The PDZ-GEF Dizzy regulates cell shape of migrating macrophages via Rap1 and integrins in the Drosophila embryo

From worms to humans, cell migration is an essential cellular process:during embryonic development cells migrate to specific locations, where they subsequently differentiate to form tissues and organs. Then later in adult organisms, cell migration is implicated in tissue renewal and immune response. Besides its crucial role during normal development of organisms, cell migration also contributes to pathological processes such as cancer.

Migrating cells undergo a continuous cycle of integrated cellular events that are initiated by migration-promoting cues. Such cues lead to a polarization of the cell and to the formation of protrusions in the direction of migration. The protrusions adhere to the substrate via transmembrane receptors linked to the cytoskeleton. The sites of adhesion then provide traction during the contraction of the cell, leading to the forward movement of the cell body. At the rear of the cell, adhesion sites become disassembled,allowing the cell to detach and to efficiently migrate toward guidance cues(Friedl and Wolf, 2003; Lauffenburger and Horwitz,1996; Ridley et al.,2003). Thus, cell adhesion and its regulation are of crucial importance for migrating cells.

Many different factors regulate cell adhesion during cell migration,including cell surface receptors that mediate adhesion. A major family of these receptors are the integrins, heterodimers composed of one α and one β subunit, which link the migratory substrate to the cytoskeleton(Bökel and Brown, 2002; Hynes, 1992). Integrin-mediated adhesion can be regulated by modulation of their affinity to ligands or by changing their local concentration at the membrane. This regulation occurs either at the outside of the cell by ligands or from the inside by cytoplasmic signals (Kinbara et al., 2003; Liddington and Ginsberg, 2002; van der Flier and Sonnenberg, 2001). Important mediators of the latter signals toward the integrins are small GTPases. For instance, Rap1 is involved in the regulation of integrin-mediated cell adhesion in several cases(Caron et al., 2000; Katagiri et al., 2000; Reedquist et al., 2000). In Drosophila the molecule is required for many aspects of morphogenesis, including invagination of embryonic mesoderm, migration of mesoderm precursors and positioning of adherens junctions(Asha et al., 1999; Boettner et al., 2003; Knox and Brown, 2002). However, Rap1 has not yet been linked to the activity of integrins in Drosophila.

The functional state of small GTPases such as Rap1, the active GTP-bound versus the inactive GDP-bound state is determined by two classes of proteins:G-nucleotide exchange factors (GEFs) and GTPase-activating proteins (GAPs)(Ridley, 2001). While GAPs are responsible for the inactivation of small GTPases, GEFs are the activating components and stimulate specific small GTPases.

We have performed an EP misexpression screen(Rørth, 1996) in order to identify genes involved in the regulation and in the execution of cell migration of embryonic macrophages in Drosophila. Here we report the gene dizzy (Gef26 - FlyBase) encoding the Drosophila PDZ-GEF to be required for proper cell shape and cell migration of macrophages in the Drosophila embryo. Furthermore, we show that Dizzy has the capacity to induce cell shape changes in migrating macrophages depending on the function of Rap1 and βPS integrins. Our data suggest that the Drosophila PDZ-GEF Dizzy is a GEF for Rap1,regulates integrin-dependent adhesion via Rap1, and stabilizes cellular protrusions during the migration of embryonic macrophages in Drosophila.

dizzy^EP, synonyms EP(2)388, pdz-gef ² (Lee et al.,2002); dizzy^P, synonyms l(2)k13720, pdz-gef ¹ (Lee et al.,2002); Df(2L)BSC5, breakpoints 26B1-2;26D1-2 (Deal and Cook, FlyBase); UAS-cd2(Dunin-Borkowski et al.,1995); da-gal4 (Hinz et al., 1994); srph>cd2, recombinant of srph-gal4 and UAS-cd2; UAS-gfp^S65T(Bloomington #1521, #1522); UAS-actin::gfp(Verkhusha et al., 1999); UAS-rap1^WT, UAS-rap1^V12, UAS-rap1^N17(Boettner et al., 2003); rap1^P5709 (Knox and Brown, 2002); UAS-rho^N19 (Strutt et al., 1997), UAS-Rac^N17,UAS-cdc42^N17 (Luo et al.,1994); UAS-p35 (Hay et al., 1994); mys^XG43(Bunch et al., 1992);Δ 2-3 Ki (Robertson et al.,1988). dizzy^Δ1/CyO,actGFP;srph-gal4 UAS-gfp^S65T, UAS-gfp^S65T;srph-gal4 UAS-gfp^S65T and dizzy^EP/UAS-gfp^S65T;srph-gal4 UAS-gfp^S65T/+, for time-lapse recordings.

Fifteen to 30 virgins of the stock srph>cd2 were crossed with three to seven males of one genotype of the EP collection (Szeged). Embryos were collected and immunostained for CD2 following the methods described in Hummel et al. (Hummel et al.,1997), were mounted in methyl salicylate and analyzed for macrophage migration, cell form and cell number.

The P-element insertion of dizzy^EP was mobilized by crossing-in Δ2-3, and the F₂ generation was scored for semi-lethality in trans to Df(2L)BSC5. In total, 11 lines were established with strong eye and wing phenotypes of adult escapers;seven were used for this study: dizzy^Δ1 and dizzy^Δ5 remove 174 and 1187 bp upstream of the former insertion site, dizzy^Δ7 and dizzy^Δ8 2253 bp and 2632 bp around the site, and dizzy^Δ3, dizzy^Δ10 and dizzy^Δ12 2328, 2380 and 2969 bp downstream only. One revertant (dizzy^P-RV4), obtained upon mobilization of the P-element of dizzy^P, has lost the insertion completely and is fully viable.

Qi He and colleagues introduced the gene name dizzy. dizzy is identical to Gef26 and dPDZ-GEF(Lee et al., 2002; Wang et al., 2006).

For the tissue-specific downregulation of dizzy function, we generated a transgene allowing the expression of dizzydouble-stranded RNA under Gal4 control. A fragment of the dizzy EST AT08279 was amplified by PCR, using the 5′-primer GCAGTTAAAAAGATGCTATCGCTG and the 3′-primer TGTTCAAGTTGCGGGTACCGCGT. This fragment contains a portion of exon 3, the exon 4 and a part of exon 5. Following the strategy of Nagel et al.(Nagel et al., 2002), a 485 bp long piece of the fragment (positions 4328 to 4812 of cDNA DQ423241) was first cloned in pHIBS and subsequently cloned in normal and in reverse orientation,separated by the Hairless intron, in pUAST(Brand and Perrimon, 1993). This construct was introduced in the Drosophila germline by P-element-mediated transformation and gave several transgenic lines(UAS-ds.dizzy). Two of these were chosen, which led to lethality when ubiquitously expressed under control of da-gal4.

To examine the role of integrins in cell shape changes induced by dizzy^EP or by Rap1^V12, we crossed mys^XG43/FM7_ftz::lacZ;+;srph>cd2females with w/Y;dizzy^EP;+ or with w/Y;UAS-rap1^V12;UAS-rap1^V12/TM6Bmales. Immunostaining with anti-CD2 antibodies visualized the morphology of the macrophages. With progeny carrying FM7_ftz::lacZ, all embryos gave the cell shape phenotype, but that phenotype was only visible in about 50% of embryos lacking FM7_ftz::lacZ.

Embryos for histochemistry were fixed and immunostained according to standard procedures. Primary antibodies: mouse α-CD2 (Serotec), rabbitα-βGal (Cappel), mouse α-βGal (Promega), rabbitα-pFak^Tyr397 (Biosource), mouse α-βPS(Brower et al., 1984) (DSHB). Secondary antibodies: labeled with Alexa 488, Alexa 555 (Molecular Probes) or biotin (Jackson Labs). The signal of the biotinylated antibodies was enhanced using Vectastain `Elite' and detected either with DAB or with the TSA technique (PerkinElmer). Embryos stained with fluorescent dyes were counterstained with DAPI, mounted in Vectashield and documented with a Leica SP2 confocal system on an inverted Leica DM IRBE microscope. DAB-labeled embryos were mounted in araldite and analyzed on a Zeiss Axioplan 2 equipped with a Progress 3012 camera (Jenoptik).

In-situ hybridization was performed essentially as described in Tautz and Pfeifle (Tautz and Pfeifle,1989). DIG-labeled dizzy RNA probes were generated from EST RH54455, for Pez (CG9493) from EST RE59091 and for Cpr(CG11567) from EST LD46590.

Dechorionated embryos were mounted in a drop of water-saturated 3S Voltalef oil on a slide, properly oriented and covered with a coverslip based on two lateral coverslips. For time-lapse recordings, we performed xyt-sections using the Leica confocal microscope. Images were processed using IPLab (Scanalytics) and then evaluated using a software tool developed with Macromedia Director. The positions of individual cells were recorded in 4D image stacks by mouse click. Then travel distance and cell speed were calculated and the recorded tracks were visualized. Time-lapse movies were exported from IPLab, and processed in QuickTime Player. The lengths of the cellular protrusions were measured on digital photomicrographs in IPLab and exported to Excel for numerical analysis.

Genomic DNA was isolated from dizzy^EP,dizzy^P or homozygous dizzy^Δn adults,amplified by PCR and sequenced essentially following the protocol of J. Rehm(http://www.fruitfly.org/about/methods/inverse.pcr.html). For the molecular characterization of the dizzy^Δnalleles, locus-specific primers were designed, and PCR products were sequenced.

Total RNA was isolated from w embryos of Drosophila,selected for poly-A⁺ RNA and reverse-transcribed to cDNA. Then, the 5′ and 3′ ends of dizzy cDNA were amplified using the RACE systems of Invitrogen and were subcloned in TOPO TA. Two independent clones were obtained, which had an identical 5′ end. The central parts of the dizzy cDNA were amplified with gene-specific primers and then subcloned (accession number of full-length cDNA: DQ423241).

We started the functional characterization of dizzy by confirming the insertion sites of the two dizzy P-element alleles, dizzy^EP388 and dizzy^l(2)k13720(abbreviated as dizzy^EP and dizzy^P). Strikingly, even though the insertion sites are very close(Fig. 1A), the phenotypic effects of the P-elements on viability are quite different: dizzy^EP is fully viable, whereas dizzy^P is recessive lethal(Lee et al., 2002). To understand the molecular basis for this discrepancy, we isolated full-length cDNAs of dizzy and defined the transcription start site. Thereby we mapped an additional exon, termed exon 0, which is located 5′ to the first exon annotated in FlyBase (Fig. 1A). The P-elements are inserted into exon 0, a few base pairs downstream of the transcription start site. The difference between dizzy^EP and dizzy^P phenotypes is probably due to the distinct properties of the P-elements. In contrast to dizzy^P, dizzy^EP has an hsp70 promoter at its 3′ end that could allow a sufficient level of dizzy transcription controlled by endogenous enhancers.

In order to obtain loss-of-function alleles of dizzy, we mobilized the P-element of dizzy^EP and obtained a series of small deletions in the dizzy locus. Seven representative deletions were used for further analysis. dizzy^Δ1 and dizzy^Δ5 removed genomic DNA upstream of the former insertion site, including the transcription start site(Fig. 1A). dizzy^Δ7 and dizzy^Δ8deleted the entire exon 0, the transcription start site, most of the putative dizzy enhancer regions and almost the complete first intron(Fig. 1A). dizzy^Δ12 left the transcription start site intact,but removed the original translation start site and about 1.3 kb of the ORF. Therefore, it could code for a truncated Dizzy protein that lacks the cNMP,N-terminal RasGEF and PDZ domain (Fig. 1B). dizzy^Δ3 and dizzy^Δ10 are associated with smaller deletions,affecting translation start and cNMP domain only. Concerning the adult phenotype, all alleles behave like amorphic alleles, as homozygous escapers show the same phenotypes as animals carrying one of these dizzyalleles in trans to the deficiency Df(2L)BSC5: eyes are rough and reduced in size, wing blades are bent downward, and the male genitalia are distorted, probably contributing to the male sterility(Lee et al., 2002; Wang et al., 2006). Consistent with this genetic argument for the amorphic nature of the alleles is our finding that alleles with deleted transcription start sites lacked endogenous dizzy RNA expression in the embryo and therefore appeared to be null alleles (Fig. 1D). The normal expression of dizzy in the embryo was ubiquitous and of a comparably low level (Fig. 1C).

Next we tested the function of dizzy for macrophage migration in embryos either homozygous for any of the dizzy^Δnalleles, hemizygous for dizzy^Δn, or trans-heterozygous for dizzy^Δn and dizzy^P. In all cases the migration of macrophages in the embryo was similarly disturbed, specifically the migration along the ventral nerve cord (VNC). Macrophages from the anterior and from the posterior part of the embryo migrated along the midline of the VNC toward each other. Then at stages 13 and 14, wild-type macrophages surrounded the entire midline of the VNC (Fig. 2B). By contrast,macrophages of dizzy mutant embryos failed to migrate properly and did not succeed in completely surrounding the midline of the VNC at this time(arrowheads in Fig. 2A,K). In some embryos, the resulting ventral gap persisted until even later stages,whereas in others the defect disappeared(Fig. 2E). In addition,macrophages were found in an aberrant, dorsal position beneath the amnioserosa in mutant embryos after germ band retraction(Fig. 2I,K). This indicates that these macrophages did not properly adhere to the posterior end of the germ band and failed to enter the posterior germ band before and during its retraction. Later, during the phase of `central spreading' at stage 14,macrophages in wild-type embryos migrated laterally and ended up rather evenly distributed throughout the interstitial space at stage 15(Fig. 2F,H). By contrast, dizzy mutant macrophages did not achieve that even distribution, and the posterior-ventral part of the embryo contained less macrophages than the equivalent area of a wild-type embryo (Fig. 2E-H). Thus, dizzy mutant macrophages appeared to be slower and did not reach their destinations in time. Other aspects of the migration, such as the migration along the dorsal epidermis, were not noticeably affected in dizzy^Δn mutants.

dizzy is required not only for proper macrophage migration, but also for the cells to adopt their normal size and shape. In dizzymutants, macrophages formed smaller protrusions than those in wild-type embryos (Fig. 2C,D,G,H). The average lengths of the protrusions per cell were about 5 μm at stage 14 at lateral positions in fixed preparations of wild-type compared with less than half this size in dizzy mutants(Fig. 3). Hence, the analysis of the dizzy mutant phenotype demonstrates that dizzy is required for proper cell migration and proper cell form of macrophages during Drosophila embryogenesis and indicates a function of dizzyin cell adhesion.

The phenotype observed for macrophages in dizzy mutants could be due to a function of dizzy in the macrophages themselves or might be indirect, due to a requirement of dizzy in other tissues. We therefore specifically reduced dizzy function in the macrophages by expressing a dsRNA-fragment of dizzy. This expression led to the same phenotype as seen in the dizzy mutants: macrophage migration was severely delayed, macrophages failed to reach the posterior end of the germ band in time (Fig. 2M), and the cellular protrusions were smaller than in wild type(Fig. 3). We therefore conclude that the activity of dizzy is required within the macrophages for proper cell shape and motility.

Next we wondered how crucial dizzy activity is for cell shape and cell migration, and we specifically overexpressed dizzy in macrophages (Fig. 4). For this purpose we used the EP-allele dizzy^EP and directed the expression by srph-gal4 (termed dizzy^h.EPhereafter). The overexpression seen in dizzy^h.EP was confined to the dizzy transcription unit: genes neighboring dizzy, such as Pez or Cpr, were not influenced by the dizzy^EP allele. Macrophages in dizzy^h.EP embryos commenced their migration normally(Fig. 4A,B) but then formed very long protrusions, in the range of 20 μm per cell(Fig. 3, Fig. 4C,D). In addition, the protrusions of different macrophages contacted each other, resulting in the formation of a network that spanned the nervous system in a dorsoventral direction (Fig. 4C,E) or that was seen below the epidermis at lateral positions(Fig. 4G,I). In wild-type embryos, the macrophages migrating along the midline or beneath the dorsal edge of the epidermis had smaller protrusions and formed fewer or no contacts with each other (Fig. 4H,J).

Astonishingly, the motility of the macrophages was not significantly affected when dizzy was overexpressed from one copy of dizzy^EP, although the cell shape was changed dramatically. The cells migrated along their normal paths and made contacts with their normal substrates. Also, when followed by time-lapse microscopy, the cells showed the normal average speed of about 2 μm per minute(Fig. 5; see Movies 1-4 in the supplementary material). However, when the level of dizzy expression was increased further, by expression from two copies of dizzy^EP, migration of macrophages was slowed down(Fig. 4K). In embryos after stage 13, a significant gap remained at the VNC that was not closed later in embryogenesis. We conclude that the dizzy level is of crucial importance for regulation of cell shape and migration of macrophages.

Using time-lapse video microscopy, we intended to assess the origin of the large protrusions of dizzy^h.EP embryos. Live macrophages are much bigger than expected from their appearance in fixed material(Fig. 5)(Paladi and Tepass, 2004; Stramer et al., 2005). They have large, very dynamic lamellopodia, which extend in the direction of migration, multiple small filopodia and a short tail, depending on their state in the migrational cycle. These structures are not well preserved during the fixation procedure used for immunohistochemistry. Live macrophages overexpressing dizzy showed protrusions with principally the same size and the same dynamics as wild-type cells, but beyond that they also had the long cellular extensions seen in fixed preparations(Fig. 5; see Movies 1 and 3 in the supplementary material). Time series indicated that these extensions were not formed as independent entities, but originated either from retracted lamellopodia or more frequently from the tail of migrating cells(Fig. 5). This latter aspect contributes to the formation of the net-like appearance of the macrophages: in a group of cells, often only one cell at a time moved along and, upon dizzy overexpression, stayed in contact with the other, remaining cells of the group by its extended tail. Also at later stages, the long extensions were maintained in parallel to lamellopodia and did not disturb the normal local mobility of the cells (see Movie 3 in the supplementary material). We therefore suggest that dizzy contributes to the stabilization of cellular protrusions.

The phenotypes seen in macrophages of dizzy mutants, of ds.dizzy embryos and of embryos overexpressing dizzy suggest that Dizzy modulates adhesive properties of macrophages during their migration, and that the cell shape changes of macrophages reflect this modulation. As dizzy is predicted to be a GEF, based on its protein structure, we investigated by genetic means whether dizzy acts upstream of one of the small GTPases Cdc42, Rho, Rac or Rap1. Among these, we found Rap1 to be essential for the activity of Dizzy in macrophages. When we overexpressed dizzy in macrophages of zygotic rap1 mutants,these macrophages showed a substantial rescue of the cell shape changes induced by Dizzy (Fig. 6F,G). This rescue effect became even more evident in rap1 mutants that expressed dizzy from two copies of dizzy^EP in macrophages. While macrophages in these embryos looked similar to wild-type macrophages (Fig. 6I), the macrophages in rap1⁺ embryos expressing dizzy to the same extent formed a dense net of cells beneath the dorsal epidermis at the time of dorsal closure and in other regions of the embryo(Fig. 6H). This supports the argument that Dizzy acts upstream of Rap1. If Dizzy and Rap1 act in the same pathway, Rap1 overactivity should lead to a similar phenotype to the overactivity of Dizzy. To test this we expressed dominant-active rap1^V12 in macrophages. This expression led to significantly enlarged cellular protrusions(Fig. 3) and to a dense net of cells along the dorsal edge of the epidermis(Fig. 6J), very much like overexpression of dizzy from two copies of dizzy^EP (Fig. 6H). Thus, Dizzy and Rap1 appear to act in the same pathway. The amount of Rap1 was not the limiting factor in this respect, as overexpression of wild-type rap1 did not affect migrating macrophages(Fig. 6K).

If Rap1 acts downstream of Dizzy and is the relevant target for Dizzy activity, the lack of Rap1 function should give the same phenotype as the lack of Dizzy. This hypothesis is not easily tested, because Rap1 has a strong maternal contribution (Asha et al.,1999). Zygotic rap1 mutants had only a very subtle and transient macrophage migration phenotype(Fig. 6B), far weaker than the phenotype caused by the amorphic dizzy alleles. An influence on the size of the cellular protrusions was not detectable. After germ band retraction, the migration along the midline of the VNC was slightly disturbed,resulting in a small area that was free of macrophages at stage 14(Fig. 6B). This defect was compensated during later development. Other aspects of migration, such as the migration along the dorsal edge of the epidermis, were not noticeably affected in zygotic rap1 mutants. By contrast, embryos lacking both maternal and zygotic rap1 contribution are severely affected in all aspects of morphogenesis, including gastrulation (Asha et al., 1999), making it impossible to judge the specific function of rap1 for macrophage migration. We therefore used the dominant-negative Rap1^N17 to test macrophage-specific function of rap1. In embryos expressing Rap1^N17 in macrophages, the migration was severely disrupted. Few macrophages reached the posterior region of the embryo, and a major gap remained in their distribution(Fig. 7A). Additionally, the cells were smaller and had smaller extensions than wild-type cells, similar to dizzy mutant macrophages (Fig. 7C). Therefore, rap1 is unambiguously essential for proper migration of macrophages and has a similar influence on cell shape to dizzy.

Next we wanted to know whether Dizzy acts in vivo like a GEF for Rap1. To test this we simultaneously expressed Dizzy and Rap1^N17 in migrating macrophages. Rap1^N17 is thought to act as a dominant-negative form of Rap1 by sequestering the limited amount of activating GEF. If Dizzy acts as a GEF for Rap1, the simultaneous expression should rescue the rap1^N17 phenotype. Indeed, while Rap1^N17 on its own strongly disrupted the migration of macrophages,the simultaneous expression of Dizzy and Rap1^N17 substantially rescued the phenotype of Rap1^N17. Only a minimal gap was seen in the distribution of the macrophages around the VNC at stage 14, and the cells formed protrusions larger than normal (Fig. 7B,D).

Taken together, these data reveal a novel function of Rap1 in macrophages during migration. They show that active Rap1 is able to provoke cell shape changes in macrophages in a similar way to Dizzy. We have also demonstrated that Dizzy activity requires Rap1 function and have provided evidence that Dizzy is either itself a GEF for Rap1 or activates one.

Dizzy activity leads to cell shape changes in migrating macrophages, and we assume that these changes reflect modulations of adhesive properties of macrophages. Therefore, we tested whether βPS integrins are required for the cell shapes induced by Dizzy activity. mys mutants solely lacking the zygotic βPS gene function did not have a phenotype in macrophage shape or migration during embryogenesis(Fig. 8D). The mysrequirement became noticeable only in germline clones also lacking the maternal contribution (data not shown). Nevertheless, dizzyoverexpression did not lead to large protrusions and the net-like appearance of macrophages in zygotic mys mutant embryos(Fig. 8A,B). The cells had the appearance of wild-type macrophages regarding size and protrusions(Fig. 3, Fig. 7B). This shows that Dizzy requires the full function of βPS and indicates that integrins act downstream of Dizzy.

Our results above suggest that Dizzy acts via Rap1 to induce the changes of cell shape in macrophages. One therefore would expect that Rap1 stands between Dizzy and the integrins in the genetic hierarchy and that the changes in cell morphology caused by dominant-active Rap1^V12 also depend on integrin function. Indeed, in zygotic mys mutants Rap1^V12overexpression does not suffice in enlarging the protrusions(Fig. 3) or changing the cell shape of macrophages (Fig. 8F)as it does in mys⁺ embryos(Fig. 3, Fig. 8E). Thus, we conclude that in migrating macrophages Dizzy acts upstream of Rap1, and Rap1 in turn regulates cell adhesion and cell shape via a pathway that requires the function of βPS integrins.

We identified the Drosophila PDZ-GEF dizzy as an essential gene for cell shape regulation and cell migration of macrophages in Drosophila embryos. On the cellular level, Dizzy activity appears to contribute to the stabilization of cellular protrusions such as the cell tail. We showed that it requires the function of the small GTPase Rap1 and provided evidence that Dizzy behaves like a GEF for Rap1 during macrophage migration in vivo. Furthermore, the activities of Dizzy and Rap1 require the function ofβPS integrins to induce cell shape changes. These data provide the first evidence of a pathway via a PDZ-GEF and Rap1 that regulates integrin-dependent activity during embryogenesis and expands the current picture of the migration of macrophages in the Drosophila embryo(Brückner et al., 2004; Cho et al., 2002; Stramer et al., 2005; Tepass et al., 1994).

We characterized the dizzy locus and found that lack of dizzy function results in a relatively subtle phenotype in the embryo: the macrophages had cellular protrusions about half the size of normal cells in fixed preparations and did not reach their destination in the central region of the embryo in time (Figs 2, 3). Surprisingly, the difference in the size of the protrusions was not obvious in live dizzy macrophages. We assume that the difference in fixed preparations is caused by a reduced stability of protrusions in dizzymutants that leads to a diminished preservation during fixation. Also the measurement of the speed of individual dizzy macrophages showed no significant reduction in their average speed during stages 12 and 13. However,due to the stochastic behavior of individual cells during migration it is difficult to assess in live embryos whether the cells commence their migration too late or whether they stop prematurely.

The observed phenotypes in macrophage shape and migration are due to the lack of dizzy function, as they are seen in embryos homozygous or hemizygous for dizzy^P or dizzy^Δn alleles, i.e. they are independent of the genetic background. Moreover, the revertant dizzy^P-RV4, which had lost the P-element insertion, no longer showed the phenotypes. Furthermore, the dsRNA-induced reduction of dizzy function specifically in the macrophages led to the same phenotypes. Here the extent is more variable than in mutants,presumably due to the intrinsic inhomogeneity of the UAS-Gal4 system. In any case, we have no indication that inappropriate apoptosis contributes to the phenotypes. The number of macrophages is normal in the mutants throughout embryogenesis. A rescue by the expression of the pancaspase inhibitor p35 is not observed (data not shown).

Overexpression of dizzy in the macrophages led to effects on cell shape complementary to the dizzy loss of function. The protrusion length per cell was increased by a factor of four, and more protrusions per cell were visible. In live cells we saw extraordinarily long, stable extensions, which apparently connected the macrophages with each other. These protrusions were formed in addition to normal tails and leading edges,resulting in the increase of protrusion number per cell seen in fixed preparations. They seemed to originate from lamellopodia, which were incompletely retracted or often from cell tails that failed to retract properly. We propose that Dizzy normally stabilizes the protrusions,explaining the smaller protrusions in fixed preparations of dizzymutants. When overexpressed, dizzy might inhibit the proper disassembly of cellular protrusions, explaining their vast enlargement and persistence. Whether Dizzy acts in a particular subcellular location, e.g. tail versus leading edge, remains to be seen.

Experiments in mammalian cell culture demonstrated that PDZ-GEFs specifically activate the small GTPases Rap1 and Rap2(de Rooij et al., 1999; Gao et al., 2001; Kuiperij et al., 2003; Liao et al., 2001; Rebhun et al., 2000). Also recent reports from Caenorhabditis and Drosophila place PDZ-GEFs upstream of Rap1 (Lee et al.,2002; Pellis-van Berkel et al., 2005; Wang et al.,2006). Our data suggest that the PDZ-GEF Dizzy is a GEF for the small GTPase Rap1 in Drosophila macrophages: the level of Dizzy expression is crucial (Figs 2, 3 and 4), the cell shape changes induced by high levels of Dizzy in macrophages are Rap1-dependent(Fig. 6), and high levels of Dizzy rescue the phenotype of dominant-negative Rap1^N17(Fig. 7). We assume that in migrating macrophages most, if not all, of the activity of Dizzy goes through Rap1 for two reasons: (1) the appearance of macrophages of Dizzy overexpressing embryos and of embryos overexpressing dominant-active Rap1^V12 were very similar (Figs 3, 4, 6); (2) overexpression of Dizzy was effective only in rap1⁺ embryos(Fig. 6). The lack-of-function phenotypes of dizzy and rap1 were quantitatively different: rap1 mutants had a far weaker macrophage migration defect than dizzy mutant embryos because of the compensation by maternal contribution. However, dominant-negative Rap1^N17 caused a strong migrational phenotype (Fig. 7),supporting our notion that Dizzy and Rap1 act in the same pathway.

The cell shape changes induced by Dizzy and Rap1 in macrophages are strictly dependent on the function of βPS integrins(Fig. 8). Already in mys mutants that solely lack the zygotic contribution of βPS integrin, neither Dizzy nor dominant-active Rap1^V12 were able to trigger the formation of long cellular protrusions. Thus, the maternal contribution of mys in these embryos is insufficient to allow the Dizzy or Rap1 gain-of-function phenotype in the macrophages, although this contribution suffices for normal shape and migrational behavior of the cells. Apparently, the dosage of integrins is decisive in what Dizzy/Rap1 execute in the cell. Therefore, it is reasonable to assume that the stabilization of the protrusions is achieved by a positive modulation of cell adhesion.

There are two possible roles for integrins in the Dizzy/Rap1-dependent stabilization of cellular protrusions: (1) integrins are the relevant targets of the pathway, and a modulation of their activity is responsible for the cell shape changes; (2) the prime target is another cellular component, and integrins are merely required to allow the stabilization. We favor the former hypothesis, as several lines of evidence from cell culture experiments suggest that Rap1 acts in such a fashion and regulates integrin activity via an`inside-out' signaling pathway (Bos et al.,2003; Sebzda et al.,2002; Tohyama et al.,2003). Components of that pathway might be Rap1 effectors that provide the link between Rap1 and integrins. Two such effectors, RAPL and RIAM, have been described in vertebrates(Katagiri et al., 2003; Lafuente et al., 2004), but the function of the corresponding Drosophila orthologs are not yet known.

Above we proposed that Dizzy exerts a positive influence via Rap1 on the activity of integrins in Drosophila macrophages. This influence could be an immediate activation of integrins, but the nature of that activation is unclear. The level of expression or the spatial distribution of βPS integrin is not obviously changed in macrophages overexpressing Dizzy or Rap1^V12 compared with wild-type cells. We observed neither clustering nor elevated protein expression of integrins. Likewise, the level of phosphorylated focal adhesion kinase was unchanged (pFak^Tyr397,data not shown). The phosphorylation of Fak is dependent on integrin-mediated adhesion and therefore often used as a marker for integrin activity (for a review, see Mitra et al.,2005). Alternatively, the positive influence of Dizzy/Rap1 on integrin activity might be due to a repression of de-adhesion of integrins. This could be achieved by interaction with the small GTPase RhoA, which in monocytes is required for the retraction of the cell tail and appears to act by delimiting the adhesion to integrin ligands and by keeping the local levels of integrins low (Worthylake et al.,2001). However, initial experiments do not indicate a clear genetic interaction between RhoA and Dizzy in Drosophila during the migration of macrophages.

Whether as immediate activation or as repression of de-adhesion, the signaling via Dizzy/Rap1 to integrins appears to be permissive rather than instructive in nature for the guided motility of Drosophilamacrophages. We have no indication that the directionality of the migration is changed upon overexpression of Dizzy. The additional protrusions are not an indication of perturbed polarity, nor do they represent multiple leading edges. Also, macrophages in dizzy loss-of-function mutants appear to have normal polarity and migrate along the normal paths. They do not reach the central part of the embryo in time, but behave otherwise normally. Therefore,we conclude that the Dizzy/Rap1 pathway does not contribute to its directional action, but makes the migratory machinery of the cell efficient.

We are grateful to B. Boettner, N. Brown, N. Franc, U. Gaul, A. Jacinto, M. D. Martin Bermudo, A. Müller, R. Palmer and the Bloomington Stock Center for providing fly stocks and reagents. We thank Elke Müller and Tanja Mader for expert technical assistance. This work was supported by the DFG as project B11 of the SFB 446 `Zellverhalten der Eukaryoten'.