Regulation of Drosophila hematopoietic sites by Activin-β from active sensory neurons

An outstanding question in animal development, tissue homeostasis and disease is how cell populations adapt to sensory inputs. During Drosophila larval development, hematopoietic sites are in direct contact with sensory neuron clusters of the peripheral nervous system (PNS), and blood cells (hemocytes) require the PNS for their survival and recruitment to these microenvironments, known as Hematopoietic Pockets. Here we report that Activin-β, a TGF-β family ligand, is expressed by sensory neurons of the PNS and regulates the proliferation and adhesion of hemocytes. These hemocyte responses depend on PNS activity, as shown by agonist treatment and transient silencing of sensory neurons. Activin-β has a key role in this regulation, which is apparent from reporter expression and mutant analyses. This mechanism of local sensory neurons controlling blood cell adaptation invites evolutionary parallels with vertebrate hematopoietic progenitors and the independent myeloid system of tissue macrophages, whose regulation by local microenvironments remain undefined.

I n vertebrates, regulation of self-renewing blood cell populations by local organ microenvironments is poorly understood at the cellular and molecular level [1][2][3] . In a related Drosophila melanogaster model, blood cells (hemocytes), with similarities to vertebrate tissue macrophages and oligopotent hematopoietic progenitors, form resident clusters in segmentally repeated inductive microenvironments of the larval body wall, also known as Hematopoietic Pockets (HPs) [4][5][6][7] (Fig. 1a). More than 90% of these larval hemocytes of embryonic origin are macrophage-like cells (plasmatocytes), which in the larva colonize HPs and expand by proliferation in the differentiated state 4 . Based on their functional dependence on sensory neurons of the HPs for their localization and survival, and the elevated proliferation of resident hemocytes compared to those in circulation 4 , we investigated the molecular mechanism of hemocyte induction by the sensory neurons of the peripheral nervous system (PNS).
Here we identify a molecular mechanism by which Drosophila hematopoiesis is controlled by the PNS. Activin-b (Actb, Act), a Transforming Growth Factor-b (TGF-b) family ligand, is specifically produced by multidendritic neurons and chordotonal organs of the PNS, and acts to regulate the proliferation and longterm adhesion of hemocytes. PNS activation by agonist treatment drives expansion of the blood cell pool, while specific silencing of sensory neurons affects resident hemocyte number and localization. Actb plays a key role in this regulation, as evidenced by the induction of Actb expression in response to PNS activity, and a blunted response in Actb mutants. These findings shed new light on the mechanisms by which local microenvironments regulate blood cell adaptation and may integrate sensory inputs.

Results
Sensory neurons form an interface with the blood cell system. First we examined the local anatomy of PNS neurons and resident (sessile) hemocytes. High magnification imaging revealed that PNS neurons form intricate extensions to areas of resident hemocytes (Fig. 1b), suggesting an interface that allows direct neuron-to-hemocyte communication. Using a split GFP approach, GFP Reconstitution Across Synaptic Partners (GRASP) 8,9 ( Supplementary Fig. 1a), we confirmed that hemocytes are anatomically extremely close, and potentially form direct contacts with PNS neurons and glia (Supplementary Fig. 1b-e). To identify specific molecular signals that mediate this communication, we screened components of several key signalling pathways utilizing in vivo RNA interference (RNAi) 10 , searching for defects in hemocyte number and/or localization. Based on this, we focused on the role of the TGF-b family-related dSmad2 (Smox) pathway 11 in hemocyte regulation. To identify the responsible ligand, we examined expression of the putative pathway ligands Activin-b (Actb), Dawdle (daw) and myoglianin (myo) using ligand GAL4 reporters (O'Connor lab and 12,13 ). Actb was highly expressed by specific subsets of sensory neurons in the HPs, in particular the multidendritic (md) neurons and chordotonal organs 14 (Fig. 1d,e). None of the other ligands showed obvious expression in the sensory neurons. Colabelling confirmed localization of Actb-producing neurons with resident hemocytes in the HPs (Fig. 1c). In contrast, Actb was not detectably expressed in other components of the HPs, such as epidermis, muscle and oenocytes, (Figs 1c,d,6c,d,f,g). Actb was also highly expressed in motor neurons and other neurons of the central nervous system (CNS) as described previously, similar to related ligands of the TGF-b family 11,[15][16][17][18][19][20] . However, the vast majority of hemocytes in the larva, present as resident hemocytes in the HPs, were physically separated from the CNS, motor  Actb signalling promotes blood cell proliferation and adhesion.
Next we investigated the effects of Actb/dSmad2 pathway lossand gain of function (lof, gof). Studying the viable null mutant Actb ED80 (ref. 21), or silencing Actb by in vivo RNAi, driving transgene expression ubiquitously or in sensory md neurons, we found diminished hemocyte numbers in the segmental HPs ( Fig. 2a-e). Consistent with this, we observed an increase in the fraction of circulating hemocytes (Fig. 2f), and overall reduced total hemocyte numbers per larva (Fig. 2g), using a method of quantitative hemocyte retrieval from single Drosophila larvae 22 . These defects resembled hemocyte phenotypes seen on PNS neuron ablation 4 and suggested defects in hemocyte adhesion, proliferation and/or survival. Silencing of Actb in motor neurons using the driver OK6-GAL4 ( Supplementary Fig. 2d,e,h,i) did not affect the localization of hemocytes in the HPs (Supplementary Fig. 2f,g,j,k), and resulted only in minor reductions of total hemocytes during various time points of larval development, which were non-significant according to Student's t-test ( Supplementary Fig. 2l). Actb ED80 mutants showed partial penetrance (62% mutant phenotype) and were analysed sideby-side with controls as 2nd instar larvae, to avoid compensatory mechanisms that became evident in 3rd instar larvae. In all backgrounds of Actb lof, PNS neuron clusters were present and appeared normal by cell number and dendritic morphology. This suggested a role for Actb signalling in hemocytes, rather than indirect effects due to roles of Actb in the nervous system 11,15,18,19 .
To substantiate the role of Actb/dSmad2 signalling in hemocytes, we systematically determined the effects of hemocyte-autonomous RNAi silencing of the Activin type II receptor punt (put) 23 and the signal transducer dSmad2 (ref. 11). Similar to the loss of Actb in the microenvironment, knockdown of Actb pathway components in hemocytes resulted in diminished numbers of resident hemocytes in the segmental HPs ( Fig. 2h-k). Silencing of the Activin type I receptor baboon (babo) 11 showed similar albeit milder effects on hemocyte localization. Silencing of the pathway in hemocytes by knockdown of put or dSmad2 further resulted in increased fractions of circulating hemocytes (Fig. 2l) and reduced total hemocyte numbers per larva (Fig. 2m). With prolonged RNAi silencing of put and dSmad2, the reduction in total hemocytes eventually reversed in older larvae ( Supplementary Fig. 3), again implying putative compensatory mechanisms that take effect in the prolonged absence of dSmad2 signalling. Silencing of put and dSmad2 under the control of HmlD-GAL4 had no effect on lymph gland hemocytes in larvae of the developmental window studied ( Supplementary Fig. 4a-i), and no concomitant increase in the fraction of crystal cells, neither of the lymph gland nor the embryonic lineage of hemocytes, was observed (for lymph gland see Supplementary Fig. 4c,f,i). This was examined because in the larva, plasmatocytes are known to give rise to small numbers of crystal cells 24,25 . Taken together, our findings suggest that sensory neuron-produced Actb signals through the dSmad2 pathway in hemocytes, which supports hemocyte numbers and promotes hemocyte localization to the HPs.
To determine whether Actb/dSmad2 signalling has trophic and/or proliferative roles in hemocytes, we focused on the effects of Actb overexpression and babo gain of function. Moderate Actb overexpression in PNS neurons, or ectopic sites in oenocytes and epidermis using the driver Spalt (Sal)-GAL4, resulted in increased total hemocyte numbers per larva ( Fig. 3a-g), while silencing of Actb in these locations had no significant effect according to Student's t-test ( Supplementary Fig. 5a-c). The increase in total hemocyte numbers on Actb overexpression was accompanied by increased in vivo EdU incorporation, suggesting enhanced hemocyte proliferation (Fig. 3i). Consistently, silencing of put and dSmad2 in hemocytes resulted in reduced EdU incorporation of hemocytes in vivo (Fig. 3j). To substantiate a direct role of Actb in hemocyte proliferation, we also examined larval hemocytes ex vivo under conditions of Actb stimulation. Indeed, we found that Actb promoted EdU incorporation, indicative of increased hemocyte proliferation (Fig. 3k). In contrast, high overactivation of the pathway by hemocyte specific expression of the constitutively activated receptor babo-CA resulted in reduced hemocyte numbers per larva ( Supplementary Fig. 6a-c) and drove hemocytes into apopotosis ( Supplementary Fig. 6d). Neither Actb overexpression nor dSmad2 pathway silencing in hemocytes increased the rate of hemocyte apoptosis ( Supplementary Fig. 6c,d). Based on this, we conclude that the level of Actb/dSmad2 signalling may determine the nature of the hemocyte response. At moderate activation levels, Actb/dSmad2 signalling is a positive regulator of hemocyte number that promotes proliferation, while high overactivation of the pathway drives cells into apoptosis. Consequently, we anticipate that the amplitude of Actb expression is likely to be tightly regulated.
Next we examined the role of Actb/dSmad2 signalling in hemocyte localization. Ectopic expression of Actb in areas typically devoid of hemocytes, such as the Sal expressing ventral areas of the epidermis and oenocytes (Fig. 3e,f) or imaginal discs, did not result in a uniform adhesion or attraction of hemocytes, i.e., in the alternating gap areas of the epidermis where no sensory neuron clusters are located (Fig. 3e,f). Further, uniform overactivation of the pathway by hemocyte expression of babo-CA showed a largely unaffected overall pattern of resident hemocytes, despite the above-mentioned apoptosis of hemocytes ( Supplementary Fig. 6a,b). This argued against a function of Actb in hemocyte chemoattraction by gradient formation and led us instead to focus on a potential role of Actb in the induction of hemocyte adhesion. Ectopic expression of Actb produced an overall trend of decreasing the fraction of circulating hemocytes, although this effect remained statistically insignificant by t-test (Fig. 3h), and accumulation of hemocytes in ectopic areas seemed minimal ( Fig. 3c-f). Seeking a more sensitive assay to quantify hemocyte adhesion, we took advantage of the fact that resident hemocytes of the HPs can be mobilized by mechanical disturbance and spontaneously re-adhere to the body wall within 30-45 min (ref. 4). Using an established protocol for this assay 22 , we examined the adhesive properties of hemocytes under various Actb/dSmad2 pathway conditions. Indeed, dSmad2 pathway knockdowns in hemocytes, or Actb silencing in PNS neurons, diminished hemocyte re-adhesion as evidenced by increased fractions of circulating hemocytes (Fig. 4a). However, dSmad2 pathway activation did not promote increased re-adhesion, suggesting an additionally needed, rate-limiting step in hemocyte adhesion. Moreover, we found that hemocyte-autonomous knockdowns of dSmad2 pathway components, but not Actb silencing in neurons, produced a lack in hemocyte cluster formation and self-adhesion ( Fig. 4b-e). Taken together, we conclude that the Actb/dSmad2 pathway directly or indirectly promotes hemocyte adhesion to the microenvironment (Fig. 4f, Model), a process that we predict requires in addition another, rate-limiting step. dSmad2 signalling may have an additional autonomous roles in hemocyte clustering/selfadhesion, which is revealed only when the pathway is blocked in hemocytes, thereby excluding alternative signalling by other Act family ligands as might be the case in Actb lof backgrounds.   Actb links sensory neuron activity with blood cell responses.
Considering the anatomical contacts of PNS neurons with hemocytes, and the role of Actb as neuron-emanating signal that supports hemocyte proliferation and adhesion, we sought to determine whether PNS neuron activity may regulate blood cell behaviours. Using a diagnostic driver for acetylcholine production, Cha-GAL4, we confirmed previous reports that PNS neurons are cholinergic ( Fig. 5a,b) 26 . This allowed us to use the pan Acetylcholine Receptor (AChR) agonist carbachol (carbamoylcholine) for the stimulation of larval sensory neurons. Interestingly, we found that cuticle exposure of intact larvae to carbachol promoted short-term recruitment of hemocytes to the HPs within 15 min, resulting in an enhanced resident hemocyte pattern (Fig. 5c, Fig. 5e-g). As a control, we expressed the same transgenes in glia using repo-GAL4 (ref. 29), yet no hemocyte phenotypes were observed ( Supplementary Fig. 7a-d), demonstrating specificity of the observed hemocyte responses to neuronal silencing. In addition to these short-term (15 min-2 h) effects on hemocyte localization to the HPs, we examined the long-term effects of stimulation by carbachol, or neuronal silencing by Kir2.1 or UASshi dn ts , on hemocytes. Interestingly, we found that carbachol exposure over several days of larval development significantly increased total hemocyte numbers per larva compared to controls according to Student's t-test (Fig. 5h). Consistently, silencing of PNS neurons through various regimens of heat shock induction of Kir2.1 or the dominant-negative dynamin shi dn ts over 20-48 h had opposite effects, resulting in larvae with reduced total hemocyte numbers (Fig. 5h). To examine whether the carbacholinduced increase in hemocyte number was based on proliferation, we quantified in vivo EdU incorporation of hemocytes. Indeed, carbachol-exposed larvae showed about 1.5-fold increased EdU incorporation over controls, an effect which interestingly was mainly seen in younger 2nd instar larvae (Fig. 5i). Further, carbachol-exposed larvae showed an overall larger increase in the resident hemocyte population of the hematopoietic sites compared to the circulating fraction (Fig. 5j), consistent with our model proposing enhanced proliferation of resident hemocytes in the HPs. Taken together, our data suggest that PNS neuronal activity supports hemocyte localization and expansion by proliferation, and predict a molecular signal from PNS neurons that is inducible by sensory activity.
To determine whether Actb might play a role as this inducible signal, we quantified Actb expression in sensory neurons of varying states of activity. Indeed, exposure of intact Drosophila larvae to carbachol induced a substantial and dose-dependent increase in Actb expression within 2-3 h, as quantified by the expression of a UAS-luciferase transgene driven by the Actb-GAL4 reporter (Fig. 6a). For this experiment, we focused on reporter expression in the PNS by carefully removing the CNS before luciferase quantification. To examine whether the increase in Actb-GAL4 expression was specific to sensory neurons, we coexpressed a UAS-GFP transgene. GFP imaging of freshly dissected larvae confirmed increased expression in PNS neurons, and did not show signs of ectopic expression in unrelated tissues (Fig. 6b-d). In a converse experiment, we examined whether inducible expression of Kir2.1 affects expression of UAS-GFP by the reporter Actb-GAL4 in PNS neurons. Inducing expression (through tub-GAL80 ts ) in limited time windows of 22 h to visualize GFP expression yet avoiding lethality by high expression of Kir2.1, we observed induction of GFP in Actb expressing PNS neurons of controls, while this GFP signal was largely absent in parallel cohorts silenced by coexpressed Kir2.1 (Fig. 6e-g). In addition to its effects on the PNS, Kir2.1 coexpression also reduced Actb-GAL4 driven GFP expression in the CNS (Fig. 6f,g).
Finally, we asked whether Actb is required for the induction of hemocyte responses on stimulation with carbachol. Comparing Actb ED80 null mutant larvae with Actb-competent controls, we found that loss of Actb attenuated the effect of carbachol-induced blood cell expansion (Fig. 6h). This suggested that Actb plays a major role in the cholinergic regulation of hemocytes, consistent with a model of neuronal activity-induced Actb expression (Fig. 6k). However, Actb mutants showed mild albeit by t-test insignificant carbachol-induced blood cell expansion (Fig. 6h) and partial short-term hemocyte recruitment to the HPs on 15 min of carbachol exposure (Fig. 6i,j), suggesting additional inducible signal/s that may contribute to these effects. Taken together, our findings support a model in which PNS neuronal activity promotes Actb expression, which in turn drives hemocyte expansion and long-term localization to the HPs (Fig. 6k).

Discussion
This research identified Actb as one of the elusive genes that govern hemocyte proliferation in the hematopoietic sites (HPs) of the Drosophila larva, as was predicted by previous functional studies 4 . Our work further links Actb RNA expression to the level of PNS neuronal activity. This model implies that increased expression of Actb would give rise to higher levels of active Actb protein, although the formal demonstration awaits development of a suitable tool for the detection of Actb protein. In the future, it will be interesting to study specific sensory stimuli that trigger hemocyte responses. Sensory neurons of the PNS have a prime function in detecting innocuous and noxious sensory stimuli such as mechanical strain, temperature, chemicals and light 30,31 , many of which signal potentially harmful conditions that may cause tissue damage. Thus, linking the detection of challenging conditions with the adaptive expansion of the blood cell pool may be an efficient system to elevate the levels of macrophages, to remove and repair damaged tissues, enhancing the overall fitness of the animal. Because Drosophila larval hemocytes persist into the adult stage 6,7 , the mechanism of sensory neuron-induced blood cell responses may allow adaptation of the animal beyond the larval stage.
In Drosophila self-renewing hemocytes, Actb/dSmad2 signalling has diverse effects on proliferation, apoptosis and adhesion. Our ex vivo data indicate that hemocyte proliferation is likely a direct effect, which is consistent with similar roles of babo/ dSmad2 in other tissues such as Drosophila imaginal discs and brain 11,21 , and TGf-b family dependent proliferation in vertebrate systems 32,33 . Echoing our findings of babo-CA driven hemocyte apoptosis, TGF-b family mediated direct or indirect effects on apoptosis have been described in invertebrate and vertebrate systems 33,34 . Overall, TGF-b family signalling is known for its multifaceted biological roles, depending on the cellular contexts and levels of ligand stimulation, which often translate into qualitatively distinct transcriptional and other cellular responses, that are mediated by both Smad and non-Smad signalling mechanisms 32,33 . While Drosophila Actb and possibly related TGF-b family ligands are known to signal through the induction of ecdysone receptor (EcR) in some but not all Drosophila tissues 15,35 , we found no indication for a link with EcR expression in hemocytes, suggesting other signalling mechanisms in the regulation of larval blood cell responses. In the studied Drosophila system, it further remains to be seen whether Actb/dSmad2 signalling has direct or indirect effects on hemocyte adhesion, and which other rate-limiting step/s may contribute to this process. Since hemocyte-autonomous loss of dSmad2 signalling causes a more severe phenotype than Actb lof, we speculate that other Act family ligands such as daw and myo, which are expressed in various tissues including surface glia, muscle, fat body, gut, and imaginal discs 11,17,20,36,37 may partially substitute for Actb in its absence. Overall, Actb is likely to be only one player in a more complex regulatory network. Future research will identify other inducible signals from neurons that regulate neuron-blood cell communications. This is predicted from Actb mutants that only partially block carbachol-induced blood cell responses. Actb/ dSmad2 lof and pathway silencing in hemocytes also reveal an underlying ability of the cells to compensate for the lack of this  Larvae À / þ carbachol exposed from hatching onward, and larvae À / þ genetic neuronal silencing as in (f) and (g), heats shocks to induce transgenes were applied as indicated; total hemocyte counts from single larvae (2nd instars, 66-78 h AEL corresponding to 2.3-2.7 mm). For each experimental cohort, total hemocyte numbers of experiment sets relative to side-by-side control sets of larvae are shown; n ¼ 3 to 5 and comparable effect in independent repeats. (i) In vivo EdU incorporation of hemocytes from larvae exposed to À / þ carbachol (0.7 mg ml À 1 ) from 1st instar (  observations of dSmad2 lof causing Mad overactivation have been reported in the Drosophila wing disc and neuromuscular junction previously 20,38 . Larval development may comprise distinct sensitive phases for the regulation of hemocyte responses. This is supported by carbachol promoting hemocyte proliferation preferentially in the early-mid 2nd instar larva, that is, at a stage when hemocytes are still tightly localized to the HPs 4,22 . Likewise, the effects of Actb lof and pathway silencing in hemocytes are more pronounced in younger larvae, suggesting a possible stronger dependence on the pathway, in addition to the emergence of compensatory mechanisms under lof conditions over time (above). Moreover, it will be interesting to investigate whether Actb signalling may not only vary temporally, but also by the ability of cell types to produce active Actb ligand, thereby influencing signalling outcomes, consistent with the cell type specific processing known for Activins and other ligands of the TGF-b family in both invertebrates and vertebrates 16,39,40 .
Drosophila Actb has previously been studied for its role in the formation and function of neuromuscular junctions in the Drosophila larva, where Actb expressing motor neurons project axons from the CNS, reaching from the center of the larva to the muscle layers of the body wall 16,17,20 However, resident hemocytes are shielded from these areas through the muscle layers of the body wall, which also form the base of the HPs, thereby creating an anatomical space between the muscle layers and epidermis 4 where resident hemocytes and Actb expressing sensory neurons colocalize (i.e., the Hematopoietic Pocket). The model that sensory neurons signal to adjacent hemocytes in the HPs is further supported by the fact that Actb silencing in motor neurons did not affect resident hemocyte localization and had, by t-test, no significant effect on hemocyte numbers. However, we cannot completely rule out involvement of alternative or additional scenarios, for example, that experimental manipulations of PNS activity, which also feed back to the CNS, would in turn trigger a signal to motor neurons that may respond by secreting Actb and/or another factor/s, thereby influencing hemocytes and/or the PNS itself. Likewise, although we confirmed the direct effect of Actb on hemocytes ex vivo, and found no signs of altered sensory neuron morphology under Actb lof/silencing, we cannot rule out that in the larva, Actb may contribute to molecular changes in the PNS that in turn might contribute to the observed hemocyte effects.
Sensory neurons of the HPs project axons to the CNS 41 , and our work shows that hemocytes are closely adjacent to and/or form direct contacts with sensory neurons, likely along the neuron cell bodies and dendrites, suggesting the communication involves non-canonical mechanisms. In Drosophila, as in vertebrates, signal transfer along all neuronal membrane surfaces, including dendritic synapses and dendrodendritic connections, have been described 42,43 , which may also form the interface in neuron-blood cell communication. The transcriptional induction of Actb in response to sensory stimuli recalls previous reports of the transcriptional upregulation of Actb in the formation of long-term memory in both flies and vertebrates 44,45 . This suggests parallels between the neuronal regulation within the CNS, and PNS-blood cell circuits, which will be an interesting subject for future study. Based on our findings and another recent report demonstrating that transcriptional regulation of the related BMP Decapentaplegic (Dpp) in the Drosophila wing epithelium depends on the K þ channel Irk2 (ref. 46), we propose that cellular electrochemical potential may be a more general theme in the expression of TGF-b family ligands.
Our findings in the Drosophila model pioneer a new concept that has not been shown in any vertebrate system to date-the neuronal induction of self-renewing, tissue-resident blood cells. These cells correspond to the broadly distributed system of selfrenewing myeloid cells that are present in most vertebrate organs, which by lineage are completely independent from blood cell formation fueled by hematopoietic stem cells [3][4][5][6]47,48 . In vertebrates, TGF-b family ligands such as Activin A and TGF-b regulate the activity and immune functions of macrophages, and cellular and humoral immune responses, in multiple ways through autocrine and paracrine signalling 49,50 . While the autonomic neuronal and glial regulation of hematopoietic stem and progenitor cells in the bone marrow has been recognized 51-55 , the role of sensory innervation in bone marrow hematopoiesis remains unknown. Even more so, nothing is known about the role of the nervous system in the regulation of the independent, self-renewing myeloid system of tissue macrophages. However, local neurons and sensory innervation of many organs including skin, lung, heart and pancreas [56][57][58][59] and inducible changes in the self-renewal rates of tissue macrophages 2 , suggest that principles of neuronal regulation are likely also at work in vertebrates, providing a link between neuronal sensing and adaptive responses of local blood cell populations.  To bleed circulating hemocytes, larvae were opened at the ventral anterior and posterior sides, and hemolymph was allowed to leak out, avoiding to apply pressure. To prevent dislodging of resident hemocytes or contamination with lymph gland hemocytes, larvae were monitored through a fluorescence stereomicroscope. Resident hemocytes were released by opening the remainder of the larva and scraping the body wall with a needle under microscope guidance, avoiding the lymph gland. For the release of total hemocyte numbers, both procedures were combined. Hemocytes were allowed to attach to glass slides for 15-30 min, followed by standard fixation (4% PFA) and immunocytochemistry 4,5 . For larval fillet preps, larvae were pinned down on Sylgard plates and ventrally filleted in a drop of PBS. Gut, fat body and trachea were removed, and the fillet was fixed for 20 min in 4% PFA. Fillets were washed in PBS, unpinned, permeabilized, blocked and immunostained according to standard protocols, staining overnight with gentle agitation 4,5 . Lymph gland dissections and stainings were performed as described in (ref. 69), using sets of larvae 72-80 h AEL (2.5-2.8 mm). Images were obtained on a Leica DMI4000B, and live larvae and larval fillets were imaged on a Leica M205FA stereomicroscope 4 . Leica SP5 confocal microscopy was used to obtain high-resolution images. In all experiments, identical settings for experiments and controls were maintained, and images were processed in Adobe Photoshop using identical recorded action settings.

Methods
Hemocyte quantification. To quantify resident hemocytes in segmental HPs, blood cells were visualized by fluorescence microscopy and counted by external inspection of intact live larvae, excluding the compact clusters of hemocytes in the terminal segment HP. Total numbers of hemocytes per larva, and percentages of circulating hemocytes were determined by selective release and automated hemocyte quantification using ImageJ 22 4 . This was achieved by timed embryo collections and measuring larval size using Leica LAS software. For total hemocyte counts, hemocyte numbers of experiment larvae were normalized relative to backgroundmatched parallel controls. Depending on the individual experiment, baseline counts of matched controls were typically in the range of 2,000 or 3,000 hemocytes; an example of non-normalized counts over a time course can be seen in Supplementary Fig. 2. Hemocyte number averages and circulating hemocyte fraction averages from at least 4-6 larvae per genotype were calculated. The number of larvae n per genotype is listed in the figure legends, with lower numbers typically referring to the control. Phenotypes were confirmed in independent biological replicates. Where applicable, standard deviations and 2-tailed t-tests were calculated. T-test values correspond to the following: NS (not significant) P40.05; *Pr0.05; **Pr0.01; ***Pr0.001; ****Pr0.0001.
Hemocyte re-adhesion assay. We used an in vivo hemocyte re-adhesion assay 22 . Specifically, hemocytes were dislodged by paint brush manipulation, and the fraction of hemocytes not returning to the resident state was determined after 45 min incubation, using quantitative isolation of circulating and resident hemocyte populations from single larvae 22 .
Apoptosis and cell proliferation assays. EdU (5-ethynyl-2 0 -deoxyuridine, Invitrogen) was used at 10 mM for 2 h for in vivo feeding experiments, or 1 mM for 2 h in cell culture medium. Hemocytes were released by scraping the larval carcasses, and stained in multi-well cell culture dishes ex vivo 4,22 . Click-iT EdU proliferation assays (Invitrogen), and TUNEL assays to quantify apoptotic cells (In Situ Cell Death Detection Kit, Roche Diagnostics) were performed according to manufacturers' instructions and used in hemocyte experiments previously 4 . For hemocyte ex vivo Actb stimulation, Actb was generated by transiently transfecting (Effectene, Qiagen) the plasmid pAcPA-dActb 70 and a control plasmid (pAct5C-GAL4 (M.Zeidler)) each into two Drosophila cells lines, changing medium after 8 h, and collecting conditioned media 48 h after the previous media change. Hemocytes of HmlD-GAL4, UAS-GFP; He-GAL4 larvae (72-78 h AEL, corresponding to 2.5-2.7 mm) were released in S2 medium (Gibco/Invitrogen). After settling of the hemocytes (B15 min), medium was replaced by the Actb conditioned media or control media. Starting 1 h after stimulation, EdU was added for 2 h, and cells were fixed and stained as described above. Fractions of EdU positive cells among GFP positive hemocytes where quantified by Metamorph, maintaining identical settings between stimulated and unstimulated conditions. For in vivo EdU experiments, EdU positive hemocytes were quantified by Metamorph and/or manual counting.
Actb-GAL4 reporter luciferase assay and quantification of GFP expression. Actb transcription was quantified based on combining UAS-luciferase and/or UAS-CD4-tdGFP with the Activin-b enhancer reporter Actb-GAL4. Larvae were stimulated with carbachol for 2-3 h (see above). Before lysis, larvae were dissected and the anterior portion with the highly Actb-GAL4 expressing CNS was carefully removed before lysis. Luciferase levels were quantified using Bright-Glo according to the manufacturer's instructions (Promega). For quantification of GFP expression, images of controls and carbachol-stimulated fillets were taken at identical settings (Leica M205 fluorescence stereoscope or Leica SP5 confocal). Images were cropped to corresponding regions of identical areas, and image signal intensity was quantified using ImageJ. Signal intensity of carbachol-stimulated samples relative to unstimulated controls was calculated and averaged.
Data availability. The authors declare that all data supporting the findings of this study are available within the article and its Supplementary Information files or from the corresponding author on reasonable request.