2.1 A single-cell atlas of the mouse retina in the normal state and two pathological states
2.1.1 Retinal microstructure: Tertiary neuronal structure and the neurovascular unit
The retina is a highly specialized neural tissue that encodes and transmits visual information to the visual center (10). Its structure includes the following. 1) Tertiary neuron structure: Photoreceptors (rods and cones) convert photons into electrochemical signals and transmit signals to secondary neurons through neurotransmitters; interneurons (bipolar cells (BCs), horizontal cells (HCs), and amacrine cells (ACs)) process and transmit visual signals; and output neurons (retinal ganglion cells (RGCs)) conduct signals through axons to the visual cortex (11, 12). 2) The microvascular system (endothelial cells, pericytes and the outer basement membrane) forms and maintains the blood‒retinal barrier (BRB) and transports nutrients and oxygen to meet the high metabolic demand of the retina (13). 3) Macroglia (astrocytes and Müller cells) encapsulate retinal neurons and blood vessels, participate in the formation of the BRB and maintain the homeostasis of the retinal environment (13). 4) Microglia and a small number of perivascular macrophages (PVMs) act as resident immune cells to monitor changes in the retinal microenvironment (14) (Figure 1A). The retinal neurovascular unit (NVU) is composed of glial cells, vasculature, and neurons (3, 5). Since astrocytes are mainly confined to the inner retina and wrap the blood vessels and neuronal axons in the ganglion cell layer (13), the NVU of the ganglion cell layer is mainly discussed below. The cellular components include astrocytes, Müller cells, endothelial cells, pericytes, and RGCs.
2.1.2 Source of data and cell regrouping
To explore the pathological changes in the NVU in different diseases at single-cell resolution, we collected retinal data from Aire knockout (AireKO) and Ndp knockout (NdpKO) mice. Data for the analysis were derived from publicly available data uploaded by Jeremy Nathans' team (8, 9). AIRE (AutoImmune Regulator) regulates the expression of tissue-specific antigens, including retinoid binding protein (IRBP) and retinal soluble antigen (S-Ag) (15). AireKO mice develop multiorgan autoimmune disease characterized by experimental autoimmune uveitis (EAU) in the eye. NdpKO mice exhibit abnormal retinal peripheral blood vessel formation due to disturbances in the Norrin/β-catenin signaling pathway, resulting in retinal hypoxia (16, 17). NdpKO mice serve as a good model for familial exudative vitreoretinopathy (FEVR). A total of 94,622 single-cell transcripts were obtained from a comprehensive analysis of transcriptomic data from these two studies, including 4 AireKO samples, 2 NdpKO samples, and 6 wild-type (WT) samples. Based on strict quality control, 61,953 cells ultimately passed the filtration steps, of which 21,469 were from AireKO mice, 10,218 were from NdpKO mice, and 30,266 were from WT mice (Figure S1A). After unbiased clustering of the data, 53 clusters were obtained (Figure S1B-D). Based on the original article and canonical cell markers(8, 9, 18, 19), we identified resident retinal cells and infiltrating immune cells in AireKO mice (including 1) T lymphocytes, B lymphocytes, and plasma cells; 2) the monocyte lineage; and 3) NK cells) (Figure 1B). Meanwhile, some subclusters containing rods, cones and Müller cells were only present in AireKO mice (Figure 1C). This is attributed to changes in the cellular phenotype caused by the inflammatory environment, in which neuronal degeneration in uveitis has also been demonstrated in other studies(20). We also identified some strongly and specifically expressed genes in intrinsic retinal cell types (Figure 1D), providing new insights for the identification and characterization of retinal cells.
2.1.3 Sample for verification: "P14"
The retinas of seven 14-day-old WT mice obtained by Aviv Regev and Steven A. McCarroll were used to verify our conclusion and were named sample "P14" (21). The screening conditions and processing workflow were completely consistent with those used for the data in the analysis. P14 was divided into 58 subclusters, and the corresponding cell types were identified according to the original article. All clusters in P14 were used to validate the newly obtained cell markers, and a perfect fit was achieved (Figure 1E and S2A-C).
2.1.4 Differentially expressed genes of AireKO and NdpKO mice compared with WT mice
We calculated the number of differentially expressed genes (DEGs) in retinal intrinsic cell types in AireKO and NdpKO mice compared with WT mice. Differences in Müller cells were most pronounced in AireKO mice, whereas differences in endothelial cells and Müller cells were most prominent in NdpKO mice (Figure 1F). Meanwhile, in both models, the NVU was more prone to changes than the neurons of the phototransduction system.
We also show the specific up- and downregulated genes for each cell type (Figure 1G). Consistent with the original article, almost all retinal cells in AireKO mice showed upregulation of MHC-II genes, which are involved in antigen presentation and promote immune cell infiltration, thereby aggravating the inflammatory response. Notably, astrocytes, Müller cells, and endothelial cells showed downregulation of genes that maintain their own function, suggesting that all three cell types were severely damaged in AireKO mice. In NdpKO mice, Müller cells and endothelial cells showed upregulation of Vegfa and Igfbp3, which are involved in promoting neovascularization in response to hypoxia. The majority of cells in the NdpKO mice showed downregulation of genes in the Xist family, which were confirmed to inhibit hypoxia-induced cell death in cardiomyocytes(22). We speculate that the downregulation of Xist family genes is a self-protection mechanism of retinal cells.
2.1.5 Alterations in cellular communication in AireKO and NdpKO cells
CellChat was used to calculate communication signals in normal retinas to analyze cellular crosstalk(23). The cellular communication among Müller cells, astrocytes and RGCs is particularly active, and the complex signal exchange among the three cell types is conducive to maintaining the survival of RGCs (Figure 1H). We also calculated changes in the number of signals released and received by retinal cells in both models (Figure 1I). In AireKO mice, macrophages and macroglia significantly upregulated communication signals (predominantly inflammatory signals), while tertiary neurons downregulated communication signals, suggesting impairment of phototransduction. Communication signaling in Müller cells, astrocytes, and endothelial cells was upregulated in NdpKO mice, while that in tertiary neurons was almost unchanged, suggesting that the BRB was initially affected.
2.2 Exploration and validation of two subpopulations of Müller cells
Müller cells not only scavenges synaptic glutamate but also expel excess water through aquaporin 4 (Aqp4) on terminal foot processes facing the vitreous or capillary endothelium, preventing intraretinal fluid accumulation(24). Impaired Aqp4 results in decreased permeability of the plasma membrane and edema of the cell body in Müller cells; these changes are considered early pathophysiological events in diseases such as neuromyelitis optica spectrum disorders (NMOSDs)(25) and diabetic macular edema (DME)(26).
2.2.1 Two Müller cell subpopulations
Due to the integration of multiple samples, a large number of Müller cell transcripts were obtained and separated into two subpopulations with significant differences (Figure 2A, S2D and S3A). Both subpopulations expressed typical markers of Müller cells. Aqp4 was highly expressed in one subgroup and expressed at low levels in the other subgroup (Figure 2B). Therefore, these subpopulations were named Aqp4hi and Aqp4lo cells, respectively. Aqp4hi cells excrete water from the retina through high expression of Aqp4. Although both subgroups expressed transglutaminase (Glul), Aqp4hi cells had a higher expression of the glutamate-aspartate transporter (Slc1a3), suggesting a stronger ability to transport glutamate (Figure 2C). Aqp4hi cells rapidly hydrate large amounts of carbon dioxide produced by photoreceptor metabolism into bicarbonate and protons outside of the retina via carbonic anhydrase XIV (Car14) on the cell membrane(27). In addition, the adhesion molecule Ncam1 that contacts nerve cells and the connexin Cdh5 that contacts endothelial cells are also highly expressed on Aqp4hi cells. In addition, Aqp4hi cells highly expresses transcription factors of the Ap-1 family (Jun, Junb, Fos). The AP-1 family has been proven to affect a variety of cellular processes (such as differentiation, proliferation, and apoptosis), depending on the cellular environment and stimuli(28). This means that Aqp4hi cells are prone to phenotypic changes caused by environmental changes. The above enzymes and proteins are rarely expressed in Aqp4lo cells. The above evidence indicates that Aqp4hi cells are closely related to blood vessels and nerves and are the classic Müller cell type reported in previous studies; these cells are of great significance for maintaining homeostasis of the retinal environment.
Interestingly, a phototransduction process similar to that of photoreceptors was observed in Aqp4lo cells: retinal and opsin combine to form rhodopsin, which activates the transduction protein when rhodopsin absorbs light(29). Transducins in turn activate phosphodiesterase, reduce cGMP levels and regulate downstream ion channels. Recent research showed that in addition to rods and cones, intrinsically photosensitive retinal ganglion cells (ipRGCs) can directly transduce light signals(30). Although Aqp4lo cells express components of the phototransduction pathway, because the downstream neurons are not yet clear, we still cannot prove that Müller cells becomes a new type of photoreceptor cell. Notably, multiple studies have shown that in the absence of photoreceptors, Müller cells express components of the photoreceptor pathway to replenish and replace missing photoreceptors. For example, in a mouse model of retinitis pigmentosa (RP), the expression of genes in pathways involved in the maintenance of photoreceptors in Müller cells was significantly increased with the loss of rods and cones(31). Similarly, M. Goel et al. observed the expression of rhodopsin in Müller cells in mice with inherited retinal degeneration (RD)(32). Although it cannot be demonstrated that Aqp4lo cells are a class of photoreceptors other than rods and cones, our study confirms the existence of a subclass of Müller cells expressing photoreceptors in the normal retina. Furthermore, we believe that Aqp4lo cells are the Müller cells that compensate for photoreceptors in diseases involving photoreceptor degeneration. In conclusion, the role of Aqp4lo cells in normal and diseased retinas deserves further exploration.
Genomic variation analysis (GSVA) was performed on these two groups of cells, and the limma package was used to calculate the differential pathways of the two groups of cells (Figure 2D)(33). The Notch and Wnt signaling pathways were significantly enriched in Aqp4hi cells. A previous study confirmed that these studies can be activated by retinal damage to cause Müller cell proliferation(34, 35).
2.2.2 Cellular communication of Aqp4hi and Aqp4lo cells in the NVU
We describe the communication signaling pathway of the NVU structure in terms of "cell‒cell contact" and "secretory signaling". Aqp4hi cells, RGCs, and astrocytes mediate cell-to-cell contact via cell adhesion molecule (CADM), cadherin (CDH), and neural cell adhesion molecule (NCAM) (Figure 2E, S2E and S3B). Astrocytes secrete a variety of signaling factors (vascular endothelial growth factor (VEGF), Midkine, and Pleiotrophin) that act on other cells in the NVU. Both astrocytes and RGCs secrete VEGF, which acts on endothelial cells and Aqp4hi cells to regulate their function. However, astrocytes mainly secrete vegfa, while RGCs mainly secrete vegfb. Analysis of intracellular communication within the NVU suggests the following: 1) Aqp4hi cells and astrocytes encapsulate RGCs to form a glial barrier, and this close contact explains the rapid onset of reactive gliosis following RGC axonal injury(36). 2) Astrocytes are the core of signal secretion in the NVU, which is beneficial for maintaining the normal function of blood vessels and neurons. 3) Aqp4lo cells are rarely involved in cellular communication with other cells in the NVU, possibly to maintain their photoconductive properties.
2.2.3 Transcriptional regulatory mechanisms underlying the distinct phenotypes of the two Müller cell subpopulations
We then used SCENIC to infer the transcription factor (TF) regulatory information underlying the different phenotypes of the two Müller cell subclusters (Figure 2F and 2G)(37). Aqp4lo cells highly express Neurod1 to regulate photoreceptor genes (Gnat, Nrl). Aqp4hi cells express Pax6, Sox9, and Dbp to regulate adhesion factors, glutamate transport, and carbon dioxide uptake. Id1 regulates the metabolic function of Aqp4hi cells. Jund regulates other AP-1 family members.
2.2.4 Verification of the existence of two Müller cell subgroups
To verify the accuracy of the Müller cell subcluster clustering, we integrated Müller cells into the P14 samples and our analysis samples using canonical correlation analysis (CCA)(38). After regrouping the integrated data, five cell subclusters were obtained, of which cluster_0 and cluster_1 accounted for the main portion of the cells. Cluster_0 and cluster_1 correspond to Aqp4hi and Aqp4lo cells, respectively (Figure 2H and 2I). Based on the integrated data, we identified markers for Müller cell subclusters (Figure 2J). In conclusion, we identified and validated two specific Müller cell subclusters in the retina and further explored their phenotypic differences and underlying transcriptional mechanisms.
2.3 Phenotypic changes and underlying mechanisms of the NVU in two models
2.3.1 Müller_EAU cells are differentiated from Aqp4hi cells
Next, we investigated the phenotypic changes and underlying mechanisms of the NVU in AireKO and NdpKO mice. In addition to Aqp4hi and Aqp4lo cells, there is a specific group of Müller cells that exist only in AireKO and are named Müller_EAU. Correlations among the three Müller cell subgroups calculated using conserved genes of Müller cells indicate that the gene expression patterns of Aqp4hi and Müller_EAU cells are highly similar (Figure 3A). Meanwhile, Müller_EAU cells do not express photoreceptor genes. Furthermore, pseudotime analysis of the three subgroups confirmed that Müller_EAU cells are derived from the differentiation of Aqp4hi cells (Figure 3B). Therefore, we believe that intense inflammation of the retina in AireKO mice drives the phenotypic changes of Aqp4hi cells and converts them to Müller_EAU cells.
2.3.2 Phenotypic transition of Müller cells and astrocytes in AireKO mice
We next investigated the cells with the most significant DEGs in AireKO Müller cells and astrocytes (Figure 3C). We noticed that the guanylate-binding protein family (Gbp), proteasome B-type family (Psmb), and interferon-inducible protein (Ifi) were all upregulated, which was confirmed to be induced by interferon(39, 40). In addition, astrocytes specifically upregulated complement genes. GSVA showed that astrocytes and Müller_EAU cells significantly upregulated inflammatory pathways(33), of which the JAK/STAT pathway may be an important mechanism for regulating inflammation in both cell types (Figure 3D). Notably, both Müller cells and astrocytes downregulated the Notch pathway. Studies have shown that the downregulation of the Notch pathway is related to Müller cell dedifferentiation(41); however, the specific mechanism remains to be further studied. Biosynthesis of pantothenic acid and coenzyme A, which are important intermediates in cellular energy metabolism, is reduced in astrocytes and Müller cells. The former can also increase glutathione biosynthesis to slow apoptosis and cell damage(42), suggesting that the antioxidative stress capacity of glial cells is attenuated in AireKO mice.
2.3.3 Transcriptional mechanisms underlying the phenotypic transition of Müller cells and astrocytes
SCENIC analysis showed that Müller cells mainly upregulated three types of transcription factors: 1) the AP-1 family (Fos, Jun, Junb, Fosb); 2) Stats (Stat1, Stat2, Stat3); and 3) the interferon regulator family (Irf1, Irf2) (Figure 3E). The expression of Irf family members in Müller cells confirmed that the degeneration of Müller cells is directly affected by IFN-γ secreted by Th1 cells. Although astrocytes also upregulate members of the Stat and Irf families of transcription factors, they rarely express members of the AP-1 family, suggesting that AP-1 may play an important role in the phenotypic changes of only Müller cells. Astrocytes also upregulate Cebpb and the cAMP response element modulator Crem in AireKO mice to regulate immune inflammatory responses(43, 44). Downregulation of Id1 in astrocytes is associated with attenuation of self-renewal and differentiation processes(45).
SCENIC provided regulatory predictions for three classes of transcription factors (Figure 3F and S2F). The Stat family and Irf family regulate MHC-II genes and proteasomes (Psmb8 and Psmb9), among which Stat1 and Irf8 also promote the expression of interferon-inducible proteins. The network diagram indicates that Stat1 and Irf8 are the main regulators in the Stat and Irf family. Transcription factors such as the AP-1 family, Atf3, and Egr1 can also promote the release of some chemokines; however, the inflammation of Müller_EAU cells may be mainly affected by the Stat and Irf family. Notably, some genes regulated by Fos and Junb were highly expressed in normal Aqp4hi cells but significantly downregulated in cells from AireKO mice. We believe that the loss-of-function of Müller_EAU cells may be closely related to the Ap-1 family; however, the specific mechanism deserves further exploration.
2.3.4 Phenotypic transition of Müller cells in NdpKO mice
We investigated DEGs in NdpKO mice for all subclusters within the NVU (Figure 3G). Both Müller cell subgroups upregulated NADH dehydrogenase-related genes. Aqp4hi cells also upregulated Vegfa to induce neovascularization, which did not occur in Aqp4lo cells. Gene Set Enrichment Analysis (GSEA) results also confirmed the metabolic shift of Aqp4hi cells: glycolysis and oxidative phosphorylation pathways were upregulated (Figure 3H).
Endothelial cells showed impaired glucose uptake in NdpKO mice, as indicated by downregulation of multiple glucose transporters and upregulation of the lipoprotein lipase (LPL) transporter. Notably, endothelial cell-secreted basement membrane (BM) components (collagen IV and Lamin proteins) were upregulated (Figure 3G and S3C), confirming thickening of the basement membrane in the microvasculature in NdpKO mice, which impairs intercellular communication, reducing barrier function(46). Meanwhile, downregulation of claudin proteins (Ocln, Lsr, Pltp) revealed impaired tight junctions between endothelial cells (Figure 3G). Thickening of the basement membrane, together with disruption of tight junctions, suggests damage to the endothelial barrier, while other cells show only altered metabolism (Figure S3D). Therefore, we speculate that endothelial barrier disruption is the first pathological change in NdpKO mice, and Aqp4hi cells respond to this change by secreting proangiogenic factors to alleviate the hypoxic environment.
2.3.5 Transcriptional mechanisms of endothelial barrier impairment
To investigate the mechanism of endothelial barrier disruption, we further investigated the transcriptional mechanism and gene regulatory network of endothelial cells in NdpKO mice (Figure 3I and S3E). SCENIC analysis revealed activation of the transcription factors Ets1 and Mef2c, whose target genes (Dlc1, Nrp, Plvap) were upregulated in NdpKO cells and were verified to play an important role in the regulation of angiogenesis and the formation of polarized structures(47-49). On the other hand, the forkhead box (Fox) family (Foxf2, Foxp1, Foxq1) and Lef1 and its downstream target genes were all downregulated in NdpKO mice. Gene regulatory networks revealed that FOX genes and Lef1 regulate adhesion proteins (Ocln, Lsr, Pltp) and downregulate glucose transporters and extracellular matrix (ECM) components, which reveals the mechanism of endothelial barrier impairment(50, 51).
2.4 Infiltration of T cells in AireKO mice
2.4.1 Subpopulations of T cells in AireKO mice
Abundant lymphocytes infiltrate the retina of mice with EAU(52), among which Th1 cells are the main cells driving inflammation, mobilizing other immune cells by secreting IFN-γ(53, 54). NK and CD8+ T cells are among the most important pathogenic factors of Bechtel uveitis (BU): the active phase of BU is dominated by CD8+ T- and NK-cell infiltration, while other types of uveitis are dominated by CD4+ T cells(55). Since T cells and NK cells were identified as the same cluster, we regrouped this cluster to obtain NK cells, Treg cells, CD8+ T cells, and helper T cells (Th cells) (Figure 4A). Th cells included one Th0 subpopulation and three Th1 subpopulations. We identified markers for three subgroups and renamed them accordingly: Ifng+ T, Sostdc1+ T and Il10+ T cells (Figure S4A and S4B).
The pie chart shows that the severity of EAU is closely related to the degree of immune cell infiltration (Figure 4B). According to the dendrogram of the genes of each subgroup, Il10+ T cells were more similar to Ifng+ T cells, while Sostdc1+ T cells were more similar to Th0 cells (Figure 4B).
2.4.2 Inflammatory factors secreted by T and NK cells
We investigated the expression of different classes of inflammatory mediators in these immune cells. As the most important inflammatory mediator mediating retinal inflammation in uveitis, IFN-γ is mainly secreted by Ifng+ T cells (Figure 4C). In addition, Ifng+ T cells secrete a variety of chemokines that drive inflammation and tissue damage. Therefore, we consider Ifng+ T cells to be the main proinflammatory cells. Il10+ T cells, on the one hand, can secrete some inflammatory factors, like Th1 cells, to exert proinflammatory effects, and on the other hand, they can also secrete Il10 and TGFβ1 to exert anti-inflammatory effects. The release of Il10 has been confirmed to have a clear inhibitory effect on uveitis(56, 57). In contrast to Ifng+ T cells, Sostdc1+ T cells secreted fewer inflammatory factors and were phenotypically similar to Th0 cells, which may be undifferentiated Th1 cells. Compared with Ifng+ T and Il10+ T cells, Sostdc1+ T cells secreted the fewest inflammatory factors and had phenotypes similar to those of Th0 cells, which may be undifferentiated Th1 cells. NK and CD8+ T cells not only secreted chemokines (Xcl1 and CCL3) but also secreted granzyme, which induces programmed death in target cells. The cytotoxic effects of NK cells and CD8+ T cells may be an important reason why BU is more recurrent and destructive than other forms of uveitis. In addition, we observed that Ifng+ T cells also secrete proteins that regulate cytotoxic granules (Nkg7, Lgals1)(58, 59), indicating a cytotoxic effect. In uveitis, Treg cells are the main anti-inflammatory cells in the retina(60). Here, we observed that Treg cells significantly upregulated a variety of cellular receptors regulated by a variety of inflammatory factors secreted by humoral and intraretinal immune cells.
Kyoto Encyclopedia of Genes and Genomes (KEGG)(61) enrichment analysis was performed on NK- and T-cell subsets, and the enriched pathways were divided into four categories (Figure 4D). Ifng+ T and NK cells were significantly enriched in adhesion- and migration-related pathways. The Rap1, NF-κB and JAK-STAT signaling pathways were enriched in all five types of cells and are closely related to the regulation of inflammation in these cells.
2.4.3 Transcriptional mechanisms regulating T- and NK-cell inflammation
SCENIC analysis revealed that the transcription factors Runx2, Stat3 and Ets1 were significantly activated in Ifng+ T cells (Figure 4E). Gene regulatory networks (GRNs) revealed their downstream inflammatory mediators (Figure 4F and S4C). NK cells upregulate Eomes to regulate the killer cell lectin-like receptor, which is the basis for the cytotoxic effect of NK cells(59). They also upregulate Irf8, which regulates the expression of chemokines and receptors. Simultaneous activation of Runx3 in NK cells and Ifng+ T cells can promote the encoding and release of granzymes, which may be a common mechanism by which the two cell types exert cytotoxicity (Figure 4G). As one of the target genes of Maf, Il10 is specifically expressed in Il10+ T cells (Figure 4G). Therefore, Maf may become an important target for inducing Th1 cells to secrete Il10.
2.5 Monocytes and their derived cells in AireKO mice
The infiltration of CD4+ T cells in autoimmune uveitis is dependent on the activation of antigen-presenting cells (APCs). Studies have demonstrated that preventing the recruitment and/or maturation of APCs largely inhibits the generation of antigen-specific T cells(62-64).
2.5.1 MO-MA lineages, MO-DC lineages and microglia
In uveitis, monocytes (MO) in the peripheral blood are recruited to cross the BRB and differentiate into macrophages (MA) and dendritic cells (DC). In our samples, monocytes and their derived cells and microglia were mixed and aggregated into one cluster. We regrouped the cluster and obtained MO-MA lineages, MO-DC lineages and microglia according to canonical markers(65, 66) (Figure 5A).
Since macrophages and microglia exhibit similar gene expression patterns under the pathological condition of uveitis, it is difficult to distinguish them. Here, we can better differentiate mature APCs in the retina by multiple gene markers (Figure 5B).
2.5.2 Pseudotime trajectories of MO-DC lineags
We constructed a pseudotime trajectory of MO-DC lineages and divided them into monocytes with DC differentiation potential (Monocyte_DC), dendritic cell-committed monocytes (DC-committed Mo) and two mature dendritic cells: conventional DCs (cDCs) and plasmacytoid DCs (pDCs) (Figure S5A and S5B).
GeneSwitches were utilized for the MO-pDC differentiation trajectories(67), and we defined the transitions of key transcription factors along the trajectory and the order in which these transitions occurred (Figure 5C and S5C). Activation of the transcription factor Bcl11a, an early switching event on the pDC differentiation trajectory, has been shown to be an essential lineage change regulating pDC development(68). The transcription factors Runx2, Hoxa7, Spib, and Cxxc5 are subsequently activated and maintain differentiation into pDCs.
Next, we compared the phenotypic differences between cDCs and pDCs (Figure 5D). We found that many MHC-II genes and costimulatory molecules (CD80, CD86, CD40) were upregulated along the trajectory from Monocyte_DC to cDCs, revealing the gradually enhanced antigen processing and presentation ability of these cells. The high expression of actin, myosin and integrin family genes in cDCs is related to chemotaxis. We observed that the expression levels of Nfkbia and the P100/RelB dimer (Nfkb2/Relb) were all significantly elevated in cDCs. The former is a key complex in the classical NF-kB signaling pathway, and the latter is a key nuclear transcription factor in the noncanonical pathway(69). This evidence suggests that the simultaneous activation of the canonical and noncanonical NF-kB signaling pathways is an important mechanism for the differentiation and maturation of cDCs in EAU. In contrast to cDCs, pDCs downregulate inflammatory factors and adhesion proteins along the differentiation trajectory from Monocyte_DC to pDCs. Interestingly, the IFN-α receptor is highly expressed in pDCs, suggesting that their differentiation may be induced by IFN-α.
2.5.3 Pseudotime trajectories of MO-MA lineags
Likewise, we divided MO-MA lineages into monocytes with macrophage differentiation potential (Monocyte_ma), macrophage-committed monocytes (ma_committed mo), and mature macrophages (macrophages) (Figure 5E, 5F and S5D). From Monocyte_ma to macrophages, secreted inflammatory factors changed: Ccl6, Ccl9, and Il1β were downregulated, while Ccl3, Ccl4, and cathepsin D (Ctsd) were upregulated (Figure 5G and S5E).
2.5.4 Microglia in AireKO mice
Microglia are activated to exert inflammatory effects under pathological conditions(70). Here, by comparing microglia and monocyte-derived macrophages in EAU, we found that macrophages highly expressed a variety of chemokines, and their inflammatory effects were strong. This finding suggests that peripheral blood-derived macrophages, rather than microglia, are the main pathogenic macrophages in EAU (Figure 5H and S5F).
2.6 Immune regulatory network in AireKO mice
We paid particular attention to the inflammatory regulatory network mediated by immune infiltrating cells in AireKO mice. Therefore, we performed an in-depth dissection of the regulatory networks among inflammatory cells and between inflammatory cells and cells in the NVU.
2.6.1 Regulatory network among infiltrating immune cells
Interestingly, we noticed that monocytes (including Monocyte_DC and Monocyte_MA) are stimulated by a variety of inflammatory factors: 1) almost all immune cells secrete Ccl5 to act on monocytes; 2) Ifng+ T and macrophages secrete Ccl3 and Ccl4 to act on monocytes; 3) Ifng+ T secrete IFN-γ to act on monocytes; 4) MO-MA lineages act on monocytes, with Monocyte_MA secreting Ccl8 to act on two monocyte subclasses and macrophages secreting Ccl2 and Ccl12 to act on Monocyte_DC; and 5) Monocyte_DC secretes Ccl6 and Ccl9 to act on two monocyte subclasses (Figure 5I and 5J).
These results suggest that the inflammatory environment in AireKO mice induces monocytes to convert to DCs or macrophages and activate Ifng+ T cells through antigen presentation. Therefore, a closed loop of "inflammatory cell - monocyte - APC - Ifng+ T cells" is formed to produce an inflammatory cascade.
Monocytes and their derived cells act directly on the T-cell population by secreting Cxcl16, which can enhance cytotoxicity and induce higher levels of IFN-γ secretion(71). Macrophages are regulated by other immune cells due to the expression of Ccr5. NK cells are regulated by two subclasses of monocytes and macrophages via expression of Ccr2.
2.6.2 Regulatory network between inflammatory cells and cells in the NVU
It is worth noting that almost all cell subsets in the NVU upregulate MHC-II genes and secrete inflammatory factors, thus becoming key links in the inflammatory network. Similar to the previous results, all cells in the NVU act on monocytes (Figure 5K): 1) Müller_EAU and astrocytes secrete Ccl5 and Ccl8, and astrocytes also specifically release complement; 2) endothelial cells secrete Galectin-9 (lgals9); 3) RGCs release Ccl27a; and 4) pericytes release Il34. Inflammatory mediators secreted by cells in the NVU differ from infiltrating immune cells, but they both promote monocyte maturation and induce chemotaxis, adhesion, and migration.
2.6.3 Impairment of cellular normal regulatory networks in the NVU
In summary, immune cells and cells in the NVU together constitute an closed inflammatory loop of "inflammatory cells/NVU - monocytes - APC - Ifng+ T cells". Therefore, blocking the induction of monocytes by inflammatory cells and/or cells in the NVU can block the progression of inflammation and may become a new direction for the treatment of autoimmune uveitis in the future.
Furthermore, we found that cellular self-protection signaling is impaired in the NVU. The expression levels of brain-derived neurotrophic factor (BDNF) and VEGF secreted by astrocytes were decreased (Figure 5K), suggesting that the neuroprotective effect of astrocytes on RGCs and the ability to maintain vascular integrity were weakened(72).