Autophagy Regulates Müller Glial Cell Inflammatory Activation

Abstract

Purpose

We tested whether Müller cells utilize autophagy to support immune privilege in the eye.

Methods

The essential autophagy gene Atg5 was deleted in retinal Müller cells. Inflammation was induced by intravitreal injection of lipopolysaccharide (LPS) that was monitored by hematoxylin and eosin (H&E) staining, immunofluorescent confocal microscopy, and flow cytometry. Single-cell RNA sequencing was performed on retinal Müller cells isolated from control and Atg5-deficient mice. Markers of Müller cell gliosis were assessed, and cytokine production in the eye was measured. Small interfering RNA knockdown techniques were used to examine LPS-induced inflammatory pathways in culture.

Results

We observed increased and prolonged intraocular inflammation when Müller cells were autophagy (Atg5) deficient. Müller cell gliosis was significantly increased, and the retinae contained increased inflammatory mediators. Gene expression analysis revealed a heterogeneous response to LPS in Müller cells, revealing two states of activation. The normal retinae contained both basal and activated Müller cells, whereas the autophagy-deficient retinae contained only activated cells. Analysis of the gliosis markers glial fibrillary acidic protein (Gfap) and lipocalin-2 (Lcn2) confirmed this heterogeneity, as in control eyes basal and activated (gliotic) Müller glia were observed; however, with autophagy deficiency, all Müller cells were gliotic. Activated cells were largely indistinguishable between autophagy-sufficient and -deficient Müller cells. In cultured Müller cells, knockdown of Atg5 resulted in heightened mechanistic target of rapamycin (mTOR) activation, increased Gfap expression, and upregulated cytokine/chemokine production in response to LPS.

Conclusions

Autophagy regulates the activation state of Müller cells in response to LPS. Thus, autophagy restrains cellular activation and inflammation, supporting immune privilege by preventing excessive and potentially destructive immune responses.

Inflammatory responses and cellular immune reactions are the centerpiece of host defense against pathogenic insults. These reactions, although vital to clearing dangerous invaders, are often associated with nonspecific injury to nearby tissue. Most organ systems tolerate such inflammation without permanent consequences; thus, these mechanisms are generally effective. There are, however, sites in the body where bystander inflammation can threaten organ integrity, placing the entire organism at risk. Examples include the eye, the brain, and the reproductive organs, where even minor episodes of inflammation threaten vision, information processing, and reproduction, respectively. Immune privilege is the process protecting these vital organs such that immune reactions either do not proceed or are inhibited locally and systemically. In the eye, the breakdown of immune privilege can contribute to bystander damage from infection, the rejection of corneal grafts, the development of uveitis, and even retinal degenerative diseases.^1–5 Immune privilege is maintained by multiple mechanisms, including the blood–retinal barrier (BRB), the expression of membrane‐bound cell death–inducing ligands (FasL, Trail, Pd-l1) and complement regulatory proteins, the presence of soluble immunosuppressive factors (neuropeptides, transforming growth factor β) in ocular fluids, and the ability to inhibit systemic immunity to antigens released from a damaged eye (termed immune deviation). Although overt inflammation is a well-known feature in diseases such as uveitis, it is now recognized that the pathology of degenerative eye diseases such as age-related macular degeneration, glaucoma, and diabetic retinopathy, in which inflammation is not as robust, also have immunological components.³ Because of the powerful influence of immune privilege on intraocular inflammation, it is important to understand how this process modulates all retinal diseases if effective therapies are to be developed.

Müller cells are the major glial cell of the retina, spanning its entire thickness and thereby serving an important structural role. Müller cell processes interact with virtually every retinal cell and participate in key processes such as maintaining the BRB, neurotransmitter recycling, neuroprotection, and phagocytosis. Müller cells become activated (gliotic) during pathogenic insults, producing growth factors, cytokines, and neuroprotective molecules that limit the extent of damage to the retina. In contrast, excess gliosis can contribute to neurodegeneration by impeding tissue repair, downregulation of important glial functions such as barrier maintenance, and neurotransmitter recycling.⁶ Heterogeneity between neighboring Müller cells with respect to gliosis has also been observed⁷ but never sufficiently explained. Müller cells can also suppress immune responses,^8,9 suggesting that they could be key participants in immune privilege; however, the mechanisms are not known.

Macroautophagy (hereafter autophagy) is a lysosomal degradation pathway that maintains cellular homeostasis under basal and stress conditions by catabolizing cellular constituents to produce energy and building blocks. Degradation by autophagy also restricts viral infections and bacterial replication and delivers immunostimulatory determinants (Toll-like receptor and T-cell receptor ligands) to the immune system.¹⁰ The loss of autophagy can lead to damaging inflammation in many tissues.¹¹ Although the role of autophagy in the retina has been studied,^12,13 its role in Müller cell function, homeostasis, gliosis, and immune privilege is not known.

Because immune privilege, Müller cells, and autophagy all have immunomodulatory properties critical to ocular homeostasis, we tested the hypothesis that Müller cells utilize the autophagy pathway to support the anti-inflammatory nature of the eye (i.e., immune privilege). Utilizing an acute endotoxin uveitis model, we demonstrated that the absence of Müller cell autophagy leads to enhanced inflammation, a sustained gliotic response and revealed a novel perspective regarding Müller glial heterogeneity. These findings suggest an important role for autophagy in restraining inflammatory responses in the retina, thereby contributing to its immune privilege.

Materials and Methods

Mice

Slcla3-cre/ERT/Glast mice were obtained from The Jackson Laboratory (#012586; Bar Harbor, ME, USA). This strain contained the rd8 mutation¹⁴ that was removed by backcrossing to C57BL/6J (#000664; The Jackson Laboratory). Atg5^f/f were a gift from Noboru Mizushima, MD, PhD (Tokyo, Japan).¹⁵ Mice were housed in a barrier facility operated under a standard 12-hour light/12-hour dark cycle in facilities maintained by the Division of Comparative Medicine of Washington University. Experiments were carried out in strict accordance with recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health and in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Littermate controls, Atg5^f/f (cre^–), were used in all experiments. Group sizes are indicated in the figure legends.

Antibodies and Reagents

All antibodies and reagents used are listed in Supplementary Table S1.

Tamoxifen Injection

The 4-hydroxy-tamoxifen (4-HT) was prepared by dissolving it in sunflower oil at 37°C for 24 hours with rotation in the dark at a concentration of 20 mg/mL. We found that three consecutive daily doses of 50 µg/g body weight 4-HT in 4-week-old mice was optimal when we examined Müller cell autophagy 28 days after the 4-HT injection.¹⁶

Lipopolysaccharide Injection for EIU

Lipopolysaccharide (LPS; 0.5 µg/eye) was delivered to the vitreous cavity in 3 µL PBS using a Hamilton syringe fitted with a 33-gauge needle as described previously.¹⁷

Histological Analysis

Eyes were removed and fixed in 10% formalin overnight at 4°C. Tissues were then embedded in paraffin, and sagittal sections of 5-µm thickness were cut starting at the optic nerve and collected every 100 µm. Three sections were collected at each cutting point, for a total of 12 sections, and stained with hematoxylin and eosin (H&E). Endotoxin-induced uveitis (EIU) grading was performed on H&E sections of whole eyes from Atg5^f/f and Atg5^i∆Müller by a masked observer. The accumulative values for each sample were collected based on the score assigned by each of the following criteria: presence or absence of inflammatory cells in the anterior chamber (0, none; 1, <10; 2, >10) or posterior chamber (0, none; 1, ≤10; 2, >10); retinal edema (0, no; 1, yes); red blood cells within retinal tissue (0, no; 1, yes); inflammatory cells within retinal tissue (0, no; 1, yes); and tissue destruction (0, none; 1, moderate/localized; 2, severe). Four to seven eyes per group were examined, 12 sections per eye (420 × 180 µm field of view around the optic nerve).

Immunofluorescence Staining

The superior sides of eyes were marked with a permanent marker and then enucleated. Tissues were prepared and stained as previously described.¹⁸ Images were captured using a FluoView FV1000 confocal microscope (Olympus, Tokyo, Japan). Samples were collected from each group of mice.

Mean Fluorescence Intensity

Confocal RGB images were transformed into 8-bit grayscale images, and the total mean fluorescence intensity of six regions of interest (ROIs) in each slide was analyzed with ImageJ software (National Institutes of Health, Bethesda, MD, USA) as described previously.¹⁹ The mean grayscale value of background areas outside of the ROIs were also measured to correct for non-specific staining. Data represent the mean ± standard error.

SDS-PAGE and Western Blot Analysis

Retina and transfected cell lysates were prepared as previously described.²⁰ Image Studio software (LICORbio, Lincoln, NE, USA) was used to quantify band intensity relative to actin or glyceraldehyde-3-phosphate dehydrogenase (Gapdh) loading control for both chemiluminescence and fluorescence western blot analyses. Data represent mean ± standard error.

Cytokine Assays

Cytokine assays were performed on retinal lysates (prepared as for the western blot analyses) and tissue culture supernatants using a bead-based multiplex immunoassay performed by the Bursky Center for Human Immunology and Immunotherapy Programs at Washington University.

Flow Cytometry

Flow cytometry of inflammatory cells was performed on retina single-cell suspensions prepared using a Papain Dissociation System (Worthington Biochemical, Lakewood, NJ, USA) with the protocol provided. Samples were run on a FACScan (Becton Dickinson, Franklin Lakes, NJ, USA), and cells were collected and analyzed using CellQuest and Rainbow software (Becton Dickinson). A total of four to six retinal samples were collected for each group, and the data represent the mean ± standard error.

Single-Cell RNA Sequencing

Single-cell suspensions were generated from pooled isolated retinas for each group (n = 6) with a Papain Dissociation System using the protocol provided. Single-cell RNA sequencing (scRNA-seq) was performed using the 10x Genomics (Pleasanton, CA, USA) platform and reagents (Chromium Single Cell 3' Reagent Kits, v3.1 Chemistry). Libraries were sequenced on a NovaSeq 6000 sequencing system (Illumina, San Diego, CA, USA) at the McDonnell Genome Institute at Washington University School of Medicine. Raw fastq files were processed and mapped using Cell Ranger 6.0.0 (10x Genomics). Further analysis was performed using the R package Seurat v5.2.1 (The R Foundation for Statistical Computing, Vienna, Austria).²¹ The full code for this analysis is available at https://github.com/p-ruzycki/Atg5KO_LPS. Pathway enrichment analysis was conducted using Gene Set Enrichment Analysis (GSEA) v4.3.3.

Tissue Culture and Cell Line

In vitro Atg5 knockdown studies were conducted in the rat Müller cell line (rMC1) using transfection protocols previously described for RPE cells.²⁰ Prior to the start of any LPS assay, 1 mL of fresh media was added to each well. For LPS assays, 5 µL of a 1-mg/mL LPS solution was added. For western blot analysis, adherent cells were washed with ice-cold PBS (pH 7.4) and lysed with 200 µL of ice-cold lysis buffer (used to prepare retinal lysates). Each well was then harvested and lysates prepared as previously described.²⁰

Statistical Analysis

The Mann–Whitney U test was performed (**P < 0.05) for EIU. A paired two-tailed Student's t-test was used to test significance between treatments for western blot, mean fluorescence intensity, and fluorescence-activated cell sorting (FACS). P < 0.05 was considered statistically significant.

Data Availability

Raw and processed scRNA-seq data are available in the Gene Expression Omnibus (accession no. GSE293132). Analyzed data can be accessed via the Broad Institute Single Cell Portal (accession no. SCP3003; https://singlecell.broadinstitute.org/single_cell/study/SCP3024).

Results

Generation of Atg5-Deficient Müller Glial Cells

We examined the role of autophagy in Müller cells by deleting the essential autophagy gene Atg5. To do this, we crossed the Müller cell-specific Glast-Cre-ERT mouse with a strain carrying an Atg5 conditional allele (Atg5^f/f).¹⁵ Three consecutive daily doses of 50 µg/g body weight 4-HT in 4-week-old mice was optimal to block Müller cell autophagy (Atg5^i∆Müller) when examined 28 days after 4-HT injection. The effectiveness of autophagy ablation was assessed by immunofluorescent confocal microscopy (IFCM) using p62/Sqstm1 accumulation as a surrogate (Atg5-specific antibodies have proven unreliable for this purpose). We observed that p62/Sqstm1 was increased in Müller cells of Atg5^i∆Müller mice, demonstrating that autophagy was disrupted (Supplementary Fig. S1). Unless otherwise stated, all uveitis experiments began 28 days after the 4-HT injections (8-week-old mice). During our experiments, we did not observe changes in retinal morphology, loss of function (by electroretinography), or compromise of the BRB in unchallenged Atg5^i∆Müller mice (data not shown).

EIU Model

EIU is an animal model of acute intraocular inflammation induced by injection of the LPS component of the Gram-negative bacterial cell wall into the eye. EIU is a self-limiting inflammation that peaks at 18 to 24 hours and resolves in the next 24 to 48 hours. Although cited as a model of anterior segment inflammation, injection directly into the vitreous cavity leads to an acute posterior uveitis.²² We injected 0.5 µg LPS into the vitreous cavity of control (Atg5^f/f) and Atg5^i∆Müller mice and assessed inflammation at 24, 48, and 72 hours and at 5 days by H&E staining. Figure 1 shows representative sections of the optic nerve and mid-ventral regions of the retina. Although inflammation was detected in control eyes (Fig. 1A, 24 hours), it had diminished by 48 hours and was resolved by 72 hours, demonstrating the powerful role of immune privilege in the model. No inflammatory cells were detected at 72 hours or 5 days post LPS in control eyes. In contrast, inflammation was much more severe in Atg5^i∆Müller eyes at 24 hours and 48 hours (Fig. 1B). Inflammation persisted at 72 hours and was still evident at 5 days. In addition, there was evidence of focal retinal damage associated with inflammatory cell infiltration in some Atg5^i∆Müller eyes (Fig. 1B, 5 days mid-ventral). We quantified vitreous inflammation at each time point using an accumulative scoring system we developed (see Materials and Methods). These data demonstrate that inflammation was more severe in the Atg5^i∆Müller eyes compared to control, even at 5 days after injection of LPS.

Figure 1.

Increased inflammation in the presence of Atg5-deficient Müller glial cells. (A, B) H&E stains from control (Atg5^f/f) mice (A) and Atg5^i∆Müller mice (B) documenting the inflammatory infiltrates in the retina vitreous cavity at 24, 48, and 72 hours and at 5 days following LPS injection (n = 4–7 mice per group). Shown are representative sections of the optic nerve and mid-ventral regions of the retina at each time point. GCL, ganglion cell layer; ONL, outer nuclear layer. Arrows show retinal infiltration of inflammatory cells. *Vitreous inflammation. Scale bar: 25 µm.

We examined the makeup of the inflammatory infiltrate in the retinae isolated from the eyes of control and Atg5^i∆Müller mice by flow cytometry (Figs. 2B–D). Cells were gated on CD45 and/or CD11b and then analyzed for neutrophils (Ly6G), macrophages (F4/80, Ly6C^low/int), and inflammatory monocytes (F4/80⁺, Ly6G⁺). At 24 hours, an increased number of neutrophils and inflammatory monocytes were detected in the Atg5^i∆Müller retinae, but no obvious alterations in subpopulation composition were observed compared to control. The number of CD45⁺ and F4/80⁺ macrophages (Fig. 2C) did not attain statistical significance. Analysis at 48 and 72 hours gave similar results (not shown).

Figure 2.

Quantification of inflammation in Atg5^f/f and Atg5^i∆Müller mice. Eyes were injected with 0.5 µg LPS. EIU grading was performed on H&E sections of whole eyes from Atg5^f/f and Atg5^i∆Müller mice by a masked observer at 24, 48, and 72 hours and at 5 days, as described in the text. (A) The accumulative values for each sample were collected based on the score assigned by each of the following criteria. *P < 0.05, Mann–Whitney U test. (B–D) Analysis by flow cytometry 24 hours after LPS or PBS injection for neutrophils (CD45⁺, Ly6G⁺), macrophages (CD45⁺, F4/80⁺), and inflammatory monocytes (CD45⁺, F4/80⁺, Ly6G⁺). *P < 0.05; n = 4 to 6 retinae per group.

We then examined cytokine and chemokine changes in the local microenvironment to assess the possible contribution of Müller cell autophagy (Fig. 3). When we compared control with Atg5^i∆Müller mice, we noted that, at 6 hours, Il-6, leukemia inhibitory factor (Lif), chemokine (C-X-C motif) ligand 2 (Cxcl2), and C-C motif chemokine ligand 2 (Ccl2) were elevated. At 24 hours Il-1β, Il-6, Lif, and Ccl2 were increased in Atg5^i∆Müller eyes. By 48 hours, only Il-1β and Lif remained elevated. Thus, cytokines that attract immune cells (Ccl2, Cxcl1) and promote inflammation (Il-6, Il-1β) were elevated in the eye when Atg5-deficent Müller cells were present. Lif is a neuroprotective and anti-inflammatory cytokine in the Il-6 family that increases in response to the heightened inflammation,^23,24 but its role in EIU is currently unknown. These results (Figs. 1²–3) demonstrate compromised immune privilege and increased inflammation in presence of autophagy-deficient Müller glial cells.

Figure 3.

Cytokines in retinal lysates. Following LPS injection into the vitreous cavity of Atg5^f/f and Atg5^i∆Müller mice, retinal lysates were prepared at 6, 24, and 48 hours after the LPS injection and tested for cytokines and chemokines by Luminex arrays. *P < 0.05; n = 5 to 8 samples per group. (A) Il-1β (B) Il-6 (C) Lif (D) Cxcl2 (E) Ccl2.

Protein Expression in Retinal Lysates

We examined retinal lysates to detect changes in retinal/Müller glia proteins. Although we did not detect overt changes in Kir4.1 (not shown), cellular retinaldehyde-binding protein (Cralbp; not shown), or glutamine synthetase (Glul) (Figs. 4A, 4D), we detected significant changes in p62/Sqstm1 (Figs. 4A, 4B) and glial fibrillary acidic protein (Gfap) (Figs. 4A, 4C) in Atg5^i∆Müller mice compared to control (Fig. 4). In addition, we detected the accumulation of Lc3b-I in retinal lysates, likely from loss of autophagy in the Atg5^i∆Müller Müller cells (Figs. 4A, 4E). Thus, altered autophagy and increased gliosis can be detected in whole retinal lysates. Note the increase in Gfap banding detected in response to LPS, particularly in the Atg5^i∆Müller retina. This is further discussed below.

Figure 4.

Western blots for retinal proteins in Atg5^f/f and Atg5^i∆Müller retinae. Mice received an intravitreal injection of either PBS (−) or LPS (+), and retinal lysates were prepared 6, 24, and 48 hours later. (A) Blots were probed for p62/Sqstm1, Gfap, Glul, and Lc3b. Two representative samples for each mouse strain are shown from a total of three blots performed. (B–E) Relative expression of p62/Sqstm1, Gfap, Glul, and Lc3b from three experiments. *P < 0.05; n = 6 to 10 samples. Data were obtained using chemiluminescence western blotting. The legend applies to the graphs in B to E.

Müller Cell Gliosis

Increased Gfap expression is a hallmark of reactive gliosis in Müller cells (which do not constitutively express this protein). Gliosis is a neuroprotective response that can limit the tissue damage in response to retina damage and inflammation.^6,7 We then monitored Gfap expression by IFCM and western blotting in control and Atg5^i∆Müller retinae over the course of 5 days after LPS injection. As shown in Figure 5A, Gfap was increased at 24 and 48 hours but returned to control levels by 72 hours in the control mice. In Atg5^i∆Müller mice, Gfap increased substantially at 6, 24, 48, and 72 hours. Gliosis returned to normal levels by 5 days after LPS injection. Increased protein expression was confirmed by western blotting at each time point (Fig. 5B). Quantification of Gfap western blots verified increased protein levels (Fig. 5C). Also notable in the Atg5^i∆Müller retinae was an increase in the number of bands detected by the Gfap antibody. Gfap cleavage (or the presence of breakdown products) is indicative of increased cellular activation as has been observed in astrocytes.²⁵ This suggests that Müller glia are more highly activated with Atg5 deficiency.

Figure 5.

Gfap expression in Atg5^f/f and Atg5^i∆Müller retinae. Mice were injected with either PBS or LPS in the vitreous cavity, and eyes were harvested at 6, 24, 48, and 72 hours and at 5 days. (A) Representative confocal images of the retina stained for Gfap (n = 3). (B) Western blot for Gfap expression at the same time points. (C) Quantification of Gfap expression in fluorescent western blots from two separate experiments. *P < 0.05; n = 4 to 8. Scale bar: 25 µm.

Gene Expression Analysis

Increased ocular inflammation and Müller cell activation prompted us to explore the gene expression profiles in retinal cells from control and Atg5^i∆Müller mice by performing scRNA-seq. We chose the 24-hour time point, as there are substantial differences in the inflammatory response between the two strains (Fig. 1). Primary analysis clearly defined all major retinal neuronal cell types (Supplementary Figs. S2A, S2B), and subsequent subclustering of non-neurons resolved Müller glia, epithelial, and immune response–related cells (T cells, monocytes, and macrophages) (Supplementary Figs. S2C–S2E). Direct comparison of transcriptional profiles within immune cells did not identify differences in the identity or inflammatory state of cells that had infiltrated Atg5^i∆Müller versus Atg5^f/f retinae at this time point (data not shown), consistent with the FACS analysis of the LPS-induced inflammatory cell infiltrates (Fig. 2).

We then focused on the expression profiles of only the isolated Müller glia and identified two clusters that shared expression of many Müller cell markers but displayed distinct representation of cells from Atg5^f/f and Atg5^i∆Müller animals (Figs. 6A, 6B). Although cells from the Atg5^f/f mice were present in both clusters 1 and 2, Atg5^i∆Müller cells were essentially restricted to cluster 1. Further analysis of cells within these two Müller glial expression states identified hundreds of differentially expressed genes (Fig. 6C, Supplementary Table S2), that by Gene Ontology analyses describe cluster 1 as a more activated/gliotic Müller cell transcriptional state (Supplementary Table S3). Cluster 1 was enriched for genes involved in ribosome biogenesis, cellular aging, and inflammatory and interferon responses. These activated Müller glia (cluster 1) from both control and Atg5^i∆Müller datasets gained expression of the canonical gliosis marker Gfap, as well as known immunomodulatory genes such as the cytokines/chemokines Ccl2, Ccl7, Lcn2, Cxcl2, and Lif (Figs. 6C–E). Activated Müller glia (cluster 1) lost expression (or had significantly reduced expression) of genes important for known Müller cell functions that were retained in the cluster 2 fraction (basal) of control Müller cells such as Kcnj10 (Kir4.1), Aqp4, Car14, Glul, Rlbp1⁷, and Crb1 (Figs. 6C–E). Finally, direct transcriptomic comparison showed that Atg5^i∆Müller and Atg5^f/f cluster 1 activated cells were largely indistinguishable (Supplementary Fig. S2F, Supplementary Table S4). These data together suggest that the loss of autophagy within Müller glia does not impact the ability of Müller glia to mount a gliotic response but rather impacts the decision of individual cells to respond to inflammatory stimuli resulting in all Müller cells becoming activated.

Figure 6.

Müller glial gene expression by scRNA-seq. (A) Uniform manifold approximation and projection (UMAP) of the scRNA-seq identified two transcriptionally distinct states of Müller glia upon stimulation by LPS. (B) Bar plot shows the proportion of cells within each transcriptionally distinct cluster in control (Atg5^f/f) and Atg5^i∆Müller samples. (C) Volcano plot displays differentially expressed genes between cells in cluster 1 versus cluster 2. The x-axis represents the log₂ fold change (FC; cluster 1 vs. 2), and the y-axis shows the –log₁₀ false discovery rate (FDR). Significantly differentially expressed genes (FC ≥ 2 and FDR ≤ 0.05) are shown in red. Genes of interest are labeled and highlighted in black, and non-significant genes are shown in gray. (D) UMAP feature plots show the expression levels of differentially expressed genes across Müller glia states. The top panel displays Crb1, Glul, and Rlbp1, which are enriched in cluster 2 cells, and the bottom panel shows Gfap, Lif, and Ccl2, which are enriched in cluster 1 cells. The color intensity represents expression levels, with darker shades indicating higher expression. (E) Heat map displays the relative expression of genes of interest between cells within the indicated genotype/clusters. Colors represent row Z-scores, with red and blue indicating relatively higher and lower expression, respectively.

Heterogeneity Detected by Protein Expression

We more closely examined this heterogeneity utilizing co-stains for gliosis along with the known Müller cell marker Cralbp. As shown in Figure 7, we found that autophagy ablation in control retinas revealed increased Gfap/Cralbp stain within a subpopulation of Cralbp⁺ cells after LPS injection (Fig. 7C). However, we observed substantial upregulation of Gfap/Cralbp co-staining in the Atg5^i∆Müller eyes injected with LPS (Fig. 7D). We further documented the heterogeneous gliotic response by examining lipocalin-2 (Lcn2). Lcn2 is upregulated during inflammation and can be also used a marker for gliosis.^26,27 Figure 8D shows increased Lcn2 in response to LPS in the controls, but many more Lcn2⁺ cells were observed in the Atg5^i∆Müller mice. We quantified expression to emphasize this point (Fig. 8E). These results mirror gene expression (Fig. 6) and Gfap staining (Fig. 5), demonstrating more activated Müller cells in the Atg5-deficient Müller cells and heterogeneous Müller cell states in normal retina. Data in Figures 6 to 8 suggest that, under control conditions, only a portion of Müller glia respond to LPS; however, after ablation of autophagy nearly all cells become involved in the gliosis. This further emphasizes the importance of autophagy in determining Müller cell heterogeneity in the response to LPS.

Figure 7.

(A–D) Co-expression of Cralbp and Gfap in Müller cells. Atg5^f/f mice (A, C) and Atg5^i∆Müller mice (B, D) were injected in the vitreous cavity with PBS (A, B) or LPS (C, D). Retinae were harvested at 24 hours and co-stained for Gfap and Cralbp. Representative confocal images are shown. INL, inner nuclear layer; n = 5.

Figure 8.

(A–D) Expression of Lcn2 in Müller cells. Atg5^f/f mice (A, C) and Atg5^i∆Müller mice (B, D) were injected in the vitreous cavity with PBS (A, B) or LPS (C, D). Retinae were harvested at 24 hours and stained for Lcn2. Representative confocal images are shown. (E) Mean fluorescent intensity of Lcn2 expression in confocal images. Six ROIs per sample were quantified. *P < 0.05; n = 4.

In Vitro Analysis of Müller Glial Cells

We explored the role of autophagy in Müller cell cytokine production and gliosis using the rat Müller glia cell line rMC1. Cultured cells were transfected with control small interfering RNA (siRNA) or siRNA targeting Atg5. We then treated the cells with 0.5-µg/mL LPS, harvested supernatants 6 and 24 hours later, and measured cytokines by Luminex chemokine/cytokine arrays (R&D Systems, Minneapolis, MN, USA) (Fig. 9). At 6 hours we detected increased Cxcl1 and Cxcl2, both neutrophil chemoattractants. Ccl2 (which attracts macrophages) was also increased (Fig. 9A). Assays at the 24-hour time point revealed increased production of Ccl3 (a macrophage cytokine), Cxcl2 (neutrophil chemokine), and proinflammatory Il-6 and Tnf-α (Fig. 9B). Thus, blocking autophagy leads to increased production of cytokine/chemokines in Müller cells. (Insets in Figs. 9A and 9B are representative western blots showing Atg5 knockdown success.)

Figure 9.

In vitro cytokine levels in cultured rMC1 cell line. rMC1 cells were transfected with either control or Atg5 siRNA for 3 days and then treated with 5 µg/mL LPS. (A, B) Culture supernatants and cell lysates were collected at 6 hours (A) and 24 hours (B). Cytokines and chemokines in culture supernatants were quantified using a rat-specific Luminex array. Inserts are western blots of cell lysates showing deletion of Atg5 protein in Atg5 siRNA-transfected cells at both time points. *P < 0.05; n = 6.

Autophagy is regulated by the mechanistic target of rapamycin (mTOR) pathway, which also controls immune and inflammatory responses.²⁸ Therefore, we examined mTOR activation in response to LPS following knockdown of Atg5. Data in Figure 10 show that we were successful in significantly reducing Atg5 (Fig. 10A), and this led to an increase in phosphorylated mTOR at activating residue Ser2448 (Fig. 10B). We also detected phosphorylation of p70S6K, a downstream target of the mTOR complex (Fig. 10C). Knockdown of Atg5 also led to increased Gfap expression, confirming our in vivo observations that lost autophagy increases Gfap and gliosis (Fig. 10D).

Figure 10.

Activation of mTOR in Atg5-deleted rMC1 cells treated with LPS. rMC1 cells were transfected with either control or Atg5 siRNA for 3 days. Cell lysates were collected at 20, 40, 60, and 80 minutes after treatment with 5 µg/mL LPS, as well as media alone. (A–D) Lysates were run on SDS gels, transferred to polyvinylidene fluoride membranes, and probed for expression of Atg5 (A), phos-mTOR^Ser2448 and total mTOR (B), phospho-p70 S6 kinase (Thr389) and total p70S6K (C), and Gfap (D). Samples were run in either duplicate or triplicate, and the relative expression of each protein from two separate experiments is shown in the corresponding graph. *P < 0.05.

Discussion

The eye has a limited capacity for regeneration such that immune-mediated inflammation could have destructive consequences leading to blindness. Accordingly, the eye limits immune responses and inflammation by a process known as immune privilege. Our understanding of the mechanisms of immune privilege has evolved over the 70+ years since the concept was described in Medawar's investigations of skin grafts,²⁹ although similar observations are much older.³⁰ Medawar observed that tissues transplanted to the anterior chamber of the eye enjoyed prolonged survival compared to other areas, and he referred to the eye as being “immune privileged” due to its isolation from the immune system. Although physical barriers such as the BRB are significant, immune privilege is now recognized as multiple mechanisms collaborating to limit damaging inflammation that can threaten vision.^3,4

Although cell types such as the RPE and corneal endothelium are known participants in immune privilege, little is known about the role of the Müller glia. Because autophagy also has anti-inflammatory properties,^10,11 we tested the idea that autophagy in Müller cells might play a role. We did this by examining an intraocular inflammatory response model where uveitis was induced by LPS in the presence of Müller glial cells deficient in autophagy. When we compared these cells to control, autophagy-sufficient Müller cells, we observed that without autophagy inflammation was more severe and persisted for 5 days after LPS injection. In some cases, retinal damage was observed. Heightened proinflammatory cytokines and chemokines in the retinal tissue were detected, as was increased Müller cell gliosis. This suggested that autophagy in Müller cells plays a key role in immune privilege by controlling the severity of the inflammatory response in the eye.

The analysis of gliosis in Müller glia revealed a novel perspective on Müller cell heterogeneity. When autophagy was intact, only a portion of Müller cells were gliotic. However, with autophagy deficiency, nearly all cells attained a gliotic phenotype as determined by expression of the canonical gliosis marker Gfap. This was emphasized by our scRNA-seq analysis of Müller cells isolated from autophagy-sufficient and -deficient eyes. In control eyes, increased inflammatory gene expression was observed in a portion of Müller cells, whereas in the absence of autophagy nearly all cells expressed increased inflammatory gene expression. We can hypothesize that, in response to LPS, the normal retina activates a responsive subpopulation of Müller glia but maintains a significant basal population. The basal population, which retains gene expression of important glial homeostatic genes, such as Kir4.1, Cralbp1, and Aqp4, might sustain overall Müller cell/retinal function, helping return the retina to homeostasis. When autophagy is ablated, all Müller cells activate, leaving the retina without proper support and at the mercy of a destructive inflammatory response. Activated Müller cells made increased amounts of inflammation-promoting molecules with or without autophagy, but autophagy deficiency led to increased numbers of activated cells producing high levels of inflammatory molecules. We originally hypothesized that there might be gene expression differences between Müller glia with or without autophagy; however, the scRNA-seq analysis ruled out this idea.

It is known that Müller glia do not represent a homogeneous group of cells, and they do not respond to pathogenic insults in the same manner.^31,32 Heterogeneity between neighboring Müller cells in the same region with respect to gliosis has also been observed,^7,33 but this has never been sufficiently explained. Our description of two states of Müller glial cells (basal and activated) that are regulated by the autophagy pathway is a novel perspective on heterogeneity. There are several places where heterogeneity with respect to autophagy could play a critical role in the Müller cell response to inflammation (these are not mutually exclusive). First, autophagy may be preventing Müller cell activation only in some cells but promoting activation is others. This may be influenced by cell contacts or interactions with the blood vessels that influence Müller cell functions (explaining possible regional heterogeneity). Second, autophagy might mediate the return of the retina to homeostasis after inflammatory insult, and this function is lost in autophagy-deficient Müller cells. In the absence of autophagy, the cells continue to be activated and produce chemokines and neurotoxic molecules, leading to retinal damage. Third, autophagy could directly regulate molecules that participate in the BRB and osmotic balance. However, it is noteworthy that we did not detect extensive alterations in the BRB when Müller cells were autophagy deficient. Other than increased inflammation with Atg5 deficiency, no evidence for long-term BRB changes were observed (data not shown).

In cultured Müller glial cells, we observed increased proinflammatory cytokines with siRNA knockdown of Atg5. These findings mirrored our gene expression analysis confirming that cytokine/chemokines that attract inflammatory cells (Ccl2, Cxcl1, Cxcl2) and promote inflammation (Il-6, Tnf-α) are increased in the absence of autophagy. Further analysis showed that the loss of autophagy activated the mTOR pathway, leading to increased cytokine production and confirming the well-known balance between the mTOR and autophagy pathways.²⁸ Thus, the autophagy and mTOR pathways represent potential therapeutic targets for inflammatory eye diseases. It is notable that we found no evidence of the involvement of autophagy in the degradation of Gfap in cultured cells (data not shown).

We conclude that autophagy deficiency in Müller cells leads to enhanced intraocular inflammation through activation of the mTOR pathway. Our data also defined a novel perspective on Müller glial heterogeneity based on their activation state. Importantly, autophagy restrains cellular activation and inflammation, supporting immune privilege by preventing excessive and potentially destructive immune responses.