CoChat

Mapping the temporal and spatial dynamics of the human endometrium in vivo and in vitro

Open access
Garcia-Alonso, Luz, Handfield, Louis-François, Roberts, Kenny, Nikolakopoulou, Konstantina, Fernando, Ridma C., Gardner, Lucy, Woodhams, Benjamin, Arutyunyan, Anna, Polanski, Krzysztof, Hoo, Regina, Sancho-Serra, Carmen, Li, Tong, Kwakwa, Kwasi, Tuck, Elizabeth, Lorenzi, Valentina, Massalha, Hassan, Prete, Martin, Kleshchevnikov, Vitalii, Tarkowska, Aleksandra, Porter, Tarryn, Mazzeo, Cecilia Icoresi, van Dongen, Stijn, Dabrowska, Monika, Vaskivskyi, Vasyl, Mahbubani, Krishnaa T., Park, Jong-eun, Jimenez-Linan, Mercedes, Campos, Lia, Kiselev, Vladimir Yu., Lindskog, Cecilia, Ayuk, Paul, Prigmore, Elena, Stratton, Michael R., Saeb-Parsy, Kourosh, Moffett, Ashley, Moore, Luiza, Bayraktar, Omer A., Teichmann, Sarah A., Turco, Margherita Y., Vento-Tormo, Roser · 2021 · Nature Genetics

Published version of a preprint

The DOI matches its Crossref record; 70 of the 70 references it cites were found in a registry or an index.

Abstract

The endometrium, the mucosal lining of the uterus, undergoes dynamic changes throughout the menstrual cycle in response to ovarian hormones. We have generated dense single-cell and spatial reference maps of the human uterus and three-dimensional endometrial organoid cultures. We dissect the signaling pathways that determine cell fate of the epithelial lineages in the lumenal and glandular microenvironments. Our benchmark of the endometrial organoids reveals the pathways and cell states regulating differentiation of the secretory and ciliated lineages both in vivo and in vitro. In vitro downregulation of WNT or NOTCH pathways increases the differentiation efficiency along the secretory and ciliated lineages, respectively. We utilize our cellular maps to deconvolute bulk data from endometrial cancers and endometriotic lesions, illuminating the cell types dominating in each of these disorders. These mechanistic insights provide a platform for future development of treatments for common conditions including endometriosis and endometrial carcinoma.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/s41588-021-00972-2

Cites

70 of 70 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Single-cell profiling of the human uterus. a, Schematic illustration of the human uterus showing the different layers and the morphological changes seen throughout the menstrual cycle with respect to tissue sampling. b, UMAP projections of scRNA-seq data from a total of 15 individuals. c, UMAP representations colored by menstrual phase. d, UMAP of subclustered immune populations. e, UMAP projections of snRNA-seq data from a total of four individuals in the proliferative phase. f, Radial representation of the cosine distance similarity for single cells obtained from snRNA-seq to the centroids of cell types defined by scRNA-seq. g, Estimated amount of mRNA (color intensity) contributed by each cell population to each spot (color) shown over the hematoxylin and eosin (H&E) image of the secretory endometrium (A30, 152811 slide). h, Dot plot showing log2-transformed expression of genes expressed in fibroblast and stromal subsets. i, Estimated amount of mRNA (color intensity) contributed by each cell population to each spot (color) shown over the H&E image of proliferative (A30, 152810 slide) and secretory (A30, 152811 slide) endometrium. Art, artery; DC, dendritic cell; fibro, fibroblast; Gland, glandular; ILC, innate lymphoid cell; Lymph, lymphoid; Lumen, lumenal; Mac, macrophage; PV, perivascular; T, T cell; uM, uterine macrophage; uNK, uterine natural killer cell; uSMC, uterine smooth muscle cell.
Fig. 1 Single-cell profiling of the human uterus. a, Schematic illustration of the human uterus showing the different layers and the morphological changes seen throughout the menstrual cycle with respect to tissue sampling. b, UMAP projections of scRNA-seq data from a total of 15 individuals. c, UMAP representations colored by menstrual phase. d, UMAP of subclustered immune populations. e, UMAP projections of snRNA-seq data from a total of four individuals in the proliferative phase. f, Radial representation of the cosine distance similarity for single cells obtained from snRNA-seq to the centroids of cell types defined by scRNA-seq. g, Estimated amount of mRNA (color intensity) contributed by each cell population to each spot (color) shown over the hematoxylin and eosin (H&E) image of the secretory endometrium (A30, 152811 slide). h, Dot plot showing log2-transformed expression of genes expressed in fibroblast and stromal subsets. i, Estimated amount of mRNA (color intensity) contributed by each cell population to each spot (color) shown over the H&E image of proliferative (A30, 152810 slide) and secretory (A30, 152811 slide) endometrium. Art, artery; DC, dendritic cell; fibro, fibroblast; Gland, glandular; ILC, innate lymphoid cell; Lymph, lymphoid; Lumen, lumenal; Mac, macrophage; PV, perivascular; T, T cell; uM, uterine macrophage; uNK, uterine natural killer cell; uSMC, uterine smooth muscle cell.CC BY · PubMed Central
Fig. 2 Temporal and spatial dynamics of endometrial epithelial cells. a, Schematic illustration of epithelial subsets in the differentiated endometrium highlighting the anatomical location of the glandular and lumenal epithelia. b, UMAP of subclustered and subsampled epithelial populations. c, UMAP of subclustered and subsampled epithelial populations colored by their menstrual phase. d, Dot plot showing the log2-transformed expression of genes characteristic of each epithelial subset. e, Number of mRNA molecules per spot (color intensity) confidently assigned to each epithelial subpopulation (color) in the proliferative phase (A13, 152810 slide). f, High-resolution large-area imaging of a section of proliferative endometrium, stained with in situ hybridization (smFISH) for WNT7A and LGR5 (SOX9+LGR5+ epithelial markers). White arrowheads indicate lumenal and glandular regions shown at higher magnification (right). Representative image of four proliferative endometrial samples from four different donors. Scale bars: left, 250 μm; other, 25 μm. g, Number of mRNA molecules per spot (color intensity) confidently assigned to each epithelial subpopulation (color) in the early-proliferative phase (A30, 152807 slide). h, Validation of KRT5, COX1 (marker of lumenal cells) and SCGB2A2 (marker of glandular population) with IHC in endometrial tissue (proliferative and secretory phases). Nuclei are counterstained with hematoxylin. Scale bars, 250 μm. Representative images of three proliferative and three secretory endometrial samples from six different donors.
Fig. 2 Temporal and spatial dynamics of endometrial epithelial cells. a, Schematic illustration of epithelial subsets in the differentiated endometrium highlighting the anatomical location of the glandular and lumenal epithelia. b, UMAP of subclustered and subsampled epithelial populations. c, UMAP of subclustered and subsampled epithelial populations colored by their menstrual phase. d, Dot plot showing the log2-transformed expression of genes characteristic of each epithelial subset. e, Number of mRNA molecules per spot (color intensity) confidently assigned to each epithelial subpopulation (color) in the proliferative phase (A13, 152810 slide). f, High-resolution large-area imaging of a section of proliferative endometrium, stained with in situ hybridization (smFISH) for WNT7A and LGR5 (SOX9+LGR5+ epithelial markers). White arrowheads indicate lumenal and glandular regions shown at higher magnification (right). Representative image of four proliferative endometrial samples from four different donors. Scale bars: left, 250 μm; other, 25 μm. g, Number of mRNA molecules per spot (color intensity) confidently assigned to each epithelial subpopulation (color) in the early-proliferative phase (A30, 152807 slide). h, Validation of KRT5, COX1 (marker of lumenal cells) and SCGB2A2 (marker of glandular population) with IHC in endometrial tissue (proliferative and secretory phases). Nuclei are counterstained with hematoxylin. Scale bars, 250 μm. Representative images of three proliferative and three secretory endometrial samples from six different donors.CC BY · PubMed Central
Fig. 3 Epithelial signatures in endometrial disorders. a, Heatmaps showing the relative contribution of single-cell-derived signals from healthy endometrium (rows) in explaining the bulk transcriptomes of 430 endometrioid and 122 serous endometrial adenocarcinomas from TCGA (columns). b, Analysis of the 313 endometrial adenocarcinoma TCGA samples that exhibit SOX9+LGR5+ or SOX9+LGR5− exposure above the intercept value. A Kruskal–Wallis test was performed on cohorts of samples regrouped in association with cancer stages from I to IV (n = 201, 26, 72 and 14, respectively). The SOX9+LGR5+ exposure depends on the stage of the tumor. Wilcoxon tests show that the increased value associated with later stages is significant for each stage partition (horizontal lines denote binary partitions: n = 201 versus 112, n = 227 versus 86, n = 299 versus 14). Black dots are individual exposure values. Boxplots represent quartiles while whiskers extend up to 1.5 times interquartile range (IQR) beyond each box to encapsulate extrema. c, Boxplots showing normalized expression levels of epithelial marker genes in endometrium and peritoneum from control donors and patients with endometriosis from GSE141549. Expression in red, white or black peritoneal lesions is compared with endometrium and normal peritoneum by two-sided Wilcoxon test (not significant (NS): P > 0.05). Boxplots represent quartiles and whiskers extend up to 1.5 times IQR beyond each box to encapsulate extrema. For the proliferative and secretory comparisons, the number of independent biological samples was, respectively: control endometrium (n = 17 and n = 25), control peritoneum (n = 4 and n = 8), peritoneum red lesions (n = 2 and n = 7), peritoneum white lesions (n = 5 and n = 4) and peritoneum black lesions (n = 6 and n = 5). E, endometrium; E Pat, endometrium (patient); Expr, expression; P, peritoneum; P lesion B, peritoneal lesion black; P lesion R, peritoneal lesion red; P lesion W, peritoneal lesion white; P Pat, nonlesional control peritoneum (patient).
Fig. 3 Epithelial signatures in endometrial disorders. a, Heatmaps showing the relative contribution of single-cell-derived signals from healthy endometrium (rows) in explaining the bulk transcriptomes of 430 endometrioid and 122 serous endometrial adenocarcinomas from TCGA (columns). b, Analysis of the 313 endometrial adenocarcinoma TCGA samples that exhibit SOX9+LGR5+ or SOX9+LGR5− exposure above the intercept value. A Kruskal–Wallis test was performed on cohorts of samples regrouped in association with cancer stages from I to IV (n = 201, 26, 72 and 14, respectively). The SOX9+LGR5+ exposure depends on the stage of the tumor. Wilcoxon tests show that the increased value associated with later stages is significant for each stage partition (horizontal lines denote binary partitions: n = 201 versus 112, n = 227 versus 86, n = 299 versus 14). Black dots are individual exposure values. Boxplots represent quartiles while whiskers extend up to 1.5 times interquartile range (IQR) beyond each box to encapsulate extrema. c, Boxplots showing normalized expression levels of epithelial marker genes in endometrium and peritoneum from control donors and patients with endometriosis from GSE141549. Expression in red, white or black peritoneal lesions is compared with endometrium and normal peritoneum by two-sided Wilcoxon test (not significant (NS): P > 0.05). Boxplots represent quartiles and whiskers extend up to 1.5 times IQR beyond each box to encapsulate extrema. For the proliferative and secretory comparisons, the number of independent biological samples was, respectively: control endometrium (n = 17 and n = 25), control peritoneum (n = 4 and n = 8), peritoneum red lesions (n = 2 and n = 7), peritoneum white lesions (n = 5 and n = 4) and peritoneum black lesions (n = 6 and n = 5). E, endometrium; E Pat, endometrium (patient); Expr, expression; P, peritoneum; P lesion B, peritoneal lesion black; P lesion R, peritoneal lesion red; P lesion W, peritoneal lesion white; P Pat, nonlesional control peritoneum (patient).CC BY · PubMed Central
Fig. 4 Cell signaling in glandular and lumenal epithelium. a, Heatmaps showing TFs differentially expressed in ciliated (top) and secretory (bottom) epithelial lineages. Color is proportional to log-transformed fold change; asterisks highlight TFs whose targets are also differentially expressed (that is, differentially activated TFs). b, Unbiased clustering of epithelial subsets using Visium data. Spot colors represent cluster assignment based on Louvain clustering of spots assigned to epithelial subsets. Spots assigned to one of the clusters (represented in light gray in the figure) were excluded from the analysis due to the low percentage of epithelial cells in the spot after visual inspection. c, Heatmap showing log-transformed fold change of differentially expressed genes between the three main clusters defining the lumenal, functional and basal epithelial regions. d, High-resolution large-area imaging of a representative section of secretory-phase endometrium, with pseudocolor intensity proportional to smFISH signal for NOTCH2. Representative image of three endometrial samples from three different donors. e, smFISH quantification in three full-thickness secretory-phase endometrial samples. Plots represent smFISH spot intensity in glands divided by gland area at increasing distances from the lumen. Approximate lumen range is marked in yellow. Source data
Fig. 4 Cell signaling in glandular and lumenal epithelium. a, Heatmaps showing TFs differentially expressed in ciliated (top) and secretory (bottom) epithelial lineages. Color is proportional to log-transformed fold change; asterisks highlight TFs whose targets are also differentially expressed (that is, differentially activated TFs). b, Unbiased clustering of epithelial subsets using Visium data. Spot colors represent cluster assignment based on Louvain clustering of spots assigned to epithelial subsets. Spots assigned to one of the clusters (represented in light gray in the figure) were excluded from the analysis due to the low percentage of epithelial cells in the spot after visual inspection. c, Heatmap showing log-transformed fold change of differentially expressed genes between the three main clusters defining the lumenal, functional and basal epithelial regions. d, High-resolution large-area imaging of a representative section of secretory-phase endometrium, with pseudocolor intensity proportional to smFISH signal for NOTCH2. Representative image of three endometrial samples from three different donors. e, smFISH quantification in three full-thickness secretory-phase endometrial samples. Plots represent smFISH spot intensity in glands divided by gland area at increasing distances from the lumen. Approximate lumen range is marked in yellow. Source dataCC BY · PubMed Central
Fig. 5 Interrogation by CellPhoneDB v.3.0 of ligands and receptors mediating epithelial differentiation. a, Adaptation of our cell–cell communication tool that considers spatial cellular dynamics and is available at https://github.com/Ventolab/CellphoneDB. b, Schematic illustration of receptors and ligands involved in WNT and NOTCH signaling. c, Dot plots showing expression of CellPhoneDB v.3.0 relevant ligands in epithelial, stromal and fibroblast populations with cognate receptors in epithelial subsets. Only significant interactions (fold change > 0.02 and FDR < 0.005) are represented. The color of the arrows corresponds to the pathways whose ligand–receptor partners are involved, as shown in b. d, Estimated proportions of DKK1 in the early-proliferative phase (A30, 152807 slide). e, Schematic illustration of our proposed model for temporal and spatial distribution of epithelial and stromal subsets across the menstrual cycle. The proliferative phase is dominated by a WNT environment that promotes regeneration. Compartmentalization of WNT and NOTCH signaling during the secretory phase promotes efficient differentiation toward the ciliated and secretory lineages.
Fig. 5 Interrogation by CellPhoneDB v.3.0 of ligands and receptors mediating epithelial differentiation. a, Adaptation of our cell–cell communication tool that considers spatial cellular dynamics and is available at https://github.com/Ventolab/CellphoneDB. b, Schematic illustration of receptors and ligands involved in WNT and NOTCH signaling. c, Dot plots showing expression of CellPhoneDB v.3.0 relevant ligands in epithelial, stromal and fibroblast populations with cognate receptors in epithelial subsets. Only significant interactions (fold change > 0.02 and FDR < 0.005) are represented. The color of the arrows corresponds to the pathways whose ligand–receptor partners are involved, as shown in b. d, Estimated proportions of DKK1 in the early-proliferative phase (A30, 152807 slide). e, Schematic illustration of our proposed model for temporal and spatial distribution of epithelial and stromal subsets across the menstrual cycle. The proliferative phase is dominated by a WNT environment that promotes regeneration. Compartmentalization of WNT and NOTCH signaling during the secretory phase promotes efficient differentiation toward the ciliated and secretory lineages.CC BY · PubMed Central
Fig. 6 In vitro responses of endometrial organoids to ovarian hormones are similar to in vivo epithelial changes. a, Experimental timeline of endometrial organoid cultures. Organoids were derived in ExM and then subjected to hormonal stimulation with estrogen (E2) followed by estrogen + progesterone (P4) + cAMP and prolactin (PRL). The time points at which organoids were collected for scRNA-seq are marked with an asterisk. b, UMAP projections of scRNA-seq data identify major cellular populations. c, UMAP representations colored by days after hormonal stimulation (top) or by treatments (bottom). d, Dot plot showing log2-transformed expression of selected genes that distinguish the main cell populations. e, IHC to validate markers of the secretory population, SCGB2A2 and HEY1, and combined staining for FOXJ1 and acetylated α-tubulin in control (undifferentiated) and differentiated (hormonally stimulated) organoids. Black arrowheads indicate ciliated cells with FOXJ1-positive nuclei. Nuclei are counterstained with hematoxylin. Scale bars: 250 μm (red), 200 μm (black). Representative images of three endometrial organoids from three different donors. f, Predicted epithelial subsets of endometrial organoids using a logistic classifier. g, Heatmaps showing TFs differentially expressed in ciliated and secretory lineages. Color is proportional to log-transformed fold change, with asterisks highlighting TFs whose targets are also differentially expressed (that is, differentially activated TFs). h, Cells able to respond to progesterone derived from a clonal organoid culture (E001 individual) (Methods) are colored from left to right: (1) cluster labels as in Extended Data Fig. 8e; (2) Palantir pseudotime; (3) probability of cells to progress toward the ciliary lineage; and (4) probability that the cell differentiates toward the secretory lineage. NH, no hormone.
Fig. 6 In vitro responses of endometrial organoids to ovarian hormones are similar to in vivo epithelial changes. a, Experimental timeline of endometrial organoid cultures. Organoids were derived in ExM and then subjected to hormonal stimulation with estrogen (E2) followed by estrogen + progesterone (P4) + cAMP and prolactin (PRL). The time points at which organoids were collected for scRNA-seq are marked with an asterisk. b, UMAP projections of scRNA-seq data identify major cellular populations. c, UMAP representations colored by days after hormonal stimulation (top) or by treatments (bottom). d, Dot plot showing log2-transformed expression of selected genes that distinguish the main cell populations. e, IHC to validate markers of the secretory population, SCGB2A2 and HEY1, and combined staining for FOXJ1 and acetylated α-tubulin in control (undifferentiated) and differentiated (hormonally stimulated) organoids. Black arrowheads indicate ciliated cells with FOXJ1-positive nuclei. Nuclei are counterstained with hematoxylin. Scale bars: 250 μm (red), 200 μm (black). Representative images of three endometrial organoids from three different donors. f, Predicted epithelial subsets of endometrial organoids using a logistic classifier. g, Heatmaps showing TFs differentially expressed in ciliated and secretory lineages. Color is proportional to log-transformed fold change, with asterisks highlighting TFs whose targets are also differentially expressed (that is, differentially activated TFs). h, Cells able to respond to progesterone derived from a clonal organoid culture (E001 individual) (Methods) are colored from left to right: (1) cluster labels as in Extended Data Fig. 8e; (2) Palantir pseudotime; (3) probability of cells to progress toward the ciliary lineage; and (4) probability that the cell differentiates toward the secretory lineage. NH, no hormone.CC BY · PubMed Central
Fig. 7 WNT and NOTCH signatures dictate endometrial epithelial differentiation. a, Experimental timeline of endometrial organoid cultures. Organoids were treated with inhibitors to either NOTCH (DBZ or DAPT) or WNT (IWP-2 or XAV939) upon initiation of hormonal stimulation. R-spondin-1 (RSPO-1) was omitted from ExM in the presence of WNT inhibitors. Collection time points for scRNA-seq are highlighted with asterisks. b, UMAP plots for scRNA-seq samples after either WNT or NOTCH inhibition. c, UMAP representations colored by inhibitor treatments (top) or hormonal stimulation (bottom). d, Bar plots showing enrichment of cells in ciliated and secretory clusters after NOTCH or WNT inhibition compared with untreated controls, analyzed with unpaired z-tests. e, IHC for acetylated α-tubulin (ciliary marker) and glycodelin (PAEP). Scale bars, 200 μm. Representative images of endometrial organoids derived from three different patients. Blue arrowheads indicate ciliated cells, orange arrowheads indicate secretory cells and green arrowheads indicate glandular secretions. f, Dot plot showing the log2-transformed expression of genes characteristic of endometrial secretions in epithelial subsets. g, Radial representation of the cell type probabilities predicted by a logistic model trained on epithelial cells in vivo. The linear projection shows cells in each corner whenever a cell is predicted to belong to a given class with a probability of 1. h, Volcano plots representing differentially expressed genes within the secretory lineage in two comparisons: (1) cells cultured with and without WNT inhibitor; progesterone is present in the media; and (2) cells cultured with and without hormones; WNT inhibitor is present in the media. TFs that are significant in the in vivo dataset are highlighted. i, Heatmap showing differential activities of TFs significant in the in vivo analysis. Ctrl, control; NOTCHi, NOTCH inhibitor; WNTi, WNT inhibitor.
Fig. 7 WNT and NOTCH signatures dictate endometrial epithelial differentiation. a, Experimental timeline of endometrial organoid cultures. Organoids were treated with inhibitors to either NOTCH (DBZ or DAPT) or WNT (IWP-2 or XAV939) upon initiation of hormonal stimulation. R-spondin-1 (RSPO-1) was omitted from ExM in the presence of WNT inhibitors. Collection time points for scRNA-seq are highlighted with asterisks. b, UMAP plots for scRNA-seq samples after either WNT or NOTCH inhibition. c, UMAP representations colored by inhibitor treatments (top) or hormonal stimulation (bottom). d, Bar plots showing enrichment of cells in ciliated and secretory clusters after NOTCH or WNT inhibition compared with untreated controls, analyzed with unpaired z-tests. e, IHC for acetylated α-tubulin (ciliary marker) and glycodelin (PAEP). Scale bars, 200 μm. Representative images of endometrial organoids derived from three different patients. Blue arrowheads indicate ciliated cells, orange arrowheads indicate secretory cells and green arrowheads indicate glandular secretions. f, Dot plot showing the log2-transformed expression of genes characteristic of endometrial secretions in epithelial subsets. g, Radial representation of the cell type probabilities predicted by a logistic model trained on epithelial cells in vivo. The linear projection shows cells in each corner whenever a cell is predicted to belong to a given class with a probability of 1. h, Volcano plots representing differentially expressed genes within the secretory lineage in two comparisons: (1) cells cultured with and without WNT inhibitor; progesterone is present in the media; and (2) cells cultured with and without hormones; WNT inhibitor is present in the media. TFs that are significant in the in vivo dataset are highlighted. i, Heatmap showing differential activities of TFs significant in the in vivo analysis. Ctrl, control; NOTCHi, NOTCH inhibitor; WNTi, WNT inhibitor.CC BY · PubMed Central
Extended Data Fig. 1 Quality control of the scRNA-seq datasets. a, Experimental workflow for the generation of cellular profiling of the uterus. In short, single-cell suspensions were obtained following two protocols: (i) collagenase treatment to enrich for the stromal fraction (ii) collagenase followed by trypsin to enrich for the glandular fraction. In addition, tissue blocks were processed for single-nuclei RNA sequencing (snRNA-seq) and Visium experiments. b, Single-cell RNA sequencing (ScRNA-seq) data analysis strategy. In short, quality control was performed at the cell and gene level on the matrices generated by STARsolo. To integrate data from distinct individuals, data was batch corrected by each sample using scVI. After defining cell clusters, those clusters containing a high proportion of low-quality cells and doublets (defined by scrublet) were excluded. Re-clustering was performed on epithelial, endothelial and immune cells. c, UMAP (uniform manifold approximation and projection) of scRNA-seq data from all tissue samples. Clusters corresponding to doublets, low QC cells and epithelial cells from the cervix were further excluded from the analysis. d, Dot plot showing log2-transformed expression of specific markers for the population labelled as ‘cervix’, absent in organ donor samples. Contamination from the cervix is possible due to the biopsy procedure. e, UMAP representations coloured by menstrual stage, biopsy type, menstrual day, tissue type, donor ID and cell cycle phase. f, UMAP of sub-clustered endothelial populations. g, Dot plot showing log2-transformed expression of selected genes that distinguish the main cell populations. h, Dot plot showing log2-transformed expression of selected immune cell markers. uSMC = uterine smooth muscle cell; PV = perivascular; eS = non-decidualised endometrial stromal cells; dS = decidualised endometrial stromal cells; uM = uterine macrophages; uNK = uterine Natural Killer cells, T = T cells, ILC = Innate lymphoid cells, DC = Dendritic cells; scRNA-seq = single-cell RNA sequencing, IHC = Immunohistochemistry.
Extended Data Fig. 1 Quality control of the scRNA-seq datasets. a, Experimental workflow for the generation of cellular profiling of the uterus. In short, single-cell suspensions were obtained following two protocols: (i) collagenase treatment to enrich for the stromal fraction (ii) collagenase followed by trypsin to enrich for the glandular fraction. In addition, tissue blocks were processed for single-nuclei RNA sequencing (snRNA-seq) and Visium experiments. b, Single-cell RNA sequencing (ScRNA-seq) data analysis strategy. In short, quality control was performed at the cell and gene level on the matrices generated by STARsolo. To integrate data from distinct individuals, data was batch corrected by each sample using scVI. After defining cell clusters, those clusters containing a high proportion of low-quality cells and doublets (defined by scrublet) were excluded. Re-clustering was performed on epithelial, endothelial and immune cells. c, UMAP (uniform manifold approximation and projection) of scRNA-seq data from all tissue samples. Clusters corresponding to doublets, low QC cells and epithelial cells from the cervix were further excluded from the analysis. d, Dot plot showing log2-transformed expression of specific markers for the population labelled as ‘cervix’, absent in organ donor samples. Contamination from the cervix is possible due to the biopsy procedure. e, UMAP representations coloured by menstrual stage, biopsy type, menstrual day, tissue type, donor ID and cell cycle phase. f, UMAP of sub-clustered endothelial populations. g, Dot plot showing log2-transformed expression of selected genes that distinguish the main cell populations. h, Dot plot showing log2-transformed expression of selected immune cell markers. uSMC = uterine smooth muscle cell; PV = perivascular; eS = non-decidualised endometrial stromal cells; dS = decidualised endometrial stromal cells; uM = uterine macrophages; uNK = uterine Natural Killer cells, T = T cells, ILC = Innate lymphoid cells, DC = Dendritic cells; scRNA-seq = single-cell RNA sequencing, IHC = Immunohistochemistry.CC BY · PubMed Central
Extended Data Fig. 2 Quality control of the snRNA-seq datasets. a, SnRNA-seq data analysis strategy. Prior to data integration, ambient RNA was removed. b, UMAP (uniform manifold approximation and projection) of snRNA-seq data from all tissue samples. Clusters corresponding to doublets and low QC cells were further excluded from the analysis. c, UMAP representations coloured by sample ID and donor ID. d, Dot plot showing log2-transformed expression of selected genes that distinguish the main cell populations. uSMC = uterine smooth muscle cell; PV = perivascular; eS = non-decidualised endometrial stromal cells; snRNA-seq = single-nuclei RNA sequencing.
Extended Data Fig. 2 Quality control of the snRNA-seq datasets. a, SnRNA-seq data analysis strategy. Prior to data integration, ambient RNA was removed. b, UMAP (uniform manifold approximation and projection) of snRNA-seq data from all tissue samples. Clusters corresponding to doublets and low QC cells were further excluded from the analysis. c, UMAP representations coloured by sample ID and donor ID. d, Dot plot showing log2-transformed expression of selected genes that distinguish the main cell populations. uSMC = uterine smooth muscle cell; PV = perivascular; eS = non-decidualised endometrial stromal cells; snRNA-seq = single-nuclei RNA sequencing.CC BY · PubMed Central
Extended Data Fig. 3 Quality control of the Visium slides. a, Haematoxylin and eosin staining of the slides in the Visium arrays. Two individuals were selected: proliferative phase A13 and secretory phase A30. Two sections 100 μm apart were analysed. Lumenal epithelium was well preserved in individual A30 and in a small region of A13. b, Scatter plots show the number of genes over the number of counts, where each dot is a feature of the Visium slide. Plots are coloured by the percentage of mitochondrial genes. c, Bar plots showing number of genes on each of the samples. A bimodal distribution corresponding to endometrium and myometrium was shown on each of the cases. d, Visualisation of the number of genes on the Visium slides of sample A30, 152807 slide. A zone with low quality is highlighted in the image. This is probably caused by a technical artifact. No pattern like this was seen in other samples. e, Unbiased clustering of Visium spots defined by Louvain algorithm. f, Estimated amount of mRNA (colour intensity) contributed by each cell population to each spot (colour) shown over the H&E image of the proliferative (A13, 152806 slide) and secretory (A30, 152807 slide) endometrium. Endo = Endometrium; Myo = Myometrium; Epi = Epithelial; uSMC= uterine smooth muscle cell; PV = perivascular.
Extended Data Fig. 3 Quality control of the Visium slides. a, Haematoxylin and eosin staining of the slides in the Visium arrays. Two individuals were selected: proliferative phase A13 and secretory phase A30. Two sections 100 μm apart were analysed. Lumenal epithelium was well preserved in individual A30 and in a small region of A13. b, Scatter plots show the number of genes over the number of counts, where each dot is a feature of the Visium slide. Plots are coloured by the percentage of mitochondrial genes. c, Bar plots showing number of genes on each of the samples. A bimodal distribution corresponding to endometrium and myometrium was shown on each of the cases. d, Visualisation of the number of genes on the Visium slides of sample A30, 152807 slide. A zone with low quality is highlighted in the image. This is probably caused by a technical artifact. No pattern like this was seen in other samples. e, Unbiased clustering of Visium spots defined by Louvain algorithm. f, Estimated amount of mRNA (colour intensity) contributed by each cell population to each spot (colour) shown over the H&E image of the proliferative (A13, 152806 slide) and secretory (A30, 152807 slide) endometrium. Endo = Endometrium; Myo = Myometrium; Epi = Epithelial; uSMC= uterine smooth muscle cell; PV = perivascular.CC BY · PubMed Central
Extended Data Fig. 4 Spatio-temporal regulation of epithelial cells. a, UMAP (uniform manifold approximation and projection) of scRNA-seq data from epithelial cells. We performed a donor-balanced subsampling (1000 cells maximum for each donor). Clusters corresponding to doublets and low QC cells were further excluded from the analysis. b, UMAP representations coloured assigned by cell phase, donor, menstrual stage and day of the menstrual cycle. c, Number of mRNA molecules per spot (colour intensity) confidently assigned to each epithelial subpopulation (colour) in the proliferative phase (A13, 152806 slide). d, Number of mRNA molecules per spot (colour intensity) confidently assigned to each epithelial subpopulation (colour) in the secretory phase (A30, 152807 slide). e, Estimated proportion of mRNA coming from epithelial subsets in the early-proliferative phase (A30, 152807 slide).
Extended Data Fig. 4 Spatio-temporal regulation of epithelial cells. a, UMAP (uniform manifold approximation and projection) of scRNA-seq data from epithelial cells. We performed a donor-balanced subsampling (1000 cells maximum for each donor). Clusters corresponding to doublets and low QC cells were further excluded from the analysis. b, UMAP representations coloured assigned by cell phase, donor, menstrual stage and day of the menstrual cycle. c, Number of mRNA molecules per spot (colour intensity) confidently assigned to each epithelial subpopulation (colour) in the proliferative phase (A13, 152806 slide). d, Number of mRNA molecules per spot (colour intensity) confidently assigned to each epithelial subpopulation (colour) in the secretory phase (A30, 152807 slide). e, Estimated proportion of mRNA coming from epithelial subsets in the early-proliferative phase (A30, 152807 slide).CC BY · PubMed Central
Extended Data Fig. 5 Spatially-resolved single-cell transcriptomic expression of proliferative markers by smFISH. a, High-resolution large-area imaging of uterine tissue sections from three donors in the proliferative phase stained with smFISH for MKI67, combined with protein staining for EPCAM. White arrowheads indicate magnified regions demonstrating representative MKI67 expression levels across lumenal, functional, and basal epithelia. Scale bars, top = 1 mm, middle = 100 μm, bottom = 20 μm. b, Molecular integrity of uterine tissues was validated by multiplexed smFISH staining of sections for constitutively expressed genes of different typical expression levels (UBC - high; PPIB - moderate; POLR2A - low), which demonstrated strong signals irrespective of sample or region. White arrowheads indicate epithelial regions shown at higher magnification (bottom). Scale bars, top = 1 mm; below = 20 μm. Two representative donors each from the proliferative and secretory phases are shown. c, High-resolution imaging of uterine tissue sections stained with smFISH for EPCAM and LGR5, combined with protein staining of FOXJ1. White arrowheads indicate cells with dual FOXJ1 and LGR5 staining, shown magnified to the eight. Scale bars, left = 50 μm, other = 10 μm. d, High-resolution large-area imaging of four endometrial sections stained with PAEP. Scale bars = 1 mm.
Extended Data Fig. 5 Spatially-resolved single-cell transcriptomic expression of proliferative markers by smFISH. a, High-resolution large-area imaging of uterine tissue sections from three donors in the proliferative phase stained with smFISH for MKI67, combined with protein staining for EPCAM. White arrowheads indicate magnified regions demonstrating representative MKI67 expression levels across lumenal, functional, and basal epithelia. Scale bars, top = 1 mm, middle = 100 μm, bottom = 20 μm. b, Molecular integrity of uterine tissues was validated by multiplexed smFISH staining of sections for constitutively expressed genes of different typical expression levels (UBC - high; PPIB - moderate; POLR2A - low), which demonstrated strong signals irrespective of sample or region. White arrowheads indicate epithelial regions shown at higher magnification (bottom). Scale bars, top = 1 mm; below = 20 μm. Two representative donors each from the proliferative and secretory phases are shown. c, High-resolution imaging of uterine tissue sections stained with smFISH for EPCAM and LGR5, combined with protein staining of FOXJ1. White arrowheads indicate cells with dual FOXJ1 and LGR5 staining, shown magnified to the eight. Scale bars, left = 50 μm, other = 10 μm. d, High-resolution large-area imaging of four endometrial sections stained with PAEP. Scale bars = 1 mm.CC BY · PubMed Central
Extended Data Fig. 6 Expression of epithelial markers in endometrial disorders. a, Volcano plot showing upregulation of markers specific for SOX9 + LGR5 + in endometrial tumours with a SOX9 + LGR5 + signature in comparison with those with a SOX9 + LGR5- signature. Negative log fold changes similarly show the opposite pattern where markers are characteristic of the SOX9 + LGR5- population instead. b, Boxplots showing expression levels of epithelial marker genes in endometrium and peritoneum from patients with endometriosis from GSE141549. Expression in peritoneal lesions is compared to endometrium and peritoneum with two-sided Wilcox test (ns; not significant: p > 0.05). Box plots represent quartiles and whiskers extend up to 1.5 IQR beyond each box to encapsulate extrema. For the proliferative and secretory comparisons, the number of independent biological samples was, respectively: control endometrium (n = 17 and n = 25), control peritoneum (n = 4 and n = 8), peritoneal red lesions (n = 2 and n = 7), peritoneal white lesions (n = 5 and n = 4) and peritoneal black lesions (n = 6 and n = 5). E Pat = endometrium (patient); P Pat = non-lesional control Peritoneum (Patient); P lesion R = Peritoneal Lesion Red; P lesion W = Peritoneal Lesion White; P lesion B = Peritoneal Lesion Black.
Extended Data Fig. 6 Expression of epithelial markers in endometrial disorders. a, Volcano plot showing upregulation of markers specific for SOX9 + LGR5 + in endometrial tumours with a SOX9 + LGR5 + signature in comparison with those with a SOX9 + LGR5- signature. Negative log fold changes similarly show the opposite pattern where markers are characteristic of the SOX9 + LGR5- population instead. b, Boxplots showing expression levels of epithelial marker genes in endometrium and peritoneum from patients with endometriosis from GSE141549. Expression in peritoneal lesions is compared to endometrium and peritoneum with two-sided Wilcox test (ns; not significant: p > 0.05). Box plots represent quartiles and whiskers extend up to 1.5 IQR beyond each box to encapsulate extrema. For the proliferative and secretory comparisons, the number of independent biological samples was, respectively: control endometrium (n = 17 and n = 25), control peritoneum (n = 4 and n = 8), peritoneal red lesions (n = 2 and n = 7), peritoneal white lesions (n = 5 and n = 4) and peritoneal black lesions (n = 6 and n = 5). E Pat = endometrium (patient); P Pat = non-lesional control Peritoneum (Patient); P lesion R = Peritoneal Lesion Red; P lesion W = Peritoneal Lesion White; P lesion B = Peritoneal Lesion Black.CC BY · PubMed Central
Extended Data Fig. 7 Spatially-resolved single-cell transcriptomic expression of WNT and NOTCH signals. a, High-resolution large-area imaging of uterine tissue sections stained with smFISH for WNT7A, WNT5A, and LGR5. White arrowheads indicate magnified regions demonstrating spatial segregation of WNT7A (lumenal epithelial) and WNT5A (stromal) expression. Scale bars, top = 200 μm, other = 50 μm. b, High-resolution imaging of endometrial tissue sections stained with smFISH for JAG1. A comparison of representative regions of lumenal and glandular epithelium is shown for three secretory phase donors. Scale bars = 10 μm. c, Co-staining of JAG1 and HEY1. Top, solid white arrowheads indicate regions of lumenal epithelium magnified below. Below, cells showing juxtacrine expression of JAG1 and HEY1 (magenta arrowheads = JAG1highHEY1low, green arrowheads = JAG1lowHEY1high) or co-expression (white outlined arrowheads). Scale bars, top = 50 μm, other = 5 μm. d, Dot plot showing log2-transformed expression of AXIN2 expression.
Extended Data Fig. 7 Spatially-resolved single-cell transcriptomic expression of WNT and NOTCH signals. a, High-resolution large-area imaging of uterine tissue sections stained with smFISH for WNT7A, WNT5A, and LGR5. White arrowheads indicate magnified regions demonstrating spatial segregation of WNT7A (lumenal epithelial) and WNT5A (stromal) expression. Scale bars, top = 200 μm, other = 50 μm. b, High-resolution imaging of endometrial tissue sections stained with smFISH for JAG1. A comparison of representative regions of lumenal and glandular epithelium is shown for three secretory phase donors. Scale bars = 10 μm. c, Co-staining of JAG1 and HEY1. Top, solid white arrowheads indicate regions of lumenal epithelium magnified below. Below, cells showing juxtacrine expression of JAG1 and HEY1 (magenta arrowheads = JAG1highHEY1low, green arrowheads = JAG1lowHEY1high) or co-expression (white outlined arrowheads). Scale bars, top = 50 μm, other = 5 μm. d, Dot plot showing log2-transformed expression of AXIN2 expression.CC BY · PubMed Central
Extended Data Fig. 8 Quality control of organoid scRNA-seq dataset. a, UMAP (uniform manifold approximation and projection) of scRNA-seq data from all organoid samples. b, UMAP representations coloured by donor ID. The three donors correspond to E001, B044 and B080. c, Logistic regression probabilities. d, Experimental timeline for endometrial organoid cultures. Clonal organoids were derived in Expansion Medium (ExM) with CHIR99021 (CHIR) and ROCK inhibitor Y-22763 (Ri), grown in ExM and then subjected to hormonal stimulation with estrogen (E2) followed by E2 + progesterone (P4) + cyclic AMP (cAMP) and prolactin (PRL). The time points at which organoids were collected for scRNA-seq are shown with an asterisk. Representative bright field images of organoids for some of the timepoints are shown. e, UMAP projections of scRNA-seq data from two clonal organoids derived from the same individual. f, UMAP representations coloured by hormonal stimulation, cell cycle phase, individual clone, days after hormonal stimulation, sample ID. g, Dot plot showing log2-transformed expression of selected genes that distinguish the main cell populations. NH = No hormone; E2 = Estrogen; P4 = Progesterone; d = days.
Extended Data Fig. 8 Quality control of organoid scRNA-seq dataset. a, UMAP (uniform manifold approximation and projection) of scRNA-seq data from all organoid samples. b, UMAP representations coloured by donor ID. The three donors correspond to E001, B044 and B080. c, Logistic regression probabilities. d, Experimental timeline for endometrial organoid cultures. Clonal organoids were derived in Expansion Medium (ExM) with CHIR99021 (CHIR) and ROCK inhibitor Y-22763 (Ri), grown in ExM and then subjected to hormonal stimulation with estrogen (E2) followed by E2 + progesterone (P4) + cyclic AMP (cAMP) and prolactin (PRL). The time points at which organoids were collected for scRNA-seq are shown with an asterisk. Representative bright field images of organoids for some of the timepoints are shown. e, UMAP projections of scRNA-seq data from two clonal organoids derived from the same individual. f, UMAP representations coloured by hormonal stimulation, cell cycle phase, individual clone, days after hormonal stimulation, sample ID. g, Dot plot showing log2-transformed expression of selected genes that distinguish the main cell populations. NH = No hormone; E2 = Estrogen; P4 = Progesterone; d = days.CC BY · PubMed Central
Extended Data Fig. 9 NOTCH and WNT inhibition. a, Representative brightfield images of the organoids treated with DBZ, DAPT, IWP-2 or XAV939 at the end of the experiment (day 6). Scale bars, 500 μm, 200 μm, 100 μm from left to right. Black arrowheads point at folded organoids. Representative images of three endometrial organoids from three different donors. b, UMAP (uniform manifold approximation and projection) projections of scRNA-seq data from all organoid samples. c, UMAP projections of scRNA-seq data coloured by days after hormonal stimulation, individual, hormonal stimulation, inhibitor used, phase of the cell cycle and sample ID. d, quantitative real-time PCR (qRT-PCR) analysis for genes expressed in ciliated or secretory cells of hormonally-stimulated organoids treated with NOTCH inhibitors (blue) or WNT inhibitors (pink). Bar plots showing the mean with SD levels of expression relative to housekeeping genes and control conditions at day 0 (without inhibitors, black) (n = 3 different donors). Deviations to the control conditions (without inhibitors) were detected as statistically significant by Dunnett’s multiple comparisons with p-value <0.05, which defines two-sided confidence intervals for individual deviations. e, ELISA assay for glycodelin (PAEP) from supernatants of hormonally stimulated organoids treated with NOTCH (blue) or WNT inhibitors (pink). Bar plots showing the mean with SD levels of expression (n = 3 donors). Deviations to the control condition (no inhibitors, no hormones) at day 0 were detected as statistically significant with Dunnett’s multiple comparisons (p-value <0.05). Source data
Extended Data Fig. 9 NOTCH and WNT inhibition. a, Representative brightfield images of the organoids treated with DBZ, DAPT, IWP-2 or XAV939 at the end of the experiment (day 6). Scale bars, 500 μm, 200 μm, 100 μm from left to right. Black arrowheads point at folded organoids. Representative images of three endometrial organoids from three different donors. b, UMAP (uniform manifold approximation and projection) projections of scRNA-seq data from all organoid samples. c, UMAP projections of scRNA-seq data coloured by days after hormonal stimulation, individual, hormonal stimulation, inhibitor used, phase of the cell cycle and sample ID. d, quantitative real-time PCR (qRT-PCR) analysis for genes expressed in ciliated or secretory cells of hormonally-stimulated organoids treated with NOTCH inhibitors (blue) or WNT inhibitors (pink). Bar plots showing the mean with SD levels of expression relative to housekeeping genes and control conditions at day 0 (without inhibitors, black) (n = 3 different donors). Deviations to the control conditions (without inhibitors) were detected as statistically significant by Dunnett’s multiple comparisons with p-value <0.05, which defines two-sided confidence intervals for individual deviations. e, ELISA assay for glycodelin (PAEP) from supernatants of hormonally stimulated organoids treated with NOTCH (blue) or WNT inhibitors (pink). Bar plots showing the mean with SD levels of expression (n = 3 donors). Deviations to the control condition (no inhibitors, no hormones) at day 0 were detected as statistically significant with Dunnett’s multiple comparisons (p-value <0.05). Source dataCC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Cited by

See all 662

Metadata

Authors
Garcia-Alonso, L., Handfield, L.-F., Roberts, K., Nikolakopoulou, K., Fernando, R. C., Gardner, L., Woodhams, B., Arutyunyan, A., Polanski, K., Hoo, R., Sancho-Serra, C., Li, T., Kwakwa, K., Tuck, E., Lorenzi, V., Massalha, H., Prete, M., Kleshchevnikov, V., Tarkowska, A., Porter, T., Mazzeo, C. I., van Dongen, S., Dabrowska, M., Vaskivskyi, V., Mahbubani, K. T., Park, J.-E., Jimenez-Linan, M., Campos, L., Kiselev, V. Y., Lindskog, C., Ayuk, P., Prigmore, E., Stratton, M. R., Saeb-Parsy, K., Moffett, A., Moore, L., Bayraktar, O. A., Teichmann, S. A., Turco, M. Y., & Vento-Tormo, R.
Affiliations
  • Wellcome Sanger Institute (GB)
  • University of Cambridge (GB)
  • European Bioinformatics Institute (GB)
  • Uppsala University (SE)
  • Cambridge University Hospitals NHS Foundation Trust (GB)
Venue
Nature Genetics
Published
2021-12
Volume
53
Issue
12
Pages
1698-1711
ISSN
  • 1061-4036 (ISSN-L)
  • 1546-1718
OpenAlex
W3214755680
Cited by
662
References
70
Citations by year
2026: 102 · 2025: 151 · 2024: 148 · 2023: 141 · 2022: 105 · 2021: 15
FWCI
32.65
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Genetics (publishedVersion, cc-by)
  • bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Uppsala University Publications (Uppsala University) (submittedVersion, other-oa)
  • Apollo (University of Cambridge) (submittedVersion, cc-by)
  • Apollo
  • Apollo (cc-by)
Funders
  • Silicon Valley Community Foundation
  • GlaxoSmithKline
  • Biogen
  • Wellcome Trust (211276/Z/18/Z, WT206194, 211276, Wellcome) (211276/Z/18/Z, WT206194, 200841/Z/16/Z, 211276/Z/18/Z, WT206194, 200841, 200841/Z/16/Z)
  • UK Research and Innovation (MR/S036350/1)
  • Chan Zuckerberg Initiative
  • Royal Society (DH160216, RG93116, DH160216)
  • Pathological Society of Great Britain and Ireland
  • European Commission (874867, HORIZON2020)
  • Jean Shanks Foundation (No 1175)
  • Genentech
  • Horizon 2020 Framework Programme
  • Medical Research Council (MR/S036350/1)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Single-cell and spatial transcriptomics
  • Life Sciences › Immunology and Microbiology › Immunology › Reproductive System and Pregnancy
  • Health Sciences › Medicine › Obstetrics and Gynecology › Endometrial and Cervical Cancer Treatments
Keywords
human endometrium, menstrual cycle, endometrial organoids, epithelial cell fate, Wnt signaling, Notch signaling, endometrial cancer, endometriosis, endometriotic lesions, cell type deconvolution
MeSH terms
Cell Differentiation, Endometrium, Female, Humans, Menstrual Cycle, Organoids, Gonadal Steroid Hormones, Uterus, Signal Transduction, Endometrial Neoplasms, Cell Lineage, Tissue Culture Techniques, Wnt Proteins, Receptors, Notch, Transcriptome, Cellular Microenvironment, Spatio-Temporal Analysis, In Vitro Techniques

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.