Work

Deciphering primate retinal aging at single-cell resolution

Year: 2020

Type: article

Abstract: Dear Editor, The retina is a light-sensitive highly-organized tissue, which is vulnerable to aging and age-related retinal diseases. Specifically, progressive retinal degeneration leads to visual function deterioration and vision impairment in the elderly (Lin et al., 2016). In diseases such as age-related macular degeneration (AMD), retinitis pigmentosa (RP) and diabetic retinopathy (DR), pathological process lacking effective treatments profoundly and negatively impact on the quality of life in the elderly (Lin et al., 2016; Chen et al., 2019). Thus, an in-depth molecular assessment of the mechanisms driving retinal aging is of urgent scientific and medical importance. The retina comprises a layered structure with neural retina and a layer of retinal pigment epithelium (RPE). The neural retina consists of elaborate neuronal circuitry for the detection of visual information whereas the RPE and adjacent choroid layer (a vascularized and pigmented connective tissue) provide nutrition and physical and metabolic support. The neural retina layer mainly contains three demarcated cellular populations, photoreceptors (PRs), interneurons and retinal ganglion cells (RGCs). PRs, including rod and cone photoreceptors (Rod and Cone for abbreviation), convert light into an electrical signal and then transfer it to interneurons (bipolar cells, for instance). Interneurons then transmit the signal to RGCs, whose axons coalesce into the optic nerve projecting to the visual cortex. In addition, the neural retina layer contains many other cells, for example, Müller glial cells (Müller), which serve as supporting cells for the neurons. With aging, the retina undergoes a series of alterations, including physiological, structural and functional alterations, reflected as morphologic changes and/or loss of PRs and RGCs, decreased Rod- and Cone-mediated responses to light, impaired RPE function, and immune dysregulation (Lin et al., 2016; Chen et al., 2019). Due to ethical constraints and limited access to disease-free human retina tissues, in-depth studies of human retinal degeneration are not practically feasible. As a result, non-human primates (NHPs), including crab-eating macaques (Macaca fascicularis), have become a widely used primate model in retinal studies, as they share very similar genetic and physiological characteristics with humans (Li et al., 2020; Wang et al., 2020; Zhang et al., 2020). Most of the retinal cell types have been identified primarily by morphological criteria, which lacks the molecular resolution necessary for dissecting aging-related changes in the primate. With the advent of single-cell RNA sequencing (scRNA-seq) technologies, in-depth molecular surveys of cellular taxonomy and aging mechanism of primate ovary, aorta and pancreatic islet have been achieved (Li et al., 2020; Wang et al., 2020; Zhang et al., 2020). However, the effects of aging on retinal cell types including those present in neural retina and RPE, as well as cell types in choroid layer remain largely unknown. Here, we established the single-cell transcriptomic atlas of the retina and adjacent choroid in young and aged NHPs, identifying 15 cell types with distinct gene expression signatures and finding several novel markers. Our analysis reveals that oxidative stress is a major aging feature of the cells in the neural retinal layer, whereas an enhanced inflammatory response is that of RPE and choroidal cells. We also found that the RPE cell is the cell type most susceptible to aging in retina, as evidenced by the decreased cell density as well as the highest numbers of differentially expressed genes (DEGs) overlapping with genes underlying aging and aging-related retinal diseases, along with aberrant cell-cell interactions with its two adjacent layers. Altogether, our study provides the roadmap for understanding retinal aging in a NHP model at single-cell resolution, enabling the identification of new diagnostic biomarkers and potential therapeutic targets for age-related human retinal disorders. To investigate structural and molecular alterations during physiological aging, we obtained retinal and choroidal tissues from eight young and eight aged cynomolgus monkeys (Fig. S1A). A histological comparison revealed the width of the RPE-choroid layer, neural retinal layer and its internal retinal nerve fiber layer (RNFL), ganglion cell layer (GCL), inner plexiform layer (IPL), inner nuclear layer (INL), outer plexiform layer (OPL), outer nuclear layer (ONL), photoreceptor layer (PRL) remained largely unchanged in the aged retina, suggesting an overall structural stability of aged retina (Fig. 1A). However, the density of Cone and RPE cells was reduced in the aged retina, implying age-related impairment on retinal cells (Figs. 1B, S1B and S1C). Scanning electron microscopy (SEM)-based 3D reconstruction revealed a less compact structure between PRs and RPE in aged retina compared with that from young animals (Fig. 1C), indicating disconnected interactions between these two layers. In the aged RPE cells, we also observed aging-associated elevated lipofuscin level (Fig. 1D), an indicator of compromised retinal function due to disruption of digestion of photoreceptor outer segments (Kennedy et al., 1995). Together, these results demonstrate that the PR and RPE undergo compromised connections and disrupted function during physiological aging. Establishment of cynomolgus monkey single-cell transcriptome landscape of retinal and choroidal cells. (A) Left, diagram showing the histological structure of monkey retina and choroid. Middle, representative images of H&E-stained sections of retinal and choroidal tissues from young and old monkeys. Right, the quantitative data for the relative width of each layer are shown as the mean ± SEM. Scale bar, 50 μm. Young, n = 8 monkeys; old, n = 8 monkeys. One-tailed student's t-test P values are indicated. (B) The relative density of Cone and RPE cells are shown as the mean ± SEM. Young, n = 8 monkeys; old, n = 8 monkeys. One-tailed student's t-test P values are indicated. (C) Large-scale three-dimensional reconstruction of a rectangular piece of retina using automatic collector of ultrathin sections scanning electron microscopy (AutoCUTS-SEM). Left, volume electron microscopy; right, the zoomed sections selected from the position of coordinate axes labeled with blue rectangles; the enlarged areas are indicated by black dash lines and shown without coloring with Imaris9.2.1 (color-key). Scale bar, 50 μm. (D) Representative images of lipofuscin accumulation in RPE are shown on the left, and quantitative data are shown as the mean ± SEM on the right. Scale bar, 50 μm. Young, n = 8 monkeys; old, n = 8 monkeys. Two-tailed Student's t-test P value is indicated. (E) Study flowchart of single-cell RNA-seq in this study. (F) Top, t-SNE plot showing different cell types. Rod, rod photoreceptors; Cone, cone photoreceptors; Bipolar, bipolar cells; Müller, Müller glial cells; RGC, retinal ganglion cells; MG, microglia; RPE, retinal pigment epithelium; Fib, fibroblasts; EC, endothelial cells; Gra, granulocyte; DC, dendritic cells; Mac, macrophages; T, T cells; B, B cells; Mast, mast cells. Bottom, stacked bar plot showing the cell type distribution in the neural retina layer, RPE-choroid layer. (G) t-SNE plots showing gene expression signatures of representative marker genes for different cell types. The color key from grey to red indicates low to high gene expression levels. (H) Bar plots showing the expression level of representative marker genes for various cell types. All expression levels are measured using the same scale. Data are shown as mean ± SEM. (I) Immunofluorescence analysis showing specific expression of RBP4 and ARR3 in monkey Cone, THY1 in RGC, GLUL in Müller, VSX2 and PKCα in Bipolar, as well as RPE65 in RPE cells. Scale bar, 50 μm. Asterisk represents the newly identified marker. (J) Heatmap showing the top-ranked 100 (ranked by average difference among different cell types) gene expression signatures of each cell type. Each column indicates the mean expression level of the corresponding cell type, and values in each row are z-score scaled. Low to high gene expression levels are indicated by color key from blue to red. Representative GO terms are shown to the right. To dissect cell-type-specific aging-related gene expression changes, we utilized a modified single-cell tagged reverse transcription sequencing (STRT-seq) technique (Li et al., 2020; Wang et al., 2020; Zhang et al., 2020). First, we dissected the neural retina and RPE-choroid layers from eight young and eight aged animals, and subjected tissue samples to digestion and subsequent scRNA-seq (Fig. 1E). After stringent cell filtration, we retained 6,410 qualified single cells from 16 individuals for the downstream analyses (Fig. S1A and S1D; Table S1). Global gene expression profiling displayed no apparent differences in cell type distribution among monkeys of different ages and sexes, alleviating concerns about technical variations between samples (Fig. S1E). Unsupervised clustering analysis and t-distributed stochastic neighbor embedding (t-SNE) dimensionality analysis identified 15 distinct cell types characterized by their unique gene-expression signatures (Table S2). These included Rod and Cone, bipolar cells (Bipolar), RGC, Müller, and microglia (MG) in the neural retinal layer, RPE cells, as well as fibroblasts (Fib), endothelial cells (EC), granulocytes (Gra), dendritic cells (DC), macrophages (Mac), T cells (T), B cells (B) and mast cells (Mast) in the choroid layer (Fig. 1F). We mapped gene expression profiles of well-defined markers as shown in t-SNE plots and bar plots, including PDE6A for Rod, GNAT2 and ARR3 for Cone, VSX2 and PRKCΑ for Bipolar, GLUL for Müller, THY1 and NEFM for RGC, CX3CR1 for MG, TYR and RPE65 for RPE cells, as well as a panel of known markers for immune cells and other supporting cells (Figs. 1G, 1H and S1F). Of particular interest, we identified a set of novel markers of retinal cells, such as AGRP and TFPI2 for Rod, RBP4 and NPTX1 for Cone, TMEM37 and SRGAP1 for Müller, ATP1B1 and RTN1 for RGC, as well as SCGB3A2 and TPM2 for RPE cells (Figs. 1H and S1F; Table S2). RBP4 was further validated by immunofluorescence staining specific to Cone (Fig. 1I). Gene Ontology (GO) analysis of cell-type-specific markers supported the identity of diverse cell types, revealing functional characteristics of NHP retinal and choroidal cells (Fig. 1J). For example, “sensory perception of light stimulus” for Rod and Cone, “gliogenesis” for Müller, “synaptic signaling” for RGC, “pigmentation” for RPE. In addition, our detailed analysis of Müller revealed a novel Müller glial cell subtype I (Fig. S2A and S2B; Table S2), with neuronal differentiation potential as indicated by subtype-specific DEG identification and GO enrichment analysis. Trajectory analysis ordered Müller glial cell subtype I at the beginning of the pseudotime trajectory (Fig. S2C), suggesting that Müller glial cell subtype I may have developmental potential to differentiate into other neural retinal cells. In addition, higher expression levels of several retinal progenitor markers, such as ID3, SOX9 and ZFP36L2 were also observed in subtype I (Lu et al., 2020), indicative of progenitor characteristics (Fig. S2D). Next, we performed a cross-species analysis among human, monkey and mouse to investigate cell identity similarities across species. As expected, the result indicates that the monkey retina recapitulates the human single-cell retinal transcriptional profile than mouse, as indicated by a higher correlation coefficient (Fig. S2E). We next sought to explore cell-type-specific gene expression alterations associated with aging. We found similar transcriptional signatures of marker genes for each cell type between young and old monkey (Fig. S3A), demonstrating that aging had minimal effect on cell identity. Calculation of age-associated coefficient of variation (CV) demonstrated that Rod in the neural layer and Mast in the choroid layer exhibited higher transcriptional noise reflected as higher expression variations among individual aged cells than among non-aged cells (Fig. S3B). We noticed that the hotspot genes annotated in the GenAge gene set and gene sets implicated in retinal diseases including AMD and RP (Table S3) were highly expressed in Rod, Cone, RGC, RPE cells and Bipolar (Fig. S4), suggesting that these cell types may be implicated in retinal homeostasis and aging. Based on analysis of aging-associated DEGs across all retinal and choroidal cell types, we found the largest numbers of upregulated DEGs in the RPE cells, EC and RGC (1,970, 1,561 and 1,560, respectively) in aged tissues, while the most downregulated DEGs were found in Gra and B cell types (939 and 840) relative to other cell types in the aged tissues (Fig. 2A; Table S4), indicating that these cell populations are most strongly affected by aging. Through GO analysis, we noticed that increased “response to oxidative stress” and “apoptotic signaling pathway”, as well as decreased “sensory organ development”, “transcription-coupled nucleotide-excision repair” and “visual perception” were enriched for DEGs of cells in the neural retinal layer (Fig. 2B; Table S4), indicative of elevated responses to external stresses and impaired functions in the aged neural layer. In parallel, GO analysis of DEGs in cells from the RPE-choroid layer revealed increased “cell activation involved in immune response” and “autophagy”, along with decreased “regulation of mitotic cell cycle” and “regulation of cytoskeleton organization” (Fig. 2B; Table S4), demonstrating age-related elevated inflammation and compromised supporting abilities in aged RPE-choroid layer. Generally, senescent cells are known to contribute to aging and age-related diseases by generating a low-grade inflammatory state (senescence-associated secretory phenotype, SASP). Indeed, when we enumerated the senescent state of various cell types in aged tissue, we found higher SASP score in most aged cells, especially in the RPE cells and cells from choroid layer (Figs. 2C, S5A and S5B; Table S3). Next, we performed a joint comparative analysis of aging-associated DEGs and genes from the GenAge gene set, or from the retinal disease-associated gene sets to identify the high-risk DEGs for retinal degeneration (Fig. 2D). This analysis highlighted potentially important regulators among the upregulated aging-associated DEGs, including EGFR, AGBL5, STAT3 (Fig. 2D). For instance, increased EGFR expression may lead to disruption of tight junctions in RPE and diabetic retinopathy progression (Sugimoto et al., 2013), while AGBL5, a involved in and a transcription for inflammation response et al., 2013), are whose may inflammation and pathological in aged retinal The potentially important regulators we identified as downregulated aging-associated DEGs included (Fig. 2D). is associated with retinitis pigmentosa by and the expression of retinal genes et al., 2020). is of whose is implicated in the of by of genes et al., Thus, of these transcriptional regulators may result in which by into aberrant downstream may contribute to retinal aging and diseases in the aged joint comparative analysis that the RPE cells and RGC were the cell types most affected by aging and retinal diseases in the retinal as evidenced by the most enriched overlapping genes (Figs. and For example, high-risk genes for AMD including endothelial A and were upregulated in the aged RPE cells (Fig. and transcriptional alterations in various cell types of monkey retina and choroid. (A) showing upregulated and downregulated aging-associated DEGs in different cell types. DEGs present in at two cell types are top-ranked in each The at the of each panel indicates the layer corresponding to each cell type. (B) plots showing GO terms of aging-associated DEGs in each cell type. Top, cell types from the neural retinal cell types from the RPE-choroid layer. The indicates the of a corresponding GO (C) Left, density plot showing the cell distribution with different SASP in young and old monkeys. in all different cell types are and blue to cells from young and old lines the value of SASP score in cells from young and old Student's t-test P value is indicated. Right, plots showing SASP score in representative cell types including RPE cells and cells in choroid layer. Two-tailed Student's t-test P values are indicated. SASP score is as the average expression level of genes in each (D) showing the overlapping genes between aging-associated DEGs and gene sets GenAge and retinal gene Each in this indicates each and each indicates the between a gene set and a The and color the of cell types in which a gene was differentially expressed during aging. The and color the average value of of corresponding cell Gene with than three are (E) Bar plots showing GO terms of upregulated and downregulated aging-associated DEGs in RPE cells. (F) plots showing the expression level of representative downregulated aging-associated DEGs of RPE cells. Each represents a single P values obtained from are indicated. (G) showing cell-cell interactions among different cell types. The color indicates the type of interactions in aged interactions in young unchanged interactions in young and old The indicates the of interactions between a cell and the indicates the of interactions in a cell type. (H) Bar plots showing GO terms of gene between aged RPE cells and cells in aged choroid layer and gene between aged RPE cells and cells in aged neural retina layer (I) showing potentially transcriptional regulators in downregulated aging-associated DEGs of RPE cells. The indicates the of a and the indicates the of Gene of top-ranked (ranked by the of are shown As the the neural and choroid layer, RPE cells to which is by the highly and structure et al., 2019). We found that a panel of downregulated aging-associated DEGs were enriched for organization” in aged RPE cells (Fig. including = P value = a specific marker of the TPM2 = P value = a of the as well as = P value = (Fig. a nuclear may contribute to age-related aberrant cytoskeleton and in aged RPE cells, and the signaling between its two adjacent layers. we performed cell-cell analysis across all cell types (Fig. we found that cell-cell interactions were enhanced in the aged RPE cells in old tissues in its young (Fig. Table Specifically, the between aged RPE cells and cells in neural layers were enriched in and (Figs. and suggesting that cell-cell interactions the and result in loss of retinal homeostasis in aged retinal The between aged RPE cells and cells in the choroid layer, enriched in the response” and The representative were and (Figs. and of the retinal and choroidal by retinal and vision impairment et al., 2019). we transcriptional to explore regulators of aging-associated DEGs in different cell types (Figs. and we identified several downregulated transcriptional regulators aging-associated DEGs in RPE cells, such as and (Fig. of which are involved in the signaling is for RPE et al., its dysregulation is to RPE and retinal aging in the In this we the primate single-cell transcriptome atlas of retinal aging including cellular and molecular signatures in the neural retina and the RPE-choroid layer. unique gene-expression we annotated 15 different cell types and identified several novel markers. DEG analysis revealed gene expression alterations associated with increased stress response and compromised and neural function in the neural retina layer, and an elevated inflammatory response and supporting in the RPE-choroid layer. our analyses an important of the RPE and its enhanced inflammatory response and may contribute to RPE functional with and function as a for the of an in the The retina is an model to investigate and is of therapeutic However, due to ethical human retinal tissues is Most due to the that aging is a process and across physiological aging is we a understanding of the cell-type-specific molecular mechanisms that and retinal aging and aging-related retinal diseases. With of scRNA-seq technologies, it become to molecular of the aging process by at the single-cell level (Li et al., 2020; Wang et al., 2020; Zhang et al., 2020; et al., 2020). We a primate single-cell transcriptome landscape of retinal aging, which the detailed analysis of the effect aging on diverse cell types in the neural retina and the RPE-choroid layer on the and transcriptomic data using a modified technique (Li et al., 2020; Wang et al., 2020; Zhang et al., 2020). Of we characterized a Müller glial cell subtype that retinal progenitor cells with elevated levels of a panel of progenitor cell markers. with our a that Müller glial cells from a of and a of the genes upregulated in Müller glial cells their expression in during (Lu et al., 2020). In the retina, Müller glial cells a potential for to et al., to the that this cell be in retina However, the cells and cells were not identified in our study. This may be by the by the technical of these cells with tissue digestion and/or which be transcriptomic analysis of retinal tissues in cells a of light, vulnerable to (Lin et al., 2016). we found that aging-associated DEGs were enriched for external stress and in the neural layer. we also observed impaired nucleotide-excision in the aged neural layer, which may result in the loss of neural cells in the aged the RPE is for homeostasis of the as well as supporting the structure and function of the choroidal that inflammation is an response to stress or the elevated immune response and inflammation we observed in the aged RPE cells and cells in the choroid layer, that the retinal become a stress we noticed age-related accumulation of lipofuscin in the aged RPE. A study that various of lipofuscin may immune dysregulation and activation (Lin et al., to photoreceptor (Kennedy et al., 1995). We also found that cytoskeleton was impaired in aged monkey RPE, a feature of many retinal diseases, including AMD et al., 2019). Altogether, our data identified a of underlying molecular the primate retina to external and internal stresses during physiological aging, many of which may contribute to age-related retinal In we a single-cell transcriptomic landscape of primate retinal aging, which provides understanding of diverse mechanisms underlying retinal aging in Our also as a for identifying potential biomarkers for of retinal aging-associated diseases and targets for of novel therapeutic to aging-related retinal diseases. The of this contains which is to The and for many of for their in the and tissue for their in the data for and for their with Large-scale three-dimensional of the analysis was performed on the of the for at for This was supported by the and of the of the of the and of the of the of the and of and for the of the of of the of and the of of in and and that they have no of All and for the and of animals were Table quality information of single-cell RNA-seq Table cell-type-specific marker genes and DEGs between two Müller glial subtype cells. Table the GenAge gene set and retinal gene Table DEGs and corresponding GO terms in each cell type. Table information between two cell types.

Source: Protein & Cell

DOI: 10.1007/s13238-020-00791-x