ProNGF induces TNFα-dependent death of retinal ganglion cells through a p75^NTR non-cell-autonomous signaling pathway

Abstract

Neurotrophin binding to the p75 neurotrophin receptor (p75^NTR) activates neuronal apoptosis following adult central nervous system injury, but the underlying cellular mechanisms remain poorly defined. In this study, we show that the proform of nerve growth factor (proNGF) induces death of retinal ganglion cells in adult rodents via a p75^NTR-dependent signaling mechanism. Expression of p75^NTR in the adult retina is confined to Müller glial cells; therefore we tested the hypothesis that proNGF activates a non-cell-autonomous signaling pathway to induce retinal ganglion cell (RGC) death. Consistent with this, we show that proNGF induced robust expression of tumor necrosis factor alpha (TNFα) in Müller cells and that genetic or biochemical ablation of TNFα blocked proNGF-induced death of retinal neurons. Mice rendered null for p75^NTR, its coreceptor sortilin, or the adaptor protein NRAGE were defective in proNGF-induced glial TNFα production and did not undergo proNGF-induced retinal ganglion cell death. We conclude that proNGF activates a non-cell-autonomous signaling pathway that causes TNFα-dependent death of retinal neurons in vivo.

The four mammalian neurotrophins comprise a family of related secreted factors that are required for differentiation, survival, development, and death of specific populations of neurons and nonneuronal cells. Neurotrophins are produced as proforms of ∼240 amino acids that are cleaved by furins and proconvertases to yield products of ∼120 amino acids. Recent studies have indicated that nerve growth factor (NGF) and brain-derived neurotrophic factor (BDNF) can be secreted as proforms in the central nervous system (CNS) (1 –3) and demonstrated that proneurotrophins can function as potent apoptosis-inducing ligands both in vitro and in vivo (4). However, the precise mechanisms by which proneurotrophins lead to neuronal death are poorly defined.

The biological effects of neurotrophins are mediated by binding to TrkA, TrkB, and TrkC receptor tyrosine kinases and to the p75 neurotrophin receptor (p75^NTR). Trk receptors respond preferentially to mature neurotrophins whereas proneurotrophins exert their apoptotic effect via a receptor complex that contains p75^NTR and sortilin (5). The precise signaling cascades evoked by occupancy of the p75^NTR–sortilin complex remain to be elucidated, but several lines of evidence indicate that NRIF and NRAGE adaptor proteins play key roles in death signaling cascades evoked by p75^NTR (6, 7).

Previous studies have shown that neurotrophins induce cell death via p75^NTR during early retinal development (8). p75^NTR has also been implicated in light-induced photoreceptor death in adult rodents in vivo (9) and a proNGF-p75^NTR link has been proposed to facilitate apoptosis in a retinal cell line (10). Here, we investigate the role of proNGF in the adult retina and demonstrate that proNGF promotes death of retinal ganglion cells (RGCs) in vivo. Importantly, proNGF-induced RGC loss is indirect and requires the p75^NTR-dependent production of tumor necrosis factor-alpha (TNFα) by Müller glial cells. Therefore, proNGF-induced neuronal loss in the adult retina occurs through a non-cell-autonomous mechanism.

Results

ProNGF Induces Death of Retinal Ganglion Cells in Adult Rodents.

To investigate whether proNGF promotes neuronal death in vivo, we first retrogradely labeled RGCs of adult rats by applying fluorogold to the surface of the superior colliculus and then provided a single intraocular injection of proNGF or vehicle. A week later, retinal whole mounts were prepared and RGC densities were quantified. ProNGF caused a profound loss of adult rat RGCs, whereas vehicle injection had no effect on cell death (Fig. 1A). To determine if the effect of proNGF on neuronal survival was specific to the proform of this neurotrophin, we asked whether mature NGF could similarly promote RGC death and found that neuronal density was not altered by mature NGF treatment (Fig. 1A). The effect of proNGF was not species specific as proNGF also caused a marked loss of RGCs in mice subjected to intraocular proNGF injection (Fig. 1B). We conclude that elevation of proNGF levels within the retina promotes neuronal loss and used this system as a model for examining the cellular details of proNGF-induced cell death in vivo.

Fig. 1.

Exogenous proNGF leads to marked RGC loss in the adult rodent eye. Quantitative analysis is shown of RGC survival in rat (A) and mouse (B) retinas at 1 week after intraocular injection of proNGF (solid bars), vehicle (PBS, shaded bars), or NGF (bars with horizontal lines). The density of RGCs in intact, noninjected retinas is shown as reference (open bars). Data are expressed as RGC densities (RGCs/mm²; mean ± SEM; ANOVA, *P < 0.001). The number of animals used in each experiment is shown above the corresponding graph bar.

p75^NTR, Sortilin, and NRAGE Are Required for proNGF-Induced Death of Retinal Ganglion Cells.

To determine if p75^NTR was required for the loss of RGCs evoked by exogenous proNGF, we first asked whether coinjection of the p75^NTR function-blocking antibody REX (11) antagonized proNGF-induced neuronal death. Coadministration of proNGF and REX resulted in significant rescue of RGCs in mice, whereas combined proNGF and nonspecific Ig did not exert a protective effect (Fig. 2A). As an alternative approach, we assessed the apoptotic effect of proNGF in mice deficient for p75^NTR and showed that proNGF-induced loss of RGCs did not occur in p75^NTR null retinas. Together, these data indicate that proNGF binding to p75^NTR is required to induce RGC death.

Fig. 2.

p75^NTR, sortilin, and NRAGE are required for proNGF-induced death of RGCs. (A) Treatment with REX, a p75^NTR function-blocking antibody, promoted RGC survival in the presence of proNGF whereas coadministration of a nonspecific, control Ig did not exert a neuroprotective effect (ANOVA, *P < 0.001). RGCs from mice eyes deficient in p75^NTR (B), sortilin (C), or NRAGE (D) were resistant to proNGF-induced death (Student’s t test, P > 0.05). Values are expressed as RGC densities (RGCs/mm²; mean ± SEM).

p75^NTR and sortilin have been shown to form a cell surface receptor complex for proneurotrophins that is required for activation of downstream apoptotic pathways (5), and we previously demonstrated that NRAGE, a p75^NTR adaptor protein, plays an obligatory role in p75^NTR-dependent death (6). We therefore asked whether sortilin or NRAGE was required for proNGF-induced RGC loss in vivo. Our data show that RGCs within sortilin or NRAGE null mice retinas were protected from proNGF-induced death (Fig. 2 C and D). We conclude that p75^NTR, sortilin, and NRAGE are each required for proNGF-induced death of RGCs in the adult retina.

ProNGF Kills Retinal Ganglion Cells Through Glia-Mediated Production of TNFα.

A simple model to explain these results would have proNGF binding to the p75^NTR–sortilin complex on the surface of RGCs that in turn activates a cell-autonomous, NRAGE-dependent proapoptotic pathway. However, it has been reported that Müller glia are the only cells that express p75^NTR in the adult retina (9, 12 –15), implying that proNGF killing of RGCs may involve a more complex mechanism. To confirm the p75^NTR expression pattern in the adult mouse retina, RGCs were retrogradely labeled with fluorogold and then stained with antibodies specific for p75^NTR. In both naive and proNGF-treated animals, RGCs were uniformly negative for p75^NTR whereas neighboring cells and processes typical of Müller glia invariably expressed abundant quantities of this receptor (Fig. 3A). Costaining with the Müller cell-specific marker CRALBP confirmed that p75^NTR is abundantly expressed by Müller glia, but not RGCs (Fig. 3B). Previous studies have shown that sortilin is expressed by Müller glia in the mouse retina (16), and when we examined the distribution of NRAGE in this system, we found that this p75^NTR adaptor protein is abundantly expressed in Müller cell soma and processes (Fig. 3C).

Fig. 3.

p75^NTR and p75^NTR-induced TNFα are expressed by Müller cells in the adult mouse retina. (A) Confocal microscopy images show absence of p75^NTR within Fluorogold-positive RGCs, demonstrating that adult RGCs are devoid of p75^NTR. (B) Double immunolabeling with antibodies against p75^NTR (REX) and the Müller cell marker cellular retinaldehyde-binding protein (CRALBP) shows strong expression of p75^NTR in Müller cell processes surrounding RGCs. (C) Fluorescent microscopy images show that NRAGE protein colocalizes with Müller cell soma and processes, visualized with CRALBP. (D) Robust TNFα protein induction was observed in retinas exposed to proNGF compared to low, basal levels in eyes injected with vehicle or recombinant, mature NGF. (E) Confocal microscopy images show that TNFα colocalizes with CRALBP-positive Müller cells. GCL, ganglion cell layer; INL, inner nuclear layer; IPL, inner plexiform layer; ONL, outer nuclear layer; OPL, outer plexiform layer. (Scale bars: A and B, 10 μm; C–E, 50 μm.)

On the basis of these results, we hypothesized that proNGF promotes RGC death through an indirect pathway by stimulating the production of a proapoptotic factor by Müller cells. A candidate proapoptotic factor downstream of p75^NTR is tumor necrosis factor alpha (TNFα) because exogenous and endogenous TNFα can induce death of retinal neurons (17 –19) and because p75^NTR activates NF-κB, a transcription complex that is a potent inducer of TNFα production (20 –22). To address whether TNFα could play a role in proNGF-induced RGC killing, we first determined if retinal TNFα levels were increased in eyes injected with proNGF. Immunostaining showed that in eyes injected with vehicle or mature NGF, TNFα basal levels were low. In contrast, eyes injected with proNGF showed robust TNFα expression, both in cell bodies in the inner nuclear layer and within processes that extended radially across the breadth of the retina (Fig. 3D). Double immunocytochemistry using antibodies against CRALBP identified these TNFα-expressing cells as Müller glia (Fig. 3E).

The finding that proNGF stimulates TNFα production by Müller cells prompted us to ask whether RGC death induced by this proneurotrophin could be blocked by Etanercept, a recombinant TNFα antagonist in which the extracellular ligand-binding domain of TNFα receptor 2 (TNFR2) is fused to an Fc fragment (23). Fig. 4A shows that intraocular injection of Etanercept markedly blocked RGC death induced by proNGF. To rule out the possibility that Etanercept may have off-target pharmacological effects and to further substantiate a role for TNFα in proNGF-induced killing, we also examined whether proNGF led to RGC loss in TNFα null mice. Our data show that proNGF administration failed to induce RGC death in TNFα null mice (Fig. 4B), indicating that TNFα plays a crucial role in RGC death induced by proNGF.

Fig. 4.

Sortilin and NRAGE cooperate with p75^NTR to transduce proNGF-induced signaling events required for TNFα production. (A) Coadministration of proNGF with the TNFα inhibitor Etanercept leads to striking RGC neuroprotection, but RGCs did not survive in Fc-treated control retinas (ANOVA, *P < 0.001). (B) RGCs from TNFα null mice were resistant to proNGF-induced death (Student’s t test, P > 0.05). (C) Western blot analyses of retinal extracts show that whereas TNFα protein levels are below the detection limit in wild-type retinas, exposure to proNGF elicits robust TNFα protein up-regulation within 48 h of proNGF injection. ProNGF-induced TNFα production is completely lost in p75^NTR null, sortilin null, and NRAGE null animals (n = 5 mice/group). Lower blots were probed with antibodies that recognize p75^NTR, sortilin, or NRAGE, respectively, and actin was used as a control for equal protein loading.

These data indicate that proNGF causes RGC death indirectly, by stimulating production of TNFα by Müller glial cells. We therefore tested whether p75^NTR and sortilin activated an NRAGE-dependent pathway that resulted in TNFα production. For this purpose, TNFα protein levels were compared in retinas derived from wild-type mice vs. mice lacking p75^NTR, sortilin, or NRAGE following proNGF or vehicle injection. In wild-type retinas, robust TNFα protein production was detected within 48 h of proNGF administration, wheras PBS injection had no effect on TNFα levels (Fig. 4C). In contrast, proNGF-induced TNFα up-regulation did not occur in p75^NTR, sortilin, and NRAGE null mice (Fig. 4C), correlating with the failure of proNGF to induce RGC death in these null lines (Fig. 2 B–D). Importantly, intraocular injection of TNFα caused RGC death in both wild-type and p75^NTR null strains, confirming that TNFα is the actual factor that kills these neurons (Fig. S1). Together, these data indicate that sortilin and NRAGE cooperate with p75^NTR to transduce proNGF-induced signaling events required for TNFα production by Müller glial cells and that the TNFα kills RGCs through a p75^NTR-independent mechanism.

Discussion

This study reports three major findings. First, proNGF leads to striking neuronal death in the adult rodent retina in vivo. Second, proNGF mediates RGC death indirectly via a non-cell-autonomous mechanism that involves TNFα production by Müller glia. Third, proNGF-induced death requires p75^NTR, sortilin, and NRAGE, which cooperate to stimulate TNFα production by Müller cells.

p75^NTR is an important neuronal signaling protein that interacts with numerous ligands and coreceptors to exert a wide range of functions. The role of p75^NTR as an apoptotic receptor is well established as it has been shown to facilitate developmental cell death of peripheral sympathetic neurons and early retinal neurons (8, 24 –26). p75^NTR has also been implicated in cell loss following various forms of CNS injury and promotes apoptosis of cortical and hippocampal neurons, basal forebrain neurons, oligodendrocytes, and photoreceptor cells (8, 27 –31). Overall, the functional role of p75^NTR in neuronal death is best understood in neurons that endogenously express this receptor. In this study, however, we identify a unique mechanism by which p75^NTR in glial cells can profoundly influence neuronal death by activating production of neurotoxic TNFα in the adult retina in vivo.

Recent studies have indicated that proNGF, but not NGF, functions as a potent proapoptotic ligand that binds to a cell surface complex of p75^NTR and sortilin (5). Consistent with this, we showed that fully processed NGF does not share the apoptotic effect of proNGF and that proNGF is unable to induce retinal cell death in mice rendered null for sortilin. NRAGE is a crucial adaptor protein required for p75^NTR-dependent apoptosis in cell lines, in primary cells, and in vivo (6). Consistent with this, we found that NRAGE is also necessary for proNGF-induced death of RGCs within the adult retina. Previous studies have established that RGCs express abundant p75^NTR during retinal development (25) and that p75^NTR and sortilin collaborate to induce NGF-dependent apoptosis of a subset of these neurons at embryonic day (E)15 (8, 25, 26). Therefore, it is possible that proNGF induces cell death in the adult retina by directly binding to a p75^NTR–sortilin complex on adult RGCs. This is very unlikely, however, because examination of p75^NTR expression in the adult rodent retina using light and electron microscopy has established that retinal p75^NTR expression is confined to Müller glial cells (9, 12 –15). A recent study reported that sortilin is expressed by Müller glia in the adult retina (16), and the data shown here indicate that NRAGE is also expressed in these glial cells. Therefore, association of proNGF with p75^NTR and sortilin on Müller cells must induce death of RGCs through an indirect, non-cell-autonomous pathway.

Pharmacological or genetic blockade of TNFα dramatically reduced proNGF-induced RGC death, indicating that Müller cell-derived TNFα plays a crucial role in neuronal loss. There are several direct and indirect mechanisms by which TNFα produced by Müller cells could kill RGCs but two predominate. First, previous work established that TNFα can directly kill primary cortical neurons via a caspase 8-dependent mechanism in models of excitotoxicity (32, 33). It is thus possible that direct caspase activation occurs in RGCs. However, caspase-8 inhibitors that we validated in vivo and in vitro did not block proNGF-induced death of RGCs. Thus, we favor a second possibility, which involves the recent discovery that TNFα promotes the selective insertion of Ca²⁺-permeable AMPA receptors (AMPARs) into the neuronal cell surface (34, 35). Under physiological circumstances, TNFα-mediated increase in Ca²⁺-permeable AMPARs plays a role in synaptic scaling (36) but during injury, TNFα-mediated increase in these channels can facilitate death of primary hippocampal and spinal cord neurons (37, 38). Importantly, this mechanism of cell death may play a role in RGC loss after injury. For example, RGCs lacking TNFR1 are protected following optic nerve crush (39), and we recently showed that TNFα-mediated increase in cell surface Ca²⁺-permeable AMPARs leads to RGC loss in vivo (15).

Glaucoma is a group of diseases characterized by progressive optic nerve degeneration leading to visual field loss and irreversible blindness. A common characteristic of all forms of glaucoma is the death of RGCs (40). Of significant interest, TNFα and TNFR1 are up-regulated in human donor eyes with glaucoma (19, 41, 42), TNFα levels are increased in the aqueous humor of glaucoma patients (43), and TNFα gene polymorphisms have been correlated with primary open angle glaucoma (44, 45). Therefore, an important priority in future studies will be to determine if the proNGF, p75^NTR, and TNFα cascade described here contributes to retinal degeneration and optic neuropathy in preclinical models of disease and in humans.

Several recent studies have highlighted the important role that non-cell-autonomous mechanisms play in neuronal cell death, notably in models of amyotrophic lateral sclerosis, spinocerebellar ataxia, and Huntington disease (46). TNFα is a hallmark of acute and chronic neuroinflammation and has emerged as an important candidate for mediating non-cell-autonomous effects in these conditions (47). p75^NTR expressed by cholinergic neurons is required for production of an unknown factor that facilitates GABAergic neuron development (48), and p75^NTR expressed on Müller glial cells mediates light-induced photoreceptor death via a non-cell-autonomous pathway (9). The potential role of TNFα downstream of these p75^NTR-dependent events will be an interesting topic for future work.

In summary, our work demonstrates that proNGF can induce neuronal death in the retina through a non-cell-autonomous mechanism that involves the activation of p75^NTR and production of TNFα by Müller glial cells. These findings raise the possibility that non-cell-autonomous events may be a general feature of p75^NTR-dependent cell apoptosis in vivo.

Materials and Methods

Experimental Animals.

Procedures were carried out in adult C57BL/6 transgenic or wild-type littermate control mice, with the exception of results shown in Fig. 1A that were carried out in adult Sprague-Dawley rats. All animal procedures were performed in accordance with the policies on the Use of Animals in Neuroscience Research and the Canadian Council on Animal Care guidelines (49). p75^NTR (50), TNFα (51), and NRAGE (6) null mice have been previously described. To inactivate the sortilin gene in ES cells we used the recombination cloning vector pML. A 4.6-kb fragment of the 5′-flanking genomic sequence and a 3.2-kb fragment of the 3′-flanking region of sortilin were subcloned upstream and downstream, respectively, of the neomycin resistance gene within the vector. The Neomycin/G418 in the pML vector was used for positive selection. This vector contains a thymidine kinase gene (TK) that in combination with gancyclovir was used for negative selection. The targeting construct was linearized by PmeI restriction digestion and electroporated into ES cells. These G418 and gancyclovir-resistant ES cell clones were screened by Southern blot after digestion of the ES genomic DNA with HindIII. The homologous recombination resulted in the replacement of a segment between exons 2 and intron 3 of the sortilin gene with the neomycin resistance cassette. Chimeric cells were injected into C57BL/6 blastocysts giving rise to chimeric mice, which were then backcrossed to C57BL/6 mice to identify if the germ-line transmission of the mutant allele had taken place (52). These heterozygotes were backcrossed with C57BL/6 mice for a minimum of seven generations before they were used in this study. All genotypes were verified by PCR using pfuTurbo (Stratagene). The number of animals used in each experiment (n) is shown in each corresponding graph.

Intraocular Injections.

A mutant form of proNGF that is resistant to cleavage by proteases (proNGFmut-m, 25 ng/μL, Alomone Labs) was injected into the vitreous chamber of the left eye using a 10-μL Hamilton syringe adapted with a 32-gauge glass microneedle (total volume: 2 μL). The average vitreous fluid volume in adult mice is estimated to be ∼10 μL (53 –55); therefore, the concentration of proNGF that reached RGCs was 0.1 μM. ProNGF was injected alone or in combination with Etanercept (Enbrel, 25 μg/μL, Wyeth), the p75^NTR antibody REX (10 μg/μL), or an Fc control (25 μg/μL, Sigma). Control eyes were injected with mature NGF (0.1 μg/μL, human recombinant NGF, PeproTech), TNFα (200 ng/mL, murine recombinant TNFα, R&D Systems), or vehicle (PBS). Intraocular injections were performed under general anesthesia (2% Isoflurane/oxygen mixture, 0.8 L/min). The needle tip was inserted into the superior hemisphere of the eye, at a 45° angle through the sclera into the vitreous body. This route of administration avoided retinal detachment or injury to eye structures, including the iris and lens, which release factors that induce neuronal survival (56, 57).

Retrograde Labeling and Quantification of Neuronal Survival.

Retrograde labeling of RGCs was performed using Fluorogold (2%, Fluorochrome) in 0.9% NaCl, which was applied to the superior colliculus as described (58). Subsequent surgical procedures were performed at 1 week after Fluorogold application (59 –62). Mice were perfused with 4% paraformaldehyde; eyes were dissected and flat mounted vitreal side up on glass sides. Fluorogold-positive RGCs were counted in 12 retinal zones: Three areas in each eye quadrant (dorsal, ventral, nasal, and temporal) located at 0.5, 1.0, and 1.5 mm from the optic nerve head were examined (58), corresponding to a total area of 0.5 mm². Statistical analyses were performed using one-way analysis of variance (ANOVA) followed by a nonparametric test (Bonferroni’s multiple-comparison test) or by a Student’s t test, as indicated in the figure legends.

Retinal Immunohistochemistry.

Mice were perfused with 4% paraformaldehyde and retinal sections were prepared as above. Tissue sections were incubated in 3% BSA and 0.3% Triton X-100 (Sigma) to block nonspecific binding and then incubated with primary antibodies (see list below) overnight at 4 °C, followed by incubation with secondary antibodies at room temperature. Slides were mounted with SlowFade (Molecular Probes) and visualized on a Zeiss Axioskop 2 Plus microscope (Carl Zeiss Canada) or a confocal microscope (Leica Microsystems). Primary antibodies used were anti-cellular retinaldehyde-binding protein (1:1,000, gift from J. C. Saari, University of Washington, Seattle, WA), anti-p75^NTR (REX, 2 ng/μL) (63), anti-NRAGE (5 ng/μL, Orbigen), and anti-TNFα (0.4 μg/mL, Chemicon). The secondary antibodies used were sheep anti-mouse IgG (1 μg/mL, FITC conjugate, Sigma) or anti-rabbit IgG (1 μg/mL, Cy3, Jackson ImmunoResearch Laboratories).

Immunoblot Analysis.

Retinas were homogenized in lysis buffer [20 mM Tris (pH 8.0), 135 mM NaCl, 1% SDS, and 10% glycerol supplemented with protease inhibitors] and centrifuged at 14,000 rpm for 5 min. The supernatants were collected, diluted in Laemmli sample buffer (4% SDS, 10% glycerol, 0.004% bromophenol blue, 0.1 M DTT, and 0.125 M Tris, pH 6.8), and analyzed by SDS–polyacrylamide gel electrophoresis and immunoblotting following standard protocols. Primary antibodies were against p75^NTR (REX, 1:1,000) (64), NRAGE (1:1,000) (64), Sortilin (1:1,000) (BD Biosciences), TNFα (0.2 μg/mL, Santa Cruz Biotechnology), and β-actin (1:30,000) (Sigma). Secondary HRP-coupled antibodies were obtained from Jackson ImmunoResearch Laboratories. Protein signals were detected using a chemiluminescence reagent (ECL, Amersham Biosciences) followed by exposure of blots to X-Omat (Kodak) imaging film.