Embryonic origin of hemocytes and their relationship to cell death in Drosophila

Programmed cell death, or apoptosis, is a widespread phenomenon occurring during normal development of all metazoans. (Glücksmann, 1950; Saunders, 1966; Wyllie et al, 1980; Ellis et al., 1991; Raff, 1992). Experimental evidence indicates that cell death can be triggered by both intrinsic and extrinsic cues. The best studied example for intrinsically (cellautomomously) controlled cell death is the nematode Caenorhabditis elegans. Here, an invariant set of cells was defined which already at the time of their birth are ‘doomed’ to degenerate (for review, see Ellis et al., 1991). In contrast, in both vertebrate and invertebrate systems, numerous extrinsic (non cell-autonomously acting) factors that trigger cell death have been identified. Glucocorticoids stimulate death of thymocytes (Wyllie, 1980); Muellerian Inhibiting Substance (MIS) is responsible for cell death of the Muellerian ducts in male development (e.g., Tran et al., 1977); low levels of ecdysteroids initiate histolysis of neurons and muscle fibres in metamorphosing insects (Schwartz and Truman, 1984). For both vertebrate and invertebrate systems, genes have been defined that control the cell death program (for recent review, see Ellis et al., 1991; Vaux, 1993).

Cells that have undergone apoptosis are engulfed by macrophages, although other cells (e.g., epidermis) are also able to engulf their apoptotic neighbors (e.g. Wolff and Ready, 1991, for the Drosophila eye). Beside their role as ‘scavengers’, macrophages could also be more actively involved in the initiation of cell death. It had been proposed a long time ago that during metamorphosis of holometabolous insects, phagocytic cells actively trigger histolysis of larval tissues (e.g., Perez, 1910). This hypothesis was refuted in view of the fact that, in both the nervous and the muscular system of insect pupae, apoptosis begins in the absence of macrophages (e.g., Crossley, 1964). In contrast, the recent finding of Lang and Bishop (1993) that killing macrophages in developing mice by expressing diphteria toxin prevents cell death in the pupillary membrane and hyaloid vasculature of the eye suggests that macrophages are actively involved in cell death in this system.

In this paper, we have addressed the question of how macrophages develop in the Drosophila embryo and what is their relationship towards cell death occurring in this organism. In insects, macrophages represent a subpopulation of blood cells, or hemocytes, which are contained within the hemolymph space. In adults and larvae, hemocytes are formed by specialized hemopoietic tissues associated with the aorta and heart (for review, see Hoffman et al., 1979). The embryonic origin of hemocytes has been studied only in a cursory manner for a number of different insect groups. Most authors agree that the median mesoderm, i.e., the narrow band of mesoderm overlying the ventral nerve cord, gives rise to the hemocytes (Mori, 1979). A similar origin was also stated for the Dipteran Dacus by Anderson (1963), although this author emphasizes that blood cells precursors derive predominantly from the mesoderm of the head and anterior part of the trunk. Using hemocyte-specific markers, we have followed the development of these cells and their differentiation into macrophages. Our findings show that hemocytes derive from the mesoderm of the head and migrate along several invariant paths throughout the embryo. In the wild type, the large majority of hemocytes develop into phagocytic macrophages; in mutations with an increased amount of cell death [knirps (kni; Nüsslein-Volhard and Wieschaus, 1980); stardust (sdt; Tepass and Knust, 1993); fork head (fkh; Weigel et al., 1989)], the number of hemocytes turning into macrophages approaches 100%. Analysis of these mutations shows further that the absolute number of hemocytes is not increased compared to wild type, indicating that cell death does not ‘induce’ the formation of hemocytes. Finally, in contrast to the above cited vertebrate system in which macrophages seem to be required for the initiation of cell death (Lang and Bishop, 1993), we show that in the Drosophila embryo cell death occurs in the absence of macrophages. In mutations lacking macrophages [Bicaudal D (BicD; Nüsslein-Volhard et al., 1982); twist (twi) snail (sna) double mutants (Simpson, 1983; Grau et al., 1984; D. Gullberg and L. I. F., unpublished observations); torso⁴⁰²¹ (tor⁴⁰²¹; Klinger et al., 1988)], cell death is abundant.

We used the strong twi allele twi^HH07 (Nüsslein-Volhard et al., 1984), the strong sna allele sna^4.26 (Lindsley and Zimm, 1992), the strong sdt allele sdt^EH (Eberl and Hilliker, 1988), the dominant torso allele tor⁴⁰²¹ (Klingler et al., 1988), the strong fkh allele fkh¹ (Jürgens et al., 1984), the BicD allele BicD⁷¹³⁴ (Wharton and Struhl, 1989) and Oregon R as wild-type stock. Flies were grown under standard conditions and crosses were performed at room temperature or at 25°C. Egg collections were made on yeasted apple juice agar plates. Embryonic stages are given according to Campos-Ortega and Hartenstein (1985).

The following markers were used in this study: the enhancer-trap line B4-2-27 (Bier et al., 1989; Hartenstein and Jan, 1992) to label pharynx myoblasts derived from the head mesoderm; the enhancer-trap line E8-2-18 (Bier et al., 1989; Hartenstein and Jan, 1992) to label hemocytes. These enhancer-trap lines express β-galactosidase that was detected with a polyclonal anti-β-galactosidase antibody (Cappel; dilution 1:2000). The polyclonal anti-peroxidasin (anti-X; Abrams et al., 1993; Nelson et al., 1994) was used to label the hemocytes in wild type and all examined mutants. This antibody was diluted 200-fold. Antibody stainings and sections of stained embryos were done as described previously (Tepass and Knust, 1993).

Embryos for examination in the transmission electron microscope were prepared as described previously (Tepass and Hartenstein, 1994). Embryos for semi-thin sectioning were prepared in the same way. 2 μm sections were cut on an LKB Ultrotom V and stained with a toluidine blue/methylene blue/borate solution.

The base analogue BrdU, which is incorporated into replicating DNA, was applied by permeabilizing staged, dechorionated embryos with octane (Sigma) for 3 minutes and spreading them on BrdU containing Grace medium (1 mg/ml). Embryos were allowed to develop at 25°C for specific times, then collected, and fixed for 30 minutes in a mixture of 4% formaldehyde in PEMS(0.10M Pipes, 2 mM MgSO₄, 1 mM EGTA, pH 7.0) with heptane. Next they were devitellinized in methanol. After several washes in PBT, embryos were incubated for 35 minutes in 2 N HCl to denature the DNA. After this step they were washed for 30 minutes in several changes of PBT. The preparations were then incubated for 1 hour in PBT+N, followed by an overnight incubation in a monoclonal antibody against BrdU (Beckton-Dickinson) at a dilution of 1:50. For further steps of antibody labelling see Ashburner (1989).

The hemocytes can be first identified approximately 2 hours after gastrulation (late stage 10) as a discrete subpopulation of mesoderm cells located in the head of the embryo. This part of the mesoderm, which we will call procephalic mesoderm, forms two vertical plates overlying the procephalic neuroblasts. To label the hemocytes, two markers, the PlacZ construct E8-2-18 (Bier et al., 1989; Fig. 1D) and an antibody against the extracellular matrix component peroxidasin (anti-X; Abrams et al., 1993; Nelson et al., 1994), were used. These markers recognize the large majority, if not all of, the hemocytes throughout development. Thus, based on observations of whole mounts and sectioned material, all cells that can be classified as hemocytes structurally (scattered cells of round or irregular shape located in the hemocoel) express E8-2-18 and peroxidasin.

BrdU incorporation experiments show that the procephalic mesoderm forms a separate mitotic domain, which undergoes four divisions (cell cycles 14-17) during embryonic stages 811 (data not shown). After the final division, the majority of procephalic mesoderm cells (approximately 300 on either side of the embryo) are recognizable as hemocytes. Only the anteriormost portion of the procephalic mesoderm (about 50 on either side) does not give rise to hemocytes. Instead, as revealed by another enhancer trap line (B4-2-27), these cells represent the population of myoblasts of the pharyngeal musculature (Fig. 1E,F). In addition to the procephalic mesoderm, small groups of mesoderm cells in the lateral and midventral part of the gnathal segments become hemocytes (Fig. 1A,C). However, we cannot entirely rule out that these cells originate in the procephalic mesoderm from where they migrate posteriorly and ventrally into the gnathal segments.

We believe that the procephalic and gnathal mesoderm represents the only source of hemocytes found in the late embryo. Both E8-2-18 and anti-peroxidasin label about 700 cells in the head of the late stage 11 embryo. This number remains constant throughout embryogenesis. During late embryogenesis, the fat body starts expressing peroxidasin (Fig. 2J), as well as E8-2-18 (data not shown), at a low level. However, cells of the fat body form a coherent tissue, which does not significantly change in cell number and, at least during the embryonic period, is not likely to contribute to the scattered population of hemocytes.

At the beginning of germ band retraction (early stage 12), the hemocytes start spreading throughout the embryo (Figs 2, 3). Moving anteriorly and ventrally, they populate the clypeolabrum and gnathal buds. Posteriorly directed migration brings them into the tail end of the germ band, which is folded over the anterior part of the germ band during this stage and abuds the head region. A substantial portion of hemocytes remain in the dorsal head region. Germ band retraction carries the group of hemocytes that previously have entered the tail end posteriorly. In the following stages (13-14), hemocytes migrate from surface of hemocytes is not covered with a basement both ends of the embryo towards its middle. During this membrane (Fig. 4E,F). migration hemocytes follow four different routes:

(1) mid-ventrally between the ventral epidermis and the ventral nerve cord;

(2) between the dorsal surface of the ventral nerve cord and the mesoderm;

(3) along the dorsal boundary of the epidermal primordium;

(4) along the gut primordium.

By late stage 14 most parts of the embryo are rather evenly macrophages at different times during development. Thus, populated with hemocytes (Figs 2, 3). Dense clusters can be whereas most of the hemocytes located around the nervous observed in the head, as well as around the foregut and hindgut.

In early stages of development, hemocytes ultrastructurally represent typical small, round mesoderm cells (Fig. 4A).

According to the nomenclature proposed by Gupta (1985), these cells should be designated as prohemocytes.

During their migration between stages 12 and 13, prohemocytes start differentiating, developing prominent endoplasmic reticula and forming long processes, which radiate out from the cell bodies in all directions (Fig. 4B,D). Cells that show these ultrastructural characteristics have been classified as plasmatocytes (Gupta, 1985). All hemocytes that we have examined in the late embryo show the characteristics of plasmatocytes. In the following, we will use the (more widely used) general term hemocyte for this cell type.

From stage 12 onward, many hemocytes, in particular those ones surrounding the brain and ventral nerve cord, show phagocytic activity. They contain one or more vacuoles filled with dark inclusions which represent shrunken cytoplasm and pycnotic nuclei of ingested cells that have undergone apoptotic cell death (Fig. 4B). Hemocytes containing ingested cells (or cell fragments) are thereby histologically defined as macrophages. We have analyzed the temporal sequence in which macrophages appear in whole mounts of anti-peroxidasin-stained embryos. During late stage 13, approximately half of the hemocytes show phagocytic activity (Fig. 3B). Most of these early macrophages are located around the CNS and in the head, where cell death is particularly abundant (Abrams et al., 1993; Younossi-Hartenstein et al., 1993). At stage 13 and 14, hemocytes located in the interior of the embryo around the esophagus, midgut and hindgut are still small and (at least on the light microscopic level) do not yet show the inclusions that would define them as macrophages. However, towards later stages, the fraction of hemocytes turned macrophages increases and, by the end of embryogenesis, around 90% of the anti-peroxidasin-labeled cells are macrophages (Fig. 3D).

From stage 13/14 onwards, hemocytes possess an extensive rough endoplasmatic reticulum that is often dilated into lacunae (Fig. 4D), indicating a high secretory activity that is typical for cells producing extracellular matrix (see Discussion). By contrast to other tissues in the late embryo and larva, the cell surface of hemocytes is not covered with a basement membrane (Fig. 4E,F).

Most of the dead cells in the embryo are ingested and degraded by hemocytes. Only occasionally it can be observed that dead cells have been phagocytosed by other cells (e.g. epidermal cells; Fig. 4C). As described above, hemocytes convert into macrophages at different times during development. Thus, whereas most of the hemocytes located around the nervous system and in the head have ingested cellular debris by midembryogenesis, few of the hemocytes that surround the gut seem to have done so. Even later (stage 17), a certain fraction of these interiorly located hemocytes do not contain cellular debris. This could indicate that they represent a separate class of hemocytes, which do not possess phagocytic activity; alter-natively, it could be the absence of cell death from the gut tissues that is responsible for the lower rate at which hemocytes located close to the gut convert into macrophages. According to this hypothesis, all hemocytes have the potential to phagocytose and thereby become macrophages once they come into contact with degenerating cells. Since cell death is indeed rare, natively, it could be the absence of cell death from the gut if not absent, in the embryonic gut (Abrams et al., 1993; own tissues that is responsible for the lower rate at which hemocytes observation), hemocytes surrounding this structure have no located close to the gut convert into macrophages. According opportunity to phagocytose. To test this hypothesis, we to this hypothesis, all hemocytes have the potential to phago-analyzed the pattern of macrophages in fkh mutant embryos, in cytose and thereby become macrophages once they come into which most of the gut epithelium degenerates during stages 12/13 (Weigel et al., 1989). In these embryos, all interior hemocytes do become macrophages at an early stage (Fig. 5). In addition, the massive, locally circumscript cell death in the gut seems to attract hemocytes, since the proportion of interior versus superficial macrophages is significantly higher in fkh than in wild type. We conclude that all of the hemocytes stained by the anti-peroxidasin-antibody form a homogenous population with respect to their ability to become macrophages.

Hemocytes were counted in embryos carrying a mutation in the sdt or kni genes. In these mutants, the amount of cell death is dramatically increased. In sdt, most epithelial cells deriving from the ectoderm die during an extended period of time (Tepass and Knust, 1993). In kni mutant embryos, most cells located in the kni domain of expression (Nauber et al., 1988) die (own observation). Despite the high amount of cell death in both mutants, the number of anti-peroxidasin-positive cells is not significantly increased (Fig. 6). We counted (per half embryo) approximatly 345 anti-peroxidasinpositive cells for sdt (n=3; 320, 341, 373) and 320 for kni (n=2; 309, 331) versus 365 for wild type (n=4; 372, 389, 368, 331). These figures indicate that the number of hemocytes does not adapt to an increased need; the additional cell death does not ‘induce’ extra mesoderm cells to form hemocytes. To accomodate the large amount of cellular debris, macrophages in both mutations grow to a larger average size. In addition, cellular debris also becomes located extracellularly in the hemolymph.

In embryos carrying the BicD mutation, head (including procephalic and gnathal mesoderm), thorax and anterior abdomen are absent. Instead, a second posterior abdomen with reversed polarity differentiates (Nüsslein-Volhard et al., 1982). We labeled BicD mutant embryos with anti-peroxidasin antibody to investigate whether, in the absence of the head mesoderm, cells of the trunk are able to give rise to hemocytes. In BicD mutant embryos hemocytes could not be detected (Fig. 7C,D). These observations suggest that, in the absence of head mesoderm, the remaining (trunk) population of mesoderm cells is unable to produce hemocytes.

The analysis of the BicD phenotype, as well as two other mutations, twi sna double mutants and tor⁴⁰²¹, demonstrates that macrophages are not required for cell death to occur. Despite the absence of macrophages in late BicD mutant embryos, there is abundant cell death; cellular debris is located extracellularly in the hemolymph space (Fig. 7C,E). The same observation was made for twi sna double mutants (Fig. 7F) and tor⁴⁰²¹ mutants (data not shown) which, like BicD, lack head mesoderm. In these mutants, massive cell death occurs, leading to extensive clusters of cell fragments which take up much of the hemolymph space. These findings show that macrophages do not play an active role in the initiation of at least the majority of apoptosis.

Since hemocytes are cells that are not stationary but move freely in the hemolymph, it is difficult in the absence of suitable markers to determine their exact origin in the embryo. This problem has resulted in some controversy; besides the median mesoderm, on which most authors agree as the site of hemocyte formation in other insects, the coelomic sacs (i.e., The hemocytes in the Drosophila embryo, according to the lateral mesoderm) and the subesophageal body were proposed as tissues giving rise to hemocytes. As discussed by Mori (1979), these reports were almost certainly the result of the erroneous assumption that hemocytes, because they are spatially associated with coelomic sacs and subesophageal body in late embryos, have to derive from these tissues.

The median mesoderm can be defined at a stage when the cells of the formerly monolayered mesoderm regroup and form multiple layers. Dorsally, the mesoderm forms metamerically reiterated ‘sacs’ with an inner layer (splanchnopleura or visceral mesoderm) and an outer layer (somatopleura or somatic mesoderm). The mid-ventral portion of the mesoderm remains thinner than the lateral mesoderm and is defined as the median mesoderm. According to previous histological studies, part of the median mesoderm cells dissociate and become hemocytes (reviewed in Mori, 1979).

In Drosophila, the median mesoderm is a 1-to 2-cell-thick layer overlying the primordium of the ventral nerve cord during stages 11-12 and differentiates later in somatic muscle and fat body cells (Hartenstein and Jan, 1992). The present study shows that, in Drosophila, if at all, only the most anterior (gnathal) part of the median mesoderm gives rise to hemocytes. In fact it is possible that the hemocytes derive only from the procephalic mesoderm and associate with the median mesoderm of the gnathal segments during an early phase of their migration.

There exists a confusingly diverse nomenclature of insect hemocytes. The widely used classification of Gupta (1985) distinguishes between 7 classes of hemocytes. Of these, 2 classes have been described for the Drosophila larva (Rizki, 1978): the plasmatocytes and oenocytoids (Rizki calls the latter ‘crystal cells’). Plasmatocytes are large, variably shaped cells which show phagocytic activity. Oenocytoids, in contrast, do not phagocytose; they are structurally defined by inclusions of fibrous or crystalline material (Gupta, 1985). We could not detect oenocytoids in the embryo. Furthermore, at least in the background of mutations with increased cell death, all hemocytes showed phagocytic activity, which is absent from oenocytoids. These findings indicate that oenocytoids may originate shortly before hatching (late stage 17) or during postembryonic development.

The hemocytes in the Drosophila embryo, according to the morphological criteria given by Gupta (1985), are plasmatocytes. In wild-type embryos most of the hemocytes have phagocytosed dead cell or cell fragments and are therefore operationally defined as macrophages. If cell death is elevated, the number of hemocytes that have incorporated cellular debris approaches 100%. These observations suggest that the hemocytes of the Drosophila embryo are a homogeneous group of cells all of which have the ability to act as macrophages if required.

In the second half of embryogenesis, the bulk of extracellular matrix molecules are secreted (Fessler and Fessler, 1989) and basement membranes develop that cover all cell surfaces that are in contact with the hemolymph (Tepass and Hartenstein, 1994) with the exception of the cell surface of the hemocytes themselves. The hemocytes are apparently a major source of extracellular matrix molecules including papilin, laminin, collagen IV, glutactin and tiggrin (Fessler and Fessler, 1989, Kusche-Gullberg et. al., 1992, Fogerty et al., 1994). The ultrastructure of embryonic hemocytes, which is characterized by an extensive often dilated rough endoplasmatic reticulum suggesting a high secretory activity, is consistent with this conclusion.

A substantial fraction of cells die during normal embryonic development of Drosophila. An analysis of the pattern of cell death has recently been published (Abrams et al., 1993). Among the dying cells are mainly epidermal and neural precursors. Cell death is particularly abundant in the head during mid-embryogenesis. Here, contiguous regions of the head ectoderm degenerate, leading (among other mechanisms) to the drastic reorganization of the head structures during the head involution process (Younossi-Hartenstein et al., 1993). By contrast, no immediate consequence of the cell death occuring in the trunk has so far been detected. In all cases of cell death, most of the cellular debris is taken up by macrophages. In addition, similar to what has been described for the eye (Wolff and Ready, 1991), epidermal cells also occasionally phagocytose cellular debris (this study).

What is the relationship between hemocytes, macrophages and cell death in the Drosophila embryo? First, the formation of hemocytes from the head mesoderm is independent of cell death. Hemocytes are born before the onset of cell death. The number of mesoderm cells giving rise to hemocytes is intrinsically fixed and is not influenced by the amount of cell death, as evidenced by the fact that in mutants with dramatically increased cell death no addtional hemocytes are formed.

Secondly, the time and place at which hemocytes assume phagocytic activity seems to be determined by the pattern of cell death. In wild-type embryos, macrophages first appear in large numbers in the head and at the ventral aspect of the ventral nerve cord; by contrast, hemocytes located in the interior of the embryo around the gut become phagocytic late or not at all in the embryo. In the mutant fkh, in which the gut epithelium degenerates during stage 12/13 (Weigel et al., 1989), all deeply located hemocytes around the gut show phagocytic activity already at stage 13. Furthermore, the amount of cellular debris taken up by individual macrophages also depends on the extent of cell death. In mutants with increased cell death, macrophages have a larger average size and contain more vacuoles than in wild type. There seems to be a maximum volume of material a given macrophage can take up: in kni and sdt, a considerable amount of cellular debris remains extracellular, a phenomenon seen very rarely in wild type.

Finally, cell death is independent of macrophages. In three mutations that lack macrophages (BicD, tor⁴⁰²¹, twi sna double mutants), abundant programmed cell death could be observed. This result indicates that, as already stated for the cell death during metamorphosis (Wigglesworth, 1979), macrophages do not play any active part in at least the majority of cell undergoing cell death. The finding is in contrast to the recently published findings of Lang and Bishop (1993) who expressed diphteria-toxin under the control of the promoter for granulocyte/macrophage colony stimulating factor. This promoter specifically directs the expression in three subpopulations of macrophages, among them the hyalocytes found in the eye. Transgenic mice showed indeed degeneration of hyalocytes; in addition, two eye tissues, which normally are only transient, the hyaloid vasculature and the pupillary membrane, persisted, suggesting that the hyalocytes are actively involved in their removal. In conclusion, the role of macrophages during the initiation of cell death may vary among different animal groups and/or organ systems. In order to resolve this issue, more needs to be learned about the molecular machinery that initiates and carries out the cell death program.

We are grateful to Silvia Yu for technical assistance, to Rolf Bodmer, Judy Lengyel and the Bloomington Stock Center for providing fly stocks. We thank John Fessler for critical reading of the manuscript. This work was supported by NSF Grant IBN 9221407 to V. H. and by a HFSPO fellowship to U. T.