A parasitoid wasp of Drosophila employs preemptive and reactive strategies to deplete its host’s blood cells

The wasps Leptopilina heterotoma parasitize and ingest their Drosophila hosts. They produce extracellular vesicles (EVs) in the venom that are packed with proteins, some of which perform immune suppressive functions. EV interactions with blood cells of host larvae are linked to hematopoietic depletion, immune suppression, and parasite success. But how EVs disperse within the host, enter and kill hematopoietic cells is not well understood. Using an antibody marker for L. heterotoma EVs, we show that these parasite-derived structures are readily distributed within the hosts’ hemolymphatic system. EVs converge around the tightly clustered cells of the posterior signaling center (PSC) of the larval lymph gland, a small hematopoietic organ in Drosophila. The PSC serves as a source of developmental signals in naïve animals. In wasp-infected animals, the PSC directs the differentiation of lymph gland progenitors into lamellocytes. These lamellocytes are needed to encapsulate the wasp egg and block parasite development. We found that L. heterotoma infection disassembles the PSC and PSC cells disperse into the disintegrating lymph gland lobes. Genetically manipulated PSC-less lymph glands remain non-responsive and largely intact in the face of L. heterotoma infection. We also show that the larval lymph gland progenitors use the endocytic machinery to internalize EVs. Once inside, L. heterotoma EVs damage the Rab7- and LAMP-positive late endocytic and phagolysosomal compartments. Rab5 maintains hematopoietic and immune quiescence as Rab5 knockdown results in hematopoietic over-proliferation and ectopic lamellocyte differentiation. Thus, both aspects of anti-parasite immunity, i.e., (a) phagocytosis of the wasp’s immune-suppressive EVs, and (b) progenitor differentiation for wasp egg encapsulation reside in the lymph gland. These results help explain why the lymph gland is specifically and precisely targeted for destruction. The parasite’s simultaneous and multipronged approach to block cellular immunity not only eliminates blood cells, but also tactically blocks the genetic programming needed for supplementary hematopoietic differentiation necessary for host success. In addition to its known functions in hematopoiesis, our results highlight a previously unrecognized phagocytic role of the lymph gland in cellular immunity. EV-mediated virulence strategies described for L. heterotoma are likely to be shared by other parasitoid wasps; their understanding can improve the design and development of novel therapeutics and biopesticides as well as help protect biodiversity.

Introduction Parasitoid (parasitic) wasps have an obligatory relationship with their insect hosts. Engaged in a biological "arms race," each partner continuously adapts to the other to emerge alive. For reproductive success, parasitic wasps target their hosts' behavior, development and immune system. Their attack mechanisms range from biochemical warfare and mimicry, to passive evasion and active immune suppression [1][2][3]. Drosophila and their parasitic wasps are an emerging model for studying how wasps evade or suppress host defenses [4,5]. The generalist Leptopilina heterotoma (Lh) succeeds on the Drosophila species within and beyond the melanogaster group. Its close relative, L. boulardi (Lb), considered a specialist, mainly infects flies of the melanogaster group. Both wasps are highly successful on D. melanogaster; they consume its developing larval and pupal stages to emerge as free-living adults [6].
Oviposition into second-to-early-third D. melanogaster larval hosts by Lb and Lh wasps yields divergent immunological effects. Lb infection activates many components of humoral and cellular immunity: Toll-NF-κB, JAK-STAT, and the melanization pathways and their target genes are transcriptionally upregulated; there is a burst of hematopoietic proliferation and differentiation of blood cells (also called hemocytes) in the lymph gland and in circulation. If the immune responses are strong and sustained, macrophages and lamellocytes encapsulate and kill wasp eggs [7][8][9][10][11]. Lh infection in contrast suppresses immune gene expression and kills immature and mature larval hemocytes [12,13].
Lb and Lh females (and also L. victoriae, a sister species of Lh) produce discrete immunesuppressive extracellular vesicle-(EV) like structures in their venom glands (called multi-strategy extracellular vesicles, MSEVs in Lh and venosomes in Lb) [14,15]. Previously called virus like particles [16][17][18], these EVs lack clear viral features [15,19]. They are produced in the venom gland, a structure made up of the long gland and a reservoir. The secretory cells of the long gland synthesize and secrete proteins, some of which are initially incorporated into discrete non-spiked vesicle-like structures. In sister species Lh and Lv, these structures mature in the reservoir and assume a stellate morphology with 4-8 spikes radiating from the center. Mature EVs are roughly 300 nm in diameter, [14,16,[20][21][22]. Packed with more than 150 proteins, EVs are, in part, responsible for divergent physiological outcomes in infected hosts [15,19,23].
Among the most abundant in the Lh EV proteins is a 40 kDa surface/spike protein (SSp40) [20]. SSp40 shares structural similarities with the IpaD/SipD family of proteins of the gastroenteric disease-causing Gram negative bacteria, Shigella and Salmonella [15]. Similar to SSp40's localization to Lh EV spike tips, IpaD localizes to the tips of the T3 secretion injectisome, a bacterial transfer system that injects virulence proteins into mammalian cells. IpaD itself promotes apoptosis of mammalian macrophages [24,25]. These parallels between Lh SSp40 and bacterial IpaD/SipD suggest that Lh EVs may share some similarities with bacterial secretion systems. Comparative transcriptomic/proteomic approaches revealed that SSp40 and a few other EV proteins are not expressed in the Lb venom [15].
Whereas Lh EVs lyse lamellocytes within a few hours of wasp attack, Lb EVs do not have the same effect [7,16,26]. Our immune-inhibition experiments suggested that Lh's SSp40 mediates EV interactions with lamellocytes [20]. Lh infection also uniquely promotes apoptosis of larval macrophages and of lymph gland hemocytes [13]. Macrophages make up more than 95% of all hematopoietic cells while differentiated lamellocytes are rarely found in naïve hosts [8][9][10]27]. Work in the field strongly suggests that the protein activities concentrated within the Lh EVs are responsible for the destruction of these mature and immature blood cells. However, how a macro-endoparasite targets the hematopoietic system and accesses its progenitor population has not been studied. The modes of Lh EV entry into these cells and the pathways of destruction are also not well understood.
The goal of this study was to obtain a macro-level view of Lh EV interactions with cells of the larval hemolymphatic system after infection. The term hemolymph refers to the interstitial fluid that distributes hormones, peptides and other macromolecules into organs through the pumping action of an unbranched tubular heart, or the dorsal vessel. Dozens of macrophages circulate in the hemolymph. The heart lumen is surrounded by a column of paired cardiomyocytes and associated pericardial cells. This tubular structure is held in place by alary muscles [28][29][30]. Hematopoietic cells are organized in paired cell clusters (or lobes) on the dorsal vessel. In third instar larvae, the anterior-most lobes have blood cells at various stages of differentiation; the least differentiated progenitors are confined adjacent to the dorsal vessel, whereas the developing macrophages are sequestered in the cortical regions of the lobes (Fig 1A). In naïve hosts, the progenitor state is maintained by a putative niche (also called the posterior signaling center, PSC). The PSC is a tight unit of about 25-50 cells and is positioned posteriorly to the progenitors [31,32]. Upon Lb infection, the PSC reprograms hematopoiesis inducing macrophage and lamellocyte differentiation [30,[32][33][34][35][36][37][38][39]. The entire structure is covered by the acellular basement membrane [28,40].
Using Drosophila genetics, cell-specific markers and SSp40 staining as a proxy for Lh EV localization, we have pieced together a broad view of this host-parasite interaction interface. We show an abundance of Lh EVs in (a) the lumen of the larval dorsal vessel, (b) along the collagen/perlecan-based basement membrane around the dorsal vessel and surrounding clusters of lymph gland progenitors, and (c) inside the progenitor and mature macrophages. Moreover, high EV signal correlates with the disassembly of the cohesive PSC unit. PSC ablation limits EV internalization and loss of lymph gland hemocytes, while PSC inactivation via hedgehog (hh) knockdown (KD) does not have this effect. We also show that lymph gland hemocytes can phagocytose Lh EVs using the classical Rab5-mediated retrograde transport (RGT) pathway. Surprisingly, Rab5 also maintains fly macrophage identity, as Rab5 KD leads to over-proliferation, lamellocyte differentiation and tumorigenesis; Rab5 function is cell-autonomous. Thus lymph glands are not merely a source of mature blood cells but are themselves immune competent organs and can clear the immune-suppressive Lh EVs to defend the host. However, Lh EVs proactively dislodge cells of the PSC, blocking differentiation of the protective immune cells. Lh EVs target the endomembrane system of macrophages that ultimately results in their apoptosis, thus highlighting central and previously unrecognized roles of the lymph gland in cellular immunity. These observations help explain why Lh infections target the larval lymph gland. The direct EV-macrophage interactions and cellular outcomes set the stage for future molecular analyses in both the hosts and parasites.

Lh EVs are present within the larval lymphatic system
Lb17 attack triggers lamellocyte differentiation in the larval lymph gland cortex (Fig 1A-1C). At an equivalent time-point, Lh-infected lobes are significantly smaller (Fig 1D-1K); [13]. Surprisingly, unlike Antp>mCD8GFP-expressing PSCs of naïve and Lb-infected lobes that remain tightly clustered (Fig 1A-1C), PSC cells of Lh-infected hosts are dislodged and some are distributed in the body of the lobe (Fig 1D).
To understand these responses, we imaged more than 25 hosts in multiple experiments. Throughout these studies, we used a polyclonal antibody to mark SSp40, an Lh EV-specific protein [20]. Lh infection of Pxn>GFP animals reduced Pxn>GFP expression (Fig 1E-1H). (Pxn is normally active in the cortex and its expression mimics that of many other genes downregulated by Lh infection [7]). An abundance of Lh EVs was observed in anterior-and posterior-lobe hemocytes, in the dorsal vessel, and in circulating hemocytes (Fig 1F-1H and 1J-1L). This staining signal is absent in glands from naïve animals (Fig 1E and 1I). Thus, Lb and Lh attack have drastically different outcomes and Lh EVs appear to interact directly with most lymph gland hemocytes.
In our analyses across experiments, we found that the degree of tissue loss and EV distribution varies. Lobe morphologies ranged from nearly intact and filled with EVs ( Fig 1F and 1J) to damaged lobes, with few to many EVs ( Fig 1G, 1H and 1K). This variation is likely due to (a) the duration of infection (i.e., time between oviposition and dissection); (b) the injected EV dose; or (c) the dynamics of EV circulation. Dissections at later time points showed loss of almost all lymph gland hemocytes [13].
To then probe how Lh EVs enter the lymphatic system, we stained Lh-infected glands from fly strains with GFP-tagged Collagen IV (basement membrane, Viking [41]) or GFP-tagged proteoglycan core protein, Perlecan/Trol [42]. In both cases, EV puncta were clearly localized with the continuous GFP signals of these extracellular matrix (ECM) proteins along the dorsal vessel as well as in the interstitial spaces around clustered hematopoietic progenitors (yellow puncta in Fig 2A-2D' arrows). Surprisingly, punctate staining was also seen inside immature progenitors, adjacent to the dorsal vessel ( Fig 2B, 2B' 2D and 2D'). EVs were also observed inside some cardiomyocytes as evidenced by SSp40 colocalization with the mCD8GFP signal in HandΔ>mCD8GFP larvae ( Fig 2E-2F', arrows). Thus, Lh EVs associate with ECM proteins of the lymphatic system, enter the dorsal vessel lumen and even some cardiomyocytes.

Effects of Lh infection on PSC integrity
Previous studies have demonstrated that upon Lb infection, the PSC reprograms hematopoietic development and promotes lamellocyte differentiation. Similar to controls, Lb infected PSCs remain tightly clustered [35,37,38,43,44]. We found that regardless of the Lh strain, PSC cells dislodged from their normal posterior position into the body of the lobe (Figs 1D and S1A and S1B). At the time points we examined, half of the PSC cells relocated into the lobe, of which an overwhelming majority (> 95%) were present as single cells or as groups of two cells. The other half remained in their original place, although some were not as tightly clustered (Figs 1D and S1; n = 24 PSCs from 12 lymph glands). As expected, in 14 lobes from naïve controls, all PSC cells were tightly packed. Strikingly, in samples from infected animals, where the PSCs were still intact, EVs congregated in regions adjacent to the PSC, but were never found inside the PSC cells (S1A and S1B Fig).
The Slit ligand, originating from adjacent cardiomyocytes, controls PSC integrity via the Robo receptors in the PSC; Robo2 has the strongest effect [45]. Indeed, the effect of Lh infection on the PSC resembles Slit/Robo2 KD, which promotes PSC disassembly [45] (Figs 1 and S1). A cohesive PSC is important in hematopoietic development as fewer macrophages and crystal cells develop in Slit/Robo2 KD lobes compared to controls. But differing from Lh infection, Slit KD PSCs are larger and there is no apparent loss of progenitors [45]. In spite of different outcomes in the two conditions, we hypothesized that the initial steps might be shared and that Lh EVs might inactivate the Slit-Robo signal, which might explain PSC disassembly.
To test this idea, we infected animals in which the Slit-Robo pathway was manipulated to promote constitutive signaling. We found that neither expressing active Slit nor overexpressing Robo2 altered Lh EVs' ability to disassemble the PSC.

PSC-less lymph glands remain intact
PSC-less lymph glands were unable to induce lamellocyte differentiation after Lb infection [36]. We asked if ablating the PSC might similarly inhibit the Lh infection responses.  Taken together, these results suggest that the PSC plays a structural role in trafficking of EVs from either the hemolymph or the dorsal vessel into the lobes and that the cell-lethal effects of Lh EVs is distinct from the PSC's niche function.

The larval lymph gland is phagocytically competent
We next studied if EV uptake into hematopoietic cells occurs via RGT mechanisms. In Pxn>GFP; +/Bc heterozygous lymph glands, we observed blackened, dead crystal cells in the cytoplasm of the GFP-positive cortical cells (S6 Fig). This observation suggests that GFP-positive lymph gland cells are phagocytically competent. We therefore investigated if Lh EV uptake depends on Rab5, an early endosomal protein. Rab5 mediates trafficking from the plasma membrane to early endosomes [47]. In contrast to Pxn>GFP macrophages, where EV staining is bright and punctate throughout the cytoplasm (Fig 4A and 4B), Pxn>GFP, Rab5 RNAi cells show peripheral punctate staining, presumably from intact EVs, trapped in early endosomes, both in lymph gland and circulating hemocytes (Fig 4C and 4D; arrows). In lamellocytes, the EV signal is diffuse and nuclear, and Rab5 KD shows no change in staining intensity or distribution (Fig 4E and 4F; arrowheads), suggesting that Rab5-independent uptake mechanisms

Lh EVs negatively impact phagolysosomal organization in macrophages
Rab7 mediates late endosome formation and trafficking between late endosomes and lysosomes is marked by Rab7 and LAMP1, respectively [47]. To evaluate if Lh EVs impact the RGT machinery, infected glands expressing GFP-tagged Rab5, Rab7, or LAMP1 proteins were examined (Fig 5). Under our experimental conditions, Lh EVs rarely colocalized with early endosomes and Rab5 compartment morphology remained comparable to uninfected controls (only 14% of SSp40 puncta are Rab5-positive; n = 221 cells; 6 lobes; Fig 5A-5C). In contrast, Lh EVs were consistently found with GFP-Rab7 and GFP-LAMP1 and these compartments were grossly distorted (100% co-localization; n = 115 and 112 cells, Rab7 and LAMP, respectively, 6 lobes each; Fig 5D-5E"). Moreover, the Rab7/EV and LAMP1/EV signals were asymmetrically localized in Lh-affected cells. In contrast, Lb infection did not distort LAMPpositive compartments and they retained their normal morphology (Fig 5F and 5G). These observations suggest that high numbers of Lh EVs transit through early endosomes, but that they are retained in late RGT compartments including lysosomes. Thus, Lh EVs have a detrimental effect on RGT compartment integrity and this loss of integrity may promote lysosomal leakage and labilization.

Rab5 suppresses proliferation and maintains the macrophage fate
We were surprised to find that Pxn>GFP Rab5 RNAi animals developed melanized tumors (S8A and S8B Fig); the hematopoietic population is significantly expanded and lamellocyte differentiation is robust, affecting viability (S8C- S8F Fig). (Variability in viability and tumor development in Rab5 KD animals is likely due to differences in the strengths and expression patterns of the GAL4 drivers.) A similar result was observed with the expression of the dominant negative Rab5S43N protein which cannot bind GTP [48]). Rab5 KD even in the lymph gland medullary zone (TepIV>Rab5 RNAi ) resulted in lamellocyte differentiation (Fig 6A and 6B).
Hematopoietic expansion correlated with increased mitotic index (MI) in lobes of tumorbearing Pxn>GFP Rab5 RNAi animals (Fig 6C and 6D) suggesting that normal Rab5 function checks over-proliferation and ectopic progenitor differentiation. (MI = 2.2 ± 2.15 in control Pxn>GFP; 2.9 ± 1.1, and 5.5 ± 1.9 in experimental Pxn>GFP, Rab5 RNAi animals without and with tumors, respectively (n = 10 for each condition). These results suggest that Rab5 acts as a tumor suppressor and maintains hematopoietic immune quiescence.
Control "FLP-out" clones without Rab5 RNAi contained small cells that did not express integrin-β; experimental clones with Rab5 RNAi had larger, F-actin-rich cells with a typical lamellocyte morphology, that were also integrin-β-positive (Fig 6E-6H). These results suggest that Rab5's requirement in maintaining the progenitor or macrophage fate is cell-autonomous.

System-wide distribution but specific effects of Lh EVs
Parasitism by L. heterotoma has been of interest because of its ability to parasitize many Drosophila hosts and the existence of venom factors that kill host hemocytes. Lb lacks these activities (Fig 7). The discovery of an anti-lamellocyte activity intrinsic to Lh EVs provided initial insights into the critical roles of these EVs in parasitism [20,49]. However, details underlying their apoptotic effects on macrophages have been lacking. This work provides the first view into how Lh EVs rely on the host's circulation to gain system-wide distribution to not only precisely kill the available effector cells but also to pre-emptively interfere with the host's ability to produce additional effector cells. We show that lymph glands serve an important, previously unappreciated role in immunity. A majority of lymph gland cells can phagocytose Lh EVs to protect the host from their detrimental effects. EV activities in turn promote their apoptotic death by disrupting their endomembrane system. This work also lays bare new questions. Lh EVs' association with the ECM proteins around the lymphatic system cells suggests ways in which EVs might recognize and home into the lymph gland hemocytes and cardiomyocytes although the role of the ECM, the details of their entry and physiological effects on cardiomyocytes are currently unclear. As has been suggested for Slit-carrying vesicles [45], cardiac cells might provide a route for Lh EVs to converge into the vicinity of the PSCs. Once inside lymph gland hemoctytes, they simultaneously target the protective functions of macrophages and lamellocytes and their activities culminate to strongly block encapsulation (Fig 7). These strategies are not uncommon and are likely to be shared by closely related Leptopilina wasps or even unrelated virulent wasps that attack drosophilid and non-drosophilid hosts and are known to destroy their hosts' hematopoietic cells [50-54].

Lh EVs proactively block encapsulation
Studies with Lb showed that the lymph gland itself responds to wasp infection and lamellocytes differentiate from hematopoietic progenitors [8,9,55,56]. In addition to PSC's niche function in naïve animals [31,32,[57][58][59], the PSC also appears to play an anti-parasite role as Lb infection promotes lamellocyte differentiation [30,[35][36][37][38]43,60]. Given this latter role, it is reasonable to interpret Lh's effects on PSC integrity as part of a corresponding adaptive strategy that Lh has acquired during its evolutionary history.
The high Lh EV levels around the PSC and its disassembly provide novel physiological insights into PSC functions and raise intriguing mechanistic questions. A normally clustered and cohesive PSC organization is needed for proper hematopoietic differentiation in naïve animals [45]. Although Lh infection-induced PSC disassembly and hemocyte loss are observed together in fixed samples, our data from infected PSC-less animals suggest that PSC disruption might precede hemocyte death. It is possible that the PSC may somehow "recognize" foreign entities and might serve to protect the progenitor microenvironment by acting as a chemical or mechanical barrier between the vascular and hematopoietic cells. In this scenario, Lh EVs may be targeting PSC cohesiveness to inactivate this barrier function.
This interpretation is consistent with the recently discovered permeability barrier in the PSC that is breached by systemic bacterial infection. Permeability barrier in the PSC is maintained by septate junctions and their disruption is linked to increased Toll signaling, cellular immune activation and improved host survival [39]. The barrier function between the vascular and hematopoietic cells proposed here would serve to limit the ingress of structures such as microbes and EVs. The mechanisms underlying the effects of ablated PSCs are unclear but it is notable that PSC-less lobes do not respond to either Lb or Lh infections (our results and [36]). Examining whether Lb EVs/venosomes similarly interact with the lymph gland ECM, congregate around the PSC, and are phagocytosed by hemocytes will shed light on these processes. Extracellular vesicles secreted from mammalian neutrophils and endothelial cells can direct cell motility and chemotaxis [61,62]. Thus, it is possible that Lh EV activities [15] are similarly responsible for PSC disassembly. These activities may perturb host pathways required for normal PSC cohesion and function, although manipulating the Slit-Robo pathway components, or hh signaling, hypothesized to disrupt the infection process, was insufficient to block Lh's ability to disperse the PSC and attack hemocytes. Lh EVs may possess redundant or independent mechanisms to control PSC integrity and this question remains open for further research.

A central role for phagocytosis in the anti-parasite response
The ability of macrophages to ingest and kill microbes is a fundamental facet of innate immunity. Microbes have evolved to evade or escape the destructive conditions in their host cells' phagolysosomes. While most intracellular pathogens avoid fusion with lysosomes, others modify endocytic trafficking differently to survive in their host cells [63]. We have shown that like microbes, Lh EVs are endocytosed and can damage the late endocytic compartments. This suggests that their biochemical activities may distort and damage intracellular membranes although how this occurs mechanistically is unclear. A novel family of Lh EV-associated GTPases [15] are possible candidates for such activities as expression of select GTPases in yeast alter vacuolar morphologies [64]. Microbial infection of macrophages can activate apoptosis responses [65][66][67] and it may be that similar effects of Lh EVs in fly macrophages are directly linked to their apoptosis.
Lamellocytes utilize a flotillin/lipid raft dependent mechanism to internalize Lb EVs [68], and it is likely that Lh EVs use the same or a similar pathway. The significance of the nuclear SSp40 signal in lamellocytes after Lh infection is unclear; but because the signal is not punctate, Lh EVs are likely internalized via a membrane fusion step in which their vesicular character is lost. Electron microscopy results also show that unlike macrophages with membrane-enclosed endocytic vesicles containing intact Lh EVs, lamellocytes do not have such compartments, and once internalized, EVs lose structural integrity [49]. Efferocytosis of lysed lamellocytes appears to be an effector anti-parasite response and it may ultimately also be beneficial to parasite development.
Our genetic studies with Rab5 highlight the central role of the endocytic processes in the anti-parasite response. Loss of endocytic trafficking activates immune signaling ( [69][70][71]). From a physiological standpoint, it makes sense why the lymph gland is a dedicated target of wasp infections. Both key aspects of anti-wasp cellular immunity, i.e., phagocytosis of the wasp's EVs and lamellocyte differentiation, unequivocally reside in the lymph gland. These ideas can be further examined at the molecular level with the available descriptions of the Lh and Lb EV proteomes [15,19,23]. Virulence factors provide the armament for parasite success in the host/pathogen "arms race". Insights from this model host-parasite system can influence our understanding of how parasite-derived factors have shaped the immune physiology of fly hosts.

Wasp infections
y w flies were used to rear wasps. Unless specified otherwise, infections were done with either Lb17 or Lh14 [7]. LhNY [20] was used to validate results with the Lh14 strain. Ten to twelve trained female wasps were introduced to hosts from a 12-hr egg-lay. Hosts were allowed to recover after an 8-12 hr infection. Dissections were typically done one-to-two days after infection. Uninfected controls followed the same timeline. In general, longer infection regimes led to stronger responses: more lamellocytes differentiated after Lb infection and more lobe cells were lost after Lh infection. Under our experimental conditions, superparasitism by either wasp was rare and for our analyses, we avoided hosts with more than one parasite.

Confocal imaging
Mounted samples were imaged with Zeiss laser scanning confocal microscopes LSM 510 or LSM 710. For each experiment, images were scanned on the same microscope with the same software and scan settings. Images were gathered at 0. Rab5 KD animals varied with different GAL4 drivers. (F) Viability to adulthood was differentially affected. More than 100 animals were scored for each cross in panels (E) and (F