A dual role for the βPS integrin myospheroid in mediating Drosophila embryonic macrophage migration

Integrins are αβ heterodimeric cell-surface receptors that bind specific extracellular matrix (ECM) proteins or counter receptors on other cells (Hynes, 2002; Humphries et al., 2006). Besides their role in regulating cell adhesion, integrins transduce signals inside the cell that regulate actin cytoskeletal rearrangements, cell migration, cell shape, gene expression, cell proliferation and survival (Legate et al., 2009). These signals integrate with signals transduced from growth factors, cytokines and other transmembrane receptors to regulate numerous anchorage-dependent cellular properties (Hood and Cheresh, 2002).

Adult mature macrophages are strategically located throughout the body tissues where they play a central role in protecting the host and maintaining tissue homeostasis. Over the last few years, it has been appreciated that integrins are essential in regulating many of these macrophage responses, such as transmigration into the inflammatory site, migration across tissues, cytokine secretion and phagocytosis (Abram and Lowell., 2009). However, recent work has revealed strong discrepancies in the role that integrins play during leukocyte migration in two-dimensional (2D) versus three-dimensional (3D) environments. Live imaging of leukocytes migrating in vivo as well as artificial 3D matrices of fibrin and collagen has shown that while integrins are essential to overcome tissue barriers, they are dispensable for interstitial leukocyte migration. These studies highlight the need to study leukocyte migration in the natural context of a living organism (Lämmermann et al., 2008).

Embryonic macrophages play important roles throughout embryonic development, sculpting and promoting organogenesis by clearing apoptotic cells and depositing ECM molecules. Central to their role in embryogenesis, embryonic macrophages must migrate and disseminate throughout the embryo. Yet despite their pivotal role during embryonic development, little is known about how embryonic macrophage migration is regulated. Recently the use of fluorescently labelled transgenic animals and real-time in vivo confocal microscopy has facilitated the study of embryonic macrophages in vivo within the complex 3D environment of embryos (Grabher et al., 2007). In this respect, Drosophila embryonic macrophages (haemocytes) have emerged as a powerful system to study embryonic macrophage biology in vivo (Evans and Wood, 2011). Deriving from the head mesoderm, haemocytes are highly migratory cells, achieving an even distribution throughout the embryo by the end of embryogenesis (Tepass et al., 1994). During their stereotypical migrations Drosophila embryonic macrophages contact several diverse tissues surrounded by components of the ECM. Yet despite the fact that integrins are key regulators of cell migration over ECM substrates, and regardless of their central role in adult macrophages, further characterization of their requirement during embryonic macrophage migration is needed.

Here, we use Drosophila to analyse the function of integrins in embryonic macrophages. Eight β- and 18 α-subunits have been characterized in mammals, while the Drosophila genome encodes two β-subunits (βPS and βν) and five α-subunits (αPS1–5). βPS is widely expressed and forms heterodimers with all α-subunits, while the βν integrin subunit is predominantly expressed in the midgut endoderm (Brown, 2000). The αPS2 integrin subunit, inflated, is required for the invasive movement of embryonic macrophages from the head region into the tail (Siekhaus et al., 2010). In addition, integrins appear to act downstream of the Rap1 guanine nucleotide exchange factor (GEF) dizzy during haemocyte migration (Huelsmann et al., 2006), since removal of βPS integrin function can rescue haemocyte migration defects due to overexpression of dizzy. However, while haemocytes from dizzy mutant embryos displayed migration defects, a clear function for βPS integrins in this process has not yet been clearly demonstrated.

By using live imaging we have dissected cell autonomous and non-autonomous roles for integrins during haemocyte dispersal within the embryo; the βPS integrin myospheroid is required for correct development of the 3D environment in which the haemocytes are migrating and also within the haemocytes themselves for their migration. In addition we show that the αPS1 and αPS3 subunits act redundantly to mediate myospheroid function during haemocyte migration. Using high resolution, live imaging, we discover a requirement for myospheroid in the coordination of the actin and microtubule dynamics within migrating haemocytes and that a loss of integrin function results in defective inflammatory migrations to wounds as well as a failure in contact repulsion between migratory haemocytes.

In Drosophila embryonic haemocytes, αPS2 is specifically required for haemocyte movement into the tail (Siekhaus et al., 2010), indicating that either integrins are not required for other haemocyte developmental migrations or that other αPS integrins are involved. We sought to investigate the role of integrins in haemocyte migrations more broadly by removing the main β-integrin subunit, encoded by myospheroid (mys). As integrin subunits must form αβ heterodimers in order to be transported to the cell surface (Leptin et al., 1989), removing the main β-integrin subunit disrupts most integrin function within the embryo. As βPS is maternally deposited within the embryo, we generated maternal zygotic mutants to ensure complete elimination of integrin function (subsequently referred to as mys^M/Z). Initial dispersal of haemocytes from the head mesoderm, where they are specified, towards the cypeolabrum, the gnathal buds and the anterior ventral nerve cord (VNC) were unaffected by loss of mys (Fig. 1A,E). At this stage, ventrally located haemocytes undertake pvf-directed migration along the developing VNC of the embryo to occupy the length of the midline by stage 13 (Wood et al., 2006) (Fig. 1B). In mys^M/Z mutant embryos haemocyte migration at this stage is severely disrupted, with haemocytes limited to the anterior and posterior ends of the VNC (Fig. 1F). This disruption persists through to stage 15, indicating that haemocyte migration is not slowed but inhibited (Fig. 1H). Removal of mys also affected haemocyte migration along the dorsal edge of the epidermis and the visceral mesoderm (Fig. 1C,G and data not shown), and resulted in accumulation of haemocytes between the amnioserosa and yolk regions (Fig. 1F, arrowhead). These haemocyte migration defects were phenocopied in embryos lacking both maternal and zygotic Talin, a key component of integrin mediate adhesion (supplementary material Fig. S1B). In addition, removal of zygotic and maternal βv integrin function did not enhance the mys phenotypes, indicating that βv integrin plays no role in haemocyte dispersal (data not shown).

myospheroid is able to form heterodimers with all five α-subunits present in Drosophila (Brown, 2000). In order to identify which α-subunits are involved in regulating haemocyte migration, we analysed haemocyte migration along the VNC in embryos lacking the different αPS subunits. Integrin α-subunit genes show no maternal contribution (Brower et al., 1995; Stark et al., 1997), therefore zygotic mutants were used. Analysis of embryos mutant for αPS1 and αPS3 revealed a slight disruption of haemocyte migration along the VNC, with embryos showing midline segments devoid of haemocytes (supplementary material Fig. S2B,C), and in contrast to mys mutant embryos, this phenotype was not observed in stage 15 embryos (data not shown). These results show that migration was delayed but not inhibited in the absence of either of these α subunits. Only with removal of both αPS1 and αPS3 function did the posterior half of the VNC remain free of haemocytes until the end of embryogenesis (supplementary material Fig. S2D), indicating that these two α subunits function redundantly to mediate mys function during haemocyte migration along the VNC in the developing embryo.

Integrins have been previously shown to regulate cell survival in different cell types in a variety of species (reviewed in Vachon, 2011). Furthermore loss of trophic support in pvr mutants contributes to defective developmental dispersal of haemocytes (Bruckner et al., 2004). To rule out increased haemocyte death in causing the reduced number of haemocytes found on the midline, we induced haemocyte-specific expression of the apoptotic inhibitor p35 in mys mutant embryos. Blocking apoptosis failed to rescue the strong defect in developmental dispersal of these embryonic haemocytes (data not shown), indicating that mys function plays no significant role in regulating the survival of these cells and that the dispersal defects seen are not the result of reduced haemocyte numbers.

In the Drosophila embryo, mys genetically interacts with the growth cone repellent Slit to mediate axonal guidance within the developing CNS (Stevens and Jacobs, 2002). As slit mutant embryos exhibit disrupted haemocyte migration due to a failure in separation of the VNC from the overlying epithelium (Evans et al., 2010), we wondered whether this non-autonomous defect occurred in and contributed to haemocyte migration defects seen in mys mutant embryos. To examine this possibility, we used dye injections to visualise the spatial constraints encountered by haemocytes on the ventral side of the embryo (Evans et al., 2010). Dye injections of stage 15 WT embryos indicate separation of the epithelium from the VNC along the entire length of the midline (Fig. 2A, n = 17). In contrast, in mys mutant embryos the dye fails to permeate along the length of the embryo, indicating areas where the epithelium has failed to detach from the VNC (Fig. 2B, n = 14). In 64% of mys mutant embryos injected, presence of the dye coincides with areas occupied by haemocytes (Fig. 2Bi), suggesting that haemocyte migration defects in these embryos may be the result of spatial constraints within the embryo. However, in the remaining 36% of mys mutant embryos, haemocyte progression along the VNC is halted before the dye becomes more restricted, indicating that in some embryos, other reasons may underlie haemocyte migration failure (Fig. 2Bii).

In order to investigate a cell-autonomous requirement for integrins within haemocytes we expressed RNAi constructs specifically within the haemocytes using the srp-HemoGAL4 driver (Bruckner et al., 2004). Expression of RNAi transgenes for either mys or talin was sufficient to mimic haemocyte migration defects seen with loss of mys function (Fig. 2D; supplementary material Fig. S1C), demonstrating an additional requirement for integrins within the migrating haemocytes. Integrins have roles in both adhesion and in promoting intracellular signalling in response to environmental cues (Huttenlocher and Horwitz, 2011). In order to determine the primary requirement for integrins in migrating haemocytes we utilised the Tor^D/β_cyt fusion protein. Tor^D/β_cyt, consists of the cytoplasmic tail of mys fused to the extracellular and transmembrane domains of a dominant gain-of-function allele of the Torso receptor tyrosine kinase, and has been shown to block ECM adhesion mediated by endogenous integrins while activating signalling even in the absence of adhesion (Martin-Bermudo and Brown, 1999; Narasimha and Brown, 2004; Tanentzapf et al., 2006). Ectopic expression of Tor^D/β_cyt within wild-type haemocytes using the srp-HemoGAL4 driver also phenocopied loss of integrin function (Fig. 2E). These findings suggest that one of the critical functions for integrins in haemocyte developmental dispersal appears to be adhesion. To investigate in more detail the contribution of integrin requirement in both the haemocytes and the surrounding tissue during their migration along the ventral midline, we used the GAL4 system to express the βPS subunit in either haemocytes, or in their migratory substratum, or both in embryos that lack mys function. To quantify the rescue effects we defined four distinct phenotypic classes according to the number of neuromeres of the VNC devoid of macrophages and determined for each genotype the distribution among these classes (Fig. 2F–L). Expression of mys solely in the haemocytes of mys^M/Z mutant embryos resulted in a substantial rescue of the migration phenotype (Fig. 2L, srp >). Rescue of haemocyte migration was further improved when mys was coexpressed in the ventral midline, consistent with a requirement for this integrin subunit in the tissues surrounding the haemocyte at this stage of development (Fig. 2L, sim >srp >). In contrast, mutant embryos expressing mys only in the midline showed a very limited rescue of macrophage migration (Fig. 2L, sim >). Expression of mys in the midline glia cells, using slit-GAL4, was unable to rescue the migration defect (Fig. 2L, slit >) and did not enhance the rescue effect seen with expression of mys in the haemocytes (Fig. 2L, slit >srp >, compared to srp >). Taken together these results demonstrate that the βPS integrin subunit is required for migration of haemocytes along the ventral midline via a specific requirement in the haemocytes to regulate their adhesions as well as a role in the ventral nerve cord in regulating VNC–epithelial separation.

To investigate in more detail the cell autonomous role for integrins in haemocytes we used live imaging of mys mutant haemocytes as they undergo their normal developmental dispersal in the embryo. From stage 13 wild-type haemocytes undergo a segmented and highly directional lateral migration away from the ventral midline (Fig. 3A,C) (Wood et al., 2006). Even using mys zygotic mutants, in which a low level of maternal protein is present, this developmental migration was almost completely abolished (Fig. 3B,D). Tracking individual cells revealed that only a small fraction of haemocytes migrated laterally in the absence of mys function and the few that did moved at significantly slower speeds than wild-type cells (Fig. 3E,F). This is in contrast to slit mutant embryos in which laterally migrating haemocytes migrated at speeds comparable to those observed in wild-type embryos (Evans et al., 2010). Therefore, although environmental constraints may account for the reduction in numbers migrating laterally, the reduction in migration speed of those that do is likely due to a haemocyte-specific requirement for mys. Similarly, mys mutant haemocytes moved at almost half the speed of wild-type cells when migrating randomly at stage 15 (Fig. 3G). However, we did not detect differences in the directionality of migration between wild-type and mys mutant haemocytes during lateral migration or random migration, suggesting that the polarity machinery and the ability to decipher migratory cues remained intact in these cells (data not shown). To analyse cell migration velocity in mys mutant embryos in the absence of spatial constraint, we used laser ablation to create epithelial wounds in the anterior trunk of the embryo, in areas populated with haemocytes. In mys mutant embryos haemocytes were competent to respond to an inflammatory cue (Fig. 3J,K), but there was a significant reduction in the number of haemocytes recruited to wounds at the early stages of this inflammatory migration (Fig. 3L). At later time points the inflammatory response recovered to that of WT embryos. This lag in recruitment is likely a reflection of the significant reduction in migration speed exhibited by mys embryos (Fig. 3M), as the highly directional routes taken by haemocytes towards the wound are comparable to those seen in wild type (data not shown), again highlighting a role in providing a driving force for haemocyte migration rather than a pathfinding role for integrins.

Together these results indicate a haemocyte-specific requirement for integrins in both directed and random migrations. In all types of migration analysed the migration speed but not the directionality of haemocytes was affected by loss of mys function. Consistent with the above data this suggests that integrin-mediated adhesion, rather than signalling, is the predominant functional requirement during haemocyte migration.

Recent studies have shown that microtubule-mediated contact repulsion acts as a major driving force behind lateral migration of haemocytes from the midline and the subsequent maintenance of their even distribution (Stramer et al., 2010). In order to address whether integrins may play a role in this important process we used time-lapse imaging of haemocytes expressing cytoplasmic GFP to analyse the ability of mys mutant haemocytes to undergo contact repulsion. In wild-type embryos haemocyte cell–cell contact arrested migration and stimulated mutual repulsion (Fig. 4A) (see also Stramer et al., 2010). In contrast, mys mutant haemocytes remained in contact for prolonged periods; on average haemocytes spent six times longer in contact than wild-type equivalents (Fig. 4B,C). These results demonstrate a role for integrins in enabling haemocyte contact repulsion. Previous studies have shown that contact repulsion is driven by realignment of the polarized microtubule cytoskeleton in contacting cells. We therefore analysed microtubule dynamics in mys mutant haemocytes.

In Drosophila the use of fluorescent probes to label actin (mCherry–Moesin) and microtubules (GFP–CLIP170), has shown that embryonic haemocytes assemble a polarised array of microtubules that extend into lamellipodia during migration. Coalescence of stabilized microtubules drives the formation of a ‘microtubule arm’ that orients the cell in the direction of migration (Stramer et al., 2010). Formation of the microtubule arm is pivotal in polarizing haemocytes in response to external cues that drive their developmental migrations and migration towards wounds as well as enabling haemocytes to undergo contact repulsion (Stramer et al., 2010). Interestingly, in vitro experiments have revealed a role for integrins in microtubule stabilisation downstream of the small GTPase Rho (Palazzo et al., 2004). To investigate the microtubule dynamics in mys mutant haemocytes we timelapse-imaged haemocytes expressing GFP–CLIP170. Visualization of wild-type haemocytes at stage 15 reveals bundling of microtubules into an arm and close co-ordination of microtubule arm disassembly and lamellipodial retraction (Fig. 5A; supplementary material Movie 1) (Stramer et al., 2010). Analysis of mys mutant haemocytes revealed that while microtubules polarised and initially formed an arm, this structure was not maintained and rapidly collapsed within persisting lamellipodia, indicating that a loss of integrins reduces the stability of microtubule arms (Fig. 5B; supplementary material Movie 2). Despite this apparent instability there were a comparable number of microtubule arms formed by wild-type and mys mutant haemocytes (Fig. 5C), consistent with a role for integrin in maintaining a polarized microtubule arm. Closer analysis of microtubule dynamics revealed that, whereas in wild-type haemocytes microtubule arm disassembly was almost always triggered by contact with another haemocyte, in mys mutant embryos microtubule arm loss predominantly occurred independently of cell-cell contact (Fig. 5D). Microtubule arm alignment is essential for contact repulsion (Stramer et al., 2010). Therefore, this microtubule arm loss phenotype, coupled with the more general deficits in cell translocation, which would perturb movement of colliding haemocytes away from each other post-contact, presumably underlies the failures in contact inhibition also seen in mys mutant embryos.

Analysis of individual microtubule dynamics by high-speed microscopy and subsequent tracking of mCherry–CLIP to show movement of microtubule + tips towards the cell periphery showed that microtubules in mys mutant haemocytes protruded at slower speeds than in the wild type (Fig. 5E–G). In keratinocytes loss of integrin-linked kinase (ILK), a signalling protein that associates with the intracellular domain of integrins, results in a failure of peripheral microtubule to reach the cell cortex (Wickström et al., 2010). However, in haemocytes, loss of mys did not result in a significant reduction in the final distance microtubules polymerised into the lamellipodia (Fig. 5H), indicating that although slowed, individual polymerizing microtubules are as stable as in WT cells. Together this indicates a specific requirement for integrins in the stabilization of microtubules that are captured and bundled into the microtubule arm structure.

Cell migration is a highly orchestrated process, which requires co-ordination between the actin and microtubule cytoskeleton. Integrin engagement with the ECM recruits and activates numerous signalling molecules, which in turn can activate the actin polymerisation machinery, shown to be important in driving haemocyte migration (DeMali et al., 2003). To determine whether alterations in microtubule dynamics observed in mys mutant haemocytes could be the result of disruption to the actin cytoskeleton, time-lapse imaging of haemocytes expressing GFP–Moesin was conducted. Analysis of lamellipodia area over time revealed that mys mutant haemocytes form dynamic actin-rich lamellipodial protrusions comparable to those observed in the wild type (Fig. 6A–C). Cell culture experiments have shown that integrin-dependent attachment to the ECM can control transition from a spheroid to a flattened morphology, which underlies cell-spreading events that occur during cell migration (reviewed in Holly et al., 2000). The fact that mys haemocytes are able to make lamellipodial protrusions as dynamic and of the same size (Fig. 6F) as wild-type cells demonstrates that a loss of βPS does not affect the ability of haemocytes to spread in vivo. However, closer analysis of these protrusions revealed that the organization of the actin cytoskeleton within the mys haemocytes is altered with an increase in the number of microspikes (F-actin struts within lamellipodia) and filopodia (microspikes that extend beyond the lamellipodial leading edge) with respect to wild-type cells (Fig. 6D,E). Interestingly, this is in contrast to in vitro fibroblast experiments in which integrin binding triggers the formation of microspikes (Levy et al., 2003).

Embryonic macrophages play essential roles throughout embryonic development, clearing apoptotic corpses and secreting ECM components. In order to carry out these important functions, embryonic macrophages must disseminate throughout the embryo. As of yet very little is known about the molecular mechanisms governing these early macrophage migrations within the developing embryo. Utilising Drosophila haemocytes as a model system to study the role of integrins in mediating embryonic macrophage migration, we show a role for this family of transmembrane receptors in regulating several responses in embryonic macrophages: developmental dispersal, contact repulsion and chemotactic inflammatory migration towards wounds.

Integrins have always been proposed to play an essential role in the recruitment of macrophages to sights of infection or injury. However, recent work has shown that, although there is an absolute requirement for integrins in leukocyte migration in vitro, in vivo they are only required for crossing tissue barriers and are dispensable for interstitial migration within the lymph node (Lämmermann et al., 2008). Authors show that leukocyte migration is driven solely by expansion of the actin network, which drives protrusion of the leading edge independently of adhesion, allowing leukocytes to migrate autonomously in the wide variety of tissue environments encountered in vivo. These results challenge the classical view about the role of integrins, and cell adhesion in general, during cell migration within the context of a living organism. Our results show that, although integrins are not essential for the initial movements of haemocytes away from their point of origin and along the anterior ventral nerve cord, they are required for all subsequent developmental migrations within the embryo. It is likely therefore that the requirement for integrins in cell migration is cell type as well as environment specific. Whereas leukocytes show stochastic, swarming migration across a wide variety of extra cellular environments, Drosophila haemocytes undertake highly orchestrated migrations over tissues with a basement membrane. Therefore although performing common functions in vivo, environmental interactions may have forced these cell types to adapt different modes of migration. For example although leukocytes are able to switch to contraction-driven ‘squeezing’ through a confined 3D environment, fibroblasts rely solely on integrin-dependant anchorage to track along components of the extracellular matrix (Renkawitz and Sixt, 2010). This linkage-based guidance of migration could aid in the control of haemocyte developmental migration along pre-determined pathways. Interestingly, the ability of integrin mutant haemocytes to undergo very early dispersal from the head mesoderm to occupy the anterior region of the ventral midline indicates that during embryonic development haemocytes are also able to switch from an integrin-independent to an integrin-dependant mode of migration.

By analysing cell migration within the developing embryo, we have shown a dual requirement for integrins in both the migrating haemocytes and the surrounding tissues. Haemocyte migration along the ventral midline is dependent upon correct development of the VNC. In slit and robo1, 2 mutant embryos, haemocytes fail to progress along the midline due to a failure in the separation of the VNC and epidermis (Evans et al., 2010). Authors propose that this failed separation event may result from disruptions in axonal path finding and glial positioning within the VNC, leading to a failure of these structures to separate from the epidermis. We show here that mys mutant embryos also display a failure in the separation of the VNC and given that mys mutant embryos display axonal pathfinding defects (Stevens and Jacobs, 2002), the separation defect observed in mys mutants may be a consequence of this disrupted axonal wiring. Alternatively, integrin-dependant assembly of ECM components, as in several cellular contexts (Narasimha and Brown, 2004; Tanentzapf et al., 2007), may be required for correct separation of the VNC. Consistent with this, haemocytes fail to migrate along the VNC in laminin mutant embryos (Urbano et al., 2009). It would be interesting to see whether this migration defect may also be due to a failure in VNC separation.

Integrins allow traction in migrating cells by acting as a molecular clutch that links the actin cytoskeleton to the ECM, allowing the transmission of acto-myosin contraction to the substratum (Alexandrova et al., 2008). Consistent with this role for integrins, we find that removal of functional integrins from haemocytes results in a severe reduction in migratory velocity during both random and directional migration. In migrating leukocytes and fibroblasts, loss of integrins results in increased actin retrograde flow as a consequence of the integrin ‘clutch’ (Renkawitz and Sixt, 2010). Within 3D environments, integrin mutant leukocytes are able to overcome this loss of traction by increasing the polymerisation of actin, reaching migration velocities indistinguishable to WT. Although haemocyte spreading is not affected by integrin loss in vivo, perhaps attributable to the close confinement of haemocytes within the 3D context of the embryo (Tucker et al., 2011), we do however see subtle changes in the structure of the actin cytoskeleton. Surprisingly, the increase in the number of microspikes, actin bundles within the lamellipodia, seen in mys mutant haemocytes usually correlates with increased migration speed (P. K. Tucker, I. E. and W. W., unpublished data). If the compensatory mechanism observed in integrin mutant leukocytes is conserved in haemocytes, the increase in microspikes within lamellipodia could reflect an attempt to overcome slippage due to loss of the integrin ‘clutch’. It would therefore be interesting to compare retrograde flow in mys mutant haemocytes to determine whether the function of integrins between mammalian and Drosophila blood cells is conserved.

Applying this clutch theory could also aid in explaining the microtubule phenotype observed in mys mutant haemocytes. It is appreciated that migration requires close co-ordination between the actin and microtubule network, however the nature of these interactions during cell migration remains poorly understood. We have previously shown that microtubules are closely associated with actin microspikes as they extend and probe the lamellipodia of an advancing haemocyte (Stramer et al., 2010). An increase in actin retrograde flow within the lamellipodia could lead to the rapid and repeated collapse of the microtubule arm as seen in integrin mutants. Alternatively, these microtubule defects could reflect a direct requirement for integrins in regulating microtubule dynamics, via previously identified downstream targets such as ILK, FAK and diaphanous (Palazzo et al., 2004; Wickström et al., 2010).

In summary, our results highlight an essential role for integrins in mediating many embryonic macrophage functions, including developmental migration, inflammatory migration towards wounds and contact repulsion. The possibility of examining these processes in vivo in a genetically tractable organism such as Drosophila will significantly assist in the dissection of the molecular mechanisms by which integrins exert these different functions in this prominent cell type so important in many biological processes. Furthermore, in addition to their role in the clearance of apoptotic cells, macrophages are being increasingly recognised for their role in providing trophic support in many tissues, their contribution to tissue regeneration and their pivotal role in tumour angiogenesis and metastasis. Pro-metastatic roles have even been demonstrated by haemocytes in Drosophila cancer models (Cordero et al., 2010). Recent expression profiling of embryonic macrophages revealed similarities with tumour associated macrophages (TAM), independent of their tissue of origin (Rae et al., 2007). Hence, a better characterization of the biology of embryonic macrophages may also provide clues to TAM function in cancer and may lead to the identification of new molecular targets to inhibit their proangiogenic and protumoral activities in neoplasia and other diseases.

Flies were raised at room temperature. Embryos were collected from laying cages kept overnight at 25°C. For RNAi experiments laying cages were kept overnight at 29°C. The following stocks were used for fixing and staining to analyse defects in haemocyte developmental dispersal: mys^XG43 FRT101 (Bunch and Brower, 1992), rhea⁷⁹ FRT 2A (Prout et al., 1997), if^B4 (Brown, 1994), scab^IIG (Stark et al., 1997), mew^m6 (Brower et al., 1995), βν (Yee and Hynes, 1993), UAS-mys (Martin-Bermudo and Brown, 1996), UAS-Torso^D/β_cyt (Martin-Bermudo and Brown, 1999), UAS-p35 (Hay et al., 1994), srph-GAL4 (Huelsmann et al., 2006), srp-HemoGAL4 (Bruckner et al., 2004), sim-gal4 and slit-gal4 (Scholz et al., 1997), UAS-gfp^S65T, ovo^D1FRT101 (Bloomington). For the RNAi knockdown experiments, we used the following UAS-RNAi lines; UAS-RNAi mys and UAS-RNAi talin (VDRC).

For time-lapse imaging SerpentHemoGAL4 (srp-GAL4) (Bruckner et al., 2004), croquemort-GAL4 (crq-GAL4) (Stramer et al., 2005) and singed-GAL4 (sn-GAL4) (Zanet et al., 2012) were used to drive haemocyte-specific expression of the following UAS constructs (obtained from Bloomington Stock Center unless otherwise stated): UAS-GFP, UAS-GFP-Moesin (Dutta et al., 2002), UAS-mCherry-Moesin (a gift from Paul Martin, University of Bristol), UAS-mCherry-CLIP170 and UAS-GFP-CLIP170 (Stramer et al., 2010) and UAS-redstinger (Barolo et al., 2004). The constructs and drivers were used to produce the following genotypes; srp-GAL4,UAS-GFP-Moesin;UAScherryClip, srp-GAL4,UAS-GFP;crq-GAL4,UAS-GFP, srp-GAL4,UAS-mCherry-Moesin;crq-GAL4,UAS-GFP-CLIP170, srp-GAL4,UAS-mCherry-Moesin;crq-GAL4,UAS-mCherry-Moesin, sn-GAL4,UAS-LifeActGFP and srp-GAL4,UAS-redstinger;crq-GAL4,UAS-redstinger.

Antibody staining of embryos was performed using standard procedures. We used the following primary antibodies: Rb-GFP 1/100 (Molecular Probes), rabbit anti-β-Gal (1:6000; Cappel), rabbit anti-Srp (1:1000; (Riechmann et al., 1998), rabbit anti-Crq (1:1000; (Franc et al., 1996), mouse anti-βPS (1:300; DSHB, Iowa). Alexa-conjugated secondary antibodies used were Alexa Fluor 488 (green), Alexa Fluor 568 (red) (Molecular Probes™). For non-fluorescent staining, embryos were incubated in biotinylated secondary antibodies followed by incubation with Elite ABC complex (Vector Laboratories) and revealed with DAB (Gibco-BRL). Images were collected with a Zeiss Axioplan 2 microscope or a Leica TCS-SP2 confocal microscope.

Embryos were prepared and mounted as previously described (Wood and Jacinto, 2005). To analyse random migration stage 15 red stinger-labelled haemocytes were imaged on Zeiss Axioplan 2 wide-field imaging system. Contact repulsion of stage 15 haemocytes expressing GFP was captured using a Zeiss 510 confocal laser-scanning microscope with a plan apochromat 63×/1.4 oil objective. All other imaging was conducted on a spinning disc confocal microscope (Ultraview; PerkinElmer).

Due to anterior localization of mys mutant haemocytes, stage 15 WT and mys^XG43 mutant embryos were wounded in the anterior half of the embryo between the ventral midline and lateral lines of haemocytes by laser ablation (Wood et al., 2002).

Stage 15 embryos were prepared and mounted ventral side up as per live imaging before 2.5 mg/ml 70 kDa Rhodamine–dextran (Molecular Probes/Invitrogen, Carlsbad, CA, USA) was injected between the epidermis and VNC as previously described (Evans et al., 2010). Embryos were then imaged live as described above.

Cell tracking and cell area measurements were performed using image J (NIH). Graphs and statistical analysis was carried out using Prism for Mac (Graph Pad). Unless otherwise stated, haemocyte migratory behaviour was analysed using images acquired from five embryos of each genotype.

We thank Bloomington Stock Center for fly stocks and for sharing valuable reagents. The institutional support from the Junta de Andalucía to the CABD is acknowledged.