A single cell atlas of human cornea that defines its development, limbal progenitor cells and their interactions with the immune cells

Abstract

Purpose: Single cell (sc) analyses of key embryonic, fetal and adult stages were performed to generate a comprehensive single cell atlas of all the corneal and adjacent conjunctival cell types from development to adulthood.

Methods: Four human adult and seventeen embryonic and fetal corneas from 10 to 21 post conception week (PCW) specimens were dissociated to single cells and subjected to scRNA- and/or ATAC-Seq using the 10x Genomics platform. These were embedded using Uniform Manifold Approximation and Projection (UMAP) and clustered using Seurat graph-based clustering. Cluster identification was performed based on marker gene expression, bioinformatic data mining and immunofluorescence (IF) analysis. RNA interference, IF, colony forming efficiency and clonal assays were performed on cultured limbal epithelial cells (LECs).

Results: scRNA-Seq analysis of 21,343 cells from four adult human corneas and adjacent conjunctivas revealed the presence of 21 cell clusters, representing the progenitor and differentiated cells in all layers of cornea and conjunctiva as well as immune cells, melanocytes, fibroblasts, and blood/lymphatic vessels. A small cell cluster with high expression of limbal progenitor cell (LPC) markers was identified and shown via pseudotime analysis to give rise to five other cell types representing all the subtypes of differentiated limbal and corneal epithelial cells. A novel putative LPCs surface marker, GPHA2, expressed on the surface of 0.41% ± 0.21 of the cultured LECs, was identified, based on predominant expression in the limbal crypts of adult and developing cornea and RNAi validation in cultured LECs. Combining scRNA- and ATAC-Seq analyses, we identified multiple upstream regulators for LPCs and demonstrated a close interaction between the immune cells and limbal progenitor cells. RNA-Seq analysis indicated the loss of GPHA2 expression and acquisition of proliferative limbal basal epithelial cell markers during ex vivo LEC expansion, independently of the culture method used. Extending the single cell analyses to keratoconus, we were able to reveal activation of collagenase in the corneal stroma and a reduced pool of limbal suprabasal cells as two key changes underlying the disease phenotype. Single cell RNA-Seq of 89,897 cells obtained from embryonic and fetal cornea indicated that during development, the conjunctival epithelium is the first to be specified from the ocular surface epithelium, followed by the corneal epithelium and the establishment of LPCs, which predate the formation of limbal niche by a few weeks.

Conclusions: Our scRNA-and ATAC-Seq data of developing and adult cornea in steady state and disease conditions provide a unique resource for defining genes/pathways that can lead to improvement in ex vivo LPCs expansion, stem cell differentiation methods and better understanding and treatment of ocular surface disorders.

Keywords: Conjunctiva; Cornea; Embryonic and fetal eye; Keratoconus; LSCs dysplasia; Limbal epithelial cells (LECs); Limbal epithelial expansion; Limbal progenitor cells (LPCs); Limbal stem cells (LSCs); Ocular surface; Single cell ATAC-Seq; Single cell RNA-Seq.

Graphical abstract

Fig. 1

scRNA-Seq analysis of adult human cornea and conjuctiva (see also Figures S1-S12 and Table S1). A) Microphotograph of adult human ocular surface before scRNA-Seq analysis (left panel) and UMAPs of each adult human cornea and conjunctiva (middle panel). Following data integration and analysis, an integrated UMAP revealing the presence of 21 cell clusters was generated (right panel); B) Violin plots showing the presence of key markers for stem, progenitor and differentiated cells in the epithelial, stromal, and endothelial compartments. Abbreviations for panel 1A: BV – blood vessels, CB – corneal basal epithelium, CS – corneal superficial epithelium, CSB – corneal suprabasal epithelial cells, CjB – conjunctival basal epithelium, CjS – conjunctival superficial epithelium, CE-corneal endothelium, CSK – corneal stroma keratocytes, CSSCs – corneal stromal stem cells, FCEC – fibroblastic corneal endothelial cells, IC1 - immune cells 1, IC2 – immune cells 2, LF – limbal fibroblasts, LSB – limbal suprabasal epithelial cells, LSK – limbal stroma keratocytes, LNCPs – limbal neural crest progenitors, LPCs –limbal progenitor cells, LS – limbal superficial epithelium, LV – lymphatic vessels, Mel - melanocytes, RBC – red blood cells, Additional abbreviations for panel 1B: Conj– conjunctival Ep-epithelium, Fib-fibroblasts, End-endothelial.

Fig. 2

Novel markers for LPCs. A) Violin plots showing the expression of five transcripts (GPHA2, CASP14, MMP1, MMP10 and AC093496.1), which are highly and predominantly expressed in LPCs (cluster 9); B-D) IF analysis showing overlapping GPHA2 and KRT15 expression (B, B′, B′), rare co-localisation of GPHA2 with Ki67 expression shown by white arrow (C) and overlapping GPHA2 and MMP10 expression in the limbal crypts (D); E-G) IF showing expression of ΔNp63 (E), p27 (F) and CEBPδ (G) in the limbal crypts; H–I) faint background like-staining of GPHA2 in the conjunctival (H) and central cornea region (I); J-L) MMP1 is expressed in the conjunctival (J), limbal epithelium (K) and corneal epithelium(L); M − P), MMP10 expression is found in the limbal (M), conjunctival (O) and corneal epithelium (P); KRT15 expression in the limbal and corneal epithelium shown in N and Q panels represent sister sections to MMP10 expression in M and P panels; R) Representative negative control immunostaining. Representative images from four different human cornea and conjunctiva samples are shown. Scale bars: 50 μm. Nuclear staining indicated by Hoechst in blue colour.

Fig. 3

GPHA2 supports LPC undifferentiated state in vitro. A-C) Expression of GPHA2 and overlap with KRT15, ΔNp63 and Ki67 respectively in ex vivo expanded LECs, passage 1; D) Downregulation of GPHA2 and KRT15 expression and upregulation of differentiated corneal epithelial markers, KRT3, KRT12, MUC1 and MUC16 upon air liquid induced differentiation of LECs; E) Flow activated cell sorting combined with qRT-PCR and clonal analyses indicate enrichment of cells with holoclone forming ability in the GPHA2+ enriched cell fraction; F) Quantitative RT-PCR analysis showing knockdown of GPHA2 expression in ex vivo expanded LECs resulting in a significant decrease in KRT15 and increase in KRT3 and KRT12 expression; G) CFE is significantly reduced upon GPHA2 knockdown . E–G: Data presented as mean ± SEM, n = 3 for panels E–F and n = 7 for panel G. Statistical significance was assessed using one-way Anova with Dunnet Multiple Comparison Tests, *p < .05; **p < .001, ***p < .001, ****p < .0001.

Fig. 4

scATAC-Seq of adult human cornea and conjuctiva (see also Figure S13, Table S3). A) UMAP of adult human cornea and conjunctiva epithelial clusters and limbal neural crest progenitors; B-E) Schematic single cell chromatin accessibility of KRT14 (B), CASP14 (C), MUC22 (D) and EHD1 (containing a distal enhancer for GPHA2) (E) in the human cornea and conjunctiva epithelial clusters and LNCPs.

Fig. 5

scRNA-Seq of ex vivo expanded human LECs and comparison to adult human cornea and conjunctiva (see also Figure S16 and Table S6). A) UMAP of ex vivo expanded LECs integrated with the four adult human cornea/conjunctiva samples shown in Fig. 1A, revealing the presence of three additional cell clusters found in the ex vivo expanded LECs. Although all cluster annotations are the same as in Fig. 1A, during the integration process the original 18 and 11 were combined in the integrated cluster 11; B) Cell cycle distribution of additional clusters 1–3; C) Comparative heatmap showing differentially expressed genes between the three additional clusters found in the ex vivo expanded LECs.

Fig. 6

Expansion of LECs in vivo leads to loss of GPHA2 expression and acquisition of markers associated with proliferative limbal basal epithelial cells (see also Figures S17, S18 and Table S8). A) Representative photo showing a human cornea with a limbal dysplasia; B) UMAP of limbal dysplasia sample integrated with ex vivo expanded LECs (Fig. 5A) and the four adult human cornea/conjunctival samples (Fig. 1A) showing the presence of six additional clusters in the cornea with limbal dysplasia. Although all cluster annotations are the same as in Fig. 1A, during the integration process the original cluster 7 was combined with original cluster 2 in the integrated cluster 6 and clusters 18 and 11 were combined in the integrated cluster 11; C) Comparative heatmap showing the differentially expressed genes between the six additional clusters found in the cornea with limbal dysplasia; D) Comparative heatmap showing differentially expressed genes between LPCs (cluster 9) and the proliferative basal limbal epithelium I-III corresponding to additional clusters 1, 3, 5 in the cornea with limbal dysplasia.

Fig. 7

scRNA-Seq of keratoconus samples (see also Figures S19, S20 and Tables S9-S12). A) UMAP of central cornea samples obtained from two patients with keratoconus and one unaffected subject integrated with the four adult human cornea/conjunctiva samples shown in Fig. 1A. Although all cluster annotations are the same as in Fig. 1A, during the integration process the original, cluster 19 and cluster 20 matched just one cluster in the integrated clustering (shown as cluster 19), as did cluster 15 and 17 (shown as cluster 15); B) Comparative analysis between the unaffected and keratoconus cornea showing the percentage of cells in each cluster.

Fig. 8

scRNA-Seq of embryonic and fetal cornea and conjunctiva from 10 to 21 PCW with cluster annotations (see also Table S13 and Figures S21-S26).

figs1

figs2

figs3

figs4

figs5

figs6

figs7

figs8

figs9

figs10

figs11

figs12

figs13

figs14

figs15

figs16

figs17

figs18

figs19

figs20

figs21

figs22

figs23

figs24

figs25

figs26