The phosphoinositide phosphatase MTM-1 regulates apoptotic cell corpse clearance through CED-5–CED-12 in C. elegans

Tissue homeostasis and organ formation requires a physiological balance that regulates proliferation and death of cells (Hanahan and Weinberg, 2000). Once a single cell is committed to die, neighboring or specialized phagocytic cells recognize, internalize and degrade the cell corpse. Cell corpse clearance ensures that corpses are not able to release harmful intracellular contents into the surrounding tissue, which could evoke inflammation and autoimmune disease (Nagata et al., 2010; Savill and Fadok, 2000).

In the nematode Caenorhabditis elegans, greater than 20 genes involved in the recognition, internalization and degradation of apoptotic cell corpses have been described. Loss-of-function (lf) alleles of these genes result in the accumulation of persistent cell corpses in the soma and/or in the hermaphrodite gonad. Genetic and biochemical analyses placed the engulfment genes in three partially redundant pathways that might converge at the level of the small GTPase CED-10 (Rac1 in mammals), which is followed by a ‘linear’ phagosome maturation pathway that is required for the degradation of engulfed cell corpses (Fullard et al., 2009; Kinchen, 2010).

In a first pathway, CED-7 (ABCA1) and CED-1 (MEGF10) have been proposed to function as receptors for the recognition of dying cells, either directly, or via the bridging molecule TTR-52 (Wang et al., 2010; Wu and Horvitz, 1998b; Zhou et al., 2001b). The adaptor protein CED-6 (GULP) physically binds to and might transduce signal(s) from CED-1 further downstream to the small GTPase CED-10, and additionally regulates phagosome maturation through the large GTPase DYN-1 (Dynamin) (Kinchen et al., 2005; Liu and Hengartner, 1998; Yu et al., 2006).

In a second signaling cascade, two small Rho GTPases act in a serial manner: UNC-73 (Trio) acts as a guanosine exchange factor (GEF) for the small GTPase MIG-2 (RhoG) (deBakker et al., 2004). Active (i.e. GTP-loaded) MIG-2 regulates corpse removal by modulating plasma membrane recruitment of the unconventional bipartite CED-5 (Dock180)–CED-12 (Elmo) GEF complex (Gumienny et al., 2001; Wu and Horvitz, 1998a). The GEF complex is stabilized further by the adaptor molecule CED-2 (CrkII), which also promotes the activation of CED-10 (Akakura et al., 2005; Gumienny et al., 2001). GTP-bound CED-10 initiates extensive cytoskeletal rearrangements, a requirement for the engulfment of cell corpses (Kiyokawa et al., 1998; Reddien and Horvitz, 2000).

Several receptors in the second signaling pathway have been described, with MOM-5 (Frizzled) proposed to act as the major receptor important for embryonic corpse recognition, signaling through GSK-3 (GSK3β), APR-1 (APC) and CED-2 to activate CED-10, providing a link between corpse recognition and activation of CED-10 (Cabello et al., 2010). Additionally, the two integrins INA-1 (Integrin α) and PAT-3 (Integrin β) play a redundant role in corpse recognition and might also recruit CED-2 to the phagocytic cup in a phosphotyrosine-dependent manner through SRC-1 (Src) (Hsu and Wu, 2010).

A third pathway has been described in which ABL-1 (Abl kinase) regulates cytoskeletal remodeling through ABI-1 (Abi), potentially through CED-10 and in a yet unknown CED-10-independent way (Hurwitz et al., 2009).

Once CED-10 is activated, downstream signaling towards the cytoskeleton will lead to an extensive re-orchestration within the phagocytic cell. As a result, the cell corpse will be fully internalized and its components will be degraded and recycled (Kinchen, 2010; Nagata et al., 2010). Following corpse internalization, the GTPase activating protein (GAP) SRGP-1 inactivates CED-10, thereby turning engulfment signaling off (Neukomm et al., 2011).

Cytoskeletal rearrangements are associated with altered membrane dynamics on the plasma membrane (Saarikangas et al., 2010). Polyphosphoinositides (PPIn), phosphorylated derivatives of phosphatidylinositol (PtdIns) (see Mitchell et al., 2006), are the most versatile and crucial regulators of cytoskeletal dynamics, signal transduction, membrane trafficking and phagosome formation (Simonsen et al., 2001; Yeung et al., 2006). Interestingly, several proteins involved in corpse removal were shown to contain features predicted to bind phospholipids. For example, the mammalian CED-5 ortholog, Dock180 (now known as Dock1), binds PtdIns(3,4,5)P₃ and PtdIns(3,5)P₂ via its DHR1 domain. However, only a limited number of kinases and phosphatases linking PPIn to phagocytosis of apoptotic cells have been described (Leverrier et al., 2003; Liu et al., 2005; Zou et al., 2009).

In this paper, we present a genetic screen that allowed us to identify 13 new mutations that enhance cell corpse clearance in C. elegans. We show that the strongest suppressor mutation is a viable, hypomorphic allele of mtm-1, a PPIn phosphatase identified independently by Zou et al. in an RNAi screen for negative regulators of cell corpse clearance (Zou et al., 2009). Using our allele, we confirm that MTM-1 functions upstream of the CED-5–CED-12 GEF for CED-10, and show that MTM-1 acts in parallel to MIG-2 in the regulation of this GEF complex. We demonstrate that MTM-1, similar to its human homolog myotubularin 1, has biochemical phosphatase activity towards both PtdIns3P and PtdIns(3,5)P₂, and that expression of the human homolog can rescue the C. elegans mutant phenotype. Finally, we show that, in addition to its negative regulation of corpse internalization, MTM-1 function is required for proper maturation of phagosomes containing cell corpses and for recycling of CED-1 back to the plasma membrane. Our findings suggest that during internalization MTM-1 modulates CED-10 activity by controlling the amount of plasma membrane PtdIns(3,5)P₂, which provides docking sites for the recruitment and/or activation of the CED-5–CED-12 GEF complex. Later, during phagosome maturation, MTM-1 might act on the same lipid intermediates to control CED-1 receptor recycling.

C. elegans strains were grown at 20°C as described previously (Brenner, 1974). The wild-type strain used was Bristol N2. The following alleles were used: LGI: mtm-1(op309), mtm-1(ok742), gla-1(op234), ced-12(k149), ced-12(bz187), ced-12(oz167), ced-1(e1735) and ced-1(n1995). LGIII: ced-6(n1813), ced-6(tm1826), ced-6(op360), ced-7(n1892), ced-7(n1996), ced-7(n2690) and unc-119(ed3). LGIV: ced-2(n1994), ced-2(e1752), ced-10(n3246), ced-10(n1993), ced-5(n1812), ced-5(tm1949), ced-5(n2002), ced-3(n2433), ced-3(n717) and ced-3(op149). LGX: mig-2(mu28), mig-2(gm103gf) and dyn-1(n4039). All alleles are described in WormBase (http://www.wormbase.org/).

Integrated arrays [containing unc-119-(+)] used were: opIs110[P_ced-1::yfp::act-5], opIs195[P_mtm-1(short)::yfp::mtm-1(genomic)], opIs220[P_eft-3::dyn-1(genomic)::yfp], opIs222[P_eft-3::gfp::fyve::fyve], opIs265[P_mtm-1(long)::cfp::mtm-1(genomic)], opIs266[P_mtm-1(long)::yfp::mtm-1(op309,genomic)], opIs282[P_ced-1::yfp::rab-5], opIs334 [P_ced-1::yfp::fyve::fyve] nIs96[P_lin-11::gfp; lin-15(+)] and bcIs39[P_lim-7::ced-1::gfp; lin-15(+)]. Extrachromosomal arrays used were: opEx1280[P_mtm-1(long)::gfp::hmtm1(cDNA); unc-119(+)], opEx1303[P_ced-1::yfp::rab-7; unc-119(+)] and opEx1465[P_lst-4::lst-4c(genomic)::yfp; unc-119(+)]. Balancers were: szT1[lon-2(e678)] (I;X).

Staged L4 gla-1(op234); ced-6(n1813) mutants were mutagenized [1 mM n-ethyl-n-nitrosourea (ENU) in M9 for 4 hours] and allowed to recover overnight at 15°C (De Stasio and Dorman, 2001). Worms were transferred to plates to lay eggs for 4-6 hours at 20°C, then burned off; ~3 days later (when F1 progeny had reached adulthood) adults were counted and removed. F2 progeny were incubated for 24-36 hours post-L4/adult molt and stained with Acridine Orange (AO) as described previously (Kinchen et al., 2005; Neukomm et al., 2011).

For two-factor mapping, op309 males were crossed into mapping triples (dpy-5 I; rol-6 II; lon-1 III or unc-5 IV; dpy-11 V; lon-2 X) and F2 homozygote op309 hermaphrodites were tested for segregated markers in the F3 generation (AO staining in a gla-1; ced-6 background). Homozygote dpy-5 mutants were never observed indicating tight linkage to the middle of chromosome I.

For three-factor mapping, op309 was mapped onto the unc-74 dpy-5 genomic segment, closer to the left side, as 11 out of 15 Unc-non-Dpy worms and five out of 15 Dpy-non-Unc worms were suppressed by op309 (corpse numbers in L1 heads in a ced-6 background). This interval contains 203 genes.

For sequencing, mtm-1(op309) animals contain a G to A substitution (introducing an EcoRI site) in exon 4 compared with wild type.

RNAi was performed as described previously (Kamath and Ahringer, 2003; Neukomm et al., 2011). Among all 203 candidate genes tested, AO⁺ germ cell corpses reappeared solely in mtm-1(RNAi)-treated animals.

For embryonic apoptotic cell corpses, mixed embryos were mounted on a 3% agar pad and indicated stages scored for persistent cell corpse in whole bodies using a differential interference contrast (DIC) microscope. Larval L1 head apoptotic cell corpses: Scores were performed as described (Neukomm et al., 2011). For germ cell corpses, >20 either 12, 24 or 36 hour post-L4/adult molts were mounted (3% agar) and anesthetized (5 mM levamisole in M9). Refractive apoptotic germ cell corpses were scored using a DIC microscope and, if applicable, by fluorescent halo scoring using an epifluorescence microscope.

For Pn.aap cell survival, Pn.aap cells were scored as described previously (Reddien et al., 2001).

RNA isolation and cDNA synthesis was performed as described previously (Neukomm et al., 2011).

Transgenic worms were generated as described previously (Kinchen et al., 2005; Neukomm et al., 2011).

4D microscopy, lineaging and cell corpse persistence measurement Fertilized single-cell eggs were isolated, mounted on a coverslide (containing 0.1% poly-l-lysine) in M9 and sealed with vaseline. Using OpenLab Imaging software and a Leica DMR-A2 DIC microscope, 40 z-stack pictures (≈0.8 μm distance) through the embryo were taken every 30 seconds for 7 hours. The first 12 apoptotic cells in the AB lineage were followed using Virtual Wormbase (http://www.biosci.ki.se/groups/tbu/software) using tiff files. Onset of cell death was defined as the time between single-cell embryo (t_1-cell) and cell death (onset of refractivity, t_lentil), i.e. t_lentil – t_1-cell. Corpse persistence was defined as the time between onset of death (t_lentil) and cleared corpse (t_engulfed), i.e. t_engulfed – t_lentil (Hoeppner et al., 2001).

Recombinant GST::MTM-1 proteins (wild-type and G106E) were expressed in E. coli CodonPlus BL21 bacteria, lysed by sonication (lysis buffer: 10 mM Tris, 50 mM NaCl, 50 mM KCl, 10% glycerol, 1 mM DTT, 100 ng/μl lysozyme, PIC, pH 7.0), centrifuged (10,000 g for 10 minutes) and purified with glutathione Sepharose 4B beads (GE Healthcare) in lysis buffer containing 1% Triton for 2 hours. Recombinant proteins were washed (10 mM Tris, 50 mM NaCl, 50 mM KCl, 10% glycerol, 0.1% NP40, 1 mM DTT, pH 7), eluted (100 mM Tris, 150 mM NaCl, 30 mM glutathion, pH 7) and dialysed (10 mM Tris, 50 mM NaCl, 50 mM KCl, 10% glycerol, pH 7).

Phosphatase activity using fluorescent substrate was performed as described previously (Rohde et al., 2009) with slight modifications. Increasing concentrations of recombinant proteins were incubated with 2.5 μg of C6-BODIPY-FL-PtdIns3P or C6-BODIPY-FL-PtdIns(3,5)P₂ (Echelon Bioscience) in 50 mM ammonium acetate, pH 6.0 (30°C, 30 minutes) and the reaction stopped with chloroform/methanol (1/1). Fluorescent lipids (upper phase) were dried (nitrogen), resuspended [methanol/isopropanol/acetic acid (5/5/2)], separated by thin-layer chromatography [chloroform/methanol/acetone/acetic acid/water (7/5/2/2/2)] and visualized under ultraviolet light.

L2 larvae were mounted (2% agarose) in 0.2% NaN₃ in M9. Images were taken with a Zeiss LSM710 confocal microscope and GFP::FYVE::FYVE punctae scored with ImageJ (NIH), n=25.

GST or GST::PH_CED-12 were expressed in BL21(DE3) CodonPlus bacteria (Stratagene) at 4°C overnight, then harvested as described previously (Kinchen et al., 2008). Lipid dot blots (Echelon Bioscience) were incubated with ~1.0 μg/ml of protein overnight, then developed as described previously (Park et al., 2007).

Worms were high-pressure frozen (EM Pact2, Leica Microsystems) using flat specimen carriers (indentation, 1.5 mm×0.2 mm). The hole of the carrier (dedicated for pressure transmission) was filled with 1-hexadecene and the cavity of the carrier with a droplet of PBS. Worms were transferred to the PBS droplet, and the majority of PBS was drawn off (filter paper) leaving the worms in a small PBS volume. 1-Hexadecene was added (on top), the specimen immediately frozen and freeze-substituted in anhydrous acetone containing 2% OsO₄ (Leica EM AFS2). Specimens were kept successively at −90°C, −60°C, and −30°C (8 hours each), then changed at a rate of 30°C per hour to reach room temperature, (RT). After 1 hour, OsO₄ was removed with anhydrous acetone and the specimens gradually embedded in 33% Epon/Araldite (EA) in anhydrous acetone (overnight, 4°C), 66% EA in anhydrous acetone (6 hours, 4°C) and 100% EA (1 hour, RT) prior to polymerization (60°C, 48 hours). Thin sections were stained with aqueous 2% uranyl acetate and Reynolds lead citrate and imaged (Phillips CM 100 TEM, FEI, Netherlands) using a Gatan Orius CCD camera and digital micrograph acquisition software (Gatan, Germany). Quantification was performed as described in Yu et al. (Yu et al., 2006).

Primers are listed in Table S3 in the supplementary material. Plasmids are listed in Table 1. If not otherwise stated, ‘LazyBoy’ plasmids were used as described previously (Neukomm et al., 2011).

In order to identify negative regulators of apoptotic cell corpse clearance, we developed a genetic screen for mutants with increased engulfment activity. The vital dye Acridine Orange (AO) stains internalized apoptotic cells in both Drosophila melanogaster and C. elegans in vivo (see Fig. S1 in the supplementary material) (Abrams et al., 1993; Gumienny et al., 1999; Kinchen et al., 2008). AO staining was abrogated in engulfment deficient mutants (Fig. 1A,B; see Fig. S1A in the supplementary material). We surmised that a partial restoration of engulfment activity would lead to the reappearance of internalized (thus AO⁺) corpses, which can be observed readily under a fluorescence dissecting microscope. After screening 45,000 haploid genomes, we isolated 13 suppressor mutations that restored AO staining significantly (Fig. 1; see Fig. S1 in the supplementary material), suggesting that they might have restored, at least partially, engulfment activity. Consistent with this hypothesis, all isolated mutants also contained partially reduced numbers of persistent apoptotic cell corpses in 4-fold embryos (see Fig. S1C in the supplementary material). Here, we describe our characterization of one of these suppressors, op309.

op309 mutants showed the most pronounced ced-6 suppressor phenotype, based on the reappearance of AO⁺ germ cell corpses and corpse numbers in 4-fold embryos. We therefore looked in more detail at the kinetics of cell corpse clearance during embryonic development. Whereas dying cells were engulfed rapidly in wild-type embryos, removal of cell corpses was delayed in engulfment-deficient ced-6 animals, resulting in the accumulation of persistent cell corpses in late embryos and freshly hatched L1 larvae (Fig. 1E,F). By contrast, fewer numbers of persistent cell corpses were observed in op309; ced-6 double mutants at all stages examined, suggesting that op309 results in reduced persistent cell corpse numbers throughout development.

A reduction in persistent cell corpse numbers could, in principle, be achieved not only through a restoration of engulfment activity but also through reduced apoptosis or a delayed initiation of cell death, as found in ced-8 mutants (Stanfield and Horvitz, 2000). Whereas ~12 cells remained alive in the anterior pharynx of ced-3(lf) mutants, no surviving extra cells were observed in op309 mutants (Metzstein et al., 1998) (see Fig. S2A in the supplementary material). Furthermore, we observed a normal onset of programmed cell death of the first 12 apoptotic cells in the AB lineage (see Fig. S2B in the supplementary material), as assessed by 4D microscopy (Schnabel et al., 1997; Sulston et al., 1983). We therefore conclude that op309 does not interfere with developmental apoptosis.

Next, we asked whether a more efficient cell corpse clearance could explain the reduced ced-6 persistent cell corpse phenotype in op309 mutants. Using 4D microscopy, we found that the first 12 cells undergoing apoptosis in wild-type embryos were usually engulfed within 30 minutes upon onset of death (see Fig. S2C and Table S1 in the supplementary material) (Hoeppner et al., 2001; Yu et al., 2006). In ced-6 mutants, three quarters of the dying cells (27/36 cells) were engulfed with normal kinetics, whereas the remaining quarter either showed a significant delay in engulfment (3/36 cells) or remained unengulfed until the end of the recording (6/36 cells). Strikingly, op309; ced-6 double mutants showed nearly wild-type engulfment kinetics, and not a single corpse remained unengulfed. We thus conclude that op309; ced-6 mutants regained significant engulfment activity, which results in a reduction of persistent cell corpse numbers.

The op309 mutation could also suppress the internalization defect of apoptotic germ cell corpses in ced-6 mutants, albeit more weakly than in the soma: op309; ced-6 double mutants had fewer germ cell corpses than ced-6 single mutants, and conversely show a partial reappearance of corpses within acidified phagosomes, as measured by AO staining (Fig. 1D; data not shown) and various markers of the phagosome maturation pathway (see Fig. S2D in the supplementary material).

Through a combination of genetic mapping, RNA interference and DNA sequencing, we were able to identify mtm-1, the C. elegans homolog of human myotubularin 1, as the candidate gene mutated in op309 mutants. We identified a single point mutation in exon 4 in op309 mutants (GGA → GAA), which changes a conserved glycine to glutamic acid at codon 106 (G106E) (see Figs S3 and S4 in the supplementary material). There is another mtm-1 mutant available, ok742, which, unlike superficially normal op309 mutants, arrest development as L2 larvae, suggesting that mtm-1 function is more strongly impaired in the deletion allele ok742 than in the point mutant op309 (Fig. 3A,B). mtm-1(ok742) failed to complement the op309 suppressor phenotype (Fig. 2B), suggesting that op309 is an allele of mtm-1.

Interestingly, Zou and co-workers independently identified mtm-1 in an RNAi-screen for suppressors of persistent cell corpses in ced-1 mutants (Zou et al., 2009). Using our viable mtm-1(op309) allele, we confirmed the involvement of MTM-1 as a negative regulator of corpse internalization as shown previously by Zou et al. and expanded further the characterization of mtm-1 function in C. elegans, as described below.

The C. elegans genome contains a large number of operons (Blumenthal et al., 2002; Spieth et al., 1993). mtm-1 is the last member of operon 1220 (CEOP1220), which consists of four genes (Fig. 2A). We expressed mtm-1(+) under the control of both a short ‘operon promoter’ as well as a long ‘comprehensive mtm-1 5′-regulatory region’ which was sufficient to rescue the mtm-1 engulfment phenotype (Fig. 2B). Importantly, similar constructs driving MTM-1(G106E) failed to rescue, confirming that the G106E substitution impairs normal function of MTM1 and is likely to be the cause of the op309 phenotype.

The human homolog MTM1 shares 46% identity and 64% similarity with C. elegans MTM-1 at the protein level (see Fig. S4 in the supplementary material). To determine whether the molecular function of MTM-1 is conserved through evolution, we expressed hMTM1 under the control of the long C. elegans mtm-1 promoter (opEx1280). hMTM1 efficiently rescued mtm-1(op309) mutants, indicating that human myotubularin can substitute functionally for C. elegans MTM-1 in the regulation of cell corpse clearance.

In a previous study, a 1.5 kb mtm-1 promoter fragment drove MTM-1 expression in a few head and tail neurons (Xue et al., 2003). We re-addressed this question using our rescuing transgenic lines. Both short and long promoters led to a broad expression of MTM-1 in many tissues (see Movie 1 and Fig. S5A-D in the supplementary material). As previously described, we found that MTM-1 is abundantly expressed at the cell cortex (e.g. on the apical membrane of intestinal cells; see Fig. S5D in the supplementary material), with the rest of the protein being distributed throughout the cytosol (Laporte et al., 2002a). The broad expression of MTM-1 and its subcellular localization suggest other functions beside its role in cell corpse clearance, consistent with the lethal phenotype of the deletion allele mtm-1(ok742).

mtm-1 belongs to a large and disease-associated family of polyphosphoinositide (PPIn) 3-phosphatases, the myotubularins, which act on phosphatidylinositol-3-phosphate (PtdIns3P) and phosphatidylinositol-3,5-diphosphate [PtdIns(3,5)P₂] in mammals (Blondeau et al., 2000; Laporte et al., 2002b; Taylor et al., 2000; Tronchere et al., 2004; Walker et al., 2001). In addition to the phosphatase domain, all myotubularins contain a conserved N-terminal GRAM-PH (glucosyltransferases, Rab-like GTPase activators and myotubularins, pleckstrin-homology) domain, which has been suggested to mediate interaction with PPIn and membranes (Laporte et al., 2001).

Several mammalian members of the myotubularin family contain inactive phosphatase domains (Laporte et al., 2003); we thus asked whether C. elegans MTM-1 possesses phosphatase activity. Bacterially expressed C. elegans MTM-1 protein readily showed 3-phosphate phosphatase activity towards both PtdIns3P and PtdIns(3,5)P₂ in vitro. Interestingly, the phosphatase activity was strongly reduced, but not abrogated in MTM-1(G106E) (Fig. 3C).

To confirm the biochemical activity of MTM-1 in vivo, we visualized levels of one of the MTM-1 PPIn targets, PtdIns3P, by a transgenic strain that broadly expresses a fluorescent 2xFYVE domain in vivo (opIs222[P_eft-3::gfp::fyve::fyve]). Consistent with our in vitro results, op309 mutants and mtm-1(RNAi)-treated animals had significantly higher numbers of PtdIns3P-positive vesicles compared with control strains (Fig. 3D,E).

The residual in vitro phosphatase activity of MTM-1(G106E) suggests that op309 is not a null allele. To test this genetically, we compared the strength of the engulfment defect caused by op309 and by the deletion allele ok742, which completely lacks the phosphatase domain and thus possibly represents a null allele. Interestingly, we did not observe any significant difference between the two mtm-1 mutations in our assay (Fig. 3F). Taken together, our results suggest that C. elegans MTM-1 is a PPIn 3-phosphatase and that mtm-1(op309) is a strong reduction-of-function allele with reduced enzymatic activity.

The results above are surprising insofar as the op309 mutation does not affect the catalytic phosphatase domain, but rather a residue in the GRAM-PH domain. We surmise that the G106E mutation alters the overall conformation of MTM-1 or the oligomerization that was shown to be important for allosteric activation (Schaletzky et al., 2003) and thereby interferes with the enzymatic activity.

The myotubularin family is grouped into six subfamilies, based on catalytic activity as well as domain content. Each subfamily is represented by a single gene in C. elegans: mtm-1, mtm-3, mtm-5, mtm-6, mtm-9 and mtm-12 (Laporte et al., 2003). To determine whether other myotubularin family members might also be involved in the regulation of cell corpse clearance, we tested existing mutants (or RNAi-treated animals) for their ability to suppress ced-6 mutants. MTM-1 was the only family member able to dramatically suppress apoptotic cell corpse persistence (see Fig. S6 in the supplementary material). Thus, we conclude that mtm-1 is likely to be the major myotubularin involved in the regulation of cell corpse clearance.

To determine where mtm-1 acts within the engulfment signaling pathway, we performed an in depth set of double and triple mutant analysis. Both the viable mtm-1(op309) and the lethal mtm-1(ok742) mutations were able to suppress loss-of-function alleles of ced-1, ced-6 and ced-7 (Table 2A), suggesting that mtm-1 acts downstream or in parallel to the ced-1, ced-6 and ced-7 pathway.

By contrast, mtm-1 was unable to suppress the engulfment defect of mutations in ced-2, ced-5 and ced-12 mutant worms, three genes that act in a signaling pathway in parallel to ced-1, ced-6 and ced-7 (Table 2B). These findings suggest that MTM-1 functions in this pathway, probably upstream of ced-2, ced-5 and ced-12. mtm-1 was also able to suppress two different hypomorphic ced-10 alleles: n1993, a mutation in the C-terminal CAAX prenylation motif (affecting localization rather than function), and n3246 (impaired nucleotide binding). However, this ced-10 suppression was ced-5 dependent (ced-10 ced-5 versus mtm-1; ced-10 ced-5), indicating that mtm-1 suppression of ced-10 mutants requires ced-5 function (and that these two mutant proteins can be activated by CED-5 to some degree). Taken together, our results clearly suggest that mtm-1 acts in parallel to ced-7, ced-1 and ced-6, and upstream of the ternary ced-2–ced-5–ced12 GEF complex, which in turn modulates ced-10 activity.

We next tested whether mtm-1 might function at the same step as the GTPase mig-2, which is known to act as a positive regulator of CED-12 (deBakker et al., 2004). mig-2(mu28), an early stop allele, did not revert the ced-6 suppressor phenotype of mtm-1 (Table 3), suggesting that mtm-1 acts downstream or in parallel to mig-2.

We also analyzed the rare gain-of-function allele mig-2(gm103gf), which is equivalent to transforming mutations in codon 12 of Ras. The MIG-2(G16E) mutation prevents GTP hydrolysis and thus renders the protein constitutively active (Forrester and Garriga, 1997; Zipkin et al., 1997). We surmised that this mutation, by constitutively activating the CED-5–CED-12 GEF complex, should, like mtm-1(op309), be able to partially compensate for defects in the ced-1/ced-6/ced-7 signaling pathway. Consistent with this hypothesis, the mig-2(gm103gf) mutation strongly suppressed the engulfment defect of ced-1, ced-6 and ced-7 mutants (Table 3).

Strikingly, we observed that loss of mtm-1 function and gain of mig-2 function could cooperate to generate an even stronger suppression of engulfment defects. This additive effect was observed in ced-1, ced-6 and ced-7 mutant worms. Consistent with its proposed role as a regulator of the bipartite CED-5–CED-12 GEF complex, mig-2(gm103gf) could not suppress ced-5 and ced-12 mutants. By contrast, mig-2(gf) effectively suppressed ced-2 mutants, suggesting that MIG-2 regulates CED-5–CED-12 in a CED-2-independent manner.

Taken together, our results suggest that MIG-2 and MTM-1 act in parallel to each other, one as a positive and the other as a negative regulator of the CED-5–CED-12 GEF complex. Moreover, the differential response of ced-2 mutants to mig-2 and mtm-1 suggest that these two proteins influence CED-5–CED-12 via different molecular processes (Fig. 6A).

How does MTM-1 regulate the CED-5–CED-12 GEF complex? PPIn have been implicated in the recruitment and activation of GEFs and GAPs (Saarikangas et al., 2010). As MTM-1 contains 3-phosphate phosphatase activity towards at least two different PPIn, it is tempting to speculate that MTM-1 influences engulfment signaling by modulating plasma membrane PPIn levels and subsequently CED-5–CED-12 recruitment. Consistent with this hypothesis, previous work has shown that Dock180, the mammalian homolog of CED-5, contains a phospholipid binding domain (DHR1), which interacts in vitro with PtdIns(3,5)P₂ and PtdIns(3,4,5)P₃ and mediates membrane translocation of the Dock180/ELMO GEF complex (Côté et al., 2005). However, we could not detect interaction between the CED-5 DHR1 domain and any phospholipid in vitro (data not shown).

CED-12 contains a pleckstrin homology (PH) domain which is required for the GEF function with CED-5 (Zhou et al., 2001a). Many PH domains have been shown to bind to PPIn (Lemmon, 2008), suggesting that the CED-12 PH domain might also contribute to GEF recruitment to the plasma membrane. To test this hypothesis directly, we cloned and expressed a GST CED-12 PH domain fusion (GST::PH_CED-12), and measured its lipid binding ability and specificity on dot blots (Fig. 4), where PtdIns4P, PtdIns(3,5)P₂ and PtdIns(3,4,5)P₃ were strongly bound by PH_CED-12.

These observations led us to speculate that the PH domain of CED-12 might not only stabilize the nucleotide free CED-5–CED-10 complex (Ravichandran and Lorenz, 2007), but probably also contributes to the phosphoinositide-dependent recruitment or activation of the CED-5–CED-12 GEF complex at phagocytic cups.

We originally focused on mtm-1 as the candidate gene mutated in op309 because mtm-1(RNAi), like the op309 mutation, can suppress the internalization defect of ced-6 mutants, and thereby lead to a reappearance of internalized (or AO⁺) germ cell corpses in the adult germline (see above). Surprisingly, as part of a control experiment, we noticed that knockdown of mtm-1 in wild-type animals can, by itself, also interfere with cell corpse clearance, as mtm-1(RNAi) animals often showed increased numbers of germ cell corpses compared with control strains (Fig. 5A). Importantly, we observed this phenotype in three independent mtm-1(RNAi) clones (see Fig. S7 in the supplementary material), which argues against (but does not exclude) an off-target effect as a cause of this additional germ cell corpse phenotype (Rual et al., 2007).

To determine whether these germ cell corpses are not engulfed, or are engulfed but not degraded, we analyzed the gonads of staged mtm-1(RNAi)-treated adult hermaphrodites using transmission electron microscopy (TEM) (Fig. 5B,C). In wild-type gonads, apoptotic germ cells pinch off the germline syncytium and are rapidly internalized and degraded by the surrounding somatic sheath cells (Gumienny et al., 1999). In ced-1; ced-5 double mutant animals, which are defective in cell corpse recognition and internalization, apoptotic cells accumulate as uninternalized corpses between the syncytial germline and the surrounding sheath cells (Fig. 5C). By contrast, all cell corpses observed in mtm-1(RNAi) animals were readily internalized and found within sheath cell phagosomes.

gla-3 and gla-1 function in the hermaphrodite germline (Kritikou et al., 2006; Lettre et al., 2004). Loss of gla gene function superficially results in a similar germ cell corpse phenotype as observed in mtm-1(RNAi) treated animals: in gla-3 mutants, all germ cell corpse observed localized inside sheath cells (Fig. 5C). To determine whether the increase in internalized germ cell corpses observed in mtm-1(RNAi) is due to either increased germ cell apoptosis or reduced degradation in the somatic gonad, we measured germ cell corpse numbers in animals defective for cell corpse internalization. Whereas gla-1; ced-6 double mutants had significantly more germ cell corpses than either single mutant (many cells died, none could be engulfed), germ cell corpse number in mtm-1(RNAi); ced-6 animals was not increased compared with mtm-1(RNAi) alone (see Table S2 in the supplementary material). We therefore concluded that germ cell apoptosis is not affected, but rather that germ cell corpses fail to be properly degraded in mtm-1(RNAi) treated animals.

To decipher whether the germ cell corpse degradation defect observed in mtm-1(RNAi) animals is due to a loss of MTM-1 function in the dying or the engulfing cell, we repeated the RNAi-mediated mtm-1 knockdown in two specific mutant backgrounds: rrf-1(pk1417), which are defective in somatic RNAi; and mut-7(pk204), in which germline RNAi is abrogated (Ketting et al., 1999; Sijen et al., 2001). As expected, gla-1(RNAi) and ced-3(RNAi) were still effective in somatic RNAi-deficient rrf-1 mutants, but failed to induce any change in the germline-defective mut-7 background. By contrast, mtm-1(RNAi) resulted in increased germ cell corpse numbers only in germline-defective mut-7 animals, implying that mtm-1 function is required in the soma, probably in the gonadal sheaths cells, for effective germ cell corpse degradation.

The results above indicate that mtm-1(RNAi) reduces cell corpse clearance at a stage post corpse internalization. To determine at which stage mtm-1 acts during phagosome maturation, we applied RNAi-mediated knockdown of gla-1 and mtm-1 and compared the recruitment of various fluorescent phagosome maturation reporters (Kinchen et al., 2008; Nieto et al., 2010; Yu, 2008) (see Fig. S8 in the supplementary material). Interestingly, many more phagosomes in mtm-1(RNAi)-treated animals still contained high levels of CED-1, compared with control strains. However, because late-stage markers still efficiently labeled corpses in mtm-1(RNAi) animals, we conclude that loss of mtm-1 does not generally block phagosome maturation, but rather might interfere, directly or indirectly, with CED-1 receptor recycling.

Our findings suggest that MTM-1 is required at least twice during the process of cell corpse clearance. First, MTM-1 negatively regulates cell corpse recognition and internalization, probably by limiting recruitment or activation of the CED-5–CED-12 GEF complex at the phagocytic cup. Following internalization, MTM-1 function is required once more, this time to promote, directly or indirectly, recycling of the CED-1 receptor.

Loss of engulfment activity, besides leading to corpse persistence, also promotes cell survival (Hoeppner et al., 2001; Reddien et al., 2001). Conversely, increased engulfment activity leads to reduced sick cell survival (Neukomm et al., 2011). We thus determined whether loss of mtm-1 function can also lead to increased elimination of cells on the verge of death, by scoring Pn.aap cell survival in weak ced-3 reduction-of-function (rf) mutants showing limited caspase activity (see Fig. S9A in the supplementary material). Loss of mtm-1 function reduced survival of the supernumerary Pn.aap cells by ~50% in ced-3(rf) mutants (see Fig. S9B in the supplementary material). Importantly, this increased killing activity was dependent on a functional engulfment machinery, as loss of mtm-1 had no effect in either ced-5 or ced-6 mutant animals. Our results demonstrate that loss of mtm-1 function, probably by over-activation of the engulfment signaling cascade(s), can promote the removal of cells that are viable but fated to die.

Here, we described our identification and characterization of mtm-1, a PPIn phosphatase that acts at multiple steps to regulate apoptotic cell clearance in C. elegans. First, consistent with results from Zou and colleagues (Zou et al., 2009), who independently identified mtm-1 as a negative regulator of corpse internalization in an RNAi screen, we found that loss of mtm-1 partially restores phagocytic activity in mutants defective in corpse recognition and internalization. Epistatic analyses indicated that MTM-1 controls internalization by negatively regulating the CED-5–CED-12 GEF complex together with CED-2, which promotes GTP-loading of the CED-10 GTPase. Zou et al. described a ced-2 suppression by mtm-1. We did not observe such a suppression in our assay, which could be explained by the different developmental stages used for epistatic analyses (4-fold embryos versus L1 heads, respectively). Expression of human MTM1 partially rescued mtm-1 mutants, suggesting a conserved molecular function for this gene in the regulation of Rac1-dependent cytoskeletal rearrangements. Second, we showed that loss of mtm-1 leads to the accumulation of internalized but undegraded cell corpses. Our observations suggested that MTM-1 is required for proper maturation of phagosomes containing apoptotic corpses, in particular for the recycling of the apoptotic cell corpse receptor CED-1.

Biochemical analysis of the mutant protein revealed that MTM-1, like its mammalian ortholog, has phosphoinositide 3-phosphatase activity towards PtdIns3P and PtdIns(3,5)P₂ in vitro. Interestingly, CED-5 and CED-12, the two subunits of the CED-10 GEF complex, can both bind PtdIns(3,5)P₂.

PtdIns(3,5)P₂ is a low abundance PPIn, commonly present at 0.1% or less of total cellular PPIn (Michell et al., 2006). However, PtdIns(3,5)P₂ is present in all eukaryotic cells so far examined and thus presumably contributes to widely conserved cell function(s). Surprisingly, the membrane binding domain of CED-5, DHR1, binds with approximately the same affinity to PtdIns(3,4,5)P₃ and to PtdIns(3,5)P₂ in vitro (Côté et al., 2005). We similarly found that the PH membrane-binding domain of CED-12 also binds to PtdIns(3,5)P₂ in vitro. It is thus tempting to propose that recruitment or activation of the CED-5–CED-12 GEF complex might be mediated through binding to PtdIns(3,5)P₂ generated on the inner leaflet of the phagocytic cup. Through hydrolysis of PtdIns(3,5)P₂ to PtdIns5P, MTM-1 would antagonize GEF recruitment and/or activation, possibly as part of a negative feedback loop (Fig. 6B). Consistent with this hypothesis, overexpression of dominant active Rac1 in mammalian cells recruits hMTM1 to Rac1-induced membrane ruffles, where hMTM1 might dephosphorylate PtdIns(3,5)P₂, thereby blocking further GEF complex recruitment (Laporte et al., 2002a). As we failed to detect any enrichment of a functional CED-12::GFP reporter around phagocytic cups or early internalized corpses in the adult gonad (data not shown), we suspect that localized activation, rather than physical recruitment, might be the major mode of CED-5–CED-12 regulation by PPIn. Alternatively, CED-5–CED-12 association with the phagocytic cup might simply be too transient to be detected by this method.

Our analysis of an MTM-1 reporter construct revealed a broad MTM-1 expression in C. elegans. We found that the protein is often abundant at the plasma membrane, consistent with the model presented above. However, we also observed significant amounts of MTM-1 in the cytosol, suggesting that the protein might also be acting on other cellular membranes. Indeed, PtdIns3P and PtdIns(3,5)P₂, the two known substrates for MTM-1, are present on the surface of various intermediates of the endosomal pathway (Robinson and Dixon, 2006). Evidence that MTM-1 also hydrolyzes PPIn in the endosomal pathway is provided by the increase in PtdIns3P-positive vesicles in hypodermal cells of mtm-1(op309) and mtm-1(RNAi) animals.

In addition to promoting internalization, we found that knockdown of mtm-1 interfered with phagosome maturation, leading to the accumulation of internalized, but undegraded apoptotic germ cell corpses within phagosomes of the engulfing sheath cell. Consistent with previous observations, not only phosphorylation, but also dephosphorylation, of PPIn by PPIn phosphatases is required for the proper maturation of endosomes and phagosomes, e.g. via the phosphoinositide 3-kinase VPS-34 (Cremona et al., 1999; Kinchen et al., 2008; Zou et al., 2009). However, as we could detect phagosomes at all stages of maturation, loss of mtm-1 is unlikely to block maturation at a specific step. Rather, we suggest, based on the increase in CED-1-positive phagosomes in mtm-1(RNAi) animals, that MTM-1 is required for efficient recycling of CED-1 from the phagosome back to the plasma membrane. Interestingly, loss-of-function mutations in snx-1 and snx-6, two PX-domain containing subunits of the C. elegans retromer complex, led to a similar defect in germ cell corpse degradation and the accumulation of internalized, CED-1-positive phagosomes (Chen et al., 2010). Whether MTM-1 and the retromer complex, which is known to bind to PPIn, act in the same molecular pathway remains to be determined.

Intriguingly, we did not observe any persistence of embryonic cell deaths in mtm-1(RNAi) animals. There are several possible explanations for this observation, ranging from differences in RNAi sensitivity to differential requirements for MTM-1 in phagosome maturation. For example, whereas embryonic cells rarely need to engulf more than one apoptotic cell, the gonadal sheath cells take up and degrade several hundred apoptotic germ cells over the adult life of the animal (Gumienny et al., 1999; Sulston et al., 1983). It is thus possible that the phagosome maturation pathway in sheath cells is more sensitive to interference. Consistent with this hypothesis, animals treated with vps-34(RNAi), in which persistent cell corpses are readily observed in the germline, but fail to be observed in early embryos (Kinchen et al., 2008) (our unpublished results), present experimental evidence for increased RNAi susceptibility in sheath cells compared with the soma.

The myotubularin family of PPIn 3-phosphatases, which dephosphorylate PtdIns3P and PtdIns(3,5)P₂), are associated with a variety of human syndromes (such as peripheral Charcot-Marie-Tooth (CMT) neuropathies with or without associated glaucoma, myotubular myopathy) as well as impaired spermatogenensis and azoospermia in mice (Azzedine et al., 2003; Bolino et al., 2000; Firestein et al., 2002; Houlden et al., 2001; Laporte et al., 2003; Nicot and Laporte, 2008; Senderek et al., 2003). However, the molecular mechanisms causing these phenotypes are not understood. As each myotubularin subfamily is represented by a single member in C. elegans, further study of myotubularin function in nematodes might shed new light into the cellular and molecular roles of this important family of proteins.

We thank the Hengartner lab for comments and discussions on this manuscript and Andres Kaech and Therese Bruggmann for technical support (TEM analyses). This work was supported by grants from the Swiss National Science Foundation, The Ernst Hadorn Foundation and the European Union (FP5 project APOCLEAR) (M.O.H.), the American Heart Association and American Cancer Society (J.M.K.), e-rare and ANR (ANR-07-BLA-0065-01) (A.-S.N., H.T., B.P. and J.F.L.), and by grants from Collège de France, the Association Française contre les Myopathies and Fondation Recherche Médicale (DEQ20071210538) (J.F.L.).