Perilipin 2-positive mononuclear phagocytes accumulate in the diabetic retina and promote PPARγ-dependent vasodegeneration

Abstract

Type 2 diabetes mellitus (T2DM), characterized by hyperglycemia and dyslipidemia, leads to nonproliferative diabetic retinopathy (NPDR). NPDR is associated with blood-retina barrier disruption, plasma exudates, microvascular degeneration, elevated inflammatory cytokine levels, and monocyte (Mo) infiltration. Whether and how the diabetes-associated changes in plasma lipid and carbohydrate levels modify Mo differentiation remains unknown. Here, we show that mononuclear phagocytes (MPs) in areas of vascular leakage in DR donor retinas expressed perilipin 2 (PLIN2), a marker of intracellular lipid load. Strong upregulation of PLIN2 was also observed when healthy donor Mos were treated with plasma from patients with T2DM or with palmitate concentrations typical of those found in T2DM plasma, but not under high-glucose conditions. PLIN2 expression correlated with the expression of other key genes involved in lipid metabolism (ACADVL, PDK4) and the DR biomarkers ANGPTL4 and CXCL8. Mechanistically, we show that lipid-exposed MPs induced capillary degeneration in ex vivo explants that was inhibited by pharmaceutical inhibition of PPARγ signaling. Our study reveals a mechanism linking dyslipidemia-induced MP polarization to the increased inflammatory cytokine levels and microvascular degeneration that characterize NPDR. This study provides comprehensive insights into the glycemia-independent activation of Mos in T2DM and identifies MP PPARγ as a target for inhibition of lipid-activated MPs in DR.

Keywords: Diabetes; Inflammation; Macrophages; Microcirculation; Ophthalmology.

Figure 1. PLIN2⁺ MPs are found in regions with retinal plasma leakage in a patient with DR.

(A–G) Representative immunofluorescence (IF) images of human flat-mounted retinas from 4 postmortem DM donors. (A) IF staining of ALB (plasma protein) and COL4 (vessels) within a healthy vasculature area of human retina. Scale bar: 200 μm. (B) IF staining for ALB and COL4 within an area showing vascular leakage. Scale bar: 200 μm. (C) IF staining for ALB, IBA1, and PLIN2 within an area with a leaking microaneurysm. Scale bar: 40 μm. (D) IF staining for COL4, IBA1 (MPs), and PLIN2 (neutral lipid droplets) within an area with a microaneurysm. Scale bar: 30 μm. (E and F) IF staining for IBA1 and PLIN2 and labeling with UEA1 (vessels). (E) IF staining for IBA1 and PLIN2. The arrows show a PLIN2^– (blue) and a PLIN2⁺ (red) MP within the same region. Scale bar: 10 μm. (F) Representative image outside the area with the microaneurysms. Scale bar: 200 μm. (G) Representative image within the area of the microaneurysms. (H) Total MP count and PLIN2⁺ MP count around microaneurysms in 4 different DM donors (each color represents a different donor). (I) Percentage of PLIN2⁺ MPs within microaneurysms containing at least 1 PLIN2⁺ MP. (J) Schematic depiction of a ruptured retinal microaneurysm leading to plasma leakage in a DR retina.

Figure 2. PA stimulation triggers the expression of key DR markers by MPs.

(A) Schematic representation of Mo isolation, treatment, and preparation. (B–D) Chromatographic analysis of FA chain composition after BSA (n = 3) or PA (n = 3) treatment. (B) Mean percentage of PA (C16:0) chains relative to total FAs. P s were calculated using Welch’s t test. (C) Mean percentage of monounsaturated FAs (MUFAs), polyunsaturated FAs (PUFAs), and saturated FAs (SFAs) (minus PA). (D) Mean percentage of ω−3 and ω–6 PUFAs. (C and D) P values were calculated using multiple Welch’s t tests corrected for multiple comparisons using the Holm-Šidák method. (E) Heatmap representation of the percentage of individual FA chains relative to total FA chains (minus PA). (F) RT-qPCR quantification of PLIN2 after treatment with BSA (n = 4), PA (n = 4), or PA_CH3 (n = 4). P values were determined by 1-way Welch’s ANOVA (P = 0.0021) followed by Dunnett’s T3 multiple-comparison test. (G–I) RNA-Seq transcriptomics analysis after BSA (n = 3) or PA (n = 3) treatment. (G) Scatter plot of the mean TPM value for all transcripts detected after BSA (x axis) or PA (y axis) treatment. The red and blue dots represent transcripts upregulated with a log₂ FC of 4 or higher or a log₂ FC of 4 or lower. (H) Heatmap representation of the log₂ variance stabilizing transformation (vst) of the top 10 upregulated and downregulated transcripts. (I) GO enrichment analysis representing the fold enrichment of the 528 transcripts with a log₂ FC of 2 or higher; red dots represent pathways related to lipid metabolism. Numbers in parenthesis represents the number of genes regulated by PA stimulation. Resp., response; Neg., negative; reg., regulator; act. activation; lipoprot., lipoprotein; Cell., cellular; stim., stimulation; Pos., positive; org., organization; diff., differentiation; drvd, derived. (J) Heatmap representation of the log₂ vst of transcripts with a log₂ FC of 4 or higher and belonging to the GO pathway “fatty acid metabolic process.” (K) Schematic representation of the biological function of the markers selected as a signature of lipid exposure.

Figure 3. T2DM plasma, but not glucose, induces a lipid-associated phenotype in MPs.

(A) Schematic representation of donor phenotyping and group attribution, plasma preparation, and naive Mo treatment. (B and C) RT-qPCR quantification of healthy donor naive Mos treated for 18 hours with donor plasma from ND individuals (n = 10 [blue dots]) or patients with T2DM (n = 27, no DR [light pink dots], NPDR [pink dots], PDR [dark red dots]). (B) Violin plot representation of the relative expression of the indicated genes in response to individual donor plasma exposure. Dashed lines represent the median and quartiles. P values were determined using a 2-tailed Mann-Whitney U test. (C) Simple linear regression representation of PDK4, ACADVL, ANGPTL4, and CXCL8 (y axis) and PLIN2 expression (x axis). Correlations between expression levels were analyzed using Spearman’s correlation; the linear regression equation, Spearman’s r [95% CI], and 2-tailed P values are given below each correlation graph. (D) Schematic representation of naive Mo treatment with PA and increasing concentrations of glucose. (E) Scatter plot representation of RT-qPCR expression of selected markers in healthy donor naive Mos treated for 18 hours (or 42 hours for ANGPTL4) with either BSA (unbound BSA, blue dots) or PA (BSA-bound PA, red dots) and various concentrations of glucose. Values represent the mean ± SEM of 4 independent cultures. Statistical differences were analyzed by 2-way ANOVA interaction, and P values for the PA and glucose treatments are given below each graph.

Figure 4. Lipid-associated MPs produce inflammatory cytokines.

(A) Schematic representation of naive Mo treatment with increasing concentrations of PA. (B) Nonlinear regression representation of the RT-qPCR relative expression of selected markers of healthy donor naive Mos treated for 42 hours with increasing concentrations of PA (0–500 μM). Values represent the mean ± SEM of 5 independent culture points. Goodness of fit is indicated by R² (0.9577, 0.9413, 0.7958, 0.9067, and 0.8298, respectively). (C) Schematic representation of 2-phase treatment of naive Mos to test the long-term effect of PA. (D) Scatter plot representation of RT-qPCR expression of PLIN2 and CXCL8 in healthy donor naive Mos subjected to the 2-phase treatment. Values represent the mean ± SEM of a minimum of 4 independent culture points. P values were determined by 1-way Welch’s ANOVA (P = 0.0010 for PLIN2; P = 0.0002 for CXCL8) followed by Dunnett’s T3 multiple-comparison test. (E) Schematic representation of the preparation CM of Mos. (F) Scatter plot of cytokines and growth factor protein concentrations in pg/mL detected by multiplex measurements in CM of naive Mos from a healthy donor differentiated with PA (PA-bound BSA, y axis) or BSA (unbound BSA, x axis) (left graph) or 25 mM glucose plus BSA (y axis) or 5 mM glucose plus BSA (x axis) (right graph).

Figure 5. Lipid-associated MPs show vasodegenerative properties.

(A) Schematic representation of the preparation of CtlCM and PAstimCM^PAfree from healthy donor Mos treated with PA (PA-bound BSA) or BSA (unbound BSA). (B and C) HUVECs were stimulated with either CtlCM or PAstimCM^PAfree. (B) Epifluorescence images of HUVEC nuclei stained with Hoescht (white). Scale bars: 100 μm. (C) Scatter plot of normalized nuclei counts. Values represent the mean ± SEM of 6 independent culture points. The P value was determined using Welch’s 2-tailed t test. (D–G) Capillary degeneration was quantified in an ex vivo assay (48). (E) Time-course of the mean ± SEM sprout number between day 4 (D4) and day 8 (D8) in the 3 treatment groups: basal medium (n = 10, black), CtlCM (n = 7, blue), and PAstimCM^PAfree (n = 7, red). (F) Violin plot of the log₂ FC of sprout numbers between paired day-6 and day-8 rings; dots represent individual aortic rings, and dashed lines represent the median and quartiles. P values were determined by 1-way Welch’s ANOVA test (P < 0.0001) followed by Dunnett’s T3 multiple-comparison test. (G) Aortic rings and sprouts treated with CtlCM or PAstimCM^PAfree on day 6 and stained with COL4 on day 8. Left: Epifluorescence micrographs of COL4 (white). Scale bars: 500 μm. Right: Higher-magnification confocal micrographs of COL4 (green). Scale bars: 200 μm. Nuclei were stained with Hoechst (blue). (H) Schematic of the biological changes in Mos after lipid exposure and their acquired properties.

Figure 6. Inhibition of PPARγ signaling normalizes the PA-induced lipid-associated phenotype.

(A) Scatter plot of RT-qPCR quantification of CXCL8 in THP-1 cells transfected with the indicated siRNA and treated with BSA (unbound BSA) or PA (PA-bound BSA). P values were determined by 1-way Welch’s ANOVA test (P = 0.0136) followed by Dunnett’s T3 multiple-comparison test. (B) Left: Graph of high-confidence predicted PPAR target genes (49) differentially expressed in naive Mos treated with PA. Middle and right: Heatmaps of the expression level (TPM) of curated PPARα target genes with a log₂ FC of 2 or higher (middle) and curated PPARγ target genes with a log₂ FC of 2 or higher (right). (C) Schematic of the stimulation of naive Mos with PA or BSA and PPAR modulators. (D–G) Scatter plots of RT-qPCR quantification of CXCL8 in healthy donor naive Mos treated with either PA or BSA and with (D) fenofibric acid PPARα agonist (P values for interaction, P = 0.0790; PA effect, P < 0.0001; and drug effect, P = 0.0760); (E) GW6471 PPARα antagonist (P = 0.0570, P < 0.0001, and P = 0.9448); (F) pioglitazone PPARγ agonist (P = 0.0027, P = 0.0002, and P = 0.0021); or (G) T0070907 (T007) PPARγ antagonist (all P < 0.0001). Interactions and P values and were determined by 2-way ANOVA followed by Tukey’s multiple-comparison test. (H) Scatter plots of RT-qPCR quantification of the indicated genes in healthy donor naive Mos treated with PA plus T007. P values were determined using a t test with Welch’s correction. (I) Representative images and scatter plot of TUNEL⁺ HUVECs after stimulation with CtlCM, PAstimCM^PAfree, or PAstim+T007-CM^PA&T007-free. P values were determined by 2-way Welch’s ANOVA (P < 0.0046) followed by Dunnett’s T3 multiple-comparison test. (J) Violin plot of the log₂ FC of sprout numbers between paired rings before and after stimulation with CtlCM, PAstimCM^PAfree, or PAstim+T007-CM^PA&T007free. Dashed lines indicate the median and quartiles. P values were determined by 1-way Welch’s ANOVA (P < 0.0001) followed by Dunnett’s T3 multiple-comparison test.