Genetic evidence for cell death mediated by nerve growth factor and the neurotrophin receptor p75 in the developing mouse retina and spinal cord

The neurotrophin receptor p75 (p75^NTR) was the first member to be identified of a family of cell death-inducing receptors, including in particular Fas/Apo-1/CD95 and tumor necrosis factor receptor 1 (for a review, see Nagata, 1997). P75^NTR was initially described as the receptor for nerve growth factor (NGF) (Johnson et al., 1986; Radeke et al., 1987), but following the discovery of the NGF-related brain-derived neurotrophic factor and neurotrophin-3, it became clear that p75^NTR also binds these neurotrophins (RodrÍguez-Tebar et al., 1990, 1992). While the in vivo role of p75^NTR remained unclear for a long time, recent evidence indicates that it is involved in regulating programmed cell death during neurogenesis (for reviews, see Dechant and Barde, 1997; Barker, 1998). In particular, experiments using blocking antibodies have indicated that both NGF and p75^NTR participate in the elimination of large numbers of cells during the time when the first retinal ganglion cells begin to form axons in the chick retina (Frade et al., 1996). These experiments were performed at early developmental stages, before detectable expression of trkA, a tyrosine kinase receptor known to bind NGF and to mediate its anti-apoptotic properties (for a review, see Barbacid, 1994). As in other tissues where Fas ligand and tumor necrosis factor are typically found in activated macrophages, NGF in its killing function was localised to microglial cells invading the developing retina (Frade and Barde, 1998).

In line with these in vivo findings, NGF has also been shown to kill cultured p75^NTR-expressing rat oligodendrocytes (Casaccia-Bonnefil et al., 1996) or chick trigeminal neurons (Davey and Davies, 1998), and to cause cell death in explants of otic vesicles (Frago et al., 1998).

A role for p75^NTR in mediating cell death in vivo is supported by observations made in the basal forebrain of mice carrying a deletion in the p75^NTR gene (van der Zee et al., 1996; Yeo et al., 1997). These mice have more cholinergic neurons compared with wild-type animals, suggesting that p75^NTR participates in the elimination of some of these neurons in wild-type animals. In addition, overexpression of the cytoplasmic domain of p75^NTR in transgenic mice leads to considerable cell losses in many areas of the developing central nervous system (CNS), revealing that this receptor has a substantial death potential in vivo (Majdan et al., 1997).

So far, there is no direct genetic evidence for the role of NGF in cell death in vivo. This study examines and compares the effects of mutations in the ngf and p75^NTR genes in the developing murine retina and spinal cord. Cell-type specific markers were also used to identify the phenotype of some of the cells eliminated by NGF- and p75^NTR-dependent mechanisms.

Original breeding pairs of mice homozygous for a mutation targeted to the third exon of the p75^NTR gene (p75^NTR−/−; Lee et al., 1992) were obtained from the Jackson Laboratory (Bar Harbor, ME). No NGF binding to p75^NTR is observed in such mice (Lee et al., 1992). The background of this strain was a mixture of Sv129 and Balb/c. p75^NTR−/− mice were back-crossed with wild-type mice from one of the parental strains (Sv129) to generate heterozygous F₁ progeny, which were inbred to produce F₂ progeny consisting of all three genotypes. Pregnant females were identified by the presence of a vaginal plug. The day of plug observation was designated embryonic day (E)0.5. Embryo littermates, staged as described by Kaufman (1992), consisting of wild-type controls (p75^NTR+/+) and homozygous for the p75^NTR mutation (p75^NTR−/−), were used in the present study. Wild-type embryos were also used for the analysis of cell death during the normal development of the retina and brachial spinal cord. Genotypes were determined by PCR of genomic DNA obtained from the tail using the following primers: P75KOD2 (5′CCC CTT CCC AGC CTC TGA GC3′), P75U1 (5′AGC CGT GCA AGC CGT GCA CC3′), and P75D1 (5′AGG GTA GGC ACG GGT CCA CG3′). P75KOD2 is complementary to a sequence of the PGK promoter region (bp 767-786; Adra et al., 1987) present in the fragment containing the neomycin gene used for disrupting the exon III of the p75^NTR gene (Lee et al., 1992). The sequence of P75U1 is present in the 5′ region from the exon III of the rat p75^NTR gene (bp 358-377; Radeke et al., 1987) and is identical with the human p75^NTR sequence (bp 355-374; Johnson et al., 1986). The sequence of P75D1 is complementary to a sequence present in the 3′ region of the exon III of the rat p75^NTR gene (bp 596-615; Radeke et al., 1987) and is also identical with the human p75^NTR sequence (bp 593-612; Johnson et al., 1986). P75U 1 and P75D1 flank the BstXI site used for the disruption of the mouse exon III of the p75^NTR gene (Lee et al., 1992). A 258 bp product corresponding to the mouse p75^NTR gene was detected in p75^NTR+/+ embryos, and a product of about 600 bp was detected in p75^NTR−/− embryos. Both products were observed in p75^NTR+/− mice. Genomic PCR reactions containing the three primers were incubated for 30 cycles at 95°C (30 seconds)/62°C (30 seconds)/72°C (1 minute). Original breeding pairs of mice heterozygous for a null mutation in the ngf gene (ngf+/−; Crowley et al., 1994) were obtained from Genentech, Inc. (South San Francisco, CA). The strain background was C57BL/6. ngf+/− mice were mated to produce progeny consisting of all three genotypes. Pregnant females were identified by presence of a vaginal plug (the day of plug observation being designated E0.5). Embryo littermates, staged as described by Kaufman (1992), consisting of wild-type controls (ngf+/+) and homozygous for the ngf mutation (ngf−/−), were used in the present study. Genotypes were determined by PCR of genomic DNA obtained from tail using the following primers: NGFKOU2 (5′CCG TGA TAT TGC TGA AGA GC3′), NGFU6 (5′CAG AAC CGT ACA CAG ATA GC3′) and NGFD1 (5′TGT GTC TAT CCG GAT GAA CC3′). NGFKOU2 corresponds to the sequence bp 825-844 (Beck et al., 1982) from the neomycin insert used for targeting the mouse ngf gene (Crowley et al., 1994). The sequence of NGFU6 is present in the 5′ region from the exon IV of the mouse ngf gene (bp 346-365; Scott et al., 1983). The sequence of NGFD1 is complementary to a sequence present in the 3′ region from the exon IV of the mouse ngf gene (bp 955-974; Scott et al., 1983). NGFU6 and NGFD1 are flanking the BsmI site used for the replacement of a 150 bp BstEII-BsmI fragment at the beginning of exon IV of the mouse ngf gene with the Neo construct (Crowley et al., 1994). A 629 bp product corresponding to the mouse ngf gene was detected in ngf+/+ embryos, and a product of about 900 bp was detected in ngf− /− embryos; both products were observed in ngf+/− mice. Genomic PCR reactions containing the 3 primers were incubated for 32 cycles at 95°C (30 seconds)/59°C (30 seconds)/72°C (1 minute).

Affinity-purified IgGs from an anti-human p75^NTR polyclonal antiserum (Promega) were used at 2 μg/ml (spinal cord) and 10 μg/ml (eye). The previously described affinity-purified anti-Islet-1 (ISL1) polyclonal antibody (Thor et al., 1991; a gift from T. Edlund) was used at 1:250 dilution.

Whole embryos were fixed overnight in PBS/4% paraformaldehyde. Then, following incubation for 2 days at 4°C in 100 mM sodium phosphate buffer (pH 7.3) containing 30% sucrose, the head or the brachial region (cervical cord segments 4-8 and toracic cord segment 1) were isolated, embedded in OCT compound (Tissue-Tek, Miles Inc, Elkhart, IN) and sectioned at 8 μm using a cryostat. Sections were collected on 3-aminopropyl-trimethoxysilane-coated slides (Fluka), blocked with 100 mM glycine containing 0.3% H₂O₂ in PBS for 30 minutes, followed by 1 hour in PBS containing 0.1% Tween-20 and 10% goat serum (PTG) (see Frade et al., 1997). Following an overnight incubation at 4°C with the appropriate antibody in PTG and 5 washes with PBS/0.1% Tween-20 (PT), the sections were incubated for 1 hour with biotinylated sheep anti-rabbit IgG species-specific whole antibody (Amersham) diluted 1:300 in PTG. After 5 washes as above, the sections were treated for 1 hour with streptavidin-peroxidase (Amersham) diluted 1:500 in PTG, washed with PT as above, and then with PBS. The antigen-antibody complex was then revealed with 0.05% 3-3′-diaminobenzidine tetrahydrochloride (DAB) (Sigma) solution containing 0.03% H₂O₂ (Merck) and stopped with water. P75^NTR staining was intensified as described by Merchenthaler et al. (1989). Briefly, sections were incubated for 5 minutes with 0.05% 3-3′-diaminobenzidine tetrahydrochloride (DAB) (Sigma)/0.06% NiCl₂ (Sigma) solution containing 0.03% H₂O₂ (Merck). Following 2 washes with 1% sodium acetate for 1 minute each, the sections were immersed in developer for 2-3 minutes. The developer was freshly prepared by slowly adding 1 ml of 5% sodium tungstate (Sigma) and 1 ml of 0.2% ascorbic acid (Sigma) into 8 ml of a solution prepared as follows. 4.82 g of sodium acetate (Merck) were dissolved in 50 ml of distilled water. Then, 0.56 ml of acetic acid (Merck), 10 ml of 1% silver nitrate (Merck) and 1 ml of 1% of N-acetylpyridinium chloride (Merck) were added under vigorous stirring. After 4 hours at 4°C, the solution was filtered. Then, 6 ml of 1% Triton X-100 (Sigma) was added and the volume adjusted to 80 ml with water. Sections were finally washed twice (for 10 minutes) with 1% acetic acid, dehydrated with alcohol and toluene, and mounted with Entellan (Merck).

Tissues were fixed in 4% paraformaldehyde/PBS overnight at 4°C, then incubated in 100 mM phosphate buffer (pH 7.3) containing 30% sucrose. Cryopreserved 10 μm sections were stained with the ‘in situ cell death detection’ kit (Boehringer Mannheim), following the manufacturer’s instructions.

Cell death was quantified in neural retinae using an enzyme immunoassay (Boehringer Mannheim, catalogue number 1544 675) based on a combination of antibodies recognizing histones and DNA as described previously (Frade et al., 1996). Neural retinae were homogenized in 200 μl PBS containing 1 mM phenylmethylsulfonyl fluoride (Sigma) and centrifuged at 20,000 g for 10 minutes. A sample of the supernatant was used to quantify proteins by standard methods and the rest diluted 1:10 in the buffer provided by the manufacturer and used as described (Frade at al., 1996). Absorbance values were normalized with respect to the values obtained with control retinae.

Counts of TUNEL-positive nuclei and ISL1-positive cells were carried out on at least 10 sections from each embryo, followed by averaging the values per section. Since one-way ANOVA analysis between p75^NTR+/+ and ngf+/+ embryos showed non-significant differences in all the cases, these values were grouped.

Cell death in early stages of mouse retina development has been described using conventional histological methods (see in particular Silver and Robb, 1979). Starting at E13.5, programmed cell death was quantified at various embryonic ages using a technique that measures soluble nucleosomes (see also Frade et al., 1996; Frago et al., 1998). Even though substantial variability was observed, somewhat higher levels were found to occur between E15.5 and E17.5 than either before or after this period (Fig. 1A). This time period roughly coincides with the peak of retinal ganglion cell generation (Sidman, 1961). The distribution of dying cells in the retina was monitored using TUNEL staining, and some labelled nuclei were observed mostly located within the neuroepithelium of the central retina, close to the optic nerve exit (Fig. 1B,C). The small number of dying cells (compared with previous observations in the avian retina using similar techniques, see Frade et al., 1996) presumably explains the variability observed in the nucleosome assay.

We then examined whether p75^NTR is expressed in the retina during the peak of programmed cell death. Sections of E15.5 mouse retinae were incubated with a p75^NTR antiserum, and staining could be observed, mostly associated with the retina ganglion cells and their axons located in the central retina (Fig. 2A,B). No staining was detectable in the retinae prepared from p75^NTR mutant animals (Fig. 2C).

The levels of cell death in E15.5 retinae were then compared between wild-type animals and embryos homozygous for mutations in the ngf or p75^NTR genes. On average, significantly less cell death was observed in the retinae of both mutants (17 p75^NTR and 11 ngf mutants) compared with wild-type animals (Fig. 3), though as with wild-type animals (see above), the levels of cell death were found to be quite variable.

In addition to the retina, cell death during early stages of neurogenesis has also been described in numerous other structures including the spinal cord (for work in avian embryos, see in particular Homma et al., 1994; Yaginuma et al., 1996).

TUNEL-positive nuclei were counted on consecutive sections of mouse brachial spinal cord at stages roughly corresponding to those previously examined in the chick. Between E10.5 and E13.5, a period during which commissural axons grow within the mouse spinal cord towards the floor plate (Serafini et al., 1996), TUNEL-positive nuclei were found scattered through the ependymal zone, including the roof plate, and the mantle zone (Table 1). The ependyme consists mostly of proliferating precursors and newborn neurons that migrate to the mantle zone, the future gray matter in the adult. Of note is the clear reduction in the number of apoptotic nuclei within the ependymal zone as development proceeds. Unlike the ependymal region, the mantle zone showed constant numbers of apoptotic nuclei until E13.5. The increase observed in the ventral portion of the spinal cord at E13.5 corresponds to the beginning of the target-dependent cell death of motor neurons (Lance-Jones, 1982; Oppenheim et al., 1986).

We next examined by immunohistochemistry the distribution of p75^NTR in the brachial spinal cord at E11.5. While no detectable staining could be observed in the spinal cord of p75^NTR−/− embryos (Fig. 4C), large numbers of p75^NTR immunoreactive cells were detected in the mantle zone. Only a few cells were labelled in the ependymal zone. Cells in the roof plate were also stained (Fig. 4A,B).

TUNEL-positive nuclei were quantified at E11.5 in wild-type, as well as in p75^NTR−/− and ngf−/− mutant embryos (Fig. 5A-C; Table 2). In both, a significant decrease in the number of apoptotic nuclei was noted in the ependymal and in the mantle zones. Presumably as a result of decreased cell death in the mutants, the thickness of the mantle zone was clearly increased (Fig. 6A,E,I). To begin to address the question of the identity of the supernumerary cells in the mutants, we used ISL1 antibodies known to stain specific cell types in the developing vertebrate spinal cord, including in particular motoneurons in the ventral horn (Ericson et al., 1992; Tsuchida et al., 1994). Even though the developing ventral cord is larger in both mutants, the number of ISL1-positive cells was similar to that found in control spinal cords (Table 2; Fig. 6B,D,F,H,J,L). However, more ISL-1 positive cells were present in the dorsal region of the spinal cord in both mutants, compared with wild type controls (Table 2; Fig. 6B,C,F,G,J,K). ISL-1-positive cells were also observed in the intermediate part of the ependymal zone, outside of the area in which they can be detected in wild-type embryos (Fig. 6G,K).

As developmental cell death is particularly well documented in dorsal root ganglia (DRG), a structure that is also known to express p75^NTR and to contain NGF-dependent neurons, we also determined the number of TUNEL-positive nuclei in this structure. At E11.5, the number of TUNEL-positive cells in wild-type or p75^NTR mutant animals was not different (Table 2; Fig. 5A,B). However, there was a clear increase in the number of dying cells in the DRGs of ngf−/− compared to control embryos (Table 2; Fig. 5A,C).

Our results with mice carrying mutations in the ngf and p75^NTR genes indicate that a sizeable proportion of the early cell death observed in the developing retina and in the spinal cord is mediated by NGF acting through p75^NTR. In the CNS, both mutants showed a phenotype that is indistinguishable with regard to decreased cell death. As NGF is a well-known ligand of p75^NTR, the present results using a genetic approach thus further support the notion that the interaction between NGF and p75^NTR causes cell death in the CNS. The decrease in cell death in the spinal cord of the mutants was accompanied by an increase in the thickness of the mantle zone, and some of the supernumerary cells could be characterised on the basis of their immunoreactivity for ISL1.

Cell death in the developing mouse neural retina is closely associated with the generation of retinal ganglion cells and with the initial phase of axonal elongation (Silver and Hughes, 1973; Silver and Robb, 1979), in line with observations in avian embryos (Cuadros and Rios, 1988). In agreement with these previous histological observations, we found that most TUNEL-positive cells are located close to the optic disc, suggesting that this cell death could serve the purpose of creating spaces for axons forming the optic nerve (Silver and Robb, 1979; Cuadros and Rios, 1988). Compared with the chick retina, we note that the much smaller, more slowly developing mouse retina shows considerably less cell death. At any time point, only small numbers of TUNEL-positive cells were observed in the mouse, and as a result, substantial variability was observed in the values obtained in the nucleosome assay. In some experiments (data not shown), we also compared cell death between the two retinae of one animal and found different levels of cell death, suggesting that neither genetic background nor different stages of development are likely explanations to account for variability. We note that using histological methods, Homma et al. (1994) previously described variability in the number of pyknotic cells in the chick spinal cord. In view of this problem, we analysed relatively large numbers of both wild-type and mutant embryos, and found upon averaging a clear and significant reduction in cell death in both the ngf and the p75^NTR mutants. This result is in line with our previous observations in avian retinae using an antibody-based approach (Frade et al., 1996). Both the neutralisation of endogenous NGF and the injection of antibodies blocking the binding of NGF to p75^NTR (also expressed by newly generated retinal ganglion cells) led to a decrease in programmed cell death.

Previous work by Oppenheim and colleagues has shown that in the chick embryo, clusters of dying cells are already observed at stage 16 (ca. 51-56 hours; Hamburger and Hamilton, 1951) in the brachial region of the spinal cord (Homma et al., 1994). In addition, extensive cell death in the cervical spinal cord of chick embryos was noted at E4-E5, shortly after motoneurons exit the cell cycle (Yaginuma et al., 1996). We did not observe in the mouse clusters of TUNEL-positive nuclei in the brachial spinal cord but instead, scattered dying cells throughout the ependymal zone and the mantle zone between E10.5-E12.5. During this time period, commissural axons grow towards the floor plate through the mantle zone (Serafini et al., 1996), suggesting that, as occurs in the retina, the purpose of early cell death in the mantle zone may be to create space for axons. This possibility has been previously discussed with regard to commissural axons and cell death in the ventral and floor plate regions of chick embryos (Homma et al., 1994). As many cells in the developing spinal cord express p75^NTR, we examined spinal cords from p75^NTR or ngf mutant mice and compared them with control littermates. A striking increase in the thickness of the mantle zone was observed, accompanied by a decrease in TUNEL-positive cells, mostly associated with this zone. Using an antibody to ISL1, an early motoneuron marker, we were surprised to see that the number of ISL1-positive cells was not increased in the ventral portion of the spinal cord, even though the thickness of this area was considerably increased in both the p75^NTR and ngf mutants. While the phenotype of the supernumerary cells remains to be established, the dorsal region of both mutant spinal cords showed an increase in the number of ISL1-positive cells compared to the wild-type littermate controls. In the chick, these cells have been described as interneurons (Liem et al., 1997), and our results suggest that a significant proportion of this population is eliminated during normal development by mechanisms depending on NGF and p75^NTR. In the chick, early cell death in the spinal cord has been hypothesized to represent a form of negative selection of inappropriate phenotypes or precursor cells (Homma et al., 1994).

Cell death caused by NGF corresponds to the time when post-mitotic neurons are being generated, many of which expressing p75^NTR. In this study, p75^NTR could be localised to retina ganglion cells and their axons in the central region of the E15.5 retina and in many cells of the mantle zone of the E11.5 brachial spinal cord. Fewer cells were labelled in the ependymal zone. As previously observed in the chick retina (Frade et al., 1996), it is clear that many more cells express p75^NTR than die, and it is possible that the expression of p75^NTR by newly born neurons may mark them for potential elimination by NGF-dependent mechanisms. P75^NTR expression may introduce the option to kill cells that need to be eliminated when they are in the way of growing axons. Their remnants would be likely to be eliminated by the phagocytic function of microglial cells that also carry the ligand activating p75^NTR, NGF, as previously demonstrated in the avian retina (Frade and Barde, 1998). In the retina and in the mantle zone of the spinal cord, the primary trigger of cell death may then be axonal outgrowth, a possibility that can be tested by examining appropriate mutants. In this context, it is of interest to note that cell death is drastically reduced in the mutation ocular retardation or^J (Silver and Robb, 1979). The mutation affects the Chx10 gene (Burmeister et al., 1996) and retinal ganglion cell axons are unable to exit from the eye and to form the optic nerve (Silver and Robb, 1979).

While expression of p75^NTR by post-mitotic neurons is necessary to cause NGF-mediated cell death, it cannot be used to predict whether or not neurons will be eliminated. Not only does the proportion of p75^NTR-expressing neurons in the CNS far exceed the number of dying cells, but also in the peripheral nervous system, there is no evidence that the expression of p75^NTR correlates in any way with NGF-mediated cell death. For example in the DRG, while dying cells have been observed in this and numerous previous studies, we did not observe any reduction of cell death at E11.5 in ngf−/− or p75NTR−/− mice. On the contrary, the ngf−/− mutant showed an increase in cell death, in agreement with previous results by others (see for example White et al., 1996). Presumably, the dying cells also express the trophic receptor trkA, and its activation by NGF is likely to inhibit any proapoptotic action of p75^NTR (see, for example, Yoon et al., 1998). It is interesting to note that in yet other structures such as the sympathetic ganglia, where trkA is abundantly expressed, recent results indicate that BDNF seems to take on the role of NGF as a pro-apoptotic factor (Bamji et al., 1998). These authors have also provided evidence for the fact the killing action of BDNF requires p75^NTR (Bamji et al., 1998).

In conclusion, the phenotype of mice carrying mutation in the ngf or the p75^NTR genes is indistinguishable with regard to early cell death in the CNS. The temporal correlation between the period of axonal outgrowth and cell death in the retina and the mantle zone of the spinal cord is suggestive of a mechanism allowing for rapid cell killing, thus generating space for axonal tracts to develop. It is also possible that early cell death in neuroepithelia may contribute to the elimination of inappropriate neuronal phenotypes, as previously suggested by Oppenheim and colleagues (Homma et al., 1994).

We are grateful to T. Edlund for the gift of antibodies used in the present work. This work was supported in part by the European Union, Biotechnology Programme (PL 960024).