Activity regulates programmed cell death of zebrafish Rohon-Beard neurons

Proper development and organization of the nervous system requires an extensive amount of programmed cell death (PCD; Cowan et al., 1984; Clark, 1985; Oppenheim, 1991). Perturbations that decrease neuronal PCD during embryogenesis result in malformations of the central nervous system (Kuida et al., 1996). By contrast, in the adult nervous system, PCD is less prominent, and, when present, is associated with neurodegenerative disorders. Deciphering the mechanisms that underlie neuronal PCD will contribute to our understanding of normal brain development as well as neurodegenerative disorders.

The zebrafish embryo provides several advantages for study of PCD in the developing vertebrate nervous system. At a gross level, the normal developmental changes that occur in neuronal morphology can be monitored in live embryos with a stereomicroscope. The optical transparency of the embryo allows easy visualization of dying cells in live embryos. In addition, a vast collection of DNA and antibody probes allows for more detailed examination of the consequences of cell death on pattern formation and neural differentiation (Metcalfe et al., 1990; Marusich et al., 1994). Furthermore, the feasibility of forward genetic approaches raises the possibility of identifying novel genes and mechanisms that contribute to neuronal PCD in the developing vertebrate nervous system.

During a recent large-scale mutagenesis screen of the zebrafish embryo, several mutations were identified that led to massive cell degeneration within the developing nervous system (Abdelliah et al., 1996; Furutani-Seiki et al., 1996; Daly and Sandell, 2000). The majority of mutations increased the extent of PCD via activation of a cell suicide program (apoptosis). In some instances, cell death was limited to a restricted region of the central nervous system, suggesting that early patterning defects were primary consequences of the mutations (Furutani-Seiki et al., 1996). A previously isolated mutant, ned-1 (generated by γ-irradiation)(Grunwald et al., 1988) led to massive neural cell death in most neurons, with the exception of early born primary neurons. Primary neurons, which include several populations of spinal neurons and the reticulospinal Mauthner cells, were spared and did not show enhanced cell death. The zebrafish neural degeneration mutations will aid in genetic dissection of the signals and biochemical pathways that are involved in neuronal PCD and ultimately pathological degenerative conditions.

Although the zebrafish neural degeneration mutants involve perturbations in cell death pathways that affect several neuronal populations, one interesting population appeared to be unaffected. This particular neuronal population, Rohon-Beard (RB) cells, normally undergo extreme PCD and exist only transiently in zebrafish as well as other teleost and amphibian species (Lamborghini, 1987; Hensey and Gautier, 1998; Reyes, 2000; Williams et al., 2000). By 5 days post fertilization (dpf), RB cells are completely eliminated from wild-type zebrafish (Henion et al., 1996; Reyes, 2000; Williams et al., 2000). Transient neuronal populations are not unique to fish and frog species, and are represented in mammals by Cajal-Retzius cells of the cerebral cortex (Derer and Derer, 1990; Frotscher, 1997).

Despite the fact that all RB cells die during embryogenesis, no mutations have been isolated so far that prevent RB cell death. This is somewhat surprising, given the ease of studying RB cell differentiation and function. For example, RB cells are easy to recognize in live embryos using differential interference contrast optics (Grunwald et al., 1988) or in fixed preparations with a variety of immunological and molecular probes (Henion et al., 1996). In addition, their sensory function can be readily examined using a simple and rapid touch sensitivity test (Granato et al., 1996).

Although the zebrafish genetic mutants have not yet provided information regarding RB PCD, recent evidence suggests that neurotrophins are involved (Williams et al., 2000). Antibodies that block neurotrophin 3 (NT3) function increase the number of dying RB cells. Conversely, application of exogenous NT3 diminishes the extent of RB PCD. Withdrawal of neurotrophins can trigger cell death via a caspase-dependent pathway (Lievremont et al., 1999). Indeed, inhibition of caspases decreases the number of dying RB cells, indicating that PCD of RB neurons involves the activity of these proteolytic cell death enzymes (Williams et al., 2000).

Involvement of neurotrophins in RB PCD raises the possibility that neural activity also regulates the cell death pathway because activity affects expression and secretion of neurotrophins (Ghosh et al., 1994; Schinder and Poo, 2000). In addition, neural activity alters levels of intracellular Ca²⁺, which can regulate cell death independently of neurotrophins (Larmet et al., 1992). No information exists about the role of activity in regulating cell death in the zebrafish nervous system.

We have examined the role that activity plays in the zebrafish embryo in RB cell death. Neural activity was perturbed either by use of a genetic mutant (mao) or a pharmacological blocker (tricaine). The mao mutant was isolated in a screen for motility mutants (Granato et al., 1996) and revealed by its insensitivity to tactile stimulation. Motor function appeared normal, suggesting that the deficit resides in mechanosensory neurons. Subsequent physiological examination indicated that the relevant mechanosensory neurons, RB cells, in mao mutants lacked voltage-gated Na⁺ current and could not fire action potentials, thereby accounting for the behavioral deficit (Ribera and Nüsslein-Volhard, 1998). Tricaine is an anesthetic agent that preferentially blocks Na⁺ current (Frazier and Narahashi, 1975; Wang et al., 1994). We found that block of Na⁺ current, either genetically or pharmacologically, reduced RB cell death. Furthermore, the extents to which the mao mutation and tricaine treatment reduce cell death were similar. These results indicate that activity is required for the normal progression of cell death in RB cells.

The mao mutant line was kindly provided by Dr Hans-Georg Frohnhöfer of the Tübingen Stockcenter. The original ethynirosourea-induced mutation was created in fish of the Tü strain (Granato et al., 1996; Haffter et al., 1996). This line has been outbred for several generations with several wild-type strains and the behavioral phenotype persists in different genetic backgrounds (Granato et al., 1996; Ribera and Nüsslein-Volhard, 1998). The results reported here similarly did not depend upon the genetic background. We used three different wild-type strains for control studies (Tü, WIK and PS; the latter are fish from a local petstore); similar results were obtained with all three strains. Adults and embryos were housed at 28°C in a fish facility maintained in the UCHSC Center for Animal Laboratory Care. Embryos were developmentally staged according to previously described methods (Kimmel et al., 1995). 22-24 hpf embryos were manually dechorionated and transferred to Embryo Medium (EM; Westerfield, 1994) containing 0.002% 1-phenyl-2 thiourea (PTU) to inhibit pigment formation. Tricaine treatment consisted of raising 24 hpf embryos at 28°C in EM-PTU containing 0.006% tricaine (ethyl 3-aminobenzoate, methansulfonic acid; MS222; Sigma-Aldrich); controls were raised in EM-PTU at 28°C. Tricaine-treated and control fish were examined every 3 hours at room temperature to determine when the touch response was blocked in treated fish (typically by 30-32 hours post fertilization; hpf). Embryos received fresh embryo media (with the appropriate blockers, depending upon the experimental group) every 24 hours.

Embryos were fixed overnight at 4°C in 4% paraformaldehyde/ phosphate-buffered saline containing 0.1% Tween-20 (PBST). Fixed embryos were washed and stored in PBST at 4°C. mao mutant embryos were identified on the basis of their touch insensitivity (Granto et al., 1996). Within a clutch of embryos, we clipped the tails of mutant or sibling embryos. The tail clipping allowed distinction of mutants from siblings (or wild type from tricaine-treated) while allowing them to be co-processed in the same reaction vial to ensure procedural consistency.

For immunocytochemical processing, embryos were first washed in PBST. Permeabilization was achieved by a 60 minute incubation in distilled H₂O, followed by a 20 minute exposure to 100% acetone at 4°C and a final 10-60 minute collagenase (1 mg/ml) treatment. Embryos were blocked for 1 hour in 10% normal calf serum/PBST and then incubated overnight at 4°C in 10% serum/PBST containing primary antibody. The primary antibodies anti-Hu (referred to here as HuAntibody, HuA; gift from Dr M. Marusich, University of Oregon), zn-12 (Developmental Hybridoma Bank, University of Iowa) and anti-acetylated tubulin (aat, Sigma-Aldrich) were used at 1:1000, 1:500 and 1:1000, respectively. The next morning, embryos were washed in PBST for at least 5 hours at room temperature and then incubated overnight (4°C) with a biotinylated, goat anti-mouse secondary antibody (Vector ABC Elite Kit, 1:250 dilution). Embryos were then washed in PBST for at least 2 hours and incubated in avidin/biotin (A/B) solution for 40 minutes. After the A/B incubation, embryos were washed in acetate buffer (50 mM, pH 5.0) for 2 hours. Embryos were then transferred to fresh acetate buffer solution containing 0.04% 3-Amino-9 Ethylcarbazol (AEC, Sigma-Aldrich) and 0.05% N,N-dimethyl-formamide (DMF, Sigma) in multi-well tissue culture dishes (Kaplow, 1974). The chromogenic reaction was catalyzed by adding 0.01% H₂O₂ and stopped within 2-10 minutes by washing in PBST for at least 30 minutes.

Double-immunolabeling experiments were performed using the aat and zn-12 primary antibodies; the former recognizes a mature form of tubulin, whereas the latter binds to an extracellular sugar epitope on the plasma membrane (Metcalfe et al., 1990). The protocol described above was modified slightly and fluorescent secondary antibodies were used. Embryos were first incubated overnight with the zn-12 antibody (1:500). The next day, embryos were washed with PBST and then incubated overnight at 4°C in a biotinylated, goat anti-mouse secondary antibody (1:250, Vector ABC Elite Kit). On day 3, embryos were washed for 2-3 hours in PBST. To reveal zn-12 immunoreactivity, we applied rhodamine conjugated-streptavidin (Jackson ImmunoResearch, 1:400) for 5 hours at room temperature. After confirming positive zn-12 labeling, we washed embryos overnight at 4°C in PBST. On day 4, embryos were washed for 8-10 hours in PBST and then incubated with the aat primary antibody (1:1000) overnight at 4°C. On day 5, embryos were washed for 2 hours in PBST and then incubated in fluorescent secondary antibody (1:500, Alexa 488, Molecular Probes) at room temperature for 3-5 hours. Embryos were washed and stored in PBST (4°C) until examination. Similar results were obtained when the aat antibody was presented first and followed by the zn-12 antibody.

Terminal transferase dUTP nick-end labeling (TUNEL) staining was combined with whole-mount immunocytochemistry as described previously (Reyes, 2000) with the following two modifications: proteinase K (1 μg/ml, Sigma Aldrich) treatment (37°C) was longer (45-60 minutes) and embryos were not cleared in glycerol.

Embryos that had been processed for aat immunoreactivity were squash-mounted laterally and viewed on an Eclipse TE200 inverted Nikon Microscope with a 20× or 40× objective and photographed using a Princeton Instruments digital camera. Images of embryos that were processed for both zn-12 and aat immunoreactivities were collected on a Nikon PCM-2000 laser-scanning confocal microscope and digital images were acquired at various focal planes.

For anti-Hu visualization, we embedded embryos dorsal side up in 1.25% low-melting point agarose (Gibco) and stored them in individual wells of tissue culture dishes until the RB neurons in each embryo were counted (see below). Each embryo was given a code designation so that cell counts could be made without knowledge of the embryo type (e.g. mutant versus sibling control). After cell counts had been performed, we mounted some embryos laterally to examine the dorsal root ganglia.

Embryos were processed for immunocytochemistry and coded by one investigator, while another investigator counted RB somata. RB somata were identified on the basis of position and size in the spinal cord and strong HuA immunoreactivity. HuA recognizes an RNA binding protein that is expressed in several spinal cord neuron populations and abundantly in RB cells, as well as in the later appearing dorsal root ganglia (Marusich et al., 1994). RB somata were identified by their dorsal position, large size and intense HuA immunoreactivity (Grunwald et al., 1988; Henion et al., 1996; Ribera and Nüsslein-Volhard, 1998). The region that was analyzed extended from spinal cord segment 2 to the most caudal segment above the yolk sac extension. All data are presented as mean±s.e.m. Levels of statistical significance were assessed by an unpaired two-tailed Student’s t-test. P< 0.05 was considered to be indicative of statistical significance.

Zebrafish embryos first display a response to tactile stimulation of the trunk at ∼1 dpf (Grunwald et al., 1988; Saint-Amant and Drapeau, 1998; Saint-Amant and Drapeau, 2000). At this time, primary sensory neurons, known as RB cells, mediate the touch response in the trunk (Clarke et al., 1984; Soffe, 1991). One day later, dorsal root ganglion neurons begin to innervate the skin and thus potentially also contribute to the touch response. RB and dorsal root ganglion cells co-exist for a few days. By 5 dpf, however, all RB cells have been eliminated and dorsal root ganglion now solely mediate the touch response in the trunk (Reyes, 2000; Williams et al., 2000).

We used whole-mount immunocytochemistry to monitor changes in the distribution of aat immunoreactivity as RB cells differentiate and die. Normal development and function of RB cells requires that they extend processes peripherally into the skin. The peripheral processes terminate in free endings that possess mechanosensitive channels, the site of the initial transduction event (Clarke et al., 1984). We examined the distribution of aat immunoreactivity during the period of maturation of RB peripheral processes (Fig. 1). In wild-type embryos, aat immunoreactivity was present in RB major peripheral processes at the time embryos first display an immediate response to tactile trunk stimulation (27 hpf). Between 27-36 hpf, aat immunoreactivity was continuous within the major peripheral process (Fig. 1A,B). As development progressed, the pattern of aat immunoreactivity changed. By 40 hpf, aat immunoreactivity became varicose rather than continuous (Fig. 1C). With further development, varicosities became more prevalent; none of the major peripheral processes exhibited the continuous, smooth aat immunoreactivity characteristic of earlier stages (Fig. 1D). At any given stage, aat immunoreactivity in peripheral processes was similar, regardless of their rostrocaudal position (data not shown).

The pattern of aat immunoreactivity within the spinal cord also changed during development. At 27 hpf, RB somata and early forming spinal tracts contained aat immunoreactivity. Nine and 13 hours later, aat immunoreactivity appeared more intense within spinal tracts. However, at 48 hpf, aat immunoreactivity within the spinal cord was substantially reduced. Only the beaded aat labeling of RB peripheral processes described above was obvious at 48 hpf.

In summary, between 27-48 hpf, the distribution of aat immunoreactivity changes in both RB major peripheral processes and the spinal cord. Thus, aat immunocytochemistry reflects aspects of RB cell differentiation that occur within the cytoskeleton of the peripheral processes.

Recently, Williams et al. and Reyes demonstrated that zebrafish RB cells undergo massive programmed cell death (PCD) and are almost completely eliminated by 72 hpf (Williams et al., 2000; Reyes, 2000). At this time, all RB cells have retracted or are retracting their processes (Reyes, 2000). It is possible that redistribution of aat immunoreactivity into varicosities occurs as a result of process degeneration. However, process retraction is not prominent at 48 hpf (Reyes, 2000), a time when aat immunoreactivity already appears prominently as a collection of varicosities (Fig. 1D). To examine whether the formation of aat immunoreactive varicosities and process fragmentation occur simultaneously, we compared the distribution of aat to that of a cell surface glycoprotein recognized by the zn-12 monoclonal antibody (Metcalfe et al., 1990). RB peripheral processes of wild-type embryos were thus double-labeled for aat and zn-12 immunoreactivities. The zn-12 antibody revealed a complex network of peripheral processes (Fig. 2A). By contrast, aat immunoreactivity was restricted to the major peripheral processes (Fig. 2B). At 60 hpf, aat immunoreactivity in RB major peripheral processes was varicose (Fig. 2C, green), while zn-12 immunoreactivity appeared as a mesh of fine continuous fibers (Fig. 2C, red). In the major peripheral processes that exhibited both aat and zn-12 immunoreactivities, aat immunoreactive varicosities were present. By contrast, zn-12 immunoreactivity in these processes was continuous, as would be typical of an intact process that had not yet fragmented. Similar findings were obtained from 48 hpf embryos (data not shown). These data indicate that redistribution of aat immunoreactivity precedes the fragmentation of processes and therefore is an earlier event in the program of RB cell death.

mao mutants do not respond to touch, although they can swim (Granato et al., 1996). In mao mutants, RB cells lack voltage-gated Na⁺ current and do not fire action potentials, thus accounting for the behavioral phenotype (Ribera and Nüsslein-Volhard, 1998). To examine the possibility that the mao mutation affects RB cell differentiation and/or death, we analyzed the distributions of aat immunoreactivity in RB peripheral processes and the spinal cord of mao mutant and sibling control embryos. At 36 hpf, no differences were noted in aat immunoreactivity between wild-type, mao mutant or sibling control embryos (compare Fig. 1B with Fig. 3A,C). However, at 48 hpf, the distribution of aat immunoreactivity in RB peripheral processes of mao mutants differed dramatically from that of either sibling control or wild-type embryos (Fig. 1D, Fig. 3). Specifically, in mao mutants, aat immunoreactivity in RB peripheral processes retained its continuous, smooth distribution (Fig. 3D). Aat immunoreactivity remained continuous in RB peripheral process of mao mutants as late as it was examined (72 hpf; data not shown).

Within the spinal cord, aat immunoreactivity did not differ between mutant and sibling embryos at any developmental stage. In all embryos, aat immunoreactivity was prominent in the spinal cord at 27-36 hpf but substantially reduced by 48 hpf.

Formation of aat immunoreactive varicosities in RB peripheral processes is suppressed in the mao mutant, raising the possibility that PCD is attenuated in mao mutants. Accordingly, we relied on RB morphology and HuA immunoreactivity to count RB somata at several stages during the normal period of RB cell elimination. HuA immunoreactivity also reveals dorsal root ganglia as they begin to appear (∼48 hpf).

At 36 hpf, HuA immunoreactivity revealed little difference in the number of RB somata in mao mutants versus sibling embryos (Fig. 4A, top). Cell counts indicated that there was no significant difference in the number of RB somata between mao mutant and sibling embryos (Fig. 4B, left).

At 72 hpf, very few or no RB cells were present in sibling control embryos (Fig. 4A, bottom). However, in mao mutants, HuA immunoreactivity revealed the persistence of numerous RB cells. On average, twice as many RB cells were present in mutants compared with their sibling controls (Fig. 4B, right). By contrast, in both mao mutant and sibling embryos, HuA immunoreactivity is present in the dorsal root ganglia (Fig. 4C).

In addition to being molecularly unique cells, RB cells possess distinguishing morphological features (large soma, dorsal position) that permit their identification directly in fixed and live preparations of zebrafish embryos (Grunwald et al., 1988). Visualization of 72 hpf embryos with Hoffman optics indicates that RB cells are more numerous in mao mutant when compared with sibling embryos (Fig. 5). These data indicate that the mao mutation leads to a reduction in RB cell PCD.

An early manifestation of entry into the PCD pathway is DNA fragmentation (Kerr et al., 1972; Searle et al., 1982). DNA fragmentation can be revealed in zebrafish using TUNEL staining (Reyes, 2000; Williams et al., 2000). At 36 hpf, wild-type embryos show a high incidence of positive TUNEL labeling in the RB cell population (Williams et al., 2000). In mao mutants, significantly fewer TUNEL-positive RB cells are present at 36 hpf compared with sibling control embryos (Fig. 6). Thus, mao mutants displayed attenuations in PCD at early (TUNEL), intermediate (redistribution of aat immunoreactivity in RB peripheral processes) and late (soma number) stages.

Our results indicate that the mao mutation reduced the extent of normally occurring RB PCD. The mao mutation leads to elimination of the majority of voltage-gated Na⁺ current and suppression of action potentials in RB cells (Ribera and Nüsslein-Volhard, 1998). The mao gene may directly and independently affect both RB cell Na⁺ current and PCD. Alternatively, the mao mutation might directly control RB electrical activity, which in turn regulates PCD. To distinguish between these possibilities, we tested whether pharmacological suppression of Na⁺ current would have effects on PCD similar to those of the mao mutation.

Embryos were placed in a low dose of tricaine (0.006%) at 24 hpf, just before the time that the mao touch-insensitive phenotype is first detected. This dose of tricaine abolished touch sensitivity when assessed at 32 hpf by exposing embryos to a series of tactile stimuli. By contrast, the frequency of cardiac contractions appeared unaffected compared with those of wild-type embryos. Furthermore, embryonic development continued and embryos appeared normal (Fig. 7A).

In tricaine-treated embryos, aat immunoreactivity remained continuous in RB peripheral processes (Fig. 7B, bottom) and resembled the pattern observed in both 36 hpf wild-type embryos and 48 hpf mao mutant embryos (Fig. 1B, Fig. 3D). By contrast, the distribution of aat labeling within the peripheral processes of 60 hpf wild-type embryos was beaded (Fig. 7B, top). Tricaine treatment had no effect on the normal developmental changes of aat immunoreactivity within the spinal cord. These results indicate that tricaine treatment mimicked the effects of the mao mutation on developmental regulation of aat immunoreactivity in RB peripheral processes.

Counts of RB cells in tricaine-treated embryos revealed that the reduction of PCD was similar to that produced by the mao mutation (Fig. 8). By 72-78 hpf, RB somata were almost or completely absent from wild-type and mao sibling control embryos. By contrast, in tricaine-treated embryos or mao mutants, about half of the original RB cell population was still present (Fig. 9). Furthermore, the effects of the mao mutation and tricaine-treatment were first apparent at the same developmental stage, 60 hpf. At this time, both mao mutants and tricaine-treated embryos retain 10% more RB somata than sibling control or wild-type embryos.

While activity often acts to promote neuronal survival, we find that activity accelerates PCD of RB cells. RB cells acquire the ability to fire Na⁺-dependent overshooting action potentials by ∼1 dpf, as a result of developmental regulation of voltage-gated Na⁺ current (Ribera and Nüsslein-Volhard, 1998). We blocked neural activity either genetically (mao mutant) or pharmacologically (tricaine). The effects of the mao mutation on RB Na⁺ current are evident as early as 1 dpf, and lead to an obvious behavioral phenotype – touch insensitivity. For pharmacological suppression of Na⁺ current, we treated embryos with tricaine beginning at 1 dpf, in order to mimic functionally and temporally the effects of the mao mutation. These manipulations reduced the number of TUNEL-positive RB neurons, redistribution of aat immunoreactivity in the periphery and the normal developmental loss of RB cell somata.

Our results indicate that block of neural activity affects three properties of RB cells that are evident between 27-78 hpf. DNA fragmentation and cell body elimination are conventional markers of PCD. We propose that redistribution of acetylated-tubulin within the major peripheral processes of RB cells also reflects ongoing PCD and reports a stage that is between DNA fragmentation and cell body loss.

Discrete regions of swelling or beading form in neuronal processes as they degenerate either in pathological conditions or in response to injury (Ramon y Cajal, 1928; Delisle and Carpenter, 1984; Gold, 1987). Changes in the membrane cytoskeleton underlie the formation of the swellings (Ochs et al., 1996; Ochs et al., 1997; Hall et al., 1997b). Our data indicate that there is a reorganization of the membrane cytoskeleton of peripheral processes of dying RB cells. The redistribution of tubulin clearly precedes the eventual fragmentation of peripheral process and might be a prerequisite for process retraction.

RB cells display an extreme form of normally occurring PCD, because they all die in zebrafish, as well as in several other species (Humphrey, 1944; Lamborghini, 1987; Kollros and Bovbjerg, 1997). RB cells exist only transiently and their function is assumed by later developing dorsal root ganglion neurons, raising the possibility that RB and dorsal root ganglion neurons compete for a common survival factor. However, RB cell death begins before dorsal root ganglion differentiation (Williams et al., 2000). Similarly, we find that the effects of the mao mutation on DNA fragmentation, an early index of PCD, are apparent as early as 36 hpf. At this time, dorsal root ganglion neurons are not yet detectable, even with immunological probes (Williams et al., 2000). These data suggest that the effects of activity blockade arise in Rohon-Beard cells rather than dorsal root ganglion cells.

Neuronal activity can regulate cell survival. Often, activity promotes neuronal survival, and blockade of activity leads to an increase in developmental PCD. Genetic knockout of the voltage-gated Na⁺ channel α-subunit, Na_v1.2, leads to profound CNS apoptosis, particularly in the neocortex and brainstem (Planells-Cases et al., 2000). Similarly, transmission blockade at the chick ciliary ganglion enhances cell death (Maderdrut et al., 1988). In these cases, activity appears to be required for survival of specific neuronal populations.

There are, however, examples of activity enhancing cell death, similar to what we report for zebrafish Rohon-Beard cells. A classic case is motoneuron cell death in the chick embryo. Here, block of synaptic transmission reduces the extent of normally occurring PCD (Pittman and Oppenheim, 1979; Oppenheim et al., 2000). Activity block also prevents death in the isthmo-optic nucleus (Pequignot and Clarke, 1992).

Although it is clear that activity can have diverse effects on cell survival, little is known about the underlying mechanisms. With regard to the case of RB cells, we find that the mao mutation and pharmacological suppression of Na⁺ current reduce the loss of RB somata to a similar extent (Fig. 9). However, neither manipulation of Na⁺ current completely prevents death. Instead, the timecourse of PCD appears to be delayed (Fig. 9). Indeed, in 8 dpf mao mutants, no RB somata are detected (data not shown).

In wild-type embryos, the activation of the Na⁺ current leads to elevation of intracellular Na⁺, consequent depolarization, followed by elevation of intracellular Ca²⁺. Previous studies have demonstrated that elevations in intracellular Na⁺ suffice to induce cell death. In rat superior cervical ganglion neurons, increases in intracellular Na⁺ promote apoptosis; furthermore, cell death was not blocked by reduction of extracellular Ca²⁺ or buffering of intracellular Ca²⁺ (Koike et al., 2000). Cerebellar granule neurons also die in response to over-stimulation of Na⁺ channels, even when Ca²⁺ channels and other routes of Ca²⁺ entry are blocked (Dargent et al., 1996). In worms, mutations in degenerin genes leads to constitutive activation of ion channels, elevation of intracellular Ca²⁺ and death of mechanosensory neurons (Hall et al., 1997a). The latter report is particularly intriguing given that RB cells also function as mechanosensory neurons. Perhaps, mechanosensory cells are more vulnerable to Na⁺ overload.

Elevations of intracellular Ca²⁺ also suffice to promote cell death. For example, elevations of intracellular Ca²⁺ can activate Ca²⁺-dependent proteolytic enzymes; these proteolytic enzymes cleave precursor forms of caspases, enzymes that mediate the cell death program (Salvesen and Dixit, 1997). Thus, elevations of intracellular Ca²⁺ might lead to activation of caspases and initiation of the cell death program.

Thus, it is possible that elevations in either intracellular Na⁺ or Ca²⁺ might accelerate PCD of RB cells. Although our data do not allow us to distinguish between these possibilities, it is clear that both activity as well as neurotrophins regulate PCD of RB cells (Williams et al., 2000; this study). Whether and how these factors interact has yet to be determined.

We thank Dr Hans-Georg Frohnhöfer and the Tübingen Stockcenter for providing the mao mutant line; Ryan Heiser for expert fish maintenance; Dr Darren Gilmour for generous assistance with immunocytochemical procedures; Drs S. Rock Levinson and Robert Tanguay for use of microscopes and imaging systems; Drs Judith Eisen and Rosie Reyes for protocols; Drs Tom Finger, Joan Hooper, S. Rock Levinson, Robert Tanguay and Sukumar Vijayaraghavan for comments on the manuscript; and members of the laboratory for discussion. This work was supported by NIH grants F32-MH12748 (K. R. S.) and T32-NS07083 and NS38937 (A. B. R.).