodegeneration, immunity, immunotherapy

#### INTRODUCTION

Aging is characterized by the time-dependent deterioration of cellular function and fitness of an organism, accompanied by an increased susceptibility to diseases (1). This decline in function is inexorable, and is a key risk factor for a number of human disease pathologies such as diabetes, cancer, cardiovascular disorders, and neurodegenerative diseases (2). As the world's geriatric population continues to grow at an exceptional rate (3), a substantial economic burden is placed on the healthcare system to deal with the development of age related diseases. Therefore, understanding the mechanism of longevity and identifying targets to improve health during aging is of paramount importance.

Over the last decade there have been a number of attempts to explain the phenomenon of aging (1). Different processes that affect aging can be categorized into nine hallmarks that are shared by aging and age-related diseases (4). These include: altered intercellular communication, stem cell exhaustion, cellular senescence, mitochondrial dysfunction, deregulated nutrient sensing, genomic instability, telomere attrition, loss of proteostasis (protein homeostasis), and epigenetic alterations. These co-occur as an organism ages and are extensively interconnected. However, there are several questions regarding the interconnectedness of these hallmarks. One such question is the role of inflammation and its impact on age-related disorders (5). There is an age-dependent decline in immune response, this phenomenon is termed as "immunosenescence" (6, 7). This

#### Edited by:

*Susanna Valanne, Tampere University, Finland*

#### Reviewed by:

*Shinya Yamamoto, Baylor College of Medicine, United States Edward Giniger, National Institutes of Health (NIH), United States Arvind Shukla, National Institutes of Health (NIH), United States, in collaboration with reviewer EG*

\*Correspondence:

*Srishti Arora srishti.arora@bioch.ox.ac.uk Petros Ligoxygakis petros.ligoxygakis@bioch.ox.ac.uk*

#### Specialty section:

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

Received: *13 December 2019* Accepted: *15 June 2020* Published: *22 July 2020*

#### Citation:

*Arora S and Ligoxygakis P (2020) Beyond Host Defense: Deregulation of Drosophila Immunity and Age-Dependent Neurodegeneration. Front. Immunol. 11:1574. doi: 10.3389/fimmu.2020.01574*

**188**

process is marked by a reduction in the adaptive immune response and was initially believed to be a consequence of a progressive rise in low-grade chronic pro inflammatory status known as "inflammageing" (8–10). More precisely, as the thymus releases the last pool of naïve T-cells and that pool gets depleted each time there is an infection, older people (>65) become progressively weaker in their T-cell responses (5–7). Increase in innate immune pro-inflammatory levels has been seen as a "balancing act" to counter adaptive response reduction. Nevertheless, recent studies have observed that the two processes are mutually maintained and affect one another (11). For instance, depletion of adaptive immune cells strengthens the innate immune response causing inflammageing; similarly, the increased innate immune inflammatory mediators leads to a reduction in the number of adaptive immune cells causing immunosenescence. Cumulative studies have also indicated an age-dependent change in the innate immune cell types that leads to an overall decrease in their ability to collaborate in the initiation of the adaptive immune response [reviewed by (6)]. Inflammageing is often a result of non-resolving or sterile inflammation and is thought to be as a possible underlying basis for most age related diseases such as infections, cancer, autoimmune disorders, and chronic inflammatory diseases (12). Additionally, cell senescence in other tissues generates cytokines that signal the necessity for these cells to be removed by macrophages, to avoid what is called "by-stander senescence" propagating in tissues. This further enhances age-dependent systemic pro-inflammatory activity (5–9). The above, happens to everyone in the context of healthy aging. Some however, are "high responders" and develop neurodegeneration. Does predisposition to different levels of immunity influence this? This review focuses on the role of immunity in neurodegenerative disorders. The primary goal is to understand the importance of immune regulation and its role in aging and age-related diseases. We start by exploring the correlation between immunity and neurodegeneration connected to specific disorders. Then we explore the advantages and disadvantages of non-vertebrate animal models in studying aging and neurodegeneration. We continue with a brief introduction to Drosophila immune response and central nervous system. Finally, we conclude with studies suggesting a role for the immune system as a modulator of basal levels of age-dependent neurodegeneration and talk about the need to explore the potential role of negative regulators in immunotherapy.

#### IMMUNITY AND NEURODEGENERATION

While the correlation between inflammation and neurodegeneration is well-known (13), whether inflammation is one of its causes or a consequence remains unclear. Inflammation can be triggered as a consequence through the production of apoptotic factors and cytokines signaling during neuronal death However, immune cells produce neurotoxic cytokines that could cause death of neurons (14). Initially, the activation of the immune response in the central nervous system (CNS) was believed to be responsible for the elimination of infectious agents and the clearing of debris after injury, suggesting a neuroprotective role of inflammation. A positive role of antimicrobial peptide (AMP) production and aging has been suggested by Loch et al. (15). Genome wide association studies (GWAS) have reported the activation of numerous genes of the inflammatory pathway during aging (16). Age is the greatest risk factor for neurodegenerative disorders and age-related chronic activation of the immune response is a shared feature among many neurodegenerative disorders (17). However, the cause of this sterile inflammation is still unknown.

Studies in animal models indicate the importance of inflammation in several neurodegenerative disease pathologies (18). Altering expression of Cdk5 protein kinase (Cdk5α) leads to disruption in autophagy that in turn leads to upregulation of AMP and age-dependent degeneration of dopamine neurons in Drosophila (19). Neuroinflammation has been a crucial factor for the pathogenesis of diseases such as Alzheimer's Disease (AD) (20). Microglia, the resident innate immune cells of the CNS are shown to be chronically activated around these plaques. It is believed that the uncontrolled inflammation of these cells leads to the secretion of multiple neurotoxic factors such as inflammatory mediators and reactive oxygen species by glial cells that aggravate the pathology of the disease (14). It is further demonstrated that mutations in microglial protein TREM2, PLCG2, and ABI3 increase the risk for AD (21). Additionally, molecular and pathological interaction studies have established glial expression of TREM2/TYROBP as a key factor in tau mediated neurodegeneration (22). Activated microglia is suggested as a potential marker to detect AD before the appearance of plaques (23). Additional risk genes for late onset AD connected to microglia and immunity have been identified recently (24). Genetic analysis of these late-onset AD risk genes identified a transcriptional network of 12 largely microglial genes that form a transcriptional network (25). Six of these (OAS1, LAPTM5, ITGAM, ABI3, PLCG2, SPI1) have good Drosophila homologs expressed in the nervous system (our unpublished observations).

Drosophila models illustrate the importance of the Toll mediated NF-κB response in the neurotoxicity cause by the presence of Aβ42, an isoform of the beta amyloid protein. Down regulation of this immune pathway was shown to reduce the pathological activity of Aβ42 (26). Evidence in both human and animal model studies have illustrated the correlation between inflammation and Parkinson's disease (PD) (27). Mechanisms of neuronal dysfunction such as mitochondrial dysfunction and oxidative stress have been linked to pathogenesis of PD (28). Dopaminergic neurons (DA) in the midbrain are shown to be sensitive to pro inflammatory cytokines, reactive oxygen species, and chemokines such as TNF-α and IFN that exacerbating neuronal lesions (29). Additionally, there is a rich population of microglia in the substantia nigra, which is the region of the brain that shows the most DA neuron loss in PD patients (30). Studies have observed correlation between deposition microglial activation and alpha-synuclein making microglia an attractive therapeutic target (31, 32). However, PD is considered as a condition that is hypothesized to starts in the intestine as

chronic inflammation, then may transfer α-synuclein to the brain through the vagus nerve (33).

Transgenic mice lacking the TNF- receptor demonstrate a reduction in the TH-immunoreactivity after being exposed to MPTP (1-methyl-4-phenyl-1,2,3,6-tetrahydropyridin). Prolonged use of non-steroidal anti-inflammatory drugs such as ibuprofen is shown to reduce risk of PD (27). Consistent with these studies, an increase in circulating cytokines and increased microglial activation have also been linked to early stages of Huntington's disease (34, 35). However, production of pro inflammatory molecules is not just limited to microglia. Other types of glia cells such as astrocytes are also used to investigate the progression of inherited ALS (36). Studying these processes in humans is extremely challenging. Therefore, to explore the processes that govern aging requires accessible model systems that help provide critical insights into the cellular and molecular levels of aging.

#### ANIMAL MODELS OF AGING

There are several challenges in studying primate subjects in aging research. These include a number of ethical issues, environmental factors, as well as their relatively long lifespan. Consequently, aging researchers have turned to unicellular or small animal models to investigate the genetic and physiological mechanisms related to human aging and longevity. These models allow us to better control for several intrinsic and extrinsic factors such as uniformity in background genetics, large sample sizes, genetic tractability as well as environmental factors such as managed nutrient availability and chemically-defined diets. These conditions make mechanistic analysis easier and ultimately help identify novel pharmaceutical targets. Some of the popular models used in aging research include: the budding yeast (Saccharomyces cerevisiae), the nematode worm (Caenorhabditis elegans), the fruit fly (Drosophila melanogaster), and mouse (Mus musculus).

#### Saccharomyces cerevisiae

Studies in model organism have identified conserved pathways that influence the rate of aging (37). The simplest organism that can be used to study eukaryotic aging is Brewer's yeast or Saccharomyces cerevisiae. This single celled living organism shares a number of genes with humans, out of which a significant number carry out the same function in both organisms (38). This includes mechanisms that facilitate pathogen recognition during its vegetative development (18). Fungi possess a class of cytosolic NOD (Nucleotide Oligomerization Domain)- like receptors or NLRs are responsible for self and non-self-recognition. These fungal receptors share homology with the effector domains of several plant or animal NLRs and provide a unique opportunity to explore infectious host-pathogen interactions (39).

Aging in yeast can be studied using two different models. The first, replicative lifespan (RLS) that describes the total number of cell divisions a single mother (virgin) cell undergoes, the second, the chronological lifespan (CLS) that represents the length of time a cell can stay viable in a post mitotic state (40). CLS shows an elevation in DNA damage that is a characteristic that resembles that of post mitotic cellular aging in humans (41). However, both models rely on nutrient availability and negatively affect each other (42). The small genomic size and rapid generation time of 3 h makes yeast a great model for high throughput screening and exploiting genetic interactions that are thought to be involved in human aging (42). It is also used to study the effects of dietary restriction (43), oxidative stress (44), and target of rapamycin (TOR) nutrient response pathway on age related phenotypes (42, 45) Moreover, yeast models are used to study a number of age related diseases including Werner syndrome, (41) Huntington's disease (46), Alzheimer's disease (47), Parkinson's disease (48). The lessons from yeast have given us valuable insights into how stress and aging are modulated in higher organisms. However, yeast lacks the complexity of a higher eukaryotic cell and intercellular interactions that are of major importance in aging and age-related disorders.

#### Caenorhabditis elegans

Another invaluable model system to study aging is the small nematode worm, Caenorhabditis elegans. C. elegans have a short lifespan of about 2–3 weeks at 20◦C. This small worm grows to be about a millimeter in length and displays complex behavior such as avoidance behavior when exposed to pathogens. It is compatible with a wide range of genetic techniques including chemical mutagenesis screens, CRISPR, and RNAi. Unlike yeast, it allows us to study tissue-to-tissue communication by tissuespecific transgenic expression and knockdown techniques in a multicellular context. C. elegans lack an adaptive immune response and are devoid of any migratory innate immune cells. Instead the protective immune response relies on three lines of defense. The first is avoidance behavior in which the worm can discriminate between different species of bacteria by recognizing odors of specie specific molecules such as cyclic pentadepsipeptide biosurfactant serrawettin W2 produced by Serratia marcescens (49). The second line of defense consists of physical barriers. The strong exoskeleton of C. elegans is made up of collagen and chitin that creates a physical barrier limiting the entry of potential pathogens. Additionally, a pharyngeal grinder prevents pathogens from accessing the intestines. The third and final line of defense is the humoral response which involves the activation of conserved signaling pathways (including MAP kinase cascades) that leads to the production of several antimicrobial peptides (50, 51).

C. elegans allow us to experimentally demonstrate the roles of several other conserved processes in aging such as caloric restriction, mitochondrial pathways energy metabolism, endocrine signaling, and signal transduction, the stress response, protein translation, and gene expression in aging. (52). At the convergence of immunity and aging, recent studies have shown the role of innate immunity regulated by p38 signaling and the transcription factor ATF-7 as responsible for the lifespan extension caused during dietary restriction (53). However, C. elegans are evolutionarily distant from humans and has a very different nervous system organization of just 302 neurons leading to behaviors unique to its lifestyle (54). Nevertheless, studies of neuronal cell death in worms has implicated proteins very closely related to mammalian calpains and cathepsins (55). Calpains are a family of calcium regulated cysteine proteases that are highly expressed in neurons. They affect a wide range of cellular functions including cell division, proliferation, migration, and death. In neurons, these proteases have been linked to synaptic plasticity and neurodegeneration (56). The calpain mediated cleavage of carbonylated Hsp70.1 due to oxidative stress leads to loss of lysosomal integrity and rupture. Among the contents of the lysosome, a hydrolytic enzyme (cathepsins) is also released that takes over the role of a "death-executing proteases" by degrading several cellular proteins (55, 57). This Calpainmediated cleavage of Hsp70.1 helps elucidate the importance of proteolysis in neuronal death and serves as a promising target for preventive interventions of neuronal death (57). Thus, at the cellular level, worm genetics can provide new insights into brain cell death.

#### Mus musculus

While yeast and worms have broadened our understanding into the cellular mechanisms of aging and age-dependent neurodegeneration, they fail to replicate system level neurological changes that occur during human aging. Therefore, mammalian model organisms are essential to unravel these complex mechanisms. What makes the mouse an indispensable model is the easy genetic manipulation, short lifespan, low-cost (compared to other primate models), and considerable similarities with human physiological and cellular function (58). Many mouse models of human aging have been developed and characterized, including models for Werner syndrome (59), Ataxia telangiectasia (60), Alzheimer's (61), and Parkinson's diseases (62). Although the degree of complexity of the mouse brain is lower than that of a human, there are several cellular similarities of the nervous system. They show complex behaviors and are a good tool to measure cognitive changes in neurodegenerative disorders (63).

Studies in mice with Alzheimer's disease have highlighted rapamycin as a valid therapeutic approach for prevention or treatment of AD (64, 65). Mouse models that display lifespan extension and rise in delayed aging phenotypes are in line with the observation that DNA metabolism influences aging. Furthermore, factors such as caloric restriction and defects in genome maintenance have been investigated in mice (66). Li et al. demonstrated the role of chronic high fat diet in mice causes loss of neuronal stem cells in the hypothalamus via IKKβ/NFκB activation that eventually leads to obesity and pre-diabetes (67). The hypothalamus the neuroendocrine functional center of the body is also responsible for systemic aging though NF-κB signaling (25) and thus provide a potential therapeutic approach to combat age and age related disorders (68). This seems to be an evolutionary conserved component of aging as NF-κB in the brain is a major life span determinant in Drosophila as well [(41), see below]. Even though laboratory mice are an admirable model to study some age- related phenotypes, they do not fully mimic the mechanism. Laboratory mice are inbred and age very quickly. They invest more in reproduction and less in somatic maintenance and therefore do not display the tradeoff between fecundity and longevity observed in humans. They also do not recapitulate age related disease pathologies as seen in humans (e.g., Werner diseases models and amyloid plaques in AD mice models).

#### Drosophila melanogaster

Over the last two decades, Drosophila has developed as a powerful tool to investigate human disease mechanisms. It has orthologs of ∼65% of all genes causing heritable diseases in humans (69, 70), making it an attractive model organism to address novel lines of inquiry for human diseases (71). Moreover, for every one of these genes, the fly will most of times have one copy while humans will normally have a group of genes with the same function. The fruit fly is small, has a low cost of rearing and is easy to manipulate in the laboratory. It has short generation time of 10 days at 25◦C, a relatively short lifespan and produces a large number of eggs which boosts statistical relevance of the data obtained. Drosophila shows complex behavioral phenotypes including social aggregation, re-enforced learning as well as sleep activity that help address questions of brain function. Transgenic fly lines can be created using numerous sophisticated genetic and molecular tools such as insertions of P-elements (72) CRISPR, RNAi silencing, tissue specific GAL4-UAS expression system. Additionally, genome-wide genetic screening, genomewide analyses with deep sequencers, such as RNA-seq and ChIPseq, and metabolomics analyses allows us to enquire the cellular and molecular mechanisms of aging and age-related diseases. Drosophila has been crucial in the discovery and understanding of innate immune signaling and the development of the nervous system. The fly exhibits multiple physiological changes associated with aging and age related diseases such as reduced locomotive ability (73, 74), impaired learning and memory (75), progressive decline in intestinal barrier function (76), increased inflammation (77), reduced reproductive capacity, and altered neuronal function (78, 79). Additionally, several environmental manipulations such as effects of dietary restrictions are easy to observe (80).

The process of development of neurons is conserved from flies to humans. Drosophila has a relatively complex nervous system that is separated from the rest of the body with the blood brain barrier built by glial cells and neurons (81). It's CNS contains about 200,000–300,000 neurons can be histologically divided into two distinct regions (82): the neuronal cell cortex, formed by all the neuronal cell bodies, and a synapse dense neuropil, to which all the dendrites and axons project (83). The fly brain is a sophisticated structure that has several substructures: including the antennal lobes, the mushroom bodies, the central body complex, the protocerebrum, the optic lobes, the posterior slope, and lateral deutocerebrum. Sensory organs and the musculature send signals to the CNS via peripheral nerves. The neurons in these associated structures are supported with glial cells. Apart from being the resident immune cells for the CNS, glial cells are responsible for maintaining ionic homeostasis, recycling neurotransmitters, and for the formation of the blood brain barrier (83). Drosophila glial cells can be largely categorized on the basis of their location and/or morphology. There are six morphologically and molecularly distinct glial subtypes; perineurial glia (PG), subperineurial glia (SPG), cortex glia (CG), ensheathing glia (EG), astrocytes-like glia(ALG), and wrapping glia (EG in the PNS) (84). The surface of CNS and the peripheral nerves are covered with a thick carbohydraterich lamella secreted by perineural glia (PG) and macrophages (85). This PG layer is discontinuous and forms glia–glia pleated septate junctions (pSJs) with subperineural glial cells (SPG) that lie directly below the PG layer. These cells establish the Blood Brain Barrier (BBB) and separate the neuronal elements from the potassium-rich hemolymph. Apart from the BBB, the peripheral nerves have a specialized form of ensheathing glia called wrapping glia, that encloses motor and sensory axons (86). Deeper in the CNS beneath the BBB lie cortex glia, ensheathing glia, and astrocytes-like glia which are closely associated to neurons (87). Cortex glia or cell body associated glia are found within the cell cortex and invade the space between neuronal cell bodies. These cells are in contact with the tracheas and the BBB, suggesting that they are likely responsible to transfer nutrients and gases from the hemolymph to neurons (81). The ensheathing glial cells compartmentalize the brain by forming a sheath around the neuropil (88). These cells are responsible for the phagocytoses of axonal debris (89). As the name suggests astrocyte-like glial cells are functionally and morphologically similar to mammalian astrocytes (90). They are responsible for the maintenance of neurotransmitter homeostasis and in regulating circadian rhythm (91).

Additionally, a large amount of effort has been exerted in creating many distinct Drosophila models for a range of neurological disorders such as Parkinson's disease (PD) (92), Alzheimer's disease (AD) (93, 94), and polyglutamine diseases (polyQ) (95, 96). Many of these diseases are caused by abnormal production or accumulation of different proteins such as the accumulation of Lewy bodies in PD, amyloid plaques in AD, and inclusions in polyQ diseases. These protein defects are not normally observed in Drosophila. However, they can be artificially produced in flies by introducing human genes into the genome and over expressing them in neurons through the UAS/GAL4 system (97). This is both an advantage and a disadvantage. This technique helps replicate human-like morphological lesions of these diseases and devise screens to genetically identify mutations that suppress the extend of the resulting lesions. However, it is difficult to distinguish the immune responses to such protein build-up from mere nonspecific stress responses due to overproduction of an exogenous protein (98).

### INTRODUCTION TO DROSOPHILA IMMUNITY

The innate immune system, an immune reaction with broad specificity, is an organism's first line of defense. It is centered on receptors, which target conserved features of microbial invaders and expeditiously activate downstream cascade to destroy pathogens (99). In jawed vertebrates and some jawless fishes (lampreys) this activation leads also to the induction of adaptive immunity. Unlike those vertebrate categories however, insects lack an adaptive immune system and therefore rely on a relatively sophisticated set of innate defense responses for their survival. The development and function of these reactions are shown to be shared with higher organisms and can be used to study innate immunity and inflammation in humans (100).

Due to the wide range of genetic manipulation techniques it offers, Drosophila has been a powerful model to study innate immunity (101). It utilizes a wide range of actions to form effective barriers against pathogens, first of which is a local immune response. This includes the elimination of incoming pathogens by constitutive secretion of AMPs and by reactive oxygen species (ROS) in barrier epithelia such as gut, genitals, cuticle (102, 103), followed by a cellular response which includes engulfment, entrapment, and melanization of the pathogen (104– 106). The final response is marked by the rapid synthesis of antimicrobial peptides (AMPs) in the haemocytes and the fat body. The AMPs are regulated by two signaling pathways: The Toll pathway, which was the first in the family of Toll-like receptors discovered in a wide range of organisms from sea urchins to humans, and the IMD pathway homologous to the tumor necrosis factor receptor 1 (TNFR1) in mammals (107).

The Toll-mediated responses are triggered by bacterial or fungal infection which leads to the activation of two Rel transcription factors, Dif and Dorsal that regulate hundreds of genes [reviewed in (100)] (**Figure 1**). Apart from its role in immunity, the Toll pathway plays a crucial role in the determination of the dorsal-ventral polarity during Drosophila early embryogenesis (108, 109). In order to initiate the Toll response, bacterial or fungal pathogens, are sensed by receptors in the form of Peptidoglycan Recognition Proteins (PGRPs; in this case PGRP-SA) or Glucan Binding Proteins (GNBP1, GNBP3) or through cleavage of endogenous proteases (such as Persephone or Psh) (110–112). The next step gives rise to an extracellular proteolytic cascade that culminates in the proteolytic cleavage and activation of the Toll receptor ligand Spatzle (Spz) leading to its activation (113). Spatzle binds to the Toll receptor that recruits MyD88 through its TIR domain, which further interacts with Tube and Pelle through their respective death domains (DD) and promote the phosphorylation of Cactus. Cactus is the Drosophila IkB homolog and is bound to Dorsal and/or Dif, inhibiting their activity and nuclear localization. Once degraded, Dorsal and Dif translocates to the nucleus and ultimately leads to the transcription of AMPs and other target genes [reviewed in (114)].

Another evolutionarily conserved singling cascade is the immune deficiency (Imd) pathway (**Figure 1**), the activation of which is achieved with the help of two PGRP receptors namely, PGRP-LC and PCRP-LE. These receptors bind better to DAP-type peptidoglycan present on Gram-negative bacteria and Gram-positive bacilli and trigger a response, that finally leads to the activation of Rel/NF-κB transcription factor Relish (115). The activated PGRP receptors initiates a signaling cascade and recruits a protein complex containing Imd, a death domain protein, dFadd, and Dredd. Dredd is the caspase-8 homolog, is activated by Iap2 (inhibitor of apoptosis 2) and cleaves Imd to reveal a site important for its role as a transient signaling hub. Iap2 then ubiquitnates Imd and recruits Tab2/Tak1 complex to the reaction site that phosphorylates the Drosophila IKK complex (116). Relish is activated by two events: the phosphorylation of

its N terminal by Tab2/Tak1 complex and the cleavage of its C terminal ANK repeats domain by Dredd. The activated N terminal (Rel-68) enters the nucleus and initiates transcription of AMPs (116).

The improper activation of the immune signaling pathways is associated with inflammation, cancer, and neurodegeneration (77, 117), and leads to developmental defects during ontogenesis (118–120). To prevent the harmful consequences of unwarranted activation, the pathway is firmly regulated by extracellular and intracellular proteins.

#### Negative Regulation of IMD

The IMD immune response is tightly regulated at many levels, first of which is the dilution of the activating signal. This is done by the secreted PGRP-SC and PGRP—LB amidase that breakdown bacterial peptidoglycan into non-stimulatory fragments in the extracellular matrix (119, 121). On the plasma membrane, the three PGRP-LC isomers interact with each other to suppress spontaneous dimerization and reduce the number of functioning receptors (122). Additionally, PGRP-LF binds with PGRP-LC to form non-signaling heterodimers and down regulate the response (123). Intracellularly, PIMS/pirk/rudra coimmunoprecipitates with Imd causing a disruption in its association with the cytoplasmic tail of PGRP-LC receptor leading to its the depletion from the membrane. Pirk mutant have constitutively activated immune response and are short lived. Moreover, flies over expressing pirk have a reduced Imd response (124–126). Dnr1 inhibits the activity of Dredd caspase by promoting its proteolytic degradation. Dnr 1 mutants have shorter lifespans and exhibit age-dependent neuropathology (127). The activity of the Dredd caspase is also impaired by Caspar, which inhibit Dredd-dependent modification of Relish and further blocking its translocation to the nucleus (13). The activity of NF-kB is regulated by several ubiquitinmediated interactions, deregulation of these factors cause chronic inflammation and cancer (128). Further negative regulators of the Imd pathway include SkpA, dUSP36, CYLD, POSH, Trabid, and transglutaminase (TG). Drosophila ubiquitin-specific protease 36 (dUSP36) inhibits the K63-polyubiquitinated Imd build up and promotes its degradation (129). It is a negative regulator of the pathways as silencing dUSP36 constitutively activates the IMD signaling pathway. This activation is lost in germ free flies leading to the hypothesis that this interaction might be microbiome dependent (130). K63-linked ubiquitination of dTAK1 is monitored by another negative regulator of IMD known as Trabid. Flies lacking this protein show a remarkable increase in the amount of the IMD target, Diptericin with a dramatic reduction of the lifespan (131). A further negative regulator of TAK1, POSH (Plenty-of-SH3s) prevents engagement with the JNK scaffold (132). Downstream, the cylidromatosis disease homolog dCYLD interacts with the Drosophila IKKγ homolog Kenny and disrupts downstream signaling (133). In its absence, triglyceride and AMP levels increase (133). At the level of Relish finally, SkpA and the proteasome-ubiquitin pathway (134) and Transglutaminase (135) suppress NF-κB activity.

#### PREDISPOSITION TO AN OVERACTIVE IMMUNITY CAUSES NEURODEGENERATION

Human genetics and animal model research have illustrated the correlation between innate immunity in the brain and the pathogenesis of neurodegenerative disorders (**Table 1**). Increasing amount of evidence suggests that the accumulation of aggregated proteins is only a part of the pathology of neurodegenerative disorders and not the full story [reviewed in (143)]. Increasing evidence suggests the role of the immune system as an aetiological mechanism that influences not only the pathology of the diseases but also modulates basal levels of age dependent neurodegeneration in the context of healthy aging (144).

#### Loss of Negative Regulation of IMD

Cao et al. illustrated that chronic activation of the immune response in the wild type Drosophila brains causes neurodegeneration (127). Flies with loss of function mutations in a Relish repressor gene dnr-1 show signs of early neurodegeneration and an increase in the number of Relish target genes transcripts in the fly brain. The authors suggest the cause of this neurodegeneration as AMP-associated toxicity caused by constitutive expression or an extremely high level of AMPs present in neurons or glial cells. The study also revealed that the overexpression of AMPs in nervous tissue can cause neurodegeneration and established a causative relationship between neurodegeneration and IMD signaling. However, overexpression of AMPs brings expression to much higher levels than the dnr-1 mutant and therefore more work is needed to prove this point. Nevertheless, the possibility of a neuroprotective role of negative regulators in neuronal viability is clearly suggested here (127).

A similar result were obtained by Kounatidis et al., who demonstrated an age-dependent increase in NF-kB- controlled immune activity in Drosophila in the context of healthy aging (77). Most of it was dependent on the microbiota, as germ free flies had much reduced age-dependent AMP increase compared to conventionally reared insects (77). Nevertheless, there was consistently a 2–4x age-dependent AMP increase in germ free flies as was a clear sterile inflammation in the brain. Moreover, the loss of Trabid, Pirk, and TG in neuronal tissue resulted in shortening of lifespan, locomotion defects, and the formation of brain lesions. This phenotype was rescued once Relish was suppressed in glial cells of these flies. In wild type flies, suppressing relish in glial cells resulted in lifespan extension. Therefore, genetic predisposition to higher immune levels with mutation in trbd, pirk, and tg led to early neurodegeneration and curtailed lifespan (77).

#### Autophagy and Immunity

An interesting connection between autophagy, immunity and neurodegeneration was recently made by the observation that mutants for the Cdk5 protein kinase have increased AMP expression in the brain and loss of dopaminergic neurons. This happens because loss of Cdk5 disrupts autophagy and this results in increased levels of immunity (19). This point is important since autophagy seems to be necessary and sufficient to drive the increase in AMP levels. Given the dysregulation of Cdk5 and innate immunity in human neurodegeneration and the conserved role of this kinase in the regulation of autophagy, this sequence of events is likely to resemble what happens in humans (19). However, the connection between immunity and autophagy remains largely unexplored. One indication is the interaction between Kenny (IKKγ) and the autophagy protein Atg8, which targets Kenny for selective degradation. Loss of Atg8, "releases" Kenny, enhancing IKK signaling, and resulting in chronic IMD induction (145).

#### Neurodegenerative Disease Models

The Penetrating traumatic brain injury (pTBI) model show a greater expression of AMP genes and an over activation of the innate immune response in both young and older flies (136). The positive interaction between pTBI and aging was further supported by the high expression of Imd negative regulators in older pTBI fly brains. The study indicated that aging exasperates the immune response caused by pTBI and causes neurodegeneration. Additionally Yorkie, a co-activator TABLE 1 | Summary of papers at the junction of immunity and neurodegeneration in *Drosophila*.


*We only list those showing a causative link between immune activity or immune-related signaling and neurological phenotype.*

of Hippo pathway was also shown to reduce polyglutamine (PolyQ)-mediated neurodegeneration by negatively regulating Toll and Imd pathways via cactus and relish, respectively (137).

Recently, transcription of innate immune genes were observed as the prominent response to paraquat in a Drosophila model of PD (138). Interestingly, Relish knock down in dopaminergic neurons conferred resistance to paraquat and rescued both motility defects and loss of dopaminergic neurons. The study indicates that the immune reaction might not be protective and indicate potential drug targets for preventing neuronal loss during PD. Immunity induced neurodegeneration can explain the neurodegenerative phenotypes observed in both ataxia– telangiectasia mutated (ATM) gene and retinal degeneration in norpA (no receptor potential) mutants (139). Reduction in the ATM kinase activity in the glial cells may be responsible for the increased innate immune response through protein phosphorylation and cause neurodegeneration in these mutants. Furthermore, retinal degeneration in norpA mutant flies was shown to be dependent on Relish and Dredd.

#### Neuroprotective Roles of Immunity

In addition to the role of the long-term heightened IMD signaling in causing neurodegeneration, there is also a neuroprotective aspect of glial signaling components connected to immunity. Ray et al. showed the neuroprotective role of the glial engulfment receptor, Draper, in Drosophila model of AD (140). Overexpression of glial draper reverses amyloid (Aβ) accumulation along with AD associated behavior phenotypes. They also show that protein degradation pathways are expressed downstream to Draper in response to amyloid accumulation. This supports the theory that glial cells may be responsible for the clearance of neurotoxic amyloid peptides in the brain through a Draper/JNK/STAT92E signaling cascade (140). Draper is also observed to have a significant role in clearance of damaged axons. Purice et al. observed an age-dependent decline in the levels of Draper that causes dysfunctional glial engulfment in older flies (141). Dying neurons activate Toll receptor ligand, Spz, in the cortex glia, that further drives the expression of Draper to ensure efficient clearance of the neuron (142).

#### Gut-Brain Axis

Recent studies have also focused on the role of gut- brain crosstalk and neurodegeneration. Wu et al. highlighted the effect of enteric infection in AD progression (146). Gut dysbiosis in AD mutant flies caused an increase in haemocyte recruitment to the brain and activation of TNF-JNK mediated neurodegeneration. Neurodegeneration and reduction in lifespan were rescued in flies with genetically depleted Eiger (an activator for JNK pathway) in the brain, further supporting the hypothesis. Westfall et al. explored how symbiotic and probiotic formulation can influence gut brain signaling and delay the progression of AD (147).

#### Limitations

Needless to say, no model system is without limitations. It is important to note that like most invertebrate organisms, the fruit fly is evolutionarily distant from humans and does not accurately mimic all the neurodegenerative phenotypes observed in human diseases such as tau aggregates and plaques (148). Drosophila lacks an adaptive immune response making it difficult to recapitulate complex changes in the immune response, that might take place during aging. Additionally, the hemolymph of the fly contains primitive hemocytes which cannot undergo DNA rearrangement and somatic hypermutation like mammalian lymphocytes. Unlike mammals, flies do not possess microglia rather, all glial cells can perform microglial tasks; such as engulfing neuronal corpses during development (84). However, this restricts studies that attempt to understand the complicated relationship between the immune system and the relation between neuroprotection and neurodegeneration. Nevertheless, the studies summarized here highlight evidence suggesting that the immune system plays an important role in neurodegenerative disorder in Drosophila. Two key contributors to lifespan reduction and neuropathy are overproduction of AMPs and impaired phagocytosis. Even though animal models do not represent the diseases completely (for example in the lack of direct orthologs for the human proteins prone to aggregation in AD or PD), comparative studies of brain development and the innate immune response have demonstrated significant evolutionary conserved mechanisms between vertebrates and invertebrates. Moreover, the deregulation of innate immunity as etiology for neurodegeneration stands in Drosophila even in the absence of tau or β-amyloid. There is a large therapeutic potential of immunomodulation and therapeutic immunization (149) to help combat the development of such diseases by screening fast in whole animal models such as the fly. Moreover, since the role of immune activity in microglia and astrocytes in neurodegeneration is well-documented, we could envisage that negative regulators of immunity could be potential candidates for early interventions.

#### REFERENCES


#### CONCLUSION

The precise mechanism of the development of neurodegenerative diseases is still unknown and this presents a challenge for the development of treatments and therapies. Currently, therapies focus only on treating isolating disease symptoms such as protein accumulation, sleep disturbances, memory loss, or behavioral changes. Additionally, disease modifying therapies are largely unsuccessful and there is need for more drug candidates to enter the pipeline. Since most of the cases of neurodegeneration are only diagnosed after severe neuronal loss. Exploring preclinical symptoms as potential therapy can facilitate the development of treatments for the early symptoms of the disease. Aberrant immune regulation resulting in chronic inflammation long before neurological symptoms manifest themselves may be at the root of these diseases. We believe that Drosophila represents an ideal compromise between its relevance to humans and its demographic power and genetic tractability, making it a model of choice for understanding mechanistic aspects of age-related neurodegeneration.

### AUTHOR CONTRIBUTIONS

SA wrote the first draft and drew the figures. PL worked on subsequent drafts with SA. All authors contributed to the article and approved the submitted version.

### FUNDING

This work was supported by the European Research Council (310912 Droso-Parasite), project grant BB/P00569/1 from the Biological and Biotechnological Sciences Research Council UK and the EP Abraham Cephalosporin Trust Fund Grants CF 319 and 333, all to PL.

#### ACKNOWLEDGMENTS

We thank all members of the PL's lab for critical reading and comments on the manuscript.


immune response to monomeric DAP-type peptidoglycan. Nat Immunol. (2006) 7:715–23. doi: 10.1038/ni1356


of microbes. These pep- tides are synthesized in both surface epithelial ti. Curr Biol. (2002) 12:1728–37. doi: 10.1016/S0960-9822(02)01214-9


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

Copyright © 2020 Arora and Ligoxygakis. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Immune Control of Animal Growth in Homeostasis and Nutritional Stress in Drosophila

#### Preethi P<sup>1</sup> , Ajay Tomar 1,2, Sukanya Madhwal 1,3 and Tina Mukherjee<sup>1</sup> \*

1 Institute for Stem Cell Science and Regenerative Medicine (inStem), Bangalore, India, <sup>2</sup> The University of Trans-Disciplinary Health Sciences and Technology, Bangalore, India, <sup>3</sup> Manipal Academy of Higher Education, Manipal, India

A large body of research implicates the brain and fat body (liver equivalent) as central players in coordinating growth and nutritional homeostasis in multicellular animals. In this regard, an underlying connection between immune cells and growth is also evident, although mechanistic understanding of this cross-talk is scarce. Here, we explore the importance of innate immune cells in animal growth during homeostasis and in conditions of nutrient stress. We report that Drosophila larvae lacking blood cells eclose as small adults and show signs of insulin insensitivity. Moreover, when exposed to dietary stress of a high-sucrose diet (HSD), these animals are further growth retarded than normally seen in regular animals raised on HSD. In contrast, larvae carrying increased number of activated macrophage-like plasmatocytes show no defects in adult growth when raised on HSD and grow to sizes almost comparable with that seen with regular diet. These observations imply a central role for immune cell activity in growth control. Mechanistically, our findings reveal a surprising influence of immune cells on balancing fat body inflammation and insulin signaling under conditions of homeostasis and nutrient overload as a means to coordinate systemic metabolism and adult growth. This work integrates both the cellular and humoral arm of the innate immune system in organismal growth homeostasis, the implications of which may be broadly conserved across mammalian systems as well.

#### Edited by:

Dan Hultmark, Umeå University, Sweden

#### Reviewed by:

Michelle L. Bland, University of Virginia, United States Michael J. Williams, Uppsala University, Sweden

> \*Correspondence: Tina Mukherjee tinam@instem.res.in

#### Specialty section:

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

Received: 12 November 2019 Accepted: 10 June 2020 Published: 31 July 2020

#### Citation:

P P, Tomar A, Madhwal S and Mukherjee T (2020) Immune Control of Animal Growth in Homeostasis and Nutritional Stress in Drosophila. Front. Immunol. 11:1528. doi: 10.3389/fimmu.2020.01528 Keywords: myeloid cells, high sugar, metabolism, inflammation, stress, innate immunity, insulin

## INTRODUCTION

The immune system comprises circulating cells and blood-forming tissues whose main function is combating infections. The development of this system is metabolically expensive and often associated with trade-offs with other physiological functions such as reproductive fitness (1, 2) and survival, especially in conditions of nutrition challenge (3). Development of a robust immune system and its impact on animal growth has been described in several studies across animal models. Decreased immune function in flies with increased body mass (4) or improved resistance with reduced competitive ability on a poor diet (5) are some studies illustrating this robust connection. Nevertheless, any understanding of animal growth from the standpoint of immune homeostasis is poorly explored. Drosophila is a well-established and a conserved model system for addressing questions pertinent to blood development (6) and mechanisms regulating organismal growth (7) mechanisms. In this study, we have used Drosophila to explore the consequences of altering immune homeostasis early in animal life on organismal metabolism and growth control and the implications of nutrient overload in this phenomenon.

Drosophila blood cells akin to vertebrate myeloid cells perform functions central to the maintenance of general animal physiology that includes wound healing response (8), antimicrobial functions (9), hypoxia response (10), innate immunity, and response to wasp-parasitization (11). Of the three different types of blood cells prevailing within the Drosophila larvae, the platelet-like crystal cells are implicated in wound healing and hypoxia response, whereas lamellocytes are involved in the response to parasitic wasps. The phagocytic plasmatocytes constitute 95% of the differentiated mature cell type. These phagocytic blood cells, akin to vertebrate macrophages, perform functions relevant for clearance of apoptotic cells and invading particles, neuronal pruning, tissue remodeling, and antimicrobial functions (12). Immune cells in Drosophila are derived from immune progenitor cells whose development, much like in vertebrates, is derived from two distinct waves of hematopoiesis: the primitive and the definitive. The primitive wave of hematopoiesis occurs in the early embryonic stage where the first pool of blood precursors gets specified from embryonic head mesoderm (13, 14). These hematopoietic precursors proliferate and differentiate into mature hemocytes and constitute the larval circulatory and sessile pools of blood cells detected in the larvae (15) and later in adult stages (12). Definitive hematopoiesis initiates in a larval hematopoietic organ called the lymph gland, which gets specified at the late stages of embryonic development (13, 14). The lymph gland comprises multipotent undifferentiated blood progenitor cells that proliferate and mature to give rise to differentiated blood cells during larval stages of development. By the early pupal stage, the blood progenitor cells completely differentiate, after which the lymph gland disintegrates to release these mature hemocytes into circulation contributing to immune cells in the adult fly (6, 12). By early pupal stage, blood cells complete differentiation into hemocytes, after which the lymph gland disintegrates to release these cells into circulation in the adult fly.

The cues that regulate blood development and homeostasis in Drosophila are of both local as well as systemic origin (6). The systemic cues include environmental (odors and sensory stimulation) (16, 17) and of nutritional origins (18), the latter, more relevant to this study. During blood development, blood progenitor cells directly sense amino acids and insulin to sustain their maintenance. Starvation or loss of insulin signaling results in the differentiation of progenitors and activation of inflammatory responses, recapitulating a diabeticlike condition (18). Nutrient-rich conditions (19) or any change in the physiological state of the developing larvae (3) have been shown to alter immune cell numbers as well. As immune cells undergo functional maturation, the macrophagelike plasmatocytes perform lipid-scavenging functions and exert systemic control on glucose homeostasis and survival on lipid-rich diet (20). Taken together, these studies provide evidence of nutrient-dependent modulation of immune cell development, homeostasis, and signaling. What remain unclear are the underlying contributions of the immune cell changes on animal physiology in modulating nutrient conditions. We hypothesize immune cells as effectors of coordinating metabolic homeostasis under these conditions and they are necessary for organismal homeostasis.

Animal growth is a complex adaptive process that is dependent on extrinsic nutrient conditions, and is intricately linked with cues of both developmental (21, 22) and nutritional origins (7, 23, 24). These cues coordinate a central growth program that ensures animals achieve their size and proportion within their respective developmental time scale (21). This crosstalk is facilitated by long-range signaling molecules originating from the brain and fat body in Drosophila to coordinate the scaling of animal size in response to nutrient availability (7) uptake, and utilization (21, 25). Indeed, several recent studies have demonstrated a complex interplay of insulin signaling with innate immune pathways in growth and nutritional homeostasis (26–29). These studies have positioned the fat body as the major organ responsible for sensing and storing nutrients, in addition to its immune effector functions as a central member of the innate immune system.

Cells of the innate immune system also sense microbial load, integrate metabolic inputs, and alter nutrient allocation and organismal growth when performing pathogenic clearance functions (3, 30, 31). These functions are very similar to roles performed by the fat body, and we hypothesize immune cells as regulators of metabolism and organismal metabolic homeostasis not only in conditions of immune challenge but also in homeostasis or modulating nutrient environments. Development of metabolic disorders such as diabetes and obesity with altered immune cell activity and function are in agreement with this idea (32). In this study, we test this hypothesis and observe that blood cells are necessary to coordinate systemic metabolism and animal growth in homeostasis and in conditions of nutrient overload. Loss or gain of blood cells early in larval development affected adult growth. Our experiments suggest a role for blood cells in the control of fat body innate immune homeostasis and insulin sensitivity. These findings indicate that immune cell activity, as opposed to their number, orchestrates organismal growth homeostasis especially in conditions of dietary excess.

#### RESULTS

#### Drosophila Larval Blood Cells Function to Control Growth of Adult Flies

We aim to explore non-immune homeostatic functions of mature immune cells. To initiate this investigation, the impact of hemocyte ablation from early Drosophila larval stages on larval metabolism and development was assessed. Hemolectin1Gal4 (Hml<sup>1</sup> >) (**Figure 1a**) was used as the driver to express the pro-apoptotic gene, hid in blood cells. Expressing UAShid specifically in blood cells leads to killing of a majority of blood cells (**Figures 1b–f**) (33). UAS-hid control larvae showed no changes in blood cell numbers or overall larval growth (**Supplementary Figures 1a–c**), confirming that the dramatic loss of immune cells was specific to UAS-hid expression and not a consequence of leaky or non-autonomous UAS-hid transgene expression. We ensured the specificity of Hml1Gal4 driver line by conducting lineage analysis using G-TRACE (34). This

FIGURE 1 | loss in blood cells as evident from (d,e) DAPI, (d′ ,e′ ) Hml expression (Hml<sup>1</sup> <sup>&</sup>gt;UAS-GFP), and (d′′ ,e ′′ ) Phalloidin stainings. (d′ ′′,e ′ ′′) Merge of all the channels. (f) Graphical representation of total and Hml<sup>+</sup> blood cells in control, Hml<sup>1</sup> >UAS-GFP, and Hml<sup>1</sup> >UAS-GFP/UAS-hid larvae. Control (Hml<sup>1</sup> >UAS-GFP/+) total number of blood cells/mm<sup>2</sup> (n = 16, 409.8 ± 149.4) and Hml-GFP positive cells/mm<sup>2</sup> (n = 16, 328.8 ± 108.8). Hml<sup>1</sup> >UAS-GFP/UAS-hid show significantly less number of total blood cells/mm<sup>2</sup> (n = 16,148.3 ± 50.04, \*\*\*p-value = 0.0003, in comparison with control total blood cells) and less Hml-GFP-positive cells (n = 16, 13.11 ± 4.0, \*\*\*p-value < 0.0001 in comparison with control Hml<sup>+</sup> cells). (g–k) Blood cell ablation affects adult growth. (g–h′ ) Representative images showing adult size defect. In comparison with (g,g′ ) control adult (Hml<sup>1</sup> >/+) (g) female and (g′ ) male fly, (h, h′ ) Hml<sup>1</sup> >/UAS-hid (h) female and (h′ ) male adult flies are smaller in size. (i–j′ ) This growth defect is also shown seen in wing span areas. Representative adult wings from (i,i′ ) controls (Hml<sup>1</sup> >/+) and (j,j′ ) Hml<sup>1</sup> >/UAS-hid adults. (k) Quantification of wingspan areas. Hml<sup>1</sup> >/+ (female n = 100, 1.42 ± 0.03, male n = 1.00, 1.1 ± 0.08) and Hml<sup>1</sup> >/UAS-hid (female n = 100, 1.04 ± 0.03 (\*\*\*p-value < 0.0001), male n = 100, 0.74 ± 0.07 (\*\*\*p-value < 0.0001).

approach confirmed hid expression (both in real time and lineage based) in blood cells alone, without any expression detected in other larval tissues (**Supplementary Figures 1d–h**′′). Hml1Gal4 is seen in differentiating and mature populations of larval immune cells, which are essentially the plasmatocytes (**Supplementary Figures 1i–i**′′) (35, 36), making Hml1Gal4 a reliable driver line to specifically modulate immune cells and assess systemic changes.

Driving UAS-hid with Hml<sup>1</sup> >resulted in an overall reduction in blood cells. We report a dramatic loss of Hml positive (Hml+) hemocytes (**Figures 1b–f**) with a small increase in Hml negative (Hml−) blood cells (**Figures 1d–f**). These results are consistent with published reports (33). Loss of Drosophila larval hemocytes using this strategy dramatically affected adult fly sizes (**Figures 1g–h**′ ). This was not a consequence of any major defect in larval development in Hml<sup>1</sup> >/UAS-hid animals (**Figure 1c** compared with **Figure 1b**). However, any minor differences cannot be ruled out. To estimate the degree of adult growth retardation, wing span areas of both female and male flies were measured. We observed a significant reduction in wing span areas of Hml<sup>1</sup> >/UAS-hid adults (**Figures 1i–k**), indicating a role for Hml<sup>+</sup> blood cells in animal growth control.

Next, we assessed the temporal requirement of larval blood cells in growth control. To address this, we conducted blood cell ablations at mid L2/early L3 time point. Using the temperaturesensitive form of Gal80, the expression of Hml<sup>1</sup> > was regulated to drive UAS-hid transgene expression from mid L2 larval time point until wandering L3 after which these animals were dissected (**Supplementary Figure 2a**). While this temporal expression of UAS-hid was sufficient to successfully eliminate Hml<sup>+</sup> larval blood cells (**Supplementary Figures 2a–c**), the conditional loss of blood cells post L2 phase of larval development did not result in defective adult growth (**Supplementary Figures 2d–g**′ ), as assessed by adult fly sizes (**Supplementary Figures 2d–e**′ ) or wing span areas (**Supplementary Figures 2f–h**). Together, these results show that Hml<sup>+</sup> cells play a critical role in the systemic control of animal growth at the early phase of the larval development rather than later in larval life.

#### Blood Cells Regulate Insulin Signaling

Insulin signaling is a central regulator of animal growth (37). To understand the underlying regulation by blood cells in coordinating systemic growth, we undertook an in-depth analysis of insulin signaling in Hml<sup>1</sup> >/UAS-hid animals. We first assessed production and expression of different Dilp genes from the insulin producing cells (IPCs) of feeding L3 larvae. Dilp release from the IPCs is dependent on feeding state of the animal. As long as larvae are feeding, Dilps are produced and released from the IPCs. In non-feeding state or in conditions of nutritional deprivation such as starvation, Dilp release is inhibited leading to their accumulation in the IPCs (37). We conducted immunohistochemical and quantitative mRNA analysis of Dilp 2 and 5. Compared with control larvae, Hml<sup>1</sup> >/UAS-hid animals showed increased Dilp 2 and Dilp 5 peptide expression in the IPCs (**Figures 2a–d**, **Supplementary Figures 3a,b**). To test whether the increase was a consequence of heightened synthesis of Dilp 2 and 5, qRT-PCR analysis was done for Dilp 2 and 5 mRNA levels in feeding L3 larval brain tissues (**Figure 2e**). There was no increase in the levels of mRNA suggesting that the increase observed in Dilp 2 and Dilp 5 peptide expression was most likely a consequence of its accumulation or abrogated release. Blocking Dilp release or its accumulation in the brain IPCs is associated with hyperglycemia, which is a characteristic of reduced insulin signaling. Therefore, we tested readouts of glucose homeostasis, measuring circulating levels of glucose and trehalose as well as whole-animal glucose and glycogen and TAG levels. These biochemical assays were performed on feeding L3 larvae from control and Hml<sup>1</sup> >/UAS-hid backgrounds. As reported in conditions of reduced Dilp release, in Hml<sup>1</sup> >/UAS-hid animals, a significant increase of glucose and trehalose was observed in the circulating hemolymph (**Figures 2f,g**). Whole-animal glycogen was reduced (**Figure 2h**), whereas glucose levels were upregulated (**Figure 2i**). Further, increased levels of TAG were observed in whole larvae, circulating hemolymph, and fat body (**Figure 2l**). Neutral lipids detected via Nile red staining also confirmed increased TAG accumulation in Hml<sup>1</sup> >/UAShid fat bodies (**Figures 2j,k**). Lipid droplet size in these fat bodies was comparatively larger than seen in control conditions (**Figures 2j,k**).

Consistent with the biochemical analysis, assessment of downstream readouts of insulin signaling in the fat body also revealed a reduction in insulin signaling. Fat body glucose levels were reduced (**Supplementary Figure 3c**). Expression levels of tGPH, a membrane-associated GFP expression marker whose fluorescence is an indicator of insulin-dependent PI3K activity in living cells (38, 39), showed a reduction in Hml<sup>1</sup> >/UAS-hid larval fat bodies (**Figures 2m–o**). Insulin-mediated repression of FoxO nuclear localization and signaling (40) was also affected in Hml<sup>1</sup> >/UAS-hid larvae. FoxO protein (detected using antibodies against it), which was detected primarily in the cytoplasm of control fat body tissues, in Hml<sup>1</sup> >/UAS-hid larval fat bodies, was nuclear localized (**Figures 2p,q**). Subsequently, we also checked for the mRNA levels of FoxO targets 4EBP, InR,

insulin-producing cells (IPCs). (e) qPCR analysis for dilp2 and dilp5 in brain tissue of feeding L3 larvae does not show any change in their relative mRNA expression. (Continued) FIGURE 2 | Relative fold change is represented and statistical analysis was performed on Ctvalues (dilp2; Hml<sup>1</sup> >/+, n = 70, 4 ± 0.32; Hml<sup>1</sup> >/UAS-hid, n = 80, 3.77 ± 0.71 and dilp5; Hml<sup>1</sup> >/+, n = 70, 9.25 ± 0.18; Hml<sup>1</sup> >/UAS-hid, n = 80, 9.34 ± 0.29). (f) Hemolymph glucose levels. Hml<sup>1</sup> >/+, n = 78, 0.09 ± 0.01 and Hml<sup>1</sup> >/UAS-hid n = 78, 0.22 ± 0.06 (\*\*\*p-value < 0.0001). (g) Hemolymph trehalose levels. Hml<sup>1</sup> >/+, n = 90, 0.19 ± 0.10 and Hml<sup>1</sup> >/UAS-hid, n = 72, 0.28 ± 0.06 (\*p-value = 0.0335). (h) Whole-larvae glycogen levels. Hml<sup>1</sup> >/+, n = 6, 1.08 ± 0.12 and Hml<sup>1</sup> >UAS hid, n = 6, 0.77 ± 0.1 (\*\*\*p-value = 0.0006). (i) Whole-larvae glucose levels. Hml<sup>1</sup> >/+, n = 39, 0.11 ± 0.05 and Hml<sup>1</sup> >/UAS-hid, n = 39, 0.25 ± 0.06 (\*\*\*p-value < 0.0001). (j,k) Neutral lipid staining (Nile red) in fat bodies of (j) Hml<sup>1</sup> >/+ and (k) Hml<sup>1</sup> >UAS-hid. Compared with control, (j) Hml<sup>1</sup> >/+, (k) Hml<sup>1</sup> >UAS-hid fat bodies show more lipid droplets. (l) TAG levels measurement in whole larvae (Hml<sup>1</sup> >/+, n = 36, 1.7 ± 0.3 and Hml<sup>1</sup> >UAS hid, n = 36, 3.4 ± 0.7, \*\*\*p-value = 0.0005), hemolymph (Hml<sup>1</sup> >/+, n = 60, 0.34 ± 0.08 and Hml<sup>1</sup> >/UAS-hid, n = 66, 0.77 ± 0.16, \*\*\*p-value < 0.0001) and fat body (Hml<sup>1</sup> >/+, n = 45, 0.7 ± 0.07 and Hml<sup>1</sup> >/UAS-hid, n = 40, 2.25 ± 0.5, \*\*\*p-value < 0.0001). (m–o) tGPH expression in (m,n) fat bodies of feeding L3 larvae from (m) control (Hml<sup>1</sup> >/+) and (n) Hml<sup>1</sup> >/UAS-hid backgrounds shows reduced tGPH expression in Hml<sup>1</sup> >/UAS-hid condition. (o) Quantification of mean tGPH intensities in Hml<sup>1</sup> >/+ (n = 25, 2621 ± 486.2) and Hml<sup>1</sup> >/UAS-hid (n = 25, 1005 ± 196, \*\*\*p-value = 0.0001). (p,q) FoxO immunostaining in fat bodies of feeding L3larvae from (p) control (Hml<sup>1</sup> >/+) and (q) Hml<sup>1</sup> </UAS-hid backgrounds. As compared with (p) control, FoxO is nuclear localized in fat bodies of Hml<sup>1</sup> </UAS-hid animals. (r,s) Fat body analysis of FoxO target genes. (r) 4EBP and (s) InR mRNA expression. 4EBP is upregulated in Hml<sup>1</sup> >/UAS-hid condition. Relative fold change is represented and statistical analysis was performed on Ctvalues (4EBP: Hml<sup>1</sup> >/+, n = 60, 2.53 ± 0.91; Hml<sup>1</sup> >/UAS-hid, n = 60, 1.11 ± 0.45; \*p-value = 0.0233 and InR; Hml<sup>1</sup> >/+, n = 60, 10.15 ± 0.93; Hml<sup>1</sup> >/UAS-hid, n = 60, 9.99 ± 1.31). (t,u) Fat body analysis of inflammatory pathway target genes. (t) drs and (u) dpt mRNA expression is significantly upregulated in Hml<sup>1</sup> >/UAS-hid condition. Relative fold change is represented and statistical analysis was performed on Ctvalues (drs: Hml<sup>1</sup> >/+, n = 60, 5.40 ± 0.64; Hml<sup>1</sup> >/UAS-hid, n = 60, 2.29 ± 1.28; \*\*\*p-value = 0.0002 and dpt: Hml<sup>1</sup> >/+, n = 60, 16.79 ± 1.16; Hml<sup>1</sup> >/UAS-hid, n = 60, 5.78 ± 0.69; \*\*\*p-value < 0.0001).

and tobi in the fat body by isolating RNA followed by qRT-PCR from control and Hml<sup>1</sup> >/UAS-hid conditions. This detected a significant upregulation in the levels of 4EBP (**Figure 2r**) and mild upregulation of InR (**Figure 2s**) in Hml<sup>1</sup> >/UAS-hid conditions compared with controls. Together, these data show that the adult growth retardation was a consequence of reduced fat body insulin signaling in Hml<sup>1</sup> >/UAS-hid conditions. To determine if these changes were a result of the defect in fat body Akt signaling, we assessed levels of phosphorylated Ser505-Akt. Although immunohistochemical analysis did show a reduction in pAkt level, a quantitative analysis of its expression with respect to total Akt revealed only a minor difference in Hml<sup>1</sup> >/UAShid animals (**Supplementary Figures 3d–f**). This suggests either Akt independence (41) or phosphorylation at other sites on Akt, which are more sensitive indicators of its function (29) and remains to be addressed.

A strong connection between activation of fat body innate immune signaling leading to loss of insulin sensitivity (42, 43) and animal growth defect is well-established (29). Given the growth defect seen upon loss of immune cells, we asked if immune cell loss led to any changes in fat body innate immune signaling. For this, we investigated Toll and Imd innate immune signaling pathways in fat bodies in control and Hml<sup>1</sup> >/UAShid conditions. Expression analysis of drosomycin (drs), a Toll pathway target gene, and diptericin (dpt), an Imd pathway target gene, was undertaken by isolating RNA from the fat body of Hml<sup>1</sup> >/UAS-hid and control larvae. Compared with expression in controls, Hml<sup>1</sup> >/UAS-hid conditioned fat bodies showed a strong upregulation of both drs (**Figure 2t**) and dpt (**Figure 2u**). This indicated a robust activation of innate immune signaling pathways in the fat body on the loss of Hml<sup>+</sup> immune cells, implying a role for these cells in moderating activation of inflammatory pathways in the fat body.

### Hemocytes Control Tolerance to High-Sucrose Diet

It is known that larvae fed on a high-sucrose diet develop as smaller adults (21). High sugar stress also corresponds with a reduction in immune cell numbers (44). The sugar stress–induced growth defect is a consequence of reduction in fat body insulin signaling (39, 45). Hml<sup>1</sup> >/UAS-hid larvae with reduced Hml<sup>+</sup> immune cells demonstrated a similar growth defect, reduced insulin signaling, elevated fat body inflammation, and hyperglycemia, all of which are characteristic signs of insulin resistance. Hence, it was important to investigate how loss of Hml<sup>+</sup> immune cells influenced tolerance to additional high sugar dietary stress. To address this, larvae were raised on 25% sucrose, referred to as high-sucrose diet (HSD) in the text. Compared with the growth defect seen in control animals fed on HSD, Hml<sup>1</sup> >/UAS-hid animals on HSD showed further growth retardation (**Figures 3a–g**). The growth retardation was also detected when these animals were raised on diet with elevated fructose (**Supplementary Figures 4a–c**). This result showed that the growth defect was a consequence of high dietary sugar induced stress and was not limited to sucrose-enriched diet, suggesting a growth promoting or stress relieving function of Hml<sup>+</sup> innate immune cells in conditions of high dietary sugar.

To address if any changes in insulin signaling could explain the worsening of growth in Hml<sup>1</sup> >/UAS-hid HSD animals, we analyzed different components of insulin signaling, as described in the previous section. Estimation of expression of Dilp2 and 5 peptides (**Figures 3h–k**, **Supplementary Figures 4d,e**) and their mRNA levels in the larval brain IPCs (**Figure 3l**) showed no change between Hml<sup>1</sup> >/UAS-hid and controls. Biochemical analysis revealed a mild increase in larval hemolymph glucose in Hml<sup>1</sup> >/UAS-hid HSD animals (**Figure 3m**), but trehalose levels remained comparable with control HSD larvae (**Figure 3n**). Whole-animal glucose and glycogen failed to detect any changes in their levels (**Figures 3o,p**). Readouts of fat body insulin signaling did not reveal any difference either. Membrane tGPH expression (**Figures 3q–s**), fat body glucose levels (**Supplementary Figure 4f**), phosphorylated S505-Akt levels (**Supplementary Figures 4g,h,k**), nuclear localization of FoxO protein (**Supplementary Figures 4i,j**), and expression of FoxO target genes (**Figures 3t,u** and **Supplementary Figure 4l**) remained comparable between control HSD and Hml<sup>1</sup> >/UAS-hid HSD larvae. Thus, the growth defect seen in Hml<sup>1</sup> >/UAS-hid HSD animals was not a

units. (a–g) Blood cell ablation worsens HSD induced adult growth defect. Compared with controls (Hml<sup>1</sup> >/+) reared on (a,a′ retardation which is (c,c′ ) worsened in Hml<sup>1</sup> >/UAS-hid HSD animals. (d–g) Representative images showing adult size defect. In comparison with (d,d′ ) Hml<sup>1</sup> >/+ (Continued) FIGURE 3 | on RF, (e,e′ ) Hml<sup>1</sup> >/+ on HSD show reduction in wing sizes which is further reduced in (f,f′ ) Hml<sup>1</sup> >/UAS-hid HSD adults. (g) Quantification of wing span areas in: Hml<sup>1</sup> >/+ on RF (female n = 100, 1.4 ± 0.03, male n = 100, 1.1 ± 0.09) Hml<sup>1</sup> >/+ on HSD (female n = 100, 1.0 ± 0.04, \*\*\*p-value < 0.0001, male n = 100, 0.856 ± 0.1, \*\*\*p-value < 0.0001) and Hml<sup>1</sup> >/UAS-hid on HSD (female n = 100, 0.7 ± 0.04, \*\*\*p-value < 0.0001 and male n = 100, 0.6 ± 0.08, \*\*\*p-value < 0.0001). Two-way ANOVA analysis was performed (females \*\*\*p-value < 0.0001 and males \*\*\*p-value < 0.0001). (h–l) Dilp2 and Dilp5 analysis in feeding L3 larval brains raised on HSD. Immunostainings of (h,i) Dilp2 and (j,k) Dilp5 peptides in (h,j) control (Hml<sup>1</sup> >/+) and (i,k) Hml<sup>1</sup> >/UAS-hid larval brains do not show any change in their peptide expression in insulin-producing cells (IPCs). (l) qPCR analysis for dilp2 and dilp5 in brain tissue of feeding L3 larvae does not show any change in their relative mRNA expression. Relative fold change is represented and statistical analysis was performed on Ctvalues (dilp2; Hml<sup>1</sup> >/+, HSD, n = 100, 4.86 ± 1.27; Hml<sup>1</sup> >/UAS-hid, HSD, n = 100, 5.92 ± 1.46 and dilp5; Hml<sup>1</sup> >/+, HSD, n = 100, 9.15 ± 2.13; Hml<sup>1</sup> >/UAS-hid, HSD, n = 100, 10.66 ± 2.48). (m) Hemolymph glucose levels. Hml<sup>1</sup> >/+, HSD, n = 42, 0.27 ± 0.04 and Hml<sup>1</sup> >/UAS-hid, HSD, n = 42, 0.33 ± 0.04 (\*p-value = 0.0189). (n) Hemolymph trehalose levels. Hml<sup>1</sup> >/+, HSD, n = 36, 0.92 ± 0.22, and Hml<sup>1</sup> >/UAS-hid, HSD, n = 36, 0.28 ± 0.2. (o) Whole-larvae glucose levels. Hml<sup>1</sup> >/+, HSD n = 18, 0.24 ± 0.05 and Hml<sup>1</sup> >/UAS-hid n = 18, 0.32 ± 0.09. (p) Whole-larvae glycogen levels. Hml<sup>1</sup> >/+, HSD, n = 13, 0.6 ± 0.15 and Hml<sup>1</sup> >/UAS-hid HSD, n = 13, 0.64 ± 0.16. (q–s) tGPH expression in (q,r) fat bodies of feeding L3 larvae from (q) Hml<sup>1</sup> >/+, HSD controls and (r) Hml<sup>1</sup> >/UAS-hid, HSD backgrounds showed no change in tGPH expression. (s) Quantification of mean tGPH intensities in Hml<sup>1</sup> >/+ on HSD (n = 15, 192 ± 94) and Hml<sup>1</sup> >/UAS-hid, on HSD (n = 20, 178 ± 61). (t,u) Fat body analysis of FoxO target genes. (t) 4EBP and (u) InR mRNA expression show no difference. Relative fold change is represented and statistical analysis was performed on Ctvalues. (4EBP: Hml<sup>1</sup> >/+, HSD, n = 80, 0.29 ± 0.65; Hml<sup>1</sup> >/UAS-hid, HSD, n = 80, 0.61 ± 1.18 and InR: Hml<sup>1</sup> >/+, HSD, n = 80, 8.97 ± 1.34; Hml<sup>1</sup> >/UAS-hid, HSD, n = 80, 8.43 ± 1.91). (v,w) Neutral lipid (Nile red) staining in fat bodies of larvae raised on HSD. Compared with (v) Hml<sup>1</sup> >/+, on HSD (w) Hml<sup>1</sup> >/UAS-hid HSD animals show increased number of bigger lipid droplets. (x) TAG levels measurements in whole larvae (Hml<sup>1</sup> >/+, HSD, n = 24, 3.8 ± 1 and Hml<sup>1</sup> >UAS hid, HSD, n = 24, 4.4 ± 0.86), hemolymph (Hml<sup>1</sup> >/+, HSD, n = 60, 0.34 ± 0.08 and Hml<sup>1</sup> >/UAS-hid, HSD, n = 60, 0.7 ± 0.1, \*\*\*p-value < 0.0001), and fat body (Hml<sup>1</sup> >/+, HSD n = 30, 1.304 ± 0.5 and Hml<sup>1</sup> >/UAS-hid, HSD, n = 30, 2.5 ± 0.5, \*\*p-value = 0.0013). (y,z) qPCR analysis of (y) dpt and (z) drs mRNA expression in fat body tissue of feeding L3 larvae raised on HSD. Relative fold change is represented and statistical analysis was performed on Ctvalues (dpt: Hml<sup>1</sup> >/+ HSD n = 80, 17.96 ± 0.88; Hml<sup>1</sup> >/UAS-hid, HSD, n = 80, 5.96 ± 0.75; \*\*\*p-value < 0.0001 and drs: Hml<sup>1</sup> >/+, HSD, n = 80, 8.13 ± 0.48; Hml<sup>1</sup> >/UAS-hid, HSD, n = 80, 7.12 ± 1.11) \*indicates signifcant p values.

consequence of any dramatic change in glucose homeostasis or insulin signaling.

Immune cell ablation, however, impacted the TAG levels in these animals. Although overall larval TAG levels remained comparable, hemolymph and fat body TAG levels showed a significant increase in Hml<sup>1</sup> >/UAS-hid HSD larvae (**Figure 3x**). Nile red staining of fat bodies confirmed the increase in the TAG levels as well (**Figures 3v,w**). The lipid droplet sizes detected in Hml<sup>1</sup> >/UAS-hid HSD fat bodies were comparatively larger than seen in the control tissues (**Figures 3v,w**). Together, these results showed a defect in the fat body lipid metabolism in the Hml<sup>1</sup> >/UAS-hid HSD larvae.

We next assessed the status of fat body innate immune signaling in response to immune cell ablation in HSD condition. Interestingly, Hml<sup>1</sup> >/UAS-hid animals showed upregulation of Imd target gene, dpt (**Figure 3y**), whereas expression of Toll target gene, drs, remained comparable with control HSD fat bodies (**Figure 3z**). This suggested specific activation of the Imd pathway on loss of immune cells in HSD condition and was unlike regular dietary state, where loss of Hml<sup>+</sup> cells led to dramatic upregulation of both Toll and Imd signaling in the fat body. In conditions of dietary excess, such specific modulation of fat body inflammatory signaling (46) and influence on tolerance to metabolic toxicity (47) is reported. Our findings support this notion and implicate Hml<sup>+</sup> immune cells in moderating this specificity.

#### Increasing the Number of Activated Blood Cells in Drosophila Larvae Rescues the HSD Induced Growth Defect

We next assessed the outcome of increasing immune cell numbers on organismal metabolic state and growth homeostasis. For this, we over-expressed a constitutively active version of PDGF/VEGF-like receptor (Pvract) in blood cells using Hml<sup>1</sup> > as the driver (Hml<sup>1</sup> >/UAS PvrAct) (48). This genetic manipulation resulted in a dramatic increase in immune cell numbers and specifically of Hml<sup>+</sup> cells (**Figures 4a–c**, **Supplementary Figures 5a–b**′ ′′). This manipulation leads to the expansion of immune cells that are characteristically similar to the invasive macrophages (49). In regular food conditions, adult flies from this genetic background did not show any effect on growth phenotype and were comparable in size with control adult flies (**Supplementary Figures 5g–k**). This result suggested that although animal growth is sensitive to loss of immune cells (**Figures 1**, **2**), a mere increase in the immune cell numbers did not result in a concomitant increase in animal sizes. These data suggest that immune cells are not directly involved in scaling of animal sizes. Consistent with this notion, insulin signaling remained unaffected in Hml<sup>1</sup> >/UAS PvrAct animals (**Supplementary Figures 5c–f,l–v**).

Next, we explored the influence of increased immune cell numbers on dietary sugar–induced growth defect and tolerance. Compared with immune cell numbers of controls on HSD, Hml<sup>1</sup> >/UAS PvrAct raised on HSD had significantly higher immune cell numbers (**Figure 4d**), but not as dramatic as seen in regular condition (compared with **Figure 4c**). The proportion of Hml<sup>+</sup> immune cells was specifically increased (**Figure 4c**). Interestingly, the growth defect seen in HSD animals was dramatically restored in Hml<sup>1</sup> >/UAS PvrAct HSD genetic background (**Figures 4e–h**). Their sizes were comparable with sizes seen for controls raised on regular dietary state (**Figures 4g,g**′ compared with **Figures 4e,e**′ ). A similar trend of increased immune cell numbers was also evident with overexpression of wild-type Pvr in Hml<sup>+</sup> (Hml<sup>1</sup> >/PvrWT) blood cells (**Supplementary Figures 6a,b,d**). These animals were also significantly larger than HSD controls (**Supplementary Figures 6e–g**), but smaller than Hml<sup>1</sup> >/UAS PvrAct HSD animals (compare **Figure 4h** with **Supplementary Figure 6g**). The difference in growth restoration between Hml<sup>1</sup> >/UAS PvrAct and Hml<sup>1</sup> >/PvrWT may stem from the extent of increased numbers of activated

FIGURE 4 | HSD (female = 1.4 ± 0.05, n = 100, \*\*\*p-value < 0.0001 in comparison with Hml<sup>1</sup> >/+ reared HSD, male 1.01 ± 0.12, n = 100, \*\*p-value = 0.0064 in comparison with Hml<sup>1</sup> >/+ reared on HSD). Two-way ANOVA comparison was performed (females \*\*\*p-value < 0.0001 and males \*\*\*p-value < 0.0001). (l–k′ ) Representative wing images of adult flies showing growth restoration of Hml<sup>1</sup> >/UAS PvrAct adults on HSD. (i–j′ ) Hml<sup>1</sup> >/+on (i,i′ ) RF and (j,j′ ) on HSD and (k,k′ ) Hml<sup>1</sup> >/UAS PvrAct on HSD.

immune cells, which is much higher in Hml<sup>1</sup> >/UAS PvrAct as opposed to Hml<sup>1</sup> >/PvrWT (**Figure 4c** compared with **Supplementary Figure 6d**). Importantly, increasing immune cell numbers using other genetic manipulations did not lead to a growth restoration phenotype. Expression of a temperaturesensitive form of shibire, shits in Hml<sup>+</sup> cells (Hml<sup>1</sup> >/UAS shits), which also causes a comparable expansion of Hml<sup>+</sup> immune cells as seen in Hml<sup>1</sup> >/PvrWT (**Supplementary Figures 6a,c,d**), was insufficient to recover the growth defect of HSD. Contrastingly, Hml<sup>1</sup> >/UAS shits HSD animals were smaller and demonstrated growth retardation (**Supplementary Figures 6h–j**). These data suggested immune cell state as a key component in growth control. We conclude that immune cell activity is linked to systemic growth control as opposed to only their numbers being a regulator for growth.

We further tested other dietary sugar–induced stress, such as fructose and glucose, and observed that different sugars had varying effects on growth. Compared with the growth of control animals on sucrose-rich diet, the growth reduction was severe in high-glucose diet, whereas fructose-rich diet showed a mild growth defect (**Supplementary Figures 7a,b**); Hml<sup>1</sup> >/UAS PvrAct animals were able to restore growth in all conditions. This was not seen in Hml<sup>1</sup> >/UAS shits animals. Hml<sup>1</sup> >/UAS PvrAct condition restored growth in every dietary condition, but at differential capacities. This trend was not evident in Hml<sup>1</sup> >/UAS shits animals (**Supplementary Figures 7c–r**). This result further strengthened the importance of immune cell states in moderating dietary stress–induced growth defect.

We assessed the effect of PvrAct expression in hemocytes on peripheral insulin and inflammatory signaling in HSD condition and observed restoration of certain features of insulin resistance. The accumulation of Dilp2 and Dilp5 peptides normally seen in larval IPCs in control HSD animals was not observed in Hml<sup>1</sup> >/UAS PvrAct HSD larval brain IPCs (**Figures 5a–d**, **Supplementary Figures 8a,b**). This was a not a consequence of reduction in Dilp2 and Dilp5 mRNA levels (**Figure 5e**). Biochemical analysis of circulating larval hemolymph glucose and trehalose revealed a reduction in glucose levels, whereas trehalose remained unchanged (**Figures 5f,g**). Whole-animal glycogen was comparatively higher in Hml<sup>1</sup> >/UAS PvrAct HSD animals as compared with control groups (**Figure 5h**). However, whole-animal glucose levels remained unchanged (**Figure 5i**). These data suggested an improvement in circulating glucose and whole-animal glycogen levels in Hml<sup>1</sup> >/UAS PvrAct HSD larvae. Lipid measurements, however, showed no change and remained comparable with control HSD conditions (**Figures 5j–l**).

Readouts of peripheral fat body insulin signaling revealed restoration of some of its features. Of these, membrane tGPH levels (**Figures 5m–o**) and FoxO localization (**Figures 5p,q**) were restored in Hml<sup>1</sup> >/UAS PvrAct HSD larvae. Changes in the expression of FoxO target genes, 4EBP, InR, and tobi, were not detected (**Figures 5r,s**, **Supplementary Figure 8g**). pAKT levels also did not increase (**Supplementary Figures 8d–f**). Examining Toll and Imd pathway targets revealed an unexpected upregulation of drs (**Figure 5t**) in Hml<sup>1</sup> >/UAS PvrAct HSD animals, whereas dpt, the Imd pathway target gene, remained unchanged (**Figure 5u**). This was unlike the Hml<sup>1</sup> >/UAS hid HSD animals where a dramatic upregulation of dpt was evident without any change in drs expression.

### DISCUSSION

Much of our understanding of systemic control of animal growth is generally limited to endocrine organs. In Drosophila, the fat body (24), gut (15), and brain are highlighted as predominant nutrient sensors regulating growth (21). Because of nutrient sensing and signaling functions performed by these organs, they have been the primary focus of investigations on growth regulation. The formation of blood cells is also an energyconsuming process and has a metabolic cost on the animal. Immune cells are highly sensitive to nutrient modulation (18, 50). However, the physiological relevance of their increased sensitivity to dietary changes and impact on growth, if any, is poorly understood. Our study demonstrates the influence of immune cells on coordinating systemic metabolism and animal growth. In this regard, two important points emerge from this work: (1) immune cell states rather than their number is an important parameter in growth regulation and (2) immune cells systemically coordinate growth via the regulation of fat body inflammation and insulin signaling. These functions of blood cells support the attainment of proper adult size both in homeostasis and in conditions of sugar excess. While we focus on immune cell–fat body cross-talk, our results do not rule out any direct communication between immune cells and the brain or other organs in coordinating growth. Overall, this study positions innate immune cells as a novel player in organismal metabolic homeostasis and growth control regulation.

### Immune Cell/Fat Body Cross-Talk in Animal Growth Control

We find that in homeostatic conditions, modulating immune cell numbers in larval life impacts overall organismal metabolic state and animal growth. The loss of immune cells dramatically reduces adult growth. The metabolic and biochemical assays in these animals resembles features of systemic insulin insensitivity (51). This includes increased circulating TAGs, circulating glucose/trehalose levels with reduced whole-animal glycogen levels. Specifically, fat body metabolism, innate immune signaling, and insulin sensitivity are also affected. Past and recent findings have highlighted immune cell–mediated regulation of

(Continued)


fat body immune activation and metabolic homeostasis by secreted factors like psidin (52), Upd3 (53), and adenosine (30). These interactions between immune cells and the fat body allow nutrient allocation to regulate animal growth in nutrient overload conditions (53) or in response to infection for an effective immune response (30, 52). Our study suggests a similar cross-talk between immune cells and the fat body in the maintenance of metabolic homeostasis. We posit that immune cell activity modulates fat body insulin signaling, which is central to organismal growth control. Loss of Hml<sup>+</sup> immune cells corresponds with robust activation of fat body Toll and Imd signaling, indicating increased inflammation. This positions Hml<sup>+</sup> cells as key regulators of fat body inflammatory homeostasis. Persistent activation of Toll (29) and Imd signaling pathways in the fat body (26, 28) leads to insulin insensitivity, metabolic dysregulation, and growth defect (54). Recently published data by Shin et al. (53) have highlighted a similar role for Hml<sup>+</sup> cells in systemic control of animal growth by regulating fat body Jak/Stat signaling. Our data strengthen this notion and proposes a model where antiinflammatory inputs by immune cells function on controlling fat body innate immune homeostasis and insulin sensitivity, thereby contributing to animal growth control (43). In addition, any change of Dilp 2 and Dilp5 production or secretion from the larval brain IPCs on immune cell manipulation cannot be ruled out. This can also lead to a reduction in circulating Dilps and reduced systemic insulin signaling. Whether this is mediated by the fat body (24) or by blood cells directly requires further validation.

However, in conditions of nutrient excess (HSD), the modulation of immune cell numbers on systemic metabolism, fat body inflammation, and its insulin sensitivity does not correlate with the findings in regular dietary conditions. These data are unexpected and indicate that cross-talk between blood cells and the fat body is more complex. Ablation of Hml<sup>+</sup> blood cells further worsened the adult sizes without differences in insulin sensitivity. Fat body insulin signaling and overall glucose homeostasis were not different from what is observed in control animals raised on HSD. Interestingly, this resulted in preferential activation of the Imd pathway in the fat body, without any change in Toll signaling. On the other hand, increasing the numbers of activated immune cells (Hml<sup>1</sup> >PvrAct) relieved the symptoms of metabolic stress induced by HSD. Here, signatures of insulin resistance or fat body insulin sensitivity revealed only partial recovery, but the extent of growth restoration observed in these animals was comparable with wild types raised on a regular diet. In these growth-restored animals, a specific activation of fat body Toll signaling was evident. This is possible as activated immune cells are capable of secreting spaztle, the Toll ligand, and drive systemic Toll activation (55, 56). A similar influence of innate immune signaling on growth in nutrient overload is supported by the published literature. Peptidoglycan recognition proteins (PGRPs) are known to activate either Toll or Imd pathways, and their modulation in the fat body differentially influences animal growth. PGRP-SB2 activates the Imd pathway, and its loss in the fat body increases growth and survival on HSD. Although PGRP-SC2 negatively regulates the Imd pathway and positively regulates the Toll pathway, its loss in the fat body reduces animal size on HSD (27).

Together, these data reveal that growth is differentially controlled in dietary excess as opposed to homeostasis and is supported by the independence of animal growth from insulin signaling seen in HSD Hml<sup>1</sup> >hid and Hml<sup>1</sup> >PvrAct conditions. Our data also suggest an additional growthpromoting axis independent of insulin signaling regulated by blood cell activity. Here, we propose a model where immune cells function to balance fat body innate immune activation. This supports fat body nutrient reallocation toward the promotion of animal growth and prevents metabolic toxicity. The additional immune cell/fat body inflammatory axis compensates for reduced fat body insulin signaling in HSD conditions.

#### Immune Cell State as the Driver of Growth Control

Cross-talk between the fat body and immune cells could either be dependent on immune cell numbers (3) or activity. Our findings reveal the importance of immune cell state to sustain metabolism and growth capacity. The expression of a constitutively active form of Pvr in immune cells resulted in the activation of invasive plasmatocytes (6, 49) and improved

tolerance to dietary excess. Increasing immune cell numbers by overexpression of shits did not ameliorate the HSD growth defect. On the contrary, Hml<sup>1</sup> >/UAS shits animals were smaller. Loss of shibire function blocks exocytosis (57, 58); therefore, shibire lacking blood cells are functionally inert, unlike PvrAct-expressing hemocytes, which are more active. Immune cell states are reflective of their internal metabolic activity (25) and signaling capacities (52, 53, 56, 59). The elevated metabolic state of active immune cells may therefore provide an animal with additional means to metabolize nutrients (especially when in excess) and does not require increased immune cell number with inert metabolic states. The immune cell states may be a reflection of changes in immune heterogeneity (25, 53) or internal metabolic states as seen in development (25) or in conditions of stress (20).

#### A Model of Temporal Control of Immune Cell Function in Regulation of Animal Growth

Immune cell function in growth homeostasis is temporally controlled and is required early, before the 3rd instar. This is inferred from the pronounced deficit in adult growth after loss of immune cells in early stages of larval development and not if immune cells are depleted post 3rd instar. Although larval growth is unchanged after immune cell depletion, subtler changes in larval sizes or weight cannot be ruled out. The reduced adult size may stem from a reduction in cell growth and proliferation (60, 61). This is supported by analysis of cell densities in wing imaginal discs of wandering 3rd instar Hml<sup>1</sup> >hid larvae, which showed a reduction in wing disc cell densities with increased spacing between cells (**Supplementary Figures 2i–k**). Suppression of insulin receptor signaling late in development affects body and organ size as opposed to its early role in developmental timing. This transition from the control of developmental timing to growth occurs early in 3rd instar, when larvae reach critical size, post which development can be completed in the absence of food. Consequently, critical size is a key stage in insect development that is established in early 3rd instar larvae and sets the lower limit of final adult size. The mechanisms that measure the critical size and the organs involved in this process are largely unclear. Fat body, imaginal discs are the only predicted critical-size sensing organs thus far. Based on the early immune function apparent from this study, we hypothesize an involvement of immune cells in the assessment of larval critical weight, supported by the evidence of cross-talk in this work. The implication of immune cell/fat body interaction early in larval life could be relevant in the establishment of developmental switches or programs that time the acquisition of critical weight and the temporal shift in insulin signaling to allow growth in non-feeding larval and pupal stages (60). The comparable metabolic and growth phenotypes seen in animals with reduced Inr activity late in larval life with animals depleted for immune cells are in agreement with this hypothesis.

## CONCLUSION

The fat body functions to integrate the physiological state of the animal and determine allocation of resources in a contextdependent manner (23). Much like this tissue, immune cells, in addition to their role in sensing infection, are also effective sensors of changes in nutrient levels (3, 20, 62). However, unlike the fat body which is fixed in location, immune cells are mobile and highly dynamic. Their behavior and localization change rapidly in contexts of stress or nutrient modulation (3, 30). Our findings clearly implicate immune cells as central players coordinating global metabolic homeostasis and growth control along with the fat body. This integrates both the cellular and humoral component of the innate immune system, which may have evolved for efficient allocation of resources during infections but is also coopted in development to orchestrate systemic metabolism and growth. Metabolic disorders like diabetes, obesity, and fatty liver in mammals (62) are associated with heightened inflammation and altered immune responses. This exemplifies a connection between immune cells, altered metabolic homeostasis, and disease progression; however, its relevance in development regulation like growth remains unaddressed. Future investigations will be necessary to probe the temporal nature of this cross-talk to reveal mechanistic insights underlying developmental paradigms operating in animal growth and physiology.

## METHODS

#### Drosophila Husbandry, Stocks, Genetics, and Food

The following Drosophila stocks were used in this study: w <sup>1118</sup> (wild-type), Hml1>UAS-Gfp (S.Sinenko), domeMESO-GFP; hml-dsRed (Utpal Banerjee), Hml1>Gfp;tub gal 80ts , Hml1>Gfp;tGPH, UAS-hid/CyoGFP, UAS-PvrAct and UAS-Shits (63), UAS Pvr (BL58998), and Hml1gal4 (BL30141). All fly stocks were reared on the standard BDSC corn meal agar food medium with yeast supplementation (referred to as the regular food in the article) at 25◦C incubator unless specified. The specific composition of the regular food (RF) for 1 L is corn flour, 80 g; D-glucose, 20 g; sugar, 40 g; agar, 8 g; yeast powder, 15 g; propionic acid, 4 mL; Tego (methyl parahydroxy benzoate), 1 g (5 mL ethanol); and orthophosphoric acid, 0.6 mL. For high-sugar diet (HSD), the regular food composition was modified by supplementing the food with 25% sucrose (20 g in 100 mL of standard medium) whereas the composition of the other ingredients remained unchanged. For high-fructose diet (HFD), the regular food composition was modified by supplementing the food with 25% D-fructose (25 g in 100 mL of standard medium) while the composition of the other ingredients remained unchanged. For high-glucose diet (HGD), the regular food composition was modified by supplementing the food with 25% D-glucose (23 g in 100 mL of standard medium) whereas the composition of the other ingredients remained unchanged. All genetic crosses were set up at 25 ◦C and then transferred to 29◦C where they were grown until analysis either as larvae or as adults.

#### Embryo Collection

Embryo collections were done for 4–6 h at 25◦C. This was followed strictly for all experimental crosses. For HSD experiments, the embryos were collected on RF at 25◦C after which 40–50 embryos were carefully transferred to HSD and reared at 29◦C until analysis. Temperature-sensitive experiments with tub gal80ts were carefully monitored to maintain timings for shifting them from a non-permissive (18◦C) to permissive (29 ◦C) temperature. Specifically, for Hml<sup>1</sup> >UAS-hid;tub gal80ts , the embryo collection was conducted at 25◦C and shifted to 18◦C where they were grown until mid-2nd instar larvae followed by shift to 29◦C incubator. For Hml<sup>1</sup> >UAS shits experiments after embryo collection, the animals were grown at 29◦C until analysis.

#### Quantification of Adult Growth Phenotype

All adult flies were scored for their body sizes 2 days after eclosion. The animals were scored first by comparing their body sizes against the control Hml<sup>1</sup> >/w<sup>1118</sup> reared either on regular food or on HSD. The wingspan area of the adult animals was also scored to quantify for growth (64). The wings of flies of interest were plucked, mounted on a slide, and imaged using a bright-field microscope. Female and male wings were mounted separately and analyzed separately. The wingspan areas were calculated using Fiji software. Briefly, the circumference of the wing is marked and area is measured. Only one wing per animal has been analyzed. In every experiment, we analyzed 12–15 animals. This was done with a minimum of 10 repeats.

The wingspan area quantification for genotypes UAS hid and UAS PvrAct on RF were carried out together in multiple batches. Hence, the controls are the same in **Figure 1h** and **Supplementary Figure 5k**.

#### Immunostaining, Immunohistochemistry, and Fluorescence Quantification

For rabbit pAKT (1:400) and rabbit FoxO (1:500) staining, fat bodies from feeding 3rd instar larvae were dissected in 1× PBS, fixed in PBS containing 4% formaldehyde for 20 min at room temperature, and washed in PBS containing 0.1% Triton X-100 (PBT). Tissues were then blocked for 2 h in 0.3% PBT containing 5% NGS. Primary antibodies were incubated overnight at 4◦C and secondary antibodies for 2 h at room temperature.

For Dilp 2 (1:400) and Dilp 5 (1:800) stainings, brains were dissected from feeding L3 larvae in 1× PBS. They were fixed in 1× PBS containing 4% formaldehyde for 20 min at room temperature, and extensively washed in 1× PBS containing 0.3% Triton X-100 (PBT). Tissues were then blocked for 2 h in 0.3% PBT containing 5% NGS. Primary antibodies were incubated overnight at 4◦C and secondary antibodies for 2 h at room temperature. To quantify Dilp2 and Dilp5 levels, confocal Z series of the IPCs were obtained using a 2-µm step size and identical laser power and scan settings. Fiji software was used to generate sum-intensity 3D projections of the Z stacks (16-bit scanned images) and to measure total fluorescent intensity across the IPCs.

For staining circulating blood cells, 3rd instar larvae were bled on Teflon-coated slides (Immuno-Cell no. 2015 C 30) followed by staining protocol as previously described.

The following secondary antibodies have been used in the study at 1:500 dilution: FITC and Cy3 (Jackson Immuno Research Laboratories). Phalloidin (Sigma-Aldrich no. 94072) was used at 1:100 dilutions to stain cell morphologies and nuclei were visualized using DAPI. Samples were mounted with Vectashield (Vector Laboratories) or 70% glycerol.

#### Imaging

Immunostained images, blood cell images, and wing Disc images were acquired using Olympus FV3000 confocal microscopy system under a × 20 air or ×40 oil-immersion objective or ×60 oil-immersion objective. Bright-field and larval fluorescence images were obtained on Leica fluorescence stereomicroscope.

#### Hemocyte Isolation and Quantification

Total blood cells including circulating hemocytes and sessile pool resistant hemocytes were isolated. For this, larvae where mechanically brushed to release the sessile pool resident hemocytes into circulation, and after dissection, the cells still adhering to cuticle were scraped with the forceps as per published protocol (65).

Circulating cell numbers obtained were quantified per larvae. For each genotype, a minimum of 10 larvae were analyzed. Five images per well covering the field of view were obtained under constant magnification. The hemocytes in these views were counted manually to score for DAPI-positive (representing total blood cells), Hml-positive, and Hml-negative cells. The counts are represented as blood cell numbers per square millimeter (53).

#### Quantification of tGPH

tGPH intensity is quantified using ImageJ software as previously described in (39). Briefly, fat bodies of feeding L3 larvae were imaged using confocal microscopy and average fluorescence was measured in 25 random squared areas (10 × 30 pixels), each covering part of the plasma membrane in different cells.

### Nile Red Staining

For lipid droplet staining, wandering L3 larvae were dissected in 1× PBS and fixed in 4% formaldehyde in 1× PBS for 20 min at room temperature. Tissues were then rinsed twice with 1× PBS, incubated for 30 min in a 1:1000 dilution with 70% glycerol of 0.02% Nile red (Sigma—Cat. no. N3013). The tissues were mounted in 70% glycerol with DAPI.

#### Metabolite Measurements

Glucose, trehalose, and glycogen measurements were done in feeding 3rd instar larvae. Triglyceride measurements were done in wandering 3rd instar larvae.

Glucose and TAG assays were conducted in extracts made from whole larva (1 sample = 3 larvae) and fat body (1 sample = fat body from 5 larvae). These tissues were homogenized in 100 µL of 1× PBS using GENETIX Bead Beater to obtain the extract for glucose or TAG analysis. For hemolymph extracts, bleeds from six larvae was collected in 100 µL of 1× PBS and centrifuged at 1000 rpm to remove blood cells. The extracted samples were heat inactivated at 70◦C for 10 min and then centrifuged at 1000 rpm. The supernatant was collected and subjected to glucose analysis using Sigma GOD-POD kit (GAGO20) or TAG analysis using the Sigma Triglycerides assay reagents (T2449 and F6428) (66).

For hemolymph trehalose assays, we adapted a previously published protocol (66). Briefly, larval hemolymph from six animals was collected in 25 µL of ice-cold 1× PBS and centrifuged at 5,000 rpm for 5 min to remove blood cells. 10 µL of sample was incubated in 25 µL of 0.25 M sodium carbonate at 95◦C for 2 h in a thermal cycler, then cooled to room temperature followed by additions of 8 µL of 1 M acetic acid and 66 µL of 0.25 M sodium acetate (pH 5.2) as the digestion buffer. 1 µL of porcine trehalase (Sigma T8778) was added to 40 µL of this mixture and incubated at 37◦C overnight. The resulting glucose was analyzed and normalized to protein levels.

The trehalose measurements for genotypes UAS hid and UAS PvrAct on RF and HSD were carried out together in multiple batches. Hence, the controls are same in **Figures 5p**, **2g** (RF) and **Figures 3n**, **5g** (HSD).

Whole-larvae glycogen assay was conducted as per the instructions provided along with Glycogen Assay Kit (MAK016) (67). For this one, whole larva was homogenized in 100 µL of 1× PBS. This extract was used for the assay and the amounts were normalized to total protein levels in the same sample.

The glycogen measurements for genotypes UAS hid and UAS PvrAct on RF and HSD were carried out together in multiple batches. Hence, **Figures 5o**, **2h** (RF) and **Figures 3p**, **5h** (HSD) share the same controls.

Protein estimation was undertaken using the Thermo Scientific BCA protein assay kit (Cat. no. 23225) and Varioskan LUX Multimode Microplate Reader using skanit software was used to quantify all metabolites.

#### Cell Density Quantification

Wing disc cell density was quantified by counting the total number of nuclei stained with DAPI in five random regions (430 µm<sup>2</sup> area each) of the wing imaginal disc. This was done for five different wing discs (68).

#### RNA Extraction and RT-PCR Analysis

RNA extractions from larval fat bodies(20 larvae) and brain (35 larvae) tissues were performed as previously published (69). Briefly, RNA from these samples was extracted using Trizol reagent (Ambion by Life Technologies Cat no. 15596). For RT-PCR, RNA samples were treated with DNase I (Thermo Scientific) and converted to cDNA with SuperScript II (Invitrogen). qPCR was performed using the SYBR Green PCR Master Mix (Applied Biosystems), in the C-1000 Touch Thermal Cycler (BIO-RAD CFX384 Real-Time System) in 384-well plates (Applied Biosystems). At least three biological replicates were used for statistical analysis. The following primers were used to perform qPCR:



The mRNA quantifications for genotypes UAS hid and UAS PvrAct on HSD were carried out together in multiple batches. Hence, **Figures 3t,u,y,z**, **5r,s,t,u**, **Supplementary Figures 4l, 8g** share the same controls.

#### Protein Extraction and Western Blot Analysis

Tissues (fat bodies from 20 larvae and brains from 35 larvae) were homogenized with the help of stainless steel beads (Qiagen; 69989) in an EZ-Lyser bead beater (Genetix). Protein extraction was carried out as previously published (69). Protein estimations were done using Bradford Reagent (Sigma B6916). 10% SDS-PAGE and Western blots were performed using standard methods (69). The following antibodies were used: primary antibodies—anti-pAkt (1:1000; rabbit; CST 4054; Ser505), anti-Akt (1:1000; rabbit; CST 9272), and anti-β-tubulin (1:3000; rabbit; Abcam ab6046) and secondary antibodies—anti-rabbit IgG, HRP-linked (CST 7074). Chemiluminescent reagent Western Bright Quantum (Advansta r-03026-c50) was used for detection by iBright FL1000 (Invitrogen). For measurements of pAKT/AKT ratios, the band mean intensities of pAkt and Akt were quantified with the help of Fiji (ImageJ) and corrected for background levels, followed by calculating their ratios from which the fold change was obtained.

The immunoblot quantifications for genotypes UAS-hid and UAS PvrAct on HSD were carried out together in multiple batches. Hence, **Supplementary Figures 4k**, **8f** share the same control.

#### STATISTICAL ANALYSIS

All statistical analysis was performed using Graph Pad Prism 5 and Microsoft Excel 2010. The means were analyzed using twotailed, unpaired Student t test. Two-way ANOVA was performed between wing span areas of Hml<sup>1</sup> >/+ RF, Hml<sup>1</sup> >/+ HSD versus Hml<sup>1</sup> >/UAS-hid RF, Hml<sup>1</sup> >/UAS-hid HSD (**Figure 3g**) and Hml<sup>1</sup> >/+ RF, Hml<sup>1</sup> >/+ HSD versus Hml<sup>1</sup> >/UAS PvrAct RF, Hml<sup>1</sup> >/UAS PvrAct HSD (**Figure 4h**).

#### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the article/**Supplementary Material**.

### AUTHOR CONTRIBUTIONS

PP and TM conceptualized the project. PP, AT, and SM performed experiments, conducted the formal analysis, and contributed to the writing of the manuscript. TM contributed towards procuring funding to drive this research, overseen the project for its scientific, critical evaluation of the data and manuscript writing and editing. All authors contributed to the article and approved the submitted version.

#### FUNDING

This study was supported by the DBT-Center of Excellence grant BT/PR13446/COE/34/30/2015, DST-ECR ECR/2015/000390DBT-IYBA 2017, DBT Ramalingaswami Re-entry Fellowship and CEFIPRA awarded to TM.

#### ACKNOWLEDGMENTS

We thank Pierre Leopold for Dilp2, Dilp5 and FoxO antibodies, Raghu Padinjat lab for primers, Banerjee lab for fly stocks, Angela Giangrande, Shruthi Balachandran for useful comments on the manuscript and necessary discussion inputs and Apurva Sarin and Dasaradhi Palakodeti for critical evaluation of the manuscript. We acknowledge our imaging and fly facility. PP is a Project Assistant and AT and SM are graduate students in the Mukherjee lab.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2020.01528/full#supplementary-material

Supplementary Figure 1 | In (d–h′′ ), scale bar = 20µm. Bar graph in (c) shows mean ± standard deviation (SD). Statistical analysis applied in panel c is unpaired t-test, two-tailed in (c). "n" is the total number of larvae analyzed. (a–c) UAS-hid genetic background does not show any blood cell defect. 3rd instar larval images depicting Hml<sup>+</sup> blood cells (red, marked with RFP) in (a) control (domeMESO-GFP; hml-dsRed/+) and (b) domeMESO-GFP; hml-dsRed/UAS-hid are comparable. (c) Blood cell counts/mm<sup>2</sup> in these backgrounds (domeMESO-GFP; hml-dsRed/w1118, n = 5, 166.5 ± 111.7 and domeMESO-GFP; hml-dsRed/UAS-hid, n = 5, 165.8 ± 92.7). (d–i′′ ) Lineage analysis of Hml1gal4 using GTRACE confirms blood-specific expression. RFP (real-time Hml1gal4 expression) or GFP (lineage-based Hml1gal4 expression) was not detected in (d–d′′ ) brain, (e–e′′ ) fat body, (f–f′′ ) proventriculus, (g–g′′ ) leg disc, (h–h′′ ) wing disc. (i–i′′ ) Both RFP and GFP expression is detected in blood cells.

Supplementary Figure 2 | In (b,c,d–e′ ), scale bar = 1 mm, (f–g′ ) = 250µm, and (i,j) = 20µm. In (h, k), bar graphs show mean ± standard deviation (SD). Statistical analysis in (h, k) is unpaired t-test, two-tailed. "n" is the total number of larvae analyzed. (a) Shows the pictorial representation of the constitutive and temporal experiments that have been done using Hml<sup>1</sup> > and Hml<sup>1</sup> >; tub gal80ts, respectively. The red mark indicates the temporal expression of UAS hid in blood cells by shifting the larvae from permissive (18◦C) to non-permissive

temperature (29◦C). (b,c) Temporal expression of UAS-hid in Hml<sup>+</sup> blood cells (marked by GFP, green) causes a loss of blood cells. Compared with (b) control (Hml<sup>1</sup> >UAS-GFP; tub gal80ts/+), (c) Hml<sup>1</sup> >UAS-GFP; tub gal80ts/ UAS hid larvae show no GFP signal. (d–h) Temporal expression of UAS-hid in Hml<sup>+</sup> blood cells did not alter adult growth. Representative adult (d–e′ ) fly and (f–g′ ) wing images in (d, d′ , f, f′ ) control (Hml<sup>1</sup> >UAS-GFP; tub gal80ts/+) and (e,e′ ,g,g′ ) Hml<sup>1</sup> >UAS-GFP; tub gal80ts/ UAS hid show no growth reduction. (h) Wingspan area quantifications in Hml<sup>1</sup> >; tub gal80ts/+ (female n = 100, 1.4 ± 0.1, male n = 100, 1.05 ± 0.07) and in Hml<sup>1</sup> >; tub gal80ts/UAS-hid (female n = 100, 1.3 ± 0.1, male n = 100, 1.05 ± 0.09). (i–k) Reduced cell density in wing discs of Hml<sup>1</sup> >/UAS-hid L3 larvae. Compared with cell density seen in wing discs of (i) control (Hml<sup>1</sup> >/+), (j) Hml<sup>1</sup> >/UAS-hid cell density is reduced. (k) Quantified by counting DAPI-positive cells (white). Hml<sup>1</sup> >/+, n = 5, 11.24 ± 3, and Hml<sup>1</sup> >/UAS-hid, n = 5, 4.80 ± 0.60 and (∗∗p-value = 0.0017).

Supplementary Figure 3 | In (d,e), scale bar = 20µm. In (a–c, g), bar graphs show mean ± standard deviation (SD) and statistical analysis applied in these panels is unpaired t-test, two-tailed. "n" is the total number of larvae analyzed, FB is fat body, a.u is arbitrary units, f.c is fold change and RF indicates regular food. (a) Quantification of mean intensity of Dilp2 levels of images shown in Figures 2a,b. Control, Hml<sup>1</sup> >/+ (n = 16, 204.5 ± 149.8) and Hml<sup>1</sup> >/UAS-hid (n = 14, 1768.8 ± 794.2, ∗∗∗p-value < 0.0001). (b) Quantification of mean intensity of Dilp5 levels of images shown in Figures 2c,d. Control, Hml<sup>1</sup> >/+ (n = 16, 234.023 ± 321.72) and Hml<sup>1</sup> >/UAS-hid (n = 13, 1395.4 ± 1303.9, ∗∗p-value = 0.0019). (c) Fat body glucose levels. Hml<sup>1</sup> >/+ (n = 65, 0.095 ± 0.017) and Hml<sup>1</sup> >/UAS-hid (n = 65, 0.04 ± 0.02, ∗∗∗p-value < 0.0001). (d–f) Fat body pAKT analysis. (d,e) Immunostaining of feeding L3 larval fat bodies with anti-pAKT antibody in (d) control, Hml<sup>1</sup> >/+ and (e) Hml<sup>1</sup> >/UAS-hid backgrounds show reduced pAKT levels in Hml<sup>1</sup> >/UAS-hid condition. (f) Immunoblot analysis of pAkt/Akt ratio in fat bodies of feeding L3 larvae of control (Hml<sup>1</sup> >/+) and Hml<sup>1</sup> >/UAS-hid reveals a small difference (fold change ± SD mentioned in the blots). β-Tubulin was used as the internal loading control. (g) Relative fat body mRNA levels of tobi. Fold change is plotted, and statistical analysis was done using Ctvalues (Hml<sup>1</sup> >/+, n = 60,8.67 ± 3.57 and Hml<sup>1</sup> >/UAS-hid, n = 60, 10.68 ± 0.75).

Supplementary Figure 4 | In (a–b′ ), scale bar = 250µm and (g–j) is 20µm. In (c–f,l), bar graphs show mean ± standard deviation (SD) and statistical analysis applied for these panels is unpaired t-test, two tailed. "n" is the total number of larvae analyzed, FB is fat body, a.u is arbitrary units, f.c is fold change, HFD is high-fructose diet and HSD is high-sugar diet. (a–c) Loss of immune cells affects tolerance to high fructose diet. (a–b′ ) Representative wing images of (a,a′ ) control (Hml<sup>1</sup> >/+) on HFD and (b,b′ ), Hml<sup>1</sup> >/UAS-hid on HFD showing reduction in wing sizes of Hml<sup>1</sup> >/UAS-hid, HFD animals. (c) Quantification of wingspan areas of Hml<sup>1</sup> >/+ on HFD (female, n = 50, 1.4 ± 0.03, male n = 120, 1.1 ± 0.08) and Hml<sup>1</sup> >/UAS-hid on HFD (female n = 120, 1.4 ± 0.06, ∗∗∗p-value < 0.0001, compared with Hml<sup>1</sup> >/+ HFD females, male n = 120, 1.1 ± 0.06, <sup>∗</sup>p-value < 0.0333, compared with Hml<sup>1</sup> >/+ males). (d) Quantification of mean intensity of Dilp2 expression of representative images shown in Figures 3h,i. Control, Hml<sup>1</sup> >/+on HSD (n = 9, 2499 ± 561) and Hml<sup>1</sup> >/UAS-hid on HSD (n = 8, 2681 ± 387.4). (e) Quantification of mean intensity of Dilp5expression of representative images shown in Figures 3j,k. Control, Hml<sup>1</sup> >/+on HSD (n = 10, 1521 ± 425) and Hml<sup>1</sup> >/UAS-hid on HSD (n = 7, 1463 ± 467). (f) Fat body glucose levels. Hml<sup>1</sup> >/+ on HSD (n = 30, 0.021 ± 0.019) and Hml<sup>1</sup> >/UAS-hid on HSD (n = 30, 0.02 ± 0.01). (g,h) pAKT immunostaining in fat bodies of feeding L3 larvae on HSD of (g) control (Hml<sup>1</sup> >/+) and (h) Hml<sup>1</sup> >/UAS-hid are comparable. (i,j) FoxO immunostaining in fat bodies of feeding L3 larvae on HSD of (i) control (Hml<sup>1</sup> >/+) and (j) Hml<sup>1</sup> >/UAS-hid show similar FoxO nuclear localization. (k) Immunoblot analysis of pAkt/Akt ratio in fat bodies of feeding L3 control (Hml<sup>1</sup> >/+) and Hml<sup>1</sup> >/UAS-hid larvae raised on HSD show no change. Fold change ± SD mentioned in the blots. β-Tubulin was used as the internal loading control. (l) Relative fat body mRNA levels of tobi. Fold change is plotted, and statistical analysis was done using C<sup>t</sup> values (Hml<sup>1</sup> >/+ on HSD, n = 80, 11.83 ± 1.32 and Hml<sup>1</sup> >/UAS-hid on HSD n = 80, 11.69 ± 0.98).

Supplementary Figure 5 | In (a–f,l–n′ ,q–s′ ), and (u–u′ ) scale bar = 20µm, (g–h′ ) scale bar = 1 mm, and (i–j′ ) scale bar = 250µm. In (k,l–p,t,v), bar graphs show mean ± standard deviation (SD) and in these panels statistical analysis applied is unpaired t-test, two tailed. "n" is the total number of larvae analyzed. a.u is arbitrary unit and RF is regular food. (a–b′ ′′) Characterization of blood cells in Hml<sup>1</sup> <sup>&</sup>gt;UAS-GFP/UAS PvrAct backgrounds. Compared with blood cells in (a–a′ ′′) P et al. Innate-Immune Control of Animal Growth

control, Hml<sup>1</sup> <sup>&</sup>gt;UAS-GFP, (b–b′ ′′) Hml<sup>1</sup> >UAS-GFP/UAS PvrAct larvae have increased numbers as evident from increased (a,b) DAPI, (a′ ,b′ ) Hml (Hml<sup>1</sup> <sup>&</sup>gt;UAS-GFP) and (a′′ ,b′′ ) phalloidin stainings. (a′ ′′,b′ ′′) Merge of all the channels. (c–f) Immunostainings of (c,d) Dilp2 and (e,f) Dilp5 in L3 feeding larval brain insulin-producing cells (IPCs). As compared with (c,e) control (Hml<sup>1</sup> >/+) on RF, (d,f) Hml<sup>1</sup> >/UAS PvrAct on RF, no change in Dilp2 and Dilp5 staining is detected. Quantification of mean intensities in (l,m). (g–k) Expression of UAS PvrAct in Hml<sup>+</sup> blood cells have not altered adult growth on RF. Representative adult (g–h′ ) fly and (i–j′ ) wing images in (g,g′ ,i,i′ ) control (Hml<sup>1</sup> >/+) and (h,h′ ,j,j′ ) Hml<sup>1</sup> >/UAS PvrAct show no growth modulation. (k) Quantifications of wing span areas of control Hml<sup>1</sup> >/+ on RF (female n = 100, 1.4 ± 0.03, male n = 100, 1.1 ± 0.08) and Hml<sup>1</sup> >/UAS PvrAct on RF (female n = 100, 1.4 ± 0.06, male n = 100, 1.1 ± 0.06). (l) Mean intensity of Dilp2quantification. Control, Hml<sup>1</sup> >/+ on RF (n = 9, 446.88 ± 349.05) and Hml<sup>1</sup> >/UAS PvrAct on RF (n = 9, 649.86 ± 678.53). (m) Mean intensity of Dilp5 quantification. Control, Hml<sup>1</sup> >/+ on RF (n = 9, 411.82 ± 365.73) and Hml<sup>1</sup> >/UAS PvrAct on RF (n = 8, 474.98 ± 476.43). (n) Glucose levels were measured in whole larvae (Hml<sup>1</sup> >/+, RF, n = 18, 0.094 ± 0.036 and Hml<sup>1</sup> >/UAS PvrAct, RF, n = 15, 0.08 ± 0.025), hemolymph (Hml<sup>1</sup> >/+, RF, n = 30, 0.09 ± 0.014 and Hml<sup>1</sup> >/UAS PvrAct RF, n = 30, 0.08 ± 0.03), and fat body (Hml<sup>1</sup> >/+, RF, n = 30, 0.08 ± 0.02 and Hml<sup>1</sup> >/UAS PvrAct, RF, n = 30, 0.08 ± 0.03). (o) Glycogen level measures in whole larvae of Hml<sup>1</sup> >/+, RF (n = 6, 1.08 ± 0.12) and Hml<sup>1</sup> >/UAS PvrAct, RF (n = 6, 1.46 ± 0.16, ∗∗∗p-value = 0.0009). (p) Hemolymph trehalose levels of Hml<sup>1</sup> >/+, RF (n = 66, 0.18 ± 0.05) and Hml<sup>1</sup> >/UAS PvrAct, RF (n = 66, 0.17 ± 0.06). (q–t) Analysis of fat body insulin signaling in (q–s) control (Hml<sup>1</sup> >/+) on RF and (q ′ -s′ ) Hml<sup>1</sup> >/UAS PvrAct on RF shows no change in (q,q′ ) tGPH expression, quantified in (t, r,r′ ) pAKT immunostaining, and (s,s′ ) FoxO expression. (t) Mean tGPH intensity quantification of control, Hml<sup>1</sup> >/+ on RF (n = 25, 2621 ± 486) and Hml<sup>1</sup> >/UAS PvrAct on RF (n = 25, 2875 ± 158). (u,u′ ) Neutral lipid (Nile red) staining in fat bodies of (u) Hml<sup>1</sup> >/+ and (u′ ) Hml<sup>1</sup> >/UAS PvrAct on RF are comparable. (v) TAG levels measurements in whole larvae (Hml<sup>1</sup> >/+ on RF, n = 36, 2.01 ± 0.15 and Hml<sup>1</sup> >/UAS PvrAct on RF, n = 36, 1.98 ± 0.3) and hemolymph (Hml<sup>1</sup> >/+ on RF, n = 48, 0.37 ± 0.15 and Hml<sup>1</sup> >/UAS PvrAct on RF, n = 48, 0.23 ± 0.1).

Supplementary Figure 6 | In (a–c), scale bar = 1 mm, (e–f′ ,h–i′ ) scale bar = 250µm. In (d,g,j), bar graphs show mean ± standard deviation (SD) and in these panels statistical analysis applied is unpaired t-test, two tailed. "n" is the total number of larvae analyzed, RF indicates regular food and HSD indicates high-sucrose diet. (a–d) Larval images showing blood cells (Hml+, GFP, green) in (a) control (Hml<sup>1</sup> >/+), (b) Hml<sup>1</sup> >/UAS PvrWT , and (c) Hml<sup>1</sup> >/UAS Shits . Expression of PvrWT or Shits in Hml<sup>+</sup> cells causes increase in GFP-positive cells, quantified in (d). (d) Quantifications of total blood cell numbers and Hml<sup>+</sup> blood cells in Hml<sup>1</sup> >/+ (total number of blood cells/mm<sup>2</sup> , n = 20, 203.49 ± 41.20 and GFP-positive cells, n = 20, 152.37 ± 42.6), Hml<sup>1</sup> >/UAS PvrWT (total number of blood cells/mm<sup>2</sup> , n = 10, 403.20 ± 114.54, ∗∗∗p-value < 0.0001 and GFP-positive cells, n = 10, 334.32 ± 76.49, ∗∗∗p-value < 0.0001), and Hml<sup>1</sup> >/UAS Shits (total number of blood cells, n = 16, 503.40 ± 68.26, ∗∗∗p-value < 0.0001 and GFP-positive cells n = 16, 430.90 ± 68.14, ∗∗∗p-value < 0.0001). (e–g) Increasing immune cell numbers of an activated state restores HSD growth defect. Compared with (e,e′ ) control wing sizes, Hml<sup>1</sup> >/+ on HSD (f,f′ ), Hml<sup>1</sup> >UAS Pvr on HSD showed a significant growth rescue. (g) Quantifications of wingspan areas of Hml<sup>1</sup> >/+ on HSD (female n = 50, 1.01 ± 0.07, male n = 50, 0.9 ± 0.05) and in Hml<sup>1</sup> >/UAS PvrWT on HSD (female n = 25, 1.09 ± 0.09, ∗∗p-value = 0.0038, male n = 25, 1.03 ± 0.05, ∗∗∗p-value < 0.0001). (h–j) Increasing immune cell numbers by expressing UAS Shits did not restore HSD growth defect. Compared with (h,h′ ) control wing sizes, Hml<sup>1</sup> >/+ on HSD (i,i′ ), Hml<sup>1</sup> >UAS Shits showed a significant growth retardation. (j) Quantifications of wing span areas of Hml<sup>1</sup> >/+ on HSD (female n = 50, 1.0 ± 0.1, male n = 50,

#### REFERENCES


0.90 ± 0.11) and in Hml<sup>1</sup> >/UAS Shits on HSD (female n = 50, 0.9 ± 0.09, <sup>∗</sup>p-value = 0.0223, male n = 50, 0.78 ± 0.09, ∗∗p-value = 0.0031).

Supplementary Figure 7 | In (c–e,g–i,k–m,o–q), scale bar = 250µm. In (a,b,f,j,n,r), bar graphs show mean ± standard deviation (SD) and statistical analysis applied in these panels is unpaired t-test. "n" is total number of larvae analyzed, RF is regular food, HSD is high-sucrose diet, HFD is high-fructose diet, and HGD is high-glucose diet. (a,b) High-sugar diet causes adult growth retardation. Quantification of wing span areas of (a) females (Hml<sup>1</sup> >/+) and (b) males (Hml<sup>1</sup> >/+) reared on different sugar diets. Hml<sup>1</sup> >/+ on RF (female n = 50, 1.45 ± 0.06 and male n = 50, 1.10 ± 0.04), Hml<sup>1</sup> >/+ on HSD (female n = 50, 1.01 ± 0.08, ∗∗∗p-value < 0.0001, compared with RF females and male n = 50, 0.84 ± 0.13, ∗∗∗p-value < 0.0001, compared with RF males), Hml<sup>1</sup> >/+ on HFD (female n = 50, 1.13 ± 0.13, ∗∗∗p-value < 0.0001, compared with RF females and male n = 50, 0.84 ± 0.07, ∗∗∗p-value < 0.0001, compared with RF males), and Hml<sup>1</sup> >/+ on HGD (female n = 25, 0.74 ± 0.08, ∗∗∗p-value < 0.0001, compared with RF females and male n = 25, 0.60 ± 0.11, ∗∗∗p-value < 0.0001, compared with RF males). (c–r) Increasing immune cell numbers of an activated state restores growth defect induced by high-sugar diet as represented in wing span sizes. (c–j) HFD-induced growth defect seen in (c,g) control, Hml<sup>1</sup> >/+ females and males, is restored in (d,h) Hml<sup>1</sup> >/UAS PvrAct HSD condition but not restored in (e,i) Hml<sup>1</sup> >/UAS Shits HFD animals. Contrarily, (e) Hml<sup>1</sup> >/UAS Shits HFD females show a significant growth defect. Quantifications of wing span areas in (f) females and (j) males. (f) Wing span areas of Hml<sup>1</sup> >/+ HFD (female n = 35, 1.13 ± 0.13), Hml<sup>1</sup> >/UAS PvrAct HFD (female n = 51, 1.26 ± 0.08, ∗∗∗p-value < 0.0001), and Hml<sup>1</sup> >/UAS Shits HFD (female n = 25, 0.99 ± 0.09, ∗∗∗p-value < 0.0002) all comparisons made with Hml<sup>1</sup> >/+ HFD females. (j) Wingspan areas of Hml<sup>1</sup> >/+ HFD (male n = 25, 0.84 ± 0.07), Hml<sup>1</sup> >/UAS PvrAct HFD (male n = 25, 0.97 ± 0.07, ∗∗∗p-value < 0.0001), and Hml<sup>1</sup> >/UAS Shits HFD (male n = 25, 0.83 ± 0.07). All comparisons made with Hml<sup>1</sup> >/+ HFD males. (k–r) HGD-induced growth defect seen in (k,o) control, Hml<sup>1</sup> >/+ females and males, is restored in (l,p) Hml<sup>1</sup> >/UAS PvrAct HGD condition but not restored in (m,q) Hml<sup>1</sup> >/UAS Shits HGD animals. Quantifications of wing span areas in (n) females and (r) males. (n) Wing span areas of Hml<sup>1</sup> >/+ HGD (female n = 25, 0.74 ± 0.07), Hml<sup>1</sup> >/UAS PvrAct HGD (female n = 25, 0.96 ± 0.09, ∗∗∗p-value < 0.0001) and Hml<sup>1</sup> >/UAS Shits HGD (female n = 25, 0.77 ± 0.06). All comparisons made with Hml<sup>1</sup> >/+ HGD females. (r) Wingspan areas of Hml<sup>1</sup> >/+ HGD (male n = 25, 0.6 ± 0.11), Hml<sup>1</sup> >/UAS PvrAct HGD (male n = 25, 0.9 ± 0.05, ∗∗∗p-value < 0.0001), and Hml<sup>1</sup> >/UAS Shits HGD (male n = 25, 0.61 ± 0.08). All comparisons made with Hml<sup>1</sup> >/+ HGD males.

Supplementary Figure 8 | In (d, e), scale bar = 20µm. In (a–c,g), bar graphs show mean ± standard deviation (SD) and statistical analysis applied in these panels is unpaired t-test, two-tailed. "n" is the total number of larvae analyzed, a.u is arbitrary unit, FB is fat body, f.c is fold change, and HSD is high-sucrose diet. (a) Mean intensity of Dilp2 quantification of representative image shown in Figures 5a,b. Control, Hml<sup>1</sup> >/+ on HSD (n = 5, 2291 ± 678) and Hml<sup>1</sup> >/UAS PvrAct on HSD (n = 5, 1225 ± 578.6, <sup>∗</sup>p-value = 0.0283). (b) Mean intensity of Dilp5 quantification of representative image shown in Figures 5c,d. Control, Hml<sup>1</sup> >/+ on HSD (n = 5, 1310 ± 304) and Hml<sup>1</sup> >/UAS PvrAct on HSD (n = 5, 580.3 ± 238.7, ∗∗p-value = 0.0029). (c) Fat body glucose levels. Hml<sup>1</sup> >/+ on HSD (n = 35, 0.02 ± 0.017) and Hml<sup>1</sup> >/UAS PvrAct on HSD (n = 35, 0.09 ± 0.03, ∗∗∗p-value = 0.0005). (d–f) Fat body pAKT analysis on HSD. (d,e) Immunostaining of feeding L3 larval fat bodies with anti-pAKT antibody in (d) control, Hml<sup>1</sup> >/+, and (e) Hml<sup>1</sup> >/UAS PvrAct backgrounds shows no change. (f) Immunoblot analysis of pAkt/Akt ratio in fat bodies of feeding L3 larvae of control (Hml<sup>1</sup> >/+) and Hml<sup>1</sup> >/UAS PvrAct reveals a small increase (fold change ± SD mentioned in the blots). β-Tubulin was used as the internal loading control. (g) Fat body analysis of tobi mRNA levels on HSD. Fold change is represented and statistical analysis was done using Ctvalues (Hml<sup>1</sup> >/+ on HSD, n = 80, 11.83 ± 1.32 and Hml<sup>1</sup> >/UAS PvrAct on HSD, n = 80, 12.11 ± 0.73).

melanogaster. PLoS ONE. (2013) 8:e59530. doi: 10.1371/journal.pone.00 59530


experiments. Ecol Immunol. (2011) 25:74–80. doi: 10.1111/j.1365-2435.2010. 01800.x


by reducing insulin signaling. Proc Natl Acad Sci USA. (2009) 106:20853– 8. doi: 10.1073/pnas.0906749106


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

Copyright © 2020 P, Tomar, Madhwal and Mukherjee. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.