p75 Neurotrophin Receptor-mediated Neuronal Death Is Promoted by Bcl-2 and Prevented by Bcl-xL *

The p75 neurotrophin receptor (p75NTR) has been shown to mediate neuronal death through an unknown pathway. We microinjected p75NTR expression plasmids into sensory neurons in the presence of growth factors and assessed the effect of the expressed proteins on cell survival. We show that, unlike other members of the TNFR family, p75NTR signals death through a unique caspase-dependent death pathway that does not involve the “death domain” and is differentially regulated by Bcl-2 family members: the anti-apoptotic molecule Bcl-2 both promoted, and was required for, p75NTR killing, whereas killing was inhibited by its homologue Bcl-xL. These results demonstrate that Bcl-2, through distinct molecular mechanisms, either promotes or inhibits neuronal death depending on the nature of the death stimulus.

The p75 neurotrophin receptor (p75NTR) has been shown to mediate neuronal death through an unknown pathway. We microinjected p75NTR expression plasmids into sensory neurons in the presence of growth factors and assessed the effect of the expressed proteins on cell survival. We show that, unlike other members of the TNFR family, p75NTR signals death through a unique caspase-dependent death pathway that does not involve the "death domain" and is differentially regulated by Bcl-2 family members: the anti-apoptotic molecule Bcl-2 both promoted, and was required for, p75NTR killing, whereas killing was inhibited by its homologue Bcl-x L . These results demonstrate that Bcl-2, through distinct molecular mechanisms, either promotes or inhibits neuronal death depending on the nature of the death stimulus.
The neurotrophin receptor, p75NTR, 1 is required along with the trkA receptor for high affinity NGF binding and neuronal survival (1); however, more recently it was shown to be involved in promoting cell death by an unknown pathway. Overexpression of p75NTR increased neuronal death both in vitro and in vivo (2)(3)(4), and lowering the expression of p75NTR was shown to prevent neuronal death after growth factor withdrawal in vitro, and after sciatic nerve axotomy (5,6). Additionally, p75NTR-dependent cell death occurred after activation by NGF or brain-derived neurotrophic factor (BDNF) in cells, respectively, lacking trkA or trkB expression (7)(8)(9)(10).
p75NTR is a member of the tumor necrosis factor receptor (TNFR) superfamily, showing homology in the extracellular ligand binding domain and a cytoplasmic motif known as the "death domain," so called because of the cytotoxic actions of proteins containing the domain (11). The death signaling pathway of the TNFR family is well characterized, typically involving other "death domain"-containing proteins (12), and is not inhibited by the anti-apoptotic proteins Bcl-2 or Bcl-x L . These Bcl-2 family members are well characterized inhibitors of stress-induced cell death, blocking neuronal cell death in a variety of models (13)(14)(15).
Herein we investigate the role of the death domain and Bcl-2 family members in mediating p75NTR death signaling using microinjection technology to deliver expression plasmids into cultured neurons. We show that the p75NTR death signaling pathway is unlike the TNFR death signaling pathway as an intracellular juxta-membrane domain, and not the death domain, of p75NTR is sufficient and required to mediated death signaling. Secondly, the anti-apoptotic Bcl-2 family members that have little influence on TNFR death signaling have profound effects on the p75NTR death pathway.

MATERIALS AND METHODS
Cell Culture-Dorsal root ganglia were dissected from postnatal day zero C57Bl/6 or Bcl-2-deficient mice and plated in 3-cm tissue dishes precoated with poly-DL-ornithine (500 g/ml, Sigma) and laminin (20 g/ml, Life Technologies, Inc.) at a density of 5000 neurons/dish (5). Cells were grown in Monomed medium (CSL, Melbourne, Australia) containing 1% fetal bovine serum and leukemia inhibitory factor (LIF, AMRAD, Australia) or nerve growth factor (2.5 S NGF, 50 ng/ml, Alomone Laboratories). Survival of sensory neurons was assessed by morphological criteria (5) and propidium iodide exclusion. Heterozygous mice containing a disrupted Bcl-2 gene, (D. Loh, Roche, Japan) were mated to produce litters contains homozygous Bcl-2 knock-out mice. The genotypes of newborn mice were determined by polymerase chain reaction, and results were confirmed by staining of thymocytes with an anti-mouse Bcl-2 antibody (PharMingen) before dissection and microinjection of neurons isolated from individual mice of appropriate genotype.
Microinjection-Sensory neurons were injected into the nucleus with a solution containing plasmid (100 g/ml, with the exception of Bcl-2, which was at 50 g/ml), tetramethylrhodamine dextran ("fluoro-ruby," 0.15%, Molecular Probes), and phosphate-buffered saline. Where more than one plasmid was expressed in a single condition, only one injection of solution containing all plasmids was made, with the individual plasmid concentrations as specified above. Approximately 70 neurons/well were injected, with two or three wells comprising each condition. At least 2 h after completion of the injections, the number of fluoro-ruby containing cells that had survived the injection procedure was counted, and this provided the time zero 100% value for each well.
DNA Constructs-The plasmid containing the full-length rat p75NTR cDNA, p75NTR, is described (4), and all p75NTR plasmids are modified versions of this original expression vector. A control plasmid, p75NTRnc (no cytoplasmic domain) is missing the entire cytoplasmic domain except the membrane anchoring Lys-274 and Arg-275. p75NTRtr is truncated with an I308A substitution followed by a stop codon, deleting the entire death domain. p75NTR⌬ retains Lys-274 and Arg-275 and the last 108 amino acids creating a cytoplasmic domain including the death domain but deleted for the 33-amino acid juxtamembrane domain retained by p75NTRtr. All plasmids were constructed using polymerase chain reaction to amplify desired coding regions followed by subcloning the polymerase chain reaction products into plasmid vectors. Details of primers used in construction of these plasmids are available on request. Bcl-2 and Bcl-x L plasmids are previously described (16,17). The modified CrmA has the caspase recognition sequence modified from wild-type Leu-Glu-Ala-Asp to Asp-Gln-Met-Asp.
Yeast Two-hybrid Methods-The p75NTR death domain from Leu-342 on was cloned into pGBT9 (CLONTECH) and used to screen a * This work was supported in part by the National Health and Medical Research Council of Australia. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

RESULTS
To investigate whether p75NTR signals death through interaction with death domain containing proteins or via another pathway, we have activated the p75NTR pathway in a similar fashion to experiments that use ligand-free overexpression of TNFR to study apoptosis (18). We overexpressed p75NTR in neurons by microinjecting a rat p75NTR expression plasmid into the nucleus of mouse dorsal root ganglia sensory neurons ( Fig. 1) and cultured them in the presence of the neural cytokine, LIF, to prevent p75NTR-independent neuronal death. These neurons were chosen because we had previously shown that their death was, at least in part, mediated by p75NTR (5,6). It was found after 16 h that approximately 20 -25% of neurons injected with full-length p75NTR plasmid died ( Fig.  2A), compared with neurons injected with a control ␤-galactosidase plasmid or a truncated p75NTR protein lacking the entire cytoplasmic domain (p75NTRnc) expressed to a similar extent. The majority of the p75NTR-mediated death was observed in the first 16 h (see Fig. 3C for example), comparable  with that reported for TNFR overexpression (18).
Because the cytoplasmic tail of p75NTR is required for the cytotoxic effect, we next examined whether deletion of the death domain also abolished the ability of p75NTR to kill. Surprisingly, the neuronal death observed after expression of p75NTR with a truncated cytoplasmic tail (p75NTRtr), lacking the death domain, was not significantly different from that observed with full-length p75NTR (Fig. 2B). Furthermore, a deletion p75NTR protein (p75NTR⌬), containing the death domain but deleted for the juxta-membrane portion, did not induce significant death (Fig. 2B). Thus, the death domain was not required for p75NTR killing, and the juxta-membrane cytoplasmic tail is necessary and sufficient for killing. The p75NTR death domain has recently been shown to have a different tertiary structure to the death domain of the TNFR family and does not self-associate in vitro (19), supporting the finding that the p75NTR death domain does not function to induce death. Using the yeast two-hybrid method to find p75NTR death domain binding partners, we found no evidence of specific interaction with other death domain-containing proteins (see "Materials and Methods"). Together, these results suggest that an alternative pathway to one involving death domain adapter proteins, such as TRADD and FADD, may be responsible for p75NTR-mediated killing.
To further explore whether p75NTR killing was different from TNFR killing pathways, we examined whether overexpression of the anti-apoptotic Bcl-2 family proteins could inhibit killing by p75NTR. Bcl-2 and Bcl-x L are well characterized inhibitors of growth factor withdrawal and stress-induced apoptosis (15). However, both proteins are poor inhibitors of CD95/Apo1/Fas and TNFR-mediated apoptosis (20,21).
We found that overexpression of Bcl-x L protected neurons against p75NTR-induced death (Fig. 3A), supporting the hypothesis that p75NTR signals through an alternative pathway to TNFR-induced apoptosis. Whereas Bcl-2 overexpression alone had no effect on cell survival in the presence of LIF (Fig.  3B), Bcl-2 in combination with p75NTR overexpression, surprisingly, enhanced the neuronal death seen with p75NTR overexpression alone, yet in combination with p75NTRnc it had no effect (Fig. 3B). The results are surprising because Bcl-2 is functionally indistinguishable from Bcl-x L in most cell-death systems (16,22), yet in this assay we find that they have opposite effects (Fig. 3C). The cell death observed with p75NTR and Bcl-2 overexpression was totally ablated if the cells were cultured in NGF, confirming that signaling though trk may inhibit p75NTR activity (Fig. 3D). Bcl-2 was able to protect against neuronal death induced by NGF withdrawal (Fig. 3E). Thus, at the same expression levels in the same neuronal population, Bcl-2 was able to inhibit or enhance neuronal cell death depending on the nature of the death signal.
To determine whether the paradoxical effect of Bcl-2 on p75NTR-induced killing was related to its known anti-apoptotic activity, inactive Bcl-2 mutants were utilized (17). Like wild-type Bcl-2, expression of the Bcl-2 mutants did not affect neuronal survival. In combination with p75NTR expression, the enhanced killing effect seen with Bcl-2 co-expression was abrogated by expression of the mutant G145E Bcl-2, suggesting the protein was nonfunctional (Fig. 4A). Thus, an intact BH1 homology region is required for both the survival and death promoting activities of Bcl-2. Using a W188A mutated Bcl-2, we found that co-expression with p75NTR not only abolished the increased p75NTR killing but, more importantly, protected neurons from any p75NTR-induced death (Fig. 4B), reminiscent of that seen with Bcl-x L , despite being unable to protect against growth factor withdrawal-induced death (17). Thus, the mechanism by which Bcl-2 participates in the death pathway is separable from its function in the survival pathway.
Bcl-2 has previously been observed to increase cell death when highly expressed both in vitro and in vivo (23-25). Thus, it is possible that the high level of Bcl-2 is able to "prime" the FIG. 3. Effect of Bcl-x L and Bcl-2 expression plasmids on p75NTR-mediated killing. A, Bcl-x L and p75NTR plasmids were co-injected into neurons in the presence of LIF, and survival 16 h after injection was assessed. Bcl-x L was able to totally rescue neurons from p75NTR-mediated cell death. Note also that neurons injected with Bcl-x L had increased survival compared with Fig. 2A. B, although the survival of neurons in the presence of LIF was not affected by expression of Bcl-2 plasmid alone (compare with Fig. 2A), in combination with p75NTR there was a significant increase in cell death. The ability of Bcl-2 to promote cell death because of full-length p75NTR is not seen in the presence of p75NTRnc. C, Bcl-2 and Bcl-x L have opposite effects on the p75NTR-mediated killing within the first 16 h after injection. D, microinjection of either p75NTR alone or p75NTR and Bcl-2 constructs failed to increase neuronal death in the presence of NGF. E, death induced by withdrawal of NGF is inhibited by expression of the microinjected Bcl-2 plasmid. q, uninjected and, f, Bcl-2-injected neurons in NGF; E, uninjected and, ‚, Bcl-2-injected neurons after factor withdrawal. *, p Ͻ 0.05; ** p Ͻ 0.01; *** p Ͻ 0.001; error bars indicate S.E., n ϭ 3.
FIG. 4. Effects of survival-inactive Bcl-2 proteins on p75NTRmediated killing. Mutant Bcl-2 plasmids were injected into neurons in the presence of LIF, either alone or together with p75NTR, and the neurons were assessed for mortality 16 h after injection. A, Bcl-2 G145E expression has no effect on cell survival and does not enhance or protect against the effects of p75NTR killing. B, Bcl-2 W188A totally inhibited p75NTR killing reminiscent of Bcl-x L (compare with Fig. 3A). ** p Ͻ 0.01; ***p Ͻ 0.001; error bars indicate S.E., n ϭ 3. death pathway such that an apoptotic stimulus via p75NTR results in rapid cell death. We examined whether Bcl-2 may not only enhance p75NTR killing but may be essential for p75NTR killing to proceed by testing whether lowering endogenous Bcl-2 would inhibit p75NTR-mediated killing. We used a Bcl-2 antisense plasmid, which has previously been shown to effectively lower Bcl-2 expression after microinjection into neurons (26,27). When the Bcl-2 antisense plasmid was injected at the same time as the p75NTR plasmid, no diminution of the death signal was seen. If, however, the neurons were microinjected in the presence of NGF, to give time for the antisense to deplete endogenous Bcl-2, and then switched into LIF the next day to permit p75NTR killing, there was a significant decrease in the ability of p75NTR to kill neurons with lowered Bcl-2 (Fig. 5A); in fact, the level of killing was not significantly different from that of neurons expressing p75NTRnc (Fig. 5A). The specificity of the Bcl-2 antisense was demonstrated by its inability to affect neuronal survival when expressed alone or in conjunction with p75NTRnc (data not shown). In addition, a control vector expressing green fluorescent protein had no affect on p75NTR killing (data not shown).
The requirement for endogenous Bcl-2 for p75NTR killing was also demonstrated in neurons from newborn mice lacking Bcl-2, which showed a significant reduction in cell death induced by p75NTR compared with wild-type litter mates (Fig.  5B), supportive of the Bcl-2 antisense experiments.
To investigate whether the p75NTR-Bcl-2 death-signaling cascade was dependent on caspase activation, a modified CrmA, designed to inhibit downstream group II caspases, such as caspase 3 (28,29), was overexpressed together with p75NTR. The modified CrmA was able to block the killing induced either by p75NTR alone or by co-expression of p75NTR and Bcl-2 in wild-type neurons (Fig. 5C), suggesting that p75NTR-Bcl-2-induced apoptosis utilizes a caspase-dependent pathway.

DISCUSSION
Bcl-2 has been shown to be cleaved by caspases, with the cleavage product being capable of promoting apoptosis in vitro (30). Cleavage of Bcl-2 may occur in our system but the mutated Bcl-2 proteins would be similarly cleaved, arguing that cleavage of Bcl-2 is unlikely to be the dominant mechanism by which Bcl-2 promotes killing in our system. In addition, if caspase cleavage was the dominant mechanism, Bcl-x L might also be expected to promote death in our system as Bcl-x L has been shown to be cleaved in a similar way (31).
Bcl-2 and Bcl-x L are likely to have a very similar tertiary structure based on sequence similarity. The NMR structure of Bcl-x L has been determined, showing a hydrophobic cleft formed by seven ␣ helixes where other Bcl-2 family proteins interact (32). Bcl-x L , and Bcl-2, may function differently by specific interaction with proteins that mediate p75NTR-induced death. For example, pro-apoptotic Bcl-2 family members, Bak and Bad, interact strongly with Bcl-x L but weakly with Bcl-2 (33,34). Also, Bcl-x L , but not Bcl-2 as yet, has been shown to interact with Apaf-1, the mammalian CED-4 homologue, to regulate apoptosis via caspase-9 activation (35). Mutation of Bcl-2 at Gly-145 or Trp-188 may affect its interaction with proand anti-apoptotic proteins by conformational changes disrupting access to the cleft. This suggests that the conformation of the Bcl-2 protein is integral to the paradoxical functions observed herein and that Bcl-2 participates in the p75NTR-killing pathway and stress-induced survival by different molecular mechanisms.
The different neuronal phenotype observed in Bcl-2-and Bcl-x L -deficient animals is consistent with our results that Bcl-2 and Bcl-x L can have quite different functions in regulating neuronal death. Mice deficient for Bcl-x L have a major loss of neurons shortly after their differentiation (36), whereas Bcl-2-deficient mice show only minor loss of neurons during embryogenesis and early neonatal life (37).
The latter phenotype would support the idea that during development Bcl-2 plays two contrasting roles: under conditions where there are appropriate neuritic connections leading to trk signaling, Bcl-2 may act to enhance neuronal survival, as has been shown for trigeminal sensory neurons in vitro (26,27); alternatively, as observed, Bcl-2 may also be capable of promoting neuronal death at other stages, for example when neurons may become more dependent on neural cytokine (e.g. LIF, CNTF, OsM) support in early postnatal development (38,39). The LIF family signal through Janus tyrosine kinase/signal transducer and activator of transcription (JAK/STAT) pathways, which can be inhibited by the neurotrophin-activated mitogen-activated protein (MAP) kinase pathway (40). Thus, it is possible that the in vivo switch of neuron dependence between neurotrophins and neural cytokines may prime the neurons to cell death. Activation of p75NTR then may occur FIG. 5. Bcl-2 and caspases are required for p75NTR-induced cell death. A, neurons expressing both p75NTR and antisense Bcl-2 show a significant increase in survival compared with p75NTR alone and have no increased mortality compared with neurons expressing control plasmid p75NTRnc. B, neurons from Bcl-2-deficient mice and their in wild-type littermates were injected with either p75NTR or p75NTRnc, and the degree of killing because of p75NTR was calculated. Bcl-2-deficient neurons are significantly less susceptible to killing by p75NTR overexpression. C, neurons were injected with either modified CrmA, or p75NTR and Bcl-2, or all three together. Expression of modified CrmA alone had no effect on neuronal survival, but significantly inhibited death induced by p75NTR and Bcl-2 expression. * p Ͻ 0.05; ** p Ͻ 0.01; error bars indicate S.E., n ϭ 3.
through NGF or BDNF signaling as has been recently shown (10,41).
The apparent paradoxical actions of Bcl-2 may be a program by which rapid selection of cell survival or death occurs during neuronal development and after nerve injury. There are high levels of Bcl-2 in neurons of both the central and peripheral nervous systems during periods of developmental cell death (42), and activation of p75NTR, also expressed widely in the nervous system during development, in the presence of high Bcl-2, would lead to rapid apoptosis.