Serine supplementation suppresses hypoxia-induced pathological retinal angiogenesis

Abstract

Rationale: Pathological retinal angiogenesis with irregular and fragile vessels (also termed neovascularization, a response to hypoxia and dysmetabolism) is a leading cause of vision loss in all age groups. This process is driven in part through the energy deficiency in retinal neurons. Sustaining neural retinal metabolism with adequate nutrient supply may help prevent vision-threatening neovascularization. Low circulating serine levels are associated with neovascularization in macular telangiectasia and altered serine/glycine metabolism has been suggested in retinopathy of prematurity. We here explored the role of serine metabolism in suppressing hypoxia-driven retinal neovascularization using oxygen-induced retinopathy (OIR) mouse model. Methods: We administered wild-type C57BL/6J OIR pups with systemic serine or provided the nursing dam with a serine/glycine-deficient diet during the relative hypoxic phase, followed by analysis of retinal vasculature at postnatal (P) 17, the time of peak neovascularization. Retinas from P17 OIR pups with either systemic serine supplementation or vehicle control were subjected to metabolomics, lipidomics, proteomics, and single-cell RNA sequencing. To validate the role of mitochondrial fatty acid oxidation (FAO) and oxidative phosphorylation (OXPHOS) in mediating serine protection against OIR, we treated OIR pups with inhibitors to block FAO or OXPHOS along with either serine or vehicle. The potential transcriptional mediator and pro-angiogenic signals were validated by western blotting. Results: Systemic serine supplementation reduced retinal neovascularization, while maternal dietary serine/glycine deficiency exacerbated it. FAO was essential in mediating serine protective effects, and serine supplementation increased levels of phosphatidylcholine, the most abundant phospholipids in the retina. Serine treatment a) increased the abundance of proteins involved in OXPHOS in retinas, b) increased the expression of mitochondrial respiration-related genes, and c) decreased the expression of pro-angiogenic genes in rod photoreceptor cluster. Serine suppression of retinal neovascularization was dependent on mitochondrial energy production. High mobility group box 1 protein (HMGB1) was identified as a potential key mediator of serine suppression of pro-angiogenic signals in hypoxic retinas. Conclusions: Our findings suggest that serine supplementation may serve as a potential therapeutic approach for neovascular eye diseases by enhancing retinal mitochondrial metabolism and lipid utilization, suppressing the key drivers of uncontrolled angiogenesis.

Keywords: angiogenesis; neovascularization; retina; retinopathy of prematurity; serine.

Figure 1

Serine supplementation during the hypoxic phase of OIR decreased retinal neovascularization. Serine or vehicle was delivered by i.p. injection (0.6 µg/g, A) or oral gavage (6 µg/g) (B) to mouse littermates from P12 to P16 (relative hypoxia). At P17, retinal vaso-obliteration (VO, central area without red fluorescence) and neovascularization (NV, mid-peripheral area with high intensity of red fluorescence, highlighted in white) were examined. Body weight was monitored. n = 16-17 retinas per group (A), n = 12-16 retinas per group (B). Scale bar, 1 mm. Fold change was calculated relative to the average value of littermate vehicle controls. Unpaired t-test (or Welch's t-test) or Mann-Whitney test was used to compare the groups. ns, not significant.

Figure 2

Dietary shortage of serine and glycine worsened hypoxia-induced retinal neovascularization. In mouse OIR, the nursing dam were fed Ser/Gly-deficient or control diet from P14 to P16 (during neovessel formation). At P17, At P17, retinal vaso-obliteration (VO, central area without red fluorescence) and neovascularization (NV, mid-peripheral area with high intensity of red fluorescence, highlighted in white) were examined. n = 21-26 retinas per group. Scale bar, 1 mm. Fold change was calculated relative to the average value of control diet. Unpaired t-test (or Welch's t-test) was used to compare the groups. ns, not significant.

Figure 3

Serine treatment via FAO protected hypoxic retinas. (A) Metabolomic analysis revealed significantly altered retinal metabolites (P < 0.05) in P17 OIR neonates treated with serine (0.6 µg/g, i.p.) or vehicle from P12 to P16. Eight retinas from 4 mice were pooled to form one replicate (n = 1), with n = 3 replicates per group. A heatmap displaying the metabolites with significant changes (P < 0.05, unpaired t test) was generated using Morpheus. Lipid metabolites are highlighted in orange. The schematic was created using BioRender. (B) Metabolomic analysis identified an increase in lipid metabolites in serine-treated OIR retinas. Fold change was calculated and compared with vehicle group. Normality (histogram, QQ-plot, Sapiro-Wilk test) and variance (F-test) were confirmed. Unpaired t-test was then used to compare the groups. (C) Inhibition of fatty acid oxidation (FAO) attenuated the protective effect of serine in OIR mice. OIR neonates were treated with malonyl-CoA (15 mg/kg, i.p.) along with serine or vehicle from P12 to P16. At P17, retinal vaso-obliteration (VO, central area without red fluorescence) and neovascularization (NV, mid-peripheral area with high intensity of red fluorescence, highlighted in white) were examined. n = 12-14 retinas per group. Fold change was calculated relative to the average value of littermate vehicle controls. Unpaired t-test was used to compare the groups. ns, not significant. (D) Lipidomic analysis revealed serine treatment increased total retinal phosphatidylcholine (PtdCho) and decreased phosphatidylserine (PtdSer), phosphatidylinositol (PtdIns) levels in OIR retinas. No significant impact on phosphatidylethanolamine (PtdEtn) and sphingomyelin (CerPCho). Two retinas from one mouse were pooled to form one replicate (n = 1), with n = 6 replicates per group. The percentage of each lipid class relative to the total identified lipids was calculated. Normality (histogram, QQ-plot, Sapiro-Wilk test) and variance (F-test) were confirmed. Unpaired t-test or Welch's test was used to compare the groups. ns, not significant.

Figure 4

Serine treatment preserved mitochondrial respiration in hypoxic retinas. (A) Label-free LC-MS/MS-based proteomic analysis of P17 OIR retinas isolated from mice treated with serine (0.6 ug/g, i.p.) or vehicle control (vehicle; PBS, i.p.) from P12 to P16. Vehicle control group, n = 8; Serine, n = 8 (with two retinas pooled per mouse, n = 1). The volcano plot shows the differentially abundant proteins between serine- vs. control-treated OIR retinas. Each data point represents a unique protein, with the x-axis showing log₂ (fold change) and the y-axis representing -log₁₀ (p value). Out of 3,701 proteins identified (with at least two unique peptides), 452 proteins met the significant criteria (P < 0.05, adjusted p < 0.41). In serine- vs. control-treated group, 190 proteins were increased (highlighted in orange), and 262 proteins were decreased in abundance (highlighted in blue). (B) Gene ontology (biological process) enrichment analysis of proteins significantly enriched in serine- vs. vehicle- treated OIR retinas (190 proteins). The top 10 enriched pathways are shown, sorted by -log₁₀ (adjusted P-value). The pathway for oxidative phosphorylation (OXPHOS) was highlighted in bold, and involved proteins were indicated in orange. (C) Retinal vaso-obliteration (VO, central area without red fluorescence) and neovascularization (NV, mid-peripheral area with high intensity of red fluorescence, highlighted in white) at P17 were examined in OIR pups treated with the mitochondrial ATP synthase inhibitor oligomycin (0.25 µg/g, i.p.), combined with either serine or vehicle from P12 to P16. n = 13-16 retinas per group. Fold change was calculated relative to the average value of littermate vehicle controls. Unpaired t-test was used to compare the groups. ns, not significant. (D) BaroFuse analysis of oxygen consumption rate (OCR), reflecting mitochondrial respiration, in P17 OIR retinas. The retinas were incubated in the presence of Krebs-Ringer Solution (1 mM glucose) (control medium), Krebs-Ringer Solution (1 mM glucose) supplemented with 10 mM serine or 10 mM serine + 5 mM malate + 1 mM L-leucine + 1 mM glutamine. The glucose transporter inhibitor BAY-876 (20 µM) was added to each medium to inhibit cellular glucose uptake. Oxygen (21%) was constantly supplied to the system. n = 3-4 retinas per group from two independent experiments. The factional changes in baseline OCR are shown. Unpaired t-test was used between each test and control medium. *P < 0.05; ns, not significant.

Figure 5

Serine increased expression of genes involved in metabolism and decreased expression of genes involved in angiogenesis in rods. (A) UMAP projection displaying the different color-coded retinal cell types in serine- vs. control- treated OIR pups. Vehicle control group, n = 4; Serine, n = 4 (with two retinas pooled per mouse, n = 1). (B) Dot plots showing the normalized expression of marker genes for different retinal cell types. Markers for rods, Rho and Pde6b; bipolar cells, Vsx2 and Pcp2; Amacrine cells, Gad1 and Gad2; Müller glia, Slc3a1 and Aqp4; Cones, Opn1sw and Opn1mw. (C) Gene ontology (biological process) enrichment analysis of genes significantly differentially enriched in the rod cluster of serine- vs. control- treated OIR retinas. Pathways associated with mitochondrial respiration are highlighted in bold. The adjusted P-values for each term are shown in bar graphs (adjusted P < 0.001). The gene ratio for each pathway is represented in heatmap (right).

Figure 6

HMGB1 mediated serine suppression of neovascularization in OIR. (A) Protein levels of retinal HMGB1 from serine- vs. control- treated OIR mice were measured using western blot. β-ACTIN was used as an internal control. The band intensity of HMGB1 was normalized to β-ACTIN, and the fold change was calculated relative to the average value of the control group. Normality (QQ-plot) and F-test was first conducted. Unpaired t-test was used to compare the groups. Control, n = 6; Serine, n = 6 (with two retinas pooled per mouse, n = 1). (B) mRNA expression levels of Hmgb1 were measured by qPCR. The relative value of Hmgb1 over CypA was estimated and fold change was calculated referring to the average relative value of control group. Normality (QQ-plot) and F-test was first conducted. Mann-Whitney test was used to compare the groups. Control, n = 7; Serine, n = 6 (2 retinas from each mouse pooled as n = 1). ns, not significant. (C) All pups were treated with HMGB1 inhibitor glycyrrhizin (25 µg/g, i.p.), combined with either serine or vehicle, from P12 to P16. At P17, retinal vaso-obliteration (VO, central area without red fluorescence) and neovascularization (NV, mid-peripheral area with high intensity of red fluorescence, highlighted in white) were examined. n = 14-16 retinas per group. Fold changes were calculated relative to the average value of littermate vehicle controls. Normality (QQ-plot) and F-test was first conducted. Unpaired t-test or Mann-Whitney test was used to compare the groups. ns, not significant.

Figure 7

Schematics of the proposed pathway for serine-mediated suppression of retinal neovascularization. Serine upregulated HMGB1, which in turn enhanced mitochondrial metabolism (e.g. mitochondrial FAO, OXPHOS) to meet the metabolic demands of the developing neurons. This metabolic improvement led to a reduction in proangiogenic responses (HIF-1α induction and STAT3 activation) in the neural retina, thereby decreased the compensatory but pathological retinal neovascularization in hypoxic retina. The retinal structure diagram was generated using Biorender.com