Bax Contributes to Retinal Ganglion Cell Dendritic Degeneration During Glaucoma

The BCL-2 (B-cell lymphoma-2) family of proteins contributes to mitochondrial-based apoptosis in models of neurodegeneration, including glaucomatous optic neuropathy (glaucoma), which degrades the retinal ganglion cell (RGC) axonal projection to the visual brain. Glaucoma is commonly associated with increased sensitivity to intraocular pressure (IOP) and involves a proximal program that leads to RGC dendritic pruning and a distal program that underlies axonopathy in the optic projection. While genetic deletion of the Bcl2-associated X protein (Bax-/-) prolongs RGC body survival in models of glaucoma and optic nerve trauma, axonopathy persists, thus raising the question of whether dendrites and the RGC light response are protected. Here, we used an inducible model of glaucoma in Bax-/- mice to determine if Bax contributes to RGC dendritic degeneration. We performed whole-cell recordings and dye filling in RGCs signaling light onset (αON-Sustained) and offset (αOFF-Sustained). We recovered RGC dendritic morphologies by confocal microscopy and analyzed dendritic arbor complexity and size. Additionally, we assessed RGC axon function by measuring anterograde axon transport of cholera toxin subunit B to the superior colliculus and behavioral spatial frequency threshold (i.e., spatial acuity). We found 1 month of IOP elevation did not cause significant RGC death in either WT or Bax-/- retinas. However, IOP elevation reduced dendritic arbor complexity of WT αON-Sustained and αOFF-Sustained RGCs. In the absence of Bax, αON- and αOFF-Sustained RGC dendritic arbors remained intact following IOP elevation. In addition to dendrites, neuroprotection by Bax-/- generalized to αON-and αOFF-Sustained RGC light- and current-evoked responses. Both anterograde axon transport and spatial acuity declined during IOP elevation in WT and Bax-/- mice. Collectively, our results indicate Bax contributes to RGC dendritic degeneration and distinguishes the proximal and distal neurodegenerative programs involved during the progression of glaucoma.


Background
The BCL-2 (B-cell lymphoma-2) family of proteins contributes broadly to intrinsic, mitochondrial-based, apoptosis during development, cancer, and neurodegeneration [1]. During neurodegeneration, chronic cellular stress initiates apoptosis through the activation of Bcl-2-associated X (BAX) protein [1]. Upon activation, BAX proteins translocate from the cytosol to the mitochondrial outer membrane [2]. Within the mitochondrial outer membrane, BAX form heterodimers with BCL-2-antagonist/killer (BAK) or homodimers with other BAX proteins [3,4]. Dimerization facilitates the formation of BAX oligomers [5] that permeabilize the mitochondrial outer membrane, causing the release of cytochrome c and second mitochondrial-derived activator of caspases (SMAC) into the cytosol, promoting cell death [6,7].
Genetic and inducible models of glaucoma combined with genetic dosing of Bax and Wld S elegantly demonstrate compartmentalized degeneration [11,12,18,22]. Here, we sought to investigate the influence of Bax -/on RGC dendrites and signaling prior to significant cell death.
Using an inducible microbead model of glaucoma [24], we found Bax contributes to dendritic pruning and degradation of evoked responses of alpha RGCs signaling light increments (αON-Sustained) and decrements (αOFF-Sustained). IOP elevation decreased anterograde axonal transport and spatial acuity during glaucoma in both WT and Bax -/animals. Our results indicate Bax contributes to the reduction in RGC dendritic complexity and signaling during IOP elevation and illustrates the separability of the proximal and distal programs involved in degeneration during glaucoma.

Animals and Genotype Confirmation
All experimental procedures were approved by the Vanderbilt University Institutional Animal Care and Use Committee. We obtained 6-to 8-week-old male C57Bl/6 (WT) mice from Charles River Laboratory (Wilmington, MA). We obtained 4-to 10-week-old male mice with a targeted null mutation in the Bax gene (Bax -/-, 002994 -B6.129X1-Bax tm1Sjk ) from Jackson Laboratory (Bar Harbor, ME, [25]), We confirmed Bax -/by PCR according to the vendor's protocol, using the following primers: GTT GAC CAG AGT GGC GTA GG (Common), CCG CTT CCA TTG CTC AGC GG (Mutant forward), and GAG CTG ATC AGA ACC ATC ATG (WT forward). We obtained primers from Integrated DNA Technologies (Coralville, IA). Mice were maintained on a 12-h light-dark cycle with standard rodent chow and water available as desired.
During whole-cell current-clamp recordings, we targeted alpha-type RGCs. For 70% of cells analyzed (256 out of 364), we determined the distance from the optic nerve head. On average, cells were in the mid peripheral retina. We tested RGC responses to light and current injections (0 to 180 pA in 20 pA increments). Light responses were evoked by a light-emitting diode system (pE-4000, CoolLED, Andover, UK), and light was focused onto the retina through the microscope objective (40×, full field, 365 nm, 3 s). After physiology, retinas were fixed overnight in 4% paraformaldehyde (PFA) at −4°C. The following day, the PFA solution was removed and substituted with PBS, and we prepared retinas for immunohistochemistry.

Retinal Immunohistochemistry, Imaging, and Dendritic Morphologic Analysis
We performed immunohistochemistry on whole-mount retinas and sections from superior colliculus (SC) as previously described [26,27,[29][30][31]. Retinas were immunolabeled with the following primary antibodies: mouse-non-phosphorylated neurofilament H (SMI-32,1:1000; 801701, BioLegend, San Diego, CA) and goat-choline acetyltransferase (ChAT, AB144P,1:100, Millipore, Burlington, MA). For a subset of experiments, we determined the density of RGCs in WT and Bax -/whole mount retinas by immunolabeling against mouse-SMI-32 (1:1000; 801701, BioLegend, San Diego, CA) and goat-Brn3a (1:200, Santa Cruz Biotechnology, Dallas, TX). For SMI-32-positive and Brn3a-positive cell counts, confocal micrographs were taken with an Olympus FV1000 using a 60× objective. Four images were taken within each quadrant (superior, inferior, nasal, temporal) of the retina. Cell counts were performed manually by a masked investigator and then averaged. SMI-32 intensity was determined by creating ROIs around the soma of each SMI-32-positive cell and normalizing to background. We determined neuronal density in SC sections by immunolabeling against rabbit-NeuN (1:500,12943, Cell Signaling Technology, Danvers, MA). Z-stack images were obtained using a 40× objective with a 2× zoom. Max intensity projections were generated for each z-stack and NeuN-positive cell counts were performed manually by a masked investigator.
We imaged tissues using confocal microscopy (Olympus FV-1000). RGC dendritic arbors were montaged and manually traced using Adobe Illustrator and Adobe Photoshop, respectively. Skeletonized dendritic arbors were analyzed using Fiji (ImageJ, version 1.53c). We determined the dendritic field area using the polygon selection tool in Fiji. We determined dendritic branch points by manually counting the number of dendritic bifurcations [22,26]. Sholl analysis was performed using the Fiji plugin. For dendritic depth, we measured the distance from the start of the IPL to the inner most dendritic fluorescence signal. Six distance measurements were taken for each cell with regard to total IPL and dendritic depth, and then averaged. Dendritic depth was presented as a percentage of the relative total IPL depth for each cell.

Anterograde Axonal Transport Analysis
Mice were anesthetized with isoflurane (2.5%), and we injected 1 μL of cholera toxin subunit B 488 (CTB-488, Molecular Probes, Eugene, OR) into the anterior chamber of both eyes [26,27,29,32]. We allowed 2 days for transport of CTB-488 to the SC. After this time, we perfused mice transcardially with PBS followed by 4% PFA. After perfusion, we removed brains and cryoprotected them in sucrose (30%). We then obtained coronal midbrain sections (50 μm) using a freezing sliding microtome. After sectioning, alternating SC sections were mounted and imaged using a Nikon Ti Eclipse microscope (Nikon Instruments Inc., Melville, NY). We quantified the area of CTB-488 signal (intact transport) using a custom-written routine in ImagePro (Media Cybernetics, Bethesda, MD, [31]). Heat map surface plots show the fluorescence area relative to the total area of the SC. We confirmed successful CTB-488 uptake by RGCs in wholemount retinas by confocal microscopy.

Optic Nerve Transmission Electron Microscopy and Optic Nerve Axon Quantification
One month after IOP elevation, we perfused WT and Bax -/mice with 0.1 M cacodylate buffer followed by 2.5% glutaraldehyde in cacodylate buffer. We carefully removed eyes with the optic nerve attached. We then isolated 3-mm segments of optic nerve proximal to the globe and post-fixed them for 1 h in 2.5% glutaraldehyde in cacodylate buffer. Next, we embedded the optic nerve segments in Epon resin and obtained semi-thin (700 nm, light microscopy) and ultra-thin (70 nm, electron microscopy) cross-sections. We stained semi-thin cross-sections with 1% paraphenylenediamine (PPD; in a 1:1 mixture of methanol and 2-propanol) and 1% toluidine blue to identify myelin sheaths and glia, respectively. We imaged sections en montage using a Nikon H600L microscope equipped with a 100× oil-immersion objective, motorized X-Y-Z stage, a digital SLR camera, and differential interference contrast optics. Ultra-thin nerve cross-sections were prepared and photographed at 2700× (10 images per nerve for axon counts) and 240× magnification (one image per nerve to measure cross-sectional nerve area) using a Philips CM-12, 120-keV transmission electron microscope at Vanderbilt Cell Imaging Shared Resource Core. A naïve observer manually counted total and degenerating axons using Fiji ImageJ. We identified degenerating axons based on multilaminar myelin sheaths and diminishment of cytoskeletal content.

Behavioral Spatial Frequency Threshold Measurement
We measured spatial frequency threshold (i.e., spatial acuity) using the OptoMotry system (Cerebral Mechanics Inc., Canada) as previously described [27,29]. We determined spatial acuity by assessing the oculomotor reflex to drifting spatial frequency gratings at 100% contrast. Each spatial frequency was presented until a response (tracking) or no response (no tracking) was indicated by naïve experimenters. Mice were tested 3 times prior to microbead injection (baseline) and 2 times per week for 1 month after injections.

Statistical Analysis
We analyzed data using GraphPad (Version 9, GraphPad Software Inc, San Diego, CA) or SigmaPlot (Version 12, Systat Software Inc., San Jose, CA). We determined if data fit a normal distribution using Shapiro-Wilk tests. If data passed normality tests, we performed parametric statistics; otherwise, non-parametric statistics were performed. We identify each statistical test performed in figure captions in the "Results" section.

Bax -/-Abates RGC Dendritic Loss During Glaucoma
We verified the genotype of each Bax animal by gel electrophoresis of PCR products. Mice homozygous for the mutant Bax tm1Sjk allele (Bax -/-) showed a single band at 507 bp. Heterozygous mice for the mutant Bax tm1Sjk allele (Bax +/-) produced bands at 507 bp and 307, indicating the WT Bax allele. WT animals produced only a single band at 307 bp (Fig. 1A).
Although we find elevating IOP for 1 month does not cause significant RGC dropout, our result does not preclude the pro-degenerative impact of IOP elevation on RGC morphology and physiology. Our laboratory and others have found that RGC dendrites and voltage-gated responses are particularly sensitive to IOP elevation [22,[34][35][36]. Next, we determined if Bax contributes to the phenotypic dendritic pruning and response degradation observed in WT RGCs during IOP elevation.
After 1 month of IOP elevation, we found a significant reduction in the number of dendritic intersections (Sholl analysis, p≤0.04, Fig. 2F) and a significant, 32%, decrease in branch points (48 ± 2 vs. 32.7 ± 1.4 branch points, p<0.001, Fig. 2G) in WT αON-S RGCs. We confirmed this reduction in dendritic branching was not due to sampling bias across retinal eccentricities by measuring the distance from the optic nerve head and comparing dendritic branching versus eccentricity. Typically, we recovered WT αON-S RGCs from the mid peripheral retina (WT Ctrl: 1485±59 μm, WT 4Wk: 1443±92 μm from the optic nerve head, p=0.98, Fig. S1A). Similar to an earlier report [37], we did not detect a significant linear relationship between branch points and retinal eccentricity for WT αON-S from control (R 2 =0.002, p=0.71) or microbead-injected eyes (R 2 =0.07, p=0.21, Fig. S1B).
Bax -/increases retinal area [38]. Consequently, during whole-cell recording and dye-filling, we often targeted RGCs that were more distant from the optic nerve head than WT RGCs and even beyond the extent of WT retinas (p=0.005, Fig. S1A). Following 1 month of IOP elevation, Bax -/-αON-S RGC dendritic arbors remained intact based on the number of dendritic crossings (Sholl analysis, p≥0.11, Fig. 2F), the number of dendritic branch points (p>0.99, Fig. 2G), and dendritic field area (p>0.99, Fig. 2H). Similar to fellow WT cells, we did not detect a significant relationship between branch points and eccentricity of αON-S RGCs sampled from Bax -/control (R 2 =0.002, p=0.74) and microbead-injected eyes (R 2 =0.04, p=0.39, Fig. S1C). Overall, our findings indicate Bax -/attenuated the loss of αON-S RGC dendritic complexity during IOP elevation.
In addition to morphological features, we also identified αOFF-S RGCs by their physiologic responses [26,27,35,36]. αOFF-S RGCs from both WT and Bax -/mice produced spontaneous action potentials during darkness, which were suppressed during light onset, and generated a sustained ◂ burst of action potentials after light offset (Fig. 5A) Fig. 5B, C). Bax -/did not significantly affect RMP of αOFF-S RGCs from control eyes (p>0.99, Fig. 5D). Contrary to light responses, when we stimulated WT and Bax -/control αOFF-S RGCs with depolarizing current injections, we found mean spike rate (p≥0.17, Fig. 5E, F) and rheobase (p=0.12, Fig. 5G) to be similar.
Next, we tested the impact of IOP elevation on αOFF-S RGC responses to a series of depolarizing current injections. Like the reduction in light responses, 1 month of IOP elevation decreased spike rate at suprathreshold test currents in WT αOFF-S RGCs (p≥0.03, Fig. 3E), and significantly reduced the average spike rate to depolarizing currents (p=0.001, Fig. 5F). However, IOP elevation does not significantly affect Bax -/-αOFF-S RGC responses to depolarizing currents (p≥0.11, Fig. 5E, F) or rheobase (p>0.99, Fig. 5G). Taken together, our results show IOP elevation significantly alters WT RGC morphology and physiology regardless of cell type, and this pathology is largely undetectable in the absence of Bax.

Axonopathy Endures in Bax -/-Mice
Degradation of RGC anterograde axon transport to the superior colliculus (SC) is an early indicator of axonopathy in glaucoma [31,39]. Anterograde axon transport deficits occur prior to optic nerve axon degeneration and loss of postsynaptic target neurons of the SC [31,39]. Based on this premise, we investigated anterograde axon transport of cholera toxin subunit B (CTB-488) to the SC. Following 1 month of IOP elevation, we observed a reduction in intact anterograde transport of CTB-488 to the SC in both WT and Bax -/mice (Fig. 6A). When quantified, we found IOP elevation significantly diminished the percent of intact transport in WT (−35%, p=0.0002) and Bax -/mice (−21%, p=0.03, Fig. 6B). We did not observe a statistical difference in the percentage of intact transport in the SC of WT and Bax -/mice subjected to IOP elevation (p=0.95, Fig. 6B).
The IOP-induced loss of anterograde axon transport to the SC does not appear to be due to outright degeneration of RGC axons in the optic nerve proper or SC neurons. As noted by others, and confirmed here, Bax -/intrinsically increased optic nerve size and axon density (Fig. S2A, B). When quantified, we found Bax -/significantly increased optic nerve cross-sectional area by 59% (60420 ± 4975 vs. 96308 ± 11607 μm 2 , p=0.009, Fig. S2E). As a corollary, Bax -/also significantly increased optic nerve axon density (+56%, 0.55 ± 0.05 vs. 0.87 ± 0.02 axons/μm 2 , p=0.0005, Fig. S2F). We observed no significant difference in the percent of degenerating axons between WT (2.94%) and Bax -/-(2.12%) optic nerves from control eyes (p=0.85, Fig. S2G). Following 1 month of IOP elevation, we found optic nerve cross-sectional area (p≥0.98), axon density (p≥0.17), and the percentage of degenerating axons (p≥0.76) largely unchanged in both WT and Bax -/mice (Fig. S2E-G). The increase in RGC axon density observed in Bax -/optic nerves is supported by a modest increase in SC neuron density (p=0.09, Fig. 6C), but we did not detect a significant effect of IOP elevation on NeuN-positive cell density in the SC of either WT or Bax -/animals (p≥0.82, Fig. 6C).
We recently reported 1 month of IOP elevation worsens spatial frequency threshold (spatial acuity) in WT mice [22]. Here, we determined the influence of Bax on spatial acuity when challenged by IOP elevation. We measured the optomotor response to a series of 100% contrast sinusoidal gratings varying by spatial frequency. We found spatial acuity is similarly reduced for both WT and Bax -/mice over the course of IOP elevation (p>0.99, Fig. 6D).
We found 1 month of IOP elevation does not cause significant RGC body degeneration in either WT or Bax -/retinas (Fig. 1D-F). Similarly, we have previously found IOP elevation by microbead occlusion of the trabecular meshwork for 1 month does not cause significant RGC body dropout [22,41]. Although 1 month of IOP elevation does not produce significant RGC body elimination [22,41], RGC dendritic arbor complexity diminishes in the presence of Bax [22,27,36]. Here, we found this to be true for both αON-S (Fig. 2G) and αOFF-S RGCs from WT retinas (Fig. 4G). Bax -/reduces dendritic pruning of αRGCs during IOP elevation (Figs. 2G and 4G).
How does Bax -/protect RGC dendrites against IOPinduced degeneration? In the absence of Bax, RGCs may be less vulnerable to injury because cells are inherently smaller and less complex (Figs. 2 and 4). Previously, we have found evidence lending support to this idea. Mutation in the transient receptor potential vanilloid 1 (TRPV1) channel, which signal stress through calcium conductance, also intrinsically reduces αRGC dendritic complexity, and the dendrites of these very cells are less vulnerable to degeneration during IOP elevation [26]. However, the possibly also exists that in the absence of stress-induced TRPV1-mediated calcium signaling, dendrites are less susceptible to degeneration.
To overcome this problem of determining the influence of intrinsic developmental versus mechanistic changes on outcome measurements, which is typical when using conventional genetic models, inducible genetic constructs may be employed [42].
Bax appears to contribute to the IOP-induced reduction in light-and current-evoked responses of WT RGCs (Figs. 3 and 5). This claim is supported by evidence showing that modest activation of BAX/caspase-3 is required for the induction of NMDA receptor-mediated long-term depression (LTD) through the endocytosis of AMPA receptors [42]. As a corollary, Bax knockdown inhibits induction of LTD [42]. Interestingly, not only do we find Bax -/protects light-and current-evoked responses during IOP elevation, but the current-evoked responses of αON-S RGCs from Bax -/control eyes are significantly larger and rheobase smaller compared to WTs (Fig. 3E-G). This enhancement in current-evoked activity of αON-S RGCs from Bax -/retinas may be due to a lack of LTD. Unlike WT αON-S RGCs, current-evoked responses from Bax -/-αOFF-S RGCs are like those from WT retinas (Fig. 5E-G). In agreement with these divergent findings, only αON-S RGCs utilize NMDA receptor-mediated synaptic plasticity [52]. In regard to αOFF-S RGCs, we predict responses decline during stress due to increased metabolic burden to support electrogenic mechanisms [22]. In respect to metabolism, Bax -/may also support physiologic responses of αOFF-S RGCs, in particular, by protecting mitochondrial respiration and decreasing the amount of reactive oxygen species caused by nerve growth factor deprivation during stress [53][54][55][56][57].
As noted above, we found current-evoked responses of αON-S and αOFF-S RGCs from Bax -/control eyes are similar to or exceeded responses of their WT counterparts (Fig. 3E, F; Fig. 5 E, F). This finding is unsurprising given that smaller cells are typically more excitable [58]. However, the light responses of αON-S and αOFF-S RGCs from Bax -/control eyes were typically smaller compared to respective cell types from WT control eyes. This result could be explained by a decrease in excitatory input. This idea is supported by our finding that Bax -/reduces dendritic field complexity for both αON-S (Fig. 2F, G) and αOFF-S RGCs (Fig. 4F, G). Similarly, Bax -/reduces dendritic field area of VGluT3-positive amacrine and type 7 cone bipolar cells [59,60]. Although Bax -/decreases bipolar cell dendritic field area, the number of synapses with cone axons is similar to WT [59]. In support of these anatomical data, we find spatial acuity is naturally increased in control eyes of Bax -/mice (Fig. 6E). These data indicate, Bax -/increases spatial resolution at the expense of light sensitivity.

Conclusions
Our results indicate that Bax contributes to RGC dendritic degeneration and response degradation during glaucoma. In WT mice, 1 month of IOP elevation caused significant dendritic pruning and decline in responses to light in both αON-S and αOFF-S RGCs. Bax -/robustly reduced dendritic pruning and stabilized light signaling in αON-S and αOFF-S RGCs during IOP elevation. In agreement with previous studies, we found axonopathy endures during glaucoma in the absence of Bax.