Developing tissues produce excessive numbers of cells and selectively destroy a subpopulation through programmed cell death to regulate growth. Rapid clearance of cell corpses is essential for maintaining tissue homeostasis and preventing the release of potentially cytotoxic or antigenic molecules from dying cells, and defects in cell corpse clearance are closely associated with autoimmune and inflammatory diseases10-13
. In C. elegans
the CED-1 receptor is expressed in engulfing cells, where it acts to recognize cell corpses and drive their phagocytosis1
. CED-1 promotes engulfment through an intracellular NPXY motif, a binding site for proteins containing a phosphotyrosine-binding (PTB) domain, and a YXXL motif, a potential interaction site for proteins containing SH2 domains1
. The PTB domain adaptor protein CED-6 can bind the NPXY motif of CED-1 (ref. 14
), is required for cell corpse engulfment15
and acts in the same genetic pathway as CED-1 (ref. 16
). CED-1 ultimately mediates actin-dependent cytoskeletal reorganization through the Rac1 GTPase17
, and Dynamin modulates vesicle dynamics downstream of CED-1 during engulfment18
, but the molecular signalling cascade that allows CED-1 to execute phagocytic events remains poorly defined.
Glia are the primary phagocytic cell type in the developing and mature brain. Glia rapidly engulf neuronal cell corpses produced during development, as well as neuronal debris generated during axon pruning19,20
or during Wallerian degeneration in the adult brain21
. In Drosophila
, glial phagocytosis of these engulfment targets requires Draper, the fly orthologue of CED-1 (refs 2, 4-6
). Draper, like CED-1, contains 15 extracellular atypical epidermal growth factor (EGF) repeats, a single transmembrane domain, and NPXY and YXXL motifs in its intracellular domain ()2
Ced-6 is also required for the clearance of pruned axons4
, indicating possible conservation of the interaction between CED-1 and CED-6 in flies, but additional signalling molecules acting downstream of Draper have not been identified.
Shark binds an ITAM in the Draper intracellular domain
In a yeast two-hybrid screen for molecules interacting with the regulatory region of Shark22,23
, we identified Draper. We found that when LexA-Shark, constitutively active Src kinase and AD-Draper are present, Shark and Draper interact physically (). In the absence of Src kinase, Shark and Draper fail to interact, indicating that phosphorylation of Draper by Src may be essential for Shark-Draper interactions. We found that the Draper intracellular domain contains an ITAM (YXXI/L-X6-12
-YXXL), a key domain found in many mammalian immunoreceptors including Fc, T-cell and B-cell receptors. SFKs phosphorylate the tyrosines in ITAM domains, thereby allowing ITAM association with SH2-domain-containing signal transduction proteins including Syk and Zap-70 (refs 9, 24
). We therefore generated Y→F substitutions of the tyrosine residues within or near the Draper ITAM, and found that Tyr 949 and Tyr 934 were critical for robust Draper-Shark binding (). These correspond to the consensus tyrosine residues in the predicted Draper ITAM (). We next transfected plasmids with carboxy-terminally haemagglutinin-tagged Draper (Draper-HA) or with Draper-HA and Shark with an amino-terminal Myc tag (Myc-Shark) into Drosophila
S2 cells, immunoprecipitated with anti-HA antibodies, and performed western blots with anti-phosphotyrosine, anti-Myc and anti-HA antibodies (). We found that Myc-Shark co-immunoprecipitated with Draper-HA, and that anti-phosphotyrosine antibodies labelled a band corresponding to the position of Draper-HA that was absent in empty vector controls. Further, we found that a Y949F substitution markedly reduced Draper-Shark association (). Taken together, these data indicate Draper and Shark can associate physically through the Draper ITAM domain.
We next sought to determine whether Shark is required for glial phagocytic activity in vivo
. Severing adult Drosophila
olfactory receptor neurons (ORNs) initiates Wallerian degeneration of ORN axons. Antennal lobe glia surrounding these severed axons respond to this injury by extending membranes towards severed axons and engulfing degenerating axonal debris7
. These glia express high levels of Draper, and in draperΔ5
null mutants, glia fail to respond morphologically to axon injury, and severed axons are not cleared from the central nervous system (CNS)7
. Thus, both the extension of glial membranes to severed axons and the phagocytosis of degenerating axonal debris require Draper signalling.
We explored whether Shark function in glia is essential for glial responses to axon injury by driving an upstream activating sequence (UAS)-regulated double-stranded RNA interference construct designed to target shark (sharkRNAi) with the glial-specific repo-Gal4 driver, severing ORN axons, and assaying the recruitment of Draper and green fluorescent protein (GFP)-labelled glial membranes to severed axons. Maxillary palp-derived ORN axons project to 6 of the roughly 50 glomeruli in the antennal lobe. Within hours after maxillary palps have been ablated in control animals, Draper immunoreactivity decorates severed axons projecting to (, arrow) and within maxillary palp ORN-innervated glomeruli (), and GFP-labelled glial membranes are recruited to these severed axons (). Strikingly, knocking down Shark in glia completely suppressed these events (). We next severed antennal ORN axons; these axons project to about 44 of the 50 antennal lobe glomeruli. Antennal ablation therefore injures nearly all glomeruli in the antennal lobe and results in the majority of antennal lobe glia in control animals upregulating Draper (, open arrowhead) and undergoing hypertrophy (). We found that knocking down Shark in glia also blocked this glial response to axon injury (). Thus, Shark is essential for all axon-injury-induced changes in glial morphology and Draper expression.
Shark is required for recruitment of Draper and glial membranes to severed axons
To determine whether Shark is required for glial phagocytosis of severed axons we labelled a subset of maxillary palp ORN axons with mCD8::GFP, knocked down Shark function in glia, and assayed the clearance of severed axons. In control animals severed GFP-labelled ORN axonal debris was cleared from the CNS within 5 days (). In contrast, glial-specific sharkRNAi
potently suppressed the clearance of degenerating axons, with severed axons lingering in the CNS for at least 5 days (). We then examined whether mutations in the shark
gene affected the glial clearance of degenerating axons. The null allele of shark, shark1
, is pupal lethal25
. We therefore assayed glial responses to axon injury in shark1
heterozygous mutants, and tested for dominant genetic interactions between draperΔ5
. We found that both draperΔ5
/+ and shark1
/+ animals showed defects in glial phagocytic function: 5 days after injury, significant amounts of axonal debris remained within OR85e-innervated glomeruli () and in the maxillary nerve (). Moreover, shark1
/+ animals showed a striking suppression of glial clearance of severed axons almost equivalent to that of draperΔ5
mutants (). Thus, shark mutations dominantly suppress the glial clearance of degenerating ORN axons, and this phenotype is strongly enhanced by removing one copy of draper
. These data, taken together with our sharkRNAi
data, show that Shark is essential for the clearance of degenerating axons by glia.
Shark is required for glial clearance of severed axons from the CNS
Is Shark required for the glial clearance of neuronal cell corpses? In embryonic stage 14-15 control animals we found 24.4 cell corpses per hemisegment (). In contrast, we found that shark1 null mutants showed a marked increase in CNS cell corpses, with null mutants containing almost twice as many corpses per hemisegment (43.3 cell corpses per hemisegment; ). shark1/Df(2R)6063 mutants accumulated cell corpses at levels similar to those in shark1, indicating that this phenotype maps to shark. These cell corpse engulfment phenotypes are indistinguishable from that of draperΔ5 mutants (38.5 ± 1.68 cell corpses per hemisegment; ). We conclude that Shark, like Draper, is also essential for the efficient clearance of embryonic neuronal cell corpses by glia.
Quantification of cell corpse engulfment defects in shark and draper mutants
Because we found that Shark binds Draper only in the presence of an active Src kinase in our two-hybrid assays, we screened Drosophila
Src kinases for roles in glial phagocytic activity. Interestingly, we found that glia-specific knockdown of Src42A (src42ARNAi
) potently suppressed glial phagocytic activity: in src42ARNAi
animals, Draper was not recruited to severed maxillary palp axons (); glial hypertrophy and upregulation of Draper after antennal ablation was blocked (); and GFP-labelled severed axons lingered in the CNS for 5 days (). Knockdown of two other Drosophila
Src kinases, Btk29A and Src64B, had no effect on the glial phagocytosis of severed axons (Supplementary Fig. 1
). Thus, Src42A seems to be essential for all morphological responses of glia to axon injury and for the efficient clearance of degenerating axonal debris from the CNS.
Src42Aa is required for glial responses to axon injury and modulates Draper phosphorylation status
We predicted that Draper phosphorylation status should be sensitive to the SFK inhibitor PP2. Indeed, addition of PP2 to S2 cultures led to a decrease in the phosphorylation of Draper (, lanes 3 versus 4, 5 versus 6, 7 versus 8, and 10 versus 11) and Draper-Shark association (, lanes 5 versus 6, and 10 versus 11). Strikingly, co-transfection of Draper and Src42A led to a marked increase in Draper phosphorylation (, lane 7), which was PP2-sensitive (, lane 8) and Draper-specific (, lane 9). Draper-Shark interactions are not dependent on Shark kinase activity because kinase-dead Shark (Shark K698R) associates with Draper (, lane 10). These data indicate that Src42A may phosphorylate the Draper intracellular domain, thereby increasing the association of Shark with Draper and the activation of downstream glial phagocytic signalling.
We have identified Shark and Src42A as novel components of the Draper pathway. One potential model for Draper-Shark-Src42A interactions is that Shark and Src42A drive the recruitment of Draper to engulfment targets. However, CED-1 has been shown to cluster around cell corpses even in the absence of its intracellular domain1
. Moreover, Zap-70 and Syk bind phosphorylated ITAM domains in mammalian immunoreceptors when ITAM domains are phosphorylated by Src after ligand-dependent receptor clustering26-28
. We therefore favour a model in which the engagement of Draper with its ligand (presumably presented by engulfment targets) promotes receptor clustering, tyrosine phosphorylation of Draper by Src42A, association of Shark, and activation of downstream phagocytic signalling events. Our work suggests that Draper is an ancient immunoreceptor in which the extracellular domain is tuned to recognize modified self and the intracellular domain signals through ITAM-Src-Syk-mediated mechanisms. This is the first identification of ITAM-Src-Syk signalling in invertebrates, and it suggests that a pathway similar to Draper-Ced-1 may ultimately have given rise to ITAM-based signalling cascades in mammalian myeloid and lymphoid cells, including those regulated by Fc, B-cell and T-cell receptors.