Two Nimrod receptors, NimC1 and Eater, synergistically contribute to bacterial phagocytosis in Drosophila melanogaster

Eater and NimC1 are transmembrane receptors of the Drosophila Nimrod family, specifically expressed in haemocytes, the insect blood cells. Previous ex vivo and in vivo RNAi studies have pointed to their role in the phagocytosis of bacteria. Here, we have created a novel NimC1 null mutant to re‐evaluate the role of NimC1, alone or in combination with Eater, in the cellular immune response. We show that NimC1 functions as an adhesion molecule ex vivo, but in contrast to Eater it is not required for haemocyte sessility in vivo. Ex vivo phagocytosis assays and electron microscopy experiments confirmed that Eater is the main phagocytic receptor for Gram‐positive, but not Gram‐negative bacteria, and contributes to microbe tethering to haemocytes. Surprisingly, NimC1 deletion did not impair phagocytosis of bacteria, nor their adhesion to the haemocytes. However, phagocytosis of both types of bacteria was almost abolished in NimC1 1 ;eater 1 haemocytes. This indicates that both receptors contribute synergistically to the phagocytosis of bacteria, but that Eater can bypass the requirement for NimC1. Finally, we uncovered that NimC1, but not Eater, is essential for uptake of latex beads and zymosan particles. We conclude that Eater and NimC1 are the two main receptors for phagocytosis of bacteria in Drosophila, and that each receptor likely plays distinct roles in microbial uptake.


Introduction
Phagocytosis is an ancient and evolutionarily conserved process, generally defined as the cellular uptake of particles bigger than 0.5 lm. Phagocytosis is an important feeding mechanism in primitive and unicellular organisms, such as amoeba [1]. In higher organisms, phagocytosis is performed by dedicated cells (phagocytes) and is used as a powerful process to internalize and eliminate pathogens, as well as to trigger host inflammation [2]. Moreover, phagocytosis contributes to tissue homeostasis and embryonic development, mainly via the removal of apoptotic corpses [3]. Phagocytosis is a complex membrane-driven process guided by the actin cytoskeleton of the host phagocytic cell. It involves the recognition and subsequent binding of the microbe by surface receptors. These interactions are essential to activate intracellular signalling pathways that finally culminate in the formation of the phagosome [4]. Several studies have highlighted similarities between the phagocytic machinery of Drosophila and mammals, such as the involvement of actin and actin-related proteins [5][6][7]. Drosophila melanogaster harbours highly efficient phagocytes, called plasmatocytes, which originate from multipotent progenitors (prohaemocytes). In healthy larvae, prohaemocytes can differentiate into two mature haemocyte types: plasmatocytes and crystal cells. While the later are involved in the melanization response [8], plasmatocytes are professional phagocytes sharing functional features with mammalian macrophages, and represent the most abundant haemocyte class at all developmental stages. They play a key role in bacterial clearance during infection, as well as in the removal of apoptotic corpses [9,10]. The ability of Drosophila haemocytes to perform efficient phagocytosis relies on the expression of specific cell surface receptors that can bind particles and induce their engulfment. While many receptors have been implicated in bacterial phagocytosis, their specific involvement or individual contribution is less clear [11,12]. In this paper, we have characterized the phagocytic role of NimC1 and Eater, two EGF-like repeat Nimrod surface receptors specifically expressed in haemocytes [13,14]. The Nimrod family of proteins is characterized by the presence of epidermal growth factor (EGF)-like domains, also called 'NIM repeats'. This family comprises a cluster of 10 genes (NimA, NimB1-5 and NimC1-4) encoded by genes clustered on the chromosome II, and two related haemocyte surface receptors, Eater and Draper, encoded by genes on chromosome 3 [14,15]. Early studies have shown the implication of some Nimrod C-type proteins in bacterial phagocytosis (Eater and NimC1) [13,14,16] or engulfment of apoptotic bodies (Draper and NimC4/SIMU) [17,18]. More recently, the Eater transmembrane receptor has also been involved in haemocyte adhesion and sessility [16]. Nimrod C1 (NimC1) is a 90-kDa transmembrane protein characterized by 10 NIM repeats in its extracellular region, a single transmembrane domain and a short cytosolic tail with unknown function [14]. NimC1 has been initially identified as the antigen of a haemocytespecific antibody (P1), being involved in phagocytosis of bacteria [14]. Kurucz et al. [14] showed that NimC1 silencing by RNAi decreases Staphylococcus aureus uptake by plasmatocytes, whereas its overexpression in S2 cells enhances phagocytosis of both S. aureus and Escherichia coli bacteria and makes the cells highly adherent. Here, we generated a null mutation in NimC1 by homologous recombination (called NimC1 1 ) and revisited its function in haemocyte-mediated immunity. Moreover, we recombined the NimC1 mutation with the previously described eater 1 mutant [16], generating a NimC1 1 ;eater 1 double mutant. Using these genetic tools, we first show the involvement of NimC1 in ex vivo cell adhesion and in the regulation of haemocyte proliferation. Contrasting with previous RNAi studies [14], our ex vivo phagocytosis assays demonstrate that NimC1 is not required for phagocytosis of Gram-positive or Gram-negative bacteria. Nevertheless, we show that this Nimrod receptor contributes to the uptake of latex beads and zymosan yeast particles. The use of the NimC1 1 ;eater 1 double mutant not only reconfirmed Eater as the main Grampositive engulfing receptor, but, more importantly, revealed a synergistic action of NimC1 and Eater in microbe phagocytosis. NimC1 1 ;eater 1 haemocytes from third instar larvae, failed indeed to phagocytose any type of bacteria. Collectively, our study points to a major role of NimC1 and Eater in the phagocytosis of bacteria, and suggests that those proteins likely play distinct roles in microbial uptake, as tethering and docking receptors.

Generation of a NimC1 null mutant by homologous recombination
In order to characterize NimC1 functions, we generated a null mutant by deleting the corresponding NimC1 gene region. The deletion removes the ATG translation start site and the following 852-bp sequence. The knockout was performed in the w 1118 background, using homologous recombination [19], which also leads to the insertion of a 7.9-kb cassette carrying the white + gene (Fig. 1A,B). Functional deletion of NimC1 was confirmed by RT-PCR performed on total RNA and by P1 (anti-NimC1 antibody [14]) immunostaining (Fig. 1C,D). As NimC1 is specifically expressed in haemocytes and has been implicated in phagocytosis, we combined the NimC1 mutation with the previously described eater 1 null mutant [16], generating a double mutant NimC1 1 ;eater 1 (Fig. 1). Both NimC1 1 and NimC1 1 ;eater 1 flies were viable and did not show any developmental defect. For overexpression studies, we also generated flies containing the NimC1 gene downstream of the UAS promoter. Using these tools, we characterized the function of NimC1 focussing on haemocytes of third instar larvae.

NimC1-deficient haemocytes show adhesion defects in vitro
Eater has been involved in haemocyte adhesion and sessility [16]. Given the structural similarities between NimC1 and Eater [14], we first investigated the role of NimC1 in cell adhesion. We observed that the cell area of NimC1 1 -and eater 1 -adherent haemocytes was decreased compared to that of w 1118 wild-type control ( Fig. 2A) [16]. Notably, the cell area of NimC1 1 ;eater 1adherent haemocytes was significantly smaller than that of single mutants. Quantification analysis revealed that wild-type haemocytes have a mean cell area of 237 lm 2 , while NimC1 1 , eater 1 and NimC1 1 ;eater 1 mutants have 120, 114 and 99.7 lm 2 respectively (Fig. 2B). Image-based cytometry analysis of free-floating haemocytes revealed that the spreading defects  Cell nuclei are shown in DAPI (blue). The immunostaining was performed as previously described in [59]. Scale bar: 20 lm.   observed in our mutants were not due to an inherently smaller cell size (Fig. 2C). In order to get a deeper insight into these adhesion defects, we investigated haemocyte morphology by scanning electron microscopy (SEM). Lamellipodia are a key feature of highly motile cells, playing a central role in cell movement and migration [20]. They represent flat cellular protrusion, characterized by an enriched network of branched actin filaments. Filopodia, instead, are rather used by the cell to sense the surrounding microenvironment, and consist of parallel actin filaments that emerge from the lamellipodium. Spread plasmatocytes from wild-type larvae appeared as round adherent cells with a central bulge within the cell body, from which lamellipodia and filopodia extended (Fig. 2D). NimC1 and eater null haemocytes were still able to form narrow filopodia projections. However, both single and double mutants showed an obvious lamellipodium decreased region compared to wild-type control (Fig. 2D). Collectively, our results point to a role of NimC1 in haemocyte spreading and lamellipodia extension.
In the Drosophila larva, circulating haemocytes can attach to the inner layer of the cuticle, forming striped patterns along the dorsal vessel, and lateral patches in association with the endings of peripheral neurons [8,[21][22][23]. These subepidermal sessile compartments are known as haematopoietic pockets [21,[23][24][25][26]. Previous work has shown that eater larvae lack the sessile haemocyte compartment and have all peripheral haemocytes in circulation [16]. To further investigate whether the NimC1 deletion affects sessility, we explored haemocyte localization using the haemocyte marker HmlΔGal4>UAS-GFP by whole larva imaging and cross-section visualization. In NimC1 1 ,HmlΔGal4, UAS-GFP third instar (L3) wandering larvae, haemocytes were still able to enter the sessile state, forming dorsal and lateral patches (Fig. 2E,F). In contrast, both eater 1 and NimC1 1 ;eater 1 larvae lacked sessile haemocytes, all plasmatocytes being in circulation (Fig. 2E,F). In vivo RNAi targeting NimC1 confirmed the haemocyte adhesion defect observed with the null mutant (Fig. 2G,H). This indicates that the observed phenotypes were indeed caused by the deletion of NimC1 and not the genetic background. Altogether, our data indicate that NimC1 contributes to haemocyte adhesion ex vivo, but in contrast to Eater, it is not directly required for haemocyte sessility in vivo.

NimC1 null larvae have an increased number of haemocytes
Drosophila haematopoiesis occurs in two successive waves. A first set of haemocytes is produced during embryogenesis, giving rise to a defined number of plasmatocytes and crystal cells. This embryonic haemocyte population expands in number during the following larval stages. The second haemocyte lineage derives from the lymph gland, a specialized organ that develops along all larval stages. The lymph gland acts as a reservoir of both prohaemocytes and mature haemocytes, which are released at the onset of metamorphosis or upon parasitization [8,[27][28][29][30]. Finally, accumulating evidence suggests that the sessile haematopoietic pockets also function as an active peripheral haematopoietic niche [21,23,26]. In order to further investigate the role of NimC1, and its potential interaction with Eater in peripheral haematopoiesis, we counted by flow cytometry the number of all the peripheral haemocyte populations (i.e. both sessile and circulating). Larvae containing the haemocyte marker HmlΔdsred.nls, combined with the NimC1 and eater null mutants, were used ( Fig. 3A-C). Our study confirmed that eater L3 wandering mutant larvae have more haemocytes than the wild-type [16] (Fig. 3A). Similarly, NimC1 1 third instar larvae have 3.2 times more circulating haemocytes compared to the wild-type (Fig. 3A). As NimC1 1 L2 larvae have a wild-type like number of haemocytes, the increase in haemocyte counts in this mutant takes place at the end of larval development (Fig. 3C). Surprisingly,  haemocyte number was six times higher in NimC1 1 ; eater 1 double mutant L3 larvae (Fig. 3A), suggesting that eater and NimC1 additively regulate haemocyte counts. A higher haemocyte number was already observed in second instar larvae in the double mutant (Fig. 3C). We next investigated whether lymph glands from third instar mutant larvae had an increased number of mature haemocytes compared to wild-type. Visual count of HmlΔdsred.nls-positive cells from fixed primary lymph gland lobes revealed no major differences between mutants and wild-type, although a decreased trend in haemocyte number in single and double mutants could be observed (Fig. 3D,E). In agreement with this observation, primary lymph gland lobes of eater and NimC1 mutants showed a modest reduced area compared to control, which was not statistically significant (Fig. 3F). Nevertheless, the ratio of HmlΔdsred.nls-positive cells to the all primary lymph gland cell population (i.e. DAPI positive), was not significantly altered between wild-type and mutants (Fig. 3G).
We then decided to explore whether the increase in peripheral haemocytes count observed in our mutants ( Fig. 3A-C) was caused by a higher Phalloidin DAPI Phalloidin DAPI proliferation rate. EdU incorporation experiments revealed that NimC1 1 and eater 1 single mutants have a higher frequency of peripheral proliferating haemocytes compared to wild-type in middle L3 but not L2 larvae (Fig. 3H,I). The higher proliferation rates might, therefore, explain the increased number of haemocyte counts in both L3 wandering (Fig. 3A) and middle L3 (Fig. 3B) larvae. Interestingly, we found that both haemocyte count and mitotic rate were higher in NimC1 1 ;eater 1 in L2 and L3 larvae indicating that both receptors additively regulate haemocyte proliferation levels ( Fig. 3A-C,H,I). The higher haemocyte count in NimC1 mutant larvae was phenocopied when using an in vivo RNAi approach to silence NimC1 (Fig. 3J). Of note, overexpression of NimC1, using the HmlΔ-Gal4 plasmatocyte driver, did not increase the peripheral haemocyte count (Fig. 4A), nor their adhesion properties ( Fig. 4B-D). Overexpression of NimC1 in haemocytes from eater-deficient larvae did not rescue the lack of sessility phenotype and the ex vivo adhesion defect caused by the absence of eater (Fig. 4E). We also investigated a possible role of NimC1 in crystal cell and lamellocyte differentiation.

>UAS-NimC1 >w
Crystal cells are the second haemocyte type present in noninfected larvae, specifically involved in the melanization response and wound healing [31]. Crystal cells can be found in both the sessile and circulating state. Recent studies have shown that a fraction of those cells derive from sessile plasmatocyte by transdifferentiation [26]. Consequently, crystal cells need sessile plasmatocytes to be, themselves, sessile [16]. Lamellocytes are barely present in healthy larvae, but can differentiate from plasmatocytes [25,32] or prohaemocytes [33] in response to specific stress signals, such as parasitization. They are thought to play an essential role in encapsulation of parasitoid wasp eggs. Our study indicates that NimC1 mutants retain the ability to differentiate fully mature crystal cells (Fig. 5). Moreover, our data also show that the NimC1 deletion does not affect the ability to encapsulate parasitoid wasp eggs (Fig. 6). Finally, we did not uncover any role of NimC1 in the systemic antimicrobial response of larvae against Gram-positive (Micrococcus luteus) or Gram-negative bacteria (Erwinia carotovora carotovora), as revealed by the wild-type-like induction of Diptericin and Drosomycin gene expression, two target genes of the Imd and Toll pathways respectively [34] (Fig. 7).

NimC1 contributes with Eater to phagocytosis of bacteria
A previous in vivo RNAi approach had revealed a role of NimC1 in the phagocytosis of Gram-positive bacteria [14]. We used the NimC1 deletion to further elucidate the requirement of this receptor in bacterial uptake by performing ex vivo phagocytosis assays at two different time points (early-30 min and late-60 min). As previously reported [16], eater null mutant haemocytes were impaired in their capacity to phagocytose the Gram-positive bacterium S. aureus (Fig. 8A,B), but not the Gram-negative bacterium E. coli (Fig. 8C,D). In contrast to the previous RNAi experiments [14], loss of NimC1 affected neither the phagocytosis of S. aureus nor that of E. coli (Fig. 8A-D). However, haemocytes derived from NimC1;eater mutant larvae were not only severely impaired in the phagocytosis of S. aureus (Fig. 8A,B), as expected, but also of E. coli (Fig. 8C,D). This indicates that Eater and NimC1 contribute redundantly to the phagocytosis of Gramnegative bacteria, as the presence of Eater or NimC1 is able to compensate for the absence of the other. The use of a double mutant also revealed a contribution of NimC1 to the phagocytosis of S. aureus, although Eater plays the predominant role. To further confirm these phagocytosis defects, we extended the analysis to two additional Gram-positive (Staphylococcus epidermidis, M. luteus) (Fig. 8E,F) and one Gram-negative (Serratia marcescens) (Fig. 8G) bacteria. Phagocytosis of all those microbes was not impaired in NimC1 1 null haemocytes. However, NimC1 1 ;eater 1 double mutant haemocytes showed a strongly reduced phagocytosis for both the Grampositive bacteria S. epidermidis and M. luteus (Fig. 8E,F), and the Gram-negative bacterium S. marcescens (although statistically nonsignificant due to the high variability of the wild-type) (Fig. 8G). Those data further confirmed our initial findings (Fig. 8A-D). Interestingly, NimC1 null haemocytes showed a higher phagocytic index, when compared to wild-type, for S. epidermidis and M. luteus bacteria. We hypothesized that the absence of NimC1 could trigger a compensatory pathway in plasmatocytes, specific for certain bacteria, in order to fulfil NimC1 phagocytic functions. The signalling of this putative compensatory pathway, that would eventually finally lead to a higher bacteria uptake by plasmatocytes, might be dependent on Eater, given the dramatically reduced phagocytic ability of the double mutant.

Eater and NimC1 receptors play a critical role in adhesion to bacteria
To better understand the cause of eater 1 and NimC1 1 ; eater 1 phagocytosis defects, and thereby to elucidate the unique role of these receptors in bacterial uptake, we performed scanning and transmission (TEM) electron microscopy experiments. Both these techniques allow following the different membrane-driven events during the phagocytosis process. Haemocytes from the corresponding genotypes were incubated with either E. coli or S. aureus live bacteria for 30 min to evaluate bacterial adhesion by SEM, and to follow bacterial uptake at 60 min by TEM. In wild-type and NimC1 1 haemocytes incubated with S. aureus, we observed plasma membrane remodelling, with the formation of a phagocytic cup and pseudopod protrusions that progressively surrounded bacteria, finally leading to their engulfment (Fig. 9A,B white arrowheads). Similar observations were made for wild-type, NimC1 1 and eater 1 haemocytes incubated with E. coli bacteria (Fig. 9C,D). Surprisingly, upon incubation of eater 1 and NimC1 1 ;eater 1 haemocytes with S. aureus, no bacteria were present on the cell surface (Fig. 9A). A decreased level of bacteria adherence was also observed in NimC1 1 ;eater 1 haemocytes incubated with E. coli (Fig. 9C). In accordance with SEM experiments, transmitted electron micrographs showed numerous engulfment events in wild-type and NimC1 1 haemocytes with S. aureus bacteria (Fig. 9B, arrows), as well as for E. coli in wild-type, NimC1 1 and eater 1 haemocytes (Fig. 9D). Altogether, these experiments point to the importance of Eater in binding Gram-positive bacteria, which is consistent with a previous report [35], but also to a redundant role of NimC1 and Eater in binding Gram-negative bacteria. Furthermore, they suggest that these two receptors do not play any critical role in bacteria internalization, as NimC1;eater mutant showed (rare) engulfment events (Fig. 9B,D arrows).
To further confirm the bacteria adhesion defects, we incubated haemocytes and live fluorescent bacteria either on ice or with Cytochalasin D. Both treatments inhibit the engulfment process, without altering the binding of the bacteria to the phagocytic cell [6]. In both conditions (Fig. 9E,F), we observed less bacteria binding to plasmatocytes in the same genotypes that were defective for phagocytosis in our ex vivo assays (eater 1 for S. aureus, and NimC1 1 ;eater 1 for S. aureus and E. coli, Fig. 8A-D).

Phagocytosis of latex beads and zymosan yeast particles is impaired in NimC1 null mutants
To further understand the role of Eater and NimC1 in the phagocytosis process, we proceeded to analyse the uptake of 'neutral' latex beads particles. We also tested their role in the phagocytosis of zymosan, a compound found on the cell wall of yeast. While bacteria present at their surface-specific targets for the engulfing receptors, latex beads can be seen as nonimmunogenic particles, that do not bear any ligands for the phagocyte. We observed that the phagocytic index of latex beads was wild-type like in eater null plasmatocytes. Interestingly, plasmatocytes lacking the NimC1 receptor showed a significantly reduced ability to engulf latex beads, as well as zymosan yeast particles (Fig. 10A,B). Thus, we could uncover a phagocytic defect in the NimC1 single mutant only when using particles that do not display bacterial motifs, suggesting that bacteria can bypass NimC1, probably by recruiting other phagocytic receptors, such as Eater.
Bacteria adhesion and latex beads engulfment are not impaired in croquemort and draper mutant haemocytes The drastic effect observed with the NimC1 1 ;eater 1 double mutant on phagocytosis and bacteria adhesion led us to explore the contribution of other previously characterized phagocytic receptors using the same assays. Draper and Croquemort are two transmembrane receptors expressed by plasmatocytes, and belong to the Nimrod and CD36 family respectively [14,36]. With SIMU (NimC4), they both play a key role in the engulfment of apoptotic bodies [18,[36][37][38][39]. Moreover, a role in S. aureus phagocytosis has also been described for Draper and Croquemort, as well as in E. coli phagocytosis for Draper [17,[40][41][42]. Although we did observe a modest decrease in S. aureus phagocytosis in croquemort and draper mutants (called crq Δ and drpr Δ5 respectively), and E. coli in drpr Δ5 (Fig. 11A,B), bacteria adhesion to the haemocytes was not impaired in these mutants (Fig. 11C). This further supports a specific role of Eater as the main tethering receptor in Gram-positive bacteria phagocytosis. Moreover, drpr Δ5 and crq Δ haemocytes showed a wild-type like engulfment of latex beads (Fig. 11D), further indicating a specific role of Eater in microbe uptake, likely via the recognition of a key bacterial surface determinant.

Discussion and Conclusions
NimC1 was initially identified as an antigen for the plasmatocyte-specific monoclonal antibody P1. It belongs to the Nimrod gene family that has been implicated in the cellular innate immune response in Drosophila [43,44]. A previous study pointed to the importance of NimC1 in the phagocytosis of bacteria, since RNAi-mediated silencing of this gene resulted in decreased S. aureus uptake by plasmatocytes [14]. In the present work, we further re-evaluated the function   of the NimC1 protein by using a novel null mutant, revealing its precise role in haemocyte adhesion, proliferation and phagocytic ability. By performing ex vivo spreading assays, we observed that the cell area of adherent NimC1 null haemocytes was reduced compared to wild-type control, suggesting that NimC1 works as an adhesion molecule. Consistent with this observation, SEM on spread haemocytes of NimC1 1 mutants revealed a defect in lamellipodia extension. Spreading defects were also observed in eater 1 [16] and NimC1 1 ;eater 1 haemocytes. Thus, two structurally related Nimrod receptors, NimC1 and Eater, are involved in lamellipodia extension and haemocyte adhesion. It will be interesting to analyse, in future work, the implications of NimC1 and Eater in haemocyte migration during metamorphosis or wound healing. Our results also indicate that Eater and NimC1 additively regulate haemocyte adherence. In contrast to eater-deficient larvae, NimC1 is, however, not directly required for plasmatocyte sessility in vivo. Whether NimC1 contributes to haemocyte sessility through additional scaffold proteins has to be further investigated, even though the present evidence might favour a model where Eater is the only essential protein required for haemocyte sessility [16].
During larval development, the peripheral haemocyte population undergoes a significant proliferation, expanding by self-renewal [8,21]. Moreover, during these developmental stages, plasmatocytes are characterized by a dynamic behaviour, continuously exchanging between the sessile and circulating state. In 2011, Makhijani et al. [21] provided evidence that plasmatocyte proliferation rate is higher in the haematopoietic pockets, where haemocytes cluster on the lateral side of the larval body. At this location, sessile plasmatocytes are in contact with the endings of peripheral neurons, which are thought to provide a trophic environment to the blood cells. More recently, it has been shown that sensory neurons of the peripheral nervous system produce Activin-b, which turned out to be  an important factor in the regulation of haemocyte proliferation and adhesion [45]. By analysing the total number of haemocytes in third instar NimC1 1 or eater 1 larvae, we observed that both Eater and NimC1 negatively regulate haemocyte counts in an additive manner. EdU incorporation experiments revealed that the higher haemocyte counts in NimC1 1 ;eater 1 mutants were a consequence of an increased haemocyte proliferation rate. It is tempting to speculate that the higher proliferation rate is a secondary consequence of an adhesion defect. Indeed, adherent cells, notably when establishing contacts with other cells, are less proliferative, a process called 'contact inhibition of proliferation' [46]. Future studies should address how Eater and NimC1 contribute to both adhesion and proliferation, and the direction of causality between these two processes remains to be disentangled. Like plasmatocytes, crystal cells increase in number during larval stages. However, crystal cell proliferation is not due to a self-renewal mechanism because mature crystal cells do not divide. Instead, a recent study has shown that new crystal cells originate from transdifferentiation of sessile plasmatocytes via a Notch-Serratedependent process [26]. In the present study, we show that the NimC1 deletion does not strongly impact crystal cell formation as both sessile and circulating crystal cell populations were only mildly affected in NimC1 null larvae. Moreover, NimC1 does not affect the ability to differentiate lamellocytes and to encapsulate parasitoid wasp eggs. NimC1 was initially identified as a phagocytic receptor, mediating the uptake of S. aureus bacteria [14]. Contrary to this study, our ex vivo phagocytosis assays using the NimC1 deletion mutant revealed that the uptake of both Gram-positive and Gram-negative bacteria was not altered in NimC1 null haemocytes. We hypothesized that the RNAi approach could have targeted other phagocytic receptors, revealing a stronger phenotype not observed in the single null mutant. Strikingly, phagocytosis of both bacteria types was severely impaired in NimC1 1 ;eater 1 haemocytes, suggesting that both receptors contribute synergistically to phagocytosis of both Gram-negative and Gram-positive bacteria. At this stage, we cannot exclude that these receptors might indirectly regulate phagocytosis by controlling another receptor directly involved in bacterial recognition, although we judge this hypothesis unlikely. Consistent with our hypothesis, an RNAseq analysis of eater deficient versus wild-type haemocytes did not uncover any role of Eater in the regulation of other phagocytic receptors (data not shown).
Given the marked phagocytosis defect of the eater single mutant against S. aureus, the contribution of NimC1 was especially noticeable in the case of the Gram-negative bacterium E. coli. Our SEM approach revealed that NimC1 and Eater might contribute together to the early step of bacterial recognition, since NimC1 1 ;eater 1 double mutants showed decreased bacterial adhesion. The involvement of NimC1 in E. coli binding is consistent with previous in vitro work showing that native NimC1 binds bacteria [47]. Surprisingly, NimC1 1 and NimC1 1 ;eater 1 , but not eaterdeficient plasmatocytes, showed a significantly reduced ability to engulf latex beads and yeast zymosan particles. A recent study has also revealed a role of NimC1 in the phagocytosis of latex beads using an in vivo RNAi approach [48]. Thus, Eater and NimC1 have specific properties with regard to phagocytosis. It is interesting to address a parallel with the implication of two Nimrod receptors in bacteria tethering and docking, as shown for apoptotic cells clearance in Drosophila melanogaster [18,49]. Tethering receptors usually lack an intracellular domain and are involved in the binding to the dying cell. Docking receptors, instead, are subsequently required to activate intracellular signalling and mediate the internalization and degradation of the particle. In the fruit fly, a good example for tethering and docking receptors are SIMU/NimC4 and Draper respectively [49]. A similar dichotomy exists in vertebrates, as Stabilin 2 and TIM-4 are classified as tethering receptors, whereas the integrins aVb3 and aVb5 are grouped as docking/signalling proteins [50,51]. The involvement and cooperation of two receptors of the Nimrod family in bacterial phagocytosis raised the possibility that they might contribute via different mechanisms: binding and internalization. Our current hypothesis is that Eater might work as the main tethering receptor, required for binding to specific motifs present on the bacterial surface. Moreover, given the wild-type engulfment of latex beads in eater, this receptor might be engaged specifically for phagocytosis of microbes. Indeed, the involvement of Eater in bacterial binding was already assessed in previous studies [35], and is consistent with our assays using live fluorescent bacteria and SEM experiments. On the contrary, NimC1 could function in the activation of the subsequent intracellular signalling, maybe as a subunit of a bigger macromolecular complex. We hypothesize that in the presence of cell wall bacterial determinants, such as peptidoglycan, lipopolysaccharide or teichoic acids, microbe phagocytosis can bypass the requirement of NimC1 by providing enough 'eat me' signals to Eater. In contrast, the critical role of NimC1 in phagocytosis becomes visible with less immunogenic particles. This would explain why we do not observe any defects in the phagocytosis of S. aureus and E. coli in NimC1 single mutant, but only with latex beads (i.e. particles without any bacterial motifs).
Future studies should address how Eater and NimC1 interact, the implication of other possible phagocytic receptors and characterization of their respective ligands. Collectively, our genetic analysis using compound mutants identifies NimC1 and Eater as two critical receptors involved in the initial step of phagocytosis, and notably adhesion to bacteria. While a plethora of receptors have been identified for their role in microbial phagocytosis in Drosophila, NimC1 and Eater appear to be the best candidates to directly recognize bacterial 'eat me' signals initiating phagocytosis. Our study also provides a valuable tool to better assess the role of phagocytosis during the immune response.
Wild-type w 1118 (BL5905) flies were used as controls, unless indicated otherwise. The following fly lines were used in this study:

Gene targeting of NimC1
Gene targeting of NimC1 was performed as follows. The 5 0 and 3 0 homology arms, of 4.8 kb and 3.7 kb, respectively, were PCR amplified from genomic DNA. The 5 0 arm was inserted between NotI and NheI restriction sites, whereas the 3 0 arm was inserted between SpeI and AscI sites of the gene targeting vector pTV [Cherry]. A donor transgenic stock was generated by transformation of a starting w1118 (BL5905) stock, and used for hsFLP and hs-I-SceI-mediated gene targeting [19]. Using this method, we recorded a placed on ice in order to stop the reaction. The fluorescence of extracellular particles was quenched by adding 0.4% trypan blue (Sigma-Aldrich) diluted 1/3. Phagocytosis was quantified using a flow cytometer (BD Accuri C6) in order to measure the fraction of cells phagocytosing, and their fluorescent intensity. w 1118 larvae and HmlΔdsred.nls larvae with or without bacterial particles were used to define the gates for haemocytes and the thresholds for phagocytosed particle emission. The phagocytic index was calculated as follows: Fraction of haemocytes phagocytosingðfÞ The FITC-labelled bacteria were resuspended, centrifuged at 11 200 g, the pellet was resuspended to a final concentration of 10%, sodium azide was added as a preservative (0.1%) and the samples were kept at 4°C until use. Bacteria were washed 59 with PBS prior to the phagocytosis assay. The phagocytic activity of haemocytes was assayed with a protocol similar to [6]. Haemocytes were isolated from third instar larvae at room temperature into Shields and Sang M3 insect medium (Sigma-Aldrich) containing 5% FCS (Gibco, Thermo Fisher, Waltham, MA, USA) supplemented with 1-phenyl-2-thiourea (Sigma-Aldrich) to prevent melanization. A total of 2-3 9 10.5 haemocytes were incubated with 5-6 9 10 6 heatkilled, FITC-labelled bacteria at room temperature for 40 min in the wells of round bottomed microtitre plates (Gibco) in 100 lL. The fluorescence of extracellular bacteria was quenched by the addition of Trypan blue to the cells in 0.2% final concentration shortly before the actual measurement. The fluorescence intensity of phagocytosed FITC-labelled bacteria was analysed with a FACS Calibur equipment (BD Accuri C6, Beckton Dickinson). Phagocytic index was calculated as mentioned above. 3 Phagocytosis of green fluorescent 1 lm latex beads (Sigma-Aldrich) and AlexaFluor TM 488 Zymosan BioParticles TM (Invitrogen) was performed following the same procedure described in 1), with the exception that haemocytes without the HmlΔdsred.nls marker were used. 1 9 10 5 Zymosan BioParticles TM and 0.2 lg of latex beads were added to each sample.

Scanning electron microscopy
Samples for SEM were prepared as follows. Six wandering third instar larvae were bled into 50 lL of Schneider's insect medium (Sigma-Aldrich) containing 1 lM phenylthiourea (PTU; Sigma-Aldrich). The collected haemolymph was incubated on a glass coverslip for 20 min for spreading assay, or 30 min with bacteria for phagocytosis assay, before being fixed for 1 h with 1.25% glutaraldehyde in 0.1 M phosphate buffer, pH 7.4. Samples were then washed in cacodylate buffer (0.1 M, pH 7.4), fixed again in 0.2% osmium tetroxide in the same washing buffer and then dehydrated in graded alcohol series. Samples underwent critical point drying and Au/Pd coating (4 nm). Scanning electron micrographs were taken with a field emission scanning electron microscope Merlin, Zeiss NTS, Oerzen, Embsen, Germany.

Transmission electron microscopy
Third instar wandering larvae were bled in 50 lL of Schneider's insect medium (Sigma-Aldrich) containing 1 lM phenylthiourea (PTU; Sigma-Aldrich). The collected haemolymph was incubated with bacteria on a glass coverslip for 1 h before being fixed for 2 h with 2% paraformaldehyde + 2.5% glutaraldehyde in 0. Binding assay with live fluorescent bacteria Cytochalasin D treatment L3 wandering larvae were bled into 120 lL of Schneider's insect medium (Sigma-Aldrich) containing 1 lM phenylthiourea (PTU; Sigma-Aldrich). Haemocytes were allowed to adhere on the glass slide for 1 h before being treated for another 60 min with 1 lM of Cytochalasin D. After drug treatment, haemocytes were incubated directly on the slide with live fluorescent S. aureus or E. coli bacteria always in the presence of Cytochalasin D for 60 min. After fixation in 4% paraformaldehyde PBS, rhodamine phalloidin staining (Molecular Probes TM , Eugene, OR, USA) was performed. Finally, cells were stained with 1/15 000 dilution of 4 0 ,6-diamidino-2-phenylindole DAPI (Sigma-Aldrich) and mounted in antifading agent Citifluor AF1 (Citifluor Ltd.).

Phagocytosis inhibition by cold temperature
L3 wandering larvae were bled into cold Schneider's insect medium (Sigma-Aldrich) containing 1 lM phenylthiourea (PTU; Sigma-Aldrich) on a previously chilled glass slide. After larva bleeding, haemocytes and bacteria were incubated directly on the prechilled slide, in cold Schneider's medium, on ice for 60 min. Fixation and staining procedures were performed as described above.

Fluorescent bacteria
The E. coli GFP strain was obtained by transforming E. coli K12 with a synthetic sfGFP coding sequence cloned in a pBAD backbone (Gibco, ThermoFisher) by Gibson assembly. sfGFP induction was obtained by growing the bacteria in LB + 0.1% arabinose overnight prior to the binding assay. The S. aureus GFP strain is described in Ref. [57].
For cell area measurements, haemocytes were captured with a 209 objective on GFP, RFP and DAPI channels. Individual images were then loaded into a CellProfiler pipeline (www.cellprofiler.org). In order to define the cell area, cell nuclei were first detected using data from the DAPI channel. Cell limits were then defined by expanding the nuclei signal to the edges of the GFP channel. Cell areas were computed from this segmentation analysis, and cell area of 750 cells of each genotype was quantified.

Haemocytes visualization through larva cross sectioning
Third instar larvae of the indicated genotypes were fixed in 4% paraformaldehyde PBS for 48 h at 4°C. Afterwards, larvae were embedded using OCT medium in a Tissue-Tek cryomolds (Sakura, Alphen aan den Rijn, The Netherlands). Transverse sections of 4-5-lm thickness were cut using Leica CM1959 cryostat. Finally, sections were fixed again for 15 min in 4% paraformaldehyde PBS, prior to rhodamine phalloidin (Molecular Probes TM ) staining. Samples were imaged with an Axioplot Imager.Z1 Zeiss coupled to an AxioCam MRm camera (Zeiss).

Crystal cell counting methods
At least 10 third instar larvae were heated in 1 mL of phosphate-buffered saline (PBS) at 67°C for 20 min in Eppendorf tubes. For quantification analysis, black puncta were counted in the posterior-most segments A6, A7 and A8. Pictures were taken with a Leica DFC300FX camera (Leica Microsystems AG, Heerbrugg, Switzerland) and Leica Application Suite right after heating.
For quantification of crystal cells by flow cytometry, we crossed wild-type or mutant lzGal4>UAS-GFP flies with the corresponding HmlDdsred.nls w 1118 or mutant flies. Larvae form the resulting offspring were used to determine the number of crystal cells (lzGal4>UAS-GFP) and the ratio of crystal cells among the total haemocyte population (lzGal4>UAS-GFP / HmlDdsred.nls). Four larvae of each genotype were bled into 150 lL 19 PBS containing 1 lM phenylthiourea (PTU; Sigma-Aldrich) and 0.1% paraformaldehyde to block crystal cell rupture. Seventy-five microlitres of the haemocyte suspension was analysed by flow cytometry. Haemocytes were first selected from debris by plotting FSC-A against SSC-A on a logarithmic scale in a dot plot. Cells were then gated for singlets by plotting FSC-H versus FSC-A. FL1 and FL2 detectors were used for lzGal4>UAS-GFP and HmlΔdsred.nls events respectively.

Wounding experiment
Wandering third instar larvae were pricked dorsally near the posterior end of the animal, using a sterile needle (di-ameter~5 lm). Pictures of melanised larvae were taken 20 min after pricking, with a Leica DFC300FX camera and Leica Application Suite.

Wasp infestation and quantification of fly survival to Leptopilina boulardi infestation
For wasp infestation experiments, 30 synchronized second instar (L2) larvae were placed on a pea-sized mound of fly food within a custom-built wasp trap in the presence of three female L. boulardi (strain NS1c, described in Ref. [58]) for 2 h. Quantification of fly survival was performed as follows. Parasitized larvae were kept at room temperature and scored daily for flies or wasps emergence. The number of eclosed flies and wasps was subtracted of the initial number of exposed larvae and set as dead larvae/pupae. Pictures of melanised eggs were taken with a Leica DFC300FX camera and Leica Application Suite.

Infection experiments and qRT-PCR
Systemic infections (septic injuries) were performed by pricking third instar larvae dorsally near the posterior end of the animal using a thin needle previously dipped into a concentrated pellet (OD 600~2 00) of bacteria. After septic injury, larvae were incubated at 29°C. After 4 h, the animals were collected, and total RNA extraction was performed using TRIzol reagent (Invitrogen). RNA quality and quantity were determined using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Inc., Wilmington, DE, USA) and 500 ng of total RNA was used to generate cDNA using SuperScript II (Invitrogen). Quantitative PCR was performed on cDNA samples using the LightCycler 480 SYBR Green Master Mix (Roche, Basel, Switzerland). Expression values were normalized to RpL32.

Statistical analysis
Experiments were repeated at least three times independently and values are represented as the mean AE standard deviation (SD). Data were analysed using GRAPHPAD PRISM 7.0 (San Diego, CA, USA). P-values were determined with Mann-Whitney tests, unless indicated otherwise. For phagocytic index measurement experiments, data successfully passed a Shapiro-Wilk normality test (a = 0.05, n = 9), so that we could assume that samples follow Gaussian distribution. Therefore, significance tests were performed using Students t test.