Relaxation of mitochondrial hyperfusion in the diabetic retina via N6-furfuryladenosine confers neuroprotection regardless of glycaemic status

Abstract

The recovery of mitochondrial quality control (MQC) may bring innovative solutions for neuroprotection, while imposing a significant challenge given the need of holistic approaches to restore mitochondrial dynamics (fusion/fission) and turnover (mitophagy and biogenesis). In diabetic retinopathy, this is compounded by our lack of understanding of human retinal neurodegeneration, but also how MQC processes interact during disease progression. Here, we show that mitochondria hyperfusion is characteristic of retinal neurodegeneration in human and murine diabetes, blunting the homeostatic turnover of mitochondria and causing metabolic and neuro-inflammatory stress. By mimicking this mitochondrial remodelling in vitro, we ascertain that N6-furfuryladenosine enhances mitochondrial turnover and bioenergetics by relaxing hyperfusion in a controlled fashion. Oral administration of N6-furfuryladenosine enhances mitochondrial turnover in the diabetic mouse retina (Ins2^Akita males), improving clinical correlates and conferring neuroprotection regardless of glycaemic status. Our findings provide translational insights for neuroprotection in the diabetic retina through the holistic recovery of MQC.

Fig. 1. Neuronal degeneration is associated with mitochondrial hyperfusion in the human diabetic retina.

a, b Mitochondrial fusion was evaluated at the outer nuclear layer (ONL) of nondiabetic (ND) and diabetic (D) human donors using Cox4 immunostaining. Arrowheads indicate hyperfused mitochondrial networks. b Morphometric quantification of mitochondrial fusion (interconnectivity, average mitochondrial length and aspect ratio). Retinal micrographs from nondiabetic and diabetic human donors processed for c DAPI and d medium-wavelength opsin (M-opsin). Atrophy of cone photoreceptor outer/inner segments (arrowheads) and of synaptic terminals (arrows). e The density of DAPI⁺ nuclei at the ONL. Retinal micrographs from nondiabetic and diabetic human donors immunostained for f synaptophysin and g Calbindin. Arrowheads indicate loss of f presynaptic and g postsynaptic elements at the outer plexiform layer (OPL). h The density of synaptophysin⁺ processes at the OPL. Correlation of mitochondrial fusion (interconnectivity index) against the density of i DAPI⁺ nuclei at the ONL or j synaptophysin⁺ processes at the OPL. Regression line, significance levels (P), coefficient of determination (r²) and 95% confidence bands of the best fit line (grey) are shown. ND (n = 3 donor eyes), D (n = 7 donor eyes (b–e), n = 8 donor eyes (h)). (b–e, h) Data are presented in box-and-whisker plots with single data points (for definition of boxplot elements see “Methods” section). P values were calculated using 2-sided unpaired Student’s t-test. IS, photoreceptor inner segments; INL, inner nuclear layer; IPL, inner plexiform layer; Scale bars: 40 µm.

Fig. 2. Mitochondrial hyperfusion underpins impaired mitophagy in the retina of 8-month diabetic Ins2^Akita male mice.

a Quantification of mitophagy (mCherry-only-foci, arrowheads) and mitochondrial morphology (Fis1-GFP) using the Mito-QC reporter. b, c Mitochondrial fusion was evaluated using Fis1-GFP signal at the outer nuclear layer (ONL) of 8-month diabetic mitophagy reporter mice (mitoQC Ins2^Akita) and age-matched non-diabetic siblings (mitoQC WT). Arrowheads indicate hyperfused mitochondrial networks. c Morphometric quantification of mitochondrial fusion (interconnectivity, average mitochondrial length and aspect ratio) in MitoQC WT (n = 8 eyes) and MitoQC Ins2^Akita (n = 9 eyes). d, e Mitolysosome density (arrowheads) at the OLM-OPL of mitoQC WT mice and mitoQC Ins2^Akita (n = 10 eyes/strain). f, g Morphometric quantification of mitochondrial mass (% of Fis1⁺ signal) at the ONL of MitoQC WT (n = 8 eyes) and MitoQC Ins2^Akita (n = 9 eyes). h Example immunoblot of mitochondrial ATP-synthase and β-actin loading controls in retinal lysates of 8-month diabetic Ins2^Akita mice and age-matched WT siblings (n = 3 eyes/strain). Correlation between mitophagy and mitochondrial fusion (aspect ratio and mitochondrial interconnectivity) in i mitoQC WT and j mitoQC Ins2^Akita mice. Regression line, significance levels (P), coefficient of determination (r²) and 95% confidence bands of the best fit line (grey) are shown. i MitoQC WT (n = 25 [aspect ratio], n = 27 [mito interconnectivity]) retinal sections from n = 8 eyes. j MitoQC Ins2^Akita (n = 27 [aspect ratio], n = 29 [mito interconnectivity]) retinal sections from n = 8 eyes. Data are presented in box-and-whisker plots with single data points (for definition of boxplot elements see “Methods” section). P-values were calculated using 2-sided unpaired Student’s t-test. OLM outer limiting membrane, OPL outer plexiform layer. Scale bars: 40 µm (d, f), 20 µm (b and inset in d).

Fig. 3. Exacerbated fusion blunts mitophagy impairing mitochondrial fitness in Müller glia under hyperglycaemia.

a–f Primary Müller cells isolated from MitoQC mice (glutamine synthase immunoreactive) or g–m human MIO-M1 Müller cells, were antagonized for mitochondrial fission (P110 peptide) under physiological (NG; 5.5 mM) or elevated glucose (HG; 30.5 mM) conditions. Quantification of b, c mitochondrial fusion (interconnectivity and average mitochondrial length; arrowheads indicate fragmented mitochondria), b, d mitophagy (arrow), b, e mitochondrial mass (Fis1-GFP) and b, f TFAM expression (mean fluorescence intensity [MFI]). c–f n = 4 independent replicates. g, h Evaluation of mitochondrial membrane potential (ψm) by JC-1 dye (red, hyperpolarized; green, depolarized mitochondria). CCCP (100 μM) was added as a mitochondrial uncoupler positive control (2 hours). Gray squares in scatter-plots in g indicate % of cells with low ψm. n = 4 (CCCP), n = 9 (all other groups) independent replicates. i Representative Seahorse assay of metabolic flux (left panel) and metabolic potential (right panel) using Cell Mito Stress Test. (j–m) Quantification of oxygen consumption rate (OCR) indicative of j basal respiration, k ATP-linked respiration, l Maximum respiration and m metabolic potential. j–m n = 6 (NG + P110), n = 13 (NG, HG), n = 14 (HG + P110) independent replicates. Data are presented as (c–f, j–m) box-and-whisker plots (for definition of boxplot elements see “Methods” section) or h, i mean ± SD. P-values were calculated using One-way ANOVA with Dunnett’s multiple comparison. ECAR extracellular acidification rate. TFAM mitochondrial transcription factor A. Scale bars: 20 µm (a, b, g), 5 µm (insets in b).

Fig. 4. Exacerbated mitochondrial fusion triggers Müller glia neuroinflammation and neuronal synaptic stress under hyperglycaemia.

a, b Retinal micrographs and quantification of Vimentin radial processes at the outer (ONL-OPL) and inner (GCL-INL) retina of 8-month diabetic mice (Ins2^Akita males) and age-matched WT male siblings. Arrowheads indicate Vimentin processes at the ONL (n = 4 eyes/strain). c, d Vimentin expression (mean fluorescence intensity [MFI]) in human MIO-M1 cultures (shown in c) or mouse primary Müller cells (PMCs), antagonized for mitochondrial fission (P110 peptide) under physiological (NG; 5.5 mM) or elevated glucose (HG; 30.5 mM) conditions. d MIO-M1: n = 8 (NG, NG + P110), n = 9 (HG, HG + P110) independent replicates; PMCs n = 8 (NG), n = 9 (NG + P110, HG and HG + P110) independent replicates. e Luminex® Multiplex Assay of secreted MCP-1, VEGF-A, GM-CSF, IL-6 and IL-8 levels in MIO-M1 cell supernatants grown under NG or HG ± P110 peptide (n = 3 independent replicates). f–h Human MIO-M1 cells were induced for neuronal differentiation (48 h) and thereafter cultured for 24 h under NG or HG ± P110 peptide. f Validation of neuronal phenotype in MIO-M1 differentiated cultures via expression of heavy-chain neurofilament, β-III tubulin and the development of complex neurite networks. g, h Quantification of neurite length in different treatment groups. Arrowheads indicate neurite retraction (n = 4 independent replicates). Data are presented as b) box-and-whisker plots with single data points (for definition of boxplot elements see “Methods” section) or d, e, h mean ± SD. P-values were calculated using b 2-sided unpaired Student’s t-test or d, e, h One-way ANOVA with Dunnett’s multiple comparisons. ONL outer nuclear layer, OPL outer plexiform layer, INL inner nuclear layer, IPL inner plexiform layer, GCL ganglion cell layer. Scale bars: 60 µm (a), 40 µm (g), 20 µm (c and inset in a).

Fig. 5. Glycosylated N6-furfuryladenine (kinetin riboside) rescues mitochondrial turnover and bioenergetics in Müller glia under diabetes-mitochondrial hyperfusion.

a Molecular structure of kinetin and its metabolic conversion into the active substrate kinetin riboside triphosphate. b–d Primary Müller cells isolated from mitoQC mice or e–j human MIO-M1 cells were antagonized for mitochondrial fission (P110 peptide) under elevated glucose (HG; 30.5 mM) conditions, and treated with different concentrations of kinetin, kinetin riboside (KR), niclosamide (Nicl), urolithin-A (Uro-A) or DMSO control (0.1%). b, c Mitophagy (arrowheads) and b–d mitochondrial fusion index (as calculated by the product of mitochondrial average length and interconnectivity). c n = 3 (KR 1 µM), n = 4 (all other groups) independent replicates; d n = 4 (Nicl, Uro-A), n = 5 (DMSO, KR 0.3 µM), n = 6 (all other groups) independent replicates. e Evaluation of mitochondrial membrane potential (ψm) by JC-1 dye (red, hyperpolarized; green, depolarized mitochondria). CCCP (100 μM) was added as a mitochondrial uncoupler positive control (2 hours). n = 4 (KR 0.3 µM, KR 0.7 µM), n = 6 (CCCP), n = 8 (all other groups) independent replicates. Representative Seahorse assay of f metabolic flux and g metabolic potential using Cell Mito Stress Test. Quantification of oxygen consumption rate (OCR) indicative of h basal respiration, i ATP-linked respiration and j metabolic potential. h–j n = 13 (kinetin 1 µM), n = 14 (KR 0.3 µM), n = 19 (KR 1 µM), n = 22 (HG, Baseline, KR 0.7 µM), n = 24 (DMSO) independent replicates. Data are presented as c–g mean ± SD or h–j box-and-whisker plots (for definition of boxplot elements see “Methods” section). P-values were calculated using One-way ANOVA with Dunnett’s multiple comparisons. ECAR, extracellular acidification rate. Scale bars: 40 µm.

Fig. 6. Kinetin riboside (KR) rescues mitochondrial turnover in the diabetic retina conferring relevant neuroprotection independently of glycaemic status.

a–g Validation of effective KR dosage in vivo. KR was supplemented in drinking water (60 mg/L) to 7.5-month diabetic mitophagy reporter mice (mitoQC-Ins2^Akita males) for 2 weeks, and its bioactivity in the diabetic retina confirmed via evaluation of b, c mitophagy (arrowheads), d, e mitochondrial fusion (interconnectivity and average mitochondrial length), f mitochondrial mass (% of Fis1⁺ signal) and b, g TFAM⁺ mitochondrial nucleoids (arrows). c–g Eyes per strain and condition: mitoQC WT (n = 10 (c), n = 8 (d–g), mitoQC Ins2^Akita (n = 10 [c], n = 9 [d-g]), mitoQC Ins2^Akita + KR (n = 6 [c-g]). h–p Effective KR dosage (60 mg/L) or DMSO vehicle-control (0.1%) was supplemented in the drinking water from 4-months to 8-months of diabetes in Ins2^Akita male mice. Following treatment, glycaemic status and retinal neurodegeneration was evaluated by in vivo and post-mortem approaches. i Weight (g) and blood glucose levels (mmol/L). Mice per strain and condition: WT (n = 7 weight, n = 5 glucose levels), Ins2^Akita (n = 5 weight, n = 8 glucose levels), Ins2^Akita + DMSO (n = 4 weight, n = 5 glucose levels), Ins2^Akita + KR (n = 4 weight, n = 4 glucose levels). (j) In vivo quantification of neuroretinal thickness by SD-OCT (from GCL/NFL to IS/OS, arrow). k, l Retinal function assessed via scotopic electroretinogram (a-wave and b-wave amplitudes). m The length of cone photoreceptor segments (cone-arrestin, arrowheads). n The density of synaptophysin⁺ processes at the OPL (arrowheads). o The density of horizontal cell dendritic boutons at the OPL (calbindin, arrows). p The density of GABAergic amacrine cells at the INL (arrows). j–p Eyes per strain and condition: WT (n = 16 [j], n = 14 [k, l], n = 10 [m], n = 8 [n], n = 7 [o], n = 9 [p]), Ins2^Akita (n = 10 [j], n = 6 [k, l], n = 7 [m], n = 8 [n–p]), Ins2^Akita + DMSO (n = 8 [j], n = 6 [k, l], n = 7 [m, p], n = 5 [n, o]), Ins2^Akita + KR (n = 6 [j], n = 8 [k, l], n = 4 [m–p]). Data are presented as (c–g, k–p) box-and-whisker plots (for definition of boxplot elements see “Methods” section), or (i, j) mean ± SE. P-values were calculated using One-way ANOVA with Dunnett’s multiple comparison. IS, photoreceptor inner segments, OS photoreceptor outer segments, ONL outer nuclear layer, OPL outer plexiform layer, INL inner nuclear layer, IPL inner plexiform layer, GCL ganglion cell layer, NFL nerve fibre layer. Scale bars: 40 µm.