The macrophage lineage of hemocytes (plasmatocytes) has historically been considered a homogeneous population of cells. However, careful analysis of plasmatocyte behaviour in vivo suggested to us that this lineage might not be functionally uniform. For instance, imaging the inflammatory responses of plasmatocytes to epithelial wounds, we find that some cells close to injury sites rapidly respond by migrating to the wound, while other neighbouring cells fail to respond (Figure 1a; Video 1). We also find that plasmatocytes exhibit variation in their expression of well-characterised plasmatocyte markers such as crq-GAL4 (Figure 1b–b’; Franc et al., 1996; Stramer et al., 2005) and display a broad diversity in their migration speeds within the embryo (random migration at stage 15; Figure 1c–d). These professional phagocytes also display differences in their capacities to phagocytose apoptotic cells with some cells engulfing many apoptotic particles, whereas others engulf very few, if any (Figure 1e). Furthermore, phagocytosis of microorganisms by larval hemocytes also varies significantly from cell-to-cell in vitro (Figure 1f). These differences within the plasmatocyte lineage led us to hypothesise that this cell population is more heterogeneous than previously appreciated.

Given the diversity in plasmatocyte behaviour observed (Figure 1), we hypothesised that macrophage heterogeneity represents an evolutionarily conserved feature of innate immunity, which therefore originally evolved in the absence of an adaptive immune system. To address this and look for molecular differences between plasmatocytes, we examined transgenic enhancer reporter lines (VT-GAL4 lines) produced as part of a large-scale tilling array screen (Kvon et al., 2014) that had been annotated as labelling hemocytes (http://enhancers.starklab.org/). Based on examination of the published VT-GAL4 expression patterns, we identified VT-GAL4 lines that appeared to label reduced numbers of plasmatocytes in the embryo, reasoning that plasmatocyte subpopulations could be molecularly identified on the basis of differences in reporter expression. While a number of the enhancers appeared to label all plasmatocytes (e.g. VT41692-GAL4), we identified several that labelled discrete numbers of plasmatocytes (Figure 2a). We next confirmed that the cells labelled by these VT-GAL4 lines were plasmatocytes by using these constructs to drive expression of UAS-tdTomato in the background of a GAL4-independent, pan-hemocyte marker (srpHemo-GMA – serpent enhancer region driving expression of a GFP-tagged actin-binding domain of Moesin; Figure 2b–d). As initially predicted based on their morphology and position during embryogenesis, each of the VT-GAL4 lines marking potential subpopulations did indeed express in the hemocyte lineage (Figure 2e). These subpopulation cells were identified as plasmatocytes based upon their morphology, the absence of lamellocytes in embryos and the non-migratory nature of crystal cells (Wood and Jacinto, 2007) and could be observed to follow both the dorsal and ventral migration routes (Ratheesh et al., 2015) used by plasmatocytes during their developmental dispersal (Figure 2e). In order to quantify the proportion of cells labelled by each VT-GAL4 line, we counted the number of cells labelled on the ventral midline of the developing stage 15 embryo, using VT-GAL4 lines to drive expression from UAS-GFP. This verified reproducible and consistent labelling of discrete subsets of plasmatocytes (Figure 2f–h), suggesting that these cells represent stable subpopulations within this macrophage lineage.

To characterise these subpopulations further, their overlap with the known plasmatocyte markers Eater, Croquemort, and Simu was investigated using novel GAL4-independent VT-RFP reporters, which we generated using the same enhancer sequences and insertion sites originally characterised by the Stark lab (Kvon et al., 2014). While we were unable to detect embryonic expression of eater at this stage of development (Figure 2—figure supplement 1), reporters for crq and simu (crq-GAL4,UAS-GFP and simu-cytGFP), which also encode phagocytic receptors (Franc et al., 1996; Kurant et al., 2008), clearly labelled embryonic plasmatocytes (Figure 2—figure supplement 2). simu-cytGFP-labelled plasmatocytes with little cell-to-cell variation (Figure 2—figure supplement 2a), whereas crq-GAL4,UAS-GFP displayed considerable heterogeneity (Figure 1b–b’; Figure 2—figure supplement 2b–c). However, there was little correlation of simu or crq marker expression with subpopulation cells, since all cells expressed similar levels of simu, while both crq and VT-RFP expression appeared to vary independently of each other (Figure 2—figure supplement 2a–b). Taken together, we were able to detect discrete subpopulations of plasmatocytes in the embryo, but these subpopulations showed no clear segregation with existing plasmatocyte markers.

Having identified subpopulations of plasmatocytes in the embryo, we then examined other stages of development to see whether their presence was maintained or modulated over time. In order to exclude potential expression in non-hemocyte cells (e.g. the non-plasmatocyte cells apparent in Figure 2e), we labelled subpopulation cells specifically using a split GAL4 approach (Pfeiffer et al., 2010), employing the serpent enhancer (a well-characterised hemocyte marker; Lebestky et al., 2000; Rehorn et al., 1996) and VT enhancers to express the transcriptional activation domain (AD) and DNA binding domains (DBD) of GAL4 independently. Only when co-expressed in the same cell do the AD and DBD heterodimerise and allow expression of UAS transgenes (Figure 3—figure supplement 1a). Characterising the split GAL4 lines in the embryo via expression of the EGFP-derivative Stinger (Barolo et al., 2000) confirmed that this split GAL4 approach labels discrete subpopulations of plasmatocytes within the embryo, although with a higher proportion of cells labelled compared to the original VT-GAL4 lines (Figure 3—figure supplement 1b–d) – a difference likely due to a combination of amplification via the split GAL4 system and enhanced detectability of Stinger, which accumulates in the nucleus due to its nuclear localisation signal. Similar trends in the proportions of plasmatocytes labelled in a variety of locations across the embryo were observed for each subpopulation; for example, VT32897-labelled cells were the least frequently observed in the lateral head region and on the dorsal and ventral sides of the embryo (Figure 3—figure supplement 1b–d). As per Figure 2e, this suggests there are no clear biases between the dispersal routes undertaken by subpopulation cells.

While serpent expression decreases in hemocytes during larval stages, we found that srpHemo-AD;srpHemo-DBD in concert with UAS-stinger was sufficient to label large numbers of cells in both L1 and L2 larvae (Figure 3a–b), consistent with previous publications (Gyoergy et al., 2018). Following hatching of embryos, we cannot use cell morphology to discriminate between plasmatocytes and other hemocyte lineages (crystal cells and lamellocytes) and therefore refer to subpopulation cells as hemocytes for post-embryonic stages of development. Quantification of the numbers of subpopulation cells that could be detected using the split GAL4 system to drive expression from UAS-stinger showed that roughly 50% of serpent-positive hemocytes were labelled in L1 larvae for each VT subpopulation (Figure 3a–c). We cannot exclude the possibility that some of the cells labelled in L1 larvae are fat body cells as serpent is known to be expressed in the fat body (Rehorn et al., 1996). Therefore, it is possible that greater than 50% of hemocytes are labelled at this stage; nonetheless, a significant proportion of subpopulation cells are not labelled via these split GAL4 reporters in L1 larvae (Figure 3a–c). At this stage of development, most hemocytes are found in sessile patches attached to the body wall (Lanot et al., 2001; Makhijani et al., 2011) – this is also the case for the majority of subpopulation cells, since live imaging shows little movement relative to other cells during larval crawling, although some circulating cells could be observed (Video 2).

In contrast to the significant numbers of cells present in L1 and L2 larvae, imaging of L3 larvae containing split srpHemo-AD and VT-DBD reporters (abbreviated to VTn) revealed that fewer subpopulation cells could be detected at this stage (Figure 4a–f). This decrease in subpopulation cells does not seem to be linked to lower levels of serpent expression because blood cells are robustly labelled in positive control L3 larvae (srpHemo-AD in combination with srpHemo-DBD; Figure 4b), suggesting that serpent expression is not limiting our ability to detect subpopulation cells. Moreover, a reduction in subpopulation cells can also be seen when using the original VT-GAL4 lines to drive stinger expression at this stage (i.e. independent of a reliance on serpent expression for labelling; Figure 4—figure supplement 1a).

Using this approach, VT32897 and VT17559 labelled the most cells (Figure 4c–d), with only the occasional cell present in VT57089 larvae (Figure 4e) and cells largely absent from VT62766 larvae (Figure 4f). Labelled cells were also present in the head region, along the dorsal vessel (the fly heart) and between the salivary glands (which themselves exhibit non-specific labelling) in VT32897 larvae. The VT32897 head region cells are likely to represent sessile hemocytes, whereas cells at the remaining two sites probably correspond to serpent-positive nephrocytes and garland cells (Brodu et al., 1999; Das et al., 2008), respectively (Figure 4d). VT57089 shows additional staining in the head region (potentially the Bolwig organ; Figure 4e) and, as per the dorsal vessel-associated cells in VT32897 (Figure 4d), hemocytes can also be found in these regions when the total hemocyte population is labelled using srpHemo-AD and srpHemo-DBD in positive controls (Figure 4b; Video 3). Furthermore, these larval distributions closely resemble patterns observed using VT-GAL4 reporters, albeit with a loss of non srp-dependent labelling due to our split GAL4 approach (data not shown). Live imaging of L3 larvae confirmed that hemocytes were predominantly attached to the body wall, but that small numbers of cells could be detected in circulation (Video 4).

To quantify the proportion of hemocytes that were labelled in L3 larvae and address any bias in localisation, L3 larvae were bled and then the carcasses scraped (as described in Petraki et al., 2015) to compare numbers of cells in circulation with those more tightly adhered to tissue, respectively. This approach confirmed the low numbers of cells observed in L3 larval images and revealed no bias in subpopulation localisation, with similar proportions present in circulation and adhered to tissue for each subpopulation (Figure 4g). Quantification of subpopulation localisation along the L3 body axis suggested that VT57089 and VT62766 cells exhibit a bias toward the posterior of the larvae compared to the total population (Figure 4h–h’).

The striking decrease in proportion of subpopulation cells that can be detected in L3 larvae compared to earlier stages suggests reprogramming of cells leading to a loss of enhancer activity. To test this hypothesis, split GAL4 lines were used in lineage tracing experiments via G-TRACE (Evans et al., 2009). In this approach, current expression of reporters is marked via Red Stinger and also leads to the permanent expression of Stinger via the activity of a co-expressed FLP recombinase. As such, the expression of both fluorophores indicates current split GAL4 activity, while the expression of Stinger alone indicates historical activity in cells where expression has since ceased (Figure 4—figure supplement 1b). Analysing the total complement of current and historic subpopulation cells in L3 larvae via G-TRACE revealed that a significant proportion of subpopulation cells were positive for historical expression only (Stinger expression alone; Figure 4—figure supplement 1c). This confirms that subpopulation cells from earlier timepoints survive in vivo but change their transcriptional profile – a finding consistent with reprogramming events rather than a loss of cells themselves.

Imaging of white pre-pupae (WPP), the stage that marks the beginning of pupal development and metamorphosis, showed very similar patterns across the split GAL4 VT enhancer lines (Figure 5a–f), with a further reduction in the numbers of cells labelled. It was possible to observe the occasional cell moving in circulation within WPP, strongly suggesting these cells are hemocytes (Video 5 and Video 6). Live imaging of VT32897 WPP also confirmed association of cells with the pumping dorsal vessel (Figure 5d; Video 7). Significantly, this data indicates that the presence of subpopulations within embryos is not simply a consequence of slow accumulation of fluorescent proteins by weak drivers, since these enhancer-based reporters do not label an ever-increasing number of cells as development proceeds. Overall, the numbers of hemocytes within subpopulations that can be detected decreases over larval and early pupal stages, suggesting that plasmatocyte subpopulations are developmentally regulated and exhibit plasticity. This reprogramming could reflect specific and changing requirements for specialised plasmatocyte subpopulations across the life cycle, for example, an association with processes required for organogenesis (Charroux and Royet, 2009; Defaye et al., 2009; Regan et al., 2013). The differential localisation of some subpopulation cells also indicates the potential that molecularly and functionally different macrophage populations are present at specific tissues in the fly.

Since subpopulation cells appear associated with stages of development when organogenesis and tissue remodelling occur, we hypothesised that some hemocytes may be reprogrammed via changes in expression leading to reactivation of the enhancers that mark these subpopulations. This would enable subpopulations to return during metamorphosis. Imaging pupae at various times after puparium formation (APF) revealed that subpopulation cells re-emerged in large numbers during this stage, but with distinct dynamics between subpopulations labelled with different enhancers (Figure 6a–f). For instance, VT17559 cells are already present in substantial numbers by 18 hr APF (Figure 6c), whereas VT32897 reporter expression reappears between 24 and 48 hr APF (Figure 6d). VT57089 and VT62766 cells increased in numbers more gradually over the course of pupal development (Figure 6e–f). Different subpopulations appear present in subtly distinct locations in pupae (Figure 6). Further work will be required to understand if subpopulation specification occurs in situ or cells are specified and then migrate to these regions.

Immediately after adults hatch, large numbers of split GAL4-labelled cells can be observed across all lines and are present in selected regions that overlap with the overall adult hemocyte population (Figure 7a–e). The overall hemocyte population remains detectable as adults age (0–6 weeks; Figure 7a); however, not all subpopulations exhibit an identical localisation or dynamics during this time (Figure 7b–e). VT57089 and VT62766 cells largely disappear by 1 week (Figure 7d–e) and the majority of VT17559-labelled cells are absent by 2 weeks (Figure 7b). By contrast, VT32897 cells can be detected for at least 6 weeks of adult life and are particularly prominent in the thorax at 4 weeks (Figure 7c). Other differences in localisation are also apparent with cells particularly obvious in the legs for the VT17559 line (Figure 7b, day 1–2 weeks), whereas VT57089 and VT62766-labelled cells are more closely associated with the thorax and dorsal abdomen (Figure 7d–e, day 1). Labelled cells are also present in the proboscis for several lines (Figure 7c–e).

To quantify the proportion of blood cells labelled in adults, 1-day-old flies were dissected. Despite the large numbers of cells labelled via the split GAL4 system in adults (Figure 7b–e), the proportion of blood cells released via this technique that could be labelled using the split GAL4 system was relatively low (Figure 7f). This suggests that subpopulation cells may favour association with tissues, leaving fewer available to circulate within the hemolymph.

Overall, the distinct dynamics of subpopulation cells in pupal and adult stages (Figure 6 and Figure 7) strongly suggests that these subpopulations are at least partially distinct from each other and highlights their plasticity during development, with their presence, disappearance (via changes in expression shifting them into distinct cell states) and return correlating with changes in the biology of blood cells over the entire lifecourse. While no obvious staining was detected in the lymph gland during larval stages using the split GAL4 lines (Figure 4c–f), additional lineage-tracing analyses would be required to uncover whether cells derived from the lymph gland contribute to subpopulation cell numbers in pupae or adults.

Given that the VT lines identified above are specifically and dynamically expressed in subpopulations of hemocytes during Drosophila development, we next set out to investigate whether the labelled subpopulations are also functionally distinct using a range of immune-relevant assays. The ability of vertebrate macrophages to respond to pro-inflammatory stimuli, such as injuries, can vary according to their activation status (Arnold et al., 2007; Dal-Secco et al., 2015). To investigate this in our system, a well-established assay of inflammatory migration (Stramer et al., 2005) was employed (Figure 1a; Video 1). Strikingly, following laser-induced wounding, cells labelled by three VT-GAL4 lines (VT17559-GAL4, VT32897-GAL4 and VT62766-GAL4) showed a significantly more potent migratory response to injury. In each case, a greater proportion of labelled subpopulation cells migrated to wounds, compared to the overall hemocyte population as labelled by a pan-plasmatocyte driver (Figure 8a–c). Consistent with our results above, plasmatocytes labelled by the VT lines represent a subset of the total number of hemocytes present ventrally in stage 15 embryos (Figure 8d).

We next investigated in vivo migration speeds of the embryonic plasmatocyte subpopulations (as per Figure 1c–d). Stage 15 embryos were imaged for 1 hr and individual plasmatocyte movements were tracked (Figure 8e–f). Only the VT17559-GAL4-labelled plasmatocyte subpopulation displayed statistically significantly faster rates of migration compared to the overall plasmatocyte population (labelled using srpHemo-GAL4; Figure 8g). There were no differences in directionality (cell displacement divided by total path length) for any of the subpopulations, suggesting that the mode of migration was similar across these lines and with that of the overall population (Figure 8h).

Apoptotic cell clearance (efferocytosis) represents another evolutionarily-conserved function performed by embryonic plasmatocytes (Figure 1e). Therefore, we investigated this function in subpopulations, using numbers of phagosomal vesicles per cell as a proxy for this process (Evans et al., 2013). Cells labelled via VT17559-GAL4, VT57089-GAL4 and VT62766-GAL4 (but not VT32897-GAL4) contained fewer phagosomes than the overall plasmatocyte population (Figure 8i–k), suggesting that these discrete populations of cells are less effective at removing apoptotic cells inside the developing embryo. To confirm this result dynamically, GFP-myc-2xFYVE, a phosphatidylinositol-3-phosphate reporter (Wucherpfennig et al., 2003) was used to measure the rate of phagocytosis in subpopulation cells. This reporter rapidly and transiently localises on the surface of engulfed phagosomes in plasmatocytes (Roddie et al., 2019). All subpopulations exhibited lower rates of phagocytosis compared to the overall plasmatocyte population using this reporter (Figure 8l), suggesting the differences in numbers of phagosomes per cell result from distinct phagocytic abilities.

Finally, we examined cell size and the shape of labelled plasmatocyte subpopulations. Vertebrate macrophages are highly heterogeneous, with distinct morphologies dependent upon their tissue of residence or polarisation status (McWhorter et al., 2013; Ploeger et al., 2013; Rostam et al., 2017). We found no obvious size or shape differences between VT-GAL4-labelled cells and the overall plasmatocyte population (Figure 8—figure supplement 1a–e). This was also the case when VT-GAL4-positive cells were compared to internal controls (VT-GAL4-negative cells within the same embryos) for a range of shape descriptors (Figure 8—figure supplement 1f–i). Similarly, we were unable to detect differences in ROS levels (Figure 8—figure supplement 2) or the proportion of VT-GAL4-labelled plasmatocytes that phagocytosed pHrodo-labelled E. coli compared to controls (Figure 8—figure supplement 3), two processes associated with pro-inflammatory activation of macrophages (Benoit et al., 2008).

Taken together these data show that the subpopulations of plasmatocytes identified via the VT-GAL4 reporters exhibit functional differences compared to the overall plasmatocyte population (Table 1). Therefore, as well as displaying molecular differences in the form of differential enhancer activity, and hence reporter expression, these discrete populations of cells behave differently. This strongly suggests that these cells represent functionally distinct subpopulations and that the plasmatocyte lineage is not homogeneous. Furthermore, not all subpopulations displayed identical functional characteristics, suggesting that there are multiple distinct subtypes present in vivo, although some overlap between subpopulations seems likely. For example, VT17559-GAL4-labelled cells were more effective at responding to wounds and migrated more rapidly but carried out less phagocytosis of apoptotic cells. By contrast, VT32987-GAL4-labelled cells only displayed improved wound responses (Figure 8).

In the original study that analysed the VT-GAL4 collection, the majority of active enhancer fragments tested were found to control transcription of neighbouring genes (Kvon et al., 2014). Thus, genes proximal to enhancers that label plasmatocyte subpopulations represent candidate regulators of immune cell function (Table 2; Figure 9a). VT62766-GAL4 labels a subpopulation of plasmatocytes with enhanced migratory responses to injury (Figure 8a–c) and this enhancer region is found within the genomic interval containing paralytic (para), which encodes a subunit of a voltage-gated sodium channel (Lin et al., 2009), and upstream of the 3’ end of Calnexin14D (Cnx14D; Figure 9a). Cnx14D encodes a calcium-binding chaperone protein resident in the endoplasmic reticulum (Christodoulou et al., 1997). Alterations in calcium dynamics are associated with clearance of apoptotic cells (Cuttell et al., 2008; Gronski et al., 2009) and modulating calcium signalling within plasmatocytes alters their ability to respond to wounds (Weavers et al., 2016). Therefore, given the association of Cnx14D with the VT62766 enhancer and the potential for plasmatocyte behaviours to be modulated by altered calcium dynamics, we examined whether misexpressing Cnx14D in all plasmatocytes was sufficient to cause these cells to behave more similarly to the VT62766 subpopulation. Critically, pan-hemocyte expression of Cnx14D stimulated wound responses with elevated numbers of plasmatocytes responding to injury compared to controls (Figure 9b–c), consistent with the enhanced wound responses of the endogenous VT62766-GAL4-positive plasmatocyte subpopulation (Figure 8c). This reveals that genes proximal to subpopulation-defining enhancers represent candidate genes in dictating the biology of cells in those subpopulations. More importantly, misexpression of a subpopulation-linked gene promotes a similar behaviour to that subpopulation in the wider plasmatocyte population.

Having defined functional differences in embryonic plasmatocyte subpopulations and characterised how these populations shift during development and ageing, we sought to identify the processes via which these subpopulations were specified. In vertebrates, a range of stimuli drive macrophage heterogeneity and polarisation (Martinez and Gordon, 2014; Murray, 2017), with apoptotic cells able to polarise macrophages towards anti-inflammatory phenotypes (A-Gonzalez et al., 2017; de Oliveira Fulco et al., 2014). In the developing fly embryo, high apoptotic cell burdens impair wound responses (Armitage et al., 2020; Roddie et al., 2019), consistent with reprogramming of plasmatocytes towards less wound-responsive states. In order to test whether apoptotic cells might regulate plasmatocyte subpopulations, we exposed plasmatocytes to increased levels of apoptosis in vivo. In the developing fly embryo, both glial cells and plasmatocytes contribute to the clearance of apoptotic cells. We, and others, have previously shown that loss of repo, a transcription factor required for glial specification (Campbell et al., 1994; Halter et al., 1995; Xiong et al., 1994), leads to decreased apoptotic cell clearance by glia (Shklyar et al., 2014), and a subsequent challenge of plasmatocytes with increased levels of developmental apoptosis (Figure 10a–b; Armitage et al., 2020). Therefore, a repo mutant background represents an established model with which to stimulate plasmatocytes with enhanced levels of apoptosis.

Using srpHemo-H2A-mCherry to mark all plasmatocytes within the embryo (Figure 10c), we quantified the proportion of plasmatocytes labelled via VT-GAL4 transgenes in repo mutants compared to controls (Figure 10d–h). Increased exposure to apoptotic death shifted plasmatocytes out of each subpopulation (Figure 10d–h). Subpopulations exhibited differing sensitivities to contact with apoptotic cells, with numbers of VT62766-GAL4-labelled cells undergoing the largest decrease in a repo mutant background (Figure 10h). These results therefore reveal a mechanism via which the molecularly and functionally distinct subpopulations of plasmatocytes we have identified can be manipulated using an evolutionarily conserved, physiological stimulus (apoptotic cells) relevant to immune cell programming.

Standard cornmeal/agar/molasses media was used to culture Drosophila at 25°C (see Supplementary file 2 for ingredients). srpHemo-GAL4 (Brückner et al., 2004; Wood et al., 2006), crq-GAL4 (Stramer et al., 2005), and the GAL4-independent lines srpHemo-GMA (received from J. Bloor, University of Kent, UK), srpHemo-3x-mCherry and srpHemo-H2A-3x-mCherry (Gyoergy et al., 2018) were used to label the entire hemocyte population during embryonic development or in adults. N.b. SrpHemo-GAL4 is referred to as srp-GAL4 on graphs (for reasons of space) but this is the shorter construct more specific to hemocytes (as per Brückner et al., 2004) rather than the entire serpent promoter region. Hml(Δ)-GAL4 (Sinenko and Mathey-Prevot, 2004) was used to label larval hemocytes and da-GAL4 (Wodarz et al., 1995) was used as a ubiquitous driver line. These GAL4 lines, Vienna Tiling array GAL4 lines (VT-GAL4 lines obtained from VDRC; Kvon et al., 2014) and split GAL4 lines (see below) were used to drive expression from UAS-tdTomato, UAS-GFP (Stramer et al., 2005; Wood et al., 2006), UAS-red stinger (Davis et al., 2012), UAS-stinger, UAS-Cnx14D, UAS-GFP-myc-2xFYVE (Wucherpfennig et al., 2003), or UAS-GC3ai (Schott et al., 2017). GAL4-independent VT-RFP lines were also generated as part of this study (see below) and used to label subpopulation cells in combination with crq-GAL4,UAS-GFP (Stramer et al., 2005), eater-GFP (Sorrentino et al., 2007), and simu-cytGFP (Kurant et al., 2008). Experiments were conducted in a w¹¹¹⁸ background and the repo⁰³⁷⁰² null allele was used to expose plasmatocytes to enhanced levels of apoptotic cell death in the embryo (Armitage et al., 2020; Campbell et al., 1994; Halter et al., 1995). Both UAS-tdTomato and UAS-GFP were used to analyse subpopulations in the developing embryo in order to ensure labelling of discrete numbers of plasmatocytes was not due to positional effects of insertion sites that led to mosaic expression (Figure 2). G-TRACE flies (w;;UAS-red stinger,UAS-FLP,Ubi-p63E(FRT.STOP)Stinger; Evans et al., 2009) were crossed to split GAL4 driver lines (see below) for lineage-tracing experiments. See Supplementary file 1 for a full list of Drosophila genotypes, transgenes and the sources of the Drosophila lines used in this study.

Flies were added to laying cages attached to apple juice agar plates supplemented with yeast paste and allowed to acclimatise for 2 days before embryo collection. Plates were then changed every evening and cages incubated at 22°C overnight before embryos were collected the following morning. Embryos were collected by washing the plates with distilled water and gently disturbing the embryos with a paintbrush, after which embryos were collected into a cell strainer. Embryos were dechorionated in undiluted bleach for 1–2 min and then washed in distilled water until free from bleach. The fluorescent balancers CTG, CyO dfd, TTG, and TM6b dfd (Halfon et al., 2002; Le et al., 2006) were used to discriminate homozygous embryos after removal of the chorion.

We used the split GAL4 system (Pfeiffer et al., 2010) to restrict VT enhancer expression to serpent-positive cells. The activation domain (AD) of GAL4 was expressed using a well-characterised fragment of the hemocyte-specific serpent promoter (Brückner et al., 2004; Gyoergy et al., 2018) and the DNA-binding domain (DBD) was expressed under the control of VT enhancer regions corresponding to VT17559-GAL4, VT32897-GAL4, VT57089-GAL4, or VT62766-GAL4. High-fidelity polymerase (KAPA HiFi Hotstart ReadyMix, Roche) was used to PCR amplify VT enhancer regions from genomic DNA extracted from the original VT-GAL4 line flies, which were then TA cloned into the pCR8/GW/TOPO vector. Primers were designed according to VT enhancer sequences available via the Stark Lab Fly Enhancers website (http://enhancers.starklab.org/; Kvon et al., 2014). To make VT-DBD transgenic constructs, VT enhancers were transferred from pCR8/GW/TOPO into pBPZpGal4DBDUw (Addgene plasmid 26233) using LR clonase technology (Gateway LR Clonase II Enzyme Mix, Invitrogen).

To express the DBD and AD of GAL4 under the control of the serpent promoter (srpHemo-AD and srpHemo-DBD; also referred to as srp-AD and srp-DBD for reasons of space on graphs), these were subcloned into a vector containing an attB site and this promoter (pBS_MCS_SRPW_attB; DSPL337 – a gift from Daria Siekhaus, IST, Austria; Gyoergy et al., 2018). DBD and AD sequences along with the Drosophila synthetic minimal core promoter (DSCP) region were amplified using PCR from vectors pBPZpGal4DBDUw and pBPp65ADZpUw (Addgene clone 26234) using primers that added NotI and AvrII restriction sites (CTGATCGCGGCCGCAAAGTGGTGATAAACGGCCGGC and GATCAGCCTAGGGTGGATCTAAACGAGTTTTTAAGCAAACTCAC). These were subcloned into DSPL337 cut with NotI/AvrII (New England Biolabs) using T4 DNA ligase (Promega). Transgenic flies were generated by site-specific insertion of transgenic constructs into the VK1 attP site on chromosome 2 and/or attP2 on chromosome 3 (Genetivision).

To generate GAL4 independent VT-RFP transgenic lines, nuclear RFP was isolated by sequential digestion of pRed H-Pelican (DGRC plasmid 1203) using Acc65I and then SpeI restriction enzymes (NEB). In parallel, GAL4 was excised from pBPGUw (Addgene plasmid 17575) using the same restriction enzymes and replaced with nuclear RFP using T4 ligase. LR clonase was again used to transfer the VT enhancer regions from the PCR8/GW/TOPO gateway vectors (see above) into the nuclear RFP-containing pBPGUw destination vector. Transgenic flies were generated by PhiC31 integrase-mediated insertion of VT-RFP constructs into attP2 on chromosome 3 (Genetivision).

Live embryos were mounted ventral-side up on double-sided sticky tape in a minimal volume of Voltalef oil (VWR), after dechorionation in bleach as previously (Evans et al., 2010). High-resolution live imaging of plasmatocytes was carried out on an UltraView Spinning Disk system (Perkin Elmer) using a 40x UplanSApo oil immersion objective lens (NA 1.3). A Nikon A1 confocal microscope was used to image plasmatocyte morphology (40x CFI Super Plan Fluor ELWD oil immersion objective lens, NA 0.6) and a Zeiss Airyscan microscope (40x Plan-Apochromat oil immersion objective lens, NA 1.4) was used for imaging of embryos stained with ROS dyes.

L1 and L2 larvae were allowed to develop at 22°C from embryos laid on apple juice agar plates at the same temperature. Larvae were selected and washed in distilled water in embryo baskets, then partially anaesthetised using diethyl ether (2 min for L1 larvae, 3.5 min for L2 larvae). Larvae were then transferred to double-sided tape and covered with halocarbon oil 500 (Sigma-Aldrich). Thickness one coverslip bridges (VWR) were attached to the tape either side of larvae and another coverslip placed across these supports (over the larvae) and attached in place with nail varnish. Larvae were immediately imaged on a MZ205 FA fluorescent dissection microscope with a 2x PLANAPO objective lens (Leica) and LasX software (Leica). To quantify numbers of cells labelled via the split GAL4 system in newly hatched larvae, L1 larvae were flattened under a coverslip in a small drop of halocarbon oil 500. Overlapping images of the flattened larvae were taken using the same microscope and mosaics assembled in Adobe Photoshop. Mosaics were blinded and the number of cells expressing Stinger counted using the multipoint selection tool in Fiji. The same microscope was also used to image L3 larvae, white pre-pupae (WPP), pupae, and adults (see below).

Wandering L3 Larvae and WPP were removed from straight-sided culture bottles containing the food on which they were reared at 25°C and cleaned in distilled water. L3 Larvae were imaged in fresh ice-cold, distilled water to minimise their movements, while WPP were immobilised on double-sided tape (Scotch). For analysis of plasmatocyte populations in pupae, white pre-pupae were also collected, aged at 25°C and the pupal cases removed at a range of times after puparium formation. Dissected pupae were covered with halocarbon oil 500 to prevent desiccation during imaging. For imaging of plasmatocyte populations in adults, females were aged in vials containing cornmeal/agar/molasses media (Supplementary file 2) at 25°C, with no more than seven flies kept per vial. Flies were transferred to new food vials every 2–3 days; flies were chilled at −20°C for 4 min and imaged in a petri dish on top of ice to minimise their movements.

To analyse the distribution of subpopulation cells along the body axis in L3 larvae, the relative proportions of cells within the anterior, medial (abdominal segments A3-A5), and posterior regions were calculated. The number of cells in each region in images of L3 larvae were counted in Fiji and expressed as a fraction of the total number of cells in each larva. As a comparison to reflect the distribution of the total larval hemocyte population, images of L3 larvae with hemocytes labelled using Hml(Δ)-GAL4,UAS-GFP were analysed. Since hemocytes were too numerous to count accurately in these images, the integrated density of GFP fluorescence (mean gray value multiplied by area) was measured in each region in Fiji. The proportion of the total GFP signal in each region was then calculated per larva.

To isolate larval hemocytes, single wandering third instar were picked from bottles with a paintbrush, washed with distilled water then placed in a 75 μL drop of ice-cold S2 media, which consists of Schneider’s media (Sigma-Aldrich) supplemented with 10% heat-inactivated FBS (Gibco/Sigma-Aldrich) and 1x Pen/Strep (Gibco/Sigma-Aldrich). Larvae were then ripped open from the posterior end using size five forceps to release hemocytes into the S2 cell media. Larval carcasses were gently agitated for 5 s before being removed from the S2 media droplet. The cell suspension was then transferred to a well of a 96-well plate (Greiner) and a further 75 μL of S2 media was added per well. Hemocytes were allowed to settle for at least 90 min in a humidified box prior to fixation. Cells were then fixed for 15 min using 4% EM-grade formaldehyde (Thermo Scientific) in PBS (Oxoid). Cells were then permeabilised for 4 min using 0.1% Triton-X-100 in PBS. Following washing in PBS, nuclei and actin filaments were stained using NucBlue (two drops per ml; Invitrogen) and Alexa Fluor 647 phalloidin (1:200 in PBS; Invitrogen; Figure 7f) or Alexa Fluor 568 phalloidin (1:500 in PBS; Invitrogen; Figure 4g) for 30 min. Following a final wash step, cells were imaged using an ImageXpress Micro hi-content microscope (Molecular Devices).

To discriminate sessile and circulating populations of hemocytes, we adapted a previously described protocol (Petraki et al., 2015): wandering third instar larvae were selected, washed and placed in S2 media as above. Larvae were then bled by puncturing at their posterior and anterior ends with sterile 27G needles to release circulating hemocytes; larvae were not agitated and were left in the media droplet for up to 10 s. The bled larva was then transferred to another 75 μL drop of S2 media, while the media containing the initial bleed was transferred to a well in a 96-well plate. Bled larval carcasses were held down using a 27G needle and then jabbed/scraped with a separate needle to release sessile/adherent hemocytes. After jabbing/scraping, the carcass was removed, and the media droplet transferred to a well of a 96-well plate (Greiner). Finally, 75 μL of S2 media was added to each well and cells were allowed to settle prior to being fixed, stained and imaged as above.

For stimulation with S. cerevisiae (Figure 1), several larvae were pooled and dissected within larger droplets (75 μl per larvae used). Of this cell suspension, 75 μl was transferred into each well in a 96-well plate (Porvair) and cells allowed to adhere in a humidified box in the dark for 2 hr. After 2 hr, cells were stimulated with heat-killed S. cerevisiae particles previously stained using calcofluor staining solution (Sigma-Aldrich). S. cerevisiae (strain BY4741/accession number Y00000, Euroscarf consortium) were grown to exponential phase in YPD broth (Fisher) at 28°C. Yeast were heat killed at 60°C for 30 min, spun down and frozen at 20 × 10⁹ cells/ml. 1 × 10⁹ heat-killed yeast particles in 1 ml of PBS were stained for 30 min at room temperature (with rotation) using 15 μl of calcofluor staining solution. Stained yeast particles were washed in PBS and 1 × 10⁶ particles resuspended in 75 μl S2 cell medium, which was then added to each well of larval hemocytes for 2 hr. Cells were fixed in wells using 4% EM-grade formaldehyde in PBS for 15 min and washed in PBS. Images were taken on a Nikon Ti-E inverted fluorescence microscope using a 20x objective lens and GFP and DAPI filter sets.

To isolate hemocytes from adults, two flies per genotype (1 day post-eclosion) were anaesthetised using CO₂ and cut in half longitudinally in a 75 μl droplet of S2 media on ice. A further 75 μl of S2 media was then added and carcasses agitated by pipetting for 10 s to release hemocytes. The 150 μl of cell suspension was then transferred to a single well in a 96-well plate (Greiner). Cells were allowed to settle for 30 min before being fixed and stained as per larval hemocytes. Based on phalloidin staining and cell morphology it was assumed all adhered cells were blood cells. The percentage of labelled cells was calculated using the number of Stinger-positive cells divided by the total number of cells in images (NucBlue labelling).

Live stage 15 embryos were prepared and mounted as described above. The ventral epithelium of the embryos was ablated on the ventral midline using a Micropoint nitrogen-pulsed ablation laser (Andor) fitted to an Ultraview spinning disk confocal system (PerkinElmer) as previously (Evans et al., 2015). Pre-wound z-stacks of 30 μm were taken of superficial plasmatocytes with a 1 μm z-spacing between z-slices. Post-wound images were taken on the same settings either at 2 min intervals for 60 min (Figure 1) or at the end timepoint of 60 min (Figure 8 and Figure 9).

The proportion of plasmatocytes labelled with UAS-stinger (expression via srpHemo-GAL4 or VT-GAL4) was assessed by counting the number of labelled cells at or in contact with the wound site within a 35 μm deep volume on the ventral midline at 60 min post-wounding; this was divided by the total number of labelled cells present within the stack to calculate the percentage of plasmatocytes responding to injury. The brightfield channel was used to visualise the wound margin and only those embryos with wounds between 1000 μm² and 4000 μm² were included in analyses. Quantification was performed on blinded images in Fiji.

Embryos were prepared and mounted as previously described (Evans et al., 2010). Random migration was imaged using an Ultraview spinning disk system (PerkinElmer), with an image taken every 2 min for 1 hr with a z-spacing of 1 μm and approximately 20 μm deep from the ventral nerve cord using a 20x UplanSApo air objective lens (NA 0.8). Maximum projections were made for each timepoint (25 μm depth) and the centre of individual plasmatocyte cell bodies tracked using the manual tracking plugin in Fiji. Random migration speed (μm/min) and directionality (the ratio of the Cartesian distance to the actual distance migrated) were then calculated using the Ibidi chemotaxis plugin.

The number of apoptotic cell-containing phagosomes per plasmatocyte (averaged per embryo) was used as a read-out of apoptotic cell clearance as previously described (Evans et al., 2013). Phagocytic vesicles were counted using z-stacks of GFP-labelled plasmatocytes taken from live imaging experiments. Phagosomes were scored in the z-slice in which each macrophage exhibited its maximal cross-sectional area. Only labelled plasmatocytes present on the ventral midline of stage 15 embryos were included. Analysis was performed on blinded image stacks. This analysis does not report the absolute numbers of apoptotic corpses per cell but provides a relative read-out of the phagocytic index.

To assay rates of phagocytosis of apoptotic cells, a phosphatidylinositol-3-phosphate reporter (UAS-GFP-myc-2xFYVE; Roddie et al., 2019; Wucherpfennig et al., 2003) was expressed in all plasmatocytes (via srpHemo-AD in combination with srpHemo-DBD) or in subpopulation cells (via srpHemo-AD in combination with VT enhancers driving expression of the DBD domain). Plasmatocytes were imaged at stage 12/13 on the ventral midline and the number of ‘FYVE events’ (number of times new recruitment events could be seen to form on the surface of nascent phagosomes) per plasmatocyte per movie was scored. Only cells present in at least 15 min of 30 min movies were included in this analysis and scoring was conducted on blinded movies constructed from maximum projections of z-stacks.

For morphological analysis of plasmatocytes (Figure 8—figure supplement 1), the vitelline membrane was manually removed from individual z-slices by drawing around the inside edge of the membrane with the freehand selection tool and using the clear outside command. Maximum projections were then created of the ventral midline region in Fiji. Following this, a region of interest was manually drawn around the area of individual plasmatocytes using the polygon tool and a range of cell shape descriptors and measurements calculated using Fiji.

To stain plasmatocyte ROS levels (Figure 8—figure supplement 2), embryos containing tdTomato-labelled plasmatocytes (with expression driven using srpHemo-GAL4 or VT-GAL4) were first dechorionated and then left in water for 30 min. Stage 15 embryos were then selected and transferred to a glass vial wrapped in foil containing 1 ml peroxide-free heptane (Sigma-Aldrich) and 1 ml of 50 μM dihydrorhodamine 123 (DHR123, Sigma-Aldrich) in PBS. Embryos were shaken at 250 rpm for 30 min. Following this, embryos were removed from the interface and mixed with halocarbon oil. Embryos were orientated individually in a droplet of this oil on a glass slide and then immediately imaged using a Zeiss Airyscan microscope (40x Plan-Apochromat oil immersion objective, NA 1.4), with z-spacing of 1 μm and stacks totalling 30 μm from the surface of the vitelline membrane down through the ventral nerve cord. Embryos were exposed to 10 mM of H₂O₂ (Sigma-Aldrich) in PBS and peroxide-free heptane for 30 min prior to staining with DHR123 as a positive control. Negative control embryos were incubated in heptane/PBS alone.

To quantify ROS levels, the intensity of DHR123 staining was measured in the z-slice in which each macrophage exhibited its maximal cross-sectional area. The body of the macrophage was drawn around using the polygon tool in Fiji and then the area and mean gray value were measured in the GFP (DHR123) channel. Average mean gray value per plasmatocyte, per embryo was then plotted in Prism.

To assay phagocytosis of an immune challenge (Figure 8—figure supplement 3), dechorionated stage 15 embryos were mounted ventral-side up on a slide using double-sided Scotch tape, then dehydrated by incubating in a small container with silica beads for 7–8 min. Further dehydration was then prevented by covering embryos in a small drop of Voltalef oil. 1 mg/ml pHrodo green E. coli BioParticles (Invitrogen; resuspended in PBS) were microinjected into the anterior of stage 15 embryos to determine the phagocytic capability of labelled plasmatocytes. Needles were created by pulling 15 cm long 1 mm glass capillaries (World Precision Instruments) using a Flaming/Brown P-1000 micropipette puller (Sutter, program 51). Needle tips were snapped using forceps under high magnification to create a bevelled end. Imaging was performed 1 hr after injection using an UltraView Perkin Elmer Spinning Disk system (40x UplanSApo oil immersion objective lens, NA 1.3). The proportion of VT-GAL4 or srpHemo-GAL4-positive cells containing E. coli BioParticles was scored.

Embryos were fixed and stained as previously described (Roddie et al., 2019). Embryos containing plasmatocytes labelled via srpHemo-GMA and GAL4-driven tdTomato expression were fixed, then mounted in Dabco mountant. Control and repo mutant embryos containing plasmatocytes labelled via crq-GAL4,UAS-GFP were fixed and immunostained using mouse anti-GFP (ab1218 1:200; Abcam) and rabbit anti-cleaved DCP-1 (9578S 1:1000; Cell Signaling Technologies) to label plasmatocytes and apoptotic cells, respectively. Primary antibodies were detected using Alexa Fluor 488 goat anti-mouse IgG and Alexa Fluor 568 goat anti-rabbit IgG (Invitrogen/Molecular Probes; both used at a 1:400 dilution). Embryos were imaged on the Nikon A1 system described above.

All microscopy images were processed using Fiji (Schindelin et al., 2012). Images were blinded ahead of analysis with quantification performed on maximum z-projections, with the exception of analysis of numbers of cells labelled via VT-GAL4 lines (Figure 2h), wound responses (Figure 8c–d), apoptotic cell clearance (Figure 8k) and quantification of ROS staining (Figure 8—figure supplement 2f). Quantification was performed on blinded z-stacks for those analyses.

Statistical tests were performed using Prism 7 (GraphPad). p-Values less than 0.05 were deemed significant. Experiments were carried out across at least three independent imaging sessions with N numbers representing individual embryos, with N numbers cited for each condition in the appropriate figure legend. No outliers were excluded. Embryos were taken from laying cages containing greater than 50 adult flies of the parental genotypes, with mutant or control embryos of the correct genotype and developmental stage selected at random following dechorionation. N numbers were typically sufficiently large to enable use of parametric tests. Student’s t-test was used when comparing two experimental data sets; where multiple comparisons were required, a one-way ANOVA with Dunnett’s multiple comparisons test was performed (parametric data) or the Kruskall-Wallis with Dunn’s multiple comparisons test was used (non-parametric data). N numbers, p-values and details of statistical tests employed are reported in the appropriate figure legend. All raw numerical data can be found within the supplementary material as source data files.