. 2019 Oct;9(10):1406-1421.
Epub 2019 Jul 25.
Epigenomics and Single-Cell Sequencing Define a Developmental Hierarchy in Langerhans Cell Histiocytosis
Free PMC article
Item in Clipboard
Epigenomics and Single-Cell Sequencing Define a Developmental Hierarchy in Langerhans Cell Histiocytosis
2019 Oct .
Free PMC article
Langerhans cell histiocytosis (LCH) is a rare neoplasm predominantly affecting children. It occupies a hybrid position between cancers and inflammatory diseases, which makes it an attractive model for studying cancer development. To explore the molecular mechanisms underlying the pathophysiology of LCH and its characteristic clinical heterogeneity, we investigated the transcriptomic and epigenomic diversity in primary LCH lesions. Using single-cell RNA sequencing, we identified multiple recurrent types of LCH cells within these biopsies, including putative LCH progenitor cells and several subsets of differentiated LCH cells. We confirmed the presence of proliferative LCH cells in all analyzed biopsies using IHC, and we defined an epigenomic and gene-regulatory basis of the different LCH-cell subsets by chromatin-accessibility profiling. In summary, our single-cell analysis of LCH uncovered an unexpected degree of cellular, transcriptomic, and epigenomic heterogeneity among LCH cells, indicative of complex developmental hierarchies in LCH lesions. SIGNIFICANCE: This study sketches a molecular portrait of LCH lesions by combining single-cell transcriptomics with epigenome profiling. We uncovered extensive cellular heterogeneity, explained in part by an intrinsic developmental hierarchy of LCH cells. Our findings provide new insights and hypotheses for advancing LCH research and a starting point for personalizing therapy.
See related commentary by Gruber et al., p. 1343. This article is highlighted in the In This Issue feature, p. 1325.
©2019 American Association for Cancer Research.
Conflict of interest statement
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Figure 1. Single-cell transcriptome analysis captures cellular and molecular diversity of LCH lesions
A) Clinical presentation and incidence of Langerhans cell histiocytosis (LCH). Shown are an x-ray image of LCH bone lesions in the skull (top right), a photography of extensive LCH skin lesions (bottom right), and a hematoxylin and eosin staining of an LCH biopsy (bottom left). B) Cellular heterogeneity in an LCH biopsy (lymph node), as revealed by immunostaining for CD1A, CD207, and an antibody that specifically detects mutated BRAF
V600E protein. C) Overview the of patient biopsy samples analyzed by single-cell RNA-sequencing in this study. D) Low-dimensional projection (t-SNE plot) of the combined single-cell RNA-seq dataset across all analyzed biopsies, comprising a total of 19,044 single-cell transcriptome profiles. E) Molecular heterogeneity in LCH illustrated by low-dimensional projection (as in panel D) of all single-cell transcriptome profiles overlaid with the expression levels of selected marker genes (dark blue color indicates high expression levels). F) Cellular heterogeneity in LCH illustrated by low-dimensional projection (as in panel D), annotated with cell types inferred from marker gene expression.
Figure 2. Characteristic gene expression patterns distinguish LCH cells from other immune cells present in LCH lesions
A) Scatterplots comparing gene expression (mean normalized UMI counts plotted on logarithmic scale) between LCH cells and non-LCH immune cell populations. Differentially expressed genes specific for LCH cells are colored in red, whereas genes specific for the immune cell populations are colored in violet (monocyte/macrophages), blue (T cells), green (B cells), and orange (plasmacytoid dendritic cells). The top three enriched categories from the ARCHS4 tissue database based on Enrichr are shown below. B) Heatmap showing the expression of the genes in the LCH gene signature (genes overexpressed in LCH cells compared to at least three of the four immune cell populations in panel
A) in the LCH and immune cells analyzed (rows). Cell type and patient are indicated by the colored bars on the right. C) Enrichment analysis with Enrichr for the LCH signature genes using three databases (top to bottom: ARCHS4 Tissues, Gene Ontology, or Jensen Diseases), ordered by the Enrichr combined score. FDR-adjusted p-value: *, p < 0.1; ***, p < 0.001. Further details are provided in Supplementary Table S3. D) Gene set enrichment analysis (GSEA) for LCH cell gene expression compared to immune cells using MSigDB hallmark signatures, including a GSEA plot for the most significant gene set (MYC_TARGETS_V2, top) and all significant enrichments visualized in the bar plot (bottom). FDR-adjusted p-value: *, p < 0.1; **, p < 0.05; ***, p < 0.001
Figure 3. LCH lesions comprise proliferating LCH progenitors and multiple differentiated LCH cell subsets
A) Low-dimensional projection (t-SNE plot) of LCH cell transcriptomes, excluding non-LCH immune cells. LCH cells were clustered into 14 subsets. Dashed lines highlight the putative progenitor cell subsets (LCH-C1, LCH-C2) and the most differentiated cell subsets (LCH-C11 to LCH-C14) based on the analysis of transcriptome entropy (panel
C). B) Percentage of cells assigned to each of the 14 LCH cell subset (from panel A) for each analyzed patient biopsy sample. C) LCH cell subsets ordered by the median single-cell entropy of the corresponding single-cell transcriptomes. High entropy indicates promiscuity of gene expression, which has been described as a hallmark as undifferentiated cells. Dashed lines highlight the putative progenitor cell subsets (LCH-C1, LCH-C2) and the most differentiated cell subsets (LCH-C11 to LCH-C14) as in panel A. D) Heatmap showing expression levels of differentially expressed genes between the two LCH subsets with highest entropy and the four subsets with lowest entropy. Selected hallmark genes are highlighted for each subset. E) Enrichment analysis with Enrichr for the LCH subset signature genes (from panel D) using KEGG pathways and Gene Ontology, ordered by the Enrichr combined score. FDR-adjusted p-value: *, p < 0.1; **, p < 0.05; ***, p < 0.001. Further details are provided in Supplementary Table S3. F) Smoothed line plots displaying the relation between decreasing entropy (x-axis; ranked) and the expression of selected genes. The cell cycle genes TUBB, TUBA1B, and MKI67 showed the highest Pearson correlation ( r) between entropy and expression. A weaker correlation was found also between entropy and the expression of the canonical LCH marker gene CD1A. G) Expression of genes characteristic of (non-LCH) epidermal Langerhans cells in each LCH cell subset, displaying the mean of the scaled UMI counts of all genes in the LCH subset signature. H) Assessment of BRAF V600E mutation burden (in percent) based on allele-specific qPCR in sorted LCH-S12 cells, LCH-S1 cells, CD1A/CD207-negative (non-LCH) cells, CD1A/CD207-double-positive LCH cells, and bulk biopsy cells. All data refer to patient sample LCH_E. I) Assessment of cell clonality using the HUMARA assay in the same sorted cell populations as in panel H, and in known polyclonal and monoclonal cell lines as negative and positive controls, respectively.
Figure 4. The cell surface markers HMMR and CLEC9A identify specific LCH cell subsets
A) Low-dimensional projection (t-SNE plot) of LCH cell transcriptomes (same layout as in Fig. 3A) overlaid with the intensity of LCH-subset-specific gene expression signatures (calculated as the mean scaled UMI count for all genes in the signature). Dark red color indicates high expression levels. B) Low-dimensional projection (t-SNE plot) of the LCH cell transcriptomes (as in panel
A) overlaid with the expression of selected LCH-subset-specific marker genes (dark blue color indicates high expression levels). Expression of HMMR is specific to LCH-S1 cells, MKI67 expression overlaps strongly with both the LCH-S1 and LCH-S2 cell subsets, and CLEC9A is detected almost exclusively in LCH-S12 cells. C) Immunohistochemistry image of HMMR (brown) staining of an LCH biopsy. D) Immunofluorescence image of HMMR (green), CD1A (red) and DAPI (blue) staining of an LCH biopsy. E) Immunofluorescence image of CD1A (green), CLEC9A (red) and DAPI (blue) staining of an LCH biopsy.
Figure 5. Characteristic patterns of chromatin accessibility distinguish between LCH cell subsets
A) Scatterplots contrasting differential gene expression with differential chromatin accessibility between progenitor-like LCH-S1 cells and the more differentiated LCH subsets LCH-S12 (top) and LCH-S11 (bottom). The x-axis denotes differences in gene expression (mean normalized UMI count based on scRNA-seq) and the y-axis displays the mean difference in chromatin accessibility over all regulatory regions linked to the respective gene (mean of two replicates, mean normalized reads per million [RPM]). Colored points denote LCH-subset-specific marker genes (from Fig. 3D). B) Heatmap showing ATAC-seq signal intensity for 1,964 differentially accessible regions identified in pair-wise comparisons of one LCH cell subset against the two other subsets. Regions are grouped into six modules based on differential chromatin accessibility. ATAC-seq signal intensity scores are scaled by column (ATAC-seq peaks) for better visualization of differences. C) Enrichment analysis with Enrichr for genes linked to each of the six modules of LCH-subset-specific chromatin accessible regions, showing the top-3 most enriched terms ordered by the Enrichr combined score. FDR-adjusted p-value: *, p < 0.1; **, p < 0.05. Further details are provided in Supplementary Table S3. D) Heatmap of transcription factor binding enrichment for LCH-subset-specific modules, based on LOLA analysis using a large collection of ChIP-seq peak profiles. Top: LOLA enrichments colored by the log
2 odds ratio of enriched overlap between the region modules and ChIP-seq peaks for the corresponding transcription factor, compared to the background of all regulatory regions in the ATAC-seq dataset. Bottom: Mean expression (normalized UMI count) of the gene encoding each transcription factor in different LCH subsets (showing gene expression in LCH_E, which matches the ATAC-seq data, as well as the mean expression across all patient samples). Further details are provided in Supplementary Table S3. E) Enrichment of transcription factor binding motifs for LCH-subset-specific modules, based on HOCOMOCO motif occurrences identified using FIMO. Top: Motif enrichments colored by the log 2 odds ratio of enriched overlap between the module regions and DNA motif hits for the corresponding transcription factor, compared to the background of all regulatory regions identified in the ATAC-seq dataset. Bottom: Expression levels of the genes encoding the corresponding transcription factors (as in panel D). Right: Consensus DNA motif for selected transcription factors. FDR-adjusted p-value, Fisher’s exact test: *, p < 0.1; **, p < 0.05, ***, p < 0.005.
Figure 6. Characteristic gene regulatory networks underlie the observed LCH developmental hierarchy
A) Gene regulatory networks inferred for three LCH cell subsets (LCH-S1, LCH-S12, LCH-S11), based on single-cell transcriptome and ATAC-seq data, and the key regulators identified by the enrichment analysis in Fig. 5D and E. Nodes in the network correspond to the enriched transcription factors as well as their putative target genes (based on sequence proximity and chromatin 3D structure). Node size is proportional to both gene expression level and node out-degree (i.e., number of outgoing connections from the transcription factor) in the respective LCH subset. Edge colors indicate the module of the corresponding peak, and edge visibility is proportional to chromatin accessibility. The network layout was automatically generated using the
igraph package. A browser-based version for interactive data exploration is available in Supplementary File S1 and on the Supplementary Website: http://LCH-hierarchy.computational-epigenetics.org). B) Bar plots showing the top-10 transcription factors ranked by node importance (calculated as a combination of gene expression level and node out-degree) in the networks in panel A. C) Bar plots showing the top-10 transcription factors ranked by differential node importance (node importance in one network relative to the mean across all three networks) in the networks in panel A.
Figure 7. Speculative model of the LCH developmental hierarchy and underlying regulatory mechanisms
Schematic representation of the developmental hierarchy in LCH lesions based on the combined analysis of single-cell transcriptomes (Fig. 2 and 3) and epigenome data from prospectively sorted LCH cell subsets (Fig. 5 and 6).
All figures (7)
Single-Cell RNA Sequencing Reveals a Developmental Hierarchy in Langerhans Cell Histiocytosis.
Cancer Discov. 2019 Oct;9(10):1343-1345. doi: 10.1158/2159-8290.CD-19-0820.
Cancer Discov. 2019.
New insights into the molecular pathogenesis of langerhans cell histiocytosis.
Oncologist. 2014 Feb;19(2):151-63. doi: 10.1634/theoncologist.2013-0341. Epub 2014 Jan 16.
24436311 Free PMC article.
Langerhans cell histiocytosis in adults: Advances in pathophysiology and treatment.
Cancer Sci. 2018 Dec;109(12):3707-3713. doi: 10.1111/cas.13817. Epub 2018 Oct 30.
Cancer Sci. 2018.
30281871 Free PMC article.
Activated Conventional T-Cells Are Present in Langerhans Cell Histiocytosis Lesions Despite the Presence of Immune Suppressive Cytokines.
J Interferon Cytokine Res. 2015 Oct;35(10):831-9. doi: 10.1089/jir.2014.0190. Epub 2015 Sep 18.
J Interferon Cytokine Res. 2015.
New insights into the pathogenesis of Langerhans cell histiocytosis.
Curr Opin Pediatr. 1997 Feb;9(1):46-50. doi: 10.1097/00008480-199702000-00011.
Curr Opin Pediatr. 1997.
Research Support, Non-U.S. Gov't