Caspase Activity Is Required for Engulfment of Apoptotic Cells

Abstract

Clearance of apoptotic cells by phagocytic neighbors is crucial for normal development of multicellular organisms. However, how phagocytes discriminate between healthy and dying cells remains poorly understood. We focus on glial phagocytosis of apoptotic neurons during development of the Drosophila central nervous system. We identified phosphatidylserine (PS) as a ligand on apoptotic cells for the phagocytic receptor Six Microns Under (SIMU) and report that PS alone is not sufficient for engulfment. Our data reveal that, additionally to PS exposure, caspase activity is required for clearance of apoptotic cells by phagocytes. Here we demonstrate that SIMU recognizes and binds PS on apoptotic cells through its N-terminal EMILIN (EMI), Nimrod 1 (NIM1), and NIM2 repeats, whereas the C-terminal NIM3 and NIM4 repeats control SIMU affinity to PS. Based on the structure-function analysis of SIMU, we discovered a novel mechanism of internal inhibition responsible for differential affinities of SIMU to its ligand which might prevent elimination of living cells exposing PS on their surfaces.

INTRODUCTION

The proper elimination of unwanted or aberrant cells through apoptosis is crucial for normal development of multicellular organisms. The final important step of apoptosis is clearance of apoptotic cells by phagocytes. This is a complex and dynamic process, involving recruitment of phagocytes to the apoptotic cell by secreted “find me” signals, recognition of the cell as a target for phagocytosis through “eat me” signals on the apoptotic surface, engulfment, and finally, degradation of the apoptotic particles inside the phagosome (1–6). There are two types of phagocytes: the “professional” macrophages and immature dendritic cells and the “nonprofessional” tissue-resident neighboring cells, which are essential for apoptotic cell clearance during development (7–9). How phagocytes discriminate between healthy and dying cells is still unclear. This discrimination must be highly specific and reliable at the molecular level. Improper recognition can lead to removal of functional cells or survival of unwanted cells, leading to morphogenetic defects and pathological situations.

In mammals, a large number of transmembrane receptors and soluble bridging molecules have been shown to play a role in recognition and engulfment of apoptotic particles (10–14). Many of these molecules recognize phosphatidylserine (PS) on the outer leaflet of apoptotic cells, raising a question of specificity and competition in ligand-receptor interactions. A mechanism of “tethering and tickling” was proposed a number of years ago (7) which suggests that binding of multiple phagocytic receptors to their ligands, leading to receptor clustering, is needed for engulfment of the apoptotic cell. The high redundancy of factors acting in the mammalian systems makes it difficult to uncover the molecular and cellular mechanisms of the recognition process in vivo.

Drosophila has several receptors for apoptotic cells acting in distinct phagocytic cell populations: the CD36 homolog Croquemort (CRQ) functions specifically in professional phagocytes, the macrophages (15), whereas Draper (DRPR) and Six Microns Under (SIMU) act in macrophages and also in glia during phagocytosis of apoptotic neurons in the central nervous system (CNS) (8, 16). DRPR and SIMU belong to the recently identified Nimrod (NIM) superfamily containing phagocytic receptors (17, 18). Several members of this family have been implicated in innate immunity (bacterial phagocytosis and apoptotic cell clearance) (8, 16, 17, 19–21) where accurate recognition of the correct targets is crucial for efficient defense and normal development. The ligands of most of these receptors are unknown, and elucidation of their mechanism of action remains elusive.

The apoptotic cell death program is executed in all organisms by a specific group of cysteine proteases called caspases (22, 23). Initiator caspases start a tightly controlled proteolytic cascade to activate themselves and effector caspases, whose activation leads to apoptosis by cleavage of multiple cellular substrates (24, 25). Caspase activation in the fly embryo could be completely abrogated by deletion of a genomic region containing three proapoptotic genes, reaper, grim, and hid (H99) (26).

Here we show that during Drosophila embryonic CNS development, caspase-independent PS exposure on apoptotic neurons is not sufficient for engulfment, and additional caspase-dependent signals are required. We identified PS as a ligand on apoptotic cells for the phagocytic receptor SIMU. Our data demonstrate that SIMU recognizes and binds PS on apoptotic cells through its N-terminal EMI, NIM1, and NIM2 repeats, whereas SIMU C-terminal NIM3 and NIM4 repeats are required for proper protein localization and control SIMU affinity to PS. By dissecting the SIMU mode of action, we propose a novel internal inhibitory mechanism responsible for differential affinity of a phagocytic receptor to its ligand which is likely involved in precise uptake of appropriate targets.

MATERIALS AND METHODS

Fly strains and constructs.

The following fly strains were obtained from published sources: the repoGal4 (V. Auld), UASCD8GFP (L. Luo), Df(3L)H99 (H. Steller), elavGal4 (O. Shuldiner), UASnGFP (catalog no. 4775; Bloomington), and UAS-pro-dronc, UAS-ark and UASp35/TM6B (E. Arama) strains. Inducible full-length and truncated constructs of simu were generated by cloning of the cDNA of simu in full length or with introduced deletions into the pUASTattB vector. They also contain a C-terminal green fluorescent protein (GFP) tag. These transgenes were inserted into the attP86 site on chromosome 3R using the QC31 system (27). All strains were raised at 25°C. w1118 flies were used as a wild-type control.

SIMU protein purification and binding to apoptotic cells.

The coding sequence of the extracellular portion of the SIMU wild type, or of each of the truncated constructs, was cloned into a His-Myc tag-containing vector, pCEP-Pu (28) and stably transfected into HEK293 cells, and the tagged protein was affinity purified from the conditioned medium using Talon metal affinity beads (Clontech) according to the manufacturer's specifications. To induce apoptosis, S2 cells were treated with 75 μM Etoposide (Sigma) for 16 h. To evaluate protein binding to S2 cells, 6 μg of purified SIMU proteins was incubated with live or apoptotic S2 cells (in phosphate-buffered saline [PBS] plus 1% normal horse serum [NHS] for 1.5 h). Cells were then incubated with anti-Myc antibody (Santa Cruz Biotechnology) (1:100; 1 h) and fluorescent secondary antibody (Alexa Fluor 488; Molecular Probes) (1:400; 40 min) and analyzed by flow cytometry. For each experiment, 10,000 events were collected, and the data were analyzed using Cell Quest and FlowJo software; experiments were performed in triplicate and repeated three times, with similar results. In blocking experiments, purified recombinant annexin V (BD Biosciences) was used as a phosphatidylserine (PS) masking agent. Cells were incubated with annexin V for 15 min before adding the SIMU proteins and processing as described above. In the parallel experiments, Alexa Fluor 647-conjugated annexin V (Molecular Probes) was used as indicated by the manufacturer, following incubation of the cells with purified SIMU protein.

Phosphatidylserine binding assay.

SIMU direct binding to PS was tested by a solid-phase binding assay for phospholipids (10). Polypropylene plates were coated with PS, phosphatidylcholine (PC), or phosphatidylethanolamine (PE) (Avanti Polar Lipids) at 3 μg/ml and 100 μl/well. Purified SIMU or the truncated proteins were added, and bound proteins were detected with an anti-Myc antibody and a secondary antibody conjugated to horseradish peroxidase (HRP).

Immunohistochemistry and imaging.

Guinea pig anti-SIMU antibody was raised against SIMU wild-type protein purified from conditioned medium of SIMU-transfected HEK293 cells as described in reference 8. All embryos were stained by standard procedures using rabbit anti-activated caspase 3 (CM1/Abcam) (1:200). Mouse anti-GFP (1:100) is from Roche. Fluorescent secondary antibodies (Cy3 [Jackson ImmunoResearch] and Alexa Fluor 488 [Molecular Probes]) were used in 1:200 dilutions. All confocal images were acquired on a Zeiss LSM 700 confocal microscope using an EC Plan-Neofluar 40×/1.30 oil differential interference contrast (DIC) M27 objective and a Plan-Apochromat 20×/0.8 M27 lens. Glycerol solution (75%) was used as the imaging medium. Image analysis was performed using Zeiss LSM 700 and Imaris (Bitplane) software. To quantitate the volume of apoptotic particles, confocal stacks (5 sections; total, 7.5 μm) were acquired from the neural cortex of stage 16 ventral nerve cords.

Live imaging was carried out by dechorionating embryos (stage 15), mounting them under halocarbon oil, and injecting 2% to 3% egg volume Alexa Fluor 555-conjugated annexin V (Molecular Probes) or LysoTracker (Molecular Probes). Recording started 30 min following injection.

RESULTS

PS is a ligand of SIMU on apoptotic cells.

In order to understand the mechanism underlying recognition of apoptotic particles by phagocytic receptors, we focused on the phagocytic receptor SIMU, which is required for recognition and engulfment of apoptotic particles in vivo and binds specifically to apoptotic cells in vitro (8). To identify the SIMU ligand(s) on the apoptotic surface, we took a structure-function approach. SIMU comprises a large extracellular portion, a single transmembrane domain (TM), and a short cytoplasmic tail at the C terminus with no signaling ability (8) (Fig. 1A). The extracellular portion of the protein consists of an N-terminal EMILIN (EMI) domain and four EGF domains named Nimrod (NIM) repeats (8). To establish which domain(s) is involved in ligand recognition, we tested the ability of truncated forms of the SIMU protein (Fig. 1A) to bind to apoptotic Drosophila Schneider (S2) cells. We have previously shown that the secreted form of SIMU (SIMUΔTM) strongly binds to apoptotic S2 cells in an in vitro assay based on flow cytometry (8). Here, we show that deletion of each of the EMI, NIM1, and NIM2 domains or of the three domains together results in no binding of SIMU to apoptotic cells (Fig. 1B). Therefore, we reasoned that EMI, NIM1, and NIM2 are required for SIMU binding to apoptotic cells. In contrast, and to our surprise, a lack of the NIM3 domain or of the NIM4 domain or of their combination not only does not reduce but rather significantly increases SIMU binding to apoptotic cells (Fig. 1C), suggesting an inhibitory effect of these domains on EMI, NIM1, and NIM2 binding and no involvement in ligand binding.

Fig 1.

SIMU binding to apoptotic cells. (A) Schematic of SIMU constructs designed for expression in vitro in HEK293 cells, fused to Myc and HIS tags, or fused in vivo to GFP tags. (B and C) A representative fluorescence-activated cell sorter (FACS) experiment (left) and the quantitation graph (right) showing altered binding of truncated SIMU proteins to apoptotic S2 cells compared to SIMUΔTM. The bound protein was detected by an anti-Myc antibody followed by a fluorescently labeled secondary antibody. Note the reduced binding of truncated proteins in panel B and increased binding in panel C. The assays were performed in triplicate. Columns represent mean fluorescence values ± standard errors of the means (SEM).

NIM domains are characterized by a consensus sequence (Fig. S1A); however, small differences exist between NIM domains in SIMU. To test why NIM2 and NIM3 behave so differently and to understand whether the sequence or the position of NIM domains determines their function, we performed swapping of NIM2 and NIM3 domains. We generated SIMU full-length secreted proteins containing two NIM2 domains (an inhibitory NIM3 domain was replaced by the NIM2 domain) or two NIM3 domains (the NIM2 domain was replaced by the inhibitory NIM3 domain) (see Fig. S1A and B in the supplemental material). Interestingly, the protein with two NIM3 domains behaved similarly to the SIMUΔTM protein, indicating that the NIM3 domain in the position of the NIM2 domain functions similarly to NIM2 in binding to PS and does not inhibit SIMU binding to apoptotic cells (see Fig. S1C). However, when we tested the SIMU protein containing two NIM2 domains, we observed much stronger binding of the protein to S2 cells, indicating that substitution of the NIM3 domain with NIM2 resulted in a reduction of internal inhibition (see Fig. S1C). We obtained the same results in the enzyme-linked immunosorbent assay (ELISA), which we performed with the same swapped proteins (see Fig. S1D), suggesting that the inhibitory function of the NIM3 repeat depends on its position.

PS represents a major recognition cue for engulfment by professional and nonprofessional phagocytes (29–31). A number of phagocytic transmembrane receptors and secreted bridging molecules have been shown to bind PS (10–12, 29, 32). Moreover, PS exposure on apoptotic cells is required for phagocytic uptake and masking PS on apoptotic surfaces inhibits engulfment. Annexin V specifically binds exposed PS on apoptotic cells, and it is broadly used as a marker for apoptosis and as a PS masking agent (10, 33). In order to examine the possibility that PS is a ligand for SIMU, we tested SIMU binding to apoptotic S2 cells in the presence of annexin V. Binding of the ΔNIM3-4 mutant with strongest binding affinity to apoptotic S2 cells was significantly reduced by the presence of annexin V compared to the strong binding of this form with no annexin V addition (Fig. 2A), indicating that the EMI, NIM1, and NIM2 domains most likely bind PS on apoptotic cells. Moreover, in the corresponding experiment, significantly reduced binding of fluorescently labeled annexin V to apoptotic S2 cells was detected in the presence of SIMU in its full-length compared to the annexin V binding with no SIMU addition (Fig. 2B). We therefore tested whether SIMU directly interacts with PS by using a solid-phase binding assay for phospholipids (10). Polypropylene plates were coated with PS, phosphatidylcholine (PC), or phosphatidylethanolamine (PE) and incubated with purified SIMU in increasing concentrations. According to our results, SIMU binds specifically to PS in a dose-dependent manner (Fig. 2C). In addition, binding of SIMU truncated constructs to PS-coated plates (Fig. 2D) correlated with their binding to apoptotic cells (Fig. 1B and C), further supporting the notion that PS is an authentic SIMU ligand on apoptotic cells.

Fig 2.

SIMU binding to PS. (A) FACS experiment showing binding of ΔNIM3-4ΔTM-truncated SIMU protein to apoptotic S2 cells in the absence or presence of unlabeled annexin V as a blocking agent in a ratio of 2:1 (annexin V/SIMU). (B) FACS experiment showing binding of Alexa Fluor 647-conjugated annexin V to apoptotic S2 cells in the absence or presence of full-length SIMUΔTM as a competitor at a ratio of 2:1 (SIMU/annexin V). (C) SIMUΔTM specifically binds to PS compared to PE or PC as shown by ELISA. PS, PE, or PC is immobilized on ELISA plates, and the bound proteins are detected by the anti-Myc antibody followed by an HRP-conjugated secondary antibody. (D) Differential binding of purified truncated SIMU proteins to PS, which reflects binding of truncated proteins to apoptotic cells. The assays were performed in triplicate, and the average values are plotted with SEM.

EMI, NIM1, and NIM2 are required for SIMU function in vivo.

Our data show that EMI, NIM1, and NIM2 are required for SIMU binding to PS on apoptotic cells in vitro. To explore the function of these domains in vivo during embryonic CNS development, we tested the ability of the different truncated forms of SIMU (Fig. 1A) to rescue the phenotype of simu null mutants when expressed specifically in glial cells under the regulation of the repoGal4 driver. To examine the apoptotic cell clearance phenotype, we measured apoptotic cell volume, using an anti-activated caspase 3 (CM1) antibody, in the CNS of stage 16 embryos (8) (Fig. 3B to O). As we have previously shown, in simu mutant embryos there is an accumulation of unengulfed apoptotic particles and therefore an increase in apoptotic cell volume due to defects in the recognition and engulfment steps of phagocytosis (8) (Fig. 3E, F, and P). In concordance with the in vitro results, the transmembrane ΔEMI-NIM2 variant, which lacks all three domains required for binding to PS, did not rescue the simu mutant phenotype (Fig. 3K, L, and P), implying that EMI, NIM1, and NIM2 are required in vivo for simu function in apoptotic cell clearance. To examine each domain's role in apoptotic cell clearance, we tested secreted ΔEMI, ΔNIM1, and ΔNIM2 truncated proteins for rescue of the simu mutant phenotype. In contrast to efficient rescue by the full-length secreted SIMU (Fig. 3P), we found that none of these mutant constructs could rescue the mutant phenotype, presumably due to an interruption in binding to apoptotic cells. These results suggest that the EMI, NIM1, and NIM2 domains of SIMU are involved in PS binding during engulfment.

Fig 3.

Phenotypic analysis of the truncated simu constructs in vivo. (A to O) Projections from confocal stacks of the CNS at embryonic stage 16, ventral view; apoptotic cells are indicated in red (CM1) and glia and macrophages in green (anti-SIMU antibody). Bar, 20 μm. (A to C) In the wild-type (wt) embryo, glia (g) and macrophages (m) are labeled with anti-SIMU in green (endogenous protein). (D to F) In the simu null mutant, no SIMU expression was detected. (G to O) In rescue experiments, SIMU expression was detected only in glial cells (repoGal4 driver). (G to I) Complete rescue of the simu null phenotype with full-length SIMU. (J to O) No rescue with truncated SIMU proteins expressed in glial cells. (P) Quantification of phenotypic rescue of simu null mutants by the different simu transgenes. Columns represent mean total volume of apoptotic particles within confocal stacks of the CNS ± SEM (n = 7 to 10); asterisks indicate statistical significance versus wild-type results, as determined by one-way analysis of variance (ANOVA). ***, P < 0.001; **, P < 0.01; n.s. (not significant), P > 0.05.

Caspase-independent PS exposure is not sufficient for phagocytic uptake in the embryonic CNS.

To monitor PS exposure on apoptotic cells in the embryonic CNS, small amounts of fluorescent annexin V, which binds specifically to PS, were injected into live embryos (8). In the wild-type embryos, most of the labeled particles were found inside the glia (Fig. 4G), indicating that phagocytosis was largely not affected by the presence of annexin V. When monitoring the labeled particles by time lapse recordings, we observed that some of the particles were never engulfed, suggesting that PS exposure alone is not sufficient for phagocytic uptake (Fig. 4G; see also Movie S2 in the supplemental material).

Fig 4.

Dynamic analysis of PS exposure in the embryonic CNS. (A to F) Nuclear GFP driven with elavGal4 labels neuronal nuclei of wild-type (A and C) and H99 (D and F) embryonic CNS. Bar, 20 μm. (B, C, E, and F) PS on neuronal membranes is marked by annexin V in red. Arrows depict selected neurons with annexin V on their membranes. (G to I) Time-lapse recordings of phagocytosis in stage 16 embryos. Glia and macrophages are labeled with simu-cytGFP (green), and PS exposure is labeled by the fluorescent annexin V (red); selected frames are shown (movies are available in the supplemental material). (G) In the wild type, most of the PS-positive particles are inside the glial cells. (H) An engulfment event is marked. Some annexin V-positive particles remained unengulfed for long periods of time (arrows). (I) H99-deficient embryonic CNS with annexin V-positive particles remaining outside the glia and not engulfed.

Interestingly, when injecting annexin V into H99 embryos, which lack caspase activation, PS is still detected on some presumably living cells inside the CNS, which are not engulfed by glia (Fig. 4I; see also Movie S3 in the supplemental material). When examining the CNS in the H99-deficient embryos, we observed annexin V labeling, which surrounds the neuronal nuclei marked with elavGal4::nuclearGFP (Fig. 4D to F). This is in contrast with the annexin V signal which is concentrated predominantly on the apoptotic particles in the wild-type embryos (Fig. 4A to C). When performing time lapse recordings of the H99 embryos, we noticed that annexin V-positive cells remained unengulfed for long periods of time (see Movie S3). This suggests that caspase-independent PS exposure is not sufficient for phagocytic uptake, as has been previously shown in tissue culture experiments (31).

Since H99 deletion may result in abnormal regulation of other proteins apart from caspases, we tested the direct role of caspases in engulfment by two different approaches, including expression of activated caspases in living cells and blocking caspase activity directly by a viral P35 protein. In Drosophila, the initiator caspase-9 ortholog Dronc and the effector caspase-3 ortholog Drice are the major caspases involved in developmental apoptosis (34–37). We coexpressed in living neurons (with the elavGal4 driver) the initiator caspase Dronc together with the Drosophila Apaf-1 homolog, Ark, which have been shown to induce effector caspase activation (38). To evaluate the level of engulfment in these embryos, we used LysoTracker staining of live embryos. In wild-type embryos, LysoTracker labels almost exclusively large vesicles inside glial cells, indicating phagolysosomes (Fig. 5D to D″), whereas in H99 embryos, only small vesicles (lysosomes) are detected everywhere, suggesting that no engulfment takes place in the CNS (Fig. 5E to E″). When we compared the volume of LysoTracker-positive phagosomes in the wild type (Fig. 5D to D″ and I) with that in embryos overexpressing activated caspases in living cells (Fig. 5F to F″ and I), we detected a significantly higher volume of a LysoTracker-positive area in the elavGal4::dronc, ark embryos than in the wild type (Fig. 5I), suggesting that caspase activation in living cells leads to their engulfment.

Fig 5.

Caspase activity affects SIMU localization and phagocytosis of apoptotic cells. (A to C) Projections from confocal stacks of the CNS at embryonic stage 16. A ventral view with apoptotic particles labeled with CM1 (red) and glial cells labeled with anti-SIMU (green) is shown. Bar, 20 μm. (A to A″) Wild type. SIMU is localized in patches on glial membranes (arrows). (B to B″) H99 mutant embryo with no CM1 staining (B′) and SIMU is localized homogeneously on glial membranes (B and B″). (C to C″ and H) elavGal4::p35 embryos with much lower CM1 staining (C′, C″, H). (D to F) Projections from confocal stacks of the CNS at embryonic stage 16. A ventral view with phagosomes and lysosomes labeled with LysoTracker (red) and glial cells and macrophages labeled with simu-cytGFP (green) is shown. (D to D″) Wild type. (E to E″) H99. (F to F″) elavGal4::dronc, ark. (G to G″) elavGal4::p35. (H and I) Columns represent mean total volume of apoptotic particles (H) or LysoTracker-positive area (I) within confocal stacks of the CNS ± SEM (n = 7 to 10); asterisks indicate statistical significance versus wild-type results, as determined by one-way ANOVA. ***, P < 0.001; **, P < 0.01.

Our second approach was to inhibit activated caspases directly by using the viral protein P35, which specifically blocks effector caspase activity (39). We generated embryos expressing the UASp35 transgene in neurons by using the elavGal4 driver and evaluated the level of engulfment by comparing the volume of LysoTracker-positive phagosomes in wild-type embryos to that in the elavGal4::p35 embryos. Our results show that, consistent with the reduced volume of the CM1-positive particles in the elavGal4::p35 embryos (Fig. 5C to C″ and H), there was a lower volume of LysoTracker-positive area compared to the volume in the wild type (Fig. 5G to G″ and I), indicating that inhibition of effector caspases leads to a decrease in engulfment. Overall, these results suggest that caspase activity is required for engulfment of apoptotic cells. Therefore, we propose that in addition to PS exposure on apoptotic surfaces, other ligands which are caspase dependent are required for apoptotic cell clearance.

SIMU is localized in patches on phagocytic membranes during engulfment.

To follow SIMU-PS interactions during clearance of apoptotic neurons, we examined SIMU expression and distribution on glial membranes using an anti-SIMU antibody. We noticed that SIMU is not homogeneously distributed on phagocytic glial membranes. The patchy distribution of SIMU suggests clustering of the protein (Fig. 5A and A″). To test whether this specific pattern correlates with engulfment, we explored SIMU localization in H99-deficient embryos, in which no phagocytic activity of glia can be detected, using LysoTracker staining of live embryos (Fig. 5E′ and E″). When we examined SIMU expression in these embryos, we found homogeneous rather than patchy distribution of the protein on the phagocytic membranes (Fig. 5B and B″). This result suggests that SIMU forms clusters only during engulfment, possibly by interacting with its putative partner(s) on the phagocytic membranes, consistent with a “tethering and tickling” multiple-receptor clustering model (7). In H99 mutant embryos, inhibition of initiator and effector caspases takes place, as confirmed by a lack of staining with CM1 antibody (Fig. 5B and B″).

To address the important issue of whether ligand binding is needed for receptor clustering, we examined SIMU localization in embryos expressing the ΔEMI-NIM2 SIMU form, which does not bind to PS. In these embryos, no SIMU patches were detected on glial membranes, suggesting that initial binding of SIMU to PS is required for its clustering with other factors (Fig. 3J).

NIM3 and NIM4 are required for proper localization of SIMU on phagocytic membranes.

Based on our in vitro results, NIM3 and NIM4 domains are not required for SIMU binding to apoptotic cells and, moreover, their removal intriguingly increases the affinity of SIMU to apoptotic cells (Fig. 1C). To determine the role of these domains in vivo,we performed rescue experiments in which we tested the ability of the transmembrane SIMU variant lacking the NIM3 and NIM4 portion (ΔNIM3-4) (Fig. 1A) to rescue simu null embryos. Given the increased affinity to apoptotic cells, we expected faster or more efficient phagocytosis in the CNS of the simu; repoGAL4::ΔNIM3-4 embryos. However, the ΔNIM3-4 form did not rescue the phagocytosis phenotype of simu mutant embryos (Fig. 3M, O, and P), suggesting that NIM3 and NIM4 are required for SIMU function in vivo. Surprisingly, when we examined the distribution of the ΔNIM3-4 transmembrane protein in the CNS, we observed abnormal localization of the protein on phagocytic membranes. Wild-type endogenous SIMU protein was found on phagocytic protrusions (Fig. 6A, C, D, and F) where it is assumed to act as a recognition receptor for apoptotic particles. However, the ΔNIM3-4 mutant form was mostly absent from the phagocytic protrusions (Fig. 6B, C, E, and F) preventing contact between apoptotic particles and the protein. To test whether glial protrusions themselves are affected by the ΔNIM3-4 form, we examined glial membranes using the anti-SIMU antibody, which labels the transgenic and the endogenous forms of SIMU whereas the anti-GFP antibody detects only the transgenic form (Fig. 6). The glial membranes exhibited normal morphology when labeled with anti-SIMU in the simu wt; repoGAL4::ΔNIM3-4 embryos (Fig. 6A, C, D, and F). Therefore, the inability of the ΔNIM3-4 form to rescue the simu mutant phenotype may result from the improper localization of the mutant protein.

Fig 6.

Abnormal localization of the SIMU protein lacking NIM3 and NIM4 domains. (A to C) Projections from confocal stacks of a stage 16 embryo. A ventral view is shown. Bar, 20 μm. (D to F) Closeup of the embryonic CNS and specific glial cells (inserts). Expression pattern of repoGal4::simuΔNIM3-4GFP showing overlapping between endogenous SIMU (anti-SIMU antibody; red) and transgenic SIMUΔNIM3-4GFP (anti-GFP; green) in glial cell bodies and no overlap on the glial protrusions (arrows). (G to I) COS7 cells transfected with truncated simu constructs. SIMU is labeled with anti-SIMU (red) and DAPI (blue). (G) Full-length SIMU is localized to cell membranes. (I and H) SIMUΔNIM3-4 (I) and SIMUΔEMI-NIM2 (H) are localized to cell membranes as the full-length protein.

NIM3 and NIM4 are required for SIMU function in vivo.

To overcome the localization defect of the ΔNIM3-4-truncated form, we generated a secreted ΔNIM3-4 variant (ΔNIM3-4ΔTM) (Fig. 1A), which should act independently of protein localization on the phagocytic membrane. Given that secreted SIMU completely rescues the simu mutant phenotype (8) (Fig. 3P), we assumed that the results would reveal the role of NIM3 and NIM4 in engulfment. The ΔNIM3-4ΔTM mutant form did not rescue the simu mutant phenotype (Fig. 3P), suggesting that NIM3 and NIM4 are required for SIMU function in vivo. Moreover, we tested the ability of the individual NIM3 and NIM4 domains to rescue simu mutants. In embryos expressing ΔNIM3ΔTM or ΔNIM4ΔTM proteins, there was no rescue of the mutant phenotype (Fig. 3P), indicating that the NIM3 and NIM4 domains are both important for the in vivo function of SIMU.

In order to explore whether or not the abnormal localization of the ΔNIM3-4 truncated form in vivo is the result of an intrinsic improper protein folding, we transiently transfected COS7 cells with the full-length, ΔEMI-NIM2, and ΔNIM3-4 SIMU forms and followed protein localization using the anti-SIMU antibody. We observed the same level of expression and the same membrane localization of all the forms (Fig. 6G to I′), suggesting that the abnormal distribution of ΔNIM3-4 form in vivo may result from the impaired interaction with additional factors on the plasma membrane.

DISCUSSION

In this study, we have shown that caspase activity is needed for efficient and precise phagocytosis of apoptotic particles during embryonic CNS development. Focusing on the recognition receptor SIMU, we have demonstrated that caspase-independent PS exposure is not sufficient for clearance. Since there are no known “don't eat me” signals in Drosophila, this mechanism may explain why living cells exposing (transiently or in small amounts) some “eat me” signals on their surfaces are not removed until they are actually dying.

PS is the most abundant “eat me” signal (40, 41) and is recognized directly by different receptors in mammals (29, 42), including the TIM family (12, 43, 44), BAI1 (45), and SIMU homolog stabilin-2 (13), and by DRPR in Drosophila (46). Here we demonstrate that the fly phagocytic receptor SIMU recognizes PS on apoptotic cell surfaces, illustrating once more the high level of evolutionary conservation of apoptotic cell clearance between insects and vertebrates. Moreover, we show that SIMU specifically binds to PS on apoptotic cells in vitro through its EMI, NIM1, and NIM2 domains and that removal of these domains affects SIMU function in vivo and its distribution on glial membranes. In the embryonic CNS, SIMU protein is localized in patches on glial phagocytic membranes, suggesting clustering of SIMU with itself or other phagocytic receptors. This distribution depends on the initial binding of SIMU to PS, since the truncated ΔEMI-NIM2 SIMU does not make patches on the membranes.

Receptor clustering is thought to be important for proper and efficient removal of apoptotic particles (29). The “tethering and tickling” model (7) speaks about multiple interactions between phagocytic receptors and their ligands, which result in receptor clustering and signaling for engulfment. Elucidation of the molecular mechanisms underlying receptor clustering remains elusive, and it is unknown what the interactions are between different phagocytic receptors and whether they are influenced by ligand binding and conformation changes. In H99 embryos, lacking caspase activation, where no engulfment takes place, SIMU is equally distributed on glial membranes, indicating that SIMU clusters with itself or other receptors only during engulfment. Based on our structure-function analysis, we show that SIMU clustering during engulfment depends on its binding to PS, suggesting that clustering of phagocytic receptors depends on their binding to the ligands. What are these ligands on apoptotic cells, which are absent in H99 mutants?

Caspase involvement in the expression and exposure of “find me” and “eat me” signals is not completely understood. Particularly, PS exposure on apoptotic cells is considered caspase dependent in worms and mammals and is required for engulfment of apoptotic particles. Moreover, recent discoveries from S. Nagata's laboratory showed that PS exposure on living cells is not sufficient for their phagocytic removal in vitro (31). However, the study by van Delft and colleagues in Casp9^−/− mice showed that intact slowly dying Casp9^−/− mutant thymocytes exposed PS without caspases, which led to their noninflammatory phagocytosis (47). To test how caspase activity affects PS exposure and apoptotic cell clearance in Drosophila, we examined H99 mutant embryos, which lack caspase activation. In these embryos, we detected PS accumulation on some presumably living cells. Following these cells by time lapse recording, we observed that they were never engulfed, indicating that PS accumulation on their surface is not sufficient for clearance. Moreover, in H99 embryos, SIMU is expressed on phagocytic membranes and its ligand PS is detected on some cell surfaces but no engulfment takes place, implying that something is missing. We suggest that additional caspase-dependent ligands on apoptotic particles are missing which are recognized by their receptors on phagocytic membranes and that, akin to the “tethering and tickling” model, they participate in receptor clustering leading to engulfment. Finding these additional caspase-dependent ligands will help us uncover the molecular basis of multiple receptor/ligand interactions.

Interestingly, based on the structure-function analysis of SIMU, we discovered a novel internal inhibitory mechanism in the SIMU receptor which presumably prevents engulfment of living cells expressing PS on their surface. We found that the SIMU NIM3 and NIM4 domains are not involved in PS recognition and binding but rather inhibit SIMU binding to PS. Furthermore, these domains are required for SIMU proper localization and function in vivo in apoptotic cell clearance, suggesting their role in engulfment. We propose that they may be involved in interactions of SIMU with other phagocytic receptors or bridging molecules, which recognize additional caspase-dependent and/or -independent alterations on apoptotic surfaces. These interactions through the NIM3 and NIM4 domains may change SIMU conformation, relieve the internal inhibition, and dramatically increase SIMU affinity to PS.

Taken together, our data suggest the following model for SIMU action in apoptotic cell clearance (Fig. 7). SIMU recognizes PS on apoptotic cells with mild affinity that is crucial for not engulfing living cells, which have been shown to expose (transiently or in small amounts) PS on their surfaces (48, 49). When additional “eat me” signals are exposed on apoptotic cells, indicating that the cells are actually dying, and are recognized by additional receptors (putative SIMU partners), SIMU affinity to PS dramatically increases. We propose that this increase in affinity occurs as a result of relieving the internal inhibition by NIM3 and NIM4 through binding of the partner(s) to SIMU and changing its conformation.

Fig 7.

PS and caspase activation are required for engulfment. A schematic representation is shown of PS and additional caspase-dependent ligands on apoptotic cells that are recognized by different receptors on the phagocytic surface, which cluster and thus lead to engulfment of the apoptotic cell by the phagocyte.

Observing the phagocytic behavior of glia in the wild-type embryonic CNS, we found that glial protrusions probe the environment before engulfing an apoptotic cell (see Movie S4 in the supplemental material). Instances of phagocytes reaching the apoptotic cell without uptake suggest a threshold of receptor affinity which has to be reached and/or involvement of additional factors that are needed to be active for proper engulfment. This model suggests a mechanism for controlling the removal of only dying and not healthy cells, which is crucial for normal development and homeostasis of multicellular organisms. Finding the additional caspase-dependent ligands and understanding how apoptotic cells are marked for accurate and efficient elimination will provide new insight into the molecular and cellular mechanisms of apoptotic cell clearance.