Three sorting nexins drive the degradation of apoptotic cells in response to PtdIns(3)P signaling

Identification of four PX domain proteins that actively promote the removal of apoptotic cells

Using the Simple Modular Architecture Research Tool (SMART) program (Schultz et al., 2000), we identified genes encoding 16 FYVE domain-containing and 12 PX domain-containing proteins from the C. elegans genome (Table 1). To identify PtdIns(3)P effectors that specifically function in the removal of apoptotic cells, we analyzed the loss-of-function phenotypes of each of the 28 genes by scoring the number of persistent germ cell corpses, a readout of the cell-corpse removal defect (Ced phenotype), in the gonad of adult hermaphrodites bearing homozygous gene deletions (Table 1). When deletion alleles were not available or caused lethality, we scored animals in which the target genes were individually inactivated by RNA interference (RNAi) treatment (Materials and Methods) (Table 1). Inactivation of genes encoding FYVE domain-containing proteins did not result in an obvious Ced phenotype, except for a mild increase in the number of germ cell corpses observed in deletion mutants of C26H9A.2 (Table 1). In contrast, a large number of persistent germ cell corpses were observed from deletion mutants of F39B1.1, lst-4, snx-1, or snx-6, which encode PX domain-bearing proteins (Table 1; Figures 1B and S1, A–D). Moreover, each of these deletions also resulted in an excessive amount of somatic cell corpses in homozygous embryos throughout the mid- and late embryonic stages (Figure 1, C and D) (N. Lu and Z. Zhou, unpublished data). These results indicate that these four genes are required for the efficient removal of both somatic and germ cell corpses in C. elegans.

LST-4, SNX-1, and SNX-6 are members of the SNX-BAR subfamily of sorting nexins

C. elegans LST-4 (Lateral Signaling Target-4), SNX-1 (Sorting Nexin-1), and SNX-6 all display strong homology to proteins in the sorting nexin family, defined by the presence of an SNX-PX domain (Teasdale et al., 2001). LST-4, SNX-1, and SNX-6 also each have an additional BAR domain at their C termini and are further classified into the SNX-BAR subfamily of sorting nexins (Figure 1A) (van Weering et al., 2010). The BAR domain is a coiled-coil structure that homodimerizes and forms a rigidly curved surface that preferentially binds membranes with a certain degree of curvatures (Frost et al., 2009). Based on the domain structures and sequence homology, SNX-1 resembles mammalian SNX1 and SNX2, whereas SNX-6 resembles mammalian SNX5 and SNX6 (Figure 1A). LST-4, which possesses an additional SH3 domain at its N terminus, resembles mammalian SNX9, SNX18, and SNX33 (Figure 1A). Sorting nexins have diverse functions in endosomal trafficking events (Seet and Hong, 2006; van Weering et al., 2010). The PX and BAR domains together allow the SNX-BAR subfamily of sorting nexins to associate with membrane subdomains of high curvature and enriched for specific phosphoinositides, further stabilizing the local membrane curvature (McMahon and Gallop, 2005). We hypothesize that the three SNX-BAR proteins that we identified might act as effectors of PtdIns(3)P to assist the membrane fusion events involving phagosomes, the membranes of which display relatively low curvature (Yu et al., 2006). The characterization of F39B1.1, which encodes a Class II PI-3 kinase that we found to produce PtdIns(3)P on phagosomal membranes, will be reported elsewhere (N. Lu and Z. Zhou, unpublished data).

lst-4, snx-1, and snx-6 act in two parallel and partially redundant genetic pathways to drive the removal of apoptotic cells

To reveal the genetic interactions among lst-4, snx-1, and snx-6, we compared the severity of Ced phenotype in single-, double-, and triple-deletion mutant embryos during multiple stages throughout embryonic development (Figure 1, C and D). Each of the deletion alleles removes the majority of the protein coding sequence and introduces a stop codon or frame shift mutation, and is presumed to create a null mutation in the corresponding genes (Supplemental Figure S1, B–D; Supplemental text).

In embryos, the deletion of each gene resulted in a mild increase in the number of cell corpses. However, the numbers of persistent cell corpses increased synergistically when the deletion in either snx-1 or snx-6 was combined with the deletion in lst-4 (Figure 1, C[e, f] and D). The number of cell corpses in snx-6; snx-1 double mutants, on the other hand, is similar to that in snx-1 or snx-6 single mutants (Figure 1, C[g] and D). Moreover, the numbers of cell corpses in the lst-4; snx-6; snx-1 triple-mutant embryos is comparable to that of the lst-4; snx-6 or lst-4; snx-1 double mutants (Figure 1, C and D). Because all three are null mutations, the above results indicate that snx-6 and snx-1 are likely to act in a linear pathway, whereas lst-4 acts in another parallel pathway, and that the two pathways drive the removal of cell corpses in a partially redundant fashion (Figure 1E). We chose to further characterize mutant alleles of lst-4 and snx-1, genes that represent each of the two pathways.

Retromer complex components VPS-26 and VPS-35 are not involved in cell corpse removal

Both mammalian SNX-1 and SNX-6 are subunits of the retromer complex that mediates retrograde transport of transmembrane proteins from endosomes to the trans-Golgi network (Bonifacino and Hurley, 2008). To examine whether C. elegans SNX-1 and SNX-6 regulate the degradation of apoptotic cells as components of the retromer complex, we determined whether other known C. elegans retromer complex components were involved in the removal of cell corpses (Figure S2; Supplemental text). We found that the deletion alleles of the C. elegans vps-26 or vps-35, which encode subunits of the cargo-recognition subcomplex of the retromer, did not display any notable increase in the number of the somatic or germ cell corpse (Figure S2). These results suggest that SNX-1 and SNX-6 likely act in a protein complex distinct from the canonical retromer to regulate cell corpse removal. Similarly, SNX-3, a homologue of budding yeast Snx3/Grd19, a retromer-associating protein (Strochlic et al., 2007), does not appear to be involved in cell corpse removal (Table 1; Figure S2, Supplemental text).

LST-4 and SNX-1 pathways specifically control the degradation but not engulfment of somatic apoptotic cells

We first found that the lst-4 and snx-1 deletions resulted in the prolonged duration of cell corpses generated during embryogenesis but did not cause any increase in the number of cell death events (Supplemental text; Figure S1, F and G). The Ced phenotype thus resulted from defect(s) in either the engulfment or the degradation of cell corpses. To distinguish between these possibilities, we monitored the engulfment of embryonic cell corpse C3 using phagocytic receptor CED-1::GFP, which clustered on the extending pseudopods during engulfment and then quickly disappeared from nascent phagosomes (Materials and Methods; Figure S3B[a-i]) (Yu et al., 2006). Somatic apoptotic cells C1, C2, and C3 are located on the ventral surface, die almost simultaneously during mid-embryogenesis, and are quickly engulfed and degraded by three adjacent hypodermal cells that extend to the ventral midline (Figure 2B) (Yu et al., 2006; Lu et al., 2009). We found that both the timing of engulfment initiation and the period it took to complete engulfment are normal in lst-4; snx-1 double-mutant embryos (Figure S3, A–C). Normal pseudopod extension dynamics was also demonstrated by the dynamic localization pattern of GFP::moesin, a specific marker for filamentous actin polymerized under pseudopods (Figure S3D; Supplemental text).

We further characterized the degradation of somatic cell corpses in embryos by monitoring the gradual reduction of the volume of phagosomes containing cell corpses C1, C2, and C3 over time using GFP::RAB-7 as a marker (Figure 2B) (Yu et al., 2008). GFP::RAB-7, when expressed in engulfing cells from the P_ced-1 promoter, displays bright and diffused GFP signal in the cytoplasm, allowing a nascent phagosome that contains an apoptotic cell to be first recognized as a dark, GFP(−) hole inside an engulfing cell (Figure 2A) (Yu et al., 2008). Once GFP::RAB-7 is enriched on the phagosomal surface a few minutes later, a phagosome could be tracked as a bright GFP(+) circle until it is fully degraded (Figure 2A) (Yu et al., 2008). The duration of a phagosome was measured from the moment when engulfment is complete (the “0 min” time point in Figure 2, C–F) to the moment when the phagosome disappeared. In wild-type embryos, the duration of a phagosome typically ranges between 40 and 60 min (Figure 2, C and G). In lst-4 or snx-1 single-mutant embryos, however, the duration varied within a broader range, with 20% or 28% phagosomes lasting longer than 70 min, respectively (Figure 2, D, E, and G). In lst-4; snx-1 double-mutant embryos, the delay in degradation was further enhanced, in which 48% of phagosomes lasted longer than 70 min and 19% even lasted longer than 100 min (Figure 2, F and G). These results indicate that, in the absence of lst-4 or snx-1, phagosome maturation is impaired, and further, when both lst-4 and snx-1 are inactivated, phagosome maturation is blocked. These defects result in the persistent cell corpses observed under the DIC optics (Figure 1, C and D).

LST-4 and SNX-1 are essential for the degradation, not engulfment, of germ cell corpses

We again used GFP::RAB-7 as a phagosomal marker to determine whether the many persistent germ cell corpses observed in the gonad of lst-4 or snx-1 single-mutant adult hermaphrodites were engulfed or not (Figures 2, I–J, and S4; Supplemental text). Although engulfed and unengulfed cell corpses display similar morphology under the DIC optics, only cell corpses inside phagosomes are surrounded by a bright ring of GFP::RAB-7 that was expressed in gonadal sheath cells, the engulfing cells for germ apoptotic cells (Figures 2I[d] and S4; Supplemental text). GFP::RAB-7 also allowed us to distinguish phagosomes as dark holes inside GFP(+) sheath cell cytoplasm even in cases in which the recruitment of RAB-7 to phagosomes was blocked (Figure 2I[e]). In lst-4(tm2423) mutants, all of the large number of DIC(+) germ cell corpses (mean = 58.9) were observed as dark holes imbedded inside the gonadal sheath cells (Figure 2, I and J). Due to the large number of phagosomes accumulating in the same sheath cells, a honeycomb pattern was generated inside the GFP(+) sheath cell cytoplasm (Figure 2I[e]). In snx-1(tm847) mutants, ∼95% of DIC(+) germ cell corpses (mean = 19.1) were inside phagosomes (Figure 2, I[f] and J). These results indicate that lst-4(tm2423) and snx-1(tm847) single mutants are exclusively defective in the degradation, not the internalization of germ cell corpses. We further performed transmission electron microscopy (TEM) to determine the engulfment status of germ cell corpses in the lst-4(tm2423); snx-1(tm847) double-mutant animals, the results of which indicated that 97% of the germ cell corpses analyzed were engulfed inside gonadal sheath cells and remained undegraded (Figure S5; Supplemental text). Together, these results demonstrated that lst-4 and snx-1 mutations abolished the cell corpse degradation activity, not the engulfment activity.

LST-4 and SNX-1 are both required for the incorporation of lysosomes and early endosomes into phagosomes

To determine the specific phagosome maturation defects caused by the deletions of lst-4 and snx-1, we examined the delivery of lysosomes and early endosomes to phagosomes, which are critical events for the degradation of phagosomal contents. We monitored the incorporation of lysosomes to phagosomes using CTNS-1, an integral lysosomal membrane protein and an established lysosomal marker (Mangahas et al., 2008; Yu et al., 2008). One group of wild-type animals was treated with γ-ray irradiation, which induced extra apoptosis events for the production of a larger number of phagosomes for characterization. Irradiation increased the average numbers of germ cell corpses from 3.5 to 13.9. In wild-type animals untreated or treated with γ-ray irradiation, continuous CTNS-1::GFP(+) signal was observed on the surfaces of ∼50% of phagosomes containing germ cell corpses, indicating that lysosomes were successfully recruited to and fused to phagosomal membranes (Figure 3, A, B, and E). In contrast, the CTNS-1::GFP signal was almost absent from the surfaces of phagosomes in the gonad of lst-4 (tm2423) or snx-1 (tm847) mutants and completely lost in lst-4; snx-1 double mutants (Figure 3, C–E). A similar defect in lysosomal incorporation was observed in 1.5-fold stage embryos of these mutants (Figure 3, F and G). Meanwhile, in lst-4 and snx-1 mutant embryos, relatively normal numbers of CTNS-1::GFP-labeled lysosomal particles were present in the cytoplasm (Figure 3F), indicating that in these mutants, lysosomal biogenesis and distribution in the cytoplasm were normal.

In addition, we monitored the incorporation of early endosomes to phagosomes using the HGRS-1::GFP reporter, a specific early endosomal marker expressed in engulfing cells (Yu et al., 2006). In 1.5-fold stage wild-type embryos, on average 47.7% of cell corpses distinguishable under DIC optics are labeled with HGRS-1::GFP(+) rings, indicating that the phagosomes enveloping these cell corpses received early endosomes from the host cells (Figure 3, H[a, b] and I). In lst-4 (tm2423); snx-1 (tm847) double-mutant embryos at the same stage, the fractions of phagosomes labeled with HGRS-1::GFP dropped to 6.9% (Figure 3, H[c, d] and I). In lst-4 and snx-1 single-mutant embryos, intermediate defects of endosome incorporation were observed (Figure 3I).

The above results indicate that LST-4 and SNX-1 act together to promote the efficient recruitment and fusion of early endosomes and lysosomes to phagosomes. We propose that the defect in incorporating intracellular vesicles to phagosomes is a primary cause for the phagosome maturation defect of the lst-4; snx-1 double mutants.

LST-4 and SNX-1 actively promote the extension of membrane tubules from phagosomes

Previously, we observed that during the degradation of apoptotic cells, transient membrane tubules were frequently extended from phagosomal surfaces into the engulfing cell cytoplasm (Yu et al., 2008). Furthermore, we have established that lysosomal particles are recruited to phagosomes through two alternative routes, direct encounter with phagosomal surfaces or being captured by these phagosomal tubules, which subsequently retract and bring lysosomal particles to phagosomal surfaces (Yu et al., 2008). Because LST-4, SNX-1, and SNX-6 possess BAR domains known as sensors and inducers of membrane curvature, we analyzed the roles of LST-4 and SNX-1 in the formation of phagosomal tubules. We monitored tubule extension from phagosomes containing C1, C2, or C3 using the CTNS-1::GFP reporter, which labeled originally lysosomal membranes and subsequently phagosomal membranes once a detectable amount of lysosomes were incorporated into phagosomes (Figure 3F) (Yu et al., 2008). In wild-type embryos, phagosomal tubules were of lengths varying from ∼300 nm to ∼2.2 μm, and their durations varied from within 20 to ∼200 s (Figure 4, A–C; Supplemental Movies S1 and S2). We observed examples in which a tubule captured a nearby lysosomal particle and brought it to phagosomal surface through retraction (Figure 4A). In some of those examples, the volume and fluorescent intensity of a lysosomal particle rapidly reduced once it was attached to a membrane tubule, indicating that membrane fusion between the lysosome and tubule might have occurred before the lysosomal particle was brought to phagosomal surfaces (Figure 4A). Consistently, the successful capture of a lysosome particle by a membrane tubule was followed by a significant enrichment of the lysosomal marker CTNS-1::GFP on the phagosome (Figure 4A). Examples in which a phagosomal tubule repeatedly attempted to attach to a lysosomal particle yet eventually failed to bring the particle to phagosomal surface were also observed (Figure 4B), suggesting that first, the attachment between a tubule and a particle is unstable and transit, and secondly, attachment does not guarantee fusion. In lst-4; snx-1 double-mutant embryos, unlike in wild-type embryos (Figure 4, C and E; Movie S2), the frequency of observing phagosomal tubules was significantly reduced (Figure 4, D and E; Movie S3). In lst-4 or snx-1 single-mutant embryos, intermediate defects in the formation of phagosomal tubules were observed (Figure 4E). These results indicate that the functions of LST-4 and SNX-1 are both required for the formation and/or stabilization of phagosomal tubules. Furthermore, the lack of phagosomal tubules in lst-4; snx-1 double-mutant background was typically correlated with a slower enrichment and reduced signal intensity of CTNS-1::GFP on nascent C1, C2, and C3 phagosomes (Figure 4D), indicating that the defect in tubule formation is closely related to the inefficient targeting of lysosomes to phagosomes.

LST-4 and SNX-1 differentially control the recruitment of RAB-7 to phagosomes

In lst-4(tm2423) mutant hermaphrodite gonad, GFP::RAB-7 was absent from all phagosomes observed inside gonadal sheath cells, indicating a complete failure in recruiting RAB-7 onto phagosomal surfaces (Figure 2, I and K). In contrast, ∼63% phagosomes were labeled by GFP::RAB-7 in snx-1(tm847) mutant gonadal sheath cells (Figure 2, I and K).

In lst-4 mutant embryos, the recruitment and maintenance of GFP::RAB-7 onto phagosomes was also defective, albeit less severe (Figure 2, D and H). Several types of RAB-7 maintenance defects have been observed, of which the premature dissociation of RAB-7 appears to be the common feature (Figure 2, D[c–d] and H). Defects in RAB-7 maintenance occurred less frequently in snx-1 single-mutant embryos; in addition, the reduction of phagosomal GFP::RAB-7 level was less severe (Figure 2, E and H). The types of defects and the distribution of each type observed from the lst-4; snx-1 double-mutant embryos are similar to that observed from lst-4 single-mutant embryos. These results indicate that, between LST-4 and SNX-1, LST-4 plays a major role in recruiting and maintaining RAB-7 on phagosomes.

LST-4, but not SNX-1, helps to maintain DYN-1’s association with phagosomal surfaces

SNX-9, SNX-18, and SNX-30, the mammalian homologues of LST-4, are known to interact with mammalian dynamins (Lundmark and Carlsson, 2003; Soulet et al., 2005; Shin et al., 2007). In a yeast two-hybrid assay (Materials and Methods), we observed the physical interaction between LST-4 and C. elegans DYN-1 (Figure 5A). This specific interaction was further confirmed by coimmunoprecipitation assays of DYN-1 and LST-4 coexpressed in mammalian cell culture (Figure 5B) (Materials and Methods). During phagosome maturation, DYN-1 is transiently recruited to the surface of nascent phagosome (Yu et al., 2006). In lst-4 mutants, we found that the recruitment of DYN-1 was partially defective; on the surface of 69% of phagosomes, we detected fewer number of DYN-1::GFP puncta, almost all of which dissociated earlier from phagosomes than those observed in wild-type embryos (Figure 5, C, E, and F). On the other hand, the dynamic association pattern of DYN-1 with phagosomes was not affected by the snx-1(tm847) mutation (Figure 5, D and F). Thus, the inactivation of LST-4 but not SNX-1 appeared to impair the maintenance of DYN-1 on phagosomal surfaces. We propose that LST-4 might use multiple mechanisms to regulate phagosome maturation, one of which is to facilitate DYN-1 function.

LST-4 and SNX-1 are transiently recruited to the surfaces of nascent phagosomes and phagosomal tubules

To determine whether LST-4 and SNX-1 are under the regulation of phagosomal PtdIns(3)P, we examined the phagosomal localization of GFP-tagged LST-4 and SNX-1. LST-4(d)::GFP and SNX-1::GFP, when expressed in engulfing cells, were fully functional for promoting cell corpses degradation (Supplemental Table S1; Supplemental text), indicating that the functions of LST-4 and SNX-1 in engulfing cells are sufficient for driving phagosome maturation. In wild-type animals expressing LST-4(d)::GFP or SNX-1::GFP, each reporter was primarily localized to the cytoplasm, with a portion enriched on cytoplasmic puncta (Supplemental Figure S6, A[a] and D[a]). Importantly, in these transgenic animals, we observed the association of GFP fusion proteins to phagosomes in both adult gonad and embryos (Figure 6, A and B), suggesting that LST-4 and SNX-1 act on phagosomal surfaces to promote phagosome maturation. In wild-type embryos, we observed rapid enrichment of both LST-4(d)::GFP and SNX-1::GFP on the surface of nascent phagosomes 2–4 min after phagosome formation (Figure 6, C–E). Later, the signal intensity of LST-4(d)::GFP on phagosomes gradually reduced (Figure 6, C[e–g] and E), and that of SNX-1::GFP reduced later and more gradually (Figure 6, D and E).

In addition to phagosomal surfaces, SNX-1::GFP and LST-4::GFP were also observed on the highly dynamic membrane tubules extended from phagosomes (Figure 6, G and H; Supplemental Movie S4). Often particularly bright SNX-1::GFP and LST-4::GFP signals were observed on a few spots on phagosome surface, where the phagosomal tubules frequently emerged (Figure 6, G and H). This localization pattern is consistent with direct roles of SNX-1 and LST-4 in bending phagosome membrane and promoting phagosomal tubule formation.

PtdIns(3)P recruits LST-4 and SNX-1 to phagosomal surfaces via their PX domains

The temporal distribution patterns of SNX-1 and LST-4 on phagosomal surfaces are similar to that of PtdIns(3)P during its first of two signal intensity peaks on phagosomes (Figure 6E), suggesting that PtdIns(3)P on nascent phagosomes might act as a specific factor that recruits SNX-1 and LST-4. In in vitro protein–lipid overlap assays, both SNX-1 and LST-4 associated with a few phosphoinositide (PI) species, including PtdIns(3)P (Supplemental Figure S7; Supplemental text).

To directly test whether LST-4 and SNX-1 are PtdIns(3)P effectors in vivo, we compared the localization patterns of LST-4 and SNX-1 on phagosomes in wild-type embryos versus vps-34(h510); F39B1.1(tm3171) double-mutant embryos, in which the phagosomal PtdIns(3)P was depleted due to the inactivation of two classes of PI-3 kinases responsible for PtdIns(3)P production (N. Lu and Z. Zhou, unpublished data). Due to their transient enrichment nature, SNX-1::GFP and LST-4::GFP were visible on the surfaces of ∼19% and ∼15% of phagosomes in wild-type 1.5-fold stage embryos, respectively (Figure 6F). In contrast, in vsp-34; F39B1.1 double-mutant embryos, SNX-1::GFP and LST-4::GFP signals were completely absent from phagosomal surfaces (Figure 6F), demonstrating that PtdIns(3)P is essential for the recruitment of LST-4 and SNX-1 onto phagosomal surfaces.

We next investigated whether the PX domains in LST-4 and SNX-1 were the targets of phagosomal PtdIns(3)P. The three-dimensional structures of different PX domains display remarkable similarity, with a few highly conserved positively charged residues precisely positioned inside the binding pocket for phosphoinositides (Supplemental Figure S8A) (Ellson et al., 2002; Seet and Hong, 2006). Arginine 58 in the PX domain of p40^phox is such a residue that plays a critical role in binding PtdIns(3)P (Bravo et al., 2001). We mutated R263 and R132, the corresponding arginine residues in LST-4 and SNX-1 (Figure S8A), respectively, to glutamine, and examined the phagosomal localization as well as the in vivo functions of these mutant proteins in transgenic animals [P_ced-1lst-4(d)(PX^m)::gfp or P_ced-1snx-1(PX^m)::gfp]. LST-4(d)(PX^m)::GFP and SNX-1(PX^m)::GFP displayed diffused cytoplasmic localization pattern (Figure S6, B and E) instead of the punctate localization pattern observed from the wild-type reporters (Figure S6, A and D). In time-lapse recording experiments, no enrichment of LST-4(PX^m)::GFP or SNX-1(PX^m)::GFP to the surface of maturing phagosomes was detected over time (Figure S6, H and I; Supplemental Movies S5 and S6). These results indicate that the PX domains are essential for the association of LST-4 and SNX-1 to cellular organelles, in particular to phagosomes. Furthermore, the R263Q mutation completely abolished the activity of P_ced-1lst-4(d)(PX^m)::gfp in rescuing the lst-4 mutant phenotype (Table S1). Similarly, P_ced-1snx-1(PX^m)::gfp in snx-1 mutant background displayed reduced rescuing activity (Supplemental Table S1). These results demonstrate that the binding to PtdIns(3)P and the resulting association to phagosomes are critical for the functions of LST-4 and SNX-1 in phagosome maturation.

The BAR domain is critical for the phagosomal localization and functions of SNX-1 and LST-4

A BAR domain is composed of three long coiled-coil α helices and is capable of forming a crescent-shaped homodimer (Peter et al., 2004). The concave surface of the BAR-domain dimer, which mainly consists of the C-terminal part of helix 2 and the loop between helices 2 and 3, is enriched with basic residues and proposed to be the interface that interacts with the negatively charged membranes of particular curvatures (Supplemental Figure S8, B and C) (Peter et al., 2004). The PX and BAR domains of a number of mammalian SNX-BAR sorting nexins act together to facilitate the protein–membrane association, in a manner referred to as “coincidence detection” (Carlton et al., 2004; Pylypenko et al., 2007; Traer et al., 2007; Haberg et al., 2008; Yarar et al., 2008; van Weering et al., 2010). In LST-4 and SNX-1, we simultaneously mutated a few conserved basic residues known to be essential for the functions of their mammalian orthologues (K492E and R499E in LST-4, and R376E, K377E, and R378E in SNX-1) (Figure S8, B and C) (Carlton et al., 2004; Peter et al., 2004; Pylypenko et al., 2007) and examined the effects on the phagosomal localization of the host proteins. Similar to the mutations in the PX domains, the mutations in these BAR domains resulted in a diffused localization pattern of the GFP-tagged reporters in the cytoplasm, and further greatly reduced the GFP signals on the surfaces of nascent phagosomes (Figure S6, C, F, H, and I), indicating that BAR domains are indeed necessary for the attachment of these proteins to cellular organelles and to phagosomes. Moreover, transgenes carrying these mutations lost part of the activities toward rescuing the Ced phenotypes of the corresponding mutants (Supplemental Table S1). Together, our data indicate that both the PX and the BAR domains play important roles in promoting the association of SNX-1 and LST-4 to phagosomal surfaces and in driving phagosome maturation. The combined point mutations in both PX domain and BAR domain of SNX-1 completely abolished its rescue activity (Table S1), indicating complementary roles of both domains in phagosome maturation.

DYN-1 is another essential factor for recruiting LST-4 to phagosomal surfaces

Consistent with the physical interaction observed between LST-4 and DYN-1 in vitro, the recruitment of LST-4 to phagosomal relies on DYN-1. Unlike in wild-type embryos, in dyn-1(en9) mutants, in which DYN-1’s self-assembly and its association with phagosomes were abolished (He et al., 2010), LST-4(d)::GFP signal was not detected on the surface of maturing phagosomes (Figure 5, G and H); rather, LST-4(d)::GFP were detected on puncta that are distributed inside the host cell cytoplasm, allowing a phagosome to be visualized as a dark hole (Figure 5H). Given that the transient enrichment pattern of LST-4 overlaps with the dynamic enrichment of DYN-1 and the initial production of PtdIns(3)P on phagosomal surfaces, that LST-4 interacts with both DYN-1 and PtdIns(3)P, and that LST-4 depends on both PtdIns(3)P and DYN-1 for enrichment on phagosomal surfaces, we propose that DYN-1 and PtdIns(3)P are two essential factors that together attract LST-4 to maturing phagosomes. The accumulation of DYN-1 and LST-4 on phagosomes thus appears to be mutually dependent.

SNX-1 mediates the association of SNX-6 to phagosomal surfaces

Both mammalian SNX-1 and SNX-6 and their budding yeast counterpart (Vps5 and Vps17) interact with each other (reviewed in Bonifacino and Hurley, 2008). In yeast two-hybrid assays, we observed that C. elegans SNX-1 and SNX-6 displayed strong and specific interaction with each other but not with LST-4 (Figure 7A). In addition, all three proteins displayed self-association activities, suggesting that they each might exist in multimeric forms in vivo (Figure 7A). In in vitro pull-down assays (Materials and Methods), we further found that GST-SNX-6 expressed and affinity-purified from Escherichia coli specifically pulled down hemagglutinin (HA)-tagged SNX-1 but not LST-4 from extracts of mammalian cells ectopically expressing each of the above proteins (Figure 7B). The above results have demonstrated the specific physical interaction between SNX-1 and SNX-6.

We next investigated the functional significance of the SNX-1/SNX-6 interaction. We constructed a P_ced-1 snx-6::gfp transgene that fully rescued the Ced phenotype of snx-6 mutants (Supplemental Table S1). In wild-type embryos co-overexpressing both SNX-6::GFP and SNX-1::mRFP, a prominent level of SNX-6::GFP was detected on the surfaces of nascent phagosomes (Figure 7C). Moreover, the kinetics of SNX-6::GFP’s enrichment onto and dissociation from phagosomal surfaces was highly similar to that of SNX-1::mRFP (Figure 7C). Interestingly, in wild-type embryos expressing P_ced-1 snx-6::gfp alone, the enrichment of SNX-6::GFP on phagosomal surfaces is much reduced (Figure 7D). Furthermore, in snx-1(tm847) mutant embryos, SNX-6::GFP was not detected on nascent phagosomes (Figure 7E). These results indicate that SNX-6 is recruited to the surfaces of nascent phagosomes in an SNX-1-dependent manner, a mechanism supported by their physical interaction detected in vitro.

The ced-1 pathway controls the localization of LST-4 and SNX-1 on phagosomes

In C. elegans, the engulfment and degradation of apoptotic cells are controlled by two parallel genetic pathways (Yu et al., 2008; Zhou and Yu, 2008). The production of PtdIns(3)P on phagosomal surfaces is primarily controlled by the pathway led by phagocytic receptor CED-1 and, to a much less extent, by the second pathway composed of ced-2, ced-5, ced-10, and ced-12 (Yu et al., 2008). If LST-4 and SNX-1 act directly downstream of PtdIns(3)P, we anticipated that the phagosomal recruitment of LST-4 and SNX-1 was also under the control of the ced-1 pathway. To examine this model, we monitored the recruitment of LST-4 and SNX-1 to phagosomes in ced-1 and ced-5 null mutant embryos. In wild-type embryos, 100% of phagosomes acquired LST-4 and SNX-1 within 8 min after enclosure (Figure 8, A, E, F, and J). In contrast, in ced-1 mutant embryos, 80% of phagosomes never acquired LST-4::GFP (Figure 8, C and E) and 66.7% of phagosomes never acquired SNX-1::GFP during a 50-min recording period starting at the enclosure of a phagosome (Figure 8, H and J). In some cases, GFP signals disappeared from phagosomal surfaces much sooner than in wild-type background (Figure 8, B, E, and J), indicating an additional defect in signal maintenance. Furthermore, the LST-4(d)::GFP or SNX-1::GFP signal intensity on phagosomal surfaces was in general lower than normal, indicating another aspect of recruitment or maintenance defects (Figure 8, B and G). On the other hand, the localization patterns of LST-4 and SNX-1 on cytoplasmic puncta were not affected in ced-1 mutant embryos (Supplemental Figure S6, A[b] and D[b]), indicating that CED-1 specifically regulates the phagosomal recruitment of LST-4 and SNX-1.

In ced-5 null mutant embryos, phagosomal recruitment of LST-4(d)::GFP followed a relatively normal kinetics (Figure 8, D and E). However, 71% of phagosomes acquired SNX-1::GFP with significantly delayed kinetics but obtained nevertheless strong GFP signals (Figure 8, I and J). Consistent with these results, the defects in the maturation of C1, C2, and C3 phagosomes were much weaker in ced-5 mutants than in ced-1 mutants (Yu et al., 2006). Together, these results indicate that, like the production of phagosomal PtdIns(3)P, the recruitment of LST-4 and SNX-1 to phagosomes is primarily controlled by the ced-1 pathway, whereas the ced-5 pathway also contributes to these events.

Deletion of snx-1 does not affect the presentation of CED-1 to engulfing cell surfaces or the activity of CED-1 in promoting phagosome maturation

The disappearance of CED-1 from the surfaces of nascent phagosomes is slower in snx-1 single-mutant than in wild-type or lst-4 single-mutant embryos (Figures 9A and S3, B and C). This phenotype suggests that, in snx-1 deletion mutant embryos, phagosome maturation might be arrested at an early stage before the dissociation of CED-1 from phagosomes. Alternatively, snx-1 may directly regulate the dissociation of CED-1 from phagosomal membranes. A recent report indicated that SNX-1 and SNX-6 act in the retromer complex to recycle CED-1 from phagosomal surfaces to the plasma membrane, and in this manner regulate CED-1-mediated engulfment of apoptotic cells (Chen et al., 2010). To determine whether the prolonged presence of CED-1 on phagosomes affect the recycling of CED-1 to the plasma membrane of the host cell, we measured the level of CED-1::GFP on the plasma membrane of ABplaapppp, the engulfing cell for C3, during and after the internalization of C3, in embryos carrying enIs7 [P_ced-1ced-1::gfp], an integrated array (Venegas and Zhou, 2007). Over a 32-min time course starting from the initiation of engulfment, in both wild-type and snx-1 mutant embryos, the levels of GFP signal on the plasma membrane merely fluctuated within 15% of that at the “0 min” time point (Figure 9, A and B). Moreover, no significant differences in signal intensity levels were observed between wild-type and snx-1(tm847) backgrounds (Figure 9, A and B). This set of results indicates that, in snx-1 mutants, the prolonged existence of CED-1 on phagosomal surfaces does not significantly affect the level of CED-1 on the surface of the host cell. In addition, we observed that, throughout the entire embryogenesis, CED-1::GFP is indeed presented on the surfaces of cells expressing P_ced-1 in snx-1(tm847), snx-6(tm3790), and lst-4(n2423) single-mutant embryos as in wild-type embryos (Figure 9C and data not shown), indicating that, unlike that reported by Chen et al. (2010), CED-1’s plasma membrane presentation in embryos was not affected by the deletion of any of these three genes. Furthermore, we quantified the CED-1::GFP signal intensity in entire individual embryos at four embryonic stages, including a period spanning from 200 to 420 min after the first embryonic cell division, during which most (88 of 118) embryonic cell death events occur and cell corpses are removed (Materials and Methods) (Sulston et al., 1983). We found that, at most stages, the mean GFP intensity levels were comparable in wild-type, snx-1(tm847), snx-6(tm3790), and lst-4(tm2423) single-mutant embryos (Supplemental Table S2). Only at the late fourfold stage, which is at least 100 min after the last cell death event occurs (Sulston et al., 1983), were relatively weaker CED-1::GFP signal intensity levels observed from snx-1(tm847) and snx-6(tm3790) mutant embryos (77% and 50% of that of wild-type level, respectively) (Table S2). These results are different from the nearly complete loss of CED-1::GFP protein in snx-1(tm847) embryos reported by Chen et al. (2010), and indicate that, during the period when cell-corpse engulfment and degradation events are most active, CED-1::GFP level is not reduced by mutations in snx-1 or snx-6.

Chen et al. (2010) reported that overexpressing CED-1::GFP under the control of P_ced-1 from an integrated array (smIs34) resulted in a complete rescue of the Ced phenotype displayed by snx-1(tm847) and snx-6(tm3790) mutant embryos. In contrast, we found that the overexpression of CED-1::GFP from a reporter plasmid constructed identically and was also integrated into the genome (enIs7[P_ced-1ced-1::gfp]), did not result in any significant rescuing activity of the Ced phenotype in the snx-1(tm847), snx-6(tm3790), or lst-4(n2423) single-mutant embryos at multiple embryonic stages identical to that examined by Chen et al. (2010) (Figure 9, G–J; Supplemental text). Together, the above results strongly suggest that SNX-1, SNX-6, and LST-4 must control phagosome maturation through a mechanism other than regulating the cell-surface presentation or overall level of CED-1.

We thus directly measured whether the activity of CED-1 in promoting phagosome maturation was affected by the deletion of snx-1. As established previously, CED-1 triggers phagosome maturation through recruiting the large GTPase DYN-1 to the surface of nascent phagosomes. DYN-1 subsequently induces the robust production of PtdIns(3)P on phagosomal membranes and also recruits membrane tethering factors RAB-5 and RAB-7 to phagosomal surfaces (Yu et al., 2008; He et al., 2010). The enrichment of all of these molecules on phagosomal surfaces thus is dependent on CED-1 and could serve as readouts for CED-1 activity. We already observed that in snx-1(tm847) mutant embryos, the enrichment of DYN-1 onto the surface of nascent phagosomes was normal (Figure 5, D and F). We further monitored the presentation of PtdIns(3)P and RAB-5 on phagosomal surfaces and found no difference in the level or timing of the dynamic presence of these molecules in either wild-type or snx-1(tm847) mutant embryos (Figure 9, D–F). These results demonstrate that SNX-1 does not influence CED-1’s activity in triggering phagosome maturation, a conclusion consistent with our results that indicate that SNX-1 and LST-4 both act downstream of CED-1 for promoting phagosome maturation.

LST-4, SNX-1, and SNX-6 control the degradation but not the engulfment of apoptotic cells

The engulfment and degradation of phagocytic targets are two different cellular events that are executed through distinct mechanisms (Caron and Hall, 2001; Vieira et al., 2002). Determining which of these two events these PtdIns(3)P effectors are involved in is pivotal for further understanding their molecular functions. Two lines of experimental evidence demonstrate that LST-4 and SNX-1 specifically control the degradation, not the engulfment of apoptotic cells. First, analyses using TEM technique and fluorescence micro­scopy both strongly indicate that, in each of the snx-1(tm847) and lst-4(tm2423) deletion alleles as well as the snx-1; lst-4 double mutants, germ cell corpses were normally internalized yet remained in phagosomes without being promptly degraded. Second, time-lapse recording monitoring somatic cell corpses revealed that the kinetics of engulfment were normal in lst-4; snx-1 double mutant embryos; rather, the degradation of cell corpses was blocked. This conclusion is consistent with the observations that PtdIns(3)P, the upstream regulating molecule of SNX-1 and LST-4, is not detected on phagosomal surfaces until engulfment is complete and phagosomes are sealed (Yu et al., 2008). This conclusion is different from a recent report that concludes that SNX-1 and the retromer complex in which SNX-1 acts primarily control the engulfment of apoptotic cells (Chen et al., 2010).

SNX-1 and SNX-6 promote phagosome maturation as downstream effectors of CED-1 and in a manner independent of the canonical retromer complex

Mammalian SNX1 and SNX6 are the known members of the membrane deformation subcomplex of the evolutionarily conserved retromer complex that mediates endosome-to-Golgi retrieval of transmembrane receptors and other trafficking-related proteins (reviewed in Bonifacino and Hurley, 2008; van Weering et al., 2010). Retromer’s cargo recognition subcomplex, a heterotrimer consisting of VPS26, VPS29, and VPS35, targets specific transmembrane receptors for retrograde transport (Bonifacino and Hurley, 2008; van Weering et al., 2010). C. elegans VPS-26 and VPS-35 were known to act in the retromer-dependent recycling of transmembrane protein MIG-14/Wntless (Pan et al., 2008; Yang et al., 2008). Our observation that the inactivation of VPS-26 and VPS-35 do not result in the persistence of cell corpses indicates that the canonical retromer complex is not involved in the degradation of apoptotic cells and further suggests that the phagosome-maturation functions of SNX-1 and SNX-6 are likely executed through a distinct protein complex and a novel mechanism. We found that the deletion of vps-29 caused a mild defect in the removal of germ cell but not somatic cell corpses. Because VPS29 possesses an SNX1-interacting motif (Collins et al., 2005; Wang et al., 2005), in the gonad, this distinct protein complex might also include VPS-29 (Figure 10B).

C. elegans SNX-1 was implicated in the regulation of endosomal clathrin dynamics (Shi et al., 2009). Chen et al. (2010) proposed that SNX-1 and SNX-6 act in the retromer complex to recycle CED-1 from phagosomal surfaces back to the plasma membrane, and in this manner regulate apoptotic cell engulfment. Our results, many obtained under highly comparable experimental conditions, are different from that reported by Chen et al. (2010) and do not support this model. In particular, we observed that, in snx-1(tm847) null mutant embryos, CED-1 was presented on engulfing cell surfaces at normal levels. Furthermore, the internalization of an apoptotic cell and the prolonged presence of CED-1 on the phagosome did not appear to affect the level of CED-1 on the surface of the host cell. In addition, in snx-1(tm847), snx-6(tm3790), or lst-4(tm2423) mutant backgrounds, the overall levels of CED-1::GFP in entire embryos are similar to that in wild-type embryos at the same developmental stages during early to mid-embryogenesis, when most cell death events occur (Supplemental Table S2). More importantly, we found that, in contrast to the full rescuing activity reported by Chen et al. (2010), overexpression of CED-1 did not rescue the Ced phenotype of snx-1 or snx-6 mutant embryos. These results, together with our finding that snx-1 mutants are not defective in cell corpse engulfment, indicate that the level and, more importantly, the function of CED-1 on engulfing cell surfaces are not primarily controlled by SNX-1 or SNX-6.

The direct evidence that the snx-1 deletion does not affect CED-1’s phagosome maturation activity comes from the observation that, in snx-1 mutant embryos, the dynamic phagosomal enrichment patterns of DYN-1, PtdIns(3)P, and RAB-5, three phagosome maturation factors that are dependent on CED-1 for their enrichment on phagosomal surfaces, occur in the normal manner. On the other hand, SNX-1 is recruited to phagosomes in a PtdIns(3)P- and CED-1-dependent manner. Together, these results have established that SNX-1 acts downstream of CED-1 to control phagosome maturation (Figure 10A). Recently, several reports revealed cases in which mammalian SNX1 regulates membrane trafficking and membrane remodeling in retromer-independent mechanisms (Gullapalli et al., 2006; Nisar et al., 2010; Prosser et al., 2010). Our finding reveals a new retromer-independent function of the SNX-1/SNX-6 complex.

LST-4, SNX-1, and SNX-6 promote the fusion of lysosomes to phagosomes through two distinct molecular mechanisms

In lst-4 and snx-1 single-mutant and lst-4; snx-1 double-mutant animals, a striking set of defects reside in the incorporation of early endosomes and lysosomes to phagosomes. These defects might be the primary cause for the defects in phagosome maturation, since the recruitment and fusion of these organelles are essential for the delivery of digestive enzymes to phagosomes and the acidification of phagosomal lumen. Our further analyses indicate the presence of two distinct molecular mechanisms used by these SNX-BAR proteins in the recruitment and fusion of intracellular vesicles to phagosomes (Figure 10).

The first mechanism involves the functions of BAR domains in inducing and stabilizing membrane curvature. High membrane curvature has been proposed to facilitate fusion (Chanturiya et al., 2002; Chernomordik and Kozlov, 2003; Marrink and Mark, 2003). The C. elegans phagosomes that contain apoptotic cells, which are of 2.0–2.5-μm diameters, are relatively flat compared with various intracellular vesicles that fuse to them (Yu et al., 2006, 2008). Regional membrane bending may thus be necessary for the extensive fusion events to occur between maturing phagosomes and the smaller endosomal and lysosomal particles. BAR domains are capable of using their rigid concave surfaces to induce and/or stabilize membrane curvature (Frost et al., 2009). Both mammalian SNX1 and SNX9 are able to tubulate liposomes in vitro and induce tubular structure formation when overexpressed in cultured cells (Carlton et al., 2004; Pylypenko et al., 2007; Haberg et al., 2008). Recruitment of SNX1 to Salmonella-containing vacuoles (SCV) induces the formation of tubules from SCV surfaces (Bujny et al., 2008). Previously, it has been established that phagosomes containing immunoglobulin G–opsonized particles or apoptotic cells extend membrane tubules, which facilitate the capture and incorporation of lysosomal particles (Figure 10B) (Harrison et al., 2003; Yu et al., 2008). These tubules are an extreme form of regional high membrane curvature. Here we report that the extension of phagosomal tubules requires the additive functions of both LST-4 and SNX-1, which are enriched on tubules. The BAR domains of LST-4, SNX-1, and SNX-6 may thus directly modulate membrane curvature by initiating and/or stabilizing phagosomal tubules and smaller protrusions. We further propose that the putative LST-4 homodimer and the SNX-1/SNX-6 heterodimer might perform such functions in a cooperative manner that has not been described previously, since lst-4; snx-1 double mutants displayed a much enhanced tubule extension defect than each of the single mutants (Figure 10, B and C).

The second mechanism, which is LST-4-specific, involves the recruitment and maintenance of the membrane-tethering factor RAB-7 to phagosomal surfaces. The association of RAB-7 to phagosomes is essential for the extension of phagosomal tubules as well as the fusion between lysosomal particles and phagosomal membranes (Harrison et al., 2003; Yu et al., 2008). In the gonad of lst-4 but not snx-1 null mutant hermaphrodites, no RAB-7 signal is observed on phagosomal surfaces, indicating an essential role of LST-4 in this event. In lst-4 mutant embryos, time-lapse recording revealed frequent premature dissociation of RAB-7 from phagosomal surfaces. We thus propose that, in addition to directly promoting tubule formation, LST-4 also indirectly facilitates the incorporation of lysosomes to phagosomes through retaining RAB-7 on phagosomal surfaces.

The interaction between mammalian SNX9 and dynamin stimulates dynamin’s GTPase activity in vitro (Yarar et al., 2008) and facilitates dynamin’s membrane association in cells (Lundmark and Carlsson, 2004). In lst-4 mutant embryos, the defect in maintaining RAB-7 on phagosomes correlates well with that in maintaining DYN-1. As DYN-1 and LST-4 rely on each other for stable association with phagosomes (Figure 10B), and as the association of RAB-7 with phagosome is DYN-1-dependent (Yu et al., 2008; He et al., 2010), we propose that LST-4 acts to ensure the retention of RAB-7 on phagosomes at least partially through stabilizing the DYN-1-phagosome association.

Mammalian PtdIns(3)P-binding proteins EEA1 and Hrs1 were both implicated in the maturation of phagosomes containing latex beads (Fratti et al., 2001; Vieira et al., 2004). To our surprise, eea-1 null mutants or hgrs-1 (RNAi) are not defective in the removal of apoptotic cells (Table 1). This discrepancy may reflect the difference in the regulation of phagosome maturation in cultured mammalian cells versus C. elegans or, alternatively, in the nature of engulfed particles. Differential dynamic enrichment patterns of PtdIns(3)P on phagosomes containing different objects have been observed (Vieira et al., 2001; Yu et al., 2008), which might differentially influence the involvement of distinct PtdIns(3)P effectors.

The dynamic phagosomal association patterns of LST-4, SNX-1, and SNX-6 are regulated by complex PtdIns(3)P-mediated mechanisms

We have demonstrated that the phagosomal association of LST-4, SNX-1, and SNX-6, which is essential for their in vivo functions in promoting phagosome maturation, is dependent on the PX domain–PtdIns(3)P interaction. PX domains alone, however, often are not sufficient for targeting their host proteins to membranes (Lemmon, 2008). This seems to be the case for LST-4 and SNX-1 because the phagosomal association of both proteins also depends on their C-terminal BAR domains. BAR domains are likely to facilitate the protein-membrane association through two different mechanisms. The intrinsic dimerization of BAR domain could drive the bivalent binding of PX domain to PtdIns(3)P-containing phagosome membranes (Frost et al., 2009). In addition, the basic patches on the concave surface of the BAR domains might directly interact with negatively charged membranes (Peter et al., 2004). A number of mammalian SNX-BAR sorting nexins, including SNX-1, SNX-5, SNX9, and SNX-18, rely on the “coincidence detection” mechanism, through which the PX domain recognizes specific membrane phosphoinositides and the BAR domain senses membrane curvature, for targeting to restricted regions of membranes (Carlton et al., 2004; Pylypenko et al., 2007; Traer et al., 2007; Haberg et al., 2008; Yarar et al., 2008; van Weering et al., 2010). Our observations support the model that the PX and BAR domains act in distinct manners yet cooperatively to target LST-4 and SNX-1 to phagosomal surfaces.

We found that SNX-6 relied on the physical interaction with SNX-1 for targeting to phagosomes. The physical and genetic interactions between SNX-1 and SNX-6 indicate that SNX-1 and SNX-6 form a protein complex on phagosomal surfaces to control phagosome maturation in one of the two parallel pathways (Figure 10B). The SNX-1/SNX-6 interaction is likely to enable the formation of a BAR domain dimer that could strengthen the association of both proteins with phagosomal membranes and PtdIns(3)P.

Interestingly, LST-4 and SNX-1 associate with phagosomes transiently, concurrent only with the first of the two waves of PtdIns(3)P on phagosomes. This phenomenon suggests that PtdIns(3)P alone is not sufficient to retain the SNX-BAR proteins on phagosomal surfaces efficiently. We found that C. elegans LST-4 interacted with DYN-1 (dynamin) in vitro like its mammalian orthologue and is dependent on DYN-1 for association with phagosomal surfaces. DYN-1 is transiently recruited to the surface of nascent phagosomes prior to the first wave of PtdIns(3)P (Yu et al., 2006, 2008). We propose that DYN-1 and PtdIns(3)P together determine the transient enrichment pattern of LST-4 on phagosomal surfaces. On the other hand, phagosomal-localized LST-4 further stabilizes DYN-1’s association with phagosomes via interaction with DYN-1. Such mutual dependency strengthens the membrane association of both DYN-1 and LST-4.

Mutations and strains

C. elegans strains were grown at 20°C as previously described (Brenner, 1974). The N2 Bristol strain was used as the reference wild-type strain. Mutations and integrated arrays are described by Riddle et al. (1997) and the Worm Base (http://www.wormbase.org), except when noted otherwise: LGI, ced-1(n1735), F22G12.4(tm3221), snx-3(tm1595); LGII, rab-7(ok511), aka-1(tm389), wdfy-2(tm3806), pld-1(ok2222); LGIII, mtm-6(ok330), rabn-5(tm1555); LGIV, ced-5(n1812), lst-4(tm2423), tag-77(tm702), exc-5(rh232), rabs-5(ok1513), F13E9.1(tm1886), Y116A8C.26(tm2404); LGV, unc-76(e911), eea-1(tm933), rskd-1(tm4031), snx-6(tm3790); LGX, snx-1(tm847), F39B1.1(tm3171), F17H10.3(tm3643), enIs7[P_ced-1ced-1::gfp] (Venegas and Zhou, 2007). All ok alleles are generated by the C. elegans Gene Knockout Consortium and distributed by Caenorhabditis Genetics Center (CGC). All tm alleles were generated and provided by the National Bioresource Project of Japan. Strains carrying lst-4(tm2423), snx-1(tm847), and snx-6(tm3790) used here were outcrossed four times. Transgenic worms were generated by microinjection as previously described (Jin, 1999). The plasmids were injected with coinjection marker pUNC76 [unc-76(+)] (Bloom and Horvitz, 1997) into unc-76 (e911) mutant adult hermaphrodites and transgenic animals were identified as non-Unc animals.

Plasmid construction

The snx-1, snx-6, lst-4(d), and lst-4(e) cDNAs were amplified from a mixed-stage C. elegans cDNA library (Z. Zhou and H.R. Horvitz, unpublished data) using the PCR. The engulfing cell-specific expressing plasmids P_ced-1snx-1::gfp, P_ced-1snx-6::gfp, P_ced-1lst-4(d)::gfp, and P_ced-1lst-4(e)::gfp were constructed by cloning the snx-1, snx-6, lst-4(d), and lst-4(e) cDNAs into the multicloning sites of pZZ829, a plasmid carrying P_ced-1 (the ced-1 promotor), a 3′ gfp tag, and the unc-54 3′ UTR (Yu et al., 2006). The (R132Q), (R376E K377E R378E), and (R132Q R376E K377E R378E) mutations were introduced into P_ced-1snx-1::gfp using the QuickChange Site-directed Mutagenesis Kit (Stratagene, La Jolla, CA) to generate P_ced-1­snx-1(PX^m)::gfp, P_ced-1snx-1(BAR^m)::gfp, and P_ced-1snx-1(PX^m BAR^m)::gfp, respectively. Using the same strategy, the (R263Q) and (K492E R499E) mutations were introduced into P_ced-1lst-4(d)::gfp to generate P_ced-1lst-4(d)(PX^m)::gfp and P_ced-1lst-4(d)(BAR^m)::gfp, respectively. snx-1, snx-6, and lst-4(d) cDNAs were cloned into vector pGEX4T-1 (GE Healthcare, Piscataway, NJ) to generate pGST-snx-1, pGST-snx-6, and pGST-lst-4(d) plasmids, respectively. pGST-PX^SNX-1 (GST-PX domain of SNX-1) was constructed by cloning the cDNA encoding amino acid 75–229 of SNX-1 into vector pGEX4T-1. To generate yeast two-hybrid assay constructs, snx-1, snx-6, and lst-4(d) cDNA were first cloned into the pENTR/D-TOPO (Invitrogen, Carlsbad, CA) vector and then transferred into Gateway destination vectors pLexAgtwy or pACT2.2gtwy (Caldwell et al., 2006) by LR recombination (Invitrogen), fused with the DNA binding domain of LexA or the transcriptional activation domain of GAL4, respectively.

Nomarski DIC microscopy

DIC microscopy was performed with an Axionplan 2 compound microscope (Carl Zeiss, Thornwood, NY) equipped with Nomarski DIC optics, a digital camera (AxioCam MRm; Carl Zeiss), and imaging software (AxioVision; Carl Zeiss). Previously established protocols were used to score cell corpses under DIC microscopy (Yu et al., 2006). Somatic embryonic cell corpses were scored in the head region of embryos at different developmental stages. Germ cell corpses were scored in one of the two gonadal arms of adult hermaphrodites 48 h post-L4 stage unless otherwise stated. Time-lapse DIC microscopy was used to count the total number of apoptotic events and record the duration of each cell corpse for a 300-min time window during embryogenesis as described (Yu et al., 2006). Embryos undergoing first cell cleavage were identified and defined as “0 min” embryo. The same embryos were recorded from “160 min” to “460 min” at 3-min intervals. For each time point, DIC images of 40 z-sections were captured at 0.5-μm intervals.

Fluorescence microscopy and time-lapse recording

An Olympus IX70-Applied Precision DeltaVision microscope equipped with a Photometris Coolsnap digital camera was used to capture fluorescence images, and the Applied Precision SoftWoRx software was used for image deconvolution and processing (Yu et al., 2008). To monitor the kinetics of the subcellular localization of various GFP reporters during the engulfment and degradation of cell corpses C1, C2, and C3, the procedure followed an established protocol (Yu et al., 2008; Lu et al., 2009). Recording of the ventral surface of embryos started at 310–320 min post–first cleavage and lasted for 60–120 min in 1- or 2-min intervals, except when the extension of phagosomal tubules were monitored, when 20-s intervals were chosen and the recording lasted 20–30 min. At each time point, 10–15 serial z-sections at a 0.5-μm interval were recorded. Signs such as embryo elongation and movement were closely monitored to ensure that the embryo being recorded developed normally. Images were analyzed using the ImageJ software and fluorescence signal intensity was measured (Lu et al., 2009). The enrichment of fluorescence signal on phagosomal surfaces was considered significant when the ratio of the signal intensity on phagosomal surfaces versus the cytoplasm was ≥1.2 (Lu et al., 2009).

RNAi

RNAi experiments were performed using feeding protocol as previously described (Kamath et al., 2001). In brief, mid–L4-stage hermaphrodites were transferred to RNAi plates. After 48 h, the numbers of germ cell corpses per gonad arm were scored under DIC optics. The RNAi feeding constructs for lst-2, ZK632.12, hgrs-1, mtm-3, R11D1.10, ppk-3, and F25H2.2 were from a C. elegans RNAi library (Kamath et al., 2003). The RNAi feeding construct for Y48E1B.14 was constructed by cloning PCR amplified cDNA fragment into vector pPD129.36 using the primer pairs: 5′-atgatgaacatgtccgatcc-3′ and 5′-acgcgtcggatgcaccggatg-3′. The vector pPD129.36 was used as a control RNAi construct.

Yeast two-hybrid assay

The yeast two-hybrid assay to detect possible interaction among SNX-1, SNX-6, and LST-4 was performed as described (Vojtek and Hollenberg, 1995; Bai and Elledge, 1997). Different combinations of bait and prey constructs were transformed into yeast strain CTYLH, and transformants were selected on leucine and tryptophan drop-out synthetic plates. Interactions between baits and preys were examined at 30°C using the β-galactosidase reporter assay.

GST pull-down and coimmunoprecipitation

For GST pull-down assays, GST and GST-tagged SNX-6 were purified from E. coli lysates using Glutathione Sepharose 4B (GE Heathcare) according to the manufacturer’s instructions. snx-1 and lst-4 cDNA were cloned into pcDNA3.1/C-HA and transiently transfected into HEK 293T cell individually using Lipofectamine 2000 (Invitrogen). Transfected cells were lysed with NETN buffer (50 mM Tris-HCl, pH 8.0, 1 mM EDTA, 150 mM NaCl, 0.5% NP-40 supplemented with protease inhibitor cocktail [Roche Applied Science, Mannheim, Germany and phenylmethylsulfonyl fluoride [PMSF]). Equal amounts of cell lysates expressing SNX-1-HA or LST-4-HA were incubated with immobilized GST or GST-tagged SNX-6 at 4°C for 3 h. After three washes with NETN, proteins bound on resins were separated by SDS–PAGE and detected by Western blot with anti-HA antibody (Sigma Chemical, St. Louis, MO).

For coimmunoprecipitation assays, dyn-1 and ced-6 cDNAs were cloned into pcDNA3.1/N-HA, and the lst-4(d) cDNA into pcDNA3.1/C-Flag. Pairs of constructs expressing HA- and FLAG-tagged proteins were transiently transfected into HEK 293T cells. Approximately 500 μg transfected cell lysates in NETN buffer was incubated with 5 μg anti-HA antibody (Sigma Chemical) at 4°C for 1 h and then with Protein A agarose beads (GE Healthcare) overnight at 4°C. After three washes with NETN, the precipitated proteins were analyzed by SDS–PAGE and Western blots.