Further classify tumor cells into tumor cell states for Ewing sarcoma samples (SCPCP000015)

[allyhawkins]

Existing analysis

This analysis will be an update to the existing cell-type-ewings module.

Details of modification

After annotating which cells are tumor cells in all samples in SCPCP000015, we will want to further classify the tumor cells into distinct cell states, if applicable. In particular, we would like to annotate cells as belonging to a specific sub class of tumor cells, such as EWS-FLI1 high/low, mesenchymal like cells, proliferative tumor cells, etc.

The main goal is to label any tumor cells as belonging to a tumor cell state that has already been published and shown to be present in Ewing sarcoma tumors in the literature. This would provide further granularity to the existing tumor cell annotations in the module.

Describe your goals

As part of the cell typing module for Ewing sarcoma, we are also clustering cells. The goal is to take any clusters that contain tumor cells and label those with any cell states that we feel confident are represented.

The first thing we need to do is obtain a list of expected cell states and associated gene sets or gene signatures with those cell states. For Ewing sarcoma, cells lie on a continuum based on their EWS-FLI1 expression. The level of EWS-FLI1 expression then alters the cell phenotype, so we can search for enrichment of specific phenotypes to classify cells into EWS-FLI1 high, middle/hybrid, and low. In my preliminary search, we should expect to see the following cell states present:

• EWS-FLI1 high: High expression/repression of EWS-FLI1 targets, High HIF1alpha activity, low EMT-like phenotype (low migration, adhesion), low ECM expression, low CD73 expression
• EWS-FLI1 low: Expression of EWS-FLI1 repressed genes, mesenchymal-like, high EMT-like phenotype (high migration, adhesion), increased ECM expression, high CD73 expression, high HOXD13 activity
• EWS-FLI1 middle/hybrid: Proliferative

After creating a list of expected cell types and expected gene signatures, we could try the following first:

• Look at expression of gene signatures for specific cell states across tumor cell clusters. We would expect clusters with higher expression of the specified gene signatures to be in that cell state. This approach only works if we see strong heterogeneity within a sample and we see clusters with high and low expression of the gene signatures.
• Alternatively, we could identify marker genes for each of the tumor cell clusters compared to all other tumor cells and see if any of the specific gene programs are enriched.
• To identify proliferative cells specifically, we could look for enrichment of a proliferative gene signature as well as cell cycle stage.

Some other ideas that I have that are not fully baked and would be a bit more complicated include:

• Using normal cell references to help annotate cell states in tumor cells. In particular, Ewing sarcoma cells are hypothesized to originate from either mesenchymal stem cells (MSC) or neural crest cells. There is also literature showing that Ewing cells show varying levels of an MSC-like signature. It could be helpful to use a reference atlas that contains MSC cells and the MSC lineage and see where tumor cells match up. Some useful references that we may be able to use for this include:

  • MSCscDB
  • Single-cell transcriptome atlas of human mesenchymal stem cells exploring cellular heterogeneity
  • Fetal organ atlas
  • A human embryonic limb cell atlas resolved in space and time
  • Single‐cell study of neural stem cells derived from human iPSCs reveals distinct progenitor populations with neurogenic and gliogenic potential

• Perform NMF using all tumor cells in all libraries and look for recurrent gene expression programs. From here we could look for enrichment of specific gene signatures in those programs to assign labels. This might be a nice complementary approach to looking for specific marker gene expression to see if we come up with any phenotypes that are consistently present in both methods.

Scientific literature

In addition to the papers linked above, these may be helpful in defining existing cell states:

• Single cell RNA-sequencing of Ewing sarcoma tumors demonstrates transcriptional heterogeneity and clonal evolution
• Cancer-Associated Fibroblast-Like Tumor Cells Remodel the Ewing Sarcoma Tumor Microenvironment
• EWS::FLI1 and HOXD13 Control Tumor Cell Plasticity in Ewing Sarcoma
• Transcriptional programs define intratumoral heterogeneity of Ewing sarcoma at Single-cell resolution

Other details

At this time I do not think we will need any additional resources than what is currently in the module.
This will be completed after clustering of all libraries in SCPCP000015 is completed.