Abstract
Loss of vision in glaucoma results from the selective death of retinal ganglion cells (RGCs). Tumor necrosis factor α (TNFα) signaling has been linked to RGC damage, however, the mechanism by which TNFα promotes neuronal death remains poorly defined. Using an in vivo rat glaucoma model, we show that TNFα is upregulated by Müller cells and microglia/macrophages soon after induction of ocular hypertension. Administration of XPro1595, a selective inhibitor of soluble TNFα, effectively protects RGC soma and axons. Using cobalt permeability assays, we further demonstrate that endogenous soluble TNFα triggers the upregulation of Ca2+-permeable AMPA receptor (CP-AMPAR) expression in RGCs of glaucomatous eyes. CP-AMPAR activation is not caused by defects in GluA2 subunit mRNA editing, but rather reflects selective downregulation of GluA2 in neurons exposed to elevated eye pressure. Intraocular administration of selective CP-AMPAR blockers promotes robust RGC survival supporting a critical role for non-NMDA glutamate receptors in neuronal death. Our study identifies glia-derived soluble TNFα as a major inducer of RGC death through activation of CP-AMPARs, thereby establishing a novel link between neuroinflammation and cell loss in glaucoma.
SIGNIFICANCE STATEMENT Tumor necrosis factor α (TNFα) has been implicated in retinal ganglion cell (RGC) death, but how TNFα exerts this effect is poorly understood. We report that ocular hypertension, a major risk factor in glaucoma, upregulates TNFα production by Müller cells and microglia. Inhibition of soluble TNFα using a dominant-negative strategy effectively promotes RGC survival. We find that TNFα stimulates the expression of calcium-permeable AMPA receptors (CP-AMPAR) in RGCs, a response that does not depend on abnormal GluA2 mRNA editing but on selective downregulation of the GluA2 subunit by these neurons. Consistent with this, CP-AMPAR blockers promote robust RGC survival supporting a critical role for non-NMDA glutamate receptors in glaucomatous damage. This study identifies a novel mechanism by which glia-derived soluble TNFα modulates neuronal death in glaucoma.
Introduction
Glaucoma is the leading cause of irreversible blindness worldwide affecting >60 million people (Tham et al., 2014). The disease is characterized by progressive optic nerve degeneration resulting in vision loss. A crucial element in the pathophysiology of all forms of glaucoma is the death of retinal ganglion cells (RGCs), the neurons that convey visual information from the retina to the brain. High intraocular pressure has been identified as a major risk factor for developing the disease, but the mechanism by which elevated pressure injures RGCs is unknown. There is no cure for glaucoma and the standard of care is to lower eye pressure by medication or surgery. Many patients, however, continue to experience visual field loss even when pressure lowering treatments are implemented (Caprioli, 1997; Georgopoulos et al., 1997). A better understanding of the mechanisms involved in glaucomatous neurodegeneration triggered by ocular hypertension injury is, therefore, essential to develop effective therapeutics.
Recent data demonstrate a neuroinflammatory component in glaucoma characterized by upregulation of proinflammatory cytokines, most notably tumor necrosis factor α (TNFα; Tezel, 2013; Soto and Howell, 2014). TNFα is a pleiotropic cytokine with physiological and pathological functions that include regulation of inflammation, innate immunity, cancer, synaptic function, and neurogenesis (Marin and Kipnis, 2013). TNFα is produced as a 22 kDa membrane-bound precursor that is cleaved by ADAM-17 (a disintegrin and metalloproteinase-17) to release the soluble 17 kDa mature TNFα (Moss et al., 1997). Both the transmembrane and soluble forms of TNFα are active and play distinct biological roles (Alexopoulou et al., 2006). Soluble TNFα binds primarily to TNFR1, and mediates apoptosis and chronic inflammation (Holtmann and Neurath, 2004; Brambilla et al., 2011; Novrup et al., 2014), whereas membrane-bound TNFα displays higher affinity for TNFR2 and regulates immunity against pathogens, resolution of inflammation, and myelination (Pasparakis et al., 1996; Arnett et al., 2001; Canault et al., 2004; Olleros et al., 2005; Alexopoulou et al., 2006; Ierna et al., 2009). Consistent with this, mice expressing only transmembrane TNFα suppress the onset and progression of autoimmune demyelination while maintaining host defenses against bacterial infection, septic shock, and pulmonary fibrosis (Mueller et al., 1999; Olleros et al., 2002; Alexopoulou et al., 2006; Oikonomou et al., 2006).
TNFα signaling is linked to RGC damage. Exogenous administration of TNFα promotes RGC loss and optic nerve degeneration (Nakazawa et al., 2006; Lebrun-Julien et al., 2009b) and RGCs lacking TNFR1 are protected from mechanical injury (Tezel et al., 2004). TNFR2 deficiency was associated with increased RGC loss during retinal ischemia (Fontaine et al., 2002), yet lack of TNFR2 has been reported to protect RGCs from ocular hypertension damage (Nakazawa et al., 2006). The differential affinity of soluble versus transmembrane TNFα for its receptors, as well as the timing of injury-related TNFα signaling (Mac Nair et al., 2014) may underlie these disparate data. Recently, high-throughput characterization of the retinal proteome revealed prominent upregulation of TNFα/TNFR1 signaling in human glaucoma (Yang et al., 2011). Notably, TNFα levels are elevated in aqueous humor samples from glaucoma patients (Sawada et al., 2010; Balaiya et al., 2011) and TNFα gene polymorphisms are associated with primary open angle glaucoma (Xin et al., 2013). Despite these correlative data, the mechanism by which TNFα promotes RGC death is poorly understood.
TNFα exerts homeostatic control of synaptic strength by regulating AMPA receptor (AMPAR) trafficking in the CNS (Pribiag and Stellwagen, 2014). Administration of exogenous TNFα to hippocampal neuron cultures induces cell surface expression of GluA2-lacking AMPARs, which are calcium (Ca2+)-permeable (Beattie et al., 2002; Ogoshi et al., 2005; Stellwagen et al., 2005; Stellwagen and Malenka, 2006). Moreover, TNFα-induced expression of Ca2+-permeable-AMPARs (CP-AMPARs) exacerbates neuronal death during acute ischemia and excitotoxicity (Ferguson et al., 2008; Leonoudakis et al., 2008; Lebrun-Julien et al., 2009b; Yin et al., 2012). In this study, we investigated whether soluble TNFα is an effector of RGC death and its mechanism of action in a rat glaucoma model. Our data reveal that ocular hypertension stimulates production of glia-derived soluble TNFα, which triggers RGC surface expression of CP-AMPAR contributing to neuronal death.
Materials and Methods
Experimental animals.
All procedures were performed in male Brown Norway rats (300–400 g; Charles River Laboratories) in compliance with the guidelines of the Canadian Council on Animal Care for the Use of Experimental Animals (http://www.ccac.ca). The number of animals used in each experiment is indicated in the corresponding figure legend and the tables.
Ocular hypertension.
Unilateral elevation of intraocular pressure was induced by injection of hypertonic saline solution into an episcleral vein as previously described (Morrison model; Morrison et al., 1997; Almasieh et al., 2010). Briefly, a plastic ring was applied to the ocular equator to confine the injection to the limbal plexus and a microneedle was used to inject 50 μl of sterile 1.85 m NaCl through an episcleral vein. Following injection, the plastic ring was removed and the eyes were examined to assess the extent to which the saline solution traversed the limbal microvasculature. Polysporin ophthalmic ointment (Pfizer) was applied to the operated eye and the animal was allowed to recover. Eye pressure measurements were taken from awake animals after corneal application of one drop of proparacaine hydrochloride (0.5%, Alcon Laboratories) using a calibrated tonometer (TonoPen XL; Medtronic Solan). The tonometer was held perpendicular to the corneal surface and 10 consecutive readings per eye were taken and averaged. Intraocular pressure was measured every other day, at the same time, for the entire duration of the experiment. The mean and peak (maximum) intraocular pressure values for each eye were calculated and used to estimate the mean and peak pressure for experimental and control groups.
Drug delivery.
The following compounds were injected into the vitreous chamber in a total volume of 5 μl using a Hamilton syringe adapted with a 32 gauge glass microneedle: XPro1595 (50 μg/μl, Xencor), GYKI 52466 (250 μm, Tocris Bioscience), or Philantotoxin 343 (PhTx 343; 250 μm, Sigma-Aldrich). Control animals received sterile PBS vehicle. The tip of the needle was inserted into the superior hemisphere of the eye at a ∼45° angle through the sclera into the vitreous body to avoid retinal detachment or injury to eye structures. Surgical glue (Indermill, Tyco Health Care) was used to seal the injection site.
Quantification of RGC soma and axons.
Before induction of ocular hypertension, RGCs were retrogradely labeled with the indocarbocyanine dye DiI (1,1′-dioctadecyl-3,3,3′,3′-tetramethyl-indocarbocyanine perchlorate; Invitrogen) prepared by dissolving crystals (3%) in 0.9% NaCl. The superior colliculus was exposed and a small piece of gelfoam (GE Healthcare and Upjohn) soaked in tracer was applied to the surface. For RGC density counts, rats were deeply anesthetized and perfused transcardially with 4% paraformaldehyde (PFA), the retinas were dissected out and flat-mounted on a glass slide with the ganglion-cell layer side up. DiI-labeled RGCs were counted in three square areas at 1, 2, and 3 mm from the optic disc in each retinal quadrant for a total of 12 retinal areas encompassing a total area of 1 mm2. Macrophages and microglia that incorporated DiI after phagocytosis of dying RGCs were excluded from our quantitative analysis (Lebrun-Julien et al., 2009a). For axon counts, animals received a transcardial injection of heparin (1000 u/kg) and sodium nitroprusside (10 mg/kg) followed by perfusion with 2% PFA and 2.5% glutaraldehyde in 0.1 m PBS. Optic nerves were dissected, fixed in 2% osmium tetroxide, and embedded in epon resin. Semithin sections (0.7 μm) were cut on a microtome (Reichert) and stained with 1% toluidine blue. Images were acquired using an oil-immersion 63× objective (numerical aperture: 1.4) with an additional microscope eyepiece magnification of 2.5× for a total visual magnification of 157.5×. RGC axons were counted at 1 mm from the optic nerve head in five nonoverlapping areas of each optic nerve section, encompassing a total area of 5500 μm2 per nerve. The five optic nerve areas analyzed included: one in the center of the nerve, two peripheral dorsal, and two peripheral ventral regions. The total area per optic nerve cross-section was measured using Northern Eclipse image analysis software (Empix Imaging), and this value was used to estimate the total number of axons per optic nerve.
Retinal immunohistochemistry.
Animals were perfused transcardially with 4% PFA and retinal cryosections (16 μm) were prepared. The following primary antibodies were added to the retinal sections in blocking solution and incubated overnight at 4°C: TNFα (0.4 μg/ml; Millipore), cellular retinaldehyde-binding protein (CRALBP; 1 μg/ml; ThermoFisher Scientific), TNFR1 (2 μg/ml, Hycult Biotech), TNFR2 (1.5 μg/ml, Antibodies On-line), RNA-binding protein with multiple splicing (RBPMS; 0.01 μg/ml, PhosphoSolutions), glial fibrillary acidic protein (GFAP; 1 μg/ml, Millipore), choline acetyltransferase (ChAT; 1 μg/ml, Millipore), or GluA2 (6C4, 2 μg/ml, Millipore). Sections were washed and incubated with secondary antibodies: anti-rabbit IgG (1 μg/ml, Cy3; Jackson ImmunoResearch) or anti-mouse IgG (1 μg/ml, FITC; Sigma-Aldrich). For whole-mount immunostaining, retinas were incubated at −80°C for 15 min, and then overnight at 4°C in blocking solution (2% normal donkey serum, 2% Triton X-100 in PBS) containing an antibody against ionized calcium-binding adaptor molecule 1 (Iba1; 1.5 μg/ml, Wako Pure Chemicals) and TNFα (4 μg/ml, R&D Systems), followed by secondary donkey anti-rabbit IgG AlexaFluor 594 and anti-mouse IgG AlexaFluor 488 (6 μg/ml, Invitrogen). Retinas were mounted on glass slides using an antifade reagent (Invitrogen) for visualization with a Zeiss AxioSkop 2 Plus microscope (Carl Zeiss).
Semiquantitative RT-PCR and Q/R site editing assay.
Total RNA was isolated from individual retinas using the RNeasy Plus Mini kit (Qiagen) according to the manufacturer's instructions. First-strand cDNA synthesis was performed using SuperScript III First-Strand Synthesis System (Life Technologies) and oligo (dT)20 primers. PCR was performed using the following primers (300 nm): TNFα forward: 5′-TCCCAACAAGGAGGAGAAGTTCCC-3′, TNFα reverse: 5′-AATGGCAAATCGGCTGACGGTG-3′, TNFR1 forward: 5′-ACCCCGGCTTCAACCCCACT-3′, TNFR1 reverse: 5′-GGGTCTGCAGTGTCAAGCCGT-3′, TNFR2 forward: 5′-GTGCAGGCCCCACCGCATTT-3′, TNFR2 reverse: 5′-TGGGCCTCCGCTGTGACTCT-3′, GluA2 forward: 5′-CTATTTCCAAGGGGCGCTGAT-3′, GluA2 reverse: 5′-CAGTCCAGGATTACACGCCG-3′, β-actin forward: 5′-CACCACTTTCTACAATGAGC-3′, and β-actin reverse: 5′-CGGTCAGGATCTTCATGAGG-3′. To assess Q/R site editing, the GluA2 open reading frame containing the Q/R site was amplified using the following: forward (5′-GAATGGTATGGTTGGAGAGC-3′) and reverse (5′-CACTT TCGATGGGAGACAC-3′) primers. The expected 527 bp amplicon was obtained using a forward nested primer (5′-GCACACTGAGGAATTTGAAG-3′) and an identical reverse primer resulting in a smaller 254 bp product, which was purified (MinElute; Qiagen) and digested with Bbvl (New England Biolabs). The reaction products were separated on agarose gels and visualized using a GelDoc imaging system (Biorad). Densitometry was performed using ImageJ software (http://imagej.nih.gov) and the purified PCR products were sequenced (McGill University and Génome Québec Innovation Centre, QC).
Subcellular fractionation and Western blot analyses.
Whole retinas were isolated and homogenized in fractionation buffer: 250 mm sucrose, 20 mm HEPES, pH 7.4, 10 mm KCl, 1.5 mm MgCl2, 1 mm EDTA, and 1 mm EGTA supplemented with protease inhibitors. Samples were centrifuged for 13 min at 8700 × g to separate the nuclear fraction. The supernatant was centrifuged for an additional 6 min and 40 s at 43,000 × g to separate the mitochondria and lysosomal fraction (pellet). The supernatant containing the total membrane and cytosolic fractions was further centrifuged for 1 h at 110,000 × g; the pellet was resuspended in fractionation buffer. Retinal samples not subjected to fractionation were homogenized with an electric pestle (Kontes) in ice-cold lysis buffer: 50 mm Tris, pH 7.4, 150 mm NaCl, 1% NP-40, 5 mm Na fluoride, 0.25% Na deoxycholate, and 2 nm NaVO3 supplemented with protease and phosphatase inhibitors. Samples were separated in 7.5–15% SDS-PAGE and transferred to nitrocellulose membranes (Bio-Rad). Blots were incubated overnight at 4°C in blocking buffer (10 mm Tris, pH 8.0, 150 mm NaCl, 0.1% Tween 20, and 5% bovine serum albumin) containing each of the following primary antibodies: TNFα (1 μg/ml R&D Systems), TNFR1 (2 μg/ml, Hycult Biotech), TNFR2 (1 μg/ml, Antibodies On-line), GluA2 (4 μg/ml; Abcam), Na+/K+ ATPase (0.1 μg/ml, Thermo Scientific), glyceraldehyde 3 phosphate dehydrogenase (GAPDH; 0.5 μg/ml, Cedarlane), or β-actin (0.5 μg/ml; Sigma-Aldrich). Membranes were washed and incubated in peroxidase-linked anti-mouse or anti-rabbit secondary antibodies (0.5 μg/ml, GE Healthcare). Blots were developed with a chemiluminescence reagent (GE Healthcare) and exposed to X-OMAT imaging film (Eastman Kodak). Densitometry was performed using ImageJ software on scanned autoradiographic films obtained from a series of three independent Western blots each performed using retinal samples from distinct experimental or control groups.
Cobalt permeability assay.
Retinas were dissected out, cut in small pieces, and incubated for 30 min in oxygenated assay buffer (5 mm KCl, 2 mm MgCl2, 12 mm glucose, 20 mm bicarbonate, 139 mm sucrose, 57.5 mm NaCl, and 0.75 mm CaCl2). The tissue was then incubated in 5 mm CoCl2 and 10 mm l-glutamic acid in the presence or absence of AMPA receptor blockers GYKI 52466 (40 μm, Tocris Bioscience) or the polyamine derivative PhTx 343 (50 μm, Sigma-Aldrich). Cobalt (Co2+) was precipitated with 0.24% ammonium sulfide, and the retina was then fixed in 0.8% glutaraldehyde and sections were prepared (20 μm). Silver enhancement of the Co2+ sulfide precipitate was performed with the GE Healthcare Intense kit, and then retinal sections were rinsed, mounted, and imaged with a Zeiss Axioplan 2 imaging microscope.
Statistical analyses.
Data analysis and statistics were performed using the GraphPad Instat software by a one-way ANOVA, followed by the Tukey's multiple-comparison post hoc test, or by a Student's t test as indicated in the legends.
Results
TNFα and its receptors are rapidly upregulated in experimental glaucoma
Unilateral ocular hypertension was induced in Brown Norway rats by a single injection of hypertonic saline into an episcleral vein as described previously (Almasieh et al., 2010, 2013). This procedure leads to blockade of aqueous humor outflow, gradual increase of eye pressure and selective loss of RGC soma and axons (Fig. 1A–C). Using this model system, we first asked whether there are variations in endogenous TNFα levels following glaucoma induction. Because early changes are more likely to play a causative role in RGC loss, our analysis focused on events at 1 and 3 weeks after hypertonic saline injection. Quantitative RT-PCR of retinal samples demonstrated a gradual and significant increase of TNFα mRNA in glaucomatous retinas compared with noninjured controls (intact; Fig. 1D,E). Similarly, Western blot analysis showed a marked increase in TNFα protein levels in retinas with high intraocular pressure (Fig. 1F,G). Immunohistochemistry confirmed that although retinal TNFα was low in control noninjured eyes, it increased substantially with ocular hypertension (Fig. 1H,I). Colabeling experiments using CRALBP, a Müller cell-specific marker, demonstrated that these glial cells upregulate TNFα soon after glaucoma induction (Fig. 1J–M). On flat-mounted retinas, analysis of the inner plexiform layer using Iba1, a pan-microglia and macrophage marker (Imai and Kohsaka, 2002), revealed reduced ramifications and increased soma size in Iba1-positive cells suggestive of an activated state during ocular hypertension (Fig. 1N,O). TNFα was not detected in highly ramified, presumably quiescent, Iba1-positive cells found in noninjured retinas, but increased markedly in cells with round somata and simplified processes (Fig. 1P–U).
We then asked whether the expression of TNFα receptors, TNFR1 and TNFR2, was altered in this glaucoma model. Our findings reveal an increase of both TNFR1 and TNFR2 mRNA (Fig. 2A,B) and protein (Fig. 2C,D) detected as early as 1 week after induction of ocular hypertension. Retinal immunohistochemistry confirmed an increase in TNFR1- and TNFR2-positive cells in glaucomatous retinas (Fig. 2E,F). In the ganglion cell layer, TNFR1 and TNFR2 were upregulated in RGCs, visualized with the RGC-specific marker RBPMS (Fig. 2G,H). In addition to RGCs, displaced amacrine cells and astrocytes populate the ganglion cell layer; therefore, we investigated whether these cells also contained TNFα receptors. Colabeling of TNFR1 with GFAP revealed that some astrocytes express TNFR1 (Fig. 2I), which is consistent with a proteomic analysis of the astrocytic response in experimental glaucoma (Tezel et al., 2012). In contrast, TNFR2 was not detected in GFAP-positive astrocytes (Fig. 2J). Similarly, amacrine cells in the GCL labeled with ChAT, which represent 20% of all displaced amacrine cells in the rodent retina (Voigt, 1986; Jeon et al., 1998), did not express TNFR1 or TNFR2 (Fig. 2K,L). Quantification of the number of RGCs labeled with both RBPMS and TNFR1 or TNFR2 with respect to all RBPMS-positive cells in retinal cross sections demonstrated that 91% and 83% of RGCs in glaucomatous retinas express TNFR1 or TNFR2, respectively (Fig. 2M). Thus, RBPMS-negative cells expressing either TNFR1 and/or TNFR2 are likely to be astrocytes or a noncholinergic subtype of displaced amacrine cells. We conclude that high intraocular pressure rapidly stimulates TNFα production by retinal glia and increases TNFR1/2 expression in RGCs.
Inhibition of soluble TNFα promotes robust RGC soma and axon protection
We next sought to test whether soluble TNFα contributes to RGC loss using XPro1595, a dominant-negative protein that selectively inhibits soluble TNFα without interfering with transmembrane TNFα or TNFR (Zalevsky et al., 2007). XPro1595 was injected intraocularly at 1 and 2 weeks after induction of ocular hypertension, and RGC survival was evaluated 1 week later (3 weeks after injury onset). Cell soma and axon quantification was performed blinded to treatment. Flat-mounted retinas from eyes treated with XPro1595 showed higher densities of DiI-positive RGCs compared with control retinas treated with vehicle (PBS; Fig. 3A–C). Quantitative analysis demonstrated that XPro1595 promoted significant RGC survival (94%: 1785 ± 27 RGCs/mm2, mean ± SEM, n = 7) with respect to eyes that received vehicle (65%: 1244 ± 55 RGCs/mm2, n = 6; Fig. 3G; Table 1).
To establish whether XPro1595 increased RGC soma survival regionally or throughout the retina, RGC densities were compared in three consecutive areas at 1 mm (central), 2 mm (middle), and 3 mm (peripheral) from the optic nerve head in each retinal quadrant. Although higher RGC densities are typically found in central areas compared with the periphery, XPro1595-mediated survival was proportional in all retinal regions accounting for 32% of all RGCs (Fig. 3E). Collectively, these data indicate that XPro1595 attenuates ocular hypertension-induced loss of RGCs globally throughout the retina.
A hallmark of glaucoma is the degeneration of RGC axons in the optic nerve posterior to the lamina cribrosa; therefore, we also investigated the capacity of XPro1595 to protect axons. Analysis of optic nerve cross-sections showed a substantially larger number of RGC axon fibers with normal morphology in XPro1595-treated eyes compared with vehicle-treated controls (Fig. 3D–F). The latter displayed extensive axon degeneration including disarray of fascicular organization and degradation of myelin sheaths. Quantitative analysis confirmed that XPro1595 promoted marked protection of RGC axons (80%: 79,917 ± 1574 axons/nerve, mean ± SEM, n = 6) compared with vehicle treatment (45%: 45,345 ± 7583 axons/nerve, n = 6; Fig. 3H; Table 1).
To assess whether XPro1595 reduced intraocular pressure, which could account for its neuroprotective effect, we measured eye pressure every other day after glaucoma induction. Our results show that XPro1595 administration did not reduce eye pressure and that the pressure elevations among XPro1595-treated and vehicle-treated groups were similar (Table 2). Given that the rate of RGC death is proportional to ocular hypertension, the similar increase in intraocular pressure among groups allowed for reliable comparison of the neuroprotective effect of XPro1595 versus vehicle. Our results demonstrate that XPro1595 protects both RGC soma and axons, without altering eye pressure, suggesting a prominent role for soluble TNFα in glaucoma downstream ocular hypertension.
Ocular hypertension-induced TNFα triggers CP-AMPAR activity in RGCs
To investigate whether soluble TNFα modulates CP-AMPAR in glaucoma, we used an in situ cobalt (Co2+) permeability assay (Aurousseau et al., 2012). Co2+ is transported into neurons expressing CP-AMPAR in the plasma membrane, but it does not permeate through other Ca2+ channels or NMDA receptors (Hagiwara and Byerly, 1981; Mayer and Westbrook, 1987). Following Co2+ precipitation, the appearance of a dark brown color is visualized only in neurons containing CP-AMPAR. In noninjured retinas, Co2+ accumulation was detected solely in cells of the inner nuclear layer (INL; Fig. 4A), previously identified as horizontal and AII amacrine cells (Osswald et al., 2007). Remarkably, ocular hypertension induced massive Co2+ accumulation in the ganglion cell layer (GCL), where RGC soma are located, and in the inner plexiform layer (IPL), where dendrites are present (Fig. 4B). Displaced amacrine cells account for ∼40–50% of the total number of neurons in the GCL (Perry, 1981; Schlamp et al., 2013) hence we performed colabeling with the RGC-specific marker RBPMS (Kwong et al., 2010; Rodriguez et al., 2013) to selectively label RGCs in intact and glaucomatous retinas (Fig. 4C,D). Colocalization of Co2+ precipitate with RBPMS confirmed that Co2+ accumulated in RGCs (Fig. 4E,F).
To test whether soluble TNFα was required for increased CP-AMPAR levels leading to Co2+ accumulation, XPro1595 was injected intraocularly 2 weeks after glaucoma induction and Co2+ staining was performed 1 week later. Figure 4G shows that XPro1595 strongly inhibited Co2+ uptake by RGCs. Treatment with GYKI 52466, a noncompetitive AMPAR antagonist (Mellor, 2010), or Philantotoxin 343 (PhTx 343), a polyamine-derived compound that selectively antagonizes CP-AMPAR (Strømgaard et al., 2005), effectively blocked Co2+ uptake (Fig. 4H,I). These data indicate that Co2+ accumulation in RGCs was due to selective flux through CP-AMPAR and rule out the contribution of other glutamate receptors. We conclude that, in glaucoma, soluble TNFα triggers the CP-AMPAR activity detected in RGCs.
Retinal AMPAR GluA2 subunit is fully edited at the Q/R site in glaucoma
The Ca2+ permeability of AMPAR varies depending on whether the GluA2 subunit is present and, if so, whether it has undergone mRNA editing. GluA2 mRNA editing switches an uncharged amino acid glutamine (Q) to a positively charged arginine (R), and as a result, Ca2+ cannot be transported due to electrostatic repulsion by the arginine residues lining the AMPAR pore (Burnashev et al., 1992). We then asked whether GluA2 mRNA transcripts were edited as a mechanism accounting for AMPAR Ca2+ permeability in glaucoma. To this end, a 254 bp amplicon of rat GluA2 incorporating the Q/R site was amplified from cDNA generated by RT-PCR using retinal mRNA. Taking advantage of the presence of an additional Bbvl restriction site unique to the unedited GluA2 (Q) sequence, digestion of the PCR product resulted in a distinct band pattern consistent with GluA2 control plasmid coding for arginine at the Q/R site in both intact and glaucomatous retinas (Fig. 5A,B). In addition, parallel sequencing of the amplicon confirmed that GluA2 transcripts from both control and glaucoma samples encoded an arginine residue at the Q/R site (Fig. 5C–F). Our data demonstrate that the GluA2 subunit is fully edited in glaucoma, and strongly suggest that GluA2 editing is not the cause for increased CP-AMPAR expression.
GluA2 expression in RGCs is downregulated by ocular hypertension
Reduced expression or absence of the GluA2 subunit results in AMPARs that are permeable to Ca2+ (Dingledine et al., 1999). To determine whether TNFα-mediated loss of GluA2 could account for increased CP-AMPAR activity in RGCs, we examined GluA2 levels in glaucomatous retinas. Our results show that ocular hypertension triggers a gradual and substantial decrease of both retinal GluA2 mRNA and protein (Fig. 6A–D). Subcellular fractionation of membrane and cytosolic fractions confirmed a significant reduction of membrane-associated GluA2 in hypertensive retinas (Fig. 6E,F). The enrichment of each fraction for membrane or cytosolic proteins was confirmed using Na+/K+ ATPase or GAPDH, respectively (Fig. 6E). To establish whether GluA2 was downregulated in RGCs, retinal immunofluorescence was performed. A striking reduction of GluA2 labeling was observed in the GCL of retinas subjected to ocular hypertension (Fig. 6G,H). Costaining of GluA2 with the RGC-specific marker RBPMS confirmed that loss of GluA2 labeling reflected protein downregulation and not death of RGCs (Fig. 6G,H). We conclude that RGC levels of GluA2 are markedly reduced in glaucoma.
CP-AMPAR blockade promotes RGC soma and axon survival
To determine whether TNFα mediates RGC death by modulating plasma membrane levels of CP-AMPAR, we examined the neuroprotective effect of AMPAR channel blockers. Our data show that a single injection of GYKI 52466 or PhTx 343 promoted RGC soma and axon protection at 3 weeks after ocular hypertension compared with vehicle-treated retinas (Fig. 7A–F). Quantitative analysis was performed blinded to treatment and confirmed that GYKI 52466 and PhTx 343 resulted in striking RGC soma survival (88% and 91%, respectively) compared with vehicle (65%; Fig. 7G). Similarly, quantification of RGC axons in optic nerve cross sections revealed robust axonal protection by GYKI 52466 (70%) and PhTx 343 (73%) with respect to vehicle-treated controls (46%; Fig. 7H). Consistent with the idea that the primary site of degeneration in glaucoma is at the level of the axon, we observed more pronounced axon loss than soma loss. Nonetheless, both GYKI 52466 and PhTx protected a similar proportion of RGC soma and axons. Our data demonstrate a critical role for CP-AMPAR in RGC death in glaucoma.
Discussion
Despite data linking TNFα with glaucomatous neurodegeneration, little is known about the form of TNFα that promotes RGC death and its underlying mechanism of action. In this regard, the data presented here, using a well characterized rat glaucoma model reveal several novel findings. First, we show that ocular hypertension rapidly stimulates production of TNFα by Müller cells and microglia/macrophages, and increases TNFR1 and TNFR2 expression in RGCs. Second, we demonstrate that XPro1595, an inhibitor of soluble TNFα, effectively protects RGC soma and axons. Third, we show that TNFα stimulates the expression of CP-AMPAR in RGCs, a response that does not depend on defects in GluA2 mRNA editing but reflects selective downregulation of GluA2 by these neurons. Last, our results demonstrate that CP-AMPAR blockers promote robust RGC survival supporting a critical role for non-NMDA glutamate receptors in glaucomatous damage. Collectively, this study identifies a novel mechanism by which glia-derived soluble TNFα modulates neuronal CP-AMPA function leading to RGC death, and establishes a critical link between neuroinflammation and cell damage in glaucoma.
Elevated intraocular pressure (IOP) is a critical risk factor for developing glaucoma and it is the cornerstone of research into glaucoma pathophysiology. Here, ocular hypertension was induced by injection of hypertonic saline into episcleral veins blocking aqueous humor outflow at the iridocorneal angle and trabecular meshwork, the sites of drainage obstruction in most forms of human glaucoma. In this model, there is an excellent linear correlation between intraocular pressure and neuronal damage (Morrison et al., 1997, 2005, 2015). Inner retinal atrophy, optic nerve degeneration, and optic nerve head remodeling in this model are similar to those observed in human glaucoma (Morrison et al., 2015), thus it is an excellent in vivo paradigm of glaucomatous damage. Limitations of this model are the challenging microsurgical skill to successfully cannulate the small episcleral veins, the fluctuation in IOP responses often observed in rodent models of aqueous outflow obstruction, and the risk of anterior chamber deepening and corneal opacity in some animals (Morrison et al., 2015). Another consideration is the anatomical difference between the rat and the human optic nerve head including the composition of the lamina cribrosa (Morrison et al., 1995). Nonetheless, the cost-effective use of rats and their docile nature, allowing measurement of IOP in awake animals with only topical anesthesia, enhance the attractive features of this useful glaucoma model system.
During glaucoma, chronically reactive glial cells are thought to become a sustained source of proinflammatory cytokines (Seitz et al., 2013). Our data demonstrate that Iba1-positive microglia/macrophages are a source of TNFα in ocular hypertensive eyes. TNFα expression was observed primarily in microglia/macrophages with amoeboid shape, characteristic of a reactive state, rather than in resting cells with ramified morphology. This finding is consistent with previous reports showing TNFα expression in microglia from human glaucomatous optic nerve head (Yuan and Neufeld, 2001) and rat retinas subjected to episcleral vein cauterization (Roh et al., 2012). Of interest, high-dose irradiation leading to reduced microglial activation, and presumably decreased levels of proinflammatory mediators, attenuated RGC degeneration in a mouse model of inherited pigmentary glaucoma (Bosco et al., 2012; Howell et al., 2012). Moreover, our results demonstrate that Müller cells, the most abundant glial cell type in the retina, upregulate TNFα in response to increased eye pressure. Excitotoxic damage has also been shown to rapidly increase NFκB (nuclear factor κ B) activation and TNFα production by Müller cells (Lebrun-Julien et al., 2009b). It is currently unknown how ocular hypertension stimulates TNFα output by reactive glia but it might involve mechanisms dependent on ischemia/hypoxia. For example, cultured microglia subjected to hypoxia release more TNFα and interleukin 1β than cells exposed to normal oxygen levels (Sivakumar et al., 2011). Accumulation of toxic factors may also increase glia reactivity and subsequent neuroinflammation. Indeed, we have previously shown that the precursor form of nerve growth factor activates neurotrophin receptor p75NTR-dependent signaling mechanisms on Müller cells, leading to increased production of TNFα and subsequent RGC death (Lebrun-Julien et al., 2010).
The two active forms of TNFα, soluble and transmembrane, play distinct biological roles. We used XPro1595 to block soluble TNFα and to elucidate its role in glaucomatous neurodegeneration. XPro1595 is an engineered protein that selectively binds with soluble TNFα monomers to form inactive heterotrimers unable to interact with TNFα receptors. Therefore, XPro1595 eliminates signaling from soluble TNFα without interfering with the transmembrane form (Steed et al., 2003; Zalevsky et al., 2007). Our data demonstrate that XPro1595 effectively promotes RGC survival, without altering intraocular pressure. Given that both TNFR1 and TNFR2 are upregulated by RGCs during ocular hypertension, we speculate that blockade of soluble TNFα minimizes the detrimental effect of TNFR1 activation while preserving beneficial TNFR2-mediated signaling. Recent studies also reported a neuroprotective effect of XPro1595 in models of experimental autoimmune encephalomyelitis, spinal cord injury, Parkinson's disease, and Huntington's disease, confirming a harmful role for soluble TNFα in neurodegenerative conditions (Brambilla et al., 2011; Barnum et al., 2014; Hsiao et al., 2014; Novrup et al., 2014). Etanercept, a drug that blocks both soluble and transmembrane TNFα, has been shown to protect RGCs in a rat glaucoma model (Roh et al., 2012). However, nonselective TNFα inhibitors, including etanercept, infliximab, and adalimumab, have been linked to serious adverse effects such as impaired host defense, autoimmunity, lupus, demyelination syndromes, and congestive heart failure (van Oosten et al., 1996; Keane et al., 2001; Lee et al., 2002; Shakoor et al., 2002; Slifman et al., 2003; Sfikakis, 2010). Collectively, these findings highlight the benefits of inhibiting soluble TNFα while preserving transmembrane TNFα function during neurodegeneration.
In physiological conditions, TNFα is an important modulator of synaptic plasticity by its ability to regulate AMPA receptor trafficking. In hippocampal pyramidal neurons, TNFα strengthens synapses by rapid exocytosis of AMPAR that lack or have low stoichiometric amounts of the GluA2 subunit thus enhancing intracellular Ca2+ rises (Ogoshi et al., 2005; Stellwagen et al., 2005). We found low levels of endogenous TNFα in the uninjured adult rat retina that correlated with sparse expression of CP-AMPAR. Using a Co2+ permeability assay applied to adult intact retinas, we confirmed that CP-AMPAR expression was restricted to inhibitory interneurons, horizontal and AII amacrine cells, as shown previously (Mørkve et al., 2002; Singer and Diamond, 2003; Veruki et al., 2003; Osswald et al., 2007). Although CP-AMPAR have been reported in acutely purified neonatal RGCs and can contribute to excitotoxicity (Park et al., 2015), CP-AMPAR have not been previously detected in adult RGCs. Remarkably, we show that ocular hypertension triggers robust CP-AMPAR upregulation in these neurons. Our finding that XPro1595 attenuates Co2+ accumulation in RGCs indicates that soluble TNFα contributes to the increase of CP-AMPAR. Our data suggest that while basal levels of TNFα are required for retinal homeostasis and neurotransmission, excess soluble TNFα results in CP-AMPAR upregulation and cell death in glaucoma.
Consistent with the idea that the primary site of degeneration in glaucoma is at the level of RGC axons, we found that all glaucomatous eyes had more pronounced axon loss than cell body loss. Interestingly, however, intravitreal injection of XPro1595 or CP-AMPAR blockers effectively protected a similar proportion of RGC soma and axons suggesting a dynamic crosstalk between these compartments. The precise mechanism that renders axons more vulnerable than soma in glaucoma is still unclear, but based on our data it is tempting to speculate that a sequence of events in both compartments contributes to higher axonal susceptibility. First, ocular hypertension-induced stress at the optic nerve head inflicts an initial injury to the axon, which although still undefined in nature, might involve energy or nutrient supply deficits and activation of local degenerative pathways (Calkins, 2012; Nickells et al., 2012). Our study suggests that a rapid consequence of axonal damage is upregulation of TNFα receptors at RGC soma, which become responsive to glia-derived soluble TNFα, and in turn, upregulate cell surface CP-AMPAR. Increased Ca2+ entry to RGC soma via CP-AMPAR is likely to stimulate signaling cascades that exacerbate axonal degeneration. Indeed, excessive cytosolic Ca2+ activates Ca2+-dependent calpains and caspases that degrade components of the RGC axonal cytoskeleton slowing or blocking axonal transport (Crish and Calkins, 2011). Ca2+ overload may also lead to oxidative stress impairing mitochondrial function which, in addition to decreasing the capacity of mitochondria to buffer Ca2+, might aggravate axonal damage by disabling Na+/K+ ion pumps causing electrical failure of RGC axons (Tsutsui and Stys, 2013). Of interest, functional AMPA receptors are found at the internodal axolemma and in oligodendrocyte somata (McDonald et al., 1998; Ouardouz et al., 2009), therefore it is plausible that TNFα produced locally in the optic nerve might upregulate CP-AMPAR in RGC axons leading to loss of ionic homeostasis, lethal Ca2+ overload and axonal loss.
How do AMPAR become permeable to divalent ions in glaucoma? One possibility is defective GluA2 mRNA editing. Typically, the change from an uncharged amino acid glutamine (Q) to a positively charged arginine (R) in GluA2 is sufficient to confer Ca2+ impermeability (Burnashev et al., 1992). Abnormal mRNA processing failing to complete this process will result in a Ca2+-permeable AMPAR pore. Our findings show that retinal GluA2 is fully edited in glaucoma, ruling out a post-transcriptional editing defect as a mechanism by which AMPARs become divalent permeable. A second possibility is a decrease in GluA2 leading to AMPARs that are permeable to divalent ions. Using biochemical and immunohistochemical analyses, we demonstrate that GluA2 expression is substantially downregulated in RGCs subjected to ocular hypertension. GluA2-lacking CP-AMPAR are traditionally thought of as Ca2+ permeable channels that are blocked by exogenous and endogenous polyamines, while GluA2-containing receptors are impermeable to Ca2+ and lack polyamine sensitivity (Bowie et al., 1999; Dingledine et al., 1999). Of interest, a third class of AMPAR characterized by divalent permeability and weak sensitivity to polyamine block has been proposed to exist in some CNS regions including the retina (Bowie, 2012). Indeed, although divalent permeability through AMPAR in horizontal and AII amacrine cells was effectively blocked with the polyamine derivative PhTX 343 before eye opening, these AMPAR became insensitive to PhTX 343 by postnatal day 14 (Osswald et al., 2007). Our data show that PhTX 343 blocks divalent permeability through CP-AMPAR, and in doing so, promotes robust RGC survival in glaucomatous eyes. These observations suggest that adult retinal neurons exposed to ocular hypertension revert to an immature phenotype characterized by CP-AMPAR that are permeable to divalent cations and sensitive to polyamine block. Our findings also reveal the potential to use of polyamine-derived compounds as neuroprotective agents for glaucoma.
The molecular mechanisms by which TNFα regulates CP-AMPAR are not well understood. Early work using hippocampal neuron cultures indicated that increased cell surface AMPAR expression mediated by TNFα, insulin, or glycine requires phosphatidylinositol 3 kinase activity (Passafaro et al., 2001; Man et al., 2003; Stellwagen et al., 2005). Our data suggest that TNFα-mediated downregulation or loss of GluA2 might be a potential regulatory mechanism to increase divalent ion influx through AMPAR leading to RGC death in glaucoma. Although the process by which TNFα might regulate GluA2 levels is currently unknown, it is established that neuronal GluA2 expression is not stable, but rather closely regulated by activity, drugs, seizures, or ischemic damage often accompanied by proinflammatory cytokine release (Pollard et al., 1993; Fitzgerald et al., 1995; Prince et al., 1995; Gorter et al., 1997; Liu and Cull-Candy, 2000). During forebrain ischemia, GluA2 expression is selectively downregulated in CA1 hippocampal neurons destined to die, a process that involves suppression of GluA2 promoter activity by repressor element-1 silencing transcription factor (Pellegrini-Giampietro et al., 1992; Calderone et al., 2003). A recent study demonstrated that the interaction of GluA2 and N-ethylmaleimide-sensitive fusion protein, an ATPase involved in stabilization of GluA2 at the membrane, is disrupted by Polo-like kinase-2 leading to extensive loss of cell-surface GluA2 in primary hippocampal neurons (Evers et al., 2010). It will be of future interest to assess whether these processes are modulated by TNFα.
In conclusion, using multiple complementary techniques we demonstrate a crucial role of glia-derived soluble TNFα in glaucomatous neurodegeneration. Importantly, our data identify a novel and important mechanism of soluble TNFα-induced damage involving GluA2 downregulation and increased CP-AMPAR leading to RGC death. These findings expand our understanding of the molecular basis of RGC damage during neuroinflammation and might have implications in the design of neuroprotective strategies for glaucoma.
Footnotes
This work was supported by a team grant from the Canadian Institutes of Health Research (CIHR-VIH-105439); A.D.P. is a Chercheur National of Fonds de recherche Québec-Santé (FRQS). We thank Dr David Szymkowski (Xencor) for providing XPro1595, and Drs Ray Tesi (FPRT Bio) and Timothy Kennedy (McGill University) for helpful comments on the paper.
The authors declare no competing financial interests.
- Correspondence should be addressed to Dr. Adriana Di Polo, Department of Neuroscience, University of Montreal, CRCHUM, 900 Rue Saint-Denis, Tour Viger, Room R09.720, Montreal, QC H2X 0A9, Canada. adriana.di.polo{at}umontreal.ca