Time course of programmed cell death in Ciona intestinalis in relation to mitotic activity and MAPK signaling


  • Raffaella Tarallo,

    1. Laboratory of Biochemistry and Molecular Biology, Stazione Zoologica ‘A. Dohrn’, Naples, Italy
    Search for more papers by this author
  • Paolo Sordino

    Corresponding author
    1. Laboratory of Biochemistry and Molecular Biology, Stazione Zoologica ‘A. Dohrn’, Naples, Italy
    • Laboratory of Biochemistry and Molecular Biology, Stazione Zoologica ‘A. Dohrn’, Villa Comunale, 80121, Naples, Italy
    Search for more papers by this author


Programmed cell death (PCD) in the ascidian species Ciona intestinalis (Tunicata; Chordata) is investigated from early larvae to juvenile stages, by means of digoxigenin-based terminal deoxynucleotidyl transferase-mediated biotinylated UTP nick end labeling (TUNEL) technique. At first, PCD in the swimming larva affects trunk mesenchyme and central nervous system (CNS), then it participates extensively to metamorphosis, until it is restricted to developing organs of juveniles. Analysis of patterns of cell death and division in the larval CNS question old models on the genesis of the adult C. intestinalis brain. Upon performing immunochemical and functional assays for mitogen-activated protein kinase (MAPK) kinase kinase-1 (MEKK1), MAPK kinase 1/2 (MEK1/2), c-Jun NH2-terminal kinase (JNK), and dual phosphorylated extracellular regulated kinase 1/2 (dpERK1/2), the neurogenic competence of the larval brain appears to rely on a combinatorial regulation of PCD by the mitogen-activated protein kinase signaling cascade. These results show that, in tunicates, PCD consists of a multistep program implicated in growth and patterning with various roles. Developmental Dynamics 230:251–262, 2004. © 2004 Wiley-Liss, Inc.


Programmed cell death (PCD) is a physiological mechanism by which cells or transient structures are eliminated during development, on the grounds of their position, identity, number, or condition (Whitten, 1969; Hurle, 1988; Sanders and Wride, 1995; Jacobson et al., 1997; Hogan, 1999; Vaux and Korsmeyer, 1999; Meier et al., 2000). A common mechanism for PCD is apoptosis, which displays a typical phenotype that promotes phagocytosis by macrophages, so as to avoid inflammatory responses (Steller, 1995). Due to the pivotal role of apoptosis during development, both PCD and apoptosis are used in this report to refer to the same mechanism of cell death. As principles and molecules are fundamentally conserved since primitive diploblastic organisms, apoptosis appears to have been acquired at an early stage of evolution (Bosch and David, 1984; Cikala et al., 1999). In animals, PCD ancestry is echoed by an orderly morphogenetic regulation, with a range of specializations and similarities among and within phyla (Roccheri et al., 1997; Hensey and Gautier, 1998; Pazdera et al., 1998; Cole and Ross, 2001; Rangers et al., 2001; Seipp et al., 2001; Voronina and Wessel, 2001; Bayascas et al., 2002). Studies on apoptosis in tunicates have mainly focused, in the past decade, on asexual development of colonial forms (e.g., Lauzon et al., 1992, 1993, 2002; Cima et al., 2003). Despite recent reports providing valuable information on specific PCD phenomena during embryogenesis of solitary ascidians (Chambon et al., 2002; Jeffery, 2002a, b), the general profile is poorly defined. In this perspective, the experimental accessibility of the solitary ascidian Ciona intestinalis allows detailed analysis of cell state configurations during embryogenesis, leading to substantial descriptive information (Sordino et al., 2001; Satoh et al., 2003). Moreover, a complete determination of apoptotic arrangements in this species may serve as a reference atlas for mutagenesis screenings aiming to identify genes involved in cell cycle regulation.

To correlate PCD with morphogenesis, and to identify cellular and molecular fundaments of cell death regulation in a basal chordate (Kowalevsky, 1866; Conklin, 1905; Satoh, 1978), we have systematically documented in detail apoptotic patterns until postmetamorphic juveniles (5 days postfertilization [dpf]) of C. intestinalis, taking advantage of a highly sensitive digoxigenin-based terminal deoxynucleotidyl transferase-mediated biotinylated UTP nick end labeling (TUNEL) protocol (see Experimental Procedures section). Our work revealed that embryonic PCD is not only associated with tail resorption, as stated by Chambon et al. (2002), but it broadly operates across tissues and organ primordia and that it is temporally and spatially organized in a way that evokes a coordinated developmental control.

Because the close integration between cell proliferation and apoptosis is required for morphogenesis, we have described relevant aspects of mitotic activity during development, providing evidence for a dynamic interplay of cell states instrumental in organogenetic programs, such as secondary neurogenesis and substrate adhesion. In particular, analysis of PCD and proliferation in the larval brain fosters the idea of an unknown neurogenic region in the posterior part of the sensory vesicle, besides the recognized role of anterior sensory vesicle and neck (van Beneden and Julin, 1884; Willey, 1893; Seeliger, 1904–1905). This important observation advises a renewed attention on the histogenesis of the adult brain in C. intestinalis.

The mitogen-activated protein kinase (MAPK, or Ras/MEK/ERK) signaling cascade encodes essential components of the cell death and cell proliferation processes. At the core of the MAPK enzymatic machinery, MAPK kinase kinase-1 (MEKK1) activates different downstream targets, including the dual phosphorylated extracellular regulated kinase 1/2 (dpERK1/2) and the c-Jun NH2-terminal kinase (JNK; Mohr et al., 1998; Zhang and Liu, 2002; Lin, 2003; Satou et al., 2003). In Ciona larvae, all three proteins were found to be expressed in apoptotic and nonapoptotic territories, with consistent diversity in topographic domains and translational levels. Perturbation by a pharmaceutical antagonist of MAPK kinase 1/2 (MEK1/2), an MEKK1-driven protein kinase that can phosphorylate and activate dpERK1/2 and JNK (Yan and Templeton, 1994), demonstrated the involvement of the MAPK genetic circuitry in neural survival and regression. Patterns and functions are discussed in terms of a complex molecular and cellular network, with insights into morphogenesis and evolution of developmental mechanisms.


To study PCD during indirect development of ascidians, we have performed a time course analysis of TUNEL-positive staining in C. intestinalis, from the earliest signs of cell death, occurring at 18 hours postfertilization (hpf) in unhatched larvae, until 5 dpf juveniles. Aiming to portray rapidly evolving processes, a 15-min sampling frequency was applied to larval stages. Moreover, methodological procedures allowed optimal microscopic resolution (Chambon et al., 2002; Jeffery, 2002b). To compare apoptosis with proliferation, we generated schematic atlases of cell division in the larval trunk at the beginning and near the end of the swimming period. Some molecular facets of cell death were explored by using antibodies directed against three key members of the MAPK signaling pathway, MEKK1, dpERK1/2, and JNK, and with pharmacological blockage of MEK1/2. Results were confirmed by using different embryo batches.

PCD in the Dorsal Mesenchyme of the Larval Trunk

The first observation of TUNEL labeling occurred in larvae about to attain free-swimming life (Fig. 1A). PCD was observed at the anterior margin of the dorsal mesenchyme, in a small median subset of trunk lateral cells (TLC). Seemingly, the median PCD pattern split within bilateral TLC populations, giving rise to two populations of apoptotic cells (Fig. 1B,D,E). As soon as the larvae abandoned the chorion, these clusters were detected on the lateral surface of the central nervous system (CNS), from the posterior sensory vesicle to the visceral ganglion (Fig. 1F). Then, apoptotic patterns progressively extended toward the ventral mesenchymal pouches (trunk ventral cell, TVC; 19 hpf; Fig. 1G–I). Remarkably, PCD domains were very similar in size and position on both trunk sides, reflecting a tight bilaterality. Afterward (20–21 hpf), TUNEL-labeled cells appeared rostrally on both sides of the trunk, in lateral grooves at the boundary between the endoderm and the sensory vesicle (Fig. 1J,K). Cell shape of these dying TLCs changed during development, as they became progressively elongated (compare Fig. 1J with K). Then, PCD in TLCs vanished in the swimming larva.

Figure 1.

Terminal deoxynucleotidyl transferase-mediated biotinylated UTP nick end labeling (TUNEL) patterns in trunk lateral cells (TLC) of Ciona intestinalis larvae. A: Stained mesenchymal cells in 18 hours postfertilization (hpf) larva (small arrowheads). B: Dying TLCs in early swimming larvae. C: Transmission electron microscope transverse section across the trunk of 19 hpf larva, with apoptotic corpses in TLCs (large red arrowheads). D: Cross-sections at the level of the posterior sensory vesicle with dorsal (tlc, arrows) and ventral (tvc) mesenchymal pouches. E,F: Bilateral clusters of cell death in TLCs (E) before TUNEL patterns ingress larval trunk and approximate central nervous system subdivisions (F). G–K: First (19 hpf), TUNEL staining extended posterior–ventrally (G–I) and then (21 hpf) rostrally (J,K, large black arrowheads). Anterior is left, dorsal is up except B and E, which are dorsal views. n, neck; no, notochord; nt, neural tube; oc, ocellus; ot, otolith; p, palps; pg, primordial gut; sv, sensory vesicle; tlc, trunk lateral cells; tvc, trunk ventral cells; vg, visceral ganglion.

Apoptosis can be diagnosed at the ultrastructural level by shape abnormalities due to membrane or cytoskeletal defects, as well as nuclear fragmentation in apoptotic corpses (Kerr et al., 1972). In this study, transmission electron microscope (TEM) analysis confirmed the apoptotic nature of cell death, as chromatin bodies were observed in dorsal–lateral mesenchymal cells (Fig. 1C).

Ordered Series of Apoptotic Events During Development

During swimming life, larval trunks elongate, while organ primordia become distinguishable. Re-activation of PCD was observed shortly before settlement (22 hpf), at the posterior–dorsal margin of the sensory vesicle, in five to six photoreceptor or ependymal cells above the photosensitive organ (the ocellus; Fig. 2A). Then PCD expanded from the ocellar focus toward the specialized ependymal cells near the gravity-sensitive organ (the otolith; Nicol and Meinertzhagen, 1991) and caudally to the posterior sensory vesicle but not in the neck (Fig. 2B). Hence, apoptosis progressed caudally across the nervous system, extending in the visceral ganglion and through the nerve cord (Fig. 2B). With regard to histology, ocellar cells were all positive for TUNEL staining; in turn, the posterior sensory vesicle and the visceral ganglion were labeled only in the external layer (Fig. 2C–E). Cell death was also detected at the base of the palps (Fig. 2B). After the first PCD wave in the nervous system, a focal burst of apoptosis at the tail extremity generated a second, caudal-to-rostral, wave (in agreement with Chambon et al., 2002). Meanwhile, apoptosis expanded to mesenchyme and endoderm of the trunk in newly settled larvae (25 hpf; Fig. 2G), followed by tail withdrawal. During this period, PCD was still confined to the outer layer of the posterior sensory vesicle (Fig. 2F). Within a few hours, the tadpole larva is replaced by a barrel-shaped stage. Apoptosis was still intense in endodermal lobes and CNS, in the forming digestive system, and in a distinct rectangle of epidermis above the posterior lacuna of the trunk, where tail residues accumulate (Fig. 3A). At the same time, cell death evenly affected mesenchymal cells throughout the body. By 5 dpf, PCD was restricted to stomach, branchial slits, and siphonal borders, whereas it dropped to background levels elsewhere, e.g., in the endostyle (Fig. 3B,C). Noticeably, juvenile brain exhibited a few scattered TUNEL-stained nuclei (Fig. 3C).

Figure 2.

Cell death in Ciona intestinalis larvae at (A–E) 22 hours postfertilization (hpf) and 25 hpf (F,G). A,B: A wave of apoptosis affects the central nervous system of larvae at 22 hpf. Death signal progression starts in dorsal cells of the posterior sensory vesicle neighboring the ocellus (arrow in A), then it diffuses posteriorly to affect visceral ganglion and neural tube (asterisks in B). Note that the neck junction remains clear of terminal deoxynucleotidyl transferase-mediated biotinylated UTP nick end labeling (TUNEL) labelings. A strong signal characterizes the base of the palps (arrowhead). C–E: Transversal sections of larval trunks at the level of ocellus (C), posterior sensory vesicle (D), and visceral ganglion (E). F: Same as in D at 25 hpf, with persisting absence of cell death in dorsal and ventral neurons. G: Late larval stage features diffusion of cell death to mesenchyme and endoderm. A, B, and G, anterior is left, dorsal is up. el, endodermal lobes; psv, posterior sensory vesicle; me, mesenchyme. For other abbreviations, see legend to Figure 1.

Figure 3.

Patterns of apoptosis in whole body during metamorphosis and in early juvenile ascidiae. A: At 2 days postfertilization (dpf), a peculiar programmed cell death (PCD) domain labels a sharp rectangle of ectodermal cells at the posterior end of the trunk (arrow). Cell death is strong in endoderm. B,C: In 5 dpf juveniles, localized apoptosis affects stomach, branchial slits (asterisks in C), and openings of the siphons (arrows in B). Sparse apoptotic nuclei are observed in endostyle (B,C) and brain (C). b, brain; en, endostyle; os, oral syphon; s, stomach; td, tail debris. A, proximal is left, otherwise is down.

Spatiotemporal Analysis of Cell Proliferation

To search for spatiotemporal relationships between PCD and mitosis, we used 5′-bromo-2′-deoxy-uridine (BrdU) incorporations (Fig. 4A,C–F) and phosphorylated histone 3 (PH3) antibody (Fig. 4B,G–I) to identify dividing cells. A semiquantitative profile of cell divisions was obtained by plotting PH3-positive nuclei distribution over trunk anatomy (Fig. 5A,B). The BrdU protocol was used to estimate slow mitotic events (Fernandez et al., 2001; Salas-Vidal et al., 2001). For the sake of clarity, proliferation in the epidermis was debarred from the analysis.

Figure 4.

Mitotic activity in Ciona developmental stages, as revealed with bromodeoxyuridine (BrdU; A,C–F) pulses and phosphorylated histone 3 (PH3) antibody (B,G–I), that recognize S phase nuclei in living and fixed samples, respectively. A,B: Examples of labelings with a 30-min BrdU pulse in an early larva (18.5 hours postfertilization [hpf], A), and with PH3 antibody in a mature larva (23 hpf, B). C,D: Dorsal–lateral view (C) of 1-hr pulse BrdU signals and corresponding transversal section (D), showing cell proliferation in two columns of dorsal neurons (arrows) placed between posterior sensory vesicle and neck. E: One-hour BrdU labeling in cells of the endodermal strand, the only dividing tissue of the tail. F: Double staining with terminal deoxynucleotidyl transferase-mediated biotinylated UTP nick end labeling (TUNEL; dark) and 2-hr BrdU pulse (brown) showing densely intermingled signals in 23 hpf trunk. The anterior margin of the sensory vesicle (arrowhead) and the ocellus (arrow) differ in being proliferative and apoptotic, respectively. G: PH3 signal in presumptive precursor cells of gill slits during metamorphosis (2 days postfertilization [dpf]). H,I: Cell division is present in stomach (3 dpf, H) and endostyle (5 dpf, I). In G, H, and I, proximal is down. es, endodermal strand; gsp, gill slit primordium; td, tail debris. For other abbreviations, see legend to Figure 1.

Figure 5.

A,B: Semiquantitative maps of mitotic figures in larvae at early (A; 19 hours postfertilization [hpf]; average 10.3 phosphorylated histone 3–positive nuclei per larva; total larvae = 31) and late (B; 23 hpf; 45.9; 30) stages. S phase ectodermal cells were omitted. Anterior is left, dorsal is up. Green, endoderm; red, mesenchyme; blue, neuroectoderm; yellow, palps.

Trunks of early (19 hpf) and late (23 hpf) larvae differed in terms of organogenetic rhythms and asymmetries of tissue growth. At 19 hpf, PH3-labeled nuclei were mostly clustered at the periphery of the endoderm. Concomitantly, dividing cells were detected in the anterior sensory vesicle, neck, and, to less extent, posterior margin of the sensory vesicle and visceral ganglion (Figs. 4A, 5A). In TLC pouches, PH3 signals were mainly confined to restricted areas at the anterior and posterior–ventral boundaries (Fig. 5A). Few hours later (23 hpf), proliferation had diffused into endoderm, mesenchyme, and at the tip of the palps (Figs 4B, 5B). Mitotic nuclei were abundant in both anterior and posterior sensory vesicle, separated by nondividing ependymal cells lining the neural cavity, and in the neck. Cell division was rarely detected caudal to the visceral ganglion, and never in the nerve tube (Figs 4B, 5B). The endodermal strand was indeed the only mitotic tissue of the tail, as seen with 1-hr BrdU exposures (Fig. 4E). Notably, two columns of mitotic neurons extended dorsally from the posterior margin of the sensory vesicle to the neck region (Fig. 4C,D; Bollner and Meinertzhagen, 1993). At a mid-metamorphosis stage (3 dpf), synchronous cell divisions appeared in developing branchial slits (Fig. 4G). Proliferation occurred in stomach and, less intensively, in the endostyle of juveniles (Fig. 4H,I). The notion that cell death and proliferation patterns do overlap in tissues and organs of larvae was supported by double labelings with a 2-hr BrdU pulse followed by TUNEL technique at 23 hpf (Fig. 4F; data not shown).

MAPK Factors in the Developmental Control of PCD

To gain further insights into the MAPK cascade-dependent mechanisms that govern cell death in ascidian larvae (Chambon et al., 2002), we have correlated PCD distribution with immunostaining patterns of three MAPK enzymes, i.e., MEKK1, dpERK1/2, and JNK (Fig. 6). Moreover, we analyzed the effects of MEK1/2 inhibition on cell death at larval stages (Chambon et al., 2002). In 22 hpf C. intestinalis larvae, a cytoplasmic MEKK1 signal was observed in innermost cells of the pharynx (primitive mouth, or stomodeum), in anterior and posterior edges of the sensory vesicle, at the base of the palps, and in the neck region (Fig. 6A, data not shown). A very intense MEKK1 label characterized several neurons and axons of the visceral ganglion (Fig. 6A), wherein transversal sections showed immunopositivity in bilateral pairs of presumptive motoneurons (Fig. 6B). MEKK1 epitope aggregates were also found on nuclear membranes of cells in the endodermal strand (Fig. 6C). Likewise, synthesis of dpERK1/2 takes place in the cytoplasm of proximal palp cells, stomodeum, anterior and posterior sensory vesicle, and neck (19 hpf; Fig. 6D). Moreover, nuclear dpERK1/2 staining occurred in the epidermis overlying the sensory vesicle, atrial primordia, and muscle, notochord, and skin of the tail (22 hpf; Fig. 6E–G). Similar to MEKK1 and dpERK1/2, JNK activity was detected in pharyngeal rudiment, anterior sensory vesicle, neck, and internal neurons of the posterior sensory vesicle (Fig. 6H,I). All three epitopes were observed in endodermal lobes and the immature gut (Fig. 6A,D,H).

Figure 6.

Cytoplasmic and nuclear patterns of mitogen-activated protein kinase kinase-1 (MEKK1, A–C), dual phosphorylated extracellular regulated kinase 1/2 (dpERK1/2, D–G), and c-Jun NH2-terminal kinase (JNK, H,I) antibodies at 19 hours postfertilization (hpf, D) and 22 hpf (A–C,E–I), include several anatomical domains. A: Cytoplasmic MEKK1 labeling in a small domain at the junction between presumptive neurohypophysis and pharynx, in posterior sensory vesicle, neck (data not shown) and visceral ganglion. B: Resin section showing MEKK1 signal in two putative motoneurons of the visceral ganglion. C: MEKK1 protein synthesis in the endodermal strand. D: At 19 hpf, cytoplasmic dpERK1/2 is detected in anterior and dorsal margin of sensory vesicle, pharynx, posterior sensory vesicle, and neck region. E–G: Nuclear dpERK1/2 signal labels epidermal nuclei (asterisks in E) in a circumscribed area of the dorsal trunk above the sensory vesicle (E), as well as in muscle (F), notochord (arrows in G), and epiderm of caudal appendages. H,I: Like MEKK1 and dpERK1/2, JNK is expressed also at the rostral edge of the sensory vesicle, pharynx, and neck area. I: Of interest, JNK is observed only in dorsal and ventral neurons of the posterior sensory vesicle (long arrows; Nicol and Meinertzhagen, 1991), with immunopositive axons elongating into the posterior divisions of the larval central nervous system (red arrowhead). Despite repeated immunostainings, background did not permit clearer illustrations. All three proteins are translated in the larval endoderm. All panels are lateral views except for I, which is dorsal. asv, anterior sensory vesicle; ep, epidermal layer; es, endodermal strand; mu, musculature; ph, pharynx. For other abbreviations, see legend to Figure 1.

With the aim to functionally address MAPK regulation of cell death, we incubated embryos of different developmental stages with the MEK1/2 inhibitor, U0126. Larvae that were treated from middle tail bud stage (13–30 hpf) displayed a ubiquitous repression of PCD (Fig. 7B,C). However, ectopic TUNEL-positive nuclei were sometimes recorded in anterior and posterior walls of the sensory vesicle (average 15% larvae; n = 50; three experiments; Fig. 7B,C). When incubation started at late tail bud (from 16 to 34 hpf) or hatching stage (18–45 hpf), apoptosis was not affected in the trunk. On the other hand, TUNEL signals were observed in endodermal strand, neural tube, and epidermal sensory neurons (Fig. 7A), that is, in cell types that do not normally express dpERK1/2 (Fig. 6F,G). In all experiments, metamorphosis was arrested in 70% of the larvae.

Figure 7.

Pharmacological perturbation of mitogen-activated protein kinase (MAPK) protein kinase pathway. A: Eighteen hours postfertilization larvae incubated from late tail bud stage with a MAPK kinase 1/2 (MEK1/2) inhibitor, U0126, showed no effects in the trunk (data not shown). On the other hand, caudal apoptosis was suppressed only in mesodermal tissues and epidermis, while it was still detected in nerve cord (arrowheads), presumptive epidermal sensory neurons (black arrows) and endodermal strand (white arrows). B,C: Incubation with U0126 from middle tail bud stage caused uniform suppression of apoptotic responses. Terminal deoxynucleotidyl transferase-mediated biotinylated UTP nick end labeling (TUNEL) signals were occasionally observed in cells located in anterior (red arrowhead in B) and posterior sensory vesicle (white arrowhead in C).


In Tunicata, apoptosis has been shown to be important in blastogenetic and sexual development in the class Ascidiacea (Mukai and Watanabe, 1976; Burighel and Schiavinato, 1984; Lauzon et al., 1992; Cima et al., 2003). Originally, its role was described in tail resorption of the colonial ascidian Botryllus schlosseri with cytological criteria (Schiaffino et al., 1974). The present report portrays PCD during development of the solitary species C. intestinalis, together with a discussion on induction and morphogenetic consequences. Higher sensitivity of the digoxigenin-based TUNEL method allowed us to reveal previously undescribed aspects (compare with Chambon et al., 2002). Salient results of this study consist of (1) a complete map of PCD modules until beyond metamorphosis; (2) dissection of cell death and proliferation territories in the larval brain, with support for a novel neurogenic area; (3) spatial and functional correlation of PCD with MAPK signal transduction cascade, with emphasis on the larval CNS of C. intestinalis.

PCD in the Larval Trunk

The timing and mode of PCD activation during embryonic development are variable among tunicate species (Meinhertzhagen and Okamura, 2001; Chambon et al., 2002; Jeffery, 2002b; our unpublished data). In C. intestinalis, the first signs of apoptosis are detected in the dorsal mesenchyme of early larvae, within two bilateral pouches known as TLCs (Fig. 1), that differentiate into blood and muscles (Nishida, 1987; Mita-Miyazawa et al., 1987; Nishide et al., 1989; Hirano and Nishida, 1997). In most cases, PCD patterns on both sides of the trunk are almost mirror-like images, suggesting a coordinated and strict regulation of bilateral symmetry. Notably, cell proliferation at the dorsal and posterior–ventral margins of TLC clusters (Fig. 5A) is rather consistent with tissue growth, whereby physiological death of TLCs is putatively linked to homeostasis. Concerning the possibility that PCD is required in the context of specific TLC subclasses, surprisingly little is known about the architecture of mesenchymal pouches, albeit ultrastructural studies have formerly revealed a degree of cytological differentiation (Mancuso and Dolcemascolo, 1981; Mancuso and Giancuzza, 1983; Mancuso, 1986). While many issues concerning PCD function in TLCs remain open, it would be useful to identify which factors do specifically sensitize mesenchymal cells to undergo the cell death program.

On the other hand, evidence of mesenchymal PCD beneath the sensory vesicle is suggestive of TLC detachments. In vertebrate embryos, mesenchymal stromal cells of the kidney or retinal cells in the eye lose their connections and migrate to create a space for future events before their death (e.g., Koseki et al., 1992; Biehlmaier et al., 2001; Cole and Ross, 2001; Bard, 2002). Similarly, apoptotic TLCs could provide a physical environment for the morphogenetic agenda. A further explanation is that cell displacement might mechanically interrupt signaling between endoderm and neuroectoderm. Moreover, apoptosis in dorsal mesenchyme and in the adjacent CNS are separated by less than an hour, making it premature to rule out a link between these events. Indeed, in chick embryos, PCD in the neural tube is thought to coincide with tube closure and the loss of skin-derived signals after detachment from the overlying ectoderm (Glucksmann, 1951; Homma et al., 1994; Sanders and Wride, 1995). Mesenchymal PCD could serve an analogous role in ascidians.

PCD in the Larval Brain: Developmental Rules and Secondary Neurogenesis

Preceding larval settlement, the sensory vesicle witnesses the concomitant reactivation of apoptosis in a small subdivision of cells above the ocellus, the light-sensitive system made of one pigment-cup, 3 lens, and 17 retinal cells (Fig. 2A,C; Eakin and Kuda, 1971; Nicol and Meinertzhagen, 1991). Like in fish and amphibians, where developing eyes are early rostral PCD hallmarks (Hensey and Gautier, 1998; Cole and Ross, 2001), the ascidian ocellus defines a sharp physical boundary of neural cell death. Moreover, the photosensitive organ represents the first embryonic organ to undergo apoptosis in Ciona, and eventually to regress (Fig. 2C).

Within a short developmental window, most of the larval CNS is affected by an anterior–posterior wave of apoptosis, that emanates from the ocellus until the end of the neural tube, with some exceptions (Fig. 2B). Therefore, two apoptotic waves occur during development of the ascidian larva, first in the CNS (this study) and soon after in the tail (Chambon et al., 2002). While PCD in the neuroectoderm might spread by means of synaptic transmission (Fig. 6C,I), the caudal wave could be sustained by other processes, such as contact-mediated diffusion of molecular cues, depletion of growth factors, or mechanical forces (Cloney, 1978). However, directional cell death illustrates an ancestral strategy of metazoans, as linear progression is observed also in the embryonic nervous system of vertebrates and invertebrates (Abrams et al., 1993; Hensey and Gautier, 1998; Pazdera et al., 1998; Cole and Ross, 2001; Seipp et al., 2001).

Based on classic studies, two regions of the larval CNS, anterior sensory vesicle and neck, take part in the formation of the mature Ciona brain. First, it is widely assumed that a neurohypophysial duct opens through the pharynx and the anterior wall of the larval sensory vesicle to form the neural gland (Figs. 2B, 4A,F, 5A,B; e.g., Willey, 1893). This region has been proposed to be homologous with the pituitary gland, or neurohypophysis, of vertebrates (Julin, 1881; Elwyn, 1937; Lacalli and Holland, 1998; Lacalli, 2001; Meinertzhagen and Okamura, 2001), with the pharyngeal primordium corresponding to the adenohypophysis and the olfactory system (Willey, 1893; Torrence, 1983; Manni et al., 1999; Sorrentino et al., 2000; Romanov, 2000; Boorman and Shimeld, 2002). A second neurogenic entity refers to dorsal mitotic cells placed in the neck junction (Meinertzhagen et al., 2000), which have been claimed to contribute to the formation of ganglion rudiment and dorsal strand plexus of the adult Ciona brain (van Beneden and Julin, 1884; Willey, 1893; Seeliger, 1904–1905; Bollner and Meinertzhagen, 1993; Lacalli, 2001; Figs. 2B, 4C). Of interest, secondary neurogenesis in ascidian tunicates involves phylogenetic divergence in terms of patterning and ontogenetic strategies (see Lacalli and Holland, 1998). For instance, the neurohypophysial duct of the colonial species Botryllus schlosseri is considered to be the single source of brain precursors, generating both neural gland and cerebral ganglion (Burighel et al., 1998; Manni et al., 1999, 2001).

In this work, absence of cell death and presence of cell division in dorsal and ventral neurons of the posterior sensory vesicle is highly suggestive of a class of adult brain progenitors that had not been identified to date (Nicol and Meinertzhagen, 1991; Bollner and Meinertzhagen, 1993). Unlike the photoreceptor system, ependymal apoptosis protects the integrity of the posterior sensory vesicle of the larva, which does not completely degenerate (Fig. 2D,F,G; see Cloney, 1982; Torrence, 1983). While shedding new light on secondary neurogenesis in C. intestinalis, the hypothesis of a novel set of dividing neurons, with the ability to survive CNS degeneration during metamorphosis, will have to be tested by tracing cell fates until juvenile stages. In contrast, the visceral ganglion and the nerve cord are seemingly eradicated (Fig. 2E; van Beneden and Julin, 1884; Willey, 1893; Seeliger, 1904–1905; Scott, 1946). Peripheral apoptosis in the visceral ganglion is compatible with the elimination of the integrating centre of the CNS, which is formed by a monolayer of neurons encircling a medulla (Nicol and Meinertzhagen, 1991; Meinertzhagen et al., 2000). After metamorphosis, PCD in the juvenile brain may be interpreted as controlling the histogenetic progression toward the mature state.

Dynamic Patchwork of Cell Death and Proliferation

Spatial relationships between cell death and mitotic activity were examined by comparing distribution of PH3-positive cells with TUNEL stainings. In 23 hpf larvae, tissues facing major growth and re-patterning, such as endoderm and mesenchyme, show densely overlapping domains of PCD and cell proliferation, making it difficult to ascertain spatial patterns and differential rates of division and death. Such a theme suggests structural remodeling during metamorphosis (24–48 dpf), requiring a detailed histological classification for further interpretation (Figs. 2G, 3A, 5B). When adjacent patterns of PCD and mitosis are clearly recognized, we can conceivably attribute morphogenetic roles. Hence, whereas distal proliferation and proximal death of palp cells drive substrate adhesion (Figs. 2B, 5B), compartment selection for secondary neurogenesis is promoted by differential activation or inhibition of either physiological cell states in the larval brain (Figs. 2B, 8). Finally, as metamorphosis sets down its pace, cell death and division gradedly confine their function to stomach, endostyle, gill slits, and siphon edges of the juvenile ascidia (Fig. 3B,C).

Figure 8.

Schematic representation of the larval brain indicating proliferating neural compartments likely involved in the formation of the adult Ciona brain, compared with expression domains of mitogen-activated protein kinase (MAPK) family members MAPK kinase kinase-1 (MEKK1), dual phosphorylated extracellular regulated kinase 1/2 (dpERK1/2), and c-Jun NH2-terminal kinase (JNK). Red, nonapoptotic and proliferative domains; pale blue, apoptotic and nonproliferative domains. Dotted marks, MEKK1, dpERK1/2, and JNK; horizontal lines, MEKK1 and dpERK1/2; vertical lines, high MEKK1, low dpERK1/2, and JNK. Note spatial correlation between cell division (red) and expression of all three MAPK proteins (dots).

Role of MAPK Factors in the Regulation of Neural PCD

The mitogen-activated protein kinase (MAPK) transduction pathway is known to have an important role in a wide range of cellular processes, including apoptosis, with inductive (Yujiri et al., 1998; Sasaki and Chiba, 2001) or antagonizing properties (Houart et al., 1998; Shinya et al., 2001; Milella et al., 2002; Senger et al., 2002). Central to this pathway, MEKK1 is a MAPK kinase kinase (MAPKKK) that may up-regulate several downstream factors, including MEK1/2, JNKs, and extracellular regulated kinases (dpERK1/2; Chambon et al., 2002; Satou et al., 2003). During this analysis, we searched for topographic correspondence between distribution of MAPK proteins and cell death patterns. Overlap of different combinations of MAPK proteins with apoptotic or with nonapoptotic and/or proliferative areas enables us to postulate a functional correlation. Antibodies against MEKK1, dpERK1/2, and JNK labeled the pharynx and the anterior margin of the sensory vesicle, in a domain including the region homologous with the pituitary gland (Fig. 6A,D,H,I; Boorman and Shimeld, 2002). All three MAPK proteins were also recorded in the posterior sensory vesicle and in the neck, drawing a firm spatiotemporal association between their alike coactivations and the competence to survive of specific neural subdivisions. On the contrary, apoptotic fields in the larval nervous system appear to correlate with diverse combinations of MAPK kinases. It is interesting to note that, while dpERK1/2 and MEKK1 signals stained both neuronal and ependymal cells in the posterior sensory vesicle, including the ocellus, the JNK antibody pattern was restricted to the surviving dorsal and ventral neurons (Nicol and Meinertzhagen, 1991). As to the visceral ganglion, this neural structure displays a peculiar profile of MAPK activities, with a higher level of MEKK1 labeling in neuronal somata and corresponding axons, in relation to dpERK1/2 and JNK (Fig. 6A,B; Meinertzhagen et al., 2000; Okada et al., 2001).

The kinase activity of MEK1/2 lies directly upstream of ERK1/2, while it is induced by several protein kinases, including MEKK1 (Zheng and Guan, 1993; Yan and Templeton, 1994; Roy et al., 2002; Satou et al., 2003). As seen in Figure 7, normal PCD programs in C. intestinalis larvae are largely suppressed by means of pharmacological inactivation of MEK1/2, except for the anterior and posterior margins of the sensory vesicle, where inactivation of MEK1/2 is able to convert otherwise surviving and dividing cell types into apoptosis. In zebrafish and Xenopus, dpERK1/2 marks the anterior neural boundary, a prospective telencephalic region with inductive and patterning properties (Christen and Slack, 1999; Shinya et al., 2001) that give rise to the subpallial telencephalon. Of interest, inhibition of Ras in zebrafish induces apoptosis in the ventral telencephalon (Shinya et al., 2001). However, the hypothesis of a telencephalic forebrain in the larval CNS of ascidians is in disagreement with molecular evidence (reviewed by Holland and Holland, 1999; Lemaire et al., 2002). Inhibition of PCD in the visceral ganglion after U0126 treatment is also suggestive of dpERK1/2 implication, even though a strong MEKK1 signal could still be transduced by means of alternative pathways, maybe involving p38 (Hagemann and Blank, 2001; Zhang and Liu, 2002).

Our observations in Ciona draw a reasonable scenario by which combinations of MAPK kinases amplify a MEKK1-derived message with the aim to tune cell death in the larval CNS. In particular, an intracellular MAPK machinery is required to repress PCD in the prospective neurohypophysis, in dorsal and ventral neurons of the posterior sensory vesicle, and in the neck (Pedram et al., 1998; English et al., 1999; Chambon et al., 2002; Roy et al., 2002). It is noteworthy that ascidian cells homologous with neurohypophysis and anterior hindbrain of vertebrates are not affected by PCD, because of a signaling cascade that is conserved among chordate lineages (Katsuyama et al., 1995; Wada et al., 1998; Sabapathy et al., 1999; Kuan et al., 1999). In addition, cell survival and proliferation are found to be tightly associated with the expression profile of JNK (Fig. 8). In line with an anti-apoptotic assignment of JNK, inactivation in ependymal cells surrounding the posterior sensory vesicle could elicit PCD of these cells (Fig. 8; Lin and Dibling, 2002). Altogether, these results demonstrate that the combinatorial nature of MAPK-dependent regulation of cell death depends on cell type.

Apoptosis is positively associated with dpERK1/2 in tail tissues that are induced by inhibition of Ras (Figs 6F,G, 7A; Kim and Nishida, 2001). It has been reported that dpERK1/2 is able to lead to its translocation into the nucleus with subsequent changes in transcriptional regulation (English et al., 1999). Concomitant nuclear and cytoplasmic dpERK1/2 expression in the overlying epidermal layer and the sensory vesicle may implicate a positive feedback, mediated by undescribed factors, between an ectodermal signal and a neuroepithelial response. A similar interaction could be at the base of the epidermal modifications that precede regression of sensory vesicle in the ascidian Diplosoma macdonaldi (Fig. 6C,D; Torrence, 1983).

In summary, we have begun to explain how PCD operates during the indirect development of the ascidian C. intestinalis, and which signaling pathways are implicated in driving cell death in the larval brain. Actually, ascidian larvae may be envisaged as a continuous succession of stages, each one with its transient profile of PCD and proliferation. At the endpoints of this developmental frame, early larvae are characterized by confined apoptotic and mitotic activities (18–19 hpf), compared with even and abundant cell death and proliferation in late larvae (21–23 hpf). From this and other studies (Jeffery, 2002a, b; Chambon et al., 2002), apoptosis emerges as a cardinal force in the sexual development of ascidians. Now that PCD has been assigned the status of morphogenetic principle, we can integrate analysis of apoptotic patterns in models of metazoan phylogenies to address questions akin to developmental homologies and evolutionary pathways.



C. intestinalis specimens were collected in the Gulf of Naples. A laboratory colony was used in some experiments. Stages of development (in hours and days postfertilization, respectively, hpf and dpf) were obtained by in vitro egg fertilization with heterologous sperm in 0.22 μ-filtered seawater (FSW) in Petri dishes, where embryos grew and were staged at 18°C.

Whole-Mount TUNEL

To detect the presence of apoptotic cells during the development of C. intestinalis, we analyzed cell death in fixed tissues by means of the TUNEL method with digoxigenin-based dUTP (ApopTag Peroxidase in Situ Apoptosis Detection kit, Intergen, NY). Attempts to observe cell death in vivo with fluorescent markers (Acridine Orange, Sigma; Fluorescent-based TUNEL, Roche; Vybrant and In Vivo/Cell Death, Molecular Probes) featured poor permeability, insufficient resolution and low specificity in the trunk, which is surrounded by a double tunic envelope (see also Martínez-Álvarez et al., 2000). However, results concerning cell death in the tail, surrounded by a single coat, were consistent with digoxigenin-based TUNEL (our data; Chambon et al., 2002). C. intestinalis embryos were fixed with 4% paraformaldehyde (PFA) in MOPS and stored in 70% ETOH at −20° C. Samples were rehydrated to phosphate buffered saline (PBS) and processed following the manufacturer's instructions. After tissue labeling, samples were dehydrated to 75% glycerol and mounted on depression slides to proceed for microscopy and photography. Stained specimens were dehydrated in 70% EtOH, embedded in Epon resin, and sectioned for histology. Toluidine blue or hematoxylin and eosin were occasionally used to counterstain the sections. Images were either collected on Nikon slide films with an Axiophot (Zeiss) stereomicroscope, or by digital photomicroscopy using an AxioZeiss camera on an Axioskop (Zeiss). Images were edited with Photoshop 7.0 (Adobe).

Electron Microscopy

TEM observations were needed to confirm the ultrastructural features of cells that undergo apoptosis. As soon as they hatched (18 hpf at 18°C), larvae were fixed first in 70% EtOH, and then for 1 hr with 1% glutaraldehyde in FSW. After rinsing for 5–10 min in FSW, larvae were post-fixed in 1% osmium tetroxide in FSW for 1 hr. Rinsed again in FSW, larvae were subsequently dehydrated through an ethanol series (30 min per step), treated for 1 hr in propylene oxide, and finally embedded in Epon resin. Samples were sectioned at the Ultracut ultramicrotome (Reichert-Jung) and stained with uranyl acetate and lead citrate for observation at TEM (Philips 400).


After fixation in 4% PFA in MOPS, developmental stages of C. intestinalis were dehydrated to 70% ETOH and stored at −20°C. Rehydrated specimens were permeabilized in ice-cold acetone for 10 min at −20°C, then incubated first with primary antibodies against MEKK1, dpERK1/2, JNK (Santa Cruz Biotechnology), and PH3 (Upstate Biotechnology) and then with biotinylated secondary antibodies (Vectastain). To view epitope binding, specimens were dehydrated up to 75% glycerol in PBS. For BrdU, larvae were exposed at hatching for 30-min and 1-hr pulses in a solution containing 1 mg/ml BrdU (Sigma) and 0.001–0.0001% saponin to obviate penetration problems (Nguyen et al., 1999). Larvae treated only with saponin showed mitotic patterns as in controls. Larvae were fixed with 4% PFA in MOPS, rinsed several times in PBS and incubated for 45 min in 2 N HCl in PBS. Hence, samples were washed for 48 hr in 0.1% Triton X-100 in PBS at 4°C. Subsequently, larvae were incubated overday at 4°C with 1:2,000 anti-BrdU antibody (Sigma) in PBS–0.1% Triton X-100, and 0.5% bovine serum albumin. After 2 days of washes in 0.1% Triton X-100 in PBS at 4°C, samples were incubated for 6 hr in IB containing 1% sheep serum and 1:250 biotinylated anti-mouse IgG (Vectastain). For double BrdU–TUNEL labeling, a 2-hr BrdU pulse (21–23 hpf) was followed by BrdU immunochemistry, an overnight post-fixation with 4% PFA in PBS, and the TUNEL protocol. Image processing was as for TUNEL.

Two-Dimensional Mitotic Maps

To ascertain semiquantitative figures of cell proliferation in early and late larvae, we transformed immunohistochemical staining into cumulative maps (Photoshop 7, Adobe). The PH3 antibody was chosen instead of BrdU to measure mitotic snapshots and to prevent bias by cell movements (e.g., mesenchyme). Proliferation data were not normalized against total cell numbers, as this analysis went beyond the scope of this examination. PH3-labeled larvae were mounted on slides and precisely oriented for lateral views. In brief, histological positions of PH3-positive cells in trunks of 30 early (19 hpf) and late (23 hpf) larvae were plotted on drawings (AxioVision, Zeiss) by superimposing images with digital profiles. Histology and topographic alignment of mitotic nuclei were made unequivocal by analysis of histological sections and with care by visual inspection.

MEK1/2 Inhibition

Larvae were placed in 5-cm Petri dishes filled with 0.22 μ-FSW containing the MEK1/2-inhibitor U0126 (Promega) at 10 μM (Chambon et al., 2002). All experiments included at least 50 larvae. We treated embryos during three developmental intervals, starting at 13 (middle tail bud), 16 (late tail bud), and 18 hpf (prehatching larvae) until, respectively, 31, 34, and 45 hpf, when larvae were fixed and processed with TUNEL. U0126 was omitted in controls.


We thank I. Arnone, I. Buttino, G. Romano, R. Casotti, and G. De Vita for reagents; J.S. Joly and R. Marino for technical advices; the Microscopy Service for histology; the Marine Resources for Research Service for animal care; and the Fishery Service for collecting animals. We also thank E. Brown, P. Burighel, M. De Felice, and P. te Welscher for discussions and critical reading of the manuscript. P.S. was supported by EMBO, EC, Company of Biologists, and CNR fellowships. This study is wholeheartedly dedicated to the memory of Nigel Holder.