Function of Rho GTPases in embryonic blood cell migration in Drosophila

Cell migration is one of the most prominent aspects of animal morphogenesis. There are many different types of cell migration events in which cells move on extracellular substrates or on the surface of other cells (Montell, 1999; Rorth, 2002; Starz-Gaiano and Montell, 2004; Friedl, 2004). During Drosophila embryogenesis, macrophages migrate extensively to populate the interstitial spaces between organ primordia (Tepass et al., 1994). Embryonic macrophages appear to be engaged in two main activities: the phagocytosis of apoptotic cells and the secretion of extracellular matrix (ECM) molecules (Evans et al., 2003). While macrophage migration and cell death in the embryo are largely independent processes, it was recently shown that the effective removal of dead cells from the central nervous system by macrophages is critical for CNS morphogenesis (Sears et al., 2003). Embryonic macrophages persist into postembryonic stages and give rise to a substantial portion of the adult blood cell population that is further supplemented by hemocytes that derive from the larval lymph glands (Holz et al., 2003; Evans et al., 2003).

Embryonic blood cells (hemocytes) derive entirely from the head mesoderm (Tepass et al., 1994). Approximately 700 of those cells, called plasmatocytes, are migratory and differentiate into macrophages whereas a small stationary population of 36 hemocytes remain associated with the foregut and give rise to the crystal cells (Lebestky et al., 2000; Evans et al., 2003). The migration of macrophages during embryogenesis can be subdivided into three phases (Tepass et al., 1994; Cho et al., 2002). In phase I (stages 10 and 11) plasmatocytes initiate motility and scatter locally throughout the head region. In phase II (stages 12 to 14) plasmatocytes (which have now started to phagocytose and are thus considered macrophages) undergo a large-scale migration, following a few major migration routes, as they emerge from the head region to populate the rest of the embryo. At stage 11 the germband of the embryo is extended so that the tail is located next to the head. Macrophages enter the tail and then are carried with the retracting germband during stage 12 to populate the posterior of the embryo. As the germband retracts, macrophages also migrate beneath the amnioserosa that covers the embryo dorsally. Further, macrophages will emanate from the head and the tail and migrate along the dorsal and ventral aspects of the ventral nerve cord to populate the ventral trunk region. A late aspect of the migration of macrophages during phase II depends on the Pvr/Pvf guidance system (Brückner et al., 2004). In phase III (stages 15-17) macrophages that are scattered through the entire internal space of the embryo retaining vigorous local motility. Macrophages tend to congregate only at sites of excessive cell death, to phagocytose apoptotic cells (Abrams et al., 1993; Tepass et al., 1994).

Mechanistically, cell migration is a complex process involving the coordinated activity of several cellular compartments, including the protrusions of cytoplasmic processes at the leading edge, the subsequent forward movement of the main cell body, and the retraction of the trailing end of the cell. Traction-generating adhesive interactions with the substratum are dynamically regulated as they are newly established at the leading edge, maintained in the main cell body and dissolved at the trailing edge of the cell. Adhesion receptors, such as integrins and cadherins, are linked to the actin cytoskeleton, which generates the forces needed to propel the cell forward (Lauffenburger and Horvitz, 1996; Hall, 1998; Pollard and Borisy, 2003; Ridley et al., 2003). Small GTPases of the Rho family have been identified as key regulations of adhesion and cytoskeletal dynamics during migration (Hall, 1998; Ridley, 2001). Six Drosophila Rho proteins have been described to date: Cdc42, Rac1, Rac2, Rho1 and the two divergent Rho proteins, Rho-like (RhoL) and Mig-2-like (Mtl) (Luo et al., 1994; Harden et al., 1995; Hariharan et al., 1995; Murphy and Montell, 1996; Fehon et al., 1997; Sasamura et al., 1997; Newsome et al., 2000; Ng et al., 2002; Hakeda-Suzuki et al., 2002). Mtl is the Drosophila orthologue of the C. elegans Mig-2 Rho GTPase (Zipkin et al., 1997) and is structurally similar to both Rac and Cdc42. Functionally, Mtl appears to behave like Rac1 and Rac2 because these three GTPases act redundantly in regulating dorsal closure and axon growth and guidance. Mtl, Rac1 and Rac2 are therefore referred to as the Drosophila Rac genes (Ng et al., 2002; Hakeda-Suzuki et al., 2002). A functional similarity between RhoL [which has also been referred to as Rac3 (Sasamura et al., 1997)] and other Rho family members has not been established.

The role of Drosophila Rho GTPases in various developmental processes was initially assayed using two types of mutations: a mutation that encodes the dominant negative (DN) isoform, which has a reduced affinity for nucleotides and is thought to sequester guanine nucleotide exchange factors (GEFs), preventing them from functioning anywhere in the cell. The second mutation is a constitutively active (CA) isoform, which is permanently bound to GTP (Ridley, 2001). The expression of DN and CA isoforms of Rho GTPases in a tissue-specific manner has been a useful tool in elucidating the role of these proteins in developing tissues, particularly because the expression of DN forms eliminates the activity of maternally and zygotically derived gene products. However, comparisons of phenotypes induced by the expression of DN isoforms and phenotypes of corresponding loss-of-function mutations have revealed some inconsistencies (Luo et al., 1994; Kaufmann et al., 1998; Genova et al., 2000; Hakeda-Suzuki et al., 2002). In this study we have analyzed the function of Rho GTPases in Drosophila embryonic macrophage migration by studying embryos that misexpress DN and CA isoforms of GTPases as well as by analyzing embryos mutant for specific GTPases.

Flies were raised on standard medium. Crosses were performed at 25°C, or at 29°C if Gal4 drivers were involved. Embryos were collected on yeasted apple juice agar plates and staged according to Campos-Ortega and Hartenstein (Campos-Ortega and Hartenstein, 1997). The fly strains used in this study are listed in Table 1. Embryos that have reduced maternal and no zygotic Cdc42 expression (Cdc42^MZ) were derived from Cdc42³/Cdc42⁶ or Cdc42⁴/Cdc42⁶ females crossed to wild-type flies. Cdc42³ and Cdc42⁴ are null alleles while Cdc42⁶ is a hypomorphic allele of Cdc42 (Fehon et al., 1997; Genova et al., 2000). As Cdc42 is located on the X chromosome, 50% of males derived from these females carry a Cdc42 null allele. Double mutant germline clones for Rac1^J10 and Rac2^Δ in a wild-type or a homozygous Mtl^Δ mutant background were induced as described previously (Hakeda-Suzuki et al., 2002). Rac2^Δ and Mtl^Δ are null alleles while Rac1^J10 is a hypomorphic allele (Hakeda-Suzuki et al., 2002). For all experiments at least 30 embryos were evaluated, except for the Rac1^J10 and Rac2^Δ embryos generated in an Mtl^Δ mutant background. Of these, we recovered 16 embryos that could be evaluated, from which seven carried a balancer chromosome as indicated by a lacZ marker and thus had normal zygotic Rac expression. Nine embryos were maternally and zygotically mutant for all three Rac genes. Six of those embryos were in stages 13-15 and displayed macrophage migration defects while the remaining three embryos were in stages 16 or 17 and showed normal dispersal of macrophages.

Two macrophage-specific Gal4 lines were generated by targeted P element replacement, a technique suggested to us by Frank Laski (see also Gonzy-Treboul et al., 1995; Sepp and Auld, 1999). The X chromosomal insertion P[Gal4, w⁺]E132 was used as a donor line and gcm-lacZ and Coll-lacZ, both carrying the rosy (ry) gene as a genetic marker, as recipient lines. After selecting transposase-induced autosomal insertions of P[Gal4, w⁺], potential successful targeted transposition events were identified by examining the Gal4 expression pattern in a UAS-nGFP or UAS-lacZ background. Eight out of 21 (38%) tested lines were successful conversions of gcm-lacZ to gcm-Gal4 and three out of five (60%) tested lines were conversions of Coll-lacZ to Coll-Gal4. A recombinant chromosome containing both gcm-Gal4 and UAS-lacZ was generated to label macrophages. As Coll-Gal4 is located on the CyO balancer chromosome, genetic crosses involving Coll-Gal4 were carried out in the presence of a UAS-lacZ insertion on the third chromosome.

For observations of live embryos they were dechorionated in 50% bleach for 1 minute, rinsed with water and quickly lined up on a 22×40 mm coverslip coated with Scotch tape glue. To extract the glue, the Scotch tape was soaked in heptane. The heptane evaporates quickly when spread on a coverslip, leaving the glue behind. A rim of Vaseline was placed around the embryos and the depression created was filled with halocarbon® 56 oil (Halocarbon Product Cooperation). The coverslip was attached to a metal microscope slide with a 15×32 mm rectangular gap in its center. Embryos were imaged using an inverted microscope attached to a LSM510 Zeiss laser scanning confocal microscope. Movies were generated by taking an image every 10-20 seconds over 20- to 30-minute time periods.

Antibody stainings of embryos followed standard protocols. Embryos used for phalloidin stainings were devitellinized by hand instead of a methanol heptane mixture. Primary antibodies used were rabbit anti-β-Galactosidase (β-Gal) (1:1500; Cappel), rabbit anti-Croquemort (1:1000) (Franc et al., 1996), mouse anti-peroxidasin (1:2000) (Nelson et al., 1994). Secondary antibodies were Alexa488, Cy3 and HRP conjugated (1:500; Jackson Laboratories; Molecular Probes). Phalloidin-Oregon Green488 (1:80; Molecular Probes) was used to label F-actin. For observations with differential interference contrast microscopy, embryos were dehydrated in an ethanol series and then mounted in a mixture of two-thirds Canada balsam, one-third methylsalicylate. For observations with laser scanning confocal microscopy, embryos were mounted directly in antifade (70% glycerol in PBS containing 1 mg/ml p-phenylene diamine). Whole-mount in situ hybridization with digoxigenin-labeled DNA probes followed standard procedures. Transmission electron microscopy was carried out as described previously (Tepass and Hartenstein, 1994), and ultrathin sections were analyzed and photographed on a Hitachi H-7000 transmission electron microscope.

The Deficiency Kit collection (Bloomington Drosophila Stock Center) was used to identify loci involved in macrophage migration. Embryos were collected from individual deficiency strains and stained with macrophage markers. For all lines bearing deletions on the second chromosome and a few lines with deletions on the first or third chromosome, the embryos were stained with anti-peroxidasin antibody. In the remaining lines, macrophages were labeled in the background of the gcm-Gal4, UAS-lacZ chromosome with an anti-β-Gal antibody. For most deletions on the first and third chromosome, blue balancers were introduced to enable the identification of embryos homozygous for the deletion. For the remaining lines, homozygous mutant embryos were recognized by a consistent phenotype displayed by approximately 25% of the embryos.

To manipulate gene expression specifically in migrating hemocytes two Gal4 driver lines that induce expression of UAS controlled target genes in macrophages were generated. We used targeted P element replacement, which allows for the exchange of one P element with the sequence of another (Gonzy-Treboul et al., 1995; Sepp and Auld, 1999), to convert enhancer-trap lines that showed expression of lacZ in embryonic hemocytes into Gal4 drivers. gcm-lacZ [(Bernardoni et al., 1997) Fig. 1A-C] and Coll-lacZ [(Bellen et al., 1989) Fig. 1J-L], which carry a P element enhancer-trap insertion in glia cells missing (gcm) and the collagen IV encoding gene Cg25C, respectively, were chosen as they had the most useful expression pattern for this study. In gcm-lacZ embryos, β-Gal expression was first detected in the hemocyte primordium at the end of stage 7 (Fig. 1A) and persisted in macrophages until late embryogenesis (Fig. 1C). In addition to expression in macrophages, lacZ is also activated in some glial cells and in a segmentally repeated single-cell-wide stripes of epidermal cells located in the segmental groves (Fig. 1B and not shown). In Coll-lacZ embryos, β-Gal expression in macrophages was first detected at stage 13, although only weakly in an apparently random subpopulation of macrophages (Fig. 1J). The levels of β-Gal expression increased by stage 15 and were maintained in macrophages until the end of embryogenesis (Fig. 1L). In addition, β-Gal expression was also seen in the fat body. A similar expression pattern was reported for Cg25C (Lunstrum et al., 1988).

P element replacement was highly efficient for both lines generating gcm-Gal4 and Coll-Gal4 (see Materials and Methods). The expression of UAS-lacZ under the control of gcm-Gal4 and Coll-Gal4 was characterized and compared to lacZ expression of gcm-lacZ and Coll-lacZ (Fig. 1). In gcm-Gal4 UAS-lacZ embryos, lacZ transcript was first detected in the hemocyte primordium at stage 8 and persisted in macrophages until stage 15 (Fig. 1D-F). After stage 15, lacZ transcript declines in macrophages. The expression pattern of β-Gal protein was similar to that of the lacZ transcript. β-Gal was first detected in the hemocyte progenitors at stage 9, approximately 30 minutes later than in the gcm-lacZ line. Although the lacZ transcript began to disappear from macrophages at stage 15, β-Gal protein was maintained until stage 17 (Fig. 1G-I). gcm-Gal4 appears to drive the expression of UAS controlled reporter genes uniformly in all migrating macrophages. In gcm-Gal4 UAS-lacZ embryos both the lacZ transcript and β-Gal were also detected in glial cells and epidermal stripes. Thus, gcm-Gal4 is active in macrophage progenitor cells and in plasmatocytes/macrophages during phase I and II of their migration (stages 8-15).

In Coll-Gal4 UAS-lacZ embryos, lacZ transcript and β-Gal protein were first detected at stage 13 in a subset of macrophages (Fig. 1M,P). The number of macrophages expressing lacZ increased in older embryos, and by stage 15, all macrophages showed lacZ expression (Fig. 1N,Q). In addition, lacZ expression in Coll-Gal4 UAS-lacZ embryos was also detected in the fat body and in the salivary glands. The latter does not correlate with the expression pattern of lacZ in the original Coll-lacZ line. The activation of reporter genes by Coll-Gal4 in the salivary glands is likely a secondary effect of the Gal4/UAS system, caused by the presence of an hsp70 (salivary gland-specific) enhancer sequence in the Gal4 construct (Brand and Perrimon, 1993; Gerlitz et al., 2002). In conclusion, Coll-Gal4 is expressed in macrophages during late phase II and phase III of their migration (stages 13-17). Neither the gcm or Coll enhancer-trap lines nor the corresponding Gal4 drivers showed expression in crystal cells.

Cell migration requires the formation of membrane protrusions at the leading edge, driven in most cells by actin polymerization. Motile cells in culture most commonly display two types of protrusions - filopodia and lamellipodia (Lauffenburger and Horwitz, 1996). To characterize the shape and extensions of migrating macrophages, we examined macrophages in live embryos using time-lapse confocal microscopy. For these observations, macrophages were labeled using the gcm-Gal4 driver in combination with four to six copies of UAS-mGFP, expressing membrane-tethered GFP (Lee and Luo, 1999).

As expected, time-lapse observations revealed that macrophages are highly motile cells. Migrating macrophages extend predominantly one wide, flat cytoplasmic protrusion (Fig. 2A-P; see Movies 1 and 2 in supplementary material) that resemble the lamellipodia described for motile cells in culture. These lamellipodia are often 15-20 μm in length, about twice the diameter of the main cell body. The macrophage lamellipodia are highly dynamic, constantly changing shape by rapid extension and retraction. Extension and retraction of a large lamellipodium is achieved within 5-6 minutes. Lamellipodia extend in the direction of movement, which is illustrated in Fig. 2G-J that shows a macrophage indicated by an asterisk retracting its lamellipodium as the cell body rotates clockwise, followed by the emergence of a lamellipodium at a different site that now becomes the new leading edge and pointing in the direction of movement. A similar scenario is demonstrated in Fig. 2K-P, in which a lamellipodium can be seen progressively turning toward the left, followed by the cell body. In addition to lamellipodia, macrophages also extend multiple thin, needle-like cytoplasmic protrusions similar to filopodia (Fig. 2G-J; see Movies 1 and 2 in supplementary material). Filopodia protrude from the periphery of lamellipodia and display an exploratory behavior with quick retraction and extension. They were often seen to precede lamellipodium formation. Alternatively, filopodia-like structures are often the apparent remnants of collapsing lamellipodia. Lamellipodia are well-preserved in glutaraldehyde-fixed material (Fig. 2Q). In contrast, using DIC/Nomarski optics or laser confocal microscopy we were not able to visualize lamellipodia in macrophages fixed with formaldehyde and labeled for cytoplasmic β-Gal, or with antibodies against peroxidasin, which accumulates in the extensive endoplasmic reticulum and other compartments of the biosynthetic pathway (e.g. Fig. 2R). In such preparations, most migratory macrophages appear spindle-shaped, while macrophages that contain many phagosomes are large, rounded cells. Taken together, these observations show that migrating macrophages extend dynamic membrane extensions that are comparable to lamellipodia and filopodia. A large lamellipodium polarized in the direction of movement is the predominant extension or embryonic macrophages.

To examine the function of Rho GTPases in embryonic macrophage migration, the effects of expressing DN and CA Rho GTPase isoforms under the control of gcm-Gal4 and Coll-Gal4 drivers were examined. The phenotypes observed were compared to those of embryos carrying mutations in Rho GTPase genes.

The expression of both the DN (Rac1^N17) and CA (Rac1^V12) isoforms of Rac1 under the control of gcm-Gal4 blocked macrophage migration. Most macrophages expressing Rac1^N17 failed to migrate and remained around the foregut forming a tight cluster. This defect persisted until the end of embryogenesis (Fig. 3C,D). A few macrophages were seen in the head and tail of the embryos, suggesting that some cells had the ability to migrate, although they moved only a short distance.

Macrophages expressing Rac1^V12 showed similar, but more drastic defects (Fig. 3E,F). Most cells were unable to migrate and remained clustered anteriorly. The number of migratory macrophages was even smaller than seen upon expression of Rac1^N17. Macrophages expressing either DN or CA Rac1 were positive for macrophage-specific markers such as peroxidasin (Nelson et al., 1994), a basement membrane protein produced by macrophages (Fig. 3A-F) and Croquemort (Cqr) (Franc et al., 1996), a receptor required within macrophages for the efficient uptake of apoptotic cells (Fig. 3I-K), suggesting that the phenotype caused by altered Rac1 activity cannot be attributed to defects in cell fate determination. These findings suggest that the correct regulation of Rac activity is essential for macrophage migration.

To address the question of whether Rac function is also required during late embryonic stages when macrophages show local motility, the expression of the various Rac1 isoforms was examined using the Coll-Gal4 driver that is expressed in macrophages after they are dispersed throughout the embryo. The expression of Rac1^N17 under the control of Coll-Gal4 caused the macrophages to cluster at various sites, such as the head region, around the pharynx, midgut and hindgut, and laterally along the ventral cord. The deviation from wild type was evident at stage 16, but became more pronounced during stage 17 (Fig. 4C,D). Coll-Gal4 UAS-Rac1^V12 embryos showed an even more severe clustering of macrophages than observed with Rac1^N17. The clusters appeared larger and tighter and large areas of the embryo were devoid of macrophages (Fig. 4E,F). In addition to abnormal clustering, macrophages displayed abnormal cell shapes in response to expression of mutant Rac1 isoforms. Rac1^N17 blocked the extension of protrusions resulting in rounded cell shape (Fig. 4D), whereas Rac1^V12 caused macrophages to adopt a more elongated and spindle-like shape compared to wild type, and to display more pronounced extensions (Fig. 4F). These findings suggest that normal Rac activity is required throughout embryogenesis to control macrophage motility.

The expression of a third mutant isoform of Rac1, Rac1^L89, in macrophages caused these cells to migrate slower than in wild type although defects were mild in comparison to those caused by expression of Rac1^N17 and Rac1^V12. In wild-type stage 13 embryos when the germband is retracted, macrophages from the head and tail regions migrate toward the middle of the embryo along the ventral cord. At this stage the anterior and posterior groups of macrophages have not met in the center of the embryo, and the ventral part of the ventral cord of abdominal segments A3-A5 is devoid of macrophages (Fig. 3A). In subsequent stages, the entire ventral cord is covered by macrophages (Fig. 3B). Analyzing when macrophages meet on the ventrally on the ventral cord provides a good measure to determine the speed of macrophage migration. Macrophages in embryos that express Rac1^L89 under the control of gcm-Gal4 did not spread over the ventral surface of the ventral nerve cord at stage 13 (Fig. 3G). At stages 15, the abdominal segments A3-A7 of these embryos still remained devoid of macrophages (Fig. 3H). Expression of Rac1^L89 under the control of Coll-Gal4 induced some clustering of macrophages, and accentuated the spindle-like shape of macrophages similar to Rac1^V12 (Fig. 4G,H). Previous work has reported both loss-of-function and gain-of-function activities for Rac1^L89 (e.g. Luo et al., 1994; Kaufmann et al., 1998). Our observations suggest that Rac1^L89 functions as a weak CA isoform in macrophages.

To determine how Rac functions in regulating macrophage morphology, the filamentous actin (F-actin) content of macrophages was analyzed. In wild-type macrophages, low levels of F-actin were dispersed around the cell periphery (Fig. 5A). The membrane extensions characteristic of macrophages could not be discerned with this staining procedure. Macrophages expressing Rac1^N17 showed wild-type levels of cortical F-actin (Fig. 5B). In contrast, Rac1^V12-expressing macrophages displayed strongly elevated levels of F-actin (Fig. 5C). Marcophages that cluster around the foregut in Rac1^N17-and Rac1^V12-expressing embryos were examined with transmission electron microscopy (TEM) to analyze cell morphology. The expression of Rac1^V12 elicits the formation of an excessive number of lamellipodia (Fig. 5E), corresponding to the dramatic accumulation of F-actin, whereas Rac1^N17 was found to block the formation of membrane extensions (Fig. 5D). TEM preparations also revealed that some macrophages expressing Rac1^V12, but not macrophages expressing Rac1^N17, contain two nuclei (Fig. 5F). 15-20% of the macrophages that expressed Rac1^V12 and whose nuclei were visible in the section, were binucleate. The actual fraction of macrophages that contain two nuclei is presumably higher as the plane of sectioning may not reveal both nuclei. This observation suggests that activation of Rac blocks cytokinesis in macrophage progenitors. In addition, TEM preparations of embryos expressing either the DN or CA Rac1 isoform show that at least some of these macrophages contain apoptotic cells (Fig. 5D,E), although the number of macrophages that display phagocytotic activity is much lower than in wild-type. Taken together, our findings suggest that Rac activity plays a key role in macrophage migration by controlling actin organization and lamellipodia formation.

We also analyzed loss-of-function mutations of the three Drosophila Rac genes. Animals that lack either Rac1, Rac2 or Mtl activity are viable and do not show defects in macrophage migration. Embryos double mutant for any combination of null mutations of two Rac genes, or triple mutant embryos carrying the null mutations Rac2^Δ and Mtl^Δ as well as the hypomorphic allele Rac1^J10 did not display defects in macrophage migration (data not shown). Rac genes have maternal components of expression (Hakeda-Suzuki et al., 2002) that could support normal macrophage migration even in the absence of zygotic Rac expression. Thus, we next analyzed embryos derived from Rac1^J10 Rac2^Δ germline clones that were homozygous mutant for Rac1^J10 and Rac2^Δ, and embryos derived from Rac1^J10 Rac2^Δ germline clones in a Mtl^Δ mutant background that were homozygous mutant for all three Rac genes. The hypomorphic allele Rac1^J10 rather than a Rac1 null allele was used in these experiments as complete removal of Rac1 and Rac2 activity from the germline blocks germline development or early embryogenesis (Hakeda-Suzuki et al., 2002). Rac1^J10 Rac2^Δ maternal and zygotic mutant embryos displayed mild defects in macrophage migration as macrophages did not disperse into the ventral posterior trunk region by stage 14 or 15 (Fig. 6B). By late embryogenesis macrophages had dispersed throughout the entire ventral trunk region in these embryos. In contrast, sibling embryos that derived from Rac1^J10 Rac2^Δ germline clones but have single paternal wild-type copy of Rac1 and Rac2 show a normal distribution of macrophages (Fig. 6A). Surprisingly, embryos maternal and zygotic mutant for Rac1^J10 Rac2^Δ and Mtl^Δ displayed the same macrophage migration defects (Fig. 6C) as embryos that lack Rac2 and have reduced Rac1 activity (Fig. 6B) - macrophages did not populate the posterior trunk region by stages 13-15, but showed normal dispersal in late embryos. These results suggest that Rac1 and Rac2 act redundantly to promote macrophage migration and that Mtl does not contribute significantly to macrophage motility.

The expression of DN Cdc42^N17 or CA Cdc42^V12 under the control of either gcm-Gal4 or Coll-Gal4 did not interfere with the migration or distribution of macrophages (Fig. 7). Surprisingly, in embryos that have reduced maternal and no zygotic Cdc42 expression (hereafter referred to as Cdc42^MZ mutant embryos) the ventral cord is not covered by macrophages. Also the ubiquitous expression of Cdc42^N17, using a da-Gal4 driver, causes embryonic defects similar to those observed in Cdc42^MZ mutants including defects in macrophage migration (data not shown). Ubiquitous expression of Cdc42^V12 elicits defects in the entire embryo so severe that macrophage dispersal could not be evaluated. These findings suggest that Cdc42 is required in cells other than the macrophages to promote macrophage migration. In Cdc42 mutant embryos, germband retraction is defective (Genova et al., 2000). As most of the posterior half of the ventral cord gets populated with macrophages carried posteriorly during germband retraction, it was important to establish whether defects in germband shortening affect the spatial and temporal distribution of macrophages. To test this, hindsight (hnt) mutant embryos, which fail to undergo germband retraction (Yip et al., 1997), were analyzed for alterations in macrophage distribution. These embryos show a normal macrophage dispersal (data not shown), suggesting that germband retraction, which carries macrophages to the posterior pole of the egg in wild type, is not essential for macrophage dispersal throughout the embryo.

Macrophages expressing either Cdc42^N17 or Cdc42^V12 under the control of gcm-Gal4 and Cdc42^MZ embryos displayed wild-type cytoplasmic protrusions (Fig. 7B,E,H) and had normal F-actin content (data not shown). Unexpectedly, the expression of Cdc42^V12 under the control of Coll-Gal4 resulted in the macrophages becoming round with only short protrusions being extended (Fig. 7F). In contrast, macrophages expressing Cdc42^N17 under the control of Coll-Gal4 (Fig. 7C) did not show an altered morphology, nor did macrophages in Cdc42^MZ mutant embryos (not shown). In addition, the expression of Cdc42^V12 under the control of gcm-Gal4 caused an increase in the size of macrophages and a reduction in their number to approximately 400 (n=4; Fig. 7D) compared to wild-type embryos, which have about 700 macrophages (Tepass et al., 1994). Furthermore, TEM analysis of these embryos revealed a fraction of binucleate cells among macrophages that expressed Cdc42^V12 (Fig. 7E). This suggests that Cdc42^V12, similar to Rac1^V12, blocks cytokinesis in hemocyte progenitors, which would account for the enlargement of macrophage cells and the decrease in macrophage number in those embryos.

The expression of DN Rho1^N19 under the control of either gcm-Gal4, Coll-Gal4 or da-Gal4 did not alter the migration of macrophages, nor their distribution or morphology (data not shown). Consistent with these observations, macrophage defects were not observed in Rho1 mutant embryos. A CA isoform of Rho1 was not available, but a CA isoform of the human RhoA had been generated (Harden et al., 1999). The expression of this construct under the control of gcm-Gal4 had no effect on macrophage migration (data not shown). Thus, Rho1 has no apparent role in macrophage migration.

RhoL/Rac3 was identified as a novel Drosophila Rho protein (Murphy and Montell, 1996; Sasamura et al., 1997). The expression of DN RhoL^N25 or CA RhoL^V20 under the control of gcm-Gal4 or da-Gal4 did not interfere with the migration of macrophages. This result was unexpected as RhoL transcript was shown to be enriched in macrophages (Sasamura et al., 1997). Interestingly, in embryos with a deficiency that covers the RhoL locus (Df(3R)by416), macrophages fail to populate the ventral side of the ventral cord, causing a phenotype similar to that of sim mutants (Fig. 8B,C). This raises the possibility that RhoL may function in macrophage migration and that the RhoL^N25 and RhoL^V20 constructs are ineffective.

To identify genes whose zygotic expression is required for embryonic macrophage migration, a systematic screen of chromosomal deletions was carried out to reveal mutants in which the migration process is disrupted. We analyzed 208 deficiencies received from the Bloomington Drosophila Stock Center that cover over 80% of the Drosophila genome. Eighteen of the 208 lines produced mutant embryos with severe morphological defects or an early developmental arrest that precluded an evaluation of defects in macrophage migration. We found eight lines that produced embryos with clear defects in macrophage migration. In one group of mutants that includes Df(2L)s1402, Df(3R)e-N19 and Df(2L)ast2, macrophages fail to migrate along the ventral cord, and much of the ventral abdomen remains free of macrophages (Fig. 8D-F). This phenotype resembles that described for single minded (sim) mutants (Fig. 8B) (Zhou et al., 1995). In a second group of mutants that include Df(2L)TE29Aa-11, Df(3L)66C-G28, Df(2L)N22-14, Df(3R)mbc-R1 and Df(3R)mbc-30, little if any migration occurs and most macrophages remain in the anterior region, clustered around the foregut (Fig. 8G-K). This phenotype is similar to the phenotype induced by the expression of DN or CA Rac1 under the expression of gcm-Gal4, or the slightly weaker phenotype described for embryos that lack the Pvr receptor (Cho et al., 2002; Sears et al., 2003). In fact Df(2L)TE29Aa-11 uncovers the Pvr locus. Df(3R)mbc-R1 and Df(3R)mbc-30 are overlapping deficiencies that uncover myoblast city (mbc) the Drosophila ortholog of mammalian DOCK180, a known regulator of Rac activity during cell migration (Erickson et al., 1997; Nolan et al., 1998; Raftopoulou and Hall, 2003). To confirm mbc requirement for macrophage migration we examined embryos homozygous mutant for either of two mbc alleles (mbc^C1, mbc^D11.2). To our surprise we found that mbc mutant embryos did not show any apparent defects in macrophage migration although they displayed the previously described defects in muscle development (data not shown). This finding suggests that Df(3R)mbc-R1 and Df(3R)mbc-30 uncover a gene other than mbc that is required for macrophage migration.

In the Drosophila embryo, macrophages migrate over the internal surfaces of organs that are covered with basement membranes. The space between organs is not filled with matrix aside from the basement membrane (Tepass and Hartenstein, 1994; Tepass et al., 1994) suggesting that Drosophila macrophage migration is more related to a two-rather than a three-dimensional cell culture system. Macrophages show an elongated, polarized morphology typical for many migrating cell types. The leading edge of macrophages is characterized by a long and broad lamellipodium that often extends twice the length of the main cell body. From the trailing end, chunks of cytoplasm may be lost as the cell moves forward. The lamellipodia of Drosophila embryonic macrophages have previously been described using TEM analysis but misidentified as filopodia (Tepass et al., 1994). Lamellipodia formation is highly dynamic with a lamellipodium being extended and retracted within several minutes. While macrophages also extend filopodia, most filopodia-like structures we have observed appear to be the remnants of collapsing lamellipodia. Recent cell culture work on Drosophila S2R+ cells, which are believed to have originated from embryonic hemocytes, also reveals lamellipodia as the predominant extension. Disruption of SCAR, a downstream target of Rac, in S2R+ cells leads to the collapse of lamellipodia, leaving in their wake tread-like protrusions that resemble filopodia, but contain a branched filament network rather than a parallel bundle of filaments (Biyasheva et al., 2004).

We have generated two Gal4 driver lines that are expressed in migrating macrophages of the Drosophila embryo. Combined, both Gal4 lines promote expression of UAS-controlled target genes throughout the entire period of embryonic macrophage migration. gcm-Gal4 is expressed in hemocyte progenitors and in all macrophages up to at least stage 15, encompassing phase I and II of macrophage migration. In contrast, Coll-Gal4 drives expression initially at stage 13 in some macrophages, and high levels of expression are seen from stage 15 to the end of embryogenesis in all macrophages, representing phase III of macrophage migration. gcm-Gal4 is also expressed in glial cells and in epidermal stripes, and Coll-Gal4 in the fatbody. Thus, these driver lines are useful tools to analyze the development of macrophages and some other cell types.

gcm-Gal4 and Coll-Gal4 were used to express DN and CA mutant isoforms of the GTPases Rac1, Cdc42, Rho1 and RhoL in macrophages. Only the DN and CA forms of Rac1 interfere with macrophage migration cell-autonomously. Rac1^N17 prevents the formation of lamellipodia while Rac1^V12 causes an extensive formation of lamellipodia accompanied by increased accumulation of F-actin. Also Rac1^L89 promotes lamellipodia formation suggesting that this isoform acts as a weak activated form of Rac1. The effects of modulating Rac1 activity on actin polymerization, lamellipodia formation and cell migration in Drosophila macrophages are similar to those observed in many other cell types in invertebrates and vertebrates (Hall, 1998; Ridley, 2001), confirming a general role for Rac as a key regulator of cell migration.

The analysis of mutant embryos also demonstrates a requirement for Rac activity in macrophage migration. Embryos that lack Rac2 activity completely and have a strongly reduced Rac1 activity show a clear delay in migration. This defect is not enhanced by the complete loss of the activity of the third Drosophila Rac gene Mtl, suggesting that Mtl does not function in macrophage migration. As embryos that completely lack the function of either Rac1 or Rac2 show normal migration, Rac1 and Rac2 must act redundantly during this migration process. The migration defects caused by Rac gene mutations are substantially weaker than those elicited by the expression of Rac1^N17. One possible explanation for this discrepancy is that the hypomorphic Rac1^J10 allele, which we needed to use instead of a null allele to generate germline clones with reduced Rac activity, provides sufficient function to support largely normal migration. Alternatively, it is possible that expression of Rac1^N17 not only compromises the activity of Rac1 and Rac2 but also other cellular processes essential for migration, as is apparently the case in planar cell polarity (Hakeda-Suzuki et al., 2002). For example, Rac1, Rac2 and Cdc42 may act redundantly in macrophage migration and Rac1^N17 disrupts the activity of all three GTPases. This, and other potential scenarios need to be addressed in future studies.

How is Rac activity controlled in macrophages? Mbc, the Drosophila DOCK180 ortholog and a known upstream regulator of Rac in other cell migration processes (Duchek et al., 2001; Raftopoulou and Hall, 2004), has no essential role in macrophage migration as mbc mutant embryos do not display macrophage migration defects. Pvr, the Drosophila homologue of the mammalian PDGF/VEGF receptor, is a receptor tyrosine kinase that may act upstream of Rac. Pvr acts upstream of Rac in the migration of border cells in the Drosophila ovarian follicle (Duchek et al., 2001) and its activity is required for normal embryonic macrophage migration (Cho et al., 2002; Sears et al., 2003; Brückner et al., 2004). Pvr acts primarily as a trophic factor in macrophages as most macrophages undergo apoptosis in Pvr mutant embryos (Brückner et al., 2004). This function of Pvr is mediated largely through the Ras/MAPK pathway (Brückner et al., 2004). Consistently, we did not find evidence for macrophage cell death in Rac mutants or embryos expressing mutant Rac1 isoforms. Surviving macrophages in embryos that lack Pvr engage in `cannibalistic phagocytosis', which apparently slows their dispersal through the embryo. Blocking programmed cell death in Pvr mutants prevents macrophage cell death. In these embryos mild macrophage migration defects are observed (Brückner et al., 2004) that are similar to those we have seen in embryos that lack maternal and zygotic Rac2 and have strongly reduced activity of Rac1. Thus, the control of Rac activity by Pvr to promote macrophage migration is an attractive possibility.

Expression of DN Rho1 and CA human Rho1 did not interfere with macrophage migration, nor did null mutations in Rho1 compromise migration. To our surprise, we also did not find an effect of the DN or CA isoforms of Cdc42 - when specifically expressed in macrophages - on the migration speed or distribution of these cells, which suggests that Cdc42 is not required in macrophages for their migration. Expression of Cdc42^V12, similar to Rac1^V12, blocks cytokinesis in hemocyte progenitors indicating that its expression is effective. In other migrating cells, Cdc42 functions in several processes critical for migration including cell polarization, microtubule organization and filopodia formation (Etienne-Manneville and Hall, 2002; Fukata et al., 2003; Raftopoulou and Hall, 2003; Ridley et al., 2003). Blocking Cdc42 function in mammalian macrophages prevents filopodia formation. These macrophages are still able to migrate but fail to follow a cytokine gradient (Allen et al., 1997; Allen et al., 1998). If expression of Cdc42^N17 in Drosophila macrophages would have similar effects, one might expect migration defects, in particular during phase II of macrophage motility when large scale, most likely guided, migration takes place, which we did not observe. However, macrophages show migration defects in embryos that express Cdc42^N17 uniformly in all tissues as well as in Cdc42 mutant embryos. In both types of embryos, macrophages fail to populate the ventral trunk region. Migration may either be substantially slower in those embryos or macrophages fail to be specifically attracted to this region of the embryo. A similar migration defect has been described for embryos mutant for sim, which lack the midline cells of the central nervous system (Zhou et al., 1995), and for Pvr mutant embryos, ligands of which are expressed in the ventral nerve cord (Cho et al., 2002; Brückner et al., 2004). This raises the possibility that Cdc42 plays a role in midline development or the secretion of Pvr ligands.

Finally, we have investigated the effect on macrophage migration of 190 chromosomal deletions that cover approximately 75% of the Drosophila genome. Additional deletions studied, which together cover about 12% of the genome, were not informative because phenotypes were difficult to interpret because of severe morphological aberrations of the embryo, arrest of embryonic development prior to macrophage development or the loss of mesoderm as a consequence of the loss of genes required for mesoderm specification, such as twist and snail. We found eight lines that showed defects in macrophage distribution. For statistical reasons it is likely that most of these deletions uncover only a single gene that is essential for migration. Mutants with macrophage migration defects include deletions that uncover genes previously shown to be required for macrophage development such as Pvr (Df(2L)TE29Aa-11) (Cho et al., 2002; Sears et al., 2003), gcm and gcm2 [Df(2L)N22-14 and Df(2L)s1402 - both deficiencies uncover gcm and gcm2 (Bernadoni et al., 1997; Alfonso and Jones, 2002)]. We note, however, that the migration defects in Df(2L)N22-14 mutant embryos (by contrast to Df(2L)s1402 mutants) appear stronger than those reported for gcm and gcm2 double mutants suggesting that this deletion may uncover another gene that contributes to macrophage migration. A deletion that uncovers sim was missing from our collection. Interestingly, a deletion that covers RhoL causes migration defects, raising the possibility that RhoL is required for macrophage migration, although the expression of a DN form of RhoL did not result in migration defects. Together, these findings suggest that the Drosophila genome may encode 10-12 genes whose zygotic components of expression are essential for macrophage migration. This is a relatively small number of genes considering the known complexity of the cellular machinery involved in cell migration, suggesting that other migration factors are either maternally provided or act redundantly.

We thank Frank Laski for suggesting the P element replacement strategy to us. We are grateful to Lisa and John Fessler, Barry Dickson, Liqun Luo, Nathalie Franc, Marek Mlodzek, Denise Montell, Lynn Cooley, Nick Harden and the Bloomington Drosophila Stock Center for fly stocks and reagents. We thank Jennifer Liaw and Robert Pascal for help in early stages of this project, and we thank Dorothea Godt for critical comments on the manuscript. This work was supported by a grant from the Natural Sciences and Engineering Research Council of Canada.