Early-life peripheral infections reprogram retinal microglia and aggravate neovascular age-related macular degeneration in later life

Abstract

Pathological neovascularization in age-related macular degeneration (nvAMD) drives the principal cause of blindness in the elderly. While there is a robust genetic association between genes of innate immunity and AMD, genome-to-phenome relationships are low, suggesting a critical contribution of environmental triggers of disease. Possible insight comes from the observation that a past history of infection with pathogens such as Chlamydia pneumoniae, or other systemic inflammation, can predispose to nvAMD in later life. Using a mouse model of nvAMD with prior C. pneumoniae infection, endotoxin exposure, and genetic ablation of distinct immune cell populations, we demonstrated that peripheral infections elicited epigenetic reprogramming that led to a persistent memory state in retinal CX3CR1+ mononuclear phagocytes (MNPs). The immune imprinting persisted long after the initial inflammation had subsided and ultimately exacerbated choroidal neovascularization in a model of nvAMD. Single-cell assay for transposase-accessible chromatin sequencing (scATAC-seq) identified activating transcription factor 3 (ATF3) as a central mediator of retina-resident MNP reprogramming following peripheral inflammation. ATF3 polarized MNPs toward a reparative phenotype biased toward production of proangiogenic factors in response to subsequent injury. Therefore, a past history of bacterial endotoxin-induced inflammation can lead to immunological reprograming within CNS-resident MNPs and aggravate pathological angiogenesis in the aging retina.

Keywords: Cellular immune response; Endothelial cells; Ophthalmology.

Figure 1. Prior peripheral infection predisposes to pathological angiogenesis in the retina.

(A) Time course of Chlamydia pneumoniae (Cpn) or mock infections starting at 7 weeks. Laser-induced CNV occurred at 16 weeks, euthanasia at 18 weeks. (B) Confocal images of isolectin B4–stained (IB4-stained) laser burns with FITC-dextran–labeled CNVs. Scale bars: 20 μm. Quantification of (C) CNV area, (D) laser impact area, and (E) FITC/IB4 ratio per laser burn, all on day 14; n = 19 burns (Mock), n = 16 burns (Cpn). (F) Confocal images of mononuclear phagocytes (MNPs) stained for ionized calcium-binding adaptor molecule 1 (IBA1). Scale bars: 20 μm. (G) IBA1-positive MNP counts on day 14; n = 19 burns (Mock), n = 16 burns (Cpn). (H) Endotoxin levels (endotoxin units/mL) 5 days after infection; n = 7 per condition. Data are presented as mean ± SEM. Student’s unpaired t test (C–E, G, and H) was used. *P < 0.05; **P < 0.01; ****P < 0.0001.

Figure 2. Prior systemic exposure to endotoxin predisposes to pathological angiogenesis in the retina.

(A) Time course of peripheral LPS stimuli, where C57BL/6J mice received 1 low-dose injection of LPS (1×LPS), 4 daily injections of LPS (4×LPS), or 4 daily PBS injections (PBS) at 7 weeks old. Laser-induced CNV occurred at 11 weeks and euthanasia at 13 weeks. (B) Flow cytometry plots and percentage of viable blood monocytes in PBS (n = 5), 1×LPS (n = 5), and 4×LPS (n = 6) groups. (C) Plots and percentage of viable Ly6C^hi monocytes in PBS (n = 5), 1×LPS (n = 5), and 4×LPS (n = 6) groups. (D) Plots and percentage of viable MNPs in PBS, 1×LPS, and 4×LPS groups (n = 12 per condition). (E) Plots and percentage of viable CX3CR1⁺ microglia in PBS, 1×LPS, and 4×LPS groups (n = 9 per condition). (F) Il1b, Il6, Tgfb1, and Vegfa mRNA expression in retina/RPE-choroid-sclera complexes 4 weeks after LPS or PBS injections; n = 8 per condition. (G) CNV confocal imaging of IB4 and FITC-dextran from PBS, 1×LPS, and 4×LPS groups. Scale bars: 20 μm. Quantification of (H) CNV area, (I) laser impact area, and (J) FITC/IB4 ratio per laser burn on day 14; n = 39 burns (PBS), n = 29 burns (1×LPS), n = 23 burns (4×LPS). Data are presented as mean ± SEM. One-way ANOVA with Tukey’s multiple-comparison test (B–F and H–J) was used. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Figure 3. Prior systemic exposure to endotoxin mitigates retinal neuroinflammation.

(A) Time course of peripheral LPS stimuli, where C57BL/6J mice received injections of LPS once (1×LPS), 4 daily injections of LPS (4×LPS), or 4 daily injections of PBS (PBS) at 7 weeks. Laser-induced CNV occurred at 11 weeks and euthanasia 3 days or 2 weeks later. Naive mice received PBS injections but no laser burns. (B–H) mRNA expression in retina/RPE-choroid-sclera complexes 3 days after CNV induction (PBS, 1×LPS, and 4×LPS) of Il1b (B), Tnf (C), Il6 (D), Vegfa (E), Tgfb1 (F), Tlr4 (G), and Aif1 (H): n = 6 (naive), n = 7 (PBS), n = 7 (1×LPS), n = 7 (4×LPS). (I and J) mRNA expression in RPE-choroid-sclera complexes 3 days after CNV induction (PBS, 1×LPS, and 4×LPS) or without CNV induction (naive) relative to PBS of (I) Tnf and (J) Vegfa: n = 3 (naive), n = 8 (PBS), n = 8 (1×LPS), n = 9 (4×LPS). (K and L) Flow cytometric analyses of whole retinas and RPE-choroid-sclera complexes 3 days after CNV induction. Plots and percentage of viable mononuclear phagocytes (MNPs) (K) and CX3CR1⁺ MNPs (L) in naive (n = 9), PBS (n = 6), 1×LPS (n = 6), and 4×LPS (n = 6) groups. (M) Confocal images of IBA1-stained MNPs on day 14 of PBS-, 1×LPS-, and 4×LPS-treated groups. Scale bars: 20 μm. (N) Number of IBA1-positive MNPs around laser impact area on day 14; n = 40 burns (PBS), n = 29 burns (1×LPS), n = 23 burns (4×LPS). Data are presented as mean ± SEM. One-way ANOVA with Tukey’s multiple-comparison test (B–L and N) was used. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Figure 4. CX3CR1⁺ myeloid cells in the retina mediate proangiogenic memory after systemic exposure to LPS.

(A) Time course of Cx3cr1^CreER/+ and Cx3cr1^CreER/+:R26^iDTR/+ mice injected with 4×LPS or PBS at 7 weeks. For both Cx3cr1^CreER/+ and Cx3cr1^CreER/+:R26^iDTR/+ mice, tamoxifen (TAM) was administered i.p. starting at 6 weeks and diphtheria toxin intravitreally (ivt) at week 11 and 12. Laser-induced CNV occurred at 11 weeks, euthanasia at week 13. (B) CNV confocal images of IB4, FITC-dextran, and IBA1 staining from Cx3cr1^CreER/+ and Cx3cr1^CreER/+:R26^iDTR/+ mice with either 4×LPS or PBS injections. Scale bars: 20 μm. (C–F) Quantification of CNV area (C), IB4-stained laser impact area (D), FITC/IB4 ratio per laser burn (E), and number of IBA1-positive MNPs (F) on day 14; n = 24 burns (Cx3cr1^CreER/+ + PBS), n = 38 burns (Cx3cr1^CreER/+ + 4×LPS), n = 31 burns (Cx3cr1^CreER/+:R26^iDTR/+ + PBS), n = 31 burns (Cx3cr1^CreER/+:R26^iDTR/+ + 4×LPS). Data are presented as mean ± SEM. One-way ANOVA with Tukey’s multiple-comparison test (C–F) was used. *P < 0.05; ****P < 0.0001.

Figure 5. Adaptive immunity is not required for the proangiogenic effect of LPS-induced immune memory.

(A) Time course of Rag1^–/– mice injected with either 4×LPS or PBS at 7 weeks. Laser-induced CNV occurred at 11 weeks, euthanasia at week 13. (B) CNV confocal images of IB4, FITC-dextran, and IBA1 staining from Rag1^–/– + PBS and Rag1^–/– + 4×LPS mice. Scale bars: 20 μm. (C–F) Quantification of CNV area (C), IB4-stained laser impact area (D), FITC/IB4 ratio per laser burn (E), and number of IBA1-positive MNPs (F) on day 14; n = 31 burns (PBS), n = 37 burns (4×LPS). Data are presented as mean ± SEM. Student’s unpaired t test (C–F) was used. *P < 0.05.

Figure 6. Peripheral exposure to endotoxin induces epigenetic reprogramming of CX3CR1⁺ retina-resident microglia.

(A) Schematic of the experimental workflow for the single-cell assay for transposase-accessible chromatin sequencing (scATAC-seq) of FACS-isolated CX3CR1⁺ retinal cells from mice 3 days after CNV induction, preconditioned with either PBS or 4×LPS 1 month before, or naive retinas without CNV induction, preconditioned with PBS. (B) UMAP projections of scATAC-seq profiles of CX3CR1⁺ retinal myeloid cells (2244 cells from 20 mice) according to sample origin (left) and results of unbiased clustering using the Leiden algorithm (right). Each dot represents an individual cell. (C) Bar graphs of sample components of each cluster in B. MΦ, macrophage; Mo, monocyte. (D) Bar graphs of results of GSEA using differentially accessible regions (DARs) for each microglia cluster enriched in naive, PBS, and 4×LPS groups (C1, C2, and C3, respectively) in closed chromatin regions and open chromatin regions. Inflammation-related pathways are highlighted.

Figure 7. Peripheral exposure to endotoxin induces transcriptional reprogramming of myeloid cells.

(A) Experimental time course schematic of LPS in vivo and in vitro manipulations (B–G). C57BL/6J mice were treated with 4×LPS or PBS at 7 weeks of age and BM cells were collected at 11 weeks of age. BM cells were differentiated to BM-derived macrophages (BMDMs) with M-CSF, and BMDMs were harvested after a secondary stimulation with LPS or PBS for 4 hours. Total RNA was extracted and analyzed for gene expression by bulk RNA-seq or qPCR. (B) Volcano plot obtained from DESeq2 analysis of LPS-restimulated BMDMs from 4×LPS-prereated mice as compared with LPS-restimulated BMDMs from PBS-pretreated mice. (C) Heatmap of the top 60 most differentially expressed genes of LPS-restimulated BMDMs from 4×LPS-pretreated mice as compared with LPS-restimulated BMDMs from PBS-pretreated mice. (D) Results of GSEA of Hallmark gene sets showing those enriched in LPS-restimulated BMDMs from mice pretreated with 4×LPS as compared with LPS-restimulated BMDMs from PBS-pretreated mice (FDR < 0.1 and a nominal P value < 0.05). A positive normalized enrichment score (NES) value indicates enrichment in the 4×LPS-treated mice. (E) GSEA enrichment plots for genes related to inflammatory response and angiogenesis in LPS-restimulated BMDMs from mice pretreated with 4×LPS as compared with LPS-restimulated BMDMs from PBS-pretreated mice; n = 3. (F and G) mRNA expression in BMDMs from PBS-pretreated and 4×LPS-pretreated mice with (G) or without LPS restimulation (F): Il1b, Il6, Tnf, Tgfb1, Vegfa, Pdgfb, Postn, and Col3a1; n = 6 for all conditions. Data are presented as mean ± SEM. Comparisons between groups were analyzed using Student’s unpaired t test (F and G). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Figure 8. Prior peripheral exposure to endotoxins shifts myeloid cells toward proangiogenic polarization.

(A) Representative flow cytometry plots of M1- and M2-like macrophages in BMDMs of PBS-pretreated mice and 4×LPS-pretreated mice. (B and C) Quantification of M2-like macrophages (F4/80⁺CD11b⁺CD11c⁻CD206⁺) (B) and M1-like macrophages (F4/80⁺CD11b⁺CD11c⁺CD206⁻) (C) in BMDMs of PBS-pretreated mice and 4×LPS-prereated mice (n = 6). (D) Schematic representation of ex vivo choroid sprouting assay cocultured with BM monocytes (BM-Mo). C57BL/6J mice were treated with 4×LPS or PBS at 7 weeks of age and were subjected to laser burns at 11 weeks. BM cells were collected 3 days after laser burn and BM-Mo were isolated using immunomagnetic negative selection. Choroid pieces (RPE-choroid-sclera from peripheral retina of 5-week-old C57BL/6J mice) were seeded into 24-well plates containing Matrigel and cocultured with BM-Mo using Transwell inserts. (E) Representative images of choroid explants at 2 and 3 days of coculture with BM-Mo from PBS-pretreated and 4×LPS-pretreated mice, and without BM-Mo. (F and G) Quantitation of sprouting area at 2 (F) and 3 days (G) of coculture with BM-Mo from each group when compared with no BM-Mo. At 2 days, n = 19 (No BM-Mo), 14 (PBS BM-Mo), n = 15 (4×LPS BM-Mo). At 3 days, n = 16 (No BM-Mo), 14 (PBS BM-Mo), n = 14 (4×LPS BM-Mo). Data are presented as mean ± SEM. Comparisons between groups were analyzed using Student’s unpaired t test (B and C) or 1-way ANOVA with Tukey’s multiple-comparison test (F and G). *P < 0.05; **P < 0.01.

Figure 9. Peripheral exposure to endotoxins modulates myeloid cell response via ATF3 deregulation.

(A) Volcano plot of accessible regions with differentially accessible regions (DARs; defined by an FDR adjusted P value < 0.05, total of 51 DARs) identified between comparisons of microglial 4×LPS cluster (C3) versus microglial PBS cluster (C2) as found by CX3CR1⁺ retinal myeloid cell scATAC-seq data. (B) Cis-coaccessibility network (CCAN) links between the ATF3 promoter and distal sites in the surrounding region generated by subset analysis by cluster. Connections from microglial PBS cluster (C2) and microglial 4×LPS cluster (C3) are shown, with a minimum coaccessibility score of 0.3. (C) Chromatin accessibility in the promoter region of Atf3 in BMDMs from PBS-pretreated and 4×LPS-pretreated mice as analyzed by qPCR. n = 4 (PBS), n = 4 (4×LPS). (D) Representative immunoblots showing p-NF-κB, total NF-κB, p-c-JUN, total c-JUN, and ATF3 expression in BMDMs from PBS-pretreated and 4×LPS-pretreated mice with and without LPS restimulation. (E) Schematic representation of knockdown of Atf3 gene expression in BMDMs using siRNA. C57BL/6J mice were treated with 4×LPS or PBS at 7 weeks of age and BM cells were collected at 11 weeks of age. BM cells were differentiated into BMDMs with M-CSF. BMDMs were transfected with Atf3 or control siRNA for 24 hours and then restimulated with LPS for 4 hours. RNA was extracted for qPCR analysis. (F–J) mRNA expression in BMDMs transfected with Atf3 or control siRNA from PBS-pretreated and 4×LPS-pretreated mice with LPS restimulation: Atf3 (F), Tnf (G), Il6 (H), Vegfa (I), and Pdgfb (J); n = 6 for all groups. Data are presented as mean ± SEM. Comparisons between groups were analyzed using Student’s unpaired t test (C) or 1-way ANOVA with Tukey’s multiple-comparison test (F–J). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.