Diminished retinal complex lipid synthesis and impaired fatty acid β-oxidation associated with human diabetic retinopathy

Abstract

BACKGROUNDThis study systematically investigated circulating and retinal tissue lipid determinants of human diabetic retinopathy (DR) to identify underlying lipid alterations associated with severity of DR.METHODSRetinal tissues were retrieved from postmortem human eyes, including 19 individuals without diabetes, 20 with diabetes but without DR, and 20 with diabetes and DR, for lipidomic study. In a parallel study, serum samples from 28 American Indians with type 2 diabetes from the Gila River Indian Community, including 12 without DR, 7 with mild nonproliferative DR (NPDR), and 9 with moderate NPDR, were selected. A mass-spectrometry-based lipidomic platform was used to measure serum and tissue lipids.RESULTSIn the postmortem retinas, we found a graded decrease of long-chain acylcarnitines and longer-chain fatty acid ester of hydroxyl fatty acids, diacylglycerols, triacylglycerols, phosphatidylcholines, and ceramide(NS) in central retina from individuals with no diabetes to those with diabetes with DR. The American Indians' sera also exhibited a graded decrease in circulating long-chain acylcarnitines and a graded increase in the intermediate-length saturated and monounsaturated triacylglycerols from no DR to moderate NPDR.CONCLUSIONThese findings suggest diminished synthesis of complex lipids and impaired mitochondrial β-oxidation of fatty acids in retinal DR, with parallel changes in circulating lipids.TRIAL REGISTRATIONClinicalTrials.gov NCT00340678.FUNDINGThis work was supported by NIH grants R24 DK082841, K08DK106523, R03DK121941, P30DK089503, P30DK081943, P30DK020572, P30 EY007003; The Thomas Beatson Foundation; and JDRF Center for Excellence (5-COE-2019-861-S-B).

Keywords: Diabetes; Fatty acid oxidation; Ophthalmology; Retinopathy.

Conflict of interest statement

Conflict of interest: The authors have declared that no conflict of interest exists.

Figures

Figure 1. Flow diagram.

(A–C) Flow diagram of Pima lipidomics (A), postmortem retinal tissue lipidomics (B), and postmortem retinal tissue transcriptomics studies (C).

Figure 2. Comparing circulating lipids by diabetic retinopathy status.

(A) Distribution of individual ACs and TAGs by retinopathy. Within each diagram, the x axis shows the number of carbons, the y axis shows the number of double bonds, and the color codes within each cell represent the Z score standardized mean abundance of the corresponding lipid. Statistical tests were mixed models (2-tailed t test) that tested the effect of the study group, carbon number, and number of double bonds as the main effects and their interaction; they also adjusted for urine albumin/creatinine ratio, use of metformin, and insulin. P values refer to significance of change in relative abundance of the corresponding lipid by increase in carbon number in ACs, as well as by increase in carbon number and number of double bonds (their interaction term) in TAGs. n for no DR, mild, and moderate NDPR is 12, 7, and 9, respectively. (B) Components of long-chain ACs (C ≥ 14) and intermediate-length unsaturated and monounsaturated TAGs accurately predicted the group without retinopathy (100%), mild NPDR (100%), and moderate NPDR (88.9%) with an overall accuracy of 96.4%. Statistical test was canonical discriminant analysis of components of differential lipid factors. n for no DR, mild, and moderate NDPR is 12, 7, and 9, respectively. AC, acylcarnitine; CE, cholesteryl ester; DAG, diacylglycerol; TAG, triacylglycerol; PC, phosphatidylcholine; PE, phosphatidylethanolamine; LPC, lyso-PC; LPE, lyso-PE; pPE, plasmenyl-PE; SM, sphingomyelin; sFFA, saturated free fatty acid; uFFA, unsaturated FFA; Sat Monounsat, saturated and monounsaturated; Polyunsat, polyunsaturated; Intm, intermediate; NPDR, nonproliferative diabetic retinopathy.

Figure 3. Z score–standardized mean relative abundance of FFAs, ACs, and complex lipids by carbon numbers and double bond numbers among postmortem groups.

(A) Significantly higher abundant long chain ACs (C ≥ 14) in participants without diabetes and diabetics without retinopathy, higher unsaturated FFAs in participants without diabetes, and higher polyunsaturated longer-chain FAHFA in participants without diabetes, and their lower levels among diabetics with retinopathy. P values refer to significance of change in relative abundance of the corresponding lipid by increase in carbon number in ACs, by increase in unsaturated FFAs as compared with saturated FFAs in FFAs, and by increase in carbon number and number of double bonds (interaction term) in FAHFAs. (B) Significantly higher abundance of unsaturated DAGs, polyunsaturated TAGs, and PCs with higher carbon numbers in groups without diabetes and diabetics without retinopathy; however, a significantly lower abundance of unsaturated DAGs, polyunsaturated TAGs, PGs, and PCs is shown with higher carbon numbers in diabetes with retinopathy. Within each diagram, the x axis shows the number of carbons, and the y axis shows the number of double bonds. Statistical tests are based on mixed models tested the effect of study group, carbon number, and number of double bonds as the main effects and their interaction, adjusted for hypertension, hyperlipidemia, coronary artery disease, cancer, and kidney failure. P values refer to significance of change in relative abundance of the corresponding lipid by increase in number of double bonds in DAGs, and by increase in carbon number and number of double bonds (their interaction term) in TAGs, PGs, and PCs. n for both A and B is 19, 20, and 20 for participants without diabetes, diabetes with no DR, and diabetes with DR, respectively. AC, acylcarnitine; FAHFA, fatty acid ester of hydroxyl fatty acid; FFA, free fatty acid; DAG, diacylglycerol; TAG, triacylglycerol; PG, phosphatidylglycerol; PC, phosphatidylcholine; DM; diabetes mellitus; NPDR, nonproliferative diabetic retinopathy.

Figure 4. Group discrimination by retinal lipids in a postmortem cohort.

(A and B) Top differentially regulated lipids, triglycerides, phosphatidylcholine, and ACs in central retinal (A), and triglycerides and phosphatidylcholines in peripheral retina (B) discriminated 3 study groups. Analyses are based on canonical discriminant analysis with 100% accurate discrimination in central retina, and 98.3% in peripheral retina. (C and D) Statistically significant correlates of saturated and unsaturated fatty acids in the central and peripheral retina of postmortem cohort. Lipids in red indicate elevated levels; lipids in blue indicate suppressed levels; arrows in red indicate statistically significant direct correlation; arrows in blue indicate statistically significant inverse correlation. Overall saturated FFAs are inversely correlated with unsaturated FFAs. A higher abundance of unsaturated FFAs suggest shift of sFFAs toward uFFAs via action of desaturases. Direct correlation of uFFAs with complex glycerolipids suggests their higher incorporation in the construct of glycerolipids, while inverse correlation of sFFAs with complex lipids suggest their relatively lower incorporation in the construct of complex lipids. The net effect would be a relatively higher abundance of complex lipids in association with a higher abundance of uFFAs under normal physiological conditions. With progression toward diabetic retinopathy, diminished levels of uFFAs leads to diminished tissue levels of glycerophospholipids. n for all panels is 19, 20, and 20 for participants without diabetes, diabetes with no DR, and diabetes with DR, respectively. AC, acylcarnitine; sFFA, saturated free fatty acids; uFFA, unsaturated free fatty acids; DAG, diacylglycerols; PC, phosphatidylcholines; TAG, triacylglycerols.

Figure 5. RNA deep sequencing and qPCR.

Relative expression of the lipid metabolism gene LPCAT3 in the central (left; CR) and peripheral (right; PR) retina, analyzed by RNA deep sequencing (top) or qPCR (bottom). Similar changes were observed by both methods. In the top panels, n is 6, 6, and 4 for participants without diabetes, diabetes without DR, and diabetes with DR, respectively. In the bottom panel n is, 22, 22, and 20 for participants without diabetes, diabetes without DR, and diabetes with DR, respectively. qPCR P values are adjusted by sex, hypertension, hyperlipidemia, coronary artery disease, nephropathy, and cancer. The statistical test was ANOVA. Data are shown as mean ± SD (*P < 0.05 in DR versus ND).

Figure 6. Comparing lipid-related gene expression in central and peripheral retina by study groups.

(A and B) Relative expression (Z score) of the lipid metabolism–related genes significantly affected by diabetes and DR in the central (left; CR) and peripheral (right; PR) retina, analyzed by RNA deep sequencing. n for both panels is 6, 6, and 4 for participants without diabetes, diabetes without DR, and diabetes with DR, respectively.

Figure 7. Similar lipid alterations in serum and retinal suggesting parallel changes in the 2 compartments.

Left panel shows the serum lipidomics in moderate nonproliferative diabetic retinopathy (n = 9), and right panel shows central retinal lipidomics in diabetes with retinopathy (n = 20). The x axis in all panels represents carbon number, and the y axis represents the number of double bonds.