|Home | About | Journals | Submit | Contact Us | Français|
The elimination of unnecessary or defective cells from metazoans occurs during normal development and tissue homeostasis as well as in response to infection or cellular damage1. While many cells are removed through caspase-mediated apoptosis followed by phagocytosis by engulfing cells2, other mechanisms of cell elimination occur3, including the extrusion of cells from epithelia via a poorly understood, possibly caspase-independent process4. Here we identify a mechanism of cell extrusion that is caspase-independent and that can eliminate a subset of the Caenorhabditis elegans cells programmed to die during embryonic development. In wild-type animals, these cells die soon after their generation via caspase-mediated apoptosis. However, in mutants lacking all four C. elegans caspase genes, these cells nonetheless are eliminated by being extruded from the developing embryo into the extra-embryonic space of the egg. The shed cells exhibit apoptosis-like cytological and morphological characteristics, indicating that apoptosis can occur in the absence of caspases in C. elegans. We describe a kinase pathway required for cell extrusion involving PAR-4, STRD-1 and MOP-25.1/2, the C. elegans homologs of the mammalian tumor suppressor kinase LKB1 and its binding partners STRADα and MO25α. The AMPK-related kinase PIG-1, a possible target of the PAR-4:STRD-1:MOP-25 kinase complex, is also required for cell shedding. PIG-1 promotes shed cell detachment by preventing the cell surface expression of cell-adhesion molecules. Our findings reveal a mechanism for apoptotic cell elimination fundamentally distinct from that of canonical programmed cell death.
The CED-3 caspase is essential for nearly all programmed cell deaths that occur during C. elegans development5. However, a few cells undergo programmed cell death in ced-3 mutants5–7. We observed that some cells are eliminated from ced-3 embryos by being shed from the developing animal. The eggs of ced-3 mutants but not those of wild-type animals contained on average six shed cells that had detached during the comma stage of embryogenesis (~300 minutes post fertilization) (Figs. 1a–c and f; Supplemental Table S1). The shed cells detached at the anterior sensory depression or the ventral pocket (Figs. 2a–b) and remained within the egg (but separate from the animal) throughout embryogenesis. Embryos with a loss-of-function mutation in the Apaf-1 homolog ced-4 or the BH3 domain-encoding gene egl-1 or a gain-of-function mutation of the bcl-2 homolog ced-9 also produced shed cells (Fig. 1c), indicating that a defect in any step of the execution phase of programmed cell death can generate shed cells. As reported previously8, mutant embryos defective in engulfment (e.g., ced-1 or ced-5 embryos) contain “floaters” (Figs. 1c–d and g; Supplemental Table S1), cells that undergo CED-3-mediated apoptosis and detach from the embryo because they cannot be internalized by engulfing cells. In comparison to ced-3 shed cells, ced-5 floaters were smaller, more uniformly refractile when viewed by Nomarski optics, and less likely to aggregate into clumps of three or more cells after detachment (Figs. 1a–b and d; Supplemental Table S1). ced-3 mutations were epistatic to engulfment mutations with respect to the number of shed cells, their appearance, and their tendency to aggregate (Supplemental Table S1; data not shown). Thus, the shed cells of embryos defective in programmed cell death are genetically and morphologically distinguishable from those of embryos defective in engulfment.
The C. elegans genome encodes three additional caspase homologs: csp-1, csp-2 and csp-39. We determined that individual csp mutations did not cause the appearance of shed cells (Supplemental Fig. S1; Supplemental Table S2). Eggs from quadruple mutants lacking all four caspase genes, like ced-3 eggs, contained on average six shed cells (Figs. 1c and e), indicating that the generation of shed cells is caspase independent.
Although caspase activation can drive apoptosis, recent studies have suggested that caspases are not necessary for apoptosis3. We therefore examined csp-3; csp-1; csp-2 ced-3 (csp-Δ) shed cells for apoptotic characteristics, specifically phosphatidylserine (PS) exposure and TUNEL (terminal deoxynucleotidyl transferase dUTP nick end labeling)-reactive DNA fragments. Like floaters that undergo caspase-mediated apoptosis (Supplemental Fig. S2; ref. 8), the csp-Δ shed cells were reactive to the PS-binding protein MFG-e8 and to TUNEL staining (Figs. 1h–i). Also, ced-3 shed cells exhibited chromatin condensation (darkly staining nuclear material) and separation of the nuclear envelope double membrane in transmission electron micrographs (Fig. 1f and Supplemental Fig. S3). These apoptotic features were present in ced-5 floaters (Fig. 1g and Supplemental Fig. S3), although the cytoplasms of ced-5 floaters were more compact. We conclude that the shed cells of embryos lacking caspase activity are in many respects cytologically and morphologically apoptotic, indicating that caspases are dispensable for many cellular changes that occur during apoptosis.
The somatic cell lineage of C. elegans is essentially invariant10,11, allowing the precise identification of cell origins and fates. To determine the cellular identities of ced-3 shed cells, we recorded time-lapse videos of developing ced-3 embryos and traced the lineages of extruded cells in reverse (Figs. 2a–b; Supplemental Movie S1). We identified seven different cells eliminated by shedding from ced-3 embryos (Fig. 2c), all of which are cells that normally die during wild-type embryogenesis. This finding is consistent with our observation that ced-3 shed cells expressed egl-1 (Fig. 2d), the transcription of which initiates programmed cell death12, and with a previous report in which ABalapapaa (a cell fated to die) was observed detaching from a ced-3 embryo5. The cells that can be shed are among the first to die in the wild-type embryo: 14 cells die within the first 300 minutes of development, and seven of the eight identified shed cells are among this group of 14 cells11. Thus, specific cells fated to die early in embryogenesis can be eliminated by either canonical caspase-dependent apoptosis or by caspase-independent shedding.
To identify factors required for cell shedding, we tested genes involved in different cell-death processes and found that the generation of ced-3 shed cells did not require genes that mediate germline apoptosis, cell-corpse engulfment, necrosis, autophagy or lin-24-mediated cell death (Supplemental Tables S1 and S3). We therefore developed a screening strategy based on the hypothesis that the shed cell ABplpappap might survive and adopt a fate associated with a lineally related cell in animals doubly defective in the canonical cell-death pathway and cell shedding. The sister cell of ABplpappap generates the neuron RMEV and the excretory cell (Fig. 3a), which functions in osmoregulation13. We generated a pgp-12::cNLS::gfp transgene to express GFP specifically in the excretory cell and observed that wild-type and ced-3 animals contained a single GFP-positive cell (Figs. 3b and d). Using this reporter, we screened mutagenized ced-3 animals and identified mutants with a “Tex” (two excretory cells) phenotype. Two such Tex isolates, n5433 and n5437, are alleles of the gene pig-1 (Supplemental Fig. S4). Single mutants defective in a null allele of pig-1, gm344Δ, had one excretory cell (Fig. 3d), whereas 89% of pig-1(gm344) ced-3(n3692) mutants contained two GFP-positive nuclei that resembled the large nucleus of the excretory cell (Figs. 3c–d; Supplemental Fig. S4). Double mutants with pig-1 and ced-4, ced-9(gf) or egl-1 mutations were similar to pig-1 ced-3 animals (Supplemental Table S4). Inactivation of pig-1 by RNAi treatment phenocopied gm344, n5433 and n5437, confirming that loss of pig-1 function caused the Tex phenotype in these mutants (Supplemental Table S7).
The LIN-3 EGF ligand is expressed embryonically by the excretory cell14, and pig-1 ced-3 embryos contained an extra cell that expressed lin-3 (Supplemental Fig. S5a). Furthermore, the heads of pig-1 ced-3 animals contained large cysts (Fig. 3e) similar to those of mutants with defective excretory cell function13. Thus, pig-1 ced-3 mutants generated an ectopic excretory cell, albeit one that either was defective in osmoregulation or interfered with the function of the endogenous excretory cell.
To address whether the ectopic excretory cell of pig-1 ced-3 animals is derived from the unshed ABplpappap cell, we directly examined the fate of ABplpappap in pig-1 and pig-1 ced-3 embryos (Supplemental Movies S2 and S3). In pig-1 embryos (as in the wild-type), ABplpappap became a highly refractile cell corpse within 45 min of its generation (Fig. 3f and data not shown). By contrast, ABplpappap survived and divided approximately 115 min after it was generated in pig-1 ced-3 embryos (Fig. 3g). In the three pig-1 ced-3 embryos we examined, neither ABplpappap nor its descendants detached from the embryo, suggesting that an ABplpappap descendant gives rise to the ectopic excretory cell in pig-1 ced-3 larvae. As previously reported15, pig-1 ced-3 animals also contain ectopic RME-like neurons (Supplemental Fig. S5b; Supplemental Table S5), suggesting that when ABplpappap survives it generates both an ectopic RME-like and an ectopic excretory-like cell.
pig-1 inactivation by mutation or RNAi treatment reduced the number of shed cells in ced-3, ced-4 or ced-9(gf) embryos by nearly 75% (p < 5.0×10−7 for each pair-wise comparison, Student’s t-test; Fig. 3h and data not shown), demonstrating that pig-1 is generally required for the generation of shed cells. Given this observation, the effects of pig-1 on cell shedding are in several ways comparable to the effects of ced-3 on programmed cell death: (1) ced-3 affects most programmed cell deaths and pig-1 similarly affects most shed cells, indicating that ced-3 and pig-1 act generally to drive programmed cell death and generate shed cells, respectively; (2) like programmed cell deaths, extruded cells share morphologic and genetic properties and can be viewed as expressing a specific cell fate; and, (3) like mutations in ced-3, mutations in pig-1 cause cells that should die to expresss the fates of cells that normally survive.
pig-1 encodes a homolog of MELK, an AMPK-related serine-threonine kinase required cell autonomously for the asymmetric cell divisions of many C. elegans neuroblasts15,16. Mammalian AMPK-related kinases control metabolism and cell polarity17 and are activated via phosphorylation of a conserved threonine within their T-Loop domains by the kinase LKB118 and its complex partners STRADα and MO25α19,20. The PIG-1 T-Loop threonine (T169) was necessary for a pig-1 transgene to rescue the ectopic excretory cell defect of pig-1 ced-3 mutants, whereas changing T169 to a phosphomimetic aspartic acid bypassed this requirement (Fig. 3i), indicating that PIG-1 likely is activated by phosphorylation of its T-Loop. We therefore tested the C. elegans homologs of LKB1, STRADα and MO25α (par-4, strd-1, and paralogs mop-25.1, mop-25.2 and mop-25.3, respectively) for roles in the elimination of ABplpappap. par-4 or strd-1 inactivation caused the Tex phenotype in ced-3 mutants (Figs. 3j and k; Supplemental Table S6). Furthermore, the ok2283 deletion allele of strd-1 reduced the number of shed cells in ced-3 embryos by 67% (p = 1.9×10−10, Student’s t-test; Fig. 3h). The inactivation of both mop-25.1 and mop-25.2 was necessary to cause the Tex defect in ced-3 mutants (Fig. 3l; Supplemental Table S6), indicating redundant function. We conclude that both PIG-1 and the PAR-4:STRD-1:MOP-25 complex are required for cell shedding in C. elegans.
Our genetic analyses suggest that PIG-1 and PAR-4 function in the same pathway. The deletion mutation ok2283, a putative null allele of strd-1 (Supplementary Table S8), failed to enhance any of the pig-1 or pig-1 ced-3 defects (Fig. 4h and i; Supplementary Fig. S6; Supplementary Tables S8–9). Although, par-4 and strd-1 can act independently21, this result argues that PIG-1 and PAR-4 and its binding partners function in the same pathway. We believe that PIG-1 is the phosphorylation target of PAR-4 in cell shedding because inactivation of PIG-1 but of no other AMPK-related kinase caused the survival of ABplpappap in ced-3 mutants (Supplementary Table S7). The Tex defect of pig-1(gm344) ced-3(n3692) animals (89%) was higher than that of strd-1(ok2283); ced-3(n3692) animals (43%) (Figs. 3d; Supplementary Table S4). Thus, STRD-1 and possibly the entire PAR-4 complex are required partially for PIG-1 activation, indicating that other factors, or PIG-1 itself via autophosphorylation, also stimulate PIG-1 activity. Indeed, MELK phosphorylates itself in vitro18.
How might PIG-1 regulate cell shedding? We reasoned that shed cells must be deficient in adhesive contacts with the embryo and explored whether pig-1 modulates cell-adhesion complexes, specifically adherens junctions. In C. elegans, adherens junctions comprise a cadherin-catenin complex and a complex including DLG-1 (Discs Large) and AJM-1. Both complexes participate in the enclosure of the embryo within a layer of epidermal cells22, a major morphogenic event that coincides with the detachment of shed cells.
We examined the expression of AJM-1, DLG-1, HMP-1 (α-catenin) and JAC-1 (p120 catenin) fused to GFP in shed cells. AJM-1, DLG-1 and HMP-1 were not visible at the surface of ced-3 or ced-4 shed cells that had recently detached from either the anterior sensory depression (Figs. 4a and c; Supplemental Fig. S7b) or the ventral cleft (Fig. 4b; Supplemental Fig. S7a). (JAC-1::GFP was detectable in some shed cells; Supplemental Fig. S7c). The absence of cell-adhesion proteins was striking given their presence on adjacent epidermal cells (Figs. 4a–c; Supplemental Fig. S7). ABplpappap in ced-3 embryos remained at the ventral surface of the embryo from its generation until its detachment and never expressed HMP-1::GFP (Figs. 4d and e; data not shown). By contrast, in pig-1 ced-3 embryos, ABplpappap ingressed dorsally during ventral enclosure (Fig. 4f), matching the movement of the excretory cell away from the ventral surface14. Strikingly, ABplpappap and its descendants exhibited a uniform expression of HMP-1::GFP at the cell surface (Fig. 4g), indicating that a loss of pig-1 function results in the inappropriate expression of cell-adhesion molecules. Thus, PIG-1 might facilitate cell shedding by preventing the expression of cell-adhesion molecules on the surfaces of shed cells.
Several lines of evidence suggest that pig-1 promotes shed cell detachment through endocytosis. Many cell corpses, including the ABplpappap corpse, were not efficiently cleared in pig-1 embryos (Figs. 4h and i). The pig-1 cell corpses were encircled rapidly by the engulfment receptor CED-1 and stained positively with acridine orange, a marker of internalized corpses (Fig. 4j; Supplemental Fig. S8). These results demonstrate that the delay in clearance was post-engulfment and reflected a defect in corpse degradation, a process that requires endocytic pathway components23. Interestingly, a recent study showed that two C. elegans ARF GTPase genes, arf-1.2 and arf-6, and an ARF GTPase-activating protein (GAP) gene, cnt-2, have functions in cell-fate determination similar to those of pig-1 as well as roles in receptor-mediated endocytosis24. Mammalian Arf GTPases function in endocytosis25 and can remove cadherin complexes from the cell surface26. We noted that inactivation of Arf GTPase genes arf-1.2 or arf-3 or Arf guanine exchange factor gene grp-1 also produced ectopic excretory cells in ced-3 mutants (Fig. 4k; Supplemental Table S10).
Taken together, our observations suggest that the PAR-4 complex, PIG-1, and ARF GTPases promote the detachment of shed cells through the endocytosis-mediated removal of cell-adhesion molecules from the cell surface. Thus, the programmed elimination and apoptosis of at least eight C. elegans cells can be accomplished either through canonical caspase-mediated apoptosis involving the engulfment of dying cells or through a caspase-independent shedding mechanism that also results in apoptosis and that requires the PAR-4 complex and the AMPK-related kinase PIG-1 (Fig. 3m). These two mechanisms are functionally redundant, as ABplpappap and the other shed cells survive only in mutants in which both pathways are disrupted.
We propose that cell-shedding is an evolutionarily conserved mechanism of cell elimination. Many epithelia extrude cells constitutively to maintain tissue homeostasis, and the shed epithelial cells of vertebrates share features with the shed cells of caspase-deficient C. elegans embryos. First, like other shed epithelial cells4, shed intestinal enterocytes frequently exhibit apoptotic markers27,28, including caspase-3 activation and TUNEL-reactivity. Second, despite the apoptotic appearance of shed enterocytes, the intestinal epithelia of caspase-3−/−, Apaf-1−/−, bax−/−, bak−/−, or bcl-2 over-expressing mice are not grossly abnormal3,29, suggesting that cell extrusion is not dependent on caspase-mediated cell-killing. Third, LKB1 mutations cause Peutz-Jeghers syndrome (PJS)30, which is characterized by intestinal hamartomas (polyps) containing excess epithelial cells. It is possible that LKB1 mutations contribute to polyp formation by causing a defect in the extrusion of epithelial cells. Based on our observations of C. elegans, we predict that the PIG-1 homolog MELK could be a target of LKB1 in the gastrointestinal tract and that mutations of MELK might also impair enterocyte shedding and cause a polyposis phenotype in the mammalian intestine.
All C. elegans strains were cultured as described previously31 and maintained at 20°C unless noted otherwise. We used the Bristol strain N2 as the wild-type strain. The mutations used in our experiments are listed below:
The Pegl-1::gfp transcriptional reporter and the Pdyn-1::mfg-e8::Venus (pVV59) translation reporter have been described previously32,33. The Punc-25::mStrawberry transcriptional reporter (pSN223) was constructed using PCR to amplify a 1.8 kb fragment of the unc-25 promoter region with the primers 5’-CGAATTTTTGCATGCAAAAAACACCCACTTTTTGATC and 5’-CGGGATCCTCGAGCACAGCATCACTTTCGTCAGCAGC. The resulting amplicon was digested with SphI and BamHI and ligated into pSN199, which encodes the 4xNLS::mStrawberry fusion. pSN199 was constructed by replacing the AgeI-EcoRI fragment of pPD122.56 with the AgeI-EcoRI fragment from the plasmid mStrawberry 6. The transcriptional reporter Ppgp-12::gfp (pSN359) was constructed using PCR to amplify a 2.9 kb fragment of the pgp-12 promoter region with the primers 5’-GCAAGCTTGTGCTTGCAGTGAACCAGAAACT and 5’-CCTCTAGAATTCCATCAATTGGCTCGATCGCA. The resulting amplicon was digested with HindIII and XbaI and ligated into pPD122.56. The pig-1 rescuing construct (pDD078) was constructed using PCR to amplify three overlapping genomic fragments spanning 17.5 kb with the following primer pairs: 5’-CAGTGAGCGCGCGTAATACGACTCACTATAGGGCGAATTGGCCACATCAAATGAAAGACG and 5’-TCAGAGTTCAATATATGTTGG; 5’-TGTCTACACCACTCCAAACACC and 5’-ACGACGGCATCAGATATTCG; and, 5’-CCAAGCGCGCAATTAACCCTCACTAAAGGGAACAAAAGCTTGATGATGATGTCCGTGAGC and 5’-CATGTCCAACAATGGAATCG. The resulting amplicons were ligated into pRS426 by homologous recombination in Saccharomyces cerevisiae (yeast-mediated ligation). The pig-1(STOP) (pDD084), pig-1(T169A) (pDD085) and pig-1(T169D) (pDD088) constructs were also genetrated by yeast-mediated ligation using a SfoI-NaeI digested fragment of pDD078. pig-1(STOP) changes codons 8 and 9 to the stop codons TAG and TGA, respectively. pig-1(T169A) changes the T-Loop threonine codon at position 169 to GCG, which encodes an alanine residue. pig-1(T169D) changes the T-Loop threonine codon at position 169 to GAC, which encodes an aspartic acid residue. pDD073 (pL4440::par-4) was constructed using PCR to amplify the par-4 cDNA with the following primers: 5’-GCGCCGCGGATGGATGCTCCGTCGACATCC and 5’-GCGTCTAGACTAAGCACTATCGGTACGAG. The resulting amplicon was digested with SacII and XbaI and then ligated into pL4440.
Shed cells or floaters were counted in eggs between the 2-fold and 3.5-fold stages of development (approximately 450 to 600 min after the first cell division) using a 100× objective equipped with Nomarski differential interference contrast (DIC) optics. The number of excretory cells and excretory-like cells were counted in L3 larvae carrying the Ppgp-12::gfp transcriptional reporter using a 100× objective. A cell was scored as an excretory cell if it was located in the anterior third of the animal and its nucleus strongly expressed GFP. Persistent cell corpse and extra pharyngeal cell counts were performed as described previously34.
ced-3(n3692); nIs434 or nIs433; ced-3(n3692) L4 larvae were mutagenized with ethyl methanesulfonate (EMS) as described previously31. Using a dissecting microscope equipped for the detection of GFP fluorescence, we screened approximately 25,000 F3 progeny of the mutagenized animals for an extra GFP+ cell near the posterior bulb of the pharynx. We showed by complementation testing and DNA sequence determination that two of the mutants isolated from this screen, n5433 and n5437, were alleles of pig-1.
RNAi treatments were performed by feeding using RNAi constructs and reagents described previously35,36. Briefly, HT115 E. coli carrying RNAi clones in the pL4440 vector were cultured overnight in LB liquid media with antibiotics. NGM plates containing 1 mM IPTG and antibiotics were seeded with the LB culture and incubated 24 hrs at 20°C. Five L2 larvae were placed onto each seeded NGM plate, and the progeny of these animals were scored for either shed cells or ectopic excretory cells as described above. Control RNAi treatments were performed with the empty vector pL4440. The complete sequences of all RNAi sequences used in this study are provided in the Supporting Information.
Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) reactions were performed as described previously37,38. Fixation times (between 12 and 16 min) were optimized for labeling the floaters in ced-5(n1812); nuc-1(e1392) eggs. nuc-1(e1392) facilitates the observation of TUNEL-reactivity. Acridine orange staining of embryos was performed as described previously39.
Gravid sem-4(n1378); ced-5(n1812) or sem-4(n1971); ced-3(n717) adult animals were fixed by high pressure freezing in an HPM010 high pressure freezer by Abra. They were then substituted using an RMC FS-7500 freeze substitution system with 2% osmium, 2% water in acetone at −90°C for 5 days, warmed to −20°C over 14 hours, and then held at −20°C for 16 hrs before warming to 0°C. Samples were washed 6 × 20 min, infiltrated stepwise into Eponate 12 resin (Ted Pella), and polymerized at 60°C. The resulting blocks were sectioned at 50 nm, post stained with uranyl acetate and lead, and imaged with a JEM-1200EX II microscope (Jeol) using an AMT XR41 CCD camera. sem-4 mutations cause eggs to accumulate in the hermaphrodite gonad, allowing more embryos (including more later-stage embryos) to be analyzed per sample.
Nomarski DIC and epifluorescence micrographs were obtained using an Axioskop II (Zeiss) compound microscope and OpenLab software (Agilent). Merged DIC and epifluorescence images were generated using ImageJ software (NIH). For the time-lapse experiments, early (2 or 4-cell stage) embryos were dissected from gravid adults, mounted on a slide with a 4% agar pad, and covered with a coverslip that was sealed to the slide with petroleum jelly to prevent the preparation from drying. The developing embryos were imaged every four minutes for a total of 300 min, and at each time point a Z-stack of 50 images spaced at 0.6 µm was acquired. Confocal microscopy was performed using a Zeiss LSM 510 instrument, and the resulting images were prepared and modified using ImageJ software.
We thank Z. Zhou, T. Hirose and S. Nakano for reporter constructs; B. Castor, E. Murphy and R. Droste for determining DNA sequences; S. Dennis for screening of the deletion library; N. An for strain management; and, J. Yuan, S. Nakano, N. Paquin, C. Engert and A. Saffer for helpful discussions. The Caenorhabditis Genetic Center, which is funded by the NIH National Center for Research Resources (NCRR), provided many strains. S. Mitani kindly provided grp-1(tm1956). D.P.D. was supported by postdoctoral fellowships from the Damon Runyon Cancer Research Foundation and from the Charles A. King Trust. H.R.H. is the David H. Koch Professor of Biology at the Massachusetts Institute of Technology and an Investigator of the Howard Hughes Medical Institute.
Author contributions:D.P.D. and H.R.H. designed the experiments, analyzed the data and wrote the manuscript.
D.P.D. and V.H. performed the experiments.