|Home | About | Journals | Submit | Contact Us | Français|
Dynamins are large GTPases that oligomerize along membranes. Dynamin's membrane fission activity is believed to underlie many of its physiological functions in membrane trafficking. Previously, we reported that DYN-1 (Caenorhabditis elegans dynamin) drove the engulfment and degradation of apoptotic cells through promoting the recruitment and fusion of intracellular vesicles to phagocytic cups and phagosomes, an activity distinct from dynamin's well-known membrane fission activity. Here, we have detected the oligomerization of DYN-1 in living C. elegans embryos and identified DYN-1 mutations that abolish DYN-1's oligomerization or GTPase activities. Specifically, abolishing self-assembly destroys DYN-1's association with the surfaces of extending pseudopods and maturing phagosomes, whereas inactivating guanosine triphosphate (GTP) binding blocks the dissociation of DYN-1 from these membranes. Abolishing the self-assembly or GTPase activities of DYN-1 leads to common as well as differential phagosomal maturation defects. Whereas both types of mutations cause delays in the transient enrichment of the RAB-5 GTPase to phagosomal surfaces, only the self-assembly mutation but not GTP binding mutation causes failure in recruiting the RAB-7 GTPase to phagosomal surfaces. We propose that during cell corpse removal, dynamin's self-assembly and GTP hydrolysis activities establish a precise dynamic control of DYN-1's transient association to its target membranes and that this control mechanism underlies the dynamic recruitment of downstream effectors to target membranes.
Animal cells undergoing programmed cell death (apoptosis) are engulfed by other cells through phagocytosis and are degraded inside phagosomes. The rapid removal of apoptotic cells is important for tissue remodeling, prevention of tissue injury, and the suppression of inflammatory and autoimmune responses (Savill and Fadok, 2000 ). During the development of the nematode Caenorhabditis elegans hermaphrodites, 131 somatic cells and ~500 germ cells undergo programmed cell death and are swiftly removed by their neighboring cells (Metzstein et al., 1998 ). Apoptotic cells are easily recognized within living animals under the Nomarski differential interference contrast (DIC) optics as highly refractive, button-like objects that are referred to as “cell corpses” (Sulston and Horvitz, 1977 ; Sulston et al., 1983 ). During the removal process, an engulfing cell produces thin cytoplasmic extensions (pseudopods) to embrace the dying cell, creating a phagocytic cup; the fusion of pseudopods generates a phagosome, inside which the cell corpse is degraded (reviewed in Zhou and Yu, 2008 ). The extension of pseudopods relies on the incorporation of intracellular vesicles to phagocytic cups, which provide both lipid and protein materials necessary for membrane expansion (Touret et al., 2005 ). The degradation of apoptotic cells inside phagosomes relies on the process of phagosome maturation, during which a series of intracellular organelles including early endosomes, late endosomes, and lysosomes fuse with phagosomes, delivering digestive enzymes and promoting the acidification of phagosomal lumen (reviewed in Kinchen and Ravichandran, 2008 ; Zhou and Yu, 2008 ).
Dynamin is a conserved large GTPase that plays pivotal roles in multiple membrane-related cellular and developmental processes (Hinshaw, 2000 ; Praefcke and McMahon, 2004 ). Many lines of evidence suggest that during endocytosis, dynamin acts as a mechanochemical enzyme that drives membrane fission and the release of endocytic vesicles (Hinshaw, 2000 ; Praefcke and McMahon, 2004 ; Bashkirov et al., 2008 ; Pucadyil and Schmid, 2008 , and the references therein). In addition, dynamin and dynamin-related proteins also act to promote membrane fusion (Peters et al., 2004 ; Hoppins and Nunnari, 2009 ). Dynamin has also been proposed to act as a molecular switch (Sever et al., 1999 ) or cytoskeleton regulator (Orth and McNiven, 2003 ). In a genetic screen for mutants that are defective in both embryonic development and cell corpse removal, we identified fourteen recessive, loss-of-function alleles of dyn-1, the C. elegans dynamin gene (Yu et al., 2006 ). Our subsequent molecular and cell biological characterizations provided direct evidence that dynamin plays critical roles in phagocytosis and phagosome maturation (Yu et al., 2006 , 2008 ).
Classical dynamins, including C. elegans DYN-1, are made up of five domains: an N terminus GTPase domain, a middle domain, a pleckstrin homology (PH) domain, a GTPase effector domain (GED), and a C-terminal proline-rich domain (PRD; Figure 1A). The hydrolysis of GTP is essential for driving membrane fission (reviewed in Hinshaw, 2000 ). The PH domain targets dynamin to negatively charged lipid membranes (Salim et al., 1996 ; Zheng et al., 1996 ). The GED domain was proposed to act as an intramolecular GTPase-activating protein (Sever et al., 1999 ; Narayanan et al., 2005 ). The PRD mediates interaction with various Src homology 3-containing proteins that act as functional partners (Schmid et al., 1998 ; Dawson et al., 2006 ). The middle domain is a highly conserved domain unique for the dynamin family. This report studies the functions of the middle and GTPase domains in C. elegans DYN-1.
The dyn-1 mutant alleles that we isolated can be separated into two classes: those that bear missense mutations in either the GTPase domain (class I) or the middle domain (class II; Figure 1A). By characterizing GTPase domain mutants, we found that DYN-1 facilitates pseudopod extension and phagosome maturation by promoting the recruitment and fusion of endosomes to phagocytic cups as well as the recruitment and fusion of both endosomes and lysosomes to maturing phagosomes in a GTP-dependent manner (Yu et al., 2006 , 2008 ). Interestingly, this apoptotic cell removal activity of DYN-1 is distinct from its activity in endocytosis, demonstrated by particular dyn-1 mutant alleles that impair only one but not the other activity (Yu et al., 2006 ). Similar to DYN-1, human dynamin 2 was proposed to promote phagocytosis through regulating membrane expansion rather than membrane fission (Gold et al., 1999 ; Di et al., 2003 ). The function of dynamin in apoptotic cell removal thus suggests a less-characterized activity that promotes membrane fusion, directly or indirectly (Gold et al., 1999 ; Di et al., 2003 ; Yu et al., 2006 ). Specifically, we have found that DYN-1 acts as a mediator in the signaling pathways leading to phagosome maturation—it promotes the recruitment of small GTPase RAB-7, a membrane tethering factor, to phagosomal surfaces, and the synthesis of phosphatidylinositol-3-phosphate [PI(3)P] on phagosomal membranes (Yu et al., 2008 ). During the removal of apoptotic cells, RAB-7 is essential for the recruitment and fusion of lysosomes to phagosomes (Yu et al., 2008 ). PI(3)P is another important phagosome maturation molecule acting by attracting downstream effectors to the surface of phagosomes (Vieira et al., 2002 ). Consistent with this signal transduction role, C. elegans DYN-1 and VPS-34, a class III phosphatidylinositol (PI) 3-kinase known to convert PI to PI(3)P, were detected to interact when overexpressed in mammalian cell culture (Kinchen et al., 2008 ).
A prerequisite for the proper function of DYN-1 is its selective association with its target membranes. The association of DYN-1 to extending pseudopods and nascent phagosomes is necessary for its function in the removal of apoptotic cells (Yu et al., 2006 ). The recruitment of DYN-1 to pseudopods and nascent phagosomes is dependent on the phagocytic receptor CED-1 and its adaptor protein CED-6 (Yu et al., 2006 ). Lack of DYN-1 enrichment on phagocytic cups or nascent phagosomes, as a consequence of cell corpse removal defective (ced)-1 or ced-6 mutations, causes severe defects in engulfment and degradation of cell corpses (Yu et al., 2006 , 2008 ). However, how the dynamic association of DYN-1 to its target membranes is regulated remains poorly understood.
Dynamin monomers assemble into helical polymers in vitro, either in buffers with low ionic strength or along negatively charged lipids at physiological ionic strength (Hinshaw and Schmid, 1995 ; Sweitzer and Hinshaw, 1998 ; Stowell et al., 1999 ; Danino et al., 2004 ). Recently, point mutations of residues in the middle domain were found to impair dynamin self-assembly in vitro (Ramachandran et al., 2007 ). Self-assembly modulates dynamin activity in vitro by enhancing its affinity to lipids and its GTP hydrolysis activity (Warnock et al., 1996 ; Kavran et al., 1998 ; Klein et al., 1998 ; Takei et al., 1998 ).
Despite its extensive characterization in vitro, whether dynamin undergoes self-assembly along specific membrane regions under physiological condition has not been well demonstrated. Dynamin collar around membrane necks was only observed when its GTP hydrolysis activity was inhibited (Takei et al., 1995 ). It also remains to be established whether dynamin self-assembly is important for its in vivo function. In addition, given the distinct biochemical properties of dynamin's self-assembly and GTPase activities, it remains to be determined whether these two activities affect similar or different aspects of dynamin's in vivo function.
To answer the above-mentioned questions, we established two novel assays with which we detected DYN-1's oligomerization in vivo. We further found that high-order assembly is essential for DYN-1's function in the removal of apoptotic cells, specifically, for its association with phagosomal membranes. In addition, we found that the dissociation of DYN-1 from phagosomal membranes is GTP dependent. Together, the self-assembly and GTP hydrolysis activities establish a regulatory mechanism for the dynamic association of dynamin with its target membranes, through which dynamin regulates the phagosome association of downstream effectors.
C. elegans strains were grown at 20°C as described previously (Brenner, 1974 ). The N2 Bristol strain was used as the reference wild-type strain. Mutations used are described in Riddle et al. (1997) , except when noted otherwise: LGV, unc-76(e911); LGX, dyn-1(en9), dyn-1(en40) (Yu et al., 2006 ). Germline transformation was performed as described previously (Jin, 1999 ), in strains carrying an additional homozygous unc-76(e911) mutation, using a plasmid carrying the wild-type unc-76 genomic DNA (Bloom and Horvitz, 1997 ) as a coinjection marker. We used a previously established strategy to deplete the maternal dyn-1 product from homozygous dyn-1 mutant embryos (Yu et al., 2006 ). The homozygous mutants were maintained via an extrachromosomal array containing the dyn-1(+) genomic DNA. As a result of the rescuing activity of the transgene, the transgenic animals were normal and fertile; however, gene expression is usually repressed in the germline due to germline silencing of repetitive extrachromosomal arrays (Kelly et al., 1997 ; Yu et al., 2006 ). Among the progeny of the transgenic animals, we identified those embryos that lost the extrachromosomal array, which presumably inherited reduced amount of maternal dyn-1(+) product and do not produced zygotic dyn-1(+) product, by tracking embryos that did not carry a monomeric red fluorescent protein coinjection marker.
dyn-1 cDNA (Yu et al., 2006 ) without the stop codon was inserted into two vectors that contain heat-shock promoters Phsp-16/2 and Phsp-16/41 (Stringham et al., 1992 ) as gfp fusions to generate Phsp-16/2 dyn-1::gfp (pBZ51) and Phsp-16/41 dyn-1::gfp (pBZ52), respectively. To introduce the en9(I401F) and n4039(G204E) mutations, DNA fragments that contain the corresponding mutations were polymerase chain reaction (PCR)-amplified from lysates generated from homozygous mutant eggs. The SacI-EcoRV fragment that contained the en9(I401F) mutation replaced the wild-type fragment, generating Phsp-16/2 dyn-1(I401F)::gfp and Phsp-16/41 dyn-1(I401F)::gfp. The ClaI-NruI fragment that contained the n4039(G204E) mutation replaced the wild-type fragment, generating Phsp-16/2 dyn-1(G204E)::gfp and Phsp-16/41 dyn-1(G204E)::gfp. To introduce the en40(G40E) and R402A mutations, these mutations were first introduced individually into pBZ51 and pBZ52 by using a site-directed mutagenesis kit (Stratagene, La Jolla, CA) to generate Phsp-16/2 dyn-1(G40E)::gfp and Phsp-16/41 dyn-1(G40E)::gfp and Phsp-16/2 dyn-1(R402A)::gfp and Phsp-16/41 dyn-1(R402A)::gfp. The I401F and R402A mutations were further introduced into Phsp dyn-1(G40E)::gfp or Phsp dyn-1(G204E)::gfp plasmids to generate Phsp dyn-1(G40E I401F)::gfp, Phsp dyn-1(G204E I401F)::gfp, and Phsp dyn-1(G40E R402A)::gfp. DNA fragments that contain various mutations were also introduced into Pced-1 dyn-1::gfp (Yu et al., 2006 ) to generate Pced-1 dyn-1(G40E)::gfp, Pced-1 dyn-1(G204E)::gfp, and Pced-1 dyn-1(I401F)::gfp. To generate domain truncation mutants, DNA fragments that encode the corresponding DYN-1 protein fragments were cloned into Pced-1 dyn-1::gfp to make Pced-1 dyn-1(GTPase)::gfp, Pced-1 dyn-1(ΔPRD)::gfp, Pced-1 dyn-1(ΔGEDΔPRD)::gfp, Pced-1 dyn-1(ΔMD)::gfp, and Pced-1 dyn-1(ΔPH)::gfp. To construct plasmids for Bimolecular Fluorescence Complementation (BiFC) assays, dyn-1 cDNA were cloned into pCE-BiFC-VN173 and pCE-BiFC-VC155, a pair of BiFC vectors (Shyu et al., 2008 ), under the control of Phsp-16/41, fused, at its C terminus, to cDNAs encoding truncated Venus fragments VN173 (Venus aa 1-173) and VC155 (Venus aa 155-238), respectively, to generate pDYN-1::VN173 and pDYN-1::VC155. The I401F mutation was introduced into the above-mentioned plasmids, respectively, to generate pDYN-1(I401F)::VN173 and pDYN-1(I401F)::VC155. The Middle domain of DYN-1 was deleted from pDYN-1::VN173 and pDYN-1::VC155, respectively, to generate pDYN-1(ΔMD)::VN173 and pDYN-1(ΔMD)::VC155. The PhspbFos::VN173 was provided by C.-D. Hu (Purdue University) (Shyu et al., 2008 ). Pced-1gfp::rab-5 (pBZ141) was generated by cloning the rab-5 cDNA, which was PCR amplified from a mixed stage C. elegans cDNA library (Z. Zhou and H. R. Horvitz, unpublished data) fusing with gfp cDNA that lacked the stop codon to its N terminus under the control of Pced-1.
Wild-type and mutant forms of C. elegans dyn-1 cDNA were cloned into pFastBac1 vector (Invitrogen, Carlsbad, CA), and recombinant baculoviruses were generated using the Bac-to-Bac baculovirus expression system (Invitrogen). DYN-1 were expressed in SF21 cells and purified using ammonium sulfate precipitation and Q HP chromatography as described previously (Warnock et al., 1996 ). The fractions containing DYN-1 were detected by Western blot using anti-rat dynamin 2 antibody (BD Biosciences, San Jose, CA). The purity of the recombinant dynamin is >95% as judged by Coomassie Blue staining. Aliquots with 10% glycerol were quickly frozen in liquid nitrogen and stored at −80°C. Rat dynamin was purified from frozen rat brains (a gift from D. Sackett, National Institutes of Health) following previously described protocol (Stowell et al., 1999 ).
Polyclonal rabbit antisera against DYN-1 were generated using Escherichia coli-expressed recombinant His-tagged full-length DYN-1 as an antigen. Antibodies were affinity purified using the original antigen.
Wild-type strain N2 and different transgenic worms were harvested and boiled in SDS sample buffer (10% sucrose, 2% SDS, 0.0625 M Tis-HCl, pH 6.8, 0.0004% bromophenol blue, and 5% β-mercaptoethanol). Proteins in the supernatants were collected after centrifuge, separated by SDS-polyacrylamide gel electrophoresis (PAGE), and transferred to polyvinylidene difluoride membrane. For detecting DYN-1 and different mutant forms of DYN-1, anti-DYN-1 polyclonal antibody was incubated at a 1:2000 dilution. For detecting green fluorescent protein (GFP)-fused transgenic proteins, monoclonal anti-GFP antibody (Epitomics, Burlingame, CA) was used at a 1:2000 dilution. As the loading control, the same membranes were probed with polyclonal anti-beta-tubulin antibodies (Epitomics).
A filter assay was used to examine the GTP binding ability of wild-type and mutant dynamin (Song et al., 2004 ). DYN-1 proteins (0.25 μM) purified from SF21 insect cells (see above) were incubated with 2.5 μM guanosine 5′-O-(3-thio)triphosphate (GTPγS) containing 0.675 μCi/ml [35S]GTPγS (GE Healthcare, Little Chalfont, Buckinghamshire, United Kingdom) in GTP binding buffer (20 mM HEPES, pH 8.0, 25 mM NaCl, 0.1 mM EGTA, 1 mM MgCl2, and 1 mM dithiothreitol) and incubated at room temperature for 20 min in a volume of 60 μl. Each reaction was applied equally to two prewet nitrocellulose filters (Millipore, Billerica, MA) in a filter dot blot apparatus under vacuum (1225 sampling manifold; Millipore) and were subsequently washed with 5 ml of cold GTP binding buffer three times under vacuum. The radioactivity on the filters was measured by liquid scintillation counting. The amount of GTP bound to each sample was measured as cpm and was corrected by subtracting the cpm reading of a no-protein control. Values for each mutant DYN-1 protein were normalized to that of wild-type DYN-1.
Lipids (phosphatidylserine or total brain lipid; Avanti Polar Lipids, Alabaster, AL) solubilized in chloroform were dried in a glass tube by passing nitrogen gas and kept under vacuum overnight. Hydration of the dry lipid film was accomplished by addition of working buffer (HEPES buffer: 20 mM HEPES-NaOH, pH 7.2, 1 mM MgCl2, 2 mM EGTA, and 150 mM NaCl) to the tube to a final concentration of 1 mg/ml, followed by subsequent agitation.
To achieve a homogenous distribution of vesicles, the lipid solution was extruded using a Mini-Extruder (Avanti Polar Lipids) through a polycarbonate membrane (GE Osmonics, Minnetonka, MN). Extrusion through filters with 0.2-μm pores for 13 times yields large, unilamellar vesicles (LUVs) with a mean diameter of 220–240 nm.
Lipid tubes were generated by incubating dynamin samples (0.25 mg/ml) with 200 μM liposomes (0.2 mg/ml) for 2.5 h at room temperature. To test the effect of GTP on the DYN-1-lipid tubes, a drop of dynamin-lipid sample was placed onto the transmission electron microscopy (TEM) grid, GTP was added in excess (1 mM), followed by an immediate quenching to limit the contact time between the protein complexes and GTP to only 7–14 s.
To ensure microstructural preservation during specimen preparation, vitrified specimens were prepared in the controlled environment vitrification system, a closed chamber enabling sample preparation at well controlled temperatures and at 100% relative humidity. Experiments were done at 25°C. Specimens were prepared by placing a 5- to 8-μl drop of the sample onto a perforated carbon film supported on a TEM copper grid held by tweezers. The drop was then thinned (blotted) by a piece of filter paper wrapped on a thin metal strip to produce a thin specimen film. After blotting the specimen was plunged into liquid ethane at its freezing point (−183°C) and vitrified. The vitrified specimens were stored under liquid nitrogen (−196°C) until observation. In some experiments where contact times between GTP and dynamin-lipid tubes of only few seconds were required, on-the-grid processing was applied (Konikoff et al., 2000 ; Danino et al., 2004 ), by adding GTP to the sample on the grid (to a final concentration of 1 mM), followed by immediate blotting and vitrification.
Specimens were examined in a Philips CM120 TEM operated at an accelerating voltage of 120 kV, with an Oxford CT3500 cryo-specimen holder that maintained the vitrified specimens below −175°C. Images at nominal magnifications up to 175K times were recorded digitally by a MultiScan 791 charge-coupled device camera (Gatan, Pleasanton, CA) using the Digital Micrograph 3.4 software package, under low-dose operation to minimize electron beam radiation damage (Danino et al., 2001 ).
GTPase activity was assayed as described previously (Collins and Vallee, 1986 ; Szafer et al., 2001 ). with some modifications. Two to 5 g of activated charcoal (Sigma-Aldrich, St. Louis, MO) was weighed and washed three times in distilled water followed by one time of wash in 0.1 M NaH2PO4. Charcoal was divided into Eppendorf tubes containing 600 μl of 0.1 M NaH2PO4, so that each tube contained 5–10% charcoal (wt/vol). 10 mM [γ-32P]GTP (6000 Ci/mmol; PerkinElmer Life and Analytical Sciences, Boston, MA) was diluted in 10 mM GTP (Sigma-Aldrich) in a 1:20–40 ratio. [γ-32P]GTP/GTP was added to dynamin or dynamin-decorated lipid tubes in buffer containing 150 mM NaCl to a final concentration of 1 mM and mixed well. Immediately after GTP addition and at each time point measured, 3 μl from the reaction tube was taken into an Eppendorf containing charcoal, 0.1 M NaH2PO4 to adsorb organic material, and the reaction was stopped by incubation on ice. Samples were centrifuged in an Eppendorf microcentrifuge for 5 min at maximal speed. Radiation due to free 32P in the supernatant was assayed by scintillation counting to determine the amount of Pi released during the reaction. Maximum radiation was measured by adding 1 mM GTPγ32/GTP to 750 μl of 0.1 M NaH2PO4 in the absence of charcoal. GTPase activity was represented as the percentage of GTP hydrolyzed, which was calculated as the ratio between the cpm measured at each time point and the maximal radiation.
The defect in the removal of cell corpses was measured as the number of cell corpses remaining in the body of late fourfold stage embryos using Nomarski DIC microscopy, as described previously (Lu et al., 2009 ). Late fourfold stage embryos whose bodies made at least three turns and contained a fully developed pharyngeal grinder were scored (Yu et al., 2006 ).
Adult hermaphrodites that were 24 h post-larval stage (L)4 were allowed to lay eggs for 1 h at 20°C before removed from the plate, which was immediately placed at 30°C for 1 h. The plate was then incubated at 25°C until the embryos on the plate were scored for the number of cell corpses under the DIC optics 8 h later.
BiFC assays were performed using the protocol described in Hiatt et al. (2008) and Shyu et al. (2008) . Transgenic animals carrying pairs of BiFC constructs were raised at 20°C. Embryos on the plate were heat-shocked at 33°C for 1 h. Fluorescent images were captured under the DeltaVision microscope (Applied Precision, Issaquah, WA) 3 h afterward. Serial z-sections in 1-μm interval spanning the entire embryos were captured for each sample. Images were deconvolved and analyzed using SoftWorX software (Applied Precision).
An IX70 (Olympus, Tokyo, Japan)-DeltaVision (Applied Precision) microscope equipped with a CoolSNAP digital camera (Photometrics, Tucson, AZ), was used to capture fluorescence images, and SoftWorX software (Applied Precision) 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 90–120 min in 1- or 2-min intervals. At each time point, 14 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.
Fluorescence signal intensity was measured and images were analyzed using the ImageJ software (National Institutes of Health, Bethesda, MD) Lu et al., 2009 ). To calculate the ratio of signal intensity on the surface of phagosomes versus that in the engulfing cell cytosol, the phagosomal surface and cytoplasm were individually defined using the “Free Hand” tool, and the mean value of each selected area was used in the ratio calculation. The enrichment of fluorescence signal on the phagosomal surface was considered significant when the ratio was >1.2.
Four dyn-1 alleles, en40, en34, n4335, and n4098, represent changes of three absolutely conserved residues within the G1 motif known for binding the phosphates of guanine nucleotides, and display strong Ced phenotype (Figure 1B; Yu et al., 2006 ). Mutations n4334(G62E) and en15(P146L) are adjacent to Switch I and inside the Switch II regions, two regions that undergo major conformational changes between the GDP and GTP-bound states, respectively, and presumably also affect GTP binding or the conformational changes induced by the binding of GTP (Figure 1B). In addition, mutations n4332(P178L), en7(G204R), n4039(G204E), n4333, en37, en41(G235E), and en12(G235R) are located in the β4, β5, and β6 strands of the conserved G domain core motif closely resembling that of Ras GTPase (Figure 1B; Niemann et al., 2001 ). This group of mutations may induce structural changes of the corresponding β-strands, and subsequently affect GTP binding, hydrolysis, or the interaction of the GTPase domain with other domains. The clustering of such a large number of mutant alleles in the important motifs inside the GTPase domain underscores the importance of GTP hydrolysis for DYN-1 function during cell corpse removal.
To examine the importance of other domains of DYN-1 for the removal of apoptotic cells, we generated extrachromosomal arrays that expressed dyn-1 cDNA constructs bearing various deletions as gfp fusions under the control of the ced-1 promoter in the dyn-1(n4039) mutant worms and scored their activities in rescuing the Ced phenotype (Figure 2). In comparison to the full-length dyn-1 cDNA, which displayed 87% rescuing activity, deletions of the middle domain or the PH domain alone resulted in complete loss of the rescuing activity (Figure 2), yet did not result in significant changes in the level of the GFP fusion proteins (Supplemental Figure S1, A and C), indicating that the middle and PH domains are both essential for DYN-1's function in cell corpse removal. The deletion of the GED and PRD domains together also led to a complete loss of rescuing activity (Figure 2). However, this loss could be caused by complex reasons, which include an apparent greatly reduced protein level (Supplemental Figure S1C).
In contrast, removing the PRD domain alone only led to a partial loss of the rescuing activity (Figure 2 and Supplemental Figure S1A), indicating that this domain does not play an essential role in cell corpse removal. We focus on studying the functions of the GTPase and the middle domains in this report.
To determine what aspects of DYN-1's biochemical activities are affected by the mutations in the GTPase domain or the middle domain, we expressed wild-type and mutant forms of dyn-1 cDNAs in an insect cell expression system and purified DYN-1 proteins (Figure 3A) for all the in vitro assays described below. We performed GTP binding assays by measuring the incorporation of [35S]GTPγS into DYN-1. In addition to the representative GTPase domain mutations that we identified (Figure 1B), we also analyzed the effect of a K46A mutation, which is equivalent to K44A, a GTP binding-defective mutation in human dynamin 2 (Figure 1B). We found that the G40E and K46A mutations almost abolished the GTP binding activity of DYN-1, whereas the mutations S43F and G204E each reduces the GTP binding activity of DYN-1 to ~20% of the wild-type level (Figure 3A). In contrast, the I401F mutation did not affect the GTP binding activity of DYN-1 (Figure 3A). These results, together with the strong Ced and embryonic lethal (Emb) phenotypes observed from the corresponding homozygous mutants (see Figure 8A; Yu et al., 2006 ), indicate that GTP binding is necessary yet not sufficient for DYN-1's in vivo function.
We first examined the GTP hydrolysis activity of DYN-1 mutant proteins in a buffer of physiological ionic strength and inhibitory of dynamin self-assembly (see Materials and Methods) (Warnock et al., 1996 ). By measuring its GTP hydrolysis activity over time, we found that C. elegans DYN-1 displayed a low basal-level GTP hydrolysis activity as rat dynamin did (Figure 3B). As expected for mutant DYN-1 molecules that are severely defective for GTP binding, DYN-1(G40E) and DYN-1(K46A) displayed greatly reduced GTP hydrolysis activities in comparison with wild-type DYN-1 (Figure 3C). In contrast, the basal-level GTP hydrolysis activity of DYN-1(I401F) was similar to that of wild-type DYN-1, indicating that the intrinsic GTPase function of DYN-1 was not altered by this mutation (Figure 3C).
In vitro, dynamin oligomerization (or self-assembly), which is inducible by the presence of lipids, enhances the GTP hydrolysis activity of mammalian dynamins (Warnock et al., 1996 ). In our GTP hydrolysis assay, with the addition of negatively charged lipids, a drastic increase of the GTPase activity of wild-type DYN-1 was detected, at the same rate and level as detected from rat dynamin (Figure 3B). Thus, the enhancement of GTPase activity by lipid-stimulated assembly is conserved in C. elegans DYN-1. In contrast, we found that the presence of lipid did not further enhance the intrinsic GTPase activity of DYN-1(I401F) (Figure 3C). This lack of enhancement strongly suggests that the DYN-1(I401F) protein is deficient in self-assembly and/or lipid attachment.
We directly examined the effect of DYN-1 mutations toward the self-assembly of DYN-1 over lipids by using negative-stain TEM (data not shown) as well as cryogenic-TEM (Figure 4; see Materials and Methods). As reported previously for human dynamin (Sweitzer and Hinshaw, 1998 ; Danino et al., 2004 ), we observed that wild-type DYN-1 assembled onto lipid surfaces and transformed liposomes into long, 50-nm-diameter tubules decorated with a helical assembly of DYN-1 (Figure 4A). DYN-1(G40E) and DYN-1(K46A) behaved exactly like wild-type DYN-1 when incubated with liposomes (Figure 4B; data not shown). This result indicates that DYN-1's self-assembly activity is independent of its GTP binding or hydrolysis. In contrast, DYN-1(I401F) failed to assemble along lipid surface or to transform liposomes into tubules (Figure 4C), demonstrating that the I401F mutation specifically disrupts DYN-1's self-assembly activity and its stable association with lipids.
Previous reports showed that in cryo-electron microscopy (EM) assays similar to the assays performed here, decorated human dynamin-lipid tubes undergo conformational changes upon GTP hydrolysis, resulting in the constriction of the underlying membrane (Danino et al., 2004 ). We observed, within 7 s of the addition of GTP to lipid tubes decorated with wild-type DYN-1, an immediate constriction of lipid tubes in the radial dimension and of the helical pattern in the axial dimension, as was reported for human dynamin (Danino et al., 2004 ), and the subsequent disassembly of DYN-1, as was also observed with human dynamin (Danino, unpublished data). In consequence, bare liposomes without DYN-1 decoration became abundant (Figure 4D, black arrowheads). Twisted and constricted DYN-1-lipid complexes, which are intermediates of this dissociation process, also appeared (Figure 4D, white arrowheads). Over time, undecorated vesicles became the major population detected under cryo-EM (data not shown). In contrast, no morphological changes took place in DYN-1(G40E)- or DYN-1(K46A)-coated lipid tubes upon GTP treatment (Figure 4E; data not shown). These mutant DYN-1 proteins remained assembled on lipids as helices, maintaining the diameter of lipid tubules of 50 nm. These results support the notion that GTP hydrolysis of DYN-1 leads to the constriction of lipid tubes and dissociation of dynamin from lipid surfaces (Danino, unpublished data) (Danino et al., 2004 ).
The above-mentioned results suggest that DYN-1 undergoes self-assembly in the presence of lipids and that DYN-1 self-assembly is necessary for its enhanced GTP hydrolysis activity. Consistently, we detected a strong interaction between wild-type full-length DYN-1 in a yeast two-hybrid assay (Supplemental Figure S2). To directly visualize the interaction between DYN-1 monomers in living C. elegans, we established a BiFC assay. In this assay, GFP or its variant is split into two nonfluorescent moieties, each of which fused to a protein of interest, and the fusion proteins are coexpressed simultaneously. If the binding of the two potential interacting partners should bring two parts of GFP to a sufficiently close distance, a functional fluorophore would form, resulting in the production of fluorescence signal (Figure 5A; (Hu et al., 2002 ; Shyu et al., 2008 ). DYN-1, DYN-1(I401F), and DYN-1(ΔMD) (middle domain deletion) were tagged separately with the N-terminal (residues 1-173, VN173) or C-terminal fragments (residues 155-238, VC155) of Venus, a modified form of GFP (Nagai et al., 2002 ; Hiatt et al., 2008 ; Shyu et al., 2008 ). The BiFC complex formation is essentially irreversible: once formed, a GFP complex remains intact even if the interacting proteins dissociate from each other (Hu et al., 2002 ). To avoid the potentially detrimental cellular and developmental effects of an irreversible BiFC complex, the fusion cDNAs were engineered under the control of heat-shock promoter Phsp-16/41, which is silent at the permissive temperature for worm growing (20°C) but inducible by heat-shock (33°C), with the highest-level expression detected in intestinal cells in embryos (Stringham et al., 1992 ; Shyu et al., 2008 ). After heat-shock treatment (see Materials and Methods), we observed two distinct types of fluorescence signals in embryos and larvae expressing both DYN-1::VN173 and DYN-1::VC155 (Figure 5, B and D): I) bright, continuous fluorescent signal along the apical surface of intestinal cells (Figure 5, Ba and Da, arrowheads), and II) fluorescent puncta localized to the surface of embryos and larvae (Figure 5 Ba and Da, arrows). We interpret type II signal as nonspecific noise signal due to the fact that it was also observed in negative control embryos expressing bFos::VN173 and DYN-1::VC155 (Figure 5Be). Similar nonspecific signal was also seen in embryos expressing DYN-1(I401F) or DYN-1(ΔMD) BiFC constructs (Figure 5B, b–d). Such nonspecific signal may be derived from nonspecific aggregation of BiFC fusion proteins, similar to that reported by Kerppola (2006) and Shyu et al. (2008) .
In contrast, type I signal is specifically from the interaction between DYN-1 molecules. In embryos expressing wild-type DYN-1 BiFC constructs, this signal was observed throughout the entire intestinal track, extending from the posterior of the second pharyngeal bulb to the anus (Figure 5B, a and g). Yet, it was absent from the negative-control embryos (Figure 5Be). This specific BiFC signal representing the interaction of DYN-1 monomers along the apical surface of intestinal cells is consistent with the essential role of DYN-1 for endocytosis (Grant and Hirsh, 1999 ). In contrast, the heat-shock–induced expression of DYN-1(I401F)::VN173 and DYN-1(I401F)::VC155, although produced fusion proteins of predicted sizes and of levels equivalent to that of wild-type fusion proteins (Supplemental Figure S3), resulted in primarily no or only partial fluorescent lines along the intestinal track in wild-type embryos or larvae (Figure 5B, b and h; C; and D, c and d). The same mutant DYN-1 BiFC constructs, when expressed in dyn-1(en9) mutant embryos where no endogenous functional DYN-1 was available, created an even more prominent defect in BiFC signals—the fluorescent line along the intestinal track was absent from all embryos (Figure 5B, c and i, and C). These results indicate that the I401F mutation severely impairs DYN-1's self-assembly in vivo. Like that of DYN-1(I401F), the DYN-1(ΔMD) fusion proteins (Supplemental Figure S3) did not generate any specific BiFC signal on the apical surface of intestinal track (Figure 5B, d and j), further indicating that the middle domain is needed for DYN-1's in vivo self-interaction.
Based on the BiFC assay results, and on the observation that the GTPase domain mutations of DYN-1 induces dominant-negative effects in the removal of cell corpses (see the next section), we further developed a genetic assay that demonstrated the essential function of DYN-1 oligomerization in vivo (see the next section).
We found that the transgenic expression of dyn-1 cDNA bearing individual mutations G40E, K46A, and G204E defective in GTP binding as a gfp fusion, under the control of a pair of heat-shock promoters Phsp-16/2 an Phsp-16/41 (collectively referred to as Phsp), caused dominant-negative Ced and Emb phenotypes in an otherwise wild-type background (Figure 6; data not shown). After heat-shock treatment (see Materials and Methods), ~20 cell corpses in late fourfold stage embryos accumulate in embryos carrying these transgenes, when there are nearly no detectable cell corpses in wild-type embryos carrying no transgene (Figure 6). In two transgenic lines expressing DYN-1(G204E), 40 and 80% transgenic animals die in embryonic stage after heat-shock. Both the Ced and Emb phenotypes are similar to that observed from homozygous dyn-1 loss-of-function mutants (Yu et al., 2006 ), indicating that these mutations not only inactivate the transgenic DYN-1::GFP molecules in cis but also severely interfere with the function of endogenous wild-type DYN-1 in trans.
In contrast, the heat-shock induced expression of DYN-1(I401F) in wild-type embryos, although leading to the production of an amount of GFP fusion protein similar to DYN-1(G40E) (Figure 6A, e and f, and C), does not result in the accumulation of cell corpses (Figure 6) or embryonic lethality (data not shown). In addition, R402A, another middle domain mutation in DYN-1, which corresponds to R399A in human dynamin 1, a mutation that was reported to disrupt dynamin tetrameric structure in vitro (Ramachandran et al., 2007 ), failed to cause dominant Ced or Emb phenotypes, although the mutant protein was induced to express by heat-shock at a relatively equivalent level (Figure 6, B and C). These results indicate that although these two middle domain mutant DYN-1 molecules are inactive (Figure 2 and Supplemental Figure S1), unlike GTPase domain mutant DYN-1, they do not interfere with the function of endogenous wild-type DYN-1.
If the dominant-negative effect of DYN-1 molecules carrying mutations in the GTPase domain is caused by their interference of the endogenous wild-type DYN-1 molecules in the same polymer, a second mutation that inactivates the oligomerization activity of the very mutant molecules should suppress their dominant-negative effect. We tested this hypothesis by introducing GTPase and middle domain double mutations into the dyn-1 heat-shock induction constructs. We found that the heat-shock induced expression of dyn-1(G40E I401F) double mutant cDNA failed to induce any dominant-negative phenotype in cell corpse removal (Figure 6) or embryonic development (data not shown). DYN-1(G40E I401F)::GFP displayed a bright fluorescence signal (Figure 6Ag) and a protein level similar to that of DYN-1(G40E)::GFP or DYN-1(I401F)::GFP (Figure 6C), indicating that the lack of dominant-negative effect is not due to lack of protein production or reduced protein stability. Similarly, I401F also suppresses the dominant-negative effect of G204E in cis (Figure 6, B and C, and Supplemental Figure S4). In addition, the mutation R402A also suppresses the dominant-negative effect of G40E in cis without significantly affecting the protein level, suggesting the suppression effect is common to middle domain mutations (Figure 6, B and C; and Supplemental Figure S4). That the dominant-negative effect of GTPase domain mutant DYN-1 relies on the self-assembly activity of DYN-1 strongly indicates that the in vivo self-assembly activity of DYN-1 is essential for its functions in the removal of cell corpses and embryonic development.
To study whether self-assembly of DYN-1 plays a role in DYN-1's subcellular localization, we examined the effect of DYN-1 mutations toward the localization of DYN-1::GFP. The transgenes were placed under the control of the engulfing cell-specific ced-1 promoter (Zhou et al., 2001b ; Yu et al., 2006 ) and expressed in wild-type or the corresponding homozygous mutant embryos. We characterized DYN-1 localization in cells in which DYN-1 is essential for cellular function, including intestinal cells, where it plays an essential role in endocytosis, and hypodermal cells, where it actively promotes the engulfment and degradation of apoptotic cells (Grant and Hirsh, 1999 ; Yu et al., 2006 ).
In intestinal cells, DYN-1::GFP is enriched along the apical surface toward the intestinal lumen (Figure 7A), consistent with the active self-assembly of DYN-1 in this region detected by the BiFC assay (Figure 5). In contrast, DYN-1(I401F)::GFP seems to reside in the cytoplasm in a diffused pattern in wild-type embryos, and also in dyn-1(en9) mutant embryos, where there is no wild-type DYN-1 molecules to stimulate assembly (Figure 7B). No significant enrichment on any cellular structures, either the apical surface of intestinal cells or the surfaces of pseudopods or phagosomes, was observed in 1.5-fold embryos (Figure 7B) or embryos at other developmental stages (data not shown).
To determine the localization of DYN-1(I401F) during the engulfment and degradation of apoptotic cells, we monitored each of these processes of three particular apoptotic somatic cells, C1, C2, and C3, by using a time-lapse recording technique established previously (Yu et al., 2006 ). C1, C2, and C3, which are located close to the ventral surface of an embryo, undergo apoptosis at approximately the same time (~330 min after the first embryonic cell division), and are engulfed by their neighboring ventral hypodermal cells ABplaapppa, ABpraapppa, and ABplaapppp, respectively (Figure 7D; Yu et al., 2006 ). Throughout the engulfment and degradation processes of cell corpses C1, C2, and C3, the clustering of DYN-1::GFP signal around these cell corpses reaches its peak shortly after engulfment completes (Figure 7, E and F), as reported previously (Yu et al., 2006 , 2008 ). Within 10 min, the transient clustering disappears and a phagosome is detected as a dark hole in the engulfing cell cytosol (Figure 7, E and Fd). In embryos that express DYN-1(I401F)::GFP, however, GFP does not enrich around phagocytic cup or nascent phagosomes at any moment throughout the engulfment and degradation process (Figure 7F, g–l; data not shown), indicating that the assembly-defective I401F mutation abolishes DYN-1's ability to associate with phagocytic membranes. The above-mentioned results thus indicate that self-assembly of DYN-1 is essential for the in vivo association of DYN-1 with target membranes.
In both wild-type and dyn-1(en40) mutant embryos that expressed DYN-1(G40E)::GFP, high level GFP signal was observed on the apical surface of intestinal cells (Figure 7C), indicating that the GTP binding activity is not essential for DYN-1 association with target membranes. Interestingly, the level of enrichment of DYN-1(G40E)::GFP signal on apical surfaces of intestinal cells is higher than that of DYN-1::GFP (Figure 7, A and C), suggesting that either DYN-1(G40E) displays a higher affinity to target membrane than DYN-1 or that it does not dissociate from the target membranes efficiently. Our time-lapse recording throughout the removal processes of C1, C2, and C3 further indicates two abnormal aspects of DYN-1(G40E) localization to phagosomal surfaces. First, the peak of DYN-1(G40E)::GFP signal detected on the surface of nascent phagosomes is higher and reached later than that of DYN-1::GFP in relation to the engulfment process (Figure 7F, m–p). Second, more strikingly, after reaching the intensity plateaus, the DYN-1(G40E)::GFP signal persists on phagosomal surfaces for a much longer time period compared with DYN-1::GFP (Figure 7E; F, q–t; and G). These observations strongly indicate that there is a profound defect in DYN-1(G40E)'s dissociation from phagosomal membranes, as observed in cryo-EM experiments in vitro (Figure 4), and they further suggest that, among other defects resulted from this mutation, a defect in membrane dissociation is likely to contribute to the failure of DYN-1(G40E) to promote the removal of apoptotic cells.
In addition, in hypodermal cells, DYN-1(G40E)::GFP accumulates at a high concentration on cytoplasmic puncta, which might represent intracellular organelles (Figure 7F, m–t; Yu et al., 2006 ). The degree of aggregation of DYN-1(G40E) cytoplasmic puncta is higher than that of wild-type DYN-1 (Yu et al., 2006 ), indicating that the G40E mutation might also affect the dynamic dissociation of DYN-1 from intracellular organelles.
The dyn-1(en9) and dyn-1(en40) homozygous embryos from heterozygous mothers carry a large number of persistent cell corpses, indicating a severe cell corpse removal defect (Figure 8B, m+z− embryos) (m, maternal gene product; z, zygotic gene product). When the wild-type maternal dyn-1 product was further depleted from the homozygous mutant embryos (see Materials and Methods), the cell corpse removal defect was further enhanced (Figure 8, A and B m−z− embryos). We monitored the degradation process of C1, C2, and C3 and found that, like dyn-1(n4039(G204E)) mutant embryos (Yu et al., 2006 , 2008 ), the en9 and en40 embryos are severely defective in the degradation of apoptotic cells.
We measured the duration of phagosomes using the CED-1C::GFP reporter, a fusion protein between the cytoplasmic domain of CED-1 and GFP. CED-1C::GFP, which is expressed under the control of the ced-1 promoter, is evenly distributed in the cytoplasm of engulfing cells and allows the recognition of phagosomes as GFP(−) holes in a GFP(+) background (Figure 8B; Yu et al., 2008 ). In wild-type embryos, the degradation of phagosomal content leads to the gradual reduction of phagosome size. One-hundred percent of C1, C2, and C3 phagosomes are totally degraded within 60 min after the engulfment is completed, among which 89% are degraded within 50 min (Figure 8, C and D). On the contrary, only 21.1 and 31.8% of phagosomes in dyn-1(en9) and dyn-1(en40) embryos, respectively, were fully degraded within 60 min. In addition, 15.8 and 22.7% of phagosomes in dyn-1(en9) and dyn-1(en40) embryos, respectively, lasted >80 min (Figure 8, C and D). These results indicate that the loss of the GTP binding or self-assembly activities severely affect DYN-1's functions in the degradation of cell corpses.
Previously, we found that the phagosome maturation defects caused by DYN-1 mutations were directly related to the severe defects in the delivery and fusion of early endosomes and lysosomes to phagosomes, vesicle trafficking events that are essential for providing digestive enzymes to and the acidification of phagosome lumen (Yu et al., 2006 , 2008 ). To determine the contribution of DYN-1's self-assembly and GTP binding activities to these vesicle/phagosome fusion events, we analyzed the effects of two classes of DYN-1 mutations on the recruitment of two distinct membrane tethering factors, the small GTPases RAB-5 and RAB-7 (Cai et al., 2007 ), to phagosomal surfaces. We found previously that the function of RAB-7 on phagosomal surfaces was essential for the recruitment and fusion of lysosomes but not early endosomes to phagosomal membranes (Yu et al., 2008 ). Using an engulfing cell-specific GFP::RAB-7 reporter developed previously (Yu et al., 2008 ), we monitored the dynamic recruitment of RAB-7 to phagosomal surfaces. In wild-type worms, >90% of phagosomes were labeled with high concentrations of GFP::RAB-7 within 10 min after engulfment is completed and until their complete degradation (Figures 7E, E,9,9, Aa and B; and Supplemental Movie S1; Yu et al., 2008 ). After RAB-7 recruitment, phagosomes started to reduce in size until complete disappearance within 60 min (Figure 9, Aa and B, and Supplemental Movie S1). In contrast, in dyn-1(en9) mutant embryos, 33.3% of phagosomes remained RAB-7(−) for the entire recording period of 100 min, all of which remained undegraded throughout the recording period (Figure 9, Ab and C, a–e; and Supplemental Movie S2). In addition, on the surface of 22.2% of phagosomes, GFP::RAB-7 was briefly enriched yet quickly disappeared, indicating a defect in maintaining RAB-7 on phagosomal surfaces; likewise, these phagosomes failed to reduce in size throughout the recording time of 100 min (Figure 9A, c and d; and C, f–k). In the rest of phagosomes (44.4%), the recruitment of RAB-7 to phagosomal surfaces was delayed to various extents (Figure 9A, e and f; and C, l–q), and the degradation of the corresponding phagosomes was delayed to equivalent levels. These observations strongly indicate that DYN-1 self-assembly and targeting to phagosomal membrane are critical for the recruitment of the RAB-7 GTPase.
In contrast to the severe RAB-7 recruitment defects observed in dyn-1(en9) mutants, in dyn-1(en40) mutant embryos, the defects in the dynamic recruitment of GFP::RAB-7 to phagosomal surfaces were relatively weak and only influence a small percentage of phagosomes (Figure 9A, g–k; and D). On the surfaces of 80% of phagosomes, RAB-7 was enriched within 10 min after phagosome formation (Figure 9A, g–i; and D, a–k). In the rest of phagosomes, RAB-7 was enriched between 10 and 20 min after phagosome formation (Figure 9A, j and k; and D, l–q). These observations indicate that the DYN-1(G40E) mutant protein, although deficient in GTP binding, is still able to promote the timely association of RAB-7 to phagosomal surfaces.
The relatively normal kinetics of RAB-7 recruitment seems to correlate with the relatively short delay of phagosome maturation observed when transgenic GFP::RAB-7 is expressed in dyn-1(en40) mutant embryos. In particular, among the 12 phagosomes onto which RAB-7 was recruited within 10 min of phagosome formation, the duration of eight phagosomes was shorter than 70 min (Figure 9A, g–h). In addition, the duration of the three phagosomes on which RAB-7 was recruited between 10 and 20 min after phagosome formation ranges between 64 and 86 min (Figure 9A, j and k), merely 16–56% longer than the 55-min average phagosome duration observed in wild-type cells (Figure 9Aa). Comparing to the drastically prolonged phagosome duration observed in dyn-1(en9) mutants that also expressed GFP::RAB-7 (Figure 9A, b–f), the expression of GFP::RAB-7 seemed to partially suppress the prolonged phagosome maturation phenotype in dyn-1(en40) mutants (Figure 9A, g–k).
However, the suppression activity of GFP::RAB-7 transgene does not seem to be fully penetrant. In particular, four of the 12 phagosomes onto which RAB-7 was recruited within the normal time frame (10 min) existed for a period between 81 and 108 min (Figure 9Ai and D, f–k). This phenomenon suggests that there are other essential events for phagosome maturation that could be affected by the G40E mutation in endogenous dyn-1. We further examined the dynamic phagosomal localization of RAB-5, a tethering factor between early endosomes and phagosomes, in the dyn-1(en9) and dyn-1(en40) mutant embryos.
The small GTPase Rab5 is enriched on maturing phagosomes (Vieira et al., 2002 ; Kinchen et al., 2008 ; Kitano et al., 2008 ). Using a GFP::RAB-5 reporter expressed in engulfing cells under the control of the ced-1 promoter that we constructed, we monitored the dynamic recruitment and duration of C. elegans RAB-5 on phagosomes C1, C2, and C3. In wild-type embryos, we observed that GFP::RAB-5, which was evenly distributed in the cytoplasm, was recruited to phagosomal surfaces within 4 min after the completion of engulfment and remained there for an average of 9 min (Figures 7E and and9E,9E, a–d and t; and Supplemental Movie S3). In both dyn-1(en9) and dyn-1(en40) homozygous embryos, however, the time points when GFP::RAB-5 was first enriched on phagosomal surfaces varied in a large range from <5 min to >40 min (Figure 9Et). Only 12.5 and 26.3% of phagosomes in dyn-1(en9) and dyn-1(en40) mutant embryos, respectively, recruited GFP::RAB-5 within 5 min like those in wild-type embryos (Figure 9E). Furthermore, 25.0 and 31.6% of phagosomes remained GFP::RAB-5(−) for 20 min or longer immediately after the formation of nascent phagosomes in dyn-1(en9) and dyn-1(en40) homozygous embryos, respectively (Figure 9E). Unlike the RAB-7 recruitment defect, the severity of the RAB-5 recruitment defect was similar between the dyn-1(en9) and dyn-1(en40) mutant backgrounds (Supplemental Movies S4 and S5). In both mutants, the duration of GFP::RAB-5 on phagosomes remained within 10 min, indicating that the transient nature of RAB-5 enrichment onto phagosomes is not changed (Supplemental Figure S5).
We monitored the engulfment processes of cell corpses C1, C2, and C3 over time with the GFP::RAB-5 reporter used above. This reporter, although is not enriched on the surface of pseudopods (Figure 9E), is evenly distributed in the cytoplasm, allowing us to track the extension and closure of thin pseudopods over time (Figure 10A, a–e). To further determine the precise time point when engulfment was initiated during embryonic development, we used the time point when the two hypodermal cells ABplaapppp and ABpraapppp first made contact and started to fuse with each other at the embryonic ventral midline, a developmental event independent of cell corpse engulfment, as a reference point (Figure 10Aw; Yu et al., 2008 ). In wild-type embryos, the engulfment of C1, C2, and C3 takes on average 5 min to complete and is usually completed at least 7 min before the contact between ABplaapppp and ABpraapppp is made (Figure 10A, a–e, and w) (Yu et al., 2008 ). In ced-1 mutant embryos known to be defective in both the engulfment and degradation of cell corpses, ~28% of C3 cell corpses fail to be engulfed within 30 min (Yu et al., 2008 ). We classified the specific defects observed in the en40 and en9 mutants into four categories (Figure 10B). In the dyn-1(en40) background, among the 22 cell corpses that we monitored, four were not engulfed within the 60-min recording period, due to either the lack of pseudopod initiation or the failure of pseudopod extension and closure (Figure 10A, p–v); two were not fully engulfed until at least 6 min after ABplaapppp and ABpraapppp made contact due to either delayed pseudopod initiation or slower-than-normal pseudopod extension, whereas the rest 16 were engulfed in normal kinetics (Figure 10B). Similar defects in the four categories were also observed when CED-1C::GFP, another cytosolic reporter was used to label pseudopods (Figure 10C). In dyn-1(en9) embryos, delay of pseudopod initiation and perturbation of pseudopod extension were also observed, albeit at a lower frequency (Figure 10C). Figure 10A, f–o, shows an example of a failure of engulfment initiation observed in dyn-1(en9) mutant background.
Although the self-assembly of dynamin along lipid surface has been observed in vitro (see Introduction), the actual oligomerization in vivo under physiological conditions remains to be visualized directly (Takei et al., 1995 ; Evergren et al., 2004 ). Moreover, it is not known whether the monomeric or oligomeric form of dynamin is the active form for promoting membrane fusion.
Using C. elegans as a model system, we have developed two complementary assays that detect the oligomerization of dynamin in vivo. The BiFC assay allows us to not only directly detect the interaction between DYN-1 monomers but also determine the specific membrane region where the interaction occurs. On the other hand, the genetic suppression of the dominant-negative effect of the GTPase domain mutations by assembly-deficient mutations enables us to determine the functional importance of DYN-1 self-assembly. Whereas our biochemical assays have demonstrated that the middle domain mutation I401F resulted in a specific defect in DYN-1 self-assembly in vitro, the results of these two in vivo assays described above, together with the defects in phagosome maturation and engulfment and the 100% embryonic lethality observed from the assembly-deficient dyn-1(en9(I401F)) mutant embryos, have established that DYN-1 does assemble as polymers in C. elegans cells and that this event is essential for its functions in the removal of cell corpses and in embryonic development (Figure 11).
Overexpression of mutant DYN-1 molecules defective in GTP binding and/or hydrolysis may interfere with the functions of endogenous wild-type DYN-1 through one of the following mechanisms: 1) as monomers, mutant DYN-1 molecules may sequestrate important positive regulators from wild-type DYN-1; or 2) mutant DYN-1 molecules may generate inactive DYN-1 polymers by assembling together with wild-type DYN-1. The absolute dependence of the observed dominant-negative effect on DYN-1's self-assembly activity indicates that the generation of “poisonous polymers” is the likely mechanism among the two. This genetic suppression assay thus is an efficient method for determining whether any particular dynamin mutation impairs dynamin's self-assembly in vivo.
Our in vivo study of the I401F mutation indicates that the self-assembly of DYN-1, which relies on its middle domain, is essential for DYN-1's association with the surface of pseudopods or phagosomes or other target membranes (Figure 11A). What is the molecular relationship between DYN-1's self-assembly and the association with target membranes? Despite the obvious link between the PH domain and the membrane localization of dynamin, it was found that the PH domain alone was insufficient to specify DYN-1's localization (Labrousse et al., 1998 ). In vitro, the lipid binding affinity of dynamin's PH domain monomers seems weak but is significantly enhanced when the PH domain is oligomerized (Kavran et al., 1998 ; Klein et al., 1998 ). Among other possibilities, self-assembly may induce a conformational change and enhance dynamin's affinity for membrane lipids. Our work reported here indicates that in vivo, as observed in vitro in cryo-EM studies, dynamin oligomers display a higher affinity for membranes than monomers. By this means, self-assembly strengthens the association of dynamin with its target regions on the plasma and intracellular membranes. As upstream regulators of DYN-1, the phagocytic receptor CED-1 and its adaptor CED-6 could recruit DYN-1 to the region of action by promoting either the self-assembly of DYN-1 monomers, or the association DYN-1 oligomers with specific membrane lipids such as phosphatidylinositol 4,5-bisphosphate, or both (Figure 11B).
In contrast to the effect of the I401F mutation, we found that abolishing the GTP binding and hence the GTPase activity of endogenous DYN-1 via the G40E mutation did not affect DYN-1's self-assembly or its ability to associate with its target membranes in vivo (Figure 11A). Rather, both in vitro and in vivo, DYN-1(G40E) mutant protein failed to disappear from its target membranes, indicating a defect in the disassembly of DYN-1 oligomers or the subsequent dissociation of monomers from target membranes. DYN-1's GTP hydrolysis and self-assembly activities thus contribute opposite effects toward DYN-1's association to target membranes.
We propose that these two opposite effects together establish an autoregulatory loop that precisely controls the duration and steady-state level of DYN-1 on its target membranes (Figure 11B). According to this model, in response to an upstream signal, DYN-1 molecules undergo self-assembly, an event that leads it to associate with target membranes. Meanwhile, self-assembly enhances the GTP hydrolysis activity of DYN-1, which subsequently results in the disassembly of DYN-1 oligomers and the consequential dissociation of DYN-1 molecules from its target membranes, a process akin to ATP or GTP hydrolysis in actin and tubulin polymerization, respectively (Desai and Mitchison, 1997 ). As long as the upstream signal is present, DYN-1 monomers will once again assemble along the membrane surface. As a consequence, the cycle of DYN-1 association to and dissociation from target membranes reach equilibrium and maintain a dynamic, steady state level of DYN-1 on the target membranes. Once the upstream signal disappears, the assembly-enhanced GTPase activity of DYN-1 oligomers enables rapid dissociation of DYN-1 from target membranes, as observed from phagosome surfaces (Figures 7F and and1111B).
The dissociation of DYN-1 from its target membranes also plays an indirect, yet important role in the timely association of DYN-1 to pseudopods and phagosomes. As an essential factor for many vesicle trafficking events, DYN-1 is targeted to the plasma membrane and intracellular membranes of different identities (Hinshaw, 2000 ; Praefcke and McMahon, 2004 ). In engulfing cells, DYN-1(G40E)::GFP, whose membrane dissociation function is impaired, was heavily aggregated to cytoplasmic puncta, which were presumed intracellular vesicles; furthermore, the association of DYN-1(G40E) to pseudopods or phagosomes was significantly delayed in relation to the timing of engulfment. This delay could be a result of the delayed dissociation of DYN-1(G40E) from other cellular compartments and is likely to further affect DYN-1's function in promoting pseudopod extension and phagosome maturation.
Both self-assembly and GTPase mutations in DYN-1 cause severe defects in phagosome maturation. Previously, we have established that DYN-1 mediates the fusion of early endosomes and lysosomes to phagosomes (Yu et al., 2006 , 2008 ). To understand the functional relationship of DYN-1's self-assembly and GTP binding activities with the fusion of intracellular organelles to phagosomes, we examined events that control these fusion events. In particular, we observed common and differential effects of self-assembly and GTP binding mutations toward the phagosomal recruitment of two membrane tethering factors RAB-5 and RAB-7, which are essential for the specific association of early endosomes and lysosomes to phagosomes, respectively (Figure 11B; Vieira et al., 2002 ; Harrison et al., 2003 ; Yu et al., 2008 ).
We determined the precise temporal pattern for the enrichment of both RAB-5 and RAB-7 on phagosomal surfaces in wild-type embryos (Figures 7E and and9;9; Yu et al., 2008 ). We further found that both the self-assembly and GTP binding mutations cause substantial delay of RAB-5 recruitment to phagosomes to various degrees, indicating that the oligomerization and GTP binding of DYN-1 are both required for the timely recruitment of RAB-5, and further suggesting that the DYN-1 molecules that are capable of recruiting RAB-5 are in a polymer state localized to the surface of phagosomes and likely to be in a GTP-bound form.
In contrast, whereas the self-assembly mutation of DYN-1 nearly completely destroys the temporal recruitment pattern of RAB-7 to phagosomes, the GTP binding mutation only causes a weak and partially penetrating delay of RAB-7 recruitment. These differential effects indicate that the self-assembly and the consequential phagosomal association of DYN-1 is both necessary and sufficient for the recruitment of RAB-7 to phagosomes, whereas the GTP binding ability does not seem to make a significant contribution to this event. To our knowledge, this is the first report of these types of differential effects related to the in vivo functions of dynamin, which suggests that novel and differential molecular mechanisms are involved in dynamin-mediated recruitment of RAB-5 and RAB-7. Perhaps RAB-5 or its associating protein complex selectively interacts with the GTP-bound form of DYN-1, whereas the RAB-7 complex does not distinguish between GTP- or GDP-bound forms of DYN-1.
In wild-type embryos, RAB-5 precedes RAB-7 for associating with phagosomal surfaces (Figure 7E). In the dyn-1(en40) mutant background, however, many phagosomes that are labeled with enriched GFP::RAB-7 signal have not been labeled with GFP::RAB-5 yet, indicating that the DYN-1 mediated recruitment of RAB-5 and RAB-7 might be independent of each other. Further study of the molecular mechanisms behind the phagosomal recruitment of RAB-5 and RAB-7 will further reveal the exact roles of DYN-1 in these events.
The tight correlation between the defects in RAB-5 and RAB-7 recruitment and maintenance and the defects in phagosome maturation observed in the dyn-1(en9) mutant embryos reveals that the failure to recruit RAB-5 and RAB-7 to phagosomes is probably the major cause for the severely delayed or blocked phagosome degradation. In dyn-1(en40) mutant embryos, the strong phagosome maturation defect is also well correlated with the defect in recruiting RAB-5. In contrast, we found that overexpression of a functional GFP::RAB-7 from an extrachromosomal array, which possesses the full rescuing capacity of defects displayed by rab-7 null mutants (Yu et al., 2008 ), partially suppressed the delay of phagosome maturation in dyn-1(en40) mutants. This suppression activity is consistent with the observation that the G40E mutation does not significantly affect the recruitment of RAB-7 to phagosomal surfaces, and further suggests that in dyn-1(en40) mutants, as long as RAB-7 is associated with the phagosomal membranes, it is functional in promoting phagosome maturation. The mechanism behind the suppression effect awaits investigation. It is possible that in engulfing cells, the endogenous RAB-7 molecules are limited in number. Once overexpressed from the transgenic array, the excessive amount of RAB-7 molecules might result in the recruitment of an excessive amount of lysosomes to phagosomes, which might compensate for the lack of early endosome incorporation to phagosomes due to the lack of RAB-5 association with phagosomes in dyn-1(en40) mutants.
In summary, our current observations, together with previous studies, indicate that DYN-1 directly promotes phagosome maturation by acting on the surface of phagosomes. Furthermore, DYN-1 promotes membrane fusion primarily through controlling the recruitment of membrane tethering factors. The self-assembly and GTP binding activities of DYN-1 contribute to common and different aspects of this fusion-promoting activity. In addition, recent studies have shown that the association of dynamin polymers with membrane stabilizes membranes of high curvature (Bashkirov et al., 2008 ). The disassembly of dynamin helix promotes membrane fission (Bashkirov et al., 2008 ; Pucadyil and Schmid, 2008 ). In addition to recruiting tethering factors, DYN-1 may also act in remodeling phagosomal membranes during phagosome maturation.
During engulfment, the GTP binding and self-assembly activities also play roles for DYN-1's function in promoting pseudopod initiation and extension. The GTP-bound form of DYN-1 on engulfing cell membranes might be critical for recruiting intracellular vesicle to pseudopods. Alternatively, the membrane fission activity of DYN-1, which requires GTP hydrolysis, might act on the surface of intracellular membranes to generate sufficient amount of intracellular vesicles for fusion with pseudopods.
Given that classical dynamins possess highly conserved domain structure and amino acid sequence from C. elegans to humans, and that human dynamin 2 can partially replace the function of DYN-1 in C. elegans (Clark et al., 1997 ; Yu et al., 2006 ), the mechanisms underlying the functions of dynamin are likely to be conserved from C. elegans to humans. Our findings in C. elegans will shed lights on the mechanism of mammalian dynamins in the removal of apoptotic cells and other biological processes.
We thank C.-D. Hu for BiFC vectors; T. Wensel for help with the GTP binding assay; X. He for the DeltaVision microscope; X. He, C. Peters, and members of the Z. Z. laboratory for helpful comments. Support for Z. Z. includes National Institutes of Health grant GM-067848, the Cancer Research Institute, and the Rita Allen Foundation. D. D. acknowledges the Israel Science Foundation, the Charles H. Revson Foundation (grant 530/03), and Russell Berrie Nanotechnology Institute at the Technion for financial support. F.A.Q. acknowledges support from NIH (GM088803) and the Welch Foundation (Q581).
This article was published online ahead of print in MBC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E09-05-0440) on December 16, 2009.