Male C57BL/KsJ-db/db mice, obtained from The Chinese University of Hong Kong, were housed in the Central Animal Facility of The Hong Kong Polytechnic University. The mice were housed in groups of two in plastic cages with free access to standard rodent diet and water under a 12 h/12 h light-dark cycle. The experimental procedures were performed according to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and were approved by the Animal Subjects Ethics Sub-committee (ASESC) of The Hong Kong Polytechnic University (approval no.: 19-20/85-SO-R-STUDENT).

Fasting blood glucose levels were measured monthly from baseline to 4-month after commencing the experiment with a digital blood glucometer (Accu-Chek^® Performa, Roche Diagnostics, Basel, Switzerland). The hemoglobin A1c (HbA1c) levels, which reflect the diabetic status over the prior three to four months, were measured at baseline and at 4-month after commencing CoQ10 treatment using DCA^® Vantage Systems (Siemens Medical Solutions Diagnostics, New York, NY, USA). Blood samples were collected from the tail veins after overnight fasting. All assessments were conducted between 9 am to 10 am. Bodyweight was monitored weekly. Food and water consumption were measured thrice weekly.

Mice with a fasting blood glucose level ≥ 13.9 mmol/L at 9-week of age were included in the study. The diabetic mice were divided into two groups, group 1 mice (CoQ10 treated) received one drop (approximately 30 μl) of CoQun^® (0.1% CoQ10 and 0.5% vitamin-E TPGS in buffered isotonic solution, VISUfarma, Roma, Italy) eyedrops topically twice a day for four months, while group 2 mice (db/db control) received the same amount of phosphate-buffered saline (PBS). Following the eyedrops application, the animals were kept restrained for about 20 s to allow absorption of the eyedrops. The remaining volume was gently swiped off at the canthus.

The fasting blood glucose levels, HbA1c levels, body weight (Supplementary Figure 1), and total consumption of food and water (Supplementary Table 1) of the db/db mice in control and treatment groups did not differ statistically throughout the experiment, implying CoQ10 eyedrop treatment had a minimal effect on the glycemic status of db/db mice.

The in vivo retinal functions were assessed with full field ERGs (ffERGs) at baseline, 1-, 2-, and 4-month after CoQ10 treatment as described previously (Lam et al., 2023). For scotopic ERG, the mice were dark-adapted overnight (at least 16 h) prior to testing in the experimental room. The animals were anesthetized with a weight-based intraperitoneal injection of Ketamine 100 mg/ml (Alfasan International BV, Woerden, Holland) and Xylazine 20 mg/ml (Alfasan International BV). The corneas of both eyes were anesthetized with Proxymetacaine 0.5% (Provain-POS^®, Ursapharm, Saarbrücken, Germany) and maintained moist by a drop of 3% Carbomer 974P gel (Lacryvisc^®, Alcon, France). The pupils were dilated using a drop of solution mixed with Tropicamide 0.5% and Phenylephrine 0.5% (Mydrin^®-P, Santen, Osaka, Japan). The animals were then placed on a warming table maintained at 37°C. A gold ring electrode was placed in contact with the cornea as the active electrode. Two platinum needle electrodes were inserted subcutaneously at the base of the tail and the forehead as the ground and reference electrodes, respectively. The impedance of the active and reference electrodes was less than 10 kΩ.

Visual stimuli with white light-emitting diodes were delivered by a Ganzfeld bowl (Q450, Roland Consult, Brandenburg, Germany). Stimulation and data recording were performed using the RETI-Port system^® (Roland Consult) according to a customized protocol, with stimulus intensities varying from −4.32 log cd s/ m² to +1.3 log cd s/ m². The signals were amplified and band-pass filtered from 1 to 30 Hz and 1 to 1,000 Hz for scotopic threshold response (STR) and scotopic a- and b-wave, respectively. At the lowest intensity, 40 sweeps of response were averaged with a stimulus frequency of 0.5 Hz. The number of sweeps and stimulus frequency was reduced at higher flash intensity levels. The stimulus intensities were calibrated by a photometer (ILT1700, International Light, MA) and converted to unit photoisomerizations/rod (R*/rod), where 1 scot cd/ m² = 516 R*/rod/s. The stimulus responses within operational range of the rod cells (< −0.3 log cd s / m²) were fitted with a sigmoidal curve using the Naka-Ruston equation, a saturating hyperbolic function that describes the relationship between b-wave amplitude and flash intensity, to determine the maximum b-wave response (Bmax).

Oscillatory potentials (OPs) were isolated by digital filtering of the raw signal recorded at +1.3 log cd s/m² using a fast Fourier transform (FFT) and subsequent inverse FFTs with an algorithm computed in the free software environment R (v4.0.4). The raw data was first converted from the time domain to the frequency domain. Spectral components beyond the cut-off frequency (65 Hz to 300 Hz) were then eliminated. The inverse FFT was performed to reconstruct the OP waveform in the time domain. Implicit times of OP were measured from the stimulus onset to the peak of the OP. Individual OP amplitudes were measured from the peak to the adjacent trough. The first four major OP wavelets were included for analysis.

Photopic ERG was conducted in a subset of animals on a separate day (without dark adaptation before the test) using the same system. The animals were presented to a uniform white background of 30 cd/ m², and white light-emitting diodes flash (+0.47 log cd s/m²) was used as the stimulus. The signals were amplified and band-pass filtered from 1 to 1,000 Hz, and 40 sweeps of response were averaged with a stimulus frequency of 0.5 Hz. The amplitude and implicit time of b-wave were used for analysis.

At the end of the study period, one eye from each animal (randomly chosen) was enucleated and dissected to isolate the whole retina, which was then fixed in 4% paraformaldehyde in PBS at room temperature for one hour. Part of the harvested retinas (whole-mount) were used to quantify microglial density, while the remainder were processed into vertical retinal sections (35 μm) using a microtome (Vibratome VT1200S, Leica Microsystems, Wetzlar, Germany) as previously described (Tse et al., 2015). These samples were then blocked with 10% donkey serum in TBS (0.5% Triton X-100 and 0.1% Sodium Azide in Dulbecco’s phosphate-buffered saline, pH 7.2) at 4°C overnight to reduce nonspecific labeling.

The samples were then incubated with primary antibodies at 4°C in TBS with 3% donkey serum for four days. Following incubation, the samples were washed several times and transferred to a 3% normal donkey serum-TBS solution containing donkey-host secondary antibodies at 4°C overnight. DAPI (Invitrogen, D1306) was used to stain the nuclei in the retina. The details of antibodies used are listed in Supplementary Table 2.

A confocal laser scanning microscope (LSM800, Zeiss, Oberkochen, Germany) was used to capture confocal micrographs of the specimens using a 20×, 40×, or 63× objective. To determine the number of photoreceptor cells (including both rods and cones), the DAPI stained nuclei in the outer nuclear layer were counted over an 80 μm-segment with a cell counter in ImageJ (v1.53k, National Institutes of Health, Bethesda, MD, USA). To measure the number of cone photoreceptors and rod bipolar cells (and their synaptic terminals), the G Protein Subunit Alpha Transducin 2 (GNAT2) and Protein kinase C-α (PKCα) positive cells were counted over a 160 μm-segment. The number of rod-bipolar dendritic boutons was determined by counting the number of PKCα positive puncta at the outer plexiform layer over a 20 μm-segment. The dimensions of the outer nuclear layer and the inner nuclear layer were analyzed by ImageJ. The immunoreactivity of the vertical retinal sections against glial fibrillary acidic protein (GFAP) antibody was used to reflect the levels of glial reactivity in the retina, which was quantified based on the extent of GFAP staining using a scoring system described previously (Anderson et al., 2008; Bogdanov et al., 2014). The data of each individual animal was averaged from three retinal sections, with each group consisting of at least five animals. For quantification of microglia, the flat retinas were processed with ionized calcium binding adaptor molecule-1 (iba-1) antibody. Four regions of each flat retina around one disk diameter from the optic nerve head and two retinal regions (i.e., ganglion cell layer to inner plexiform layer and outer plexiform layer) were imaged with a 20× objective. The number of microglia (iba-1 positive cells) was counted with the cell counter in ImageJ. The data of each individual animal was averaged from the four regions per retina, each group consisting of at least four animals.

A semi-quantitative analysis approach was used to assess the MMP-9 levels in the retina of the mice in the two groups due to the limited number of samples from this longitudinal study. The MMP-9 levels were evaluated by quantifying the fluorescence intensity levels in the retinal sections after immunostaining with ImageJ. The images were first converted to gray scale. The mean fluorescence was measured across the whole retinal section imaged, and the value subtracted from that of the background. The data of each individual animal was averaged from three retinal sections, each group consisting of at least five animals. To ensure the specificity of the antibody, retinal sections of a wildtype mouse were processed using the same procedure, but omitting the primary antibody. Only trace fluorescence signals were detected in the negative control confocal images.

The mitochondrial bioenergetics of the retinas of the db/db mice in both groups were assessed by measuring the oxygen consumption rate (OCR) with a Seahorse XFe24 Extracellular Flux Analyzer (Agilent Technologies, Santa Clara, CA, USA). Four mice were chosen randomly from each of the db/db treated and db/db control groups. One eye from each animal was enucleated and dissected to harvest the whole-mount retina in ice-cold PBS. Three 1.5 mm diameter punches were obtained from the neural retina using a biopsy puncher (Miltes Instrument, Integra LifeSciences, Mansfield, MA, USA). The retinal punches were obtained adjacent to the optic nerve head to minimize any discrepancy in cell density. Each retinal punch was then carefully placed in the center of the well of an XF24 Islet capture microplate (Agilent Technologies), with the ganglion cell layer facing up and then covered with Islet Fluxpak mesh inserts. Prior to the measurement of OCR, Seahorse XF DMEM medium (103335-100, Agilent Technologies) containing 5.5 mM glucose (G6152, Sigma-Aldrich) and 1 mM sodium pyruvate (11360070, Gibco) was added to each well and the retinal punches incubated at 37°C in a non-CO² incubator for one hour to allow the temperature and pH to reach equilibrium. The OCR was then measured under basal conditions and after serial injection of 1 μM carbonyl cyanide-4-(trifluoro-methoxy) phenylhydrazone (FCCP; 15218, Cayman), and a mixture of 10 μM rotenone (13995, Cayman) and 20 μM antimycin A (A8674, Sigma-Aldrich, St. Louis, MO, USA) to determine values for basal respiration, maximal respiration, spare respiratory capacity, and non-mitochondrial oxygen consumption. Following the assays, the retinal punches in each well were lysed with EB2 lysis buffer, containing 7 M urea, 2 M thiourea, 30 mM Tris, 2% (w/v) CHAPS, and 1% (w/v) ASB14 with protease inhibitor cocktail (Roche Applied Science, Basel, Switzerland), and placed on ice. The protein concentrations were determined using the Bradford Protein Assay (Bio-Rad Laboratories Inc., St. Louis, MO, USA) according to the manufacturer’s guidelines for normalization of the OCR values. The values were then normalized to those of the control mice.

Data are presented as mean ± SEM. Shapiro–Wilk test was used to ensure the data was normally distributed prior to the use of a parametric test. For longitudinal comparison, mixed-model ANOVA was used to analyze the within-group, between-group, and interactive effects. Independent sample t-test (with Welch correction when appropriate) was used for comparison between the two groups. JASP (v0.13, Amsterdam, Netherlands) was used for the statistical analysis. Differences with a p-value < 0.05 were considered significant.

The effect of CoQ10 treatment on preserving the function of different retinal types against diabetes was assessed by ffERGs. The scotopic a-wave reflects the photoreceptor cell function, particularly that of rod cells. The amplitude and implicit time of scotopic a-wave were not statistically different between the control and CoQ10 treated groups at any of the experimental timepoints (Supplementary Figure 2). In contrast, the scotopic b-wave, an indicator of the functions of the rod cell and its post-receptoral pathway, was significantly stronger in the CoQ10 treated group than the control group at some of the tested stimulus intensities (Figure 1; see Supplementary Figure 3 for the scotopic b-wave implicit time).

The maximum rod-driven b-wave response amplitude (Bmax) was deduced by fitting the ERG responses within the rod operation range (i.e., < −0.3 log cd s / m²) with sigmoidal curves using the Naka-Ruston function (Figures 1A–D). Repeated measures ANOVA revealed a significant main effect of time and groups, as well as a significant interactive effect between time and groups (Repeated measures ANOVA, time: F = 5.304, df = 3, p = 0.003; group: F = 4.559, df = 1, p = 0.045; time × group: F = 3.630, df = 3, p = 0.018). At baseline, the difference in the Bmax value was not statistically significant between the two groups of mice (Figure 1E). After two months of CoQ10 treatment, the Bmax value of the CoQ10 treated db/db mice became significantly stronger than that of the control mice. This effect, however, was not sustained at 4-month.

Intriguingly, a significantly stronger b-wave response was observed in the CoQ10 treated group at +1.3 log cd s/m², in which the response was driven by both rod and cone cells, after two months of CoQ10 treatment compared to the control groups (2-month: db/db control = 468.75 ± 17.51 μV vs. CoQ10 treated = 611.79 ± 33.71 μV; Simple main effect analysis, p = 0.001; 4-month: db/db control = 454.02 ± 27.06 μV vs. CoQ10 treated = 541.11 ± 27.12 μV; Simple main effect analysis, p = 0.034). The representative ERG waveforms of the two groups of mice recorded at this intensity at 4-month are shown in Figure 1F. This suggests that the CoQ10 treatment may have preserved the functions of the cone photoreceptors and/ or their related postsynaptic pathway. In agreement, it was observed that CoQ10 treated db/db mice exhibited a significantly stronger photopic b-wave amplitude after receiving CoQ10 eyedrops for one month (Figure 2, Mixed model ANOVA, time: F = 3.093, df = 3, p = 0.046; group: F = 16.809, df = 1, p = 0.003; time × group: F = 1.986, df = 3, p = 0.143), though the implicit time was not significantly different (Supplementary Figure 4). The results further supported the protective effect of CoQ10 eyedrops on the cone-related pathway.

For the inner retinal function, the positive scotopic threshold response (pSTR) and oscillatory potentials (OPs), which are presumed to originate from the ganglion cells (Saszik et al., 2002) and amacrine cells (Wachtmeister, 1998), respectively, were analyzed. The representative pSTR waveforms of the two groups of mice at 4-month are shown in Figure 3A. The results revealed that the amplitude of pSTR was not statistically different between the two groups of mice (Figure 3B) but became significantly stronger in the CoQ10-treated group than the control group after 4 months of CoQ10 treatment, indicating preserved ganglion cell function (Figure 3C, please refer to Supplementary Figure 5 for the implicit time of pSTR). The OPs were obtained by digitally filtering the raw ERG signals recorded at +1.3 log cd s/m² (see Figure 3D for the representative OPs waveforms). At baseline, the control db/db mice were found to have a significantly stronger OP1 response amplitude (Figure 3E, db/db control = 48.72 ± 5.16 μV vs. CoQ10 treated = 36.17 ± 2.63 μV; Simple main effect analysis, p = 0.037). In contrast, the db/db mice in the treatment group exhibited stronger OP1 amplitude compared to their diabetic control littermates after 4 months of CoQ10 treatment (Figure 3F, db/db control = 35.13 ± 3.32 μV vs. CoQ10 treated = 48.14 ± 3.66 μV; Simple main effect analysis, p = 0.016). Implicit times of OP1-OP4 were not significantly different between the two groups of mice at baseline and 4-month (Supplementary Figure 6).

At the experimental endpoint, one eye from the animal was randomly chosen and processed for immunohistochemistry analysis as whole mount or microtome-sectioned retina. Results showed that CoQ10 treated db/db mice had a significantly thicker ONL (db/db control = 53.60 ± 2.03 μm vs. CoQ10 treated = 60.26 ± 1.96 μm; Independent sample t-test: t = −2.357, df = 11, p = 0.038) and INL (db/db control = 30.85 ± 1.55 μm vs. CoQ10 treated = 39.03 ± 0.58 μm; Independent sample t-test (Welch): t = −4.9553, df = 6.379, p = 0.002) compared to the db/db control mice (Figures 4A–C).

The photoreceptor cell density (DAPI-stained cells in theONL) of CoQ10 treated db/db mice was found to be significantlyhigher than that of the control db/db mice (db/db control = 316.42 ± 14.35 vs. CoQ10 treated = 380.81 ± 17.97; Independent sample t-test: t = −2.733, df = 11, p = 0.019). In addition, CoQ10 treated db/db mice had more cone soma (nucleus co-stained with GNAT2) then the control db/db mice (db/db control = 28.76 ± 4.10 vs. CoQ10 treated = 43.81 ± 2.09; Independent sample t-test: t = −3.422, df = 11, p = 0.006) (Figures 4D-F).

The rod-bipolar cells were examined by immunostaining with antibodies against PKCα (Figure 4G). Although the rod bipolar cell density of the CoQ10 treated db/db mice was higher than that of the control mice, the difference did not reach statistical significance (Figure 4H, db/db control = 38.01 ± 3.21 vs. CoQ10 treated = 46.60 ± 3.83; Independent sample t-test: t = −1.682, df = 11, p = 0.121). Previously, we reported a reduced number of rod-bipolar cell axon terminals and dendritic boutons in db/db mice, indicating that diabetes leads to a compromised synaptic connectivity of the second-order neurons with other retinal neurons (Lam et al., 2023). Based on these earlier findings, the number of rod-bipolar cell axon terminals and dendritic boutons in the db/db mice after CoQ10 treatment were examined. The number of rod-bipolar cell axon terminals tended to be higher in CoQ10 treated db/db mice than the control mice. However, the difference did not reach statistical significance (Figure 4I, db/db control = 55.75 ± 8.75 vs. CoQ10 treated = 90.93 ± 13.19; Independent sample t-test: t = −2.14, df = 11, p = 0.056). In contrast, the density of rod-bipolar cell dendritic boutons (Figure 4J) in the CoQ10 treated db/db mice was significantly higher than the db/db control mice (Figure 4K, db/db control = 14.69 ± 1.91 vs. CoQ10 treated = 41.85 ± 6.35; Independent sample t-test: t = −3.488, df = 10, p = 0.006), suggesting an ameliorated loss of synaptic connectivity, particularly, between photoreceptor and the second-order neurons.

Glial reactivity (gliosis) in the retina is associated with early neuronal impairment of diabetic retinopathy (Sorrentino et al., 2016). The effect of CoQ10 eyedrops conjugated with vitamin E TPGS on gliosis was investigated by processing the vertical retinal sections with GFAP antibody (Figures 4L, M), which is known to be expressed in astrocytes and reactive Müller cells (Hol and Pekny, 2015; Lee et al., 2021). The extent of gliosis was determined by a scoring system described previously (Anderson et al., 2008; Bogdanov et al., 2014). The results showed that CoQ10 eyedrops conjugated with vitamin E TPGS could significantly reduce gliosis in the retina of db/db mice (db/db control = 2.22 ± 0.34 vs. CoQ10 treated = 1.41 ± 0.16; Independent sample t-test: t = 2.302, df = 9, p = 0.047).

Microglia are the principal resident immune cell in the retina, playing an important role in immune surveillance and synapse maintenance (Wang et al., 2016). Increase in microglial density and changes in morphology have been implicated in a range of degenerative eye diseases (Fan et al., 2022). The distribution of the microglia at two retinal regions [i.e., ganglion cell layer to inner plexiform layer (GCL-IPL) and outter plexiform layer (OPL)] was studied. Results showed that retinas of db/db mice receiving CoQ10 eyedrops conjugated with vitamin E TPGS had reduced microglial density in the OPL (db/db control = 8.56 ± 0.87 vs. CoQ10 treated = 6.35 ± 0.43; Independent sample t-test: t = 2.444, df = 7, p = 0.044) but not in the GCL-IPL when compared to the db/db control mice (db/db control = 13.81 ± 1.51 vs. CoQ10 treated = 11.40 ± 0.79; Independent sample t-test: t = 1.509, df = 7, p = 0.175). In addition, the microglial cell in the retinas of the control group appeared to be less ramified than those in the treatment group (Figure 5).

The effect of CoQ10 on the mitochondrial bioenergetics of the ex-vivo retina of the db/db mice in the treatment and control group was tested after four months of CoQ10 treatment using a Seahorse XF Analyzer. Compared to the control mice, CoQ10 treated db/db mice exhibited a significantly higher basal respiration (1.373 ± 0.142-fold, p = 0.044), maximal respiration (1.589 ± 0.154-fold, p = 0.031), and spare capacity (6.133 ± 0.895-fold, p = 0.003). These results indicate a stronger mitochondrial function in the retina of db/db mice treated with CoQ10 compared to db/db control mice (Figures 6A, B).

The ameliorated retinal mitochondrial bioenergetics of the CoQ10 treated db/db mice were accompanied by an alleviated immunoreactivity of MMP-9 (Figures 6C, D). Results from immunohistochemistry showed that the retinal sections of CoQ10 treated db/db mice had a significantly lower immunoreactivity toward MMP-9 compared to the control db/db mice (db/db control = 6.44 ± 1.32 vs. CoQ10 treated = 2.49 ± 0.36; Independent sample t-test: t = 3.364, df = 10, p = 0.007).