Single-cell gene regulatory network analysis reveals new melanoma cell states and transition trajectories during phenotype switching
Preprint posted on August 08, 2019 https://www.biorxiv.org/content/10.1101/715995v2
Melanoma is the deadliest form of skin cancer. There is extreme genetic and cellular heterogeneity within each tumour as well as between patients. This, coupled with the ability of melanoma cells to switch between proliferative and invasive/drug resistant phenotypes means that identifying common cell states to target therapeutically is difficult. Despite this, two cell states have been found previously that recur across tumours and patients. One is the melanocyte-like state – marked by SOX10 and associated gene expression that controls the developmental melanocyte lineage program, and the other is the mesenchymal-like state – marked by SOX9 expression and associated with invasiveness and resistance to therapy.
In this preprint, Wouters, Kalender-Atak and colleagues aimed to investigate these states at the transcriptional level by performing single-cell RNA seq on nine diverse patient-derived melanoma cell lines, in order to identify gene regulatory networks (GRNs) common to different melanoma cell states.
First, the team performed scRNA-seq on nine patient-derived 2D melanoma cultures (MM lines), as well as the A375 melanoma cell line. Each sample clustered together after dimensionality reduction, however there were also gene signatures in common across MM lines, chiefly the SOX10-high melanocyte-like state, and the SOX9-high mesenchymal-like state. Interestingly, some melanocyte-like MM lines also expressed mesenchymal- and immune-associated genes, representing an intermediate state. This can be visualised in the tSNE plot below:
t-SNE plot of sc-RNA seq data generated using the online tool made by the authors (http://scope.aertslab.org/#/Wouters_Human_Melanoma). Each cell line forms its own cluster, however there are gene signatures in common. The SOX9 mesenchymal signature is restricted to three cell lines (blue). The other lines express SOX10, marking a melanotic phenotype (red). A subset of melanotic cell lines also express mesenchymal markers, as well as genes such as NFAT2C (green) which define a novel intermediate phenotype.
To investigate whether these transcriptional states translate to distinct tumour phenotypes, the authors conducted a single-cell migratory assay, whereby cells were seeded on a microfluidic chip containing channels for individual cells. Indeed, SOX9-expressing mesenchymal-like cells migrated fastest and the furthest along channels, whereas SOX10-expressing melanocyte-like cells migrated more slowly, with intermediate MM cells migrating somewhere between the two.
This intermediate cell state is not well understood in the literature, so the authors further investigated its transcriptional properties to identify an ‘intermediate cell state’ signature. SCENIC network inference was performed to predict which transcription factors and their targets – known as a regulon – governing each cell state. The intermediate cell states shared several regulons with both melanocyte-like and mesenchymal-like states, however some regulons including EGR3, RXRG and NFATC2 were specific to this state, suggesting this intermediate-state is a distinct cell identity. These identities were also found by Omni-ATAC-seq and in the TCGA database.
The authors next investigated a hypothesis that SOX10 disruption in would cause a transcriptional shift in melanocyte-like cells towards a more mesenchymal-like state. Indeed, bulk RNA-seq of intermediate-like cell lines at 24 hour intervals after siRNA-mediated knock-down of SOX10 showed progressive downregulation of melanocyte lineage markers, and upregulation of mesenchymal-like transcriptional signatures. By repeating SCENIC network inference along the trajectories pseudotime, the team could then order the key transcriptional events occurring after SOX10 knockdown. First, the cell cycle was paused due to inactivation of DNA polymerases and E2F/MYB transcription factors. Second, the melanocyte transcriptional program was inactivated by downregulation or pausing of melanocyte-associated transcriptional machinery. Lastly, genes associated with EMT were activated before induction of immune-related transcription factors.
Why I chose this preprint
Trying to identify clusters, common factors and tease patterns out of noisy-by-nature datasets can be quite a headache, so it is very interesting to see characterisation of common melanoma transcriptional signatures in what initially looks like very diverse samples. The authors have also made their scRNA analyses available at http://scope.aertslab.org/#/Wouters_Human_Melanoma, and it’s been fun to look at how my own genes of interest are expressed in the different cell lines. It will be interesting to see how targeting the intermediate cell state markers could change melanoma behaviour in a therapeutic context in terms of invasiveness and drug resistance.
Questions for the authors
Do you know if SOX9 knockdown or SOX10 overexpression in mesenchymal-like states would cause reprogramming towards a melanocyte-like state, and could this increase its drug sensitivity? Or are mesenchymal-like cells, or indeed melanotic cells towards the extremes of melanocyte-state more refractory to phenotype switching than intermediate-state cells?
Do you know if the established MM cell lines you used retain the degree of cellular and transcriptional heterogeneity that would be found if you were to obtain tumour cells directly from patients and compare these? Might the signatures you find be more enhanced by long term culturing in the same conditions?
How do you envision how the identification of these different states could help stratify treatments for patients, who might have different tumours at various places along the SOX10-melanotic/SOX9-mesenchymal spectrum?
Posted on: 13th September 2019Read preprint
Also in the bioinformatics category:
Regulatory context drives conservation of glycine riboswitch aptamers
|Selected by||Defne Surujon|
Variant antigen diversity in Trypanosoma vivax is not driven by recombination
|Selected by||Mariana De Niz|
GeneWalk identifies relevant gene functions for a biological context using network representation learning
|Selected by||Ramona Jühlen|
Also in the cancer biology category:
Subcellular mRNA localization regulates ribosome biogenesis in migrating cells
|Selected by||Nidhi Kanwal|
Colibactin DNA damage signature indicates causative role in colorectal cancer
|Selected by||Connor Rosen|
Mechanotransduction of strain regulates an invasive phenotype in newly transformed epithelial cells
|Selected by||William Hill|
preListsbioinformatics category:in the
Antimicrobials: Discovery, clinical use, and development of resistance
Preprints that describe the discovery of new antimicrobials and any improvements made regarding their clinical use. Includes preprints that detail the factors affecting antimicrobial selection and the development of antimicrobial resistance.
|List by||Zhang-He Goh|