The C57BL/6-Gt(ROSA)26Sor^{tm1(CAG-mCherry/GFP)Ganl/GanlH} mice were obtained from the MRC Harwell Institute, which distributes this strain on behalf of the European Mouse Mutant Archive (EMMA: www.infrafrontier.eu). The repository number is EM:11343. The C57BL/6-Gt(ROSA)26Sor^{tm1(CAG-mCherry/GFP)Ganl/GanlH} mice were originally produced from transgenic mice generated using Taconic Artemis (McWilliams et al., 2016) using ES cells from C57Bl/6NTac mice. Once at the University of Dundee, the mice were backcrossed to the C57Bl/6 J strain for more than 10 generations before distribution (I. Ganley, personal communication). Mice were housed at the University of North Texas Health Science Center and are hereafter referred to as “MitoQC” mice. The MitoQC mice were generated by inserting a CAG promoter cassette with an open reading frame for the mCherry-GFP-FIS1 fusion protein and Kozak sequence (GCCACC) into the mouse Rosa26 locus, which were maintained through a C57BL/6 J background (McWilliams et al., 2016). The fusion of mCherry-GFP to the outer mitochondrial membrane protein allows the mitochondrial network to fluoresce green and red under steady-state conditions (McWilliams et al., 2016). The GFP fluorescence is pH-dependent; therefore, when the mitochondria are delivered to the lysosomes, the GFP fluorescence is quenched due to the acidic microenvironment, leaving only the mCherry fluorescence, thereby indicating mitophagy (McWilliams et al., 2016).

Mice within the ketogenic diet experimental groups were switched to a low-carbohydrate ketogenic diet (Research Diets, catalog # D12369B) for 5 weeks, which consisted of 10.4% protein, 0.1% carbohydrate, and 89.5% fat. The two control diet groups continued to eat the standard lab chow (Diet 5,058, PicoLab Mouse Diet 20), which consisted of 23.2% protein, 55.2% carbohydrate, and 21.6% fat. The standard lab chow is pellets that were placed in the cage hopper, while the ketogenic diet is a soft dough placed in a small stainless-steel bowl. Mice were weighed before the experiment, then on days 3, 7, 10, 13, 17, 20, 30, and 35. The food was weighed on the same schedule to monitor diet consumption. A Nova Max Plus hand-held ketone testing device was used to measure ketones from tail vein blood once weekly for three mice on the ketogenic diet. Mice were split into four groups: control diet (Naive, n = 14), Control diet with ocular hypertension (OHT, n = 14), Ketogenic diet (KD, n = 13), and Ketogenic diet with OHT (KD + OHT, n = 13).

To increase intraocular pressure, mice were anesthetized with isoflurane (2.5%) delivered through a vaporizer with oxygen, then received a drop of 0.25% tropicamide (Sandoz Inc.) on the cornea to dilate the pupil, followed by 0.5% proparacaine hydrochloride eye drops (Akorn) for local pain relief. A glass-pulled micropipette was used to inject 2 μL of 8-μm-diameter magnetic beads (Bangs Laboratories, UMC4F Compel) into the anterior chamber while curved forceps stabilized the eye. A neodymium magnet was used to draw the beads into the iridocorneal angle to occlude the trabecular meshwork while the micropipette was still in place. Once the beads were in place, the micropipette was removed, and Pura-lube ophthalmic ointment was applied to the eye while the mice were placed on a warm water blanket to recover.

For intraocular pressure measurements, mice were anesthetized with isoflurane as described above, while a Tono-lab rebound tonometer (calibrated for mice; iCare) was used to take 10–20 measurements to calculate an average. The instrument-calculated running average was only used to determine the quality of the measurements through the provided standard deviation. All IOP measurements were recorded within 3 min of anesthesia to avoid anesthesia-associated IOP decline. Baseline IOP was taken prior to ocular hypertension, then every 7 days after the procedure for 4 weeks. An IOP integral was calculated by subtracting the baseline IOP from the weekly average, multiplying the difference by the number of days between measurements, and summing the week-by-week numbers. IOP integrals are expressed in mmHg-days.

Mice were anesthetized with 2.5% isoflurane and placed on a heating pad for thermal regulation; body temperature was maintained at 37°C. Eyes were topically anesthetized with 0.5% proparacaine hydrochloride (Akorn), and pupils were dilated with 0.5% tropicamide. A 32G needle connected to a reservoir filled with sterile phosphate-buffered saline (PBS) was inserted into the anterior chamber (Dibas et al., 2018). The reservoir was elevated to increase the IOP to 40 mmHg above systolic blood pressure for 30 min. After the IOP increase, the needle was removed to allow for reperfusion. Mice treated in this way served as a positive control for hypoxia analysis. Two additional groups were included in the pimonidazole experiment: Control and Control + OHT mice. Mice in the Control + OHT group (n = 3) were induced with elevated IOP using the magnetic microbead model, as described in the acute glaucoma model section above, 24 h before pimonidazole (Pimo) injection. Mice from the Control group (n = 3) underwent 15 min of isoflurane anesthesia 24 h before injection to control for any effect of anesthesia on hypoxia. Hypoxia was identified using pimonidazole hydrochloride (Hypoxyprobe Pacific Blue, Hypoxyprobe, Burlington, MA, United States), a compound that forms adducts in cells that experience a partial pressure of oxygen <10 mmHg. Pimonidazole hydrochloride (60 mg/kg) was diluted in sterile PBS and injected IP 10 min after completion of the IOP increase; mice were sacrificed 90 min later using an overdose of sodium pentobarbital (Euthasol) and perfused first with 0.1 M PBS (pH 7.4) then with 4% paraformaldehyde (Jassim and Inman, 2019). Retinas were cryoprotected in 30% sucrose and then processed as described in the section on Tissue Collection. Retinal sections were immunolabeled with anti-RBPMS antibody and an antibody against pimonidazole (Pimo) that was conjugated to Pacific Blue fluorophore, then cover-slipped with Fluoromout G (Southern Biotech) for imaging. Mitolysosome quantification was performed on Pimo-positive cells as described under Mitolysosome Quantification (below), with the exception that the blue channel analyzed was Pimo, not RBPMS. Quantification of Pimo-positive RGCs was done on images cropped in ImageJ to include only the retinal ganglion cell layer.

Mice from each treatment group (4–6 animals) were euthanized through an overdose of sodium pentobarbital (300 mg/kg, Euthasol) via IP injection, followed by transcardial perfusion with 0.1 M PBS, then 4% paraformaldehyde (PFA) for tissue fixation. Whole eyes were extracted, fixed in 4% PFA for 30 min, and then cryoprotected in 30% sucrose with 0.02% sodium azide overnight. Lenses were removed from the globes, then the globes were embedded in OCT and frozen in liquid nitrogen-cooled isopentane. Retinas were sectioned using a cryostat at 10 μm and then used for immunohistochemistry. The remaining mice from each group (8–10) were euthanized via carbon dioxide, followed by cervical dislocation. Eyes were removed by forceps and then dissected. The retina was removed, placed in a tube, and then flash frozen in liquid nitrogen until downstream mRNA and protein analysis.

During the first day of immunohistochemistry, tissues were washed using 0.1 M PBS, incubated in blocking solution (5% donkey serum, 0.5% Triton X-100 in 0.1 M PBS) for 1 h, then incubated in primary antibody diluted with the blocking solution at 4°C for 24 h. On day 2, tissues were washed again using 0.1 M PBS, incubated in secondary antibody diluted by blocking solution for 2 h at room temperature, covered to prevent light exposure, washed again, and cover-slipped using DAPI Fluoromount G (Southern Biotech) or Fluoromount G (Southern Biotech). Primary antibodies used are listed in Table 1. The secondary antibodies used were anti-rabbit Alexa Fluor 647 (1:250, 711–605-152, Jackson ImmunoResearch), anti-Chicken Alexa Fluor 647 (1:250, 703–605-155, Jackson ImmunoResearch), anti-Rabbit Dylight 405 (1:250, 711–475, Jackson ImmunoResearch), and anti-Mouse Alexa Fluor 647 (1:250, 715–605-150, Jackson ImmunoResearch). Secondary antibodies were raised in donkeys against the species appropriate to the primary antibody.

Images for RGC and Müller glia mitolysosome quantification were taken at 63× magnification on an inverted Leica DMi8 Confocal microscope. Two images were taken from each retinal section across six retinal sections per eye in each group (24 images total per mouse). For mitolysosome quantification with pimonidazole staining, images were taken at 63× magnification on a Zeiss LSM 880 confocal with 12 images total per eye from each group (24 images per mouse). Retinal sections immunolabeled with PGC-1α, BNIP3, or NIX and cover-slipped with Fluoromount G for labeling density analysis were also imaged on the Zeiss LSM 880 Confocal at 20× magnification with 12 images per eye in each group (24 images total per mouse). RGC quantification images were taken on a Leica DM4 upright fluorescent microscope using StereoInvestigator (MicroBrightfield Biosciences) at 20× magnification with two peripheral and two central images from each retinal section across a total of six retinal sections (24 images) for four eyes from each group. All representative mitolysosome images were taken at 63× magnification with 1.6× zoom on the same Zeiss confocal, then processed with deconvolution using Huygens software (Scientific Volume Imaging).

For immunolabeling analysis, we used an ImageJ macro that quantifies the pixels of a region of interest traced within a retinal section after thresholding for background that is established within each section (Dengler-Crish et al., 2014). RGCs were also quantified manually from retinal sections immunolabeled for RBPMS using regions of interest traced in ImageJ.

Retinal sections immunolabeled with RBPMS or vimentin and cover-slipped with DAPI were used to quantify mitolysosomes in the retinal ganglion cells and Müller glia. A Fiji-ImageJ macro called “mitoQC Counter” was used, with guidance from Montava-Garriga et al. (2020). This process involved splitting the color channels to allow for the selection of the channel representing Vimentin or RBPMS. After adjusting the brightness/contrast, a median filter with a radius of 5.0 pixels was applied. Next, the immunolabeled cells underwent thresholding before particle analysis for a set size of 0.01 infinity. Finally, once the channels were merged back into the appropriate colors, the mQC_counter plugin was selected with green channel assigned to 2, red channel assigned to 1, and ratio threshold set at 0.5. From there, the mitolysosome count was normalized to the cell size given by the plugin.

Flash frozen retinas were homogenized into protein by adding 100 μL of T-PER (Tissue Protein Extraction Reagent, Thermo Fisher Scientific Product # 78510) with 1 μL of 100× Halt Protease & Phosphatase Inhibitor (Thermo Fisher Scientific, #1861280), then pulling through a 21G needle to shred the tissue before centrifugation for 15 min at 10,000 g at 4°C. The supernatant was collected into a new tube and used for protein analysis. Total retinal protein was determined through the use of Pierce^™ BCA Protein Assay Kit (Thermo Fisher Scientific REF# 23225) with absorbance read using a Cytation 5 (Biotek) plate reader to allow data to be normalized by total protein level in each retinal sample. Protein analysis was performed via capillary-based electrophoresis immunoassay using the Protein Simple Jess instrument (Biotechne). Proteins analyzed and antibody concentrations are listed in Table 1 with CE designation.

mRNA was extracted from flash frozen retinas using a Qiagen kit (CAT# 74134). The isolated mRNA was then transformed into cDNA using the Verso cDNA Synthesis Kit (Thermo Fisher Scientific, AB-1453/B) for stable storage and analyzed using real-time quantitative qPCR with a QuantStudio 5 real-time PCR System instrument (Applied Biosystems). TaqMan assay probes utilized in mRNA analysis included Ppargc1a, Cat. No. Mm01208835_ml, and Actb, Cat. No. Mm01205647_g1 (Thermofisher). β-actin was used as the housekeeping gene.

Statistical analysis was performed with GraphPad Prism version 10 (La Jolla, CA, United States). Tests of normality were employed to inform downstream parametric or non-parametric analyses. For parametric comparisons across groups, one-way ANOVA or two-way ANOVA and Tukey’s multiple comparisons test were used. For non-parametric data, Brown-Forsythe and Welch ANOVA, or mixed effects, and Dunnett’s multiple comparisons tests were used. Only p < 0.05 was considered significant. The data are presented as mean ± SD.

Mice were placed on the ketogenic diet 1 week prior to induction of ocular hypertension (OHT). Our intention was to acclimate them to the diet and only use those mice who were freely eating it. We found that all MitoQC mice in the ketogenic diet groups consumed the diet; Figure 1A shows the daily average ingestion of the diet in grams charted with the blood concentration of ketones (β-hydroxybutyrate). The average ketogenic diet ingestion was 1.82 ± 64 g per day, much lower than the average amount of standard lab chow (4.19 ± 0.38 g). The high caloric density of the ketogenic diet allowed mice to eat less yet still meet or exceed the kilocalorie per day average for the mice on standard chow (Figure 1B). Mice on the ketogenic diet alone ingested significantly more calories than either the Control (p = 0.0028) or Control + OHT (p = 0.0061) groups. We also monitored weight gain or loss on the standard lab chow (control) and the ketogenic diet (Figure 1C), finding no significant differences across sex or group. Over the 5 weeks monitored, male mice on the control diet gained an average of 0.17 ± 0.15 g, while females on the control diet gained 0.046 ± 0.06 g. Male mice on the ketogenic diet gained an average of 0.23 ± 0.16 g, while females gained 0.12 ± 0.09 g. There were no significant weight gains or losses across groups.

Intraocular pressure (IOP) was measured for all groups, with mice subjected to OHT showing significantly greater IOP than control groups (p < 0.0001), beginning at 7 days and continuing to the end of the experiment at 28 days after OHT induction (Figure 2A). We calculated the IOP integral, the magnitude of the IOP injury, and found that the OHT groups had a significantly higher IOP integral than the Control or Ketogenic Diet (KD) groups (p < 0.0001). There was no difference in IOP integral for the Control-OHT group compared to the KD-OHT group (Figure 2B).

Retinal ganglion cell number in the central retina did not differ across the groups (Figure 3A) for mice subjected to OHT. In the peripheral retina, however, the Control + OHT group had significantly fewer RGCs than the Control (p = 0.022), the KD (p = 0.045), and the KD + OHT groups (p = 0.018), as shown in Figure 3B. The significantly higher numbers of peripheral RGCs in the KD + OHT group versus Control are consistent with our previous study that showed the neuroprotective effect of the ketogenic diet in the DBA/2 J model of glaucoma (Harun-Or-Rashid et al., 2018). Examples of sectioned retina from each group immunolabeled for the RGC marker RBPMS (green) taken from both the central (left) and peripheral (right) retina are shown in Figure 3C.

Using the ImageJ macro mitoQC_Counter (see Methods), we quantified the mitolysosomes (red puncta) in the RGCs from retinal sections that had been immunolabeled with RBPMS for RGCs. Only the mitolysosomes appearing within RGCs/colocalized with RBPMS were quantified. Figure 4A shows that OHT significantly increased the appearance of mitolysosomes in the RGCs (p = 0.0023) of Control + OHT retina compared to Control. The ketogenic diet (KD) also significantly increased mitolysosome number in the RGCs (p < 0.0001) compared to Control. Ocular hypertension did not have an additive effect on the increased mitolysosomes from the KD since the KD + OHT group mitolysosome number did not differ from KD alone. Both the KD and KD + OHT groups had significantly more mitolysosomes than the Control + OHT group (p < 0.0001 for KD and p = 0.0002 for KD + OHT), suggesting that the KD was a more powerful stimulus than OHT for mitolysosome generation in RGCs. Images of mitochondria (green + red fluorescence) and mitolysosomes (red fluorescence) in retinal sections immunolabeled for RGCs (blue) are shown in Figure 4B. Green and red puncta (mitochondria) are evident in the Control group RGCs, but the green fluorescence in the OHT and KD groups is largely confined to the Müller glial processes surrounding the RGCs (Figure 4B, see insets). RGCs in the Control + OHT, KD, and KD + OHT groups are filled with mitolysosomes (red puncta).

Mitolysosomes quantified in Müller glia showed that OHT and the KD had roughly the same impact on mitolysosome generation since the number of mitolysosomes in the Control + OHT, the KD, and the KD + OHT groups did not differ (Figure 5A). All three groups had significantly greater numbers of mitolysosomes than the Control group (p = 0.0036 for Control + OHT, p < 0.0001 for KD, and p = 0.0022 for KD + OHT). Figure 5B shows MitoQC mouse retinal sections from each of the groups also immunolabeled with vimentin (blue) for Müller glia. All groups show mitochondria (green and red puncta) throughout the Müller glia, with increased expansion of Müller glial basal processes after OHT (see insets for Control + OHT and KD + OHT immunofluorescence, Figure 5B), and greater numbers of mitolysosomes (red puncta) in the Control + OHT, KD, and KD + OHT groups.

Oxygen tension across retinal layers can vary based on distance from vasculature or the choroid, as well as from circumstances such as OHT (Casson et al., 2021). Previous studies have demonstrated that the IOP elevations experienced by mice exposed to the magnetic microbead model of OHT (Jassim and Inman, 2019; Jassim et al., 2022) or the DBA/2 J model of glaucoma (Jassim et al., 2021) can result in hypoxia in retinal neurons and glia. To test whether low oxygen tension experienced by retinal cells can stimulate mitophagy, we injected Naive mice or those exposed to OHT (OHT) with pimonidazole, a compound that creates covalent adducts in cells that experience a partial pressure of oxygen lower than 10 mmHg (hypoxia). Our positive control was the retina that had been exposed to 40 mmHg of IOP for 30 min. We then evaluated mitolysosome generation in retinal cells or RBPMS-positive RGCs that were pimonidazole-positive. We observed a significant increase in mitolysosomes in pimonidazole (Pimo)-positive cells throughout retina exposed to OHT (Naive vs. OHT, p < 0.0001; Figure 6A). Mitolysosome number in the OHT retina did not differ from the Positive Control. Focusing the analysis on RGCs, we observed a significant increase in mitolysosomes in the Pimo-positive and RBPMS-positive RGCs in the OHT group versus Naive (p < 0.0001, Figure 6B). Figure 6C shows Pimo labeling in the ganglion cell layer of the Positive Control and the OHT retinal sections. The Naive group shows no Pimo (blue) since no RGCs were Pimo-positive. These data indicate that Pimo-positive RGCs have a significantly higher mitolysosome number than Naive RGCs, suggesting that hypoxia is a stimulus for mitophagy in the retina.

Next, we endeavored to understand if the promotion of mitophagy would be balanced with a concomitant increase in mitochondrial biogenesis, as measured by transcriptional transactivator and mitochondrial biogenesis regulator PGC-1α. In whole retinal lysates, we measured PGC-1α protein (Figure 7A), finding no statistical differences across groups. Using mRNA isolated from the whole retina, we undertook quantitative RT-PCR for PGC-1α transcripts, also finding no change across the Control, Control + OHT, KD, or KD + OHT groups (Figure 7B). Recognizing that our mitolysosome analysis had focused on RGCs and Müller glia, not the whole retina, we immunolabeled the retina using an antibody against PGC-1α and colabeled with RBPMS for RGCs. An optical density analysis of PGC-1α in RBPMS-positive RGCs showed no significant differences across groups (Figure 7C). Figure 7D shows examples of immunolabeling for PGC-1α (red) and RBPMS (green) in each control and experimental group. Colocalization of PGC-1α with RBPMS was rare, with just one incidence shown in Figure 7D, in the section from the KD group (white arrow).

Having observed that hypoxia can be a stimulus for mitophagy in our model of OHT, we tested whether the BNIP3-NIX pathway was activated in RGCs where we had observed significant increases in mitolysosome density. BNIP3 is a member of the BH3-only subfamily whose expression is promoted by HIF-1 transcriptional activity; it localizes to the outer mitochondrial membrane after cellular stress (Mellor and Harris, 2007). BNIP3 and NIX act as mitophagy receptors that can tether the isolation membrane to mitochondria (Yamashita et al., 2024). Our analysis of whole retinal lysate yielded no changes in BNIP3, BNIP3 dimer, or NIX protein levels (Figures 8A,B,F). In order to refine our analysis, we immunolabeled the retina with BNIP3 or NIX and then measured the fraction of labeled tissue across our control and ketogenic diet groups, with or without OHT. We also observed no significant differences across groups when the immunolabeling analysis was restricted to the ganglion cell layer (Figures 8C,G). However, BNIP immunolabel (Figure 8D) was prominent in the RGCs, as denoted by asterisks. Many, but not all, RGCs contained BNIP3 immunolabel. For NIX immunolabeling (Figure 8E), colocalization with RGCs was observed in the Control, the KD, and the KD + OHT groups, but not in Control + OHT (asterisks).

Mitolysosome count was increased in the KD group compared to the Control in both RGCs and Müller glia, providing evidence that the ketogenic diet can stimulate mitophagy. In our previous study using the DBA/2 J mouse model of glaucoma, we showed that forcing cells to rely on ketones to produce cellular energy resulted in upregulation of PGC-1α and mitochondrial biogenesis (Harun-Or-Rashid et al., 2018). We theorized that the homeostatic balance of mitochondria would require increased mitophagy to accompany this upregulation in mitochondrial biogenesis. In skeletal muscle, SIRT1 is thought to promote mitochondrial biogenesis through the deacetylation of PGC-1α in fasting conditions and promote mitophagy by targeting mitophagic machinery (Palikaras et al., 2015). AMPK has also been shown to promote biogenesis through the regulation of PGC-1α (Palikaras et al., 2015), while mitophagy is upregulated by the enhancement of mitochondrial fission and TBK1-activated autophagosome engulfment in a PINK1-Parkin-independent pathway (Seabright et al., 2020). The ketogenic diet could play a role in SIRT1 activity through increased NAD⁺, which is a required substrate for SIRT1 (Maalouf et al., 2007; McCarty et al., 2015). There is also a possible pathway in which SIRT1 can promote AMPK activation through TBK1 (Lan et al., 2008). In addition to these effects, the ketogenic diet may be promoting mitophagy through upregulation of HIF-1α despite normoxic conditions, as exhibited in investigations of rat brains on this diet, therefore leading to transcription of BNIP3 and BNIP3L (NIX) (Puchowicz et al., 2008; Xu et al., 2010).

Our results indicate that glaucomatous conditions can increase mitophagy, as demonstrated by significant increases in mitolysosomes in the Control + OHT mice compared to our control (naïve) group. This is an interesting find because there have been several studies proposing impaired mitophagy as a contributor to glaucomatous pathology and an avenue for therapeutic intervention (Dai et al., 2018; Hu et al., 2018; Hass and Barnstable, 2019). One study elucidated a link between mutations in optineurin with pathology of primary open-angle glaucoma, emphasizing evidence of impaired mitophagy leading to the accumulation of damaged mitochondria (Shim et al., 2016). In DBA/2 J mice, increased levels of mitochondria with altered morphology and diminished capacity for oxidative metabolism, together with decreased LAMP1, despite increased numbers of autophagosomes, led to the belief that these mice had impaired mitophagy (Coughlin et al., 2015). In the optic nerves of ocular hypertensive rats, mitophagosome numbers increased 3 days after elevated IOP, but within 2 weeks, these numbers decreased, leading to the belief that mitophagy may be upregulated to combat stress, but with chronic elevation, mitophagic mechanisms fail and may become impaired (Dai et al., 2018). Our analysis was of one moment in time, 4 weeks after OHT, where we observed significant increases in mitophagy over control. There can be a number of possible explanations for these findings of increased mitophagy when other investigations have data suggesting the opposite. For one, we used the MitoQC mouse, an animal enabling strong, direct in vivo evidence of mitophagy—mitochondria have green and red fluorescence that turns to red fluorescence alone when the mitochondrion is engulfed in a lysosome (McWilliams et al., 2016). We evaluated mitophagy 4 weeks after the commencement of OHT, which may represent an achievement of mitochondrial homeostasis that did not exist at other times post-OHT. Many of these other studies, including one of our own (Coughlin et al., 2015), provided only circumstantial evidence of mitophagy. Additionally, other studies showed that promotion of mitophagy was neuroprotective (Dai et al., 2018; Hu et al., 2018; Hass and Barnstable, 2019), which does not rule out the existence of an already increased mitophagy level; it would just demonstrate that more mitophagy under those degenerative conditions was also beneficial. A case in point is our original ketogenic diet study, in which we showed RGC structural and functional protection with the diet (Harun-Or-Rashid et al., 2018). In the current study, we demonstrated that the ketogenic diet had an additive effect on mitolysosome production, suggesting that increasing mitophagy beyond what was promoted with OHT contributed to better outcomes for the RGCs. Without further investigation into the quality of the unrecycled mitochondria, we cannot know for certain whether mitophagy is fully functional or impaired within the glaucomatous retinas of our study; however, in alignment with the above-mentioned studies, we do show increased mitolysosomes, indicative of increased mitophagy.

Müller glia are the predominant glial cell type in the retina and are oriented radially to allow interaction across the entire retina (Vecino et al., 2016). These cells play a crucial role in retinal homeostasis by providing support for ion and water homeostasis, maintenance of the blood-retinal barrier, release of neurotrophic factors for survival, maintenance of retinal structural stability, provision of metabolic substrates, and more (Toft-Kehler et al., 2018). One suggestion of the metabolic substrate these cells can provide comes from the lactate shuttle hypothesis, which proposes that glial cells like astrocytes and Müller glia can uptake glucose and convert it to lactate to provide a metabolic substrate to RGCs (Hurley et al., 2015). Müller glia have a large concentration of mitochondria in their basal endfeet that are in close contact with RGCs (Choi et al., 2020). In light of the metabolic support these cells can provide to RGCs, we wanted to examine whether the ketogenic diet would impact mitophagy in these cells. The ketogenic diet and elevated IOP produced similarly increased mitophagy in Müller glia, indicating that the ketogenic diet was not as additive for promoting mitophagy as it was in the RGCs. Nevertheless, these results provided evidence to support that the diet does not have the same effect on mitochondrial homeostasis in RGCs and Müller glia. Müller glia show preference for glycolysis despite aerobic or anaerobic conditions (Winkler et al., 2000; Toft-Kehler et al., 2018). There is evidence that during inhibition of glycolysis, through loss of exogenous glucose, Müller glia rely on mitochondrial respiration and that simultaneous deprivation of oxygen and glucose is required to impact ATP production in these cells (Winkler et al., 2000). In addition to the flexibility these cells have for metabolic shift to maintain ATP production despite environmental conditions, there is evidence of Müller glia storing and metabolizing glycogen (PérezLeón et al., 2013). This provides a mechanism by which these cells maintain a backup supply of metabolic energy sources that can be utilized in low glucose conditions (Rombaut et al., 2023). Due to this, it is thought that Müller glia are less sensitive to acute metabolic deficits than RGCs (Silver et al., 1997). Considering the numerous mechanisms that Müller glia can utilize to maintain metabolic support, it appears that glaucomatous conditions are a sufficient stimulus to upregulate mitochondrial homeostatic processes, diminishing the potential for the ketogenic diet to induce metabolic shifts.

In contrast, within RGCs, the ketogenic diet promoted mitophagy to a greater degree than elevated IOP alone, as evidenced by a significant increase in mitolysosome number within the KD and KD + OHT groups compared to Control + OHT. Due to their sensitivity to metabolic deficits (Inman and Harun-Or-Rashid, 2017; Casson et al., 2021), RGCs may be more inclined to upregulate mitophagy via the ketogenic diet so that they may sustain mitochondrial homeostasis. The ketogenic diet does require RGCs to utilize ketone bodies for fuel, which obligates mitochondrial respiration and the attendant ROS generation that oxidative phosphorylation entails. The additional demand of the diet on the mitochondria may be the stimulus that promotes mitophagy beyond that initiated by OHT.

Through the ischemia–reperfusion experiments, we demonstrated that hypoxia provides a stimulus for mitophagy within the early stages of glaucomatous pathology (24 h after induction of elevated IOP). This was seen through increased mitolysosome number in mice with OHT (OHT) compared to our Naive (naïve) mice. In addition to this, our positive control mice showed statistically similar results as the OHT mice, suggesting that the additional IOP elevation (a difference of ~20 mmHg) did not further promote mitophagy. There could be a threshold of mitochondrial stress at which mitophagy is engaged, but at levels that can manage a wide range of elevated IOP and attendant hypoxia. As discussed above, there is evidence that increased mitophagy in early glaucomatous pathology could be a mechanism to adjust to stress produced by elevated IOP. Furthermore, hypoxia may be the stimulus leading to upregulation of mitophagy in the Control + OHT groups in the ketogenic diet experiments. Our pimonidazole experiment measured hypoxia-induced upregulation of mitophagy after 24 h in contrast to 4 weeks after OHT, yet we have evidence that hypoxic changes are likely a factor in longer-term glaucoma pathology. Pseudohypoxia, a phenomenon originally observed in diabetic conditions, refers to a metabolic state resembling hypoxia despite normal levels of oxygen (Williamson et al., 1993). A previous study revealed a link between pseudohypoxia and disrupted mitochondrial homeostasis in DBA/2 J mice (Jassim et al., 2021). In a more recent study, we confirmed chronic HIF-1α stabilization in DBA/2 J mice could induce a pseudohypoxic state and, by doing so, promote hypoxia-induced mitophagy through elevated protein levels of BNIP3 (Nsiah et al., 2023). Previously, our lab demonstrated hypoxia and increased Hif-1a transcripts in the retina 4 weeks after OHT, accompanied by the induction of autophagy (Jassim and Inman, 2019).

Interestingly, prominent pimonidazole staining seemed to be concentrated in the retinal ganglion cell layer and inner nuclear layer of the retinal sections. Our previous investigations have shown evidence of hypoxia in both RGCs and Müller glia after 4 weeks of OHT through pimonidazole localization (Jassim and Inman, 2019). Using a transgenic fluorescent reporter approach, we also observed HIF-1α stabilization after OHT in RGCs and Müller glia, with the highest reporter tdTomato labeling appearing at 6 h and 3 days (Jassim et al., 2022). The same study revealed Müller glia showed the most prominent tdTomato labeling (Jassim et al., 2022). Our relative lack of Müller glia labeling 24-h post-OHT likely reflects the lesser sensitivity of pimonidazole to HIF-1α stabilization/hypoxia as compared to the genetically encoded reporter construct (Butt et al., 2021; Jassim et al., 2022).

For potential confirmation of mitochondrial biogenesis balanced with our evidence of increased mitophagy through the ketogenic diet, we analyzed the protein and mRNA of PGC-1α within the whole retina lysate, which showed no significant difference among the four groups. Since our mitolysosome quantification was focused within immunolabeled RGCs or Müller glia, we decided to analyze the localization of PGC-1α immunohistochemical staining within the inner retinal layer, but these results also showed no significant difference. These findings contrast with our observations in a previous study in the DBA/2 J mouse model of glaucoma, where we observed that the ketogenic diet afforded neuroprotection of RGCs concomitant with PGC-1α upregulation (Harun-Or-Rashid et al., 2018). The discrepancy could be a result of the duration of increased IOP (4 weeks in the current study, while 4+ months in the DBA/2 J), the length of time on the ketogenic diet (5 weeks vs. 8 weeks), or even mouse strain (C57Bl/6 J vs. DBA/2 J). Observing significant mitophagy at 4 weeks after OHT yet unchanged PGC-1α, as we do here, suggests there may be a means by which a cell could maintain mitochondrial number without biogenesis. Analyses of mitochondria in the DBA/2 J mouse have indicated that RGC axons have steady-state numbers of mitochondria, but that those organelles are significantly smaller with age and increased IOP (Coughlin et al., 2015). In fact, glaucoma significantly reduces the volume of mitochondria per axon in the RGC axon (Kleesattel et al., 2015). Fission is one way to increase mitochondrial number, and increased fission has been observed in the DBA/2 J (Ju et al., 2008). In cancer cells, upregulation of mitochondrial dynamics can maintain mitochondrial homeostasis (Towers et al., 2021). Mitochondria undergo fission prior to mitophagy in order to be small enough for engulfment, so there is a possibility that the mitochondria undergoing fission may ultimately be destined for mitophagy. Limits to the amount of mitophagy that can take place at any one time may create a pool of smaller mitochondria with indeterminate functional capability. Little is understood about the feedback mechanisms for mitochondrial dynamics that inform biogenesis. Additional studies using the MitoQC mouse at both shorter and longer time frames than the current study, as well as an investigation of mitochondrial dynamics, should provide insight into these possibilities.

To assess whether hypoxia promoted mitophagy in the model of glaucoma used here and 4 weeks after OHT, we evaluated the BNIP3-NIX pathway of receptor-mediated mitophagy. Protein changes in BNIP3 and NIX (BNIP3L) in whole retinal lysate or through immunolabeling of the inner retinal layer resulted in no global quantitative differences in BNIP3 or NIX, nor when analysis focused on the RGCs. This argues against receptor-mediated mitophagy with the ketogenic diet or after 4 weeks of elevated IOP from the microbead model. BNIP3 is a target of HIF-1α, and its phosphorylation mediates mitophagy by enabling direct binding of BNIP3 to LC3 (He et al., 2022). However, we did find significant colocalization of BNIP3 in the RGCs across all groups, and we also observed NIX colocalization in each group except the Control + OHT retina. We did not examine BNIP3 phosphorylation state nor c-Jun N-terminal kinase 1/2, the kinase that targets BNIP3, which could provide insight and potential confirmation of our results in a future investigation. In our previous study that stabilized HIF-1α in the retina for 4 weeks, we did observe a significant increase in both BNIP3 and NIX (Nsiah et al., 2023). One potential explanation for the discrepancy here is that our model of OHT results in sporadic HIF-1α and hypoxia (Jassim and Inman, 2019), while HIF-1α stabilized using Roxadustat was a much more powerful stimulus for BNIP3 and NIX expression (Nsiah et al., 2023). These data imply that context likely dictates the mechanism of mitophagy, and that it can vary. The ketogenic diet might have also played a role in specifying the exact mechanism of mitophagy shown here. Alternatively, the lack of an apparent role for BNIP3 and NIX may indicate a mitochondrial quality control mechanism in play that is not mitophagy at all. For example, mitochondrial-derived vesicles, which contain mitochondrial membrane and could thus be studded with GFP/mCherry fluorescence in our mice, are released by mitochondria and carry oxidized protein to lysosomes (Neuspiel et al., 2008; Soubannier et al., 2012). Once engulfed in a lysosome, these mitochondrial-derived vesicles (MDVs) would appear as red puncta and thus be indistinguishable from a recycled mitochondrion. Ruling out MDVs as contributors to the mitolysosomes we observe would require future electron microscopic analysis.