Single-cell RNA sequencing reveals intratumoral heterogeneity in primary uveal melanomas and identifies HES6 as a driver of the metastatic disease

Abstract

Intratumor heterogeneity has been recognized in numerous cancers as a major source of metastatic dissemination. In uveal melanomas, the existence and identity of specific subpopulations, their biological function and their contribution to metastasis remain unknown. Here, in multiscale analyses using single-cell RNA sequencing of six different primary uveal melanomas, we uncover an intratumoral heterogeneity at the genomic and transcriptomic level. We identify distinct transcriptional cell states and diverse tumor-associated populations in a subset of the samples. We also decipher a gene regulatory network underlying an invasive and poor prognosis state driven in part by the transcription factor HES6. HES6 heterogenous expression has been validated by RNAscope assays within primary human uveal melanomas, which further unveils the existence of these cells conveying a dismal prognosis in tumors diagnosed with a favorable outcome using bulk analyses. Depletion of HES6 impairs proliferation, migration and metastatic dissemination in vitro and in vivo using the chick chorioallantoic membrane assay, demonstrating the essential role of HES6 in uveal melanomas. Thus, single-cell analysis offers an unprecedented view of primary uveal melanoma heterogeneity, identifies bona fide biomarkers for metastatic cells in the primary tumor, and reveals targetable modules driving growth and metastasis formation. Significantly, our findings demonstrate that HES6 is a valid target to stop uveal melanoma progression.

Fig. 1. Single-cell RNA-seq uncovers poor prognosis cell subpopulations.

A Experimental workflow indicating the dissociation and isolation of individual cells from primary uveal melanomas for generating single-cell RNA-seq profiles. cDNA from the individual cells were synthesized, followed by library construction and massively parallel sequencing using the 10x genomic approach. B Heatmaps of the first two principal components from the principal component analysis (PCA) based on highly variable genes in the dataset. Both cells and genes are sorted by their PC scores allowing easy exploration of the principal sources of heterogeneity in the dataset. The first ten genes with the highest or lowest absolute PC values are shown for PC1 and PC2 (yellow). C Kaplan–Meier survival plot of the ten genes with the highest PC1 values (lower left quadrant of B). D Kaplan–Meier survival plot of ten genes with the lowest PC1 values (upper right quadrant of B). E Histograms showing PC1 score, based on the top ten up and down PC1 genes, of all the single cells in each tumor. The Youden index was used as the cutoff point (red line). The percentage of cells with high and low PC1 score as well as the percentage of “poor prognosis” cells per tumors are indicated below the histogram.

Fig. 2. Single-cell RNA-seq uncovers intratumoral heterogeneity.

A Seurat analysis showing t-SNE plots of 7890 uveal melanoma cells colored by clusters. Each point represents a single cell. Red circles indicate non-malignant cells. B Kaplan–Meier survival plot for the top 20 genes of the indicated clusters. C Heatmap of inferred copy number variation (CNV) signal normalized against the topmost cluster composed by the pool of all putative non-malignant cells (no CNV variation). Cells (rows, n = 7890 cells) are ordered from non-malignant cells (NMC, n = 101 cells) to cancer cells (n = 7789 cells), from the six uveal melanomas. Chromosomal regions (columns) with amplifications (red) or deletions (blue) are shown. The additional tracks, on the right, show the associated cluster number from Seurat.

Fig. 3. scRNA-Seq identifies multiple co-existing transcriptional states and a network driven by HES6 associated with a poor prognosis.

A Heatmap of cells and regulon binary scores with hierarchical clustering inferred by SCENIC. The additional track, above, show supervised clustering by patients and the PC1 score based on the top ten genes with the highest values in PC1. The 72 regulons with the best clustering out of the 122 identified in the six primary uveal melanomas are represented. B t-SNE shows cells colored by patient (SCENIC approach). C HES6 regulon activity comprising 72 genes quantified using AUCell is represented. Regulons of predicted transcription factors in the six patients were determined to be active if they exceeded the threshold (Blue), otherwise, regulons were considered to be inactive (gray). D Kaplan–Meier survival plot of the HES6 regulon (TCGA dataset). E Kaplan–Meier survival plot of HES6 (TCGA dataset).

Fig. 4. HES6 expression controls the motile ability of primary uveal melanomas.

A Sections from tumors # A–F were labeled with RNAscope probe for HES6 (red), and images were captured by confocal microscopy. Cell nuclei (blue green). Shown are the areas of high and low heterogeneity. Scale bars represent 60 μm. For each tumor, magnification of the boxed area is shown (Right). Scale bars represent 25 μm. B Western blot and RT-QPCR analysis of HES6 in primary MP46 melanoma cells transduced with a control or HES6 adenovirus expressing HES6 (HES6 OE) for 72 h. C Human primary MP46 melanoma cells were transduced with empty (control) or HES6 expressing adenovirus (HES6 OE) for 48 h before being seeded in the upper part of the Boyden chamber. Migration was examined after 24 h. Values represent means +SD of three independent experiments. **p < 0.01. D Representative images are shown. Bar = 100 μm. E Western blot and RT-QPCR analysis of HES6 in primary Mel270 melanoma cells transfected with a control siRNA (siCtl) or two different pools of multiple siRNA targeting HES6 (siHES6#1 and siHES6#2). F Migration of Mel270 cells transfected with the indicated siRNA. ***p < 0.001. G Representative images are shown. Bar = 100 μm.

Fig. 5. HES6 signaling is a key driver of aggressive and motile phenotypes.

A QPCR analysis of HES6 in primary 92.1 melanoma cells expressing doxycycline-inducible control or HES6 shRNA in presence of 1 μg/ml doxycycline for 96 h in presence or absence of DLL4 1 μg/ml. B Migration of 92.1 melanoma cells expressing doxycycline-inducible control or HES6 shRNA in presence of 1 μg/ml doxycycline for 48 h in presence or absence of DLL4 1 μg/ml. ***p < 0.001. C Representative images are shown. Bar = 100 μm. D Description of the chicken embryo CAM assay. E 92.1 uveal melanoma cells expressing doxycycline-inducible control or HES6 shRNA were grafted on the CAM of 9-day-old (E9) chick embryos. The tumors were collected and weighted on day 18 (E18). Values represent means + SEM. F Genomic DNA is extracted from the lower CAM to evaluate the number of metastatic cells on day 18 and analyzed by qPCR with specific primers for human Alu sequences. Values represent means  ± SEM.