Vitreous fibrovascular membranes (FVMs), the hallmark of proliferative diabetic retinopathy (PDR), cause retinal hemorrhage, detachment, and eventually blindness. However, little is known about the pathophysiology of FVM. In this study, we used single-cell RNA sequencing on surgically harvested PDR-FVMs and generated a comprehensive cell atlas of FVM. Eight cellular compositions were identified, with microglia as the major cell population. We identified a GPNMB+ subpopulation of microglia, which presented both profibrotic and fibrogenic properties. Pseudotime analysis further revealed the profibrotic microglia was uniquely differentiated from retina-resident microglia and expanded in the PDR setting. Ligand-receptor interactions between the profibrotic microglia and cytokines upregulated in PDR vitreous implicated the involvement of several pathways, including CCR5, IFNGR1, and CD44 signaling, in the microglial activation within the PDR microenvironment. Collectively, our description of the novel microglia phenotypes in PDR-FVM may offer new insight into the cellular and molecular mechanism underlying the pathogenesis of DR, as well as potential signaling pathways amenable to disease-specific intervention.

Proliferative diabetic retinopathy (PDR) (1), one of the main diabetic complications, remains as one of the primary causes of severe vision loss diseases worldwide. The epiretinal fibrovascular membrane (FVM), as the hallmark of PDR, grows between the retina and posterior hyaloids of vitreous and can cause vitreous hemorrhage and tractional retinal detachment. However, the pathogenesis of the FVM in PDR is now poorly understood (2). Typically, neovascularization is initiated by hyperglycemic toxicity, retinal ischemia, and damage of the blood-retinal barrier, and is accompanied by an accumulation of extracellular matrix, resulting in the formation of FVMs that extend into the vitreous (3,4). Though the cellular composition of FVM has been studied histologically (5,6), the definition of cells was mostly based on traditional immunohistology method. On the other hand, most studies regarding gene expression of FVM have relied on bulk measurements of tissue or cultured cells (79). However, the cellular heterogeneity of FVMs cannot be fully explored by the above conventional approaches that have difficulties in revealing cell type–specific changes of gene expression.

Single-cell RNA sequencing (scRNA-seq) is a recently developed approach for molecular separation of different cell populations based on expression profiles (10). This emerging technology has thus opened the door to answering longstanding biological questions on heterogeneity of complex samples, identification of distinct cell subsets, and cell differentiation processes and regulatory signaling networks. In the current study, we used the scRNA-seq on surgically harvested FVMs of PDR patients and successfully identified eight biological cell clusters. Among them, microglia (MG), as the main cell population, was shown to present both profibrotic and fibrogenic properties, contributing to FVM formation. Moreover, cytokines that upregulated in PDR vitreous showed ligand-receptor interactions with FVM-resident MG. Taken together, we systematically provided a cellular classification of FVMs and potential targets for developing novel therapeutic interventions for PDR.

FVMs and Vitreous Sample Collection and Processing

The study was approved by the Institutional Review Boards at Ethic Committee of First Affiliated Hospital of Nanjing Medical University and registered at https://clinicaltrials.gov/ (ID NCT04682054 for single-cell RNA sequencing; ID NCT03506750 for measurement of vitreous cytokines). Informed consent with regard to FVM and vitreous samples was obtained from each patient in accordance with requirements of the First Affiliated Hospital of Nanjing Medical University Ethic Committee. Patients with PDR diagnosis who underwent necessary vitrectomy and with sufficient cell count in FVMs for single-cell sequencing were included. Patients who received preoperative anti-vascular endothelial growth factor (VEGF) or steroid injections were excluded from the study.

Cell count is an important issue for single-cell sequencing. In order to harvest as many cells as possible in the FVM, we performed a four-port, bimanual, pars plana vitrectomy. After excising the posterior hyaloid, FVMs were carefully dissected and removed from the retina with a cutter, forceps, or scissors. The entire FVMs or FVM pieces were then cut into pellets, which were then aspirated through the cutter tub into a 5-mL syringe manually. The pellets were then centrifuged at 800g, 4°C for 3 min, and stored in GEXSCOPE Tissue Preservation Solution (Singleron Biotechnologies, Jiangsu, China) at 4°C for transportation to the processing laboratory. The FVMs pellets were washed three times with PBS (10 mmol/L sodium phosphate, 0.15 mol/L NaCl, pH 7.4) at room temperature, followed by a standard two-step digestion procedure. First, collagenase type IV (Sigma-Aldrich, catalog no. C5138-100MG) was used to digest the FVM pellets for 15 min, and the tubules were sedimented by centrifugation at 200g for 5 min and washed with Hank's balanced salt solution. Second, FVMs were digested with TrypLE Express Enzyme (Thermo Fisher, catalog no. 12604013), sedimented, and washed again as in the first step. Single cells within the FVMs were then obtained by filtering the material through strainers with a mesh size of 70 µm (Falcon, catalog no. 352350). Afterward, the cells were pelleted by centrifugation at 200g for 5 min and washed twice with PBS.

For vitreous cytokines measurements, 31 patients with PDR and 12 patients with idiopathic macular hole (iMH) but no other retinal diseases were included. Approximately 0.5 to 0.8 mL undiluted vitreous fluid samples were collected by a vitreous cutter at the start of vitrectomy before intraocular infusion. Samples were collected into sterile tubes, immediately placed on ice, centrifuged at 1500 rpm for 5 min to remove the cells and debris, and then stored at −80°C until analyzed.

scRNA-Seq Library Construction and Sequencing

The single-cell suspensions (1 × 105 cells/mL in PBS) were loaded onto microfluidic devices and scRNA-seq libraries were constructed according to the Singleron GEXSCOPE protocol with GEXSCOPE Single-Cell RNA Library Kit (Singleron Biotechnologies, Jiangsu, China) (11). The individual library of each sample was diluted to 4 nmol/L and pooled for sequencing on the platform of the Illumina HiSeq X Ten platform.

Read Alignment and Gene Expression Matrix

The in-house pipeline was used to generate gene expression profiles from raw reads. Briefly, after removing R1 reads without poly-T tails, the cell barcodes and unique molecular identifiers (UMIs) were extracted. Meanwhile, the R2 reads were mapped to reference genome hg19 (STAR v2.5.4a and featureCounts 1.6.2) using STAR (v2.5.2b) reference. Reads with the same cell barcode, UMI, and gene were grouped together to calculate the number of UMIs per gene per cell. The total number of distinct UMI sequences for each gene was reported as the number of transcripts corresponding to that gene in the digital gene expression matrix.

Cluster Identification and Differential Expressed Gene Analysis

Seurat analysis was performed in R software using Seurat 3.1, tSNE, ggplot2, and dplyr. The data were normalized by scaling gene content by cell and log-normalizing, and variable features were identified. Data sets integrated in Seurat2 were aligned by performing canonical correlation analysis on the combined data sets and reduced into a lower dimensional space using dimensional reduction. Differential expressed genes (DEGs) were calculated using FindMarkers function in the Seurat package. Gene ontology (GO) enrichment analysis was performed on the basis of these DEGs by Metascape (https://metascape.org). The activity of gene signatures per cell was scored by gene set variation analysis (GSVA) (12). Gene coexpression network analysis was performed by GENIE3 (13) and visualized via Cytoscape 3.7.1.

Pseudotime Analysis

Pseudotime analysis was performed with Monocle 2 (14). During the pseudotime processing, the dimensionality of the data set was reduced by the Discriminative Dimensionality Reduction with Trees (DDRTree) algorithm on the log-normalized data set.

Cell-Cell Communication Analysis

Cellchat was used to infer the cell-cell communication networks from single-cell transcriptome data (15).

Integrative Analysis With Previous Data Sets

Expression matrix of published scRNA-seq data sets of human retina were downloaded from Gene Expression Omnibus (GEO) to extract the physiological retinal MG (Supplementary Fig. 3) (16). Two publicly available single-cell data sets of peripheral blood mononuclear cells (PBMCs) from healthy donors (matched with age and ethnicity) were also obtained from GEO (GSE175499). After creating the Seurat object, “FindIntegrationAnchors” and “IntegrateData” functions were used to perform the integration with our data set.

Immunofluorescence Staining

FVMs were harvested from 10 other PDR patients during surgery, were then fixed with 4% paraformaldehyde overnight, and were placed in 30% sucrose solution at 4°C overnight for cryoprotection. Afterward, FVMs were embedded in optimal cutting temperature compound (Tissue-Tek, Naperville, IL) and cut, kept at thicknesses of 8 μm, before being stored at −80°C. Primary antibodies used in this study were rabbit anti–Iba-1 (1:300; Wako Chemicals, Tokyo, Japan) and rat anti-fibronectin 1 (FN1; 1:100; Santa Cruz Biotechnology, Dallas, TX). Secondary anti-sera were Alexa Fluor 488 of donkey anti-rat (1:1000 dilution) and Alexa Fluor 594 of donkey anti-rabbit (1:1000 dilution). For immunofluorescence staining, the frozen sections were first washed with PBS three times (5 min/time) and blocked with 1% BSA (Sigma-Aldrich) solution at room temperature for 1 h. The sections were then incubated for 2 h with primary antibodies at 25°C. After being washed with PBS with 0.1% Tween 20 four times (10 min/time), sections were then incubated with fluorochrome-conjugated secondary antibodies for 1 h. Cell nuclei were counterstained with DAPI. Images were captured using a confocal microscope (Olympus 1 × 81 microscope).

Measurement of Vitreous Cytokines

The expression level of vitreous cytokines in PDR or MH vitreous samples were quantified with commercial kits (Supplementary Table 1). Vitreous Cytometric Bead Array 1 (Bio-Rad) for angiopoietin-2, chemokine (C-X-C motif) ligand (CXCL) 12, CXCL9, endostatin, fibroblast growth factor (FGF)-acidic, interleukin (IL)-18, macrophage colony-stimulating factor (M-CSF), migration inhibitory factor (MIF), matrix metalloproteinase (MMP)-1, MMP-3, MMP-7, MMP-8, platelet-derived growth factor (PDGF)-AB, PDGF-DD, placental growth factor (PLGF), periostin, VEGF-C, and VEGF-D; Cytometric Bead Array 2 (R&D) for IL-1b, IL-1ra, IL-2, IL-4, IL-5, IL-6, IL-7, IL-8, IL-9, IL-10, IL-12, IL-13, IL-15, IL-17A, eotaxin, FGF basic, G-CSF, GM-CSF, interferon-γ (IFN-γ), interferon-inducible protein (IP)-10, C-C motif chemokine ligand 2 (CCL2), macrophage inflammatory protein (MIP)-1a, PDGF-bb, MIP-1b, RANTES, tumor necrosis factor-α, and VEGF-A; and Cytometric Bead Array 3 (Bio-Rad) for transforming growth factor (TGF)-β1, TGF-β2, and TGF-β3. The Cytometric Bead Array experiment was a contract service offered by Wayen Biotechnologies, Inc. (Shanghai, China) with Bio-Plex MAGPIX System (Bio-Rad) in accordance with the manufacturer’s instructions. Cytokines like VEGF-B, NLRP3, and connective tissue growth factor (CTGF) were analyzed using human ELISA kits.

Data and Resource Availability

All data supporting the findings of this study are available within the article and its supplementary information files or upon request. The FVMs scRNA-seq data have been deposited in the GEO data sets under series number GSE165784 (https://www.ncbi.nlm.nih.gov).

Single-Cell Atlas of FVM Identifies MG as the Major Cell Population

Single cells were isolated from FVMs during vitrectomy in five patients with PDR. The clinical characteristics of these patients are shown in Table 1. A total of 6,894 cells passed filters with an average of 1,894 genes and 6,561 unique molecular identifier count per cell. Quality control metrics were highly reproducible between individual samples (Supplementary Fig. 1). Eight biological clusters were identified by unsupervised clustering following batch correction (Fig. 1A). The expression of collagen type I (COL1A1, COL1A2), the main components of fibrotic matrix, was restricted to cells of the mesenchymal lineages (fibroblast and pericyte), which account for only 8.2% (Fig. 1B and C and Supplementary Fig. 2A). Unexpectedly, the main cell population in FVM was identified as MG based on the combined expression of TREM2, C1QA, and GPR34 (Fig. 1B and Supplementary Fig. 2B) (17,18). It was further confirmed by the extensive expression of myeloid marker ionized calcium binding adapter molecule 1 (IBA-1) in FVM via immunofluorescence labeling (73.40 ± 6.03%) (Fig. 2). Other immune cells (FCN1+ monocyte/macrophage, CD2+ T cells, and CD1C+ dendrite cells) were also present in FVM (Fig. 1C and Supplementary Fig. 2A). The result implied the important function of immune cells, especially MG, in the pathogenesis of FVM.

Clustering of MG on FVM further identified four subclusters (Fig. 1A). Cluster MG(1) was enriched in the expression of GPNMB, LIPA, and FABP5. This subtype displayed the phenotype analogous to the activated MG accumulated around lipid plaque of diverse neurodegenerative disease (19) (Supplementary Fig. 2D). MG(3) expressed a high level of SELENOP, MRC1, and IGF1, which are often associated with anti-inflammatory activation (Supplementary Fig. 2E). In addition, we observed a CX3CR1+ cluster MG(2). CX3CR1 signaling of resident MG was previously demonstrated to maintain the homeostasis of the retina and serves as a primed responder of the surrounding insult (20). We therefore annotated the CX3CR1+ cluster as preactive state (Fig. 1D). Gene correlation network further highlighted the three independent gene networks that corresponded to each MG subtype (Fig. 1E). Of note, a cluster that highly expressed multiple cell cycle-related genes (e.g., MKI67, CENPF) was identified, indicating the presence of MG proliferation on FVM (Fig. 1C and D).

The Origin of Activated MG on FVM

Under homeostatic condition, tissue-resident MG predominate in a steady state since the embryonic stage. However, in the diseased state, especially when the interruption of blood-retinal barrier occurs, MG may activate and multiply, relying on the local expansion of MG or the recruitment of monocytes from circulation (21,22). To investigate the ontogeny of FVM-resident MG, we integrated the single-cell transcriptomic 2 data sets of PBMCs and two data sets of retinal MG with ours (Fig. 3A). The transcriptional profiles of the combined data set were mapped along a pseudotemporal trajectory. The analysis suggested the transcriptional similarity between retinal MG and FVM-resident MG, which was segregated from monocyte-to-macrophage differentiation (Fig. 3B).

To characterize the cellular dynamics during the transition of MG, we investigated the distribution of each MG subtype along pseudotime. The cycling MG and CX3CR1+ MG(2) were positioned at the starting point of the trajectory, which gradually transitioned to either MG(1) or MG(3) (Supplementary Fig. 4C). Interestingly, the monocyte-derived macrophage was also connected with the cluster of MG(2) on the Uniform Manifold Approximation and Projection (UMAP) plot, indicating the identical fate determined by two differentiation paths (Fig. 3C). We next compared the expression patterns of MG subtype-specific genes between retinal MG and FVM-resident MG. Significant downregulation of the preactive marker CX3CR1 and upregulation of the MG(1) marker GPNMB were observed in FVM-resident MG compared with retinal MG. However, the expression of MG(3) marker MRC1 were altered to a lesser extent (Fig. 3D). Together, these results indicated MG(1) as the major activation state of MG resident on FVM, which highlights the potential role of MG(1) in the pathogenesis of FVM.

Additionally, upregulated genes along the transitioning trajectory were enriched for the ontology terms that are associated with the myeloid cell migration, activation, and proliferation (Supplementary Fig. 4A and B). The result further supported the notion that MG on FVM was recruited from retina and subsequently activated to MG(1) state within the PDR-associated niche.

Profibrotic and Fibrogenic Phenotype of MG Orchestrates the Progression of FVM Formation

To further delineate the functional profile of MG(1) on FVM, we performed gene set variation analysis (GSVA) on single-cell transcriptomic data to investigate the effector pathways for the formation of FVM. GSVA revealed that fibrosis-related gene expression was significantly enriched in MG(1) compared with other cell types (Fig. 4A). The fibrosis-related signature included genes such as CTSB, SPP1, and LGALS3, which are known to regulate fibroblast activation in fibrosis (Fig. 4C). This gene signature was also found in the scar-associated macrophage recently identified in a single-cell transcriptome study of liver fibrosis (Supplementary Fig. 5) (23). Furthermore, the gene set that associated with extracellular matrix was upregulated in MG(1) as well as in mesenchymal cells (Fig. 4B). We found that fibronectin (FN1), osteonectin (SPARC), and TGF–β–induced protein (TGFBI), the extracellular matrices that previously were found abundantly in epiretinal membrane (24), were highly expressed in MG(1) (Fig. 4C and E). These associated gene signatures were weakly displayed in the retinal MG under physiological conditions (Fig. 4D). Fluorescence staining of FVM also demonstrated the presence of IBA1+ myeloid cells topographically localized in fibronectin-positive regions (Fig. 5). Taken together, our data highlighted that the activated MG on FVM represent both profibrotic and fibrogenic properties.

Intercellular Interaction of Profibrotic MG Within PDR Microenvironment

To explore the effect of PDR niche on the activation of the profibrotic MG, we first evaluated cytokines in vitreous samples from PDR and iMH patients using multiplex liquid-chip assay and ELISA (Table 2). The concentrations of VEGF-B and NLRP3 were below the limit of ELISA detection, thus the two cytokines were not included. The upregulated cytokines in PDR vitreous were also compared with publications (2529) (Supplementary Table 1). Here, we identified 23 cytokines that significantly upregulated in PDR vitreous compared with control (Fig. 6A). Among them, cytokines ANGPT2, TGFB3, and CTGF were found exclusively expressed in the mesenchymal cells of FVM. In consistent with previous studies (30), MG remained to be the major source of CCL2, CCL3, and CCL4 among all cell types (Fig. 6B). This indicated that the paracrine effect of FVM-resident cells may partially contribute to the pathological microenvironment.

We next extracted the receptors of these 23 cytokines from the CellTalker database and evaluated their expression level among different cell types on FVM. Receptors that specifically expressed in each cell type were identified (Fig. 6C). NRP1, NRP2, TGFBR1, and LRP1 were specifically expressed in both MG(1) and mesenchymal cells, whose cognate ligands VEGFA, TGFB, and CTGF have been demonstrated to induce fibrogenesis in various organs (31,32). In addition, MG(1) expressed a subset of receptors that were also enriched in monocyte and dendrite cells, such as IFNGR, CD44, CD74, CCR1, and CCR5. These receptors generally interact with the ligands IFNG, MIF, and CCL, which are critical regulators of macrophage mobilization and activation (33) (Fig. 6C). Notably, receptors CCR1, CCR5, IFNGR1, CD44, CXCR4, and NRP1 were significantly upregulated in FVM-resident MG compared with the healthy retinal MG, indicating their potential function in the activation of MG(1) state on FVM (Fig. 6D). We therefore inferred their cognate ligands VEGFA, TGFB, MIF, IFNG, and CCL as the extrinsic regulators for the activation of profibrotic MG within the PDR niche.

To further investigate the signaling of profibrotic MG that contribute to fibroblast activation, we applied CellChat to perform an unbiased ligand-receptor interaction analysis between different cell clusters (Supplementary Fig. 6). Among numerous statistically significant interactions between MG(1) and fibroblasts, SPP1 was identified as the ligand mainly expressed in MG(1) and contributed to fibroblast-dominated signaling that participates in tissue remodeling and cell-matrix cross talk (Fig. 6E). Taken together, our dissection of key ligand-receptor interactions of MG highlights the potential regulators of FVM pathogenesis within the PDR niche.

As the important hallmark of PDR, the FVM and its pathophysiological processes might be the key to understanding the mechanism of DR. In this current study, we surveyed the differential gene expression profiles of the cells, especially the MG within the FVM, at single-cell resolution for the first time. Our study also demonstrated the feasibility of scRNA-seq for comprehensive mapping of gene expression signatures of surgically incised, small-sized, human pathological tissue compared with other relatively larger tissues (e.g., retina, outflow tissues) (16,3439). Here, we identified eight different cell types in FVM tissues, among which MG was the major cell population. Importantly, we further revealed that the MG in FVMs originated from retinal MG, with its MG(1) subtype presenting both profibrotic and fibrogenic properties. These results provide the basis for a better appreciation of cell composition and functions of human FVM as well as future potential development of interventional target.

Because of the small size and limited material availability, few studies concerning the cell compositions of FVMs have been reported. Coltrini et al. (40) recently reported using quantitative RT-PCR analysis to identify two distinct idiopathic epiretinal membrane clusters based on 20 selected genes. However, the cellular heterogeneity of cell components in disease-specific FVMs and in-depth intercellular interactions cannot be fully explored. In our study, we identified eight cell clusters in FVMs: MG, monocytes, dendrite cells, fibroblasts, endothelia, pericytes, and T cells. This finding indicated the high-quality and robust data to uncover in-depth FVM-related gene expressions at single-cell level.

Most of the previous reports individually focused on specific cells on FVMs, such as retinal pigment epithelium cells, glial cells, and fibroblasts (41). It is also generally believed that the glial components of the FVM come from Müller cells and astrocytes (42). Previous histological studies usually identified cells using marker proteins, such as glial fibrillary acidic protein-positive (GFAP), glutamine synthase (GS), and vimentin for Müller cell (43). However, almost all tissues include a heterogeneous mix of cell types. Single-cell gene expression studies have enabled the characteristics of previously known cell types to be more fully defined and facilitated the identification of novel categories of cells (44). For instance, in our study, the MG cell type also expressed Müller cell canonical markers, GS and vimentin, yet the global gene expression profiles of this cell cluster defined them as MG based on their combined expression of AIF1 and C1QA. Thus, this current study may upend people’s understanding of the cellular components in FVMs. Here, MG, identified as the major cell population, was also verified by immunofluorescence staining. Previously, Zeng et al. (45) also found cells in FVMs were heavily labeled with MG markers. Besides, in experimental model of DR (46), elimination of MG and macrophages reduced retinal vascular leakage of DR (46). Of note, we not only verified the MG as the major cell population but further found that the activation of MG(1) subtype was strongly associated with FVM formation. Furthermore, the activated MG also expressed SPP1, which acts on other cell types in FVMs, participating in tissue remodeling and cell-matrix cross talk. In addition, upregulated vitreous cytokines (ligands), such as CCL, MIF, VEGF-A, and TGF-β, were found have high expression of corresponding receptor genes in MG, indicating the specific response of the resident MG to the disease microenvironment. In the future, the downstream consequence of those interactions in the progression of FVMs remains a topic of further investigation.

DR has been diagnosed and classified clinically based on the vasculature changes by fundoscopy. However, recent technological advances have confirmed neuroretina pathology prior to any detectable vascular alteration (47). These, coupled with molecular evidence, have highlighted the role of dysfunctional immune system in the pathophysiology of diabetic complications (48). In this study, we demonstrated that MG cells play a central role in end-stage diabetic retinal disorder and that interfering with the transcriptome profile of MG might modulate their functionality and pathogenicity in DR. Further studies are needed to verify the origin of the FVM-resident MG using the combined analysis of both local tissue and systematic monocytes.

Of note, this study has several limitations. Firstly, despite the difficulty of harvesting FVMs in the clinic, the sample size of FVMs sent for scRNA-seq was relatively small. Secondly, we used iMH as a control in the measurement of vitreous cytokines; vitreous floaters would be a more ideal control in future studies since macular holes have been shown to cause changes in the vitreous proteomics (49). Thirdly, there might be heterogeneity of the included FVMs concerning the varied medical history, hemoglobin A1c value, and visual prognosis (Supplementary Table 2). Finally, we did not included patients who had received preoperative anti-VEGF or steroid injections, which calls for further investigation.

In conclusion, we have generated a cell atlas of the human FVMs in PDR. Using scRNA-seq, we have emphasized that activated MG participate in the FVM formation. We believe our study highlights the feasibility of in-depth single-cell analysis on other tissues with diabetic complications, such as diabetic neuropathy and diabetic foot. Our detailed scRNA-seq also shed light on better understanding the critical role of MG and the molecular mechanism underlying FVMs and improvement of current preventative and therapeutic strategies for DR.

Z.H. and X.M. contributed equally to this work.

Clinical trial reg. nos. NCT04682054 and NCT03506750, clinicaltrials.gov

This article contains supplementary material online at https://doi.org/10.2337/figshare.18623243.

Acknowledgments. The authors thank Prof. Qianghu Wang (Nanjing Medical University) for the bio-informative assistance.

Funding. This study was supported by the National Key Research and Development Program of China (2017YFA0104101 to Q.L., and 2017YFA0104103 to S.Y.), the National Natural Science Foundation of China (81970821 to Q.L., 81900875 to Z.H., and 81870694 to S.Y.), the Special-funded Program on National Key Scientific Instruments and Equipment Development (12027808 to Z.H.), the Key Research and Development Program of Jiangsu Province (BE2018131 to S.Y.), and the Natural Science Foundation of Jiangsu Province (BK20191059 to Z.H.).

The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Z.H. and X.M. wrote the manuscript and researched data. Z.H., X.M., X.W., and T.Z. researched data. Z.H., S.Y., and Q.L. obtained funding and contributed to the design. M.C., Y.L., Z.Z., and W.F. researched data and contributed to discussion. P.X. contributed to resources and visualization. S.Y. and Q.L. reviewed and edited the manuscript. S.Y. and Q.L. contributed to discussion and reviewed and edited manuscript. S.Y. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

1.
Nawaz
IM
,
Rezzola
S
,
Cancarini
A
, et al
.
Human vitreous in proliferative diabetic retinopathy: characterization and translational implications
.
Prog Retin Eye Res
2019
;
72
:
100756
2.
Vishwakarma
S
,
Gupta
RK
,
Jakati
S
, et al
.
Molecular assessment of epiretinal membrane: activated microglia, oxidative stress and inflammation
.
Antioxidants (Basel)
2020
;
9
:
654
3.
Hu
Z
,
Su
Y
,
Xie
P
, et al
.
OCT angiography-based monitoring of neovascular regression on fibrovascular membrane after preoperative intravitreal conbercept injection. Graefe’s archive for clinical and experimental ophthalmology
.
Albrecht Von Graefes Arch Klin Exp Ophthalmol
2019
;
257
:
1611
-
1619
4.
Tamaki
K
,
Usui-Ouchi
A
,
Murakami
A
,
Ebihara
N
.
Fibrocytes and fibrovascular membrane formation in proliferative diabetic retinopathy
.
Invest Ophthalmol Vis Sci
2016
;
57
:
4999
5005
5.
El-Sabagh
HA
,
Abdelghaffar
W
,
Labib
AM
, et al
.
Preoperative intravitreal bevacizumab use as an adjuvant to diabetic vitrectomy: histopathologic findings and clinical implications
.
Ophthalmology
2011
;
118
:
636
641
6.
Ishikawa
K
,
Yoshida
S
,
Nakao
S
, et al
.
Periostin promotes the generation of fibrous membranes in proliferative vitreoretinopathy
.
FASEB J
2014
;
28
:
131
142
7.
Asato
R
,
Yoshida
S
,
Ogura
A
, et al
.
Comparison of gene expression profile of epiretinal membranes obtained from eyes with proliferative vitreoretinopathy to that of secondary epiretinal membranes
.
PLoS One
2013
;
8
:
e54191
8.
Zhou
RM
,
Wang
XQ
,
Yao
J
, et al
.
Identification and characterization of proliferative retinopathy-related long noncoding RNAs
.
Biochem Biophys Res Commun
2015
;
465
:
324
330
9.
Huang
H
,
Saddala
MS
,
Lennikov
A
,
Mukwaya
A
,
Fan
L
.
RNA-Seq reveals placental growth factor regulates the human retinal endothelial cell barrier integrity by transforming growth factor (TGF-β) signaling
.
Mol Cell Biochem
2020
;
475
:
93
106
10.
Klein
AM
,
Mazutis
L
,
Akartuna
I
, et al
.
Droplet barcoding for single-cell transcriptomics applied to embryonic stem cells
.
Cell
2015
;
161
:
1187
1201
11.
Dura
B
,
Choi
JY
,
Zhang
K
, et al
.
scFTD-seq: freeze-thaw lysis based, portable approach toward highly distributed single-cell 3′ mRNA profiling
.
Nucleic Acids Res
2019
;
47
:
e16
12.
Hänzelmann
S
,
Castelo
R
,
Guinney
J
.
GSVA: gene set variation analysis for microarray and RNA-seq data
.
BMC Bioinformatics
2013
;
14
:
7
13.
Huynh-Thu
VA
,
Irrthum
A
,
Wehenkel
L
,
Geurts
P
.
Inferring regulatory networks from expression data using tree-based methods
.
PLoS One
2010
;
5
:
e12776
14.
Qiu
X
,
Mao
Q
,
Tang
Y
, et al
.
Reversed graph embedding resolves complex single-cell trajectories
.
Nat Methods
2017
;
14
:
979
982
15.
Jin
S
,
Guerrero-Juarez
CF
,
Zhang
L
, et al
.
Inference and analysis of cell-cell communication using CellChat
.
Nat Commun
2021
;
12
:
1088
16.
Cowan
C
,
Renner
M
,
De Gennaro
M
, et al
.
Cell types of the human retina and its organoids at single-cell resolution
.
Cell
2020
;
182
:
1623
1640.e1634
17.
Hammond
TR
,
Dufort
C
,
Dissing-Olesen
L
, et al
.
Single-cell RNA sequencing of microglia throughout the mouse lifespan and in the injured brain reveals complex cell-state changes
.
Immunity
2019
;
50
:
253
271.e6
18.
Olah
M
,
Menon
V
,
Habib
N
, et al
.
Single cell RNA sequencing of human microglia uncovers a subset associated with Alzheimer’s disease
.
Nat Commun
2020
;
11
:
6129
19.
Jaitin
DA
,
Adlung
L
,
Thaiss
CA
, et al
.
Lipid-associated macrophages control metabolic homeostasis in a Trem2-dependent manner
.
Cell
2019
;
178
:
686
698.e14
20.
Zhang
Y
,
Zhao
L
,
Wang
X
, et al
.
Repopulating retinal microglia restore endogenous organization and function under CX3CL1-CX3CR1 regulation
.
Sci Adv
2018
;
4
:
eaap8492
21.
Jin
N
,
Gao
L
,
Fan
X
,
Xu
H
.
Friend or foe? Resident microglia vs bone marrow-derived microglia and their roles in the retinal degeneration
.
Mol Neurobiol
2017
;
54
:
4094
4112
22.
Ginhoux
F
,
Garel
S
.
The mysterious origins of microglia
.
Nat Neurosci
2018
;
21
:
897
899
23.
Ramachandran
P
,
Dobie
R
,
Wilson-Kanamori
JR
, et al
.
Resolving the fibrotic niche of human liver cirrhosis at single-cell level
.
Nature
2019
;
575
:
512
518
24.
Christakopoulos
C
,
Cehofski
LJ
,
Christensen
SR
,
Vorum
H
,
Honoré
B
.
Proteomics reveals a set of highly enriched proteins in epiretinal membrane compared with inner limiting membrane
.
Exp Eye Res
2019
;
186
:
107722
25.
Suzuki
Y
,
Nakazawa
M
,
Suzuki
K
,
Yamazaki
H
,
Miyagawa
Y
.
Expression profiles of cytokines and chemokines in vitreous fluid in diabetic retinopathy and central retinal vein occlusion
.
Jpn J Ophthalmol
2011
;
55
:
256
263
26.
Yu
Y
,
Zhang
J
,
Zhu
R
, et al
.
The profile of angiogenic factors in vitreous humor of the patients with proliferative diabetic retinopathy
.
Curr Mol Med
2017
;
17
:
280
286
27.
Xu
Y
,
Cheng
Q
,
Yang
B
, et al
.
Increased sCD200 levels in vitreous of patients with proliferative diabetic retinopathy and its correlation with VEGF and proinflammatory cytokines
.
Invest Ophthalmol Vis Sci
2015
;
56
:
6565
6572
28.
Urbančič
M
,
Petrovič
D
,
Živin
AM
,
Korošec
P
,
Fležar
M
,
Petrovič
MG
.
Correlations between vitreous cytokine levels and inflammatory cells in fibrovascular membranes of patients with proliferative diabetic retinopathy
.
Mol Vis
2020
;
26
:
472
482
29.
Tsai
T
,
Kuehn
S
,
Tsiampalis
N
, et al
.
Anti-inflammatory cytokine and angiogenic factors levels in vitreous samples of diabetic retinopathy patients
.
PLoS One
2018
;
13
:
e0194603
30.
Deshmane
SL
,
Kremlev
S
,
Amini
S
,
Sawaya
BE
.
Monocyte chemoattractant protein-1 (MCP-1): an overview
.
J Interferon Cytokine Res
2009
;
29
:
313
326
31.
Lopes
TG
,
de Souza
ML
,
da Silva
VD
, et al
.
Markers of renal fibrosis: how do they correlate with podocyte damage in glomerular diseases?
PLoS One
2019
;
14
:
e0217585
32.
Lipson
KE
,
Wong
C
,
Teng
Y
,
Spong
S
.
CTGF is a central mediator of tissue remodeling and fibrosis and its inhibition can reverse the process of fibrosis
.
Fibrogenesis Tissue Repair
2012
;
5
(
Suppl. 1
):
S24
33.
Sica
A
,
Mantovani
A
.
Macrophage plasticity and polarization: in vivo veritas
.
J Clin Invest
2012
;
122
:
787
795
34.
Lu
Y
,
Shiau
F
,
Yi
W
, et al
.
Single-cell analysis of human retina identifies evolutionarily conserved and species-specific mechanisms controlling development
.
Dev Cell
2020
;
53
:
473
491.e9
35.
Liang
Q
,
Dharmat
R
,
Owen
L
, et al
.
Single-nuclei RNA-seq on human retinal tissue provides improved transcriptome profiling
.
Nat Commun
2019
;
10
:
5743
36.
Voigt
AP
,
Mulfaul
K
,
Mullin
NK
, et al
.
Single-cell transcriptomics of the human retinal pigment epithelium and choroid in health and macular degeneration
.
Proc Natl Acad Sci U S A
2019
;
116
:
24100
24107
37.
He
J
,
Ou
S
,
Ren
J
, et al
.
Tissue engineered corneal epithelium derived from clinical-grade human embryonic stem cells
.
Ocul Surf
2020
;
18
:
672
680
38.
Patel
G
,
Fury
W
,
Yang
H
, et al
.
Molecular taxonomy of human ocular outflow tissues defined by single-cell transcriptomics
.
Proc Natl Acad Sci U S A
2020
;
117
:
12856
12867
39.
van Zyl
T
,
Yan
W
,
McAdams
A
, et al
.
Cell atlas of aqueous humor outflow pathways in eyes of humans and four model species provides insight into glaucoma pathogenesis
.
Proc Natl Acad Sci U S A
2020
;
117
:
10339
10349
40.
Coltrini
D
,
Belleri
M
,
Gambicorti
E
, et al
.
Gene expression analysis identifies two distinct molecular clusters of idiopatic epiretinal membranes
.
Biochim Biophys Acta Mol Basis Dis
2020
;
1866
:
165938
41.
Tosi
GM
,
Marigliani
D
,
Romeo
N
,
Toti
P
.
Disease pathways in proliferative vitreoretinopathy: an ongoing challenge
.
J Cell Physiol
2014
;
229
:
1577
1583
42.
Harada
C
,
Mitamura
Y
,
Harada
T
.
The role of cytokines and trophic factors in epiretinal membranes: involvement of signal transduction in glial cells
.
Prog Retin Eye Res
2006
;
25
:
149
164
43.
Delaunay
K
,
Khamsy
L
,
Kowalczuk
L
, et al
.
Glial cells of the human fovea
.
Mol Vis
2020
;
26
:
235
245
44.
Potter
SS
.
Single-cell RNA sequencing for the study of development, physiology and disease
.
Nat Rev Nephrol
2018
;
14
:
479
492
45.
Zeng
HY
,
Green
WR
,
Tso
MOM
.
Microglial activation in human diabetic retinopathy
.
Arch Ophthalmol
2008
;
126
:
227
232
46.
Monickaraj
F
,
Oruganti
SR
,
McGuire
P
,
Das
A
.
A potential novel therapeutic target in diabetic retinopathy: a chemokine receptor (CCR2/CCR5) inhibitor reduces retinal vascular leakage in an animal model
.
Graefes Arch Clin Exp Ophthalmol
2021
;
259
:
93
100
47.
Carpineto
P
,
Toto
L
,
Aloia
R
, et al
.
Neuroretinal alterations in the early stages of diabetic retinopathy in patients with type 2 diabetes mellitus
.
Eye (Lond)
2016
;
30
:
673
679
48.
Pan
WW
,
Lin
F
,
Fort
PE
.
The innate immune system in diabetic retinopathy
.
Prog Retin Eye Res
2021
;
84
:
100940
49.
Zhang
P
,
Zhu
M
,
Zhao
Y
, et al
.
A proteomic approach to understanding the pathogenesis of idiopathic macular hole formation
.
Clin Proteomics
2017
;
14
:
37
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at https://www.diabetesjournals.org/journals/pages/license.