Increasing Energetic Demands on Photoreceptors in Diabetes Corrects Retinal Lipid Dysmetabolism and Reduces Subsequent Microvascular Damage

Abstract

Mechanisms responsible for the pathogenesis of diabetic retinal disease remain incompletely understood, but they likely involve multiple cellular targets, including photoreceptors. Evidence suggests that dysregulated de novo lipogenesis in photoreceptors is a critical early target of diabetes. Following on this observation, the present study aimed to determine whether two interventions shown to improve diabetic retinopathy in mice-pharmacologic visual cycle inhibition and prolonged dark adaptation-reduce photoreceptor anabolic lipid metabolism. Elevated retinal lipid biosynthetic signaling was observed in two mouse models of diabetes, with both models showing reduced retinal AMP-activated kinase (AMPK) signaling, elevated acetyl CoA carboxylase (ACC) signaling, and increased activity of fatty acid synthase, which promotes lipotoxicity in photoreceptors. Although retinal AMPK-ACC axis signaling was dependent on systemic glucose fluctuations in healthy animals, mice with diabetes lacked such regulation. Visual cycle inhibition and prolonged dark adaptation reversed abnormal retinal AMPK-ACC signaling in mice with diabetes. Although visual cycle inhibition reduced the severity of diabetic retinopathy in control mice, as assessed by retinal capillary atrophy, this intervention was ineffective in fatty acid synthase gain-of-function mice. These results suggest that early diabetic retinopathy is characterized by glucose-driven elevations in retinal lipid biosynthetic activity, and that two interventions known to increase photoreceptor glucose demands alleviate disease by reversing these signals.

Figure 1

Effects of diabetes on fasting and refeeding responses of retinal cellular nutrient sensors. A–H: The 6-month–old db/db mice (A–D) or streptozotocin (STZ)–injected mice (E–H) and their respective controls were analyzed for glucose excursions and retinal phosphorylated AMP-activated kinase (AMPK) and acetyl CoA carboxylase (ACC) profiles. All mice were fasted for 6 hours and then either immediately sacrificed or allowed to refeed with chow for 1 or 3 hours. A and E: Tail vein glucose values in each group were recorded. Retinal tissue from each group was analyzed for AMPK phosphorylated at Thr 172 (pAMPK) and ACC phosphorylated at Ser 79 (pACC), with all values normalized to β-actin levels. B–D and F–H: Representative Western blot analyses are shown (B and F), and quantified values are plotted (C, D, G and H). Data were analyzed by two-way analysis of variance with Bonferroni post hoc tests. Data represent means with SEM (A, C–E, G, and H). ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, and ∗∗∗∗P < 0.0001; ^ɫP < 0.05, ^ɫɫP < 0.01, comparing post-prandial glucose values with the fasting baseline within each group. Veh, vehicle.

Figure 2

Responses of retinal cellular nutrient sensors to glucose or insulin challenge in db/db mice. A–H: The 6-month–old db/db mice or nondiabetic db/m littermate controls were fasted for 6 hours and then given an i.p. dose of dextrose (A–D) or insulin (E–H). B and F: In separate cohorts, tail vein sampling was used to measure glucose values at pre-injection baseline and 10, 30, 60, 90, and 120 minutes after the injection. A, C–E, G, and H: Mice were sacrificed after tail vein sampling, and retinal tissue was analyzed for AMP-activated kinase phosphorylated at Thr 172 (pAMPK) and acetyl CoA carboxylase phosphorylated at Ser 79 (pACC; A and E), with results quantified in relation to a control protein (C, D, G, and H). Data were analyzed by two-way analysis of variance with Bonferroni, Tukey, or Dunnett post hoc tests. Data represent means with SEM (B–D and F–H). ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, and ∗∗∗∗P < 0.0001, comparing postinjection time points with the baseline; ^ɫP < 0.05, ^ɫɫP < 0.01, ^ɫɫɫP < 0.001, and ^ɫɫɫɫP < 0.0001, comparing db/m with db/db for each time point.

Figure 3

Effects of prolonged dark adaptation on retinal cellular nutrient sensors and retinal de novo lipogenesis. The 3-month–old C57BL/6J littermate mice were randomized to housing with normal 12-hour:12-hour light-dark cycle or to 24-hour dark cycles. After 2 weeks, retinal and liver tissues were harvested and analyzed. A–H: Representative Western blot analyses for AMP-activated kinase phosphorylated at Thr 172 (pAMPK) and acetyl CoA carboxylase phosphorylated at Ser 79 (pACC) are shown for retina (A) and liver (B), with quantification of values, normalized to β-actin (C, D, F, and G). In a subgroup of mice, retinal (E) and liver (H) tissue was harvested and analyzed for fatty acid synthase (FAS) activity, normalized to total protein content. Data were analyzed by t-tests. I–N: The 6-month–old db/db mice (I–K) or streptozotocin (STZ)–treated mice (L–N) were randomized to 2 weeks of 12-hour:12-hour light-dark cycle housing or to 24-hour dark housing, and were analyzed for retinal pAMPK and pACC levels, compared with nondiabetic controls in normal light-dark cycles (either db/m for db/db mice; or vehicle (Veh)–treated littermates for STZ-treated mice). J, K, M, and N: Levels of pAMPK or pACC, normalized to β-actin, were compared across subgroups in each diabetes model. Data were analyzed by one-way analysis of variance with Tukey post hoc tests. Data represent means with interquartile ranges (C–H, J, K, M, and N). ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, and ∗∗∗∗P < 0.0001. NS, not statistically significant.

Figure 4

Effects of visual cycle inhibition on retinal cellular nutrient sensors and retinal de novo lipogenesis. The 3-month–old C57BL/6J littermate mice were randomized to treatment with retinylamine (Ret-NH₂) or vehicle, twice weekly by i.p. injection for 2 weeks. Full-field electroretinography was performed to assess effects of the drug. A–D: Scotopic responses to a –1.20 log(cd·s/m²) luminance white flash (A) and photopic responses to a 1.41 log(cd·s/m²) luminance white flash (B) are shown, with quantification of luminance-response curves of the scotopic a-waves (C) and photopic b-waves (D). Data were analyzed by two-way analysis of variance with Bonferroni post hoc tests. In the same mice, retinal tissue was then harvested and analyzed for AMP-activated kinase phosphorylated at Thr 172 (pAMPK) and acetyl CoA carboxylase phosphorylated at Ser 79 (pACC). E–G: Representative blots (E) with quantification of values, normalized to β-actin (F and G), are shown. Data were analyzed by t-tests. H–M: The 6-month–old db/db mice (H–J) or streptozotocin (STZ)–treated mice (K–M) were randomized to 2 weeks of twice-weekly Ret-NH₂ or vehicle treatment, and were analyzed for retinal pAMPK and pACC levels, compared with vehicle-treated nondiabetic controls (either db/m for db/db mice; or citrate buffer–treated littermates for STZ-treated mice). I, J, L, and M: Levels of pAMPK or pACC, normalized to β-actin, were compared across subgroups in each diabetes model. Data were analyzed by one-way analysis of variance with Tukey post hoc tests. Data represent means with SEM (C and D) or means with interquartile ranges (F, G, I, J, L, and M). ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, and ∗∗∗∗P < 0.0001. NS, not statistically significant.

Figure 5

Prevention of diabetic retinopathy by visual cycle inhibition in streptozotocin (STZ)–treated mice is dependent on inhibition of de novo lipogenesis. The 3-month–old FASR18W transgenic mice or wild-type (WT) littermate control mice were randomized to treatment with retinylamine (Ret-NH₂) or vehicle (Veh), twice weekly by i.p. injection for 2 weeks. Retinal tissue was then harvested from all mice and analyzed for AMP-activated kinase phosphorylated at Thr 172 (pAMPK) and acetyl CoA carboxylase phosphorylated at Ser 79 (pACC). A–C: Representative blots (A) with quantification of values, normalized to β-actin (B and C), are shown. D: In a subgroup of mice, retinal tissue was harvested and analyzed for fatty acid synthase (FAS) activity, normalized to total protein content. Data were analyzed by one-way analysis of variance with Tukey post hoc tests. In a separate cohort, 6-month–old STZ-treated mice (in either wild-type FAS or FAS R1812W genetic backgrounds) were treated with Ret-NH₂ or vehicle twice weekly for 3 months. E and F: Capillaries were then assessed in retinal trypsin digest preparations (E) with quantification of thin-atrophic capillaries (arrowheads; F). G and H: The number of atrophic capillaries per field was compared across groups, including vehicle-injected nondiabetic controls. Data were analyzed by one-way analysis of variance with Tukey post hoc tests. Data represent means with interquartile ranges (B–D, G, and H). ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, and ∗∗∗∗P < 0.0001. Scale bars = 50 μm (E and F). DM, diabetes mellitus; HPF, high-power field; NS, not statistically significant.