DRK/DOS/SOS converge with Crk/Mbc/dCed-12 to activate Rac1 during glial engulfment of axonal debris

Significance

Neuronal cell death or injury leads to the production of neuronal cell corpses and cellular debris that must be cleared from the nervous system to avoid inflammation or toxicity. Glia are the primary cell type responsible for clearing neuronal debris, but precisely how these cells recognize, phagocytose, and destroy this material remains poorly defined. Here we use a simple nerve injury assay to identify genes downstream of the engulfment receptor Draper that are required for efficient glial engulfment of degenerating axons. Based on the molecular and functional conservation of this pathway in mammals, this study sheds new light on pathways potentially used by mammalian glia to react to brain injury or neurodegeneration.

Abstract

Nervous system injury or disease leads to activation of glia, which govern postinjury responses in the nervous system. Axonal injury in Drosophila results in transcriptional up-regulation of the glial engulfment receptor Draper; there is extension of glial membranes to the injury site (termed activation), and then axonal debris is internalized and degraded. Loss of the small GTPase Rac1 from glia completely suppresses glial responses to injury, but upstream activators remain poorly defined. Loss of the Rac guanine nucleotide exchange factor (GEF) Crk/myoblast city (Mbc)/dCed-12 has no effect on glial activation, but blocks internalization and degradation of debris. Here we show that the signaling molecules downstream of receptor kinase (DRK) and daughter of sevenless (DOS) (mammalian homologs, Grb2 and Gab2, respectively) and the GEF son of sevenless (SOS) (mammalian homolog, mSOS) are required for efficient activation of glia after axotomy and internalization/degradation of axonal debris. At the earliest steps of glial activation, DRK/DOS/SOS function in a partially redundant manner with Crk/Mbc/dCed-12, with blockade of both complexes strongly suppressing all glial responses, similar to loss of Rac1. This work identifies DRK/DOS/SOS as the upstream Rac GEF complex required for glial responses to axonal injury, and demonstrates a critical requirement for multiple GEFs in efficient glial activation after injury and internalization/degradation of axonal debris.

Activation of glia is a hallmark of nearly all neurodegenerative diseases and neural injuries. When brain insults have occurred, glia rapidly change their morphology and gene expression profiles to invade the injury site and clear pathogens and/or neuronal debris by phagocytic engulfment (1, 2). Failure to clear debris from the CNS can result in prolonged neuroinflammation and hamper the recovery of the CNS (3, 4). However, the genetic pathways promoting glial activation after neural injuries remain poorly defined.

Genetic studies of Wallerian degeneration in Drosophila have provided important insights into glial responses to axotomy (5). Olfactory neuron axotomy results in the degeneration of axons projecting into the antennal lobe of the fly brain, where local glia sense degenerating axons, and initiate a multistep process of reactivity. Reactive glia up-regulate the transcription of the engulfment receptor Draper (drpr) and extend membranes to degenerating axons (5, 6). When at the injury site, glia internalize axonal debris and degrade it through the phagolysosomal pathway (7). Finally, glia terminate their responses by withdrawing from the injury site and down-regulating Draper, and finally return to a resting state (6).

Draper is essential for all glial responses to axonal injury. In drpr-null mutants, glia fail to respond morphologically to axonal injury, and axonal debris lingers in the brain for weeks after axotomy (5). Downstream of Draper, the small GTPase Rac1 appears to be critical in executing glial activation to axon injury, as loss of Rac1 phenocopies drpr-null mutants (7). The only Rac guanine nucleotide exchange factor (GEF) known to be required for glial engulfment of axonal debris is the noncanonical GEF Crk/myoblast city (Mbc)/dCed-12. However, in contrast to loss of Rac1, animals lacking Crk/Mbc/dCed-12 signaling exhibit relatively normal activation of glia after axotomy, with glia increasing Draper expression and extending membranes to degenerating axons, but glia then fail to internalize and degrade axonal debris (7). These data argue for a specific role for the Crk/Mbc/dCed-12 complex at the internalization/degradation phase of the glial response, and suggest that an additional Rac1 GEF must act earlier during initial activation of glial responses to axonal injury.

In an RNAi-based screen for new engulfment genes, we identified downstream of receptor kinase (drk) as a gene required for efficient glial clearance of degenerating axons. drk is best known for its role in signaling downstream of the Sevenless (Sev) receptor tyrosine kinase (RTK), where it functions with daughter of sevenless (dos) and son of sevenless (sos) to activate the small GTPase Ras (8–13). More recent studies have also linked SOS to the activation of Rac1 in the regulation of axon guidance during Drosophila embryonic CNS development (14), indicating that DRK/DOS/SOS can act upstream of multiple small GTPases. Here we show that the DRK/DOS/SOS complex plays a critical role in activation of glial responses to injury and internalization of axonal debris. Moreover, we provide genetic evidence that, at the earliest stage of glial activation, DRK/DOS/SOS function redundantly with Crk/Mbc/dCed-12 to promote Rac1 activation and initiate all steps in glial responses to axonal injury.

Results

DRK, DOS, and SOS Are Required for Glial Engulfment of Axonal Debris.

To identify new pathways required for glial engulfment of degenerating axons, we performed an RNAi-based screen for genes required in glia for clearing axonal debris after axotomy. We expressed each of ∼500 UAS-RNAi constructs from the Vienna Drosophila Resource center (15) by using the pan-glial driver repo-Gal4 (16). For animals that did not survive to adulthood, we further incorporated a temperature-sensitive version of Gal80 (Gal80^ts) (17) in the background to temporally control the induction of RNAi exclusively in the adult glia (SI Materials and Methods). We ablated maxillary palps (mps) in which a subset of olfactory receptor neurons (ORNs) were labeled with membrane-tethered GFP (OR85e-mCD8::GFP) (18) and scored axonal debris clearance 5 d after axotomy by quantifying GFP immunoreactivity of OR85e⁺ glomerulus in the antennal lobe, as previously reported (5).

In our primary, screen we found that an RNAi construct (drk^RNAi#105498) targeting DRK suppressed glial clearance of axonal debris. In control animals, the vast majority of axonal debris was cleared 5 d after axotomy (Fig. 1A). However, a significant amount of the axonal debris still lingered in the brain of drk^RNAi animals at day 5, and ultimately cleared in 20 d after axotomy (Fig. S1A; quantified in Fig. S1B), which argues for a glial role for DRK in engulfment of axonal debris. DRK is known to physically interact with the adaptor protein DOS (19) and the GEF SOS (20), which can activate downstream small GTPase such as Ras and Rac1 (8, 9, 14, 21). We therefore designed a UAS-RNAi construct (dos^RNAi#3) to knock down DOS in glia and assayed for engulfment defect. Adult glia expressing dos^RNAi#3 also exhibited a significant delay in engulfment of axonal debris (Fig. 1A), arguing that DOS also plays an important role in glial engulfment of degenerating axons. To further validate this result, we repeated this experiment with an additional RNAi line (dos^1044R-3), which targets different region of dos mRNA from dos^RNAi#3, and found similar engulfment defects 5 d after injury (Fig. S2), arguing that the phenotype was not caused by the off-target effects of RNAi. We next knocked down glial SOS by using RNAi (sos^RNAi#42849) and found that axonal debris remained uncleared in the CNS for as long as 20 d (Fig. 1A and Fig. S1A). We did not observe significant change in the expression of Draper when DRK, DOS, and SOS were knocked down respectively (Fig. S1 E and F), implying that the delayed clearance is not caused by a lower level of Draper expression. Together, these data suggest that, during glial engulfment, similar to Sev RTK signaling, DRK, DOS, and SOS interact with each other to regulate downstream small GTPase activity.

Fig. 1.

Glial clearance of axonal debris requires DRK, DOS, and SOS. (A) Glia failed to clear OR85e⁺ axonal debris in 5 d after axotomy in drk RNAi (drk^RNAi), dos RNAi (dos^RNAi), and sos RNAi (sos^RNAi) animals and the clearance defects in drk^RNAi and dos^RNAi animals were completely rescued by a GOF allele of sos, Sos^JC2. Axons from a subset of mp ORNs were labeled by OR85e-mCD8::GFP and the integrity of axons 5 d after mp ablation (5 d after injury) was examined under confocal microscope. Representative images (z-stack) are shown. (Scale bar: 30 μm.) SI Materials and Methods provides all genotypes throughout the figures. (B) Quantification of OR85e⁺ axonal debris remaining in the brain 5 d after injury. GFP immunoreactivity of each OR85e⁺ glomerulus was measured and normalized to uninjured, age-matched cohorts (as 100%). Error bars represent SEM throughout. n.s., not significant. (*P < 0.05, **P < 0.005, and ***P < 0.0001; one-way ANOVA and Bonferroni post hoc throughout unless otherwise mentioned.)

During Sev signaling, DRK and DOS couple RTK activation to stimulation of downstream small GTPase activity through SOS. We therefore speculated that increasing SOS activity could potentially compensate for the depletion of DRK and DOS. To test this hypothesis, we explored the effect of a gain-of-function (GOF) SOS allele (Sos^JC2) (8, 22) on the ability of glia to clear axonal debris when DRK or DOS was knocked down. Sos^JC2/+ animals did not exhibit any discernible clearance defect (Fig. 1A), nor did they clear debris faster than controls (Fig. S3A; quantified in Fig. S3B). However, we found the delay in clearance of axonal debris caused by drk^RNAi and dos^RNAi was completely rescued by Sos^JC2/+ (Fig. 1A; quantified in Fig. 1B). These data indicates that the DRK/DOS/SOS complex is required for efficient engulfment of axonal debris by glial cells, and that activation of SOS is sufficient to drive glia to engulf axonal debris when DRK or DOS are depleted, consistent with the notion that SOS acts downstream of DRK and DOS.

Glial DRK Is Recruited to Degenerating Axons After Injury.

After axon injury, Draper is up-regulated in glial cell and recruited to sites where glia actively engulf axonal debris (5, 6). We sought to determine whether DRK was also expressed in glia, as our RNAi data would suggest, and whether it was recruited to injury sites during glial engulfment. We used α-DRK polyclonal antibodies (11) to detect DRK expression in the adult brain. We first examined DRK localization along the maxillary nerve in the subesophageal ganglion (SOG), through which GFP-labeled mp ORN axons are projected to the antennal lobe. In control brains, we observed widespread DRK expression and, interestingly, 1 d after mp ablation, DRK was enriched along the maxillary nerve, which was absent when we drove drk^RNAi in glia (Fig. 2A), indicating that glial DRK is recruited to severed axons. Moreover, we found that Draper is required for the recruitment of DRK after injury because we could not detect up-regulation of DRK 1 d after axotomy in drpr-null (drpr^Δ5) animals (Fig. S4), suggesting that the activation of DRK/DOS/SOS requires Draper.

Fig. 2.

DRK is recruited to glial membranes surrounding the injured axons. (A) Endogenous DRK (red, α-DRK) in glia was recruited to the severed mp nerves labeled with OR85e-mCD8::GFP (green, α-GFP) 1 d after axotomy. Before injury, DRK expression was evenly distributed in the SOG but not colocalized with the OR85e⁺ maxillary nerve. One day after mp injury, DRK expression was increased around the degenerating maxillary nerves (arrowheads), which was not seen in drk^RNAi animals. Representative images (single slice) are shown. (Scale bar: 10 μm.) (B) Endogenous DRK expression in glia was increased 1 d after axotomy. Glial membranes were labeled with mCD8::GFP by glia-specific repo-Gal4 driver. Antennal ablation (removal of the third segment of antennae) was performed to induce a greater extent of axon degeneration in the antennal lobe (AL). Normally, thin glial membranes ensheath (arrows) the antennal lobe and each glomerulus (no injury, control). However, 1 d after antennal ablation, ensheathing glia became hypertrophy (1 d after injury, dashed lines), and strong DRK immunoreactivity (red) was found in hypertrophic region of ensheathing glia (yellow), but not when drk^RNAi was expressed by repo-Gal4, indicating that the increase of DRK is glia-specific. Representative images (single slice) are shown. C, cortex. (Scale bar: 10 μm.)

We next sought to determine if DRK expression was increased after antennal ablation, which leads to a dramatic increase in Draper expression and hypertrophy of glial membranes (5). We labeled glial membranes with mCD8::GFP by using the repo-Gal4 driver, and assayed glial morphology and DRK expression in control animals and animals where the third antennal segments had been ablated 1 d earlier. Consistent with our findings in the SOG, DRK immunoreactivity was dramatically increased around the antennal lobe 1 d after antennal ablation in control flies (Fig. 2B). The increase in DRK is likely a result of recruitment of DRK to glial membranes at sites of axon injury rather than up-regulation of drk gene transcription and/or translation, as DRK protein levels did not increase significantly after ablation of olfactory organs (Fig. S5). When we drove drk^RNAi in glia, there was a significant decrease of DRK immunoreactivity in the hypertrophic ensheathing glia (Fig. 2B, arrows) but not in neurons, confirming that the increase of DRK immunoreactivity comes from glia. Together, these results are consistent with the model that DRK acts downstream of Draper to promote engulfment of axonal debris.

DRK-SOS Activation Helps Glia to Internalize Axonal Debris and Activate the Phagolysosomal Program.

After axotomy, glial membranes are recruited to severed axons, where they internalize and degrade axonal debris (7). As the first step to determine where DRK impinged on glial engulfment of axonal debris, we examined glial membrane dynamics after mp injury. In healthy, uninjured animals, glia usually wrap each glomerulus with their fibrous membranes. After axon injury, glia form large numbers of membranous vesicles throughout the field of degenerating axons (Fig. 3A); these vesicles often contain internalized axonal debris, presumably to become acidic thereafter for degradation (7). However, we found that, unlike in control animals, knocking down DRK in glia by using repo-GAL4 resulted in a failure of glia to efficiently form enclosed vesicles 1 d after injury, with total vesicle number in drk^RNAi animals being reduced to ∼30% of that found in the controls (Fig. 3B). DRK activity in glial membrane vesicle formation appears to act through SOS, as the Sos^JC2/+ was sufficient to suppress the loss of glial membrane vesicles in the drk^RNAi background.

Fig. 3.

DRK-SOS activation is required for glia to internalize axonal debris. (A) Glia failed to form membrane vesicles 1 d after axotomy in drk RNAi animals, which was recued by Sos^JC2. Glial membranes were labeled by repo-Gal4 driving mCD8::GFP. Without injury, glial membranes normally loosely wrap each glomerulus (delineated by dotted lines) in the antennal lobe. However, 1 d after axon injury, glial membranes invaded the field of injured glomerulus and formed enclosed membrane vesicles (arrowhead). These membrane vesicles were largely absent in animals in which DRK expression was knocked down by repo-Gal4 (drk^RNAi). Sos^JC2 alone did not increase the number of vesicles formed 1 d after injury, but rescued the defect of glial membrane vesicle formation in drk^RNAi. Representative images (single-slice) are shown. (Scale bar: 10 μm.) (B) Quantification of the number of glial membrane vesicles in A. Because there were hardly any detectable glial membrane vesicles without axon injury, we quantified only the number of vesicles in animals at day 1 after axon injury (n ≥ 3 for all). (C) Internalization of OR85e-mCD8::GFP-labeled axonal debris was reduced when drk was knocked down in ensheathing glia but restored by Sos^JC2. Ensheathing glial membranes were labeled with mCD4::tdTomato (red) driven by TIFR-Gal4. OR85e⁺ axons were labeled with mCD8::GFP (green). Degenerating OR85e⁺ axons were often found inside glial membrane vesicles (arrows). Representative images (single-slice) are shown. (Scale bar: 10 μm.) (D) Quantification of the number of glial vesicles containing GFP-labeled axonal debris 1 d after axotomy in C (n ≥ 3 for all).

We next sought to determine whether reduced DRK function affected the ability of glia to internalize axonal debris. We labeled ensheathing glia with UAS-mCD4::tdTomato by using the TIFR-Gal4 driver (7) and compared internalization of OR85e-mCD8::GFP-labeled ORN axons in control and drk^RNAi animals. Internalized axonal debris was scored when GFP⁺ axonal debris was found within tdTomato⁺ membrane-enclosed vesicles. Before injury, tdTomato-labeled glial membrane vesicles were not detectable in the OR85e⁺ glomerulus. However, in controls, 1 d after mp ablation, glial membranes invaded the glomerulus and tdTomato-labeled vesicles containing mCD8::GFP⁺ axonal materials were found throughout this glomerulus (Fig. 3C, arrows). Expression of drk^RNAi in TIFR⁺ glia significantly reduced the number of glial vesicles containing axonal debris compared with the control (Fig. 3D). Activating SOS signaling in the drk^RNAi background through GOF Sos^JC2 resulted in glial internalization of axonal debris at WT levels. Sos^JC2 itself did not increase the total number of glial vesicles, nor did it alter the number of vesicles internalizing degenerating axons, arguing that the rescuing effect did not result from a Sos^JC2-dependent general increase in vesicle number. We therefore concluded that DRK-SOS signaling plays an important role in forming phagocytic vesicles during glial engulfment of axonal debris.

After internalization of axonal debris, glia activate the phagolysosomal pathway to degrade engulfed axonal materials (23). To explore if activation of the phagolysosomal pathway was also affected by altered DRK-SOS signaling, we examined the formation of acidified phagolysosomes surrounding severed axons by using LysoTracker red DND-99. In control uninjured brains, the area surrounding OR85e⁺ axons did not exhibit detectable LysoTracker⁺ puncta (Fig. S6A). One day after injury, LysoTracker⁺ puncta appeared robustly along the degenerating OR85e⁺ axons. Knocking down DRK in ensheathing glia resulted in ∼75% of reduction in the number of LysoTracker⁺ puncta (Fig. S6B), indicating that the loss of DRK function severely impedes glial activation of phagolysosomal pathway. We detected a slight increase in the efficiency of phagolysosome maturation with Sos^JC2, as revealed by a ∼1.5-fold increase in the number of LysoTracker⁺ puncta compared with the controls, suggesting that activation of SOS promotes phagolysosome formation. However, we found that Sos^JC2 was not sufficient to rescue the phagolysosome maturation defect caused by drk^RNAi. This lack of rescue by Sos^JC2 could indicate that the Sos^JC2 allele is not sufficiently strong to overcome the absence of DRK activity during phagolysosome formation, or that another molecule might act in a redundant fashion with DRK during this specific signaling step.

In summary, these data argue that DRK-SOS signaling plays a critical role during glial internalization of axonal debris and activation of the phagolysosomal program for degradation of axonal materials.

Rac1, but Not Ras, Is the Main Small GTPase Effector Downstream of SOS in Activating the Engulfing Glia.

SOS is an evolutionarily conserved GEF for the small GTPase Ras, and DRK/DOS/SOS were initially identified as key molecules coupling the Sev RTK to Ras activation (8–11, 13, 24). We therefore explored the possibility that Ras might act as the downstream effector after SOS activation to promote glial clearance of axonal debris. We expressed a dominant-negative (DN) form of Ras (Ras85D^N17) (25) specifically in adult glia and assayed glial engulfment after axotomy. In these animals, 5 d after mp ablation, we found that the vast majority of axonal debris was cleared (Fig. S7A), as <10% of axonal debris was left 5 d after injury (Fig. S7B). This phenotype is extremely mild compared with what we observed with dos, drk, and sos RNAi. Although we could not exclude the possibility that Ras is activated by SOS after injury and plays a minor role in clearance of axonal debris, we suspected additional GTPases might be the key downstream effectors of SOS during glial engulfment.

The small GTPase Rac1 is a potent regulator of membrane ruffling and lamellipodia formation, and is required for glial phagocytosis of degenerating axons (7). Interestingly, SOS possesses Rac1-specific GEF activity in mammals and Drosophila through its conserved N-terminal Dbl homology domain (14, 21). We therefore assayed for genetic interactions between Rac1 and SOS. In Drosophila eye development, overexpression of a DN Rac1 (Rac1^N17) results in a rough eye phenotype that can be suppressed by supplying excess upstream GEF complex proteins, Mbc and dCed-12 (26). We therefore reasoned that overactivation of SOS by use of Sos^JC2 might suppress Rac1^N17 phenotypes during glial engulfment of axonal debris. Adult-specific overexpression of Rac1^N17 in glia resulted in a potent suppression of axonal debris clearance as previously described (7) (Fig. 4A). However, when we overexpressed Rac1^N17 in Sos^JC2/+ background, we observed a ∼50% suppression of clearance defect compared with animals expressing Rac1^N17 alone (Fig. 4B). As Rac1^N17 also blocks glial membrane extension to the site of injury (7), we also examined whether Sos^JC2 could rescue this defect by scoring recruitment of Draper to severed axons 1 d after axotomy (Fig. 4 C and D). We found that, consistent with our previous study, glial expression of Rac1^N17 completely suppressed the recruitment of Draper to severed axons, but overactivation of SOS using Sos^JC2 partially restored (to ∼30% control levels) glial recruitment of Draper to axonal debris. The simplest interpretation of these data, based on interactions between SOS and RAC1 in other signaling contexts such as the Sevenless pathway, is that the GEF component SOS acts upstream of the small GTPase RAC1. However, we cannot exclude the formal possibility that SOS acts downstream of RAC1 to activate another small GTPase during glial engulfment of axonal debris.

Fig. 4.

SOS modulates Rac1 activity in glial response to axonal injury. (A) Sos^JC2 suppressed the inhibitory effects of DN Rac1 (Rac1^N17) on glial clearance of axonal debris. The clearance assay was performed as described in Fig. 1. (Scale bar: 30 μm.) (B) Quantification of data in A (n = 10 for all). (C) Sos^JC2 partially rescued Rac1^N17 and increase glial membrane (red, α-Draper) recruitment to severed OR85e⁺ glomerulus (green, α-GFP) 1 d after axotomy. (Scale bar: 10 μm.) (D) Quantification of Draper immunoreactivity at OR85e⁺ glomerulus in C. a.u., arbitrary unit. n = 10 for all.

Both DRK/DOS/SOS and Crk/Mbc/dCed-12 Activation Contribute to Rac1 Activity in Glial Response to Axon Injury.

Our previous finding argued that the GEF complex Crk/Mbc/dCed-12 acts upstream of Rac1 in glial clearance of axonal debris (7), but there were significant differences in the requirements for Rac1 vs. Crk/Mbc/dCed-12. Specifically, depletion of Rac1 activity by RNAi or use of DN constructs completely blocked glial activation: there was no recruitment of glial membranes or Draper to sites of axon injury, along with severe defects in axonal debris clearance. In contrast, depletion of Crk, Mbc, or dCed-12 resulted in a slight delay in glial recruitment to severed axons, suggesting a complementary GEF acting upstream of Rac1 might also play a role in glial response to axon injury. Based on the similarity of the phenotypes associated with inhibition of Crk/Mbc/dCed-12 and DRK/DOS/SOS (i.e., reduced vesicle formation, failure to activate the phagolysosomal pathway, and an axonal debris clearance defect), we speculated that Crk/Mbc/dCed-12 and DRK/DOS/SOS activation might act redundantly downstream of Draper to activate Rac1 not only at the early phases of the glial response (when glial membranes are recruited to severed axons), but also at later phases (during internalization and degradation of axonal debris). This predicts that simultaneous depletion of Crk/Mbc/dCed-12 and DRK/DOS/SOS should result in a complete suppression of glial activation. To test this model, we simultaneously knocked down SOS and MBC, both of which possess the enzymatic GEF activity for Rac1, specifically in adult glia, and examined the recruitment of glial membranes to the severed OR85e⁺ axons 1 d after injury (Fig. 5A). Interestingly, we found that sos^RNAi or mbc^RNAi resulted in a slightly reduced recruitment of Draper-decorated glial membranes to severed axons. However, when sos^RNAi and mbc^RNAi were simultaneously expressed, the recruitment of glial membrane to injured OR85e⁺ glomerulus was completely blocked 1 d after injury. The additive nature of the sos^RNAi and mbc^RNAi phenotypes also extended to clearance of axonal debris as well as glial hypertrophy response (Fig. S8). Thus, simultaneous blockade of DRK/DOS/SOS and Crk/Mbc/dCed-12 signaling phenocopied drpr-null mutants (5), and inhibition of Rac1 signaling (7).

Fig. 5.

DRK/DOS/SOS and Crk/Mbc/dCed-12 are redundant pathways in activating glial clearance of axonal debris. (A) Simultaneous knockdown of SOS and Mbc completely blocked glial membrane (red, α-Draper) recruitment to injured OR85e⁺ glomerulus (green, α-GFP) 1 d after injury, whereas individual RNAi expression showed mild inhibitory effects. (Scale bar: 30 μm.) (B) Quantification of data in A (n = 10 for all). (C) Sos^JC2 partially rescued the clearance defect cause by dCed-12^RNAi. The clearance assay was performed as described in the other figures, and the results were quantified in D (n = 10 for all).

We further explored if SOS activation could partially substitute for Crk/Mbc/dCed-12 during glial clearance of axonal debris by crossing Sos^JC2 into dCed-12^RNAi background. As previously shown (7), glial knockdown of dCed-12 function strongly suppressed glial clearance of axonal debris 5 d after injury (Fig. 5C). However, in a Sos^JC2/+ background, the effect of dCed-12^RNAi was attenuated (Fig. 5D), indicating that activation of SOS can partially compensate for the reduced activity from Crk/Mbc/dCed-12 to promote glial clearance of axonal debris. We conclude that Crk/Mbc/dCed-12 and DRK/DOS/SOS act in a partially redundant fashion downstream of Draper to activate Rac1 and thereby glia, and promote glial clearance of axonal debris.

Discussion

In this work, we identify Drosophila DRK, DOS, and SOS as new molecules required for glial responses to axonal injury. We show that glial depletion of DRK, DOS, or SOS results in a delay in glial responses to ORN axotomy and reduced efficiency of glial internalization and digestion of axonal debris. We observe no obvious alterations in glial morphology or expression of engulfment machinery (e.g., Draper), and demonstrate that adult-specific knockdown of DRK/DOS/SOS leads to defects in glial clearance of degenerating axons. These data indicate that DRK/DOS/SOS promote engulfment signaling in mature glia, and argues against a developmental defect causing the phenotypes we observe. Based on our observation that a dominant GOF allele of SOS (Sos^JC2) can partially suppress depletion of DRK and DOS, we propose that SOS acts genetically downstream of DRK and DOS.

Our previous work demonstrated a key role for the Drosophila GEF Crk/Mbc/dCed-12 in glial engulfment activity, with elimination of this signaling complex from glia resulting in normal glial activation (e.g., recruitment of glial membranes to axonal debris), but a failure to engulf axonal debris (7). Here we provide strong evidence that DRK/DOS/SOS and Crk/Mbc/dCed-12 act redundantly downstream of Draper at two key steps in the engulfment process (Fig. S9). First, based on the fact that simultaneous depletion of both signaling complexes phenocopies Rac1 loss of function, we propose that these complexes act redundantly to activate Rac1 and glial responses, including Draper up-regulation and extension of glial membranes to degenerating axons. Second, after glia have arrived at axonal debris, both complexes are required for the elimination of axonal debris. At this step, DRK/DOS/SOS and Crk/Mbc/dCed-12 appear to act in a nonredundant fashion to promote glial internalization of axonal debris and activation of the phagolysosomal program for degradation of internalized axonal material.

DRK/DOS/SOS signaling has been studied most intensively for its role downstream of the RTK Sev, where it acts to activate small GTPase Ras (8–13). However, consistent with our findings, in vitro and in vivo studies have also demonstrated a role for SOS in activating Rac1. In cell culture studies, SOS stimulates guanine nucleotide dissociation from Rac1 but not Cdc42 (21). In some cases, such as axon guidance in the Drosophila embryo, SOS action as a Rac1 GEF is independent of Ras activation (14), but, in other situations, SOS activation of Rac1 is coupled to stimulation of Ras (21). Based on our observation that glial expression of a DN Ras only very weakly suppresses clearance of degenerating axons, SOS activation of Rac1 during glial responses to axonal injury is likely largely independent of Ras activation.

The increase of DRK localization to glial membrane processes engulfing axonal debris, and the consequences of DRK depletion by RNAi all argue for an early role for DRK (and likely DOS and SOS) in engulfment events. Although DRK is expressed in neurons and glia (see Fig. 2), our glial-specific RNAi clearly demonstrates that DRK function is required in glia during engulfment. Western blot analysis of the brain lysates before and after axotomy suggests the increase of DRK in the glial membrane processes is likely a result of recruitment of DRK other than up-regulation of gene transcription/translation (Fig. S5). We also found that Draper is responsible for DRK localization to the site of injury, as DRK does not localize to severed nerves in drpr-null animals (Fig. S4), suggesting DRK/DOS/SOS activation requires Draper. It seems Draper activation or Draper-mediated recruitment of DRK is necessary for DRK/DOS/SOS to execute their functions, as we found Sos^JC2 failed to rescue the debris clearance defect in drpr-null animals (Fig. S10). Does DRK interact directly with Draper? Despite considerable efforts, we were unable to detect physical interactions between DRK and Draper. DRK may therefore interact very transiently or indirectly with Draper or DRK might associate with another yet unknown glial receptor that localizes to degenerating axonal materials.

The requirements we describe for DRK/DOS/SOS appear to be conserved in mammals in the context of phagocytic activity. Mammalian DRK (growth factor receptor-bound protein 2) is accumulated at the phagocytic cup during leukocyte phagocytosis (27). Mammalian DOS (Grb2-associated binding protein 2) also plays a critical role during Fcγ receptor-mediated phagocytosis (28). Recently, MEGF10 (mouse Draper) is expressed in mammalian astrocytes and essential for synaptic pruning during refinement of dorsal lateral geniculate nucleus connectivity (29). The work we present here also argues that DRK/DOS/SOS and Crk/Mbc/dCed-12 are exciting candidates downstream of MEGF10 to promote synaptic pruning.

Materials and Methods

Olfactory axon injury, adult brain dissection, sample preparation, and image analysis were previously described in ref. 5. Lysotracker staining in Drosophila adult brain was previously described in ref. 7. Fly strains and antibodies used in this study are provided in SI Materials and Methods.