Introduction

Two distinct signaling mechanisms, known as the cell-intrinsic and cell-extrinsic pathways, control activation of the proapoptotic caspase family of cysteine proteases in mammals.1 The intrinsic pathway responds primarily to developmental cues or cellular damage and triggers apoptosis through the Bcl-2 gene family and the initiator protease caspase-9.2 The extrinsic pathway mediates apoptosis in response to Fas ligand (FasL) and Apo2L/TRAIL (Apo2 ligand or tumor necrosis factor (TNF)-related apoptosis-inducing ligand), which signal respectively through the death domain (DD)-containing receptors Fas or death receptor (DR)4 and DR5. These receptors use the adaptor molecule Fas-associated DD (FADD) to recruit the initiator proteases caspase-8 and -10 into a death-inducing signaling complex (DISC), where they are activated.3, 4 Initiator caspases in turn cleave and activate effector caspases such as caspase-3, -6 and -7 to carry out programmed cell death.

While components of the intrinsic pathway apparently exist in all metazoans, the extrinsic pathway is a more recent evolutionary development. No TNF or TNF receptor (TNFR) superfamily members have been found to date in Caenorhabditis elegans. In Drosophila, a single TNF ligand (Eiger) and its associated receptor (Wengen) induce apoptosis indirectly, by activating the caspase-9 homolog DRONC through the c-Jun N-terminal kinase (JNK) pathway; Drosophila homologs of caspase-8 (DREDD) and FADD do not appear to play a role in the extrinsic apoptosis pathway.5, 6 DD-containing TNFRs have been reported exclusively in vertebrates, with examples in teleost (below), avian,7 and mammalian species.3 However, not all DD-containing TNFRs are dedicated activators of the extrinsic apoptosis pathway. For example, mammalian TNFR1 signals through the adaptor TNFR-associated DD (TRADD) and its principal role in vivo is NF-κB activation, which inhibits apoptosis.8 Only DR4, DR5, and Fas have DDs that directly bind FADD.

In mammals, the extrinsic pathway plays an important role in regulating the immune system.9 Studies with gene-knockout mice suggest that the extrinsic pathway also may be required during embryogenesis. Mice deficient in FADD, caspase-8, or the extrinsic pathway inhibitor c-FLIP (cellular FLICE inhibitory protein) all die in utero between embryonic days 10.5 and 12.5.10, 11, 12, 13 However, other observations seem to indicate that the extrinsic pathway per se is not essential for embryonic development. Ex vivo whole-embryo culture rescues all embryonic defects in caspase-8 knockout mice.14 Moreover, mice deficient for FasL or Apo2L/TRAIL signaling complete embryogenesis without significant defects.15, 16, 17, 18

To gain further insight into the evolution of the extrinsic apoptosis pathway and its importance during embryogenesis, we turned to the zebrafish (Danio rerio). Although known zebrafish expressed sequence tags (ESTs) appear to encode portions of apoptosis genes,19 the extrinsic pathway remains poorly understood in this model organism. Three putative TNF superfamily ligands have been reported, but their activities remain uncharacterized.20, 21, 22 Two DD-containing TNFRs also have been described: ovarian TNFR (zOTR), which is highly expressed in ovary,22 and hematopoietic DR (zHDR), which has been implicated in hematopoiesis.23, 24 Neither zHDR nor zOTR have unambiguous sequence homology to specific mammalian TNFRs, and their cognate ligand(s) and signaling mechanism(s) remain unknown.

Here, we systematically define the extrinsic apoptosis pathway in zebrafish and show that it resembles its mammalian counterpart. Disruption of the extrinsic pathway does not impede zebrafish embryonic development. However, ectopic expression of Apo2L/TRAIL homologs triggers apoptosis in specific embryonic tissues through zHDR and zOTR. Ligand-induced apoptosis in these tissues requires both the extrinsic and intrinsic pathways.

Results

Extrinsic apoptosis pathway homologs in zebrafish

Zebrafish possess a single FasL homolog and four Apo2L/TRAIL relatives (designated zDL1a, zDL1b, zDL2, and zDL3 for Death Ligand (DL); Figure 1a). Although zDL1a was described previously, its published sequence (Accession #AF250041) is incomplete, lacking both intracellular and transmembrane regions (full-length sequence Accession #DQ812112). The zebrafish genome also encodes two proteins (designated zTNF1 and zTNF2), previously described as zebrafish TNFα orthologs based on similarity to other teleost TNFα genes.20, 21 The genomic regions encoding zDL2 and zTNF2 in zebrafish have conserved synteny with human loci encoding Apo2L/TRAIL and TNFα, respectively (Supplementary Table 1 and Figure 1a).

Figure 1
figure 1

Structure and homology of specific TNF and TNFR superfamily members. (a) Dendrogram of TNF domains of zebrafish, mammalian, and avian ligands. (b) Schematic representation of zebrafish and mammalian DD-containing TNFRs using the domain classification system found in Bodmer et al.44 Black bars within individual modules indicate the number and relative positions of cysteine residues. The N-terminal A1 module of mammalian Fas is present in murine but not human Fas and is rendered as semitransparent. Chicken TvbS1 is included to illustrate the diversity of ECDs in the TNFRSF10 group. (c) Dendrogram of DDs of zebrafish, mammalian, and avian receptors. All sequence alignments were produced by the ClustalW algorithm

Full size image

We identified eight DD-containing TNFR superfamily members in zebrafish. Six of these – zTNFR1, zFas, zDR6, zNGFRa, zNGFRb, and zNGFRc – can be matched with mammalian orthologs based on the architecture of their extracellular domains (ECDs), a dendrogram of their DDs, and conserved synteny with human counterparts (Figure 1b and c; Supplementary Figure 1a). Mammalian homologs of zHDR and zOTR were more difficult to identify by sequence alone. Their ECDs resemble CD40, herpes !QJ;virus entry mediator (HVEM), and Fas while their DDs are most homologous to DR4, DR5, and avian TvbS1 (Figure 1b and c). Multiple independently amplified zOTR clones revealed that the GenBank sequence for this gene (Accession #AAG24365) contains a missense mutation; we therefore submitted a corrected sequence (Accession #DQ812116).

We also identified homologs of FADD, TRADD, and numerous initiator and effector caspases (Supplementary Table 1 and Figure 1b). There are at least six zebrafish caspases with protease domains more homologous to caspase-8 and -10 than to other known caspases (Supplementary Figure 1c). Both zCaspase-8a and zCaspase-8b have N-terminal death effector domains (DEDs), but only zCaspase-8a contains the QACQG active-site motif characteristic of mammalian caspase-8 and -10. zCaspase-8a-like is highly homologous to zCaspase-8a in the protease domain but has a truncated N-terminus lacking DEDs. zCaspase-Xa, -Xb, and -Xc are most homologous overall to mammalian caspase-8, but contain a QACRG motif and lack obvious N-terminal DEDs. We found a single ortholog of the long splice form of c-FLIP. Like its mammalian counterpart, zFLIP is structurally similar to caspase-8 and -10 but lacks an active-site cysteine.

Ectopic expression of zHDR, zOTR, zFas, zFADD, zTRADD, and caspases induces apoptosis

We injected zebrafish embryos with mRNAs encoding zHDR, zOTR, zFas, zDR6, or zTNFR1 to determine if these DD-containing TNFRs can trigger apoptosis. Nearly all embryos overexpressing zHDR, zOTR, or zFas displayed morphological changes consistent with apoptosis and died by 8 h postfertilization (h.p.f.; Figure 2a and b). To confirm caspase activation, we used a monoclonal antibody that recognizes the cleaved, active form of caspase-3. Immunostaining of dying embryos showed high levels of activated caspase-3 throughout the cells of the blastodisc and in the cytoplasm of the yolk cell (Figure 2a). Mock-injected embryos, or those overexpressing zDR6 or zTNFR1, did not exhibit apoptotic morphology or caspase-3 activation. These data correspond well with mammals, where TNFR1 and DR6 are implicated mainly in nonapoptotic responses, while Fas activation is directly linked to apoptosis.3

Figure 2
figure 2

Specific zebrafish extrinsic pathway genes induce cell death in developing zebrafish embryos. (a) Overexpression of zHDR, zOTR, zFADD, zCaspase-3a, and zCaspase-8a in zebrafish embryos induces apoptosis. Embryos were injected with 500 pg of the indicated mRNA and photographed at 8 h.p.f. (top panels) or fixed at 5–8 h.p.f. for antiactive caspase-3 immunohistochemistry (bottom panels). (b) Ectopic expression of zHDR, zOTR, zFas, zFADD, zTRADD, and caspases, but not zFLIP, results in embryonic mortality. Embryos were injected with 500 pg of the indicated mRNA (zHDR, zOTR, and zFas were found to be potent inducers of apoptosis and used at 200 pg) and embryonic mortality (defined as lysis of the yolk cell and/or blastomeres) was scored at 8 h.p.f. Significant mortality was observed subsequent to 8 h.p.f. for zFADD, zTRADD, and caspases, so these were also assessed at 24 h.p.f. N=number of independent experiments; n=number of embryos scored. Error bars indicate standard error of the mean between experiments. (c) Overexpression of zebrafish receptors and zFADD in human HEK-293 and HeLa cells. Cells were transfected with constructs encoding the indicated proteins and the proportion of cells remaining viable at 24 h was recorded

Full size image

Injection of zFADD, zTRADD, zCaspase-2, zCaspase-3a, zCaspase-8a, and zCaspase-9 mRNAs induced extensive apoptosis and embryonic death between 8 and 24 h.p.f. (Figure 2a and b). Embryos surviving beyond 24 h.p.f. displayed abnormal morphology and had numerous active caspase-3-staining cells in all tissues (not shown). In contrast, zFLIP, zDL1a, zDL2, zDL3, zFasL, zTNF1, or zTNF2 injection did not increase mortality (Figure 2b). Overexpression of zDL1b caused a modest amount of death by 8 h.p.f. (∼12%), but surviving embryos lacked the indiscriminate apoptosis triggered by injection of zFADD, zTRADD, or caspase homologs.

We also tested the activity of zebrafish receptors and zFADD upon transient overexpression in human cells. We transfected HeLa and human embryonic kidney (HEK)-293 cells with vectors encoding zHDR, zOTR, zFas, zTNFR1, or zFADD. After 24 h, cells overexpressing zebrafish receptors or zFADD showed substantially higher levels of cell death than those transfected with an enhanced green fluorescent protein (EGFP)-expressing control vector (Figure 2c). zHDR and zOTR elicited the most robust effect, inducing death in 50–100% of the transfected cells.

Ectopic zDL1a, zDL1b, and zDL2 expression induces apoptosis in notochord and hematopoietic cells

Although ligand overexpression did not increase mortality, zDL1a, zDL1b, and zDL2 produced striking morphological defects. Injected embryos showed greatly reduced tail lengths, flattened somites lacking a horizontal myoseptum (Figure 3a–c; Supplementary Figure 2a), and aberrant posterior vasculature (not shown). Embryos injected with zDL1a (Figure 3d) and zDL2 (not shown) also exhibited pooling of erythrocytes in the perianal region. This combination of phenotypes resembles known mutants with defects in notochord formation such as floating head and no tail (ntl).25, 26 The notochord is a rod-like structure found in all chordate embryos that functions as a major skeletal element and an essential signaling center for tissue patterning. Injection of zDL3, zFasL, zTNF1, or zTNF2 did not cause discernible defects (Figure 3b).

Figure 3
figure 3

zDL1a, zDL1b, and zDL2 selectively kill cells of the notochord and ICM during embryogenesis. (a) Overexpression of zDL1a, zDL1b, and zDL2 leads to ‘short-tail’ embryos with activated caspase-3 in notochord cells; zDL1b also activates caspase-3 in the ICM. All embryos were injected with 500 pg of mRNA except for zDL1b, which was found to be much more potent and used at 200 pg. Embryos were photographed at 48 h.p.f. (top panels) or fixed at 30 h.p.f. for antiactive caspase-3 immunohistochemistry (bottom panels). Abbreviations: n, notochord; icm, intermediate cell mass. (b) Only zDL1a, zDL1b, and zDL2 produce short-tail embryos. Embryos were injected with 500 pg of mRNA (200 pg of zDL1b) and scored at 30 h.p.f. for tail length; only embryos with unambiguous reductions in tail length were recorded as short-tail. N=number of independent experiments; n=number of embryos scored. Error bars indicate standard error of the mean between experiments. (c) Somites in zDL1a overexpressing embryos are flattened and lack a horizontal myoseptum. Embryos were injected with 500 pg zDL1a and photographed at 48 h.p.f. (d) Mature erythrocytes pool in the perianal region of zDL1a-injected embryos (arrowheads) and are completely absent in zDL1b-injected embryos. Embryos were injected with zDL1a (500 pg) or zDL1b (200 pg) and whole-mount o-dianisidine staining for hemoglobin was carried out at 48 h.p.f. Arrows indicate circulating erythrocytes in wild-type embryos. (e) Increased numbers of TUNEL-positive cells are present in the notochord and ICM of zDL1b-injected embryos. Embryos were injected with 200 pg zDL1b, fixed at 30 h.p.f., and apoptotic cells were labeled by the TUNEL method. Abbreviations: n, notochord; icm, intermediate cell mass

Full size image

Owing to the similarities between notochord mutants and ligand-injected embryos, we looked for notochord apoptosis. Injection of zDL1a, zDL1b, and to a lesser extent zDL2, induced strong activation of caspase-3 in notochord cells (Figure 3a). We verified apoptosis using the terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling (TUNEL) method, which detects apoptotic DNA fragmentation. We found numerous TUNEL-positive cells in the notochords of zDL1a, zDL1b, and zDL2-injected embryos (Figure 3e; zDL1a and zDL2 not shown). In contrast, the notochords of mock-injected embryos contained very few TUNEL-positive cells and showed no evidence of caspase-3 activation.

zDL1b-injected embryos did not show the perianal blood pooling associated with zDL1a and zDL2 overexpression in spite of displaying an identical short-tail phenotype. To determine whether mature erythrocytes were present, we stained embryos for hemoglobin using o-dianisidine. Embryos injected with zDL1a had numerous erythrocytes trapped in the perianal region while those injected with zDL1b showed virtually no o-dianisidine staining (Figure 3d). In zebrafish and other teleosts, primitive (embryonic) hematopoiesis occurs in a region of the trunk ventral to the notochord known as the intermediate cell mass (ICM).27 Activated caspase-3 staining and TUNEL revealed numerous apoptotic cells in the ICM of zDL1b-injected embryos at 30 h.p.f. (Figure 3a and e).

Expression of extrinsic pathway genes in the notochord and ICM

We analyzed the zebrafish extrinsic pathway genes for notochord and ICM-specific expression patterns using whole-mount in situ hybridization (ISH) (Figure 4a-l). At 8 h.p.f. (75% epiboly), we detected zOTR and zFADD in the embryonic shield, a dorsal population of cells which gives rise to the notochord (Figure 4d and j). By 16 h.p.f. (14-somite stage), both transcripts, as well as zHDR (Figure 4c), were found throughout the notochord. We confirmed notochord-specific expression at this stage in sectioned embryos (Figure 4e for zHDR). By 30 h.p.f., all three transcripts were limited to the posterior notochord. This dynamic expression pattern is highly reminiscent of classic notochord markers such as sonic hedgehog (shh) and ntl. zDL1a was the only ligand associated specifically with the notochord (Figure 4a). We detected expression in the embryonic shield at 8 h.p.f. and in bilateral, segmented stripes flanking the notochord at 16 h.p.f. Sectioning revealed strong zDL1a expression in cell populations directly abutting the notochord (the floorplate of the neural tube, the medial cells of the somites, and the underlying endoderm), but not in the notochord itself (Figure 4f). zHDR was the only extrinsic pathway gene expressed specifically in the ICM (Figure 4c). Consistent with previous reports,24 we first observed zHDR transcripts at 16 h.p.f. in the posterior lateral plate mesoderm, which gives rise to the ICM. By 30 h.p.f., zHDR was strongly expressed in both the ICM and circulating erythroid cells.

Figure 4
figure 4

Embryonic gene expression. (a–d; g–l) Whole-mount ISH staining (blue) for the indicated genes at 8 h.p.f. (left), 16 h.p.f. (center), and 24–30 h.p.f. (right). 16 h.p.f. embryos are shown in dorsal view and 24–30 h.p.f. embryos are shown in lateral view. Anterior is to the left. The following abbreviations are used to indicate sites of expression: a, embryonic axis; ap, anal pore; fc, forerunner cells; fp, floorplate; icm, intermediate cell mass; lpm, lateral plate mesoderm; n, notochord; op, otic placode; ov, otic vesicle; pll, posterior lateral line; pnd, pronephric duct; s, shield. (b) Staining has been allowed to develop overnight to visualize low levels of transcript throughout the body. Similar nonlocalized staining was noted for zDL1a and zDL1b, but not a control probe, following overnight incubation. (e–f) Transverse sections illustrating notochord-associated expression (purple) of zHDR and zDL1a at 16 h.p.f. Sections have been counterstained with Nuclear Fast Red. Dashed red lines in (c) and (a) indicate plane of section in (e) and (f), respectively

Full size image

Additional sites of embryonic expression

We observed embryonic expression of zebrafish extrinsic pathway genes in a number of tissues in addition to the notochord and ICM (Figure 4a-l). Only zDL3, zFasL, zTNF1, zTNF2, and zFas were not seen by whole-mount ISH during the first 48 h of development (not shown), although in some cases these transcripts were detected by RT-PCR (Supplementary Figure 3). We found strong expression of zDL1b, zDL2, and zOTR in cells of the lateral line (Figure 5a), a mechanosensory system used by fish and amphibians to detect water displacement.28 The lateral line is composed of individual sense organs called neuromasts that originate in the otic region and become spread along the body axis by long-range migration.29 We detected zOTR in the otic placode at 16 h.p.f., and both zOTR and zDL1b in the migrating posterior lateral line (PLL) primordia at 30 h.p.f. (Figure 4d and g). By 48 h.p.f., zOTR , zDL1b, and zDL2 were observed in all 7–8 PLL neuromasts and all cranial lateral line neuromasts (Figure 5a). Sectioning confirmed neuromast expression of zOTR and revealed high levels of zDL1a, zDL1b, zDL2, zOTR, and zTNFR1 in the otic vesicles at 48 h.p.f. (Figure 5b; zDL1b and zDL2 not shown). Both the otic vesicles and the lateral line neuromasts employ hair cells similar to those of the mammalian inner ear to detect stimuli.30 In light of the observed expression patterns, it is interesting to note that zebrafish hair cells undergo constant, apoptosis-mediated turnover.31

Figure 5
figure 5

Expression in the developing lateral line and CNS. (a) 48 h.p.f. embryos stained by whole-mount ISH (blue) for zDL1b, zDL2, and zOTR. Arrows indicate expression in posterior lateral line neuromasts and red arrowheads indicate expression in cranial lateral line neuromasts. (b) Transverse sections through 48 h.p.f. embryos stained by whole-mount ISH (purple) for zDL1a, zHDR, zOTR, and zTNFR1. zHDR is not detected in the body at this stage and serves as a background control. Sections have been counterstained with Nuclear Fast Red. The following abbreviations are used to indicate sites of expression: ba, branchial arch; di, diencephalon; hb, hindbrain; len, lens; mes, mesencephalon; nm, neuromast; ov, otic vesicle; pnd, pronephric duct; ret, neural retina; sc, spinal chord

Full size image

We detected expression of several zebrafish TNF superfamily ligands and receptors in the developing central nervous system (CNS). The earliest of these is zDR6, which was expressed in the floorplate of the neural tube at 16 h.p.f. and in the forebrain, midbrain, and hindbrain at 30 h.p.f. (Figure 4h). Other genes, including zDL1a, zDL1b, zDL2, zOTR, and zTNFR1, were expressed in the developing brain by 24–30 h.p.f. (Figure 4a, b, d, g, and i). We confirmed localization of TNF and TNFR transcripts to specific regions of the CNS in sectioned embryos (Figure 5b). For example, zDL1a was localized to the ventricular surface of the diencephalon, mesencephalon, and hindbrain, while zTNFR1 was more broadly expressed throughout these regions as well as in the spinal cord. We also noted expression in the developing eye, with zOTR being localized to the boundary between lens and retina, while zDL1a and zTNFR1 were expressed throughout the neural retina.

zDL1a, zDL1b, and zDL2 signal through zHDR and zOTR

The ability of zebrafish Apo2L/TRAIL homologs to induce apoptosis specifically in embryonic tissues expressing zHDR and zOTR suggested that these proteins may represent ligand-receptor pairs. To test this, we generated soluble receptor constructs in which ECDs were fused to the Fc region of immunoglobulin G1 and soluble ligands that were tagged at the N-terminus with the FLAG epitope. We were able to express FLAG-tagged zDL1a, zDL1b, zDL2, and zTNF1, as well as Fc-tagged zHDR, zOTR, and zTNFR1, in HEK-293 cells. In co-immunoprecipitation studies (Figure 6a), zDL1a bound both zHDR-Fc and zOTR-Fc, with apparent preference for zOTR. zDL2 also bound both receptors, although less strongly than an equivalent amount of zDL1a. In contrast, zDL1b selectively bound zHDR-Fc and not zOTR-Fc. Neither zDL1a, zDL1b, nor zDL2 interacted with zTNFR1-Fc, while zTNF1 did, confirming that binding was specific.

Figure 6
figure 6

Characterization of the extrinsic apoptosis pathway in zebrafish. (a) Zebrafish Apo2L/TRAIL homologs directly and selectively bind to zHDR and zOTR. Soluble FLAG-tagged ligands and receptor-Fc fusion proteins were expressed in HEK-293 cells. Conditioned media from ligand and receptor-expressing cells were mixed and immunoprecipitated using anti-FLAG M2 agarose beads. The Fc region of co-precipitating receptors was detected using an anti-human IgG-specific antibody. (b) Knockdown of zHDR prevents zDL1b-mediated notochord apoptosis while knockdown of zOTR prevents zDL1a and zDL2-mediated notochord apoptosis. Embryos were injected with 500 pg of zDL1a or zDL2 or 200 pg zDL1b. Five nanograms of zHDR-MO or zOTR-MO were coinjected as indicated. Only embryos with unambiguous reductions in tail length at 30 h.p.f. were scored as short-tail. (b, d) N=number of independent experiments; n=number of embryos scored. Error bars indicate standard error of the mean between experiments. (c) zDL1b-mediated apoptosis can be rescued by zHDR-MO, CrmA, and zBlp2, but not by zOTR-MO. Embryos were coinjected with zDL1b (200 pg) and zHDR-MO (5 ng), zOTR-MO (5 ng), CrmA (500 pg), or zBlp2 (500 pg) as indicated. Whole-mount o-dianisidine staining for hemoglobin was carried out at 48 h.p.f. (d) zDL1a and zDL1b-mediated notochord apoptosis can be inhibited by zFLIP or CrmA or by knockdown of zFADD. Embryos were injected with zDL1a (500 pg) or zDL1b (200 pg). zFADD-MO (5 ng), zTRADD-MO (5 ng), CrmA (500 pg), or zFLIP (500 pg) were coinjected as indicated. Only embryos with unambiguous reductions in tail length at 30 h.p.f. were scored as short-tail

Full size image

We analyzed these interactions in vivo using antisense morpholino oligonucleotides (MOs) targeting zHDR (zHDR-MO) and zOTR (zOTR-MO). Both MOs rescued overexpression of their respective mRNAs (Supplementary Figure 2b) and neither had nonspecific toxicity at doses as high as 10 ng (not shown). Approximately, 70% of embryos coinjected with zDL1a and a control MO developed short tails by 30 h.p.f. Coinjection of zHDR-MO did not diminish this frequency, but coinjection of zOTR-MO reduced the number of short-tailed embryos to <10% (Figure 6b). zDL1b showed the reciprocal receptor specificity: zHDR-MO almost completely blocked zDL1b-induced apoptosis in the notochord and ICM while zOTR-MO had no effect (Figure 6b and c; Supplementary Figure 2a). A dominant-negative (DN) version of zHDR (zHDR-DN) lacking the cytoplasmic DD also inhibited the action of zDL1b (not shown). Similar to zDL1a, apoptosis induced by zDL2 was blocked by zOTR-MO but not by zHDR-MO (Figure 6b). Thus, zDL1a and zDL2 signal in vivo through zOTR, while zDL1b signals through zHDR. This is consistent with the observation that all three ligands trigger apoptosis in the notochord (which expresses both zHDR and zOTR), while only zDL1b induces apoptosis in the ICM (which expresses zHDR).

zDL1a and zDL1b require zFADD and zCaspase-8-like proteases, but not zTRADD, to induce apoptosis

We assessed the importance of zFADD, zTRADD, and caspase-8-like proteases for apoptosis induction by Apo2L/TRAIL homologs. We inhibited zFADD with a splice-blocking MO (zFADD-MO) that prevents excision of the large intron between the first and second exons. Coinjection of zFADD-MO with either zDL1a or zDL1b significantly attenuated the short-tail phenotype (Figure 6d; Supplementary Figure 2a). We used a standard MO, targeting the start codon, to inhibit zTRADD (zTRADD-MO). Although coinjection zTRADD-MO fully rescued embryonic mortality arising from zTRADD overexpression (Supplementary Figure 2b), it failed to inhibit zDL1a or zDL1b-induced apoptosis (Figure 6d). The presence of multiple zebrafish caspases with protease-domain homology to caspase-8 and -10 (Supplementary Figure 1c) precluded MO knockdown. Instead, we used zFLIP and the antiapoptotic cowpox virus serpin cytokine response modifier A (CrmA). In mammals, c-FLIP interferes with DISC recruitment and activation of caspase-8,32 while CrmA inhibits caspase-8,33 caspase-10,34 and caspase-135 (the latter is involved in inflammation rather than apoptosis). Coinjection of zFLIP or CrmA significantly reduced zCaspase-8a-mediated embryonic mortality while having no effect on apoptosis induced by zCaspase-2 or the proapoptotic BH3-only protein zNoxa (Supplementary Figure 4; zFLIP not shown). This confirmed that both proteins function similarly in zebrafish and mammals. Coinjection of CrmA with zHDR or zOTR blocked receptor-induced embryonic mortality (Supplementary Figure 4). Moreover, coinjection of zFLIP or CrmA with zDL1a or zDL1b substantially reduced the incidence of the short-tail phonotype and, with zDL1b, restored circulating erythrocytes (Figure 6c and d; Supplementary Figure 2a). Hence, zDL1a and zDL1b-induced apoptosis in zebrafish requires both zFADD and caspase-8-like enzymes but not zTRADD.

Zebrafish development proceeds normally without the extrinsic apoptosis pathway

To determine the importance of the extrinsic apoptosis pathway in zebrafish embryogenesis, we injected embryos with zHDR-MO, zOTR-MO, zFADD-MO, or with mRNA encoding zHDR-DN or CrmA. In all cases, we found no gross morphological abnormalities during the first week of development. This was true even for simultaneous inhibition of zHDR and zOTR (Supplementary Figure 5a). There were no apparent defects in expression of tissue-specific markers for hemangioblasts, vasculature, notochord, floorplate, spinal cord neurons, neuromasts, the pronephric duct, or the myoseptum (not shown). There was also no alteration in tail length, confirming normal notochord development (Supplementary Figure 5b). Circulation, heart looping, and overall cardiac morphology appeared normal at 48 h.p.f. and o-dianisidine staining confirmed that fully differentiated erythrocytes were present. Thus, while the extrinsic pathway is operative in the zebrafish embryo, it is not crucial for normal development.

Ligand-induced apoptosis in the notochord and ICM requires both the extrinsic and intrinsic pathways

In some mammalian cell types, ligand-induced apoptosis requires amplification of the extrinsic signal by the intrinsic pathway,36 which can be blocked experimentally by overexpression of antiapoptotic members of the Bcl-2 gene family. Coinjection of zBlp1 (a homolog of Bcl-xL37) with zHDR or zOTR substantially reduced embryonic mortality at 8 h.p.f. (Figure 7a). Coinjection of zBlp1 also protected from death induced by enforced expression of zCaspase-8a, albeit only partially. An incomplete rescue is not surprising, given that high levels of activated caspase-8 in mammalian cells circumvent the requirement for the intrinsic pathway. Coinjection of zBlp1 or zBlp2 (a homolog of Bcl-238) with zDL1a or zDL1b markedly attenuated the short-tail phenotype and prevented notochord and ICM apoptosis (Figures 6c, 7a and b; Supplementary Figure 2a). Similarly, MO-mediated inhibition of zebrafish Mcl-1 homologs sensitized normally resistant embryonic tissues to zDL1b-induced apoptosis (Kratz et al., accompanying manuscript). These results suggest that the identified zebrafish ligands require crosstalk between the extrinsic and intrinsic pathways to trigger apoptosis in embryonic tissues.

Figure 7
figure 7

Zebrafish ligand-induced apoptosis requires the intrinsic apoptosis pathway. (a) The zebrafish Bcl-xL ortholog zBlp1 inhibits apoptosis induced by components of the extrinsic pathway. Embryos were injected with indicated mRNAs either alone or in conjunction with of zBlp1 (500 pg). Embryonic death induced by zHDR (300 pg), zOTR (300 pg), zCaspase-8a (300 pg), or zNoxa (100 pg) was scored at 8 h.p.f. (left panel) and the short-tail phenotype induced by zDL1a (500 pg) or zDL1b (200 pg) was scored at 30 h.p.f. (right panel). Only embryos with unambiguous reductions in tail length were recorded as short-tail. N=number of independent experiments; n=number of embryos scored. Error bars indicate standard error of the mean between experiments. (b) Zebrafish notochord cells and erythroblasts require the intrinsic apoptosis pathway for ligand-mediated killing. Embryos were injected with zDL1b (200 pg) alone or in combination with zBlp1 (500 pg). Embryos were photographed at 48 h.p.f. (right panels) or fixed at 30 h.p.f. for whole-mount ISH to detect the erythrocyte marker globin αe1 (left panels)

Full size image

zDL1b-induced ICM apoptosis is limited to the erythroid lineage

The ICM is the functional equivalent of the mammalian yolk sac blood islands and contains vascular, erythroid, and myeloid precursors at various stages of differentiation.39 To determine which ICM cell types respond to zDL1b, we performed whole-mount ISH and quantitative PCR (QPCR) analysis for an array of hematopoietic and vascular markers (Figure 8a). Injection of zDL1b led to a nearly complete loss of mature erythrocytes expressing embryonic α- and β-globin subunits (Figure 8b and c). In contrast, zDL1b overexpression had no effect on scl, flk1, and gata2, which are expressed much earlier in the hematopoietic program and associated with hemangioblasts, hematopoietic stem cells, and angioblasts (Figure 8b). The first unambiguous marker of erythroid commitment is the transcription factor gata1.40 Whole-mount ISH and QPCR analysis on zDL1b-injected embryos revealed complete loss of gata1 (Figure 8b and c), demonstrating that hematopoietic cells become sensitive to zDL1b-mediated apoptosis only as erythroid differentiation begins. zDL1b did not affect expression of the myeloid-specific markers draculin (not shown), pu.1 (Figure 8b), or L-plastin (Figure 8c) in either the ICM or the rostral blood island (RBI). Hence, even among cells that adopt a hematopoietic fate, only erythroid precursors respond to zDL1b by apoptosis.

Figure 8
figure 8

zDL1b selectively induces erythroblast apoptosis in the ICM. (a) Diagram of primitive hematopoiesis in zebrafish, illustrating characteristic markers. (b) zDL1b overexpression leads to loss of marker genes for erythroblasts and mature erythrocytes. Embryos were injected with 200 pg of zDL1b and fixed at 30 h.p.f. for whole-mount ISH. zDL1b-injected embryos show ICM expression of early hematopoietic markers (flk1, scl, gata2) but not markers of more mature erythrocytes (gata1, globin αe1). The myeloid-specific transcription factor pu.1 is also unaffected. Arrowhead=RBI expression; arrow=ICM expression (c) QPCR analysis of relative gata1, globin βe1, and L-plastin mRNA levels in zDL1a and zDL1b-injected embryos. Embryos were injected with zDL1a (500 pg) or zDL1b (200 pg) and total mRNA was isolated at the indicated times. mRNA levels are expressed as the fold difference between ligand-injected and control embryos (control embryos were injected with 500 pg EGFP mRNA and have a value of 1)

Full size image

Discussion

The cell-extrinsic apoptosis pathway is well characterized in mammals, but has remained poorly defined in other vertebrates. Drosophila studies show that the role of the TNF superfamily in apoptosis predates the chordate lineage. However, TNF superfamily ligands use different pathways to activate initiator caspases in fruit flies versus mammals.8 By systematically characterizing the extrinsic pathway in zebrafish, we have shown that its various basic components are highly conserved throughout vertebrate evolution (Figure 9). As in mammals, apoptosis induction through the extrinsic pathway requires FADD and at least one caspase-8-like protease. Like certain mammalian cell types, ligand-induced apoptosis in the zebrafish notochord and ICM requires crosstalk between the extrinsic and intrinsic pathways. These similarities between distant species suggest that the extrinsic apoptosis pathway coopted FADD and caspase-8 early in vertebrate evolution.

Figure 9
figure 9

Schematic representation of established ligand/receptor interactions and the intracellular machinery required for effector caspase activation in zebrafish. zDL1a binds both zHDR and zOTR in vitro but signals exclusively through zOTR when overexpressed during embryogenesis. Ligand-mediated apoptosis in the notochord and ICM can be blocked by inhibition of of caspase-8-like proteases (using CrmA or zFLIP) or the cell-intrinsic apoptosis pathway (using zBlp1 or zBlp2)

Full size image

Of the four zebrafish Apo2L/TRAIL homologs, zDL2 is most structurally similar to mammalian and avian Apo2L/TRAIL, while zDL1a and zDL1b are most homologous to a chicken ‘Apo2L/TRAIL-like’ protein (Accession #AAL23702) (Figure 1a). Chicken Apo2L/TRAIL-like, zDL1a, and zDL1b share two additional ECD cysteine residues (corresponding to L147 and Y209 in human Apo2L/TRAIL) that are not found in other homologs.41 We propose that the ancestral gene was duplicated before the divergence of ray-finned and lobe-finned fishes ∼400 million years ago, giving rise to DL1 (Apo2L/TRAIL-like) and DL2 (Apo2L/TRAIL) variants. A further chromosomal doubling event within the ray-finned fish lineage42 presumably duplicated the DL1 gene locus, producing DL1a and DL1b. Modern teleosts and birds have retained both DL1 and DL2, while the DL1 variant is absent in at least some mammals including humans and mice.

We have established that three of the four zebrafish Apo2L/TRAIL homologs interact with zHDR and zOTR. This finding is somewhat surprising, as the ECDs of both receptors differ from those of the human Apo2L/TRAIL DD-containing receptors DR4 and DR5 and the decoy receptors DcR1, DcR2, and osteoprotegerin.43, 44 Thus, receptors for this ligand do not necessarily share a conserved ECD architecture in nonmammalian species.

Mice lacking FADD, c-FLIP, or caspase-8 develop abnormal cardiac morphology between embryonic days 9.5 and 10.5 and die soon thereafter.11, 13, 14 Zebrafish embryos reach equivalent stages of heart development by ∼48 h.p.f.45 Various techniques for inhibiting the extrinsic pathway caused no cardiac defects or other morphological abnormalities in zebrafish embryos. The finding that embryonic defects in caspase-8 deficient mice can be rescued by ex vivo whole-embryo culture suggests that the primary requirement for caspase-8 in mammalian embryogenesis may be in extra-embryonic tissues rather than the embryo proper.14 Our analysis of the zebrafish, which develops ex utero and lacks extraembryonic tissues, supports the hypothesis that deficiencies in the extrinsic pathway impact very few, if any, tissues in the embryo proper.

To our knowledge, this is the first report that examines the sensitivity of embryonic tissues to ligand-induced apoptosis in any chordate. We found that full-length, transmembrane Apo2L/TRAIL homologs selectively induced apoptosis in the notochord and the ICM of zebrafish embryos. The observation that zDL1b overexpression induces apoptosis in embryonic erythroblasts is particularly intriguing, given that inhibition of zHDR stimulates erythropoiesis in adult zebrafish24 and caspase-8 deficient mice exhibit hyperemia during embryogenesis.10 Although we did not see defects in notochord development upon inhibition of the extrinsic pathway, the complementary expression patterns of zHDR and zOTR in the notochord and of zDL1a in the surrounding cells is intriguing. Perhaps zHDR, zOTR, and zDL1a prime the notochord to respond to as yet unidentified environmental cues during embryogenesis.

Materials and Methods

Data are provided as Supplementary Material.