Drosophila blood cells


For correspondence. E-mail M.Meister@ibmc.u-strasbg.fr; Tel. (+33) 03 88 41 70 32; Fax (+33) 03 88 60 69 22.


Drosophila blood cells or haemocytes belong to three lineages: plasmatocytes, crystal cells and lamellocytes. There is no equivalent of a lymphoid lineage in insects which have no adaptive immunity. Haematopoiesis is under the control of a number of transcription factors and signalling pathways (such as GATA factors, JAK/STAT or Notch pathways) most of which have homologues which participate in the control of mammalian haematopoiesis. Drosophila plasmatocytes are professional phagocytes reminiscent of the cells from the mammalian monocyte/macrophage lineage. Several receptors responsible for recognition of microorganisms or apoptotic corpses have been identified, which include a Scavenger Receptor, a CD36 homologue and a peptidoglycan recognition protein. Crystal cells contain the enzymes necessary for humoral melanization that accompanies a number of immune reactions. The production of melanin generates, as by-products, cytotoxic free radicals that are believed to participate in the killing of pathogens. Finally, lamellocytes represent a cell type that specifically differentiates after parasitism of Drosophila larvae and forms a capsule around the invader. Encapsulation together with melanization eventually kill the parasite within the capsule.


Innate immunity is phylogenetically the oldest defence system in the animal kingdom. It evolved before adaptive immunity which has arisen only recently in evolution and is the hallmark of vertebrate immunity. Like all invertebrates, insects defend themselves through innate immune mechanisms. In this context, Drosophila has been the model of choice in the two last decades, to unravel the molecular mechanisms of non-adaptive host defence. Drosophila immunity includes two complementary facets which are humoral and cellular immunity. Most of the recent progress relates to Drosophila humoral immunity. Infection by bacteria or fungi triggers the rapid and massive production in the fat body, which is the liver counterpart in insects, of a number of antimicrobial peptides and other effectors. Two major signalling pathways control this process which are called the Toll and the Imd pathways (reviewed in Hoffmann and Reichhart, 2002). The intracellular components of both pathways, as well as the extracellular cascades that lead to their activation, are currently under intense investigation in various laboratories. Opposed to this, cellular immunity, the second facet of Drosophila host defence, has only recently attracted renewed interest. We would like in this review to give an update of our current knowledge on haematopoiesis in Drosophila, and on the functions of the various blood cell or haemocyte types.

I.Haematopoiesis in Drosophila

Two phases of haematopoiesis

Drosophila haematopoiesis occurs in two phases during development, and gives rise to three haemocyte lineages that share characteristics with mammalian myeloid lineages. Drosophila, like all arthropods, has no equivalent of a lymphoid lineage. A first haematopoietic wave takes place in the second half of embryogenesis when a population of haemocytes originate in the procephalic mesoderm and migrate to colonize the whole embryo along invariant paths (Tepass et al., 1994). These cells, called plasmatocytes, act as macrophages as they eliminate apoptotic cells (Franc et al., 1996; 1999). A second population differentiates simultaneously nearby the anterior region of the gut where it remains localized around the proventriculus (Lebestky et al., 2000). These cells are crystal cells (see below) and their role in the embryo is unknown. Towards the end of embryogenesis, the precursors of the lymph glands form in the lateral mesoderm, and migrate dorsally to prefigure the first paired lobes of the organ (Rugendorff et al., 1994). The lymph glands are the main site of haematopoiesis during larval stages (Shrestha and Gateff, 1982; Rizki and Rizki, 1984; Lanot et al., 2001). They are composed of a variable number of paired lobes that are located along the dorsal vessel (Fig. 1). In the posterior lobes they contain essentially undifferentiated precursor cells called prohaemocytes. More anteriorly they mostly contain fully differentiated haemocytes which are released into the circulation. Circulating haemocytes (Fig. 1) comprise a majority of plasmatocytes which are the dedicated phagocytes, and a small proportion (<5%) of crystal cells. Crystal cells contain crystalline inclusions that correspond to enzymes necessary for humoral melanization (see below). An additional cell type is found in the anteriormost lobes of the lymph glands that displays features of intense protein synthesis. These ‘secretory cells’ are never seen in circulation: they probably correspond to a variation on the plasmatocyte theme as intermediate forms are often found in the lymph glands. Finally, a third independent haemocyte lineage exists in larvae, but is seldom observed in healthy animals. These cells called lamellocytes differentiate massively in the lymph glands after parasitization and are large flat cells devoted to encapsulation of invaders too large to be phagocytosed by plasmatocytes (Fig. 1). A commonly encountered immune threat for Diptera (flies) is parasitism by Hymenoptera (wasps) species that lay their eggs in Diptera larvae (Carton et al., 1986). The cellular reaction that fends off parasites thus involves the production of a specific, inducible haemocyte lineage.

Figure 1.

Drosophila haemocytes.
A. Scanning electron microscopy of a larval haematopoietic organ. The lymph glands are organized as paired lobes along the dorsal vessel (arrow). Anterior is to the left.
B. Circulating haemocytes stained with DAPI. Plasmatocytes are small round cells whereas lamellocytes (arrows) are larger and flattened. C. Phase-contrast microscopy of circulating crystal cells (arrows) with conspicuous crystalline inclusions.
D. Confocal image of phagocytosis of E. coli (K12 Alexa Fluor 594) by GFP-labelled plasmatocytes from a tep1-GFP larva.
E–G. Encapsulation of parasitoid wasp eggs/larvae. A wasp larva is surrounded by lamellocytes (blue nuclei in E); the cell layer becomes thicker while melanization of the parasite is initiated (arrowhead in F); eventually the wasp egg/larva is totally melanized within the lamellocyte capsule (G). Lamellocytes are visualized with a lacZ enhancer trap marker in E and F.
H–J. Transmission electron microscopy of a plasmatocyte (H), a crystal cell (I) and lamellocytes (J). Bars: 50 µm (A, E–G), 20 µm (B), 10 µm (C, D), 2 µm (H–J).

At the onset of metamorphosis, the lymph glands release large numbers of active phagocytes that are called pupal macrophages and play a crucial role in tissue remodelling as they phagocyte cells of doomed larval structures. In this process they also eat up the remainder of the lymph glands, and at later stages no haematopoietic organ can be found (Lanot et al., 2001). This modification of the properties of plasmatocytes and the simultaneous dispersal of the lymph glands is under the control of the steroid hormone ecdysone (Lanot et al., 2001; Sorrentino et al., 2002) which orchestrates all tissular modifications related to metamorphosis. In Drosophila adults, the only haemocyte type that is present is the plasmatocyte, and the pool of adult haemocytes likely derives from larval plasmatocytes. Given their morphological and functional features, it is generally considered that Drosophila plasmatocytes resemble the mammalian monocyte/macrophage lineage.

Transcriptional control of haematopoiesis

Our understanding of regulation of haemocyte proliferation and specification comes from genetic analysis of mutants (Fig. 2). Control of haemocyte proliferation was investigated at the larval stage and it was recently shown that the unique Drosophila homologue of PDGFR/VEGFR, named PVR, together with one of its three putative ligands PVF2, plays a crucial role in the control of prohaemocyte proliferation (Munier et al., 2002). This finding was essentially demonstrated by the dramatic effect of PVF2 overexpression in larvae, which resulted in a 300-fold increase in blood cell counts, due to excessive proliferation of prohaemocytes at the expense of differentiation. A comparable (40-fold) effect was obtained by pan-haematopoietic overexpression of an activated form of Ras, and this effect is mediated by the Raf/MAPK pathway (Asha et al., 2003). It is however, not documented yet whether the proliferation signal produced by PVF2/PVR is transmitted via the Ras/Rak/MAPK pathway. This proliferation effect of PVF2/PVR observed at larval stages does not apply to the embryonic stage where PVF/PVR were rather proposed to establish a guidance system responsible for the migration of the plasmatocytes throughout the embryo (Cho et al., 2002).

Figure 2.

Regulation of Drosophila haematopoiesis. Blood cell identity is conferred by the GATA factor Serpent (srp). Proliferation is controlled by the PVF2/PVR, the Ras/Raf, the Toll and the JAK/STAT pathways. Crystal cell specification is under the control of the Notch (N) pathway and the Runx1 homologue Lozenge (lz) transactivator, and antagonized by the Friend-of-GATA homologue U-shaped (ush). Plasmatocytes are specified by two Glial-cells-missing (gcm) transactivators, then further differentiate into pupal macrophages under the impulse of the steroid hormone ecdysone (Ecd). Lamellocyte specification requires wild-type functions of Notch and of the JAK/STAT pathway.

Early data have also implicated the JAK/STAT and the Toll pathways in the control of circulating haemocyte numbers as gain-of-function mutations of JAK and of Toll result in increased blood cell counts (Luo et al., 1995; Lemaitre et al., 1995), whereas loss-of-function mutations in the Toll pathway have the opposite effect (Qiu et al., 1998).

Haemocyte identity in embryos is specified by the GATA factor Serpent (Srp; Rehorn et al., 1996; Waltzer et al., 2002). A similar function for srp is likely to occur at larval stages as its expression is recorded before all differentiation markers in prohaemocytes (Lebestky et al., 2000). It is proposed that srp might be required both for the specification of haemocyte primordium within the mesoderm at an early embryonic stage, and later for gene expression during their maturation. Several transcription factors and signalling pathways have been demonstrated to govern lineage specification in Drosophila haematopoiesis. Two Zinc-finger transactivators, Gcm (Glial Cells Missing, Bernardoni et al., 1997; Lebestky et al., 2000) and Gcm2 (Alfonso and Jones, 2002) are required for plasmatocyte fate in the embryo, and this has also been documented for Gcm at the larval stage. The crystal cell fate is instructed in larvae by the Serrate/Notch pathway (Duvic et al., 2002; Lebestky et al., 2003) and the Runx1-related Lozenge transcrition factor (Lebestky et al., 2000). The Friend-of-GATA homologue U-shaped antagonises crystal cell development (Fossett et al., 2001). Transcription factors that determine lamellocyte development are not yet identified. Mutations in three genes have long been known to stimulate lamellocyte production: these are Toll and JAK dominant gain-of-function, and cactus loss-of-function mutations (Lemaitre et al., 1995; Luo et al., 1995; Qiu et al., 1998). It is however, possible that the effect of these mutations on haemocyte differentiation is indirect.

It is striking that most of the players in Drosophila haematopoiesis identified so far are counterparts of gene products that are also central to mammalian haematopoiesis, such as GATA factors, the Notch pathway, Runx1 or the JAK/STAT pathway to cite but a few examples. It appears that the same building blocks are used with different combinations to control haematopoiesis in invertebrates and vertebrates.

II. Haemocyte functions

Drosophila that are deprived of haemocytes are clearly sensitized to infection. This was shown in larvae carrying a mutation called domino which results in extremely reduced haemocyte counts. When this mutation is combined with mutations affecting humoral immunity (imd) or melanization, resistance of larvae to infection is strongly decreased compared to single mutants (Braun et al., 1998). Similarly, inactivation of adult haemocytes by massive injection of polystyrene beads reduced the ability of imd flies to fight bacterial infection (Elrod-Erickson et al., 2000). Haemocytes are responsible for a number of immune functions in Drosophila, among which phagocytosis, encapsulation and melanization have been documented. We have outlined the current data on these immune reactions below.


In vertebrates, the elimination of microorganisms and the removal of apoptotic cells are achieved by the same phagocytes. However, whereas phagocytosis of non-self particles induces inflammation, that of apoptotic cells either downregulates or suppresses inflammation (Vandivier et al., 2002). Phagocytosis occurs in several steps. The first consists in the attachment of the phagocyte to the targeted particle followed by alteration of the cytoskeleton and internalization. The engulfed target is then destroyed within phagosomes by lysosomal enzymes, reactive oxygen species and nitric oxide (Jones et al., 1999). The latter phases are likely conserved between invertebrates and vertebrates, involving namely actin-remodelling and vesicule trafficking (Rämet et al., 2002a).

It has been proposed that phagocytosis of both apoptotic cells and microorganisms requires two processes: tethering, followed by actin-dependent engulfment (reviews in Aderem and Underhill, 1999; Fadok and Chimini, 2001; Hoffmann et al., 2001; Greenberg and Grinstein, 2002). Tethering of the target to the phagocyte is achieved by receptors that will either directly recognize determinants on microorganisms or apoptotic cells, or bind through opsonizing molecules. Most of these receptors have a broad recognition spectrum for both endogenous and exogenous molecules, and may participate in the phagocytosis of both types of targets. Mammalian examples are the Scavenger Receptor family (qualified as ‘molecular fly papers’ by Krieger et al., 1993). Tethering favours a tight association between the target and the phagocyte, and clustering of the binding proteins then mediates phagocytosis. The release of the signal triggering the appropriate on/off switch of inflammation seems to be dependent on the engulfment receptor that is simultaneously recruited. A few receptors are able to initiate engulfment, and they also participate in the recognition of the target. For ingestion of microorganisms, well studied receptors are the Mannose Receptor, that binds sugar moieties on microorganisms, two proteins of the integrin family, CR3 (αMβ2) and CR4 (αXβ2) that are receptors for the opsonizing complement factor C3bi, and the Fcγ receptors that recognize antibodies. Phagocytosis of apoptotic cells is dependent on the recognition of phosphatidylserine by its receptor (PS-R), but can also be mediated by two integrins αVβ3 and αVβ5, the tyrosine kinase MER and CD91 (Henson et al., 2001; Huynh et al., 2002).

In Drosophila, plasmatocytes are responsible for the disposal of both microorganisms (Fig. 1) and apoptotic cells. The mechanisms by which they recognize and engulf their targets are poorly understood. Phagocytosis of apoptotic cells at the embryonic stage requires wild-type function of the croquemort gene, which encodes a CD36 homologue (Franc et al., 1996, 1999). In mammals CD36 is a class B Scavenger Receptor which acts in concert with the vitronectin receptor (αVβ3) and PS-R to engulf apoptotic corpses (Fadok et al., 1998). A homologue of PS-R has also been described in Drosophila but no functional studies have yet demonstrated a role for this receptor in apoptotic cell removal (Henson et al., 2001). A scavenger receptor, dSR-CI with similar broad ligand recognition as mammalian class A Scavenger Receptor was identified by Pearson et al. (1995). dSR-CI can mediate binding to both Gram-negative and Gram-positive bacteria when transfected into CHO cells and participates to some extent to the binding of these microorganisms to S2 cells, a primary Drosophila blood cell culture (Rämet et al., 2001). PGRP-LC, a Peptidoglycan Recognition Protein, mediates phagocytosis of Gram-negative, but not Gram-positive bacteria by S2 cells (Rämet et al., 2002a). Interestingly, PGRP-LC belongs to a family of 12 members in Drosophila (Werner et al., 2000). This putative membrane bound protein is involved in the induction of the fat body Imd transduction pathway, leading to the production of antibacterial peptides after an infection with Gram-negative bacteria (Gottar et al., 2002). Similarly, PGRP-SA which is a soluble PGRP, is required for the proper induction of the Toll pathway after infection with Gram-positive bacteria (Michel et al., 2001). Gottar et al. (2002) have suggested that PGRPs are sensors which define the type of microbial infection and drive the proper signalling pathway. An additional family of recognition proteins found in insects are the Gram-negative bacteria-binding proteins or GNBPs which all contain gluconase-like domains. Three cDNAs encoding GNBPs have been cloned from a Drosophila blood cell line library. Although their role in phagocytosis has not been demonstrated, it was shown that one GNBP (DGNBP-1) can bind LPS and β-1,3-glucan (Kim et al., 2000). Although there is good evidence that Croquemort, dSC-RI or PGRP-LC can mediate phagocytosis of target particles, there is no demonstration that they can by themselves trigger internalization without assistance from co-receptors.

The Drosophila genome harbours six genes encoding thiolester-containing proteins or TEPs (Lagueux et al., 2000). These proteins are related to the α2macroglobulin/complement factor C3 family as they share the 12 signature domains characteristic of these proteins, which comprise the well conserved thiolester motif region. However, TEPs clearly form a new group in this family due to a distinct cystein arrangement in their C-terminal region. All putative TEPs possess a signal peptide indicating that they are secreted proteins. The transcription of three out of six tep genes (tep1, 2 and 4) is upregulated after an immune challenge and tep1 and tep4 are mainly expressed in haemocytes (Lagueux et al., 2000; and M. Lagueux, unpubl. obs.). We have proposed that TEPs function during an immune response either as opsonins to promote phagocytosis, in a C3-like manner, or as protease inhibitors, in an α2-macroglobulin manner. The best evidence so far, in favour of the opsonin hypothesis, comes from studies in an other insect, Anopheles gambiae. Levashina et al. (2001) found similar TEP proteins in this model insect and could show that phagocytosis of Gram-negative bacteria in an Anopheles blood cell line is strongly reduced when transcription of the atep1 gene is impaired.

It is likely that plasmatocytes also send signals to other immunocompetent tissues. When larvae are infected per os with the phytopathogenic Erwinia carotovora, they respond by the activation in the fat body of the Imd pathway with subsequent production of antibacterial peptides (Basset et al., 2000). In mutant larvae with a reduced number of haemocytes, such as domino larvae for instance, the Imd pathway is not activated in response to infection. This indicates that haemocytes act as messengers that somehow convey information to the fat body in the case of bacterial infection. However, the nature of the signal is still unknown.


Some 50 hymenopteran species are reported parasites of Drosophila (Carton et al., 1986). The wasp females lay their eggs in the haemocoel of young larvae. This foreign body is detected by plasmatocytes which are the first line immune supervisors in circulation. They readily attach to the chorion of the egg (Russo et al., 1996) and a few hours later a strong cellular reaction is observed in the haematopoietic organ with enhanced proliferation, increase in crystal cell numbers (Sorrentino et al., 2002), and massive differentiation of lamellocytes (Lanot et al., 2001). Lamellocytes then form a multilayered capsule around the invader, which is ultimately accompanied by blackening due to melanization (Fig. 1). Within the capsule, the parasite is eventually killed, by asphyxiation or by the local production of cytotoxic free radicals, quinones or semiquinones (Nappi et al., 1995; 2000). The whole process of parasite encapsulation raises several intriguing questions. The first relates to the signal that is produced by plasmatocytes once they have recognized an invader which they cannot phagocyte. This signal is perceived by prohaemocytes in the lymph glands which then differentiate into an adapted haemocyte lineage. The identity of the signal has yet to be determined. Thus lamellocytes, to some extent, exhibit adaptive characteristics as they only differentiate in response to a specific immune challenge. They form several layers of cells around the parasite, to which they are attracted either by the previously attached plasmatocytes, or by wasp determinants. The nature of this mechanism is also unknown.


Humoral melanization in arthropods produces black pigment as a result of the activation of a biochemical pathway that converts tyrosine to melanin (reviews in Ashida and Brey, 1997; Söderhäll and Cerenius, 1998). Melanization is controlled by a cascade of serine proteases that ultimately cleave the zymogen prophenoloxidase to its active form. Phenoloxidase then catalyses the oxidation of phenols to quinones, which polymerize non-enzymatically to form melanin. It was shown in lepidopterans that the cascade is activated by initial recognition of non-self molecular patterns such as β-1,3-glucans, peptidoglycan or LPS. The different elements of the cascade are not yet elucidated in Drosophila, but a key control serpin has been recently identified that restricts phenoloxidase activity to the site of injury or infection (De Gregorio et al., 2002; Ligoxygakis et al., 2002). Serpin-27 A regulates the melanization cascade through the specific inhibition of prophenoloxidase processing by the terminal serine protease. Three genes encoding prophenoloxidases [Fujimoto et al., 1995; Berkeley Drosophila Genome Project (BDGP)] are present in the Drosophila genome, and in larvae they are specifically expressed in crystal cells (M. Meister, unpubl. obs.), which, once activated, readily disrupt and deliver their content into the haemolymph where the enzymes can function. A number of intermediate compounds formed during melanin synthesis are cytotoxic (see above, killing of the parasite), it is thus important to strictly control the localization of the reaction. This is achieved not only by the regulatory serpins, but also by the fact that activated phenoloxidase shows a tendency to aggregate (Ashida and Brey, 1997).

Melanization was proposed to participate in the sealing of a wound before the more elaborate epithelial wound healing process (Lai-Fook, 1966) that takes several days. This is based namely on the observation that Drosophila melanization-deficient mutants show defects in wound healing with excessive bleeding and reduced survival (Rämet et al., 2002b). These observations could indicate a role for melanization in the coagulation process, or that the proteolytic cascades leading to melanization and to coagulation share components. Nothing is known to date on molecular mechanisms responsible for coagulation in Drosophila. It will thus be challenging to determine how melanization and coagulation are activated in this model, and how closely they are related.

Extracellular matrix production

Drosophila haemocytes are known to produce abundant extracellular matrix material. Indeed many proteins have been purified from the supernatant of Kc cell cultures which are of haematopoietic origin (review in Fessler et al., 1994; Nelsson et al., 1994; Goto et al., 2001). Among these, laminins, collagen IV, tenascinm, glutactin, peroxidasin and haemolectin were subsequently found to be strongly expressed in embryonic and larval haemocytes. However, the almost complete absence of haemocytes in domino larvae does not result in apparent structural tissue defects and so the significance of this synthesis is not yet clear. Some of these molecules exhibit peculiar features, namely peroxidasin which contains a putative peroxidase domain associated with several leucine-rich repeats similar to those found in the extracellular domain of Toll family members (Nelsson et al., 1994). Haemolectin displays similarity with the von Willebrand factor which in vertebrates plays a role in haemostasis (Goto et al., 2001). However, as no mutants for both these genes are available, their function has not been assessed to date.


The field of Drosophila immunity has witnessed significant development over the last few years, thereby fuelling our general understanding of innate immunity. Opposed to this, for a long time Drosophila has failed to be a leading model in the analysis of haematopoiesis and of blood cell functions. A number of recent studies have now demonstrated that it has a role to play in the understanding of proliferation, commitment and differentiation of haematopoietic precursors, thanks to its powerful genetics and to the increasing numbers of blood cell markers. It also should help to gain insight, in the future, into molecular mechanisms of functions that are the hallmarks of myeloid-type blood cells.


The authors thank Jules Hoffmann for continued support. Work in their laboratory is supported by CNRS, NIH grant 1PO1 AI44220-02 to A. Ezekowitz and J. Hoffmann, the French Ministère de l’Education Nationale, de la Recherche et de la Technologie, EntoMed, Exelixis and l’Association de la Recherche contre le Cancer.