Background
Proteins of the
BCL2 gene family regulate mitochondrial involvement and dysfunction during the process of intrinsic apoptosis. The family is sub-classified into three distinct groups based on the presence of BCL2 Homology (BH) domains. These include full-length anti-apoptotic proteins (BCL2 and BCL2L1 (BCLX
L)) and full-length pro-apoptotic proteins (BAX and BAK), which contain all 4 BH domains, and pro-apoptotic activator proteins which contain only the BH3 domain (BH3-only proteins such as BIM, BBC3, BID, BAD, HRK, NOXA, and BMF). Both BAX and BAK are found in an equilibrium between the surface of the mitochondrial outer membrane (MOM) and the cytosol in a process that is controlled by BCL2 and/or BCL-X
L, which retrotranslocate the pro-apoptotic proteins to the cytosol [
1‐
3]. In this equilibrium, BAX is primarily cytosolic, while BAK is predominantly located on or near the MOM surface. Upon stimulation of the cell death program, one or more BH3-only proteins are either post-translationally modified or expressed. These function to antagonize anti-apoptotic proteins, while also putatively activating pro-apoptotic full-length proteins [
4,
5].
The process of BAX activation has been most extensively studied. BAX is translocated to the MOM [
6‐
8] where it undergoes a conformational change exposing a hydrophobic C-terminal domain allowing it to anchor in the membrane. BAX monomers then dimerize and these begin to oligomerize forming large molecular weight complexes that are often localized to the scission sites of fragmenting mitochondria [
9]. A principal function of BAX is to permeabilize the MOM (MOMP), which leads to the release of pro-apoptotic molecules, such as cytochrome c and SMAC/Diablo, that signal caspase activation. Studies in tissue culture cells show that MOMP occurs virtually instantaneously with the initiation of BAX translocation [
10,
11], consistent with reports that even a single activated BAX molecule can destabilize a lipid bilayer to create a proteolipid pore large enough to allow passage of cytochrome c [
12]. An exception to this timing of MOMP is during the process of anoikis, which is apoptosis that is stimulated by loss of cell–cell or cell–matrix adhesion [
13]. Loss of adhesion induces BAX translocation, but permeabilization is delayed by several hours [
14], which allows for BAX retrotranslocation once the cells regain surface contacts.
Retinal ganglion cells (RGCs) are long projection neurons of the central nervous system that transmit light-stimulated electrical signals from the retina to visual centers in the brain through axons that project into the optic nerve. These neurons are highly sensitive to both acute and chronic optic nerve damage, which causes RGC death and blindness in a variety of conditions including glaucoma. Based on prevalence statistics for glaucoma alone, RGC loss represents one of the most common forms of neurodegeneration in humans, projected to affect more than 111 million individuals world-wide by 2040 [
15]. RGCs die by executing the intrinsic apoptotic program. Evidence for this comes from studies showing that manipulating the balance of BCL2 family proteins provides nearly complete abrogation of RGC soma death in both development and after optic nerve damage. Overexpression of BCL2 or BCLX
L in RGCs yields supernumerary cells in adult mice, indicative of failure of redundant RGCs to undergo developmental apoptosis, and dramatic preservation of RGCs after axotomy and in the DBA/2 J mouse model of glaucoma [
16‐
21]. Similarly, genetic ablation of the
Bax gene yields even greater inhibition of developmental apoptosis [
22] and virtually complete elimination of RGC death in mouse models of acute optic nerve crush (ONC) [
23‐
25] and in DBA/2 J glaucoma [
26,
27]. Importantly, most neurons, including RGCs, are exclusively reliant on BAX because
Bak transcripts undergo a splice variant that introduces a premature stop codon, eliminating full length BAK as a functioning pro-apoptotic protein [
28‐
31].
Adult RGCs express select members of the
BCL2 gene family [
32]. While
Bcl2 is expressed during early development [
33], adult RGCs predominantly express
BclXL as the principal anti-apoptotic full length gene [
34]. Stimulation of RGC death by optic nerve damage is associated with activation of a kinase cascade initiated by Dual Leucine Zipper Kinase (DLK), followed by Mitogen-activated protein Kinase Kinases (MKK4/7), Jun-N-terminal Kinases (JNK2/3) culminating in the activation of the transcription factors p53 and JUN (pJUN) [
35‐
42]. These transcription factors then selectively upregulate the expression of different BH3-only genes [
43]. Genetic ablation and gene expression studies indicate that RGCs have overlapping or redundant dependence on multiple BH3-only proteins [
25,
44‐
48].
To date, few studies have examined the process and timing of BAX activation during the progression of RGC death, even though this likely represents the most critical transition point in the commitment of cells to the apoptotic pathway [
49]. Immunostaining for BAX shows increased staining of RGCs beginning as little as 30 min after acute optic nerve damage, peaking at 3–7 days [
48,
50]. While these results were originally interpreted as showing an increase in BAX expression, they likely reflect the process of translocation and concentration of cytosolic BAX at the MOM of RGCs. Supporting this, more recent studies have shown that forced expression of a GFP-BAX fusion protein in RGCs failed to spontaneously induce BAX translocation and death but confirmed that optic nerve damage induced significant translocation in cells during the 3–7 day interval [
25,
32]. Longitudinal imaging of GFP-BAX expressing RGCs using blue-light confocal scanning laser ophthalmoscopy also suggested that cells exhibiting translocation remained present for several days before clearance [
32].
Since development of neuroprotective therapies for RGCs critically relies on BAX activation, we sought to investigate the kinetics of this process in RGCs using both static and live-cell imaging methodologies. Our results demonstrate that the translocation and permeabilization two steps of BAX activation are mechanistically and temporally distinct in RGCs. While most cells irreversibly translocate, some cells exhibit spontaneous retrotranslocation of BAX in a process that resembles the interruption of anoikis, indicating that translocation and MOMP do not occur simultaneously. Further, we show that BH3-only proteins are critical to drive the translocation step, but are not sufficient to elicit permeabilization, which requires the activation of kinases distinct from the DLK-JNK axis. Together, this work highlights several points of intervention to prevent BAX activation and introduces the possibility of reversing the early stage of translocation.
Discussion
We have used a GFP-BAX fusion protein to analyze the characteristics of BAX activation in RGCs of mice after optic nerve damage. The fusion protein is a useful tool to illuminate the process of BAX activation because it can be monitored in living and fixed cells, likely without altering the characteristics of cell death executed by the endogenous proteins of the BCL2 gene family. While lowering BAX levels has a dramatic effect on the process of RGC death [
24,
25,
75], increased expression (above endogenous levels) has no impact on normal neuronal programmed cell death [
76] or the timing of RGC loss after optic nerve damage when comparing two mouse strains with different levels of endogenous BAX [
24]. Additionally, titrating GFP-BAX expression in BAX-deficient cells indicates that the maximum effect on apoptosis is reached at a threshold concentration of BAX and is not further enhanced by higher levels [
75]. A major reason why elevated BAX levels are not spontaneously pathologic is that anti-apoptotic proteins are often in large excess relative to BAX. In mouse RGCs, this difference is estimated to be between 5–tenfold greater for BCLX
L, the principal antagonist of BAX in neurons [
24,
25]. Ultimately, intrinsic apoptosis is driven by the stoichiometric interactions between the levels of BH3-only proteins that are activated or synthesized in response to a stimulus and the endogenous concentration of anti-apoptotic BCL2 gene family proteins [
5,
77].
To our knowledge this study represents the first kinetic analysis of BAX translocation and permeabilization in cells present in a complex tissue environment. The process of translocation can be initiated several days after the initial damage to the RGC axon, but once initiated proceeds rapidly in RGCs albeit with statistically slower kinetics than we have observed in tissue culture cells. Slower translocation of BAX to the MOM may be a consistent feature of RGCs regardless of the apoptotic stimulus, including exposure to STS, a kinase inhibitor that has frequently been used to assess the kinetics of BAX activation in tissue culture cells. This may be a phenomenon that is intrinsic to RGCs, but at this time we cannot rule out that this process is influenced by surrounding cells, such as supporting macro- and microglia that are responding to the damaged environment. Such responses could be the generation of trophic factors, clearance mechanisms of damaged RGC compartments, and the efforts to maintain ion homeostasis and synaptic function [
78‐
80].
We observed translocation of GFP-BAX in several RGC compartments, including the dendritic arbor and both the intraretinal and optic nerve axon. This observation suggests that BAX-mediated catabolism may play a role in degeneration of these compartments and could explain why axonal degeneration in glaucomatous DBA/2 J mice is delayed in the absence of a functional
Bax gene [
26]. BAX and/or caspase-dependent axonal degeneration has been described in several neuronal cell types [
81‐
83] and may precede SARM-1-mediated calpain-dependent catabolism, observed in other cell types [
84]. Similarly, the pruning of dendritic branches has been linked to caspase activation [
85,
86] and recent studies suggest that arbor remodeling in some RGCs is mediated by a BAX-dependent mechanism [
87].
We also observed a small percentage of cells with punctate GFP-BAX restricted to the dendritic arbor in both ONC retinas and in retinas treated with the FAK inhibitor PF573228. The arbor may be a critically susceptible region of the RGC where the process of BAX translocation is initiated and may account for why larger RGCs, which have the largest and most complex dendritic arbor branching [
88], were the earliest to exhibit BAX translocation in both ONC and FAK inhibition experiments. We speculate that once initiated, BAX translocation may then propagate to other compartments of the RGC. The propagation of apoptosis, including BAX translocation, MOMP, mitochondrial membrane depolarization, and caspase activation has been described in tissue culture cells and some neurons [
89‐
97]. With respect to BAX translocation in tissue culture cells, propagation is initiated preferentially in distal tips of processes and then travels to the soma and even into additional processes at a velocity of ~ 7 µm/sec [
89]. Within the soma, however, BAX translocation initiates simultaneously at mitochondrial foci. If RGCs also propagate BAX translocation at a similar rate, this would explain why we observed so few cells with punctate BAX restricted to the arbor, suggesting we were capturing a rare early stage of propagation. The concept that BAX translocation is initiated first in the dendritic arbor aligns with observations made by others that remodeling of this compartment represents one of the earliest morphological changes associated with RGC pathology after optic nerve damage [
98‐
101].
Our results also showed that the processes of BAX translocation and permeabilization in RGCs were separated by a significant delay. This contrasts dramatically from studies in tissue culture cells undergoing BAX-dependent apoptosis, where BAX permeabilization leading to MOMP occurs virtually simultaneously with the translocation of the first BAX molecules [
11,
102,
103]. The precise length of the delay interval in RGCs was not clear from our experiments, but we posit that it may range from several hours to days. Longitudinal imaging of GFP-BAX translocation in eyes of mice subjected to ONC suggests that cells with translocated BAX can remain present for up to 5 days before being cleared [
32].
A small proportion of RGCs exhibited first translocation of BAX followed immediately by its retrotranslocation back into the cytosol, a phenomenon that has also been documented in detached tissue culture cells that have initiated anoikis, but are then allowed to reattach with the surrounding matrix [
14,
73]. Studies on anoikis suggest that the process is initiated by the loss of function of FAK. In this pathway, FAK inhibition leads to the activation of the latent BH3-only protein BMF [
67,
104]. Anoikis also exhibits a delay between translocation and permeabilization that has been estimated to be up to 3–4 h. We hypothesized that a similar mechanism may be a critical feature of some damaged RGCs. Consistent with this, we observed a significant increase in BAX translocation in naïve retinas exposed to the FAK inhibitor PF573228. Importantly, PF573228 induced a similar level of translocation after 24 h as ONC at this time point but did not induce high levels of BAX permeabilization. We interpret the effect of PF573228 as priming BAX in RGCs without necessarily activating it. This would explain previous observations that FAK inhibition exacerbates RGC loss but only after optic nerve damage [
105,
106].
The mechanism of BAX translocation is still not elucidated but is likely tied to the activity of BH3-only proteins. A predicted primary function of BH3-only proteins is that they sequester anti-apoptotic proteins like BCLX
L allowing for a shift in the cytosolic-MOM equilibrium of pro-apoptotic counterparts [
77]. This concept is reenforced by data showing the BH3-only proteins have a stronger binding affinity to anti-apoptotic members, and only exhibit transient interactions with pro-apoptotic proteins [
77,
107]. BH3-only proteins, however, are also thought to play a direct role in BAX translocation. Evidence suggests that some BH3-only members bind to the surface of inactive BAX dimers in the cytosol to facilitate their separation into monomers and translocation to the MOM [
108,
109]. This form of BAX exposes the C-terminal membrane anchor domain allowing it to associate with Voltage-dependent anion channel type 2 (VDAC2) on the MOM. This is also the likely state of BAX and BAK in healthy cells when they are in the BCLX
L/BCL2-mediated equilibrium with the cytosol, and we predict, the state of translocated BAX in affected RGCs that have not undergone permeabilization.
Protein structural studies also indicate that BH3 mimetics play a role in the BAX permeabilization step. Mimetics can bind to a region of BAX distinct from its BH3-domain causing it to undergo a conformational change leading to unfolding of the “latch” domain that exposes hydrophobic alpha-helices that further strengthen the MOM-BAX interaction [
4,
110] and the initiation of proteolipid pores. This model is difficult to reconcile with our results using kinase inhibitors. Both SB203580 and sunitinib failed to block the expression of BH3-only genes that are linked to the DLK-JNK axis (
Hrk and
Noxa) suggesting that these BH3-only proteins (and presumably
Bbc3 and
Bim, which are also tied to p53 and pJUN activation, respectively) were active in these experiments, yet we still observed a significant attenuation of BAX permeabilization. Additionally, combined deletion of
Mkk4/7, which inhibits the expression of these BH3-only genes, effectively blocked translocation of BAX in a significant proportion of cells. Therefore, our results are more consistent with BH3-only proteins playing a critical role in the translocation of BAX but are, by themselves, not sufficient for permeabilization.
The mechanism of retrotranslocation in damaged RGCs is also not clear. In healthy cells, BAX is retrotranslocated by BCLX
L [
1], but it is unknown if this same mechanism is active in reattached cells during anoikis or in RGCs that exhibit spontaneous retrotranslocation. Overexpression of BCLX
L in both tissue culture cells and RGCs prevents BAX translocation in the first place [
21], making study of its role in retrotranslocation difficult. We predict a model, however, in which a subset of RGCs undergoing anoikis are able to recover cell–cell contacts and have sufficiently high enough levels of endogenous BCLX
L to initiate retrotranslocation. Further studies to explore this model are ongoing.
The permeabilization of the MOM by translocated BAX in anoikis is linked to the activity of p38/MAPK14 [
73], although it is not clear if this kinase modifies BAX directly, or if the mechanism is indirectly linked to p38/MAPK14 modification of an intermediate substrate. A recent report implicates DRP1 as directly interacting with BAX and mediating its activation [
111] and p38/MAPK14-dependent phosphorylation of DRP1 has been linked to other neurodegenerative conditions [
112,
113]. p38/MAPK14 has previously been shown to have a role in RGC pathology after both acute and chronic optic nerve damage [
114,
115], but the actual mechanism of neuroprotection afforded by blocking this kinase has not been studied in detail. Blocking p38/MAPK14 with SB203580 had no impact on preventing BAX translocation, but it did suppress the permeabilization of translocated BAX up to 5 days after ONC. The inability to block translocation contradicts previous reports that p38/MAPK14 is required for this process in cultured cells [
116‐
118], although these studies relied on cell fractionation experiments or used antibodies that only recognized permeabilized BAX. Alternatively, the difference between these previous observations and ours could be associated with a difference in activation mechanisms between isolated cultured cells and RGCs in a complex tissue.
Interestingly, we also observed that a low dose of sunitinib also blocked BAX permeabilization, but not translocation. Sunitinib is a broad-spectrum kinase inhibitor, but does not directly affect p38/MAPK14 activity [
74]. Additionally, at the concentration we used, we did not affect activation of the DLK-JNK axis, thus modulation of either MKK4/7 or JNK2/3, both of which have been linked to the activation of p38/MAPK14 [
119], would not be expected to be affected in our experiments. The implication of this is that low levels of sunitinib were affecting one or more kinases that have yet to be identified in the BAX activation process. Based on known selectivity measures [
74], the concentration of sunitinib we used could target as many as 99 other known kinases before reaching efficacy to inhibit DLK. Identification of other kinases in the BAX activation process in RGCs is ongoing.
Acute damage to the axons of the optic nerve leads to nearly complete loss of the entire RGC population in the mouse retina [
120], but elegant studies using single cell RNA-Seq clearly show that RGC populations exhibit varying levels of susceptibility to this damage paradigm [
121]. One interpretation of these differences in susceptibility is that damage induced by ONC is not monolithic and is instead multifactorial. For example, some RGCs may be susceptible to loss of FAK signaling, while others are more susceptible to ER stress or accumulation of reactive oxygen species, etc., as part of a continuum of damaging stimuli. Since
Bax deletion blocks essentially all RGC soma death after optic nerve injury [
23,
24,
26] it is likely the common denominator of all the proapoptotic signaling pathways in these cells after optic nerve damage. Consequently, we predict that an analysis of the pattern of BH3-only protein activation/expression would act as a signature of the critical damaging stimulus of a given RGC subtype, consistent with studies that show that the pattern of BH3-only proteins activated in apoptosis varies with the type of stimulus [
5,
122]. Ultimately, final susceptibility may be a function of the cooperation of different members of the BH3-only protein family. For example, RGCs susceptible to FAK inhibition may be more sensitive to anoikis-driven BMF activation, which would account for why a higher proportion of cells exposed to PF573228 exhibited punctate GFP-BAX without DLK-JNK driven accumulation of pJUN and its associated BH3-only proteins. Variable RGC susceptibility as a function of BH3-only protein levels also explains observations that selective deletion of these genes only provides partial protection to RGCs after ONC [
41,
44‐
48]. The implication of a model in which RGC death is a function of multiple different, and possibly sequential stimuli, is that there is no single early molecular event that can be targeted therapeutically, complicating the development of a global neuroprotective strategy. Conversely, targeting later BAX activation, particularly the permeabilization step, could offer promise as a strategy to facilitate recovery of RGCs after damage.
Methods
Animals
Adult BALB/cByJ or C57BL/6 J mice (Jackson Laboratory, Bar Harbor, ME) were housed in microisolator cages with a 4% fat diet (8604 M/R; Harland Teklad, Madison, WI) and maintained in a 12 h light/dark cycle. C57BL/6 J mice carrying dual floxed alleles of
Mkk4 and
Mkk7 were developed as previously reported [
38]. Biological variability involving sex-related differences in response to optic nerve damage was considered in the design of experiments. While no sex differences have been detected using the acute ONC protocol [
51], experimental groups were each assigned an equal mixture of male and female mice. Animals were handled in accordance with the Association for Research in Vision and Ophthalmology statement on the use of animals in research. The Institutional Animal Care and Use Committees of the University of Wisconsin and the University of Rochester approved the experimental protocols and ethical care for mice used in these experiments. Any animals that exhibited distress were removed from the study. Euthanasia was performed by intraperitoneal injection of 0.05 mL of Euthasol (Virbac AH, Fort Worth, TX). Following cessation of a detectable heartbeat, mice were cervically dislocated.
Optic nerve crush
ONC was performed as previously described [
52,
123]. Briefly, mice were anaesthetized via intraperitoneal injection of ketamine (120 mg/kg) and xylazine (11.3 mg/kg). One drop of 0.5% proparacaine hydrochloride (Akorn, Lake Forest, IL) was used to numb the experimental eye. A lateral canthotomy exposed the optic nerve and was followed by an incision in the conjunctiva at the limbal junction. The optic nerve was clamped for 3 s using self-closing N7 forceps (Roboz Surgical Instrument Co, Gaithersburgs, MD). After surgery, triple antibiotic ointment was used to cover the eye, and a subcutaneous injection of buprenorphine (0.2 mg/kg) was delivered to mitigate pain. The right eye was not surgerized and evaluated as the contralateral condition.
Intravitreal injections
After anesthesia as described above, a sterile 30G needle was used to puncture a small hole through the conjunctiva and scleral tissue. A beveled 35G Nanofil needle affixed to a Nanofil syringe (World Precision Instruments, Inc, Sarasota, FL) was inserted in the pre-poked hole to deliver desired virus, drug or vehicle. The contents were slowly delivered over 30 s, and the needle remained within the globe for an additional 30 s with caution, such that the lens was not damaged. Antibiotic ointment was placed over the injection site and buprenorphine (0.2 mg/kg) was administered to relieve pain. AAV2/2-Pgk-GFP-BAX, AAV2/2-Pgk-mCherry-BAX, and AAV2/2-Pgk-Cytochrome c-GFP were packaged by the University of North Carolina (UNC) Vector Core (Chapel Hill, NC) (1–5 × 1012 genome copies/ml). AAV2/2-CMV-Cre virus (> 1012 gc/mL) was purchased from the UNC Vector Core. All injections of virus were conducted at least 1 month prior to conducting optic nerve crush surgery.
The following inhibitors were diluted in sterile DMSO for intravitreal injection; FAK inhibitor PF573228 (Tocris Bioscience, Bristol, UK), p38/MAPK14 inhibitor SB203580 (Millipore Sigma, St. Louis, MO), and DLK inhibitor Sunitinib (Selleckchem, Houston, TX). Experimental eyes received 1 µL of PF573228 (10 µM – 25 mM), 25 µM SB203580, or 10 µM Sunitinib, while the contralateral eyes received 1 µL DMSO-only injections. The concentrations chosen were shown by others to have a biological effect on mouse and rat RGCs in vivo (PF573228 ref [
106], SB203580 ref [
114]) or provide maximal protection to hESC-derived RGCs challenged with colchicine (sunitinib ref [
36]). If mice received an inhibitor injection the same day they underwent ONC, the intravitreal injection was performed immediately following the ONC procedure. For STS treatment, eyes were injected with injected with 0.6 µL of a 10 µM solution to yield an approximate final concentration of 1 µM STS.
Clones and plasmids
The GFP-BAX construct used for tissue culture cell experiments was previously described [
75]. The HDAC3-Flag construct was a gift from Eric Verdin (Addgene plasmid no. 13819: Addgene, Cambridge, MA). The mitoBFP construct was described previously [
11,
53]. Transgenes in these plasmids were all driven by the CMV promoter. Plasmids used for generating AAV2 viral particles were derived from the parent plasmid AAV-
Pgk-Cre (Addgene plasmid no. 24593), a gift from Dr. Patrick Aebischer. The coding sequence for Cre was removed by digestion with EcoRI and SalI and was replaced with a sequence bridge of single cutting restriction sites. This bridge vector was used to subclone GFP-BAX [
75], mCherry-BAX (which was generated in the laboratory by subcloning BAX into pmCherry-C1, Clontech/Takara Bio, Kusatsu, Shiga, Japan), and cytochrome c-GFP (Addgene plasmid no. 41182), a gift from Douglas Green. All plasmids were validated by sequencing.
Tissue culture cell maintenance
661W cells (mouse retinal progenitor cells) were cultured in DMEM (Cellgro, Mediatech, Inc., Manassas, VA) supplemented with 10% FBS (Atlanta Biologicals, Norcross, GA) and 1% Antibiotic/Antimycotic (Cellgro, Mediatech, Inc.) at 37 °C and 5% CO2. Treatment with 316 nM staurosporine for 24 h induced 661W differentiation [
124]. For transfection, prewarmed media containing 2 µg each of the GFP-BAX, mitoBFP, and HDAC3-Flag plasmids was mixed with 2 volumes of Transfast reagent (E2431, Promega, Madison, WI), vortexed and incubated at room temperature for 15 min. After removal of regular media, the transfection mixture was added to cells and they were incubated at 37˚C for 1 h, after which fresh media was added to the cells and they were incubated for a further 15-24 h. Overexpression of HDAC3-Flag is selectively toxic to differentiated neurons [
125], including differentiated 661W cells [
31] and plays a critical role in the apoptotic program executed by RGCs [
57,
59]. 661W cells, originally described as RGC-5 cells from
Rattus norvegicus, were confirmed in our laboratory as
Mus musculus origin by PCR of the mitochondria d-loop [
126] and the upregulation of TUBB3 after differentiation [
31].
Live imaging and analysis
Live-cell imaging was performed using an Andor Revolution XD spinning disc confocal microscopy system (Andor, Belfast, Northern Ireland) comprised of the Nikon Eclipse Ti inverted microscope, Nikon objectives, an iXon × 3 897 EM-CCD camera, a Yokogawa CSU-X1 confocal spinning disk head, the Andor Laser Combiner with four solid state lasers, an ASI motorized stage with Piezo-Z, and an Okolab CO2 cage incubator for temperature and CO2 control at 37 °C and 5% CO2.
661W cells were plated at a density of 100,000 – 300,000 cells per well on 4-well chamber slides containing #1.5 optical grade plastic (Ibidi, Madison, WI). Cell treatments were staggered for an appropriate time for the cell type to ensure the ability to image each well on a slide. Prior to imaging, phenol-containing media was removed from the well, and replaced with a recording media prepared in-house (HBSS with 1.26 mM calcium chloride, 0.49 mM magnesium chloride, 0.4 mM magnesium sulfate, no phenol red, 4.5 g/L glucose and 10 mM HEPES) supplemented with appropriate serum concentration. Under temperature and CO2 control, time-lapse imaging was performed by collecting z-stacks consisting of 20–25 optical sections at 0.22 μm (Nyquist) taken every minute for one to two hours.
Live ex vivo imaging of mouse retina was performed with the same microscopy conditions as discussed above. Mice were first subjected to ONC surgery and then euthanized 7 days later with a lethal overdose of pentobarbital sodium. Once euthanasia was confirmed, an intravitreal injection of DRAQ5 (ThermoFisher Scientific, Waltham, MA) was administered (200 μm final concentration), then the enucleated eye globe was incubated in nutrient rich medium at 37 °C for 10 min. DRAQ5 labeling was stable for 3–4 h. To mount the retina, an eye cup was created from the eye globe, then a single relaxing cut through the retina and sclera was made in the anterior region. Without removing the dissecting scissors from the relaxed cut, in one quick motion, the lens was removed with the dissecting scissors as the retina was flipped such that the ganglion cell layer was facing down onto the coverslip bottom (#1.5) of a chamber slide (Ibidi). A second relaxing cut was made, allowing the retina to rest as flat as possible. A drop of 2% low melt agarose (ThermoFisher Scientific) medium was placed on the retina, followed by a small piece of coverglass to aid in flattening the retina for imaging purposes. Once positioned, the retina was overlayed in the agarose medium, allowed to polymerize briefly, and covered with a layer of sterile water to prevent evaporation and image drifting during acquisition. The low melt agar was made in 50% ‘recording media’ (above) and 50% Neurobasal medium (ThermoFisher Scientific) supplemented with 1X Neurocult SM1 (Stemcell Technologies, Cambridge, MA) and 10% FBS (Atlanta Biologicals). The FBS and Neurocult supplements were added after the melted agarose had been cooled to 60˚C. Both experimental and control retinas were mounted in the same way, and this process was completed in approximately 45 min from the time of euthanasia. Z-stacked images were collected every 5 min for up to three hours after the start of time-lapse imaging. DRAQ5 staining provided a control for tissue drift during imaging. Only RGCs that maintained a fixed position, based on nuclear visibility, were used for subsequent kinetic analyses.
For STS treatment, naïve animals were euthanized and the eyes injected with DRAQ5 and STS, the latter to a final concentration of 1 µM and incubated at 37˚C for 10 min. The retinas were then harvested and mounted for imaging as described above except that the media contained 1 µM STS during the imaging procedure.
All images were subject to the same background subtraction and Gaussian filtering within the IMARIS 7.7 software (Bitplane, Concord, MA), where each filter had its own constant algorithm and threshold setting. The spots function was used to identify and track GFP-BAX foci within a cell and extract the maximum fluorescence of the desired fluorophore from each surface. A subset of BAX foci was selected for computational analysis based on the longevity of the track, whether the track encompassed the translocation event and remained within the defined 3-dimensional space of the cell, and whether the BAX foci were spatially isolated. For ONC experiments, a total of 33 cells, from 7 retinas, exhibiting translocation during an imaging session were recorded. Of these cells, we were able to extract kinetic data for a minimum of 3 mitochondrial foci per cell for 27 cells (5 of which also exhibited retrotranslocation). For STS treated retinas, we were able to record 10 cells from 4 retinas, all of which were suitable for kinetic analysis.
Computational analysis
BAX translocation analysis was performed as described previously [
11]. Briefly, raw maximum fluorescence data extracted from the IMARIS program was normalized (to its initial value) for each foci within a cell to allow for comparison among many GFP-BAX foci. These data were then fit to a sigmoid curve using a linear regression model. The code using the SciPy library in the Python programming language is available in Maes et al., 2017 [
11]. Quantification for retrotranslocating BAX cells was performed by applying sigmoid functions to each phase separately (translocation and retro-translocation). The retrotranslocation curve was first fit using a gaussian function to determine the maximum point of the curve, which would define the two phases.
Retinal tissue preparation and immunostaining
Enucleated eyes were punctured on the anterior portion of the limbus with a 30 gauge needle and then immersed in 4% paraformaldehyde in PBS for 1 h. Eye cups were made by removing the anterior chamber of the eye by cutting along the limbus. In preparation for immunofluorescent staining, retinal eye cups were first wiped with the tip of a twisted Kimwipe to remove the residual vitreous and then equilibrated in 30% sucrose in PBS overnight at 4˚C. They were then subjected to 3 rounds of freeze-thawing on dry ice and blocked in PBS containing 2% goat serum and 0.1% Triton X-100 overnight at 4˚C. Primary antibodies were added to the same buffer and incubated with the eyecups for a minimum of 3 days at 4˚C. After incubation, the eyecups were washed 3 times in PBS and then incubated in PBS containing 2% goat serum and 0.1% Triton X-100 overnight at 4˚C.
For immunostaining longitudinal sections of the optic nerve, fixed eyecups containing ~ 1 mm of the optic nerve were equilibrated in 30% sucrose in phosphate buffered saline (PBS) (50 mM phosphate, pH 7.4, 150 mM NaCl) overnight at 4˚C and then embedded in Optimal Cutting Temperature media (Scigen Scientific, Gardena, CA) on dry ice. Cryosections (10 µm) were selected that included the optic nerve head and an extended portion of the optic nerve. These were first incubated in PBS at room temperature for 10 min and then blocked in PBS containing 10% horse serum (Lonza, Basel, Switzerland) and 0.1% Triton X-100 for 1 h at room temperature in humidified chambers. Sections were then covered in PBS containing 2% BSA and 0.1% Triton X-100 with primary antibody (SMI-31 see below) overnight at 4˚C. After incubation, the sections were washed 3 times in PBS and incubated in the same buffer (PBS, 2% BSA, 0.1% Triton X-100) with the secondary antibody for 1 h at room temperature.
The following primary antibody and corresponding concentrations were used: chicken anti-GFP polyclonal (1:1000) (no. ab13970, Abcam, Cambridge, UK), rabbit anti-mCherry polyclonal (1:500) (no. ab167453, Abcam), mouse anti-BAX (1:50) (6A7) monoclonal (no. sc-23959, Santa Cruz Biotechnology, Dallas, TX), rabbit anti-pJUN (Ser73) monoclonal (1:500) (D47G9) (no. 3270, Cell Signaling Technology, Danvers, MA), and mouse anti-phospho-neurofilament monoclonal (1:1000) (SMI-31) (no. NE1022, Millipore Sigma). Eyecups were then washed in PBS and incubated in a solution containing secondary antibody overnight at 4˚C. The following secondary antibodies were used at a concentration of 1:1000: polyclonal goat anti-mouse antibody that was conjugated to AlexaFluor-594 (Jackson ImmunoResearch Laboratories, West Grove, PA), polyclonal donkey anti-chicken conjugated to AlexaFluor-488 (Jackson ImmunoResearch), polyclonal goat anti-rabbit antibody conjugated to AlexaFluor-594 (Jackson ImmunoResearch). Following incubations with secondary antibodies, both eyecups and sections were washed 3 times in PBS. For whole mounting, the retinas were isolated from the eyecups and placed on glass Superfrost Plus slides (ThermoFisher Scientific) with 4 relaxing cuts to help flatten the tissue. Both sections and retinas were then mounted using Vectashield containing DAPI (Vector Laboratories, Burlingame, CA), coverslipped, and sealed with clear nail polish. Slides were protected from light and stored at 4˚C until image acquisition.
Fixed tissue imaging and analysis
For all images used for quantification, samples were imaged on a Zeiss Axioimager Z2 upright microscope (Zeiss, Oberkochen, Germany). All images were analyzed using Zen Blue software (Zeiss). Separate imaging modalities were used to quantify the percentage of cells with punctate GFP-BAX and the percentage of 6A7 positive cells. To quantify the percentage of GFP-BAX labeled cells with punctate GFP-BAX, 4 images per retinal quadrant were taken using a 20X air objective. The retinal eccentricity was held constant in the mid-periphery. For those 16 images, every GFP-BAX positive cell was counted. A minimum of 700 cells were counted for each retina. To quantify the percentage of 6A7 positive cells, a second set of images was taken of each sample. Beginning at one side of the retina, the camera was panned across the retina, and every field that contained a cell with punctate GFP-BAX was imaged. This was necessary to maximize the number of cells with punctate GFP-BAX counted per retina. An average of 94 punctate cells were counted for each retina with a range of 33–177 cells counted. The images of GFP-BAX puncta in the dendritic arbors of RGCs were imaged on a Leica SP8 (Leica Camera, Wetzlar, Germany) confocal microscope using a 1.4 NA 63X oil objective.
Realtime quantitative polymerase chain reaction
Immediately after euthanasia, eyes were enucleated and the retinas were extracted and flash frozen in microcentrifuge tubes on dry ice. A total of 12 retinas were used for each experimental group. These were split into 3 groups of 4 and their OS and OD retinas were frozen in separate tubes. Therefore, each experimental group in the experiment consists of three replicate values obtained from three mice each. RNA was extracted from the pooled retinas using a DNA/RNA/Protein extraction kit (IBI Scientific, Dubuque, IA). RNA quality was verified using a Nanodrop One spectrophotometer (ThermoFisher Scientific). All RNA samples had a A260/280 value between 2.10 and 2.13 and an A260/230 value between 2.00 and 2.30. cDNA was created by annealing 2 µg of total RNA from each sample to Oligo(dT)
15 Primers (Promega) and then incubating with MMLV-RTase (Promega) in a 40˚C water bath for 1 h. The 25 µL total volume of each reaction was diluted by addition of 200 µL of diethyl pyrocarbonate treated, nuclease free water (ThermoFisher Scientific). Quantitative PCR was performed in a Quant Studio 7 Flex Real Time PCR machine (ThermoFisher Scientific) using a customized TaqMan Array Card (ThermoFisher Scientific) that was designed to assess the abundance of the following mouse RNA transcripts:
18S, Hrk, Ecel1, Atf3, Noxa, and
Cdkn1a (p21). Each sample was assayed in triplicate and the mean of these values was used as the value assigned to each sample of pooled retinas. Changes in transcript abundance were calculated using the △△C
t method [
127] using the 18S rRNA value from each sample as a reference to calculate △C
t and then the change caused by the experimental paradigm was calculated as the difference in △C
t values between the experimental eyes and the contralateral eyes.
Statistical analysis
Calculations of sample size for statistical power were made using previous data on outcome metric variances for ONC experiments. Cohort sizes of n = 5 retinas provide a power set to 0.8 to detect a difference of 12% between groups and n = 4 retinas to detect a 20% difference. Translocation rates from live-cell imaging experiments did not fit a normal distribution and were Log2 transformed for analysis. One-way analysis of variance (ANOVA) was used to evaluate differences when comparing multiple groups. A two-sided Student’s t-test was used to compare the means of two isolated cohorts. A χ2 test was used to compare the proportions of punctate GFP-BAX cells showing accumulation of pJUN after ONC or injection of PF537228. To assess the intracellular synchronicity of both the time of initiation and the rate of translocation, we used data collected from multiple points within a single cell to determine the coefficient of variation (CoV) for each variable per cell. These statistics were normalized to a percentage (absolute value) of the mean for each cell within a single group to facilitate comparison between groups. Data was generally represented as scatter plots showing mean ± standard deviation in graphs. Each data point represents the value obtained from an individual retina unless otherwise specified. For clarity, larger datasets were graphed as Box and Whisker plots showing the mean as a red line ± standard deviation. Individual data points outside of the standard deviation were included as individual points on each graph.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.