Systematic Analysis of the Gene Expression Profile of Tumour Endothelial Cells

An Introduction to a Bioinformatics Pipeline

Thomas Mohr

HCC: The Challenge

HCC is a major health problem:

  • Over 500.000 people affected
  • Third leading cause of cancer deaths worldwide
  • Incidence is expected to rise.
  • Challenges
    • High recurrence rate after surgery
    • Bad prognosis
    • Limited treatment options due to pre-existing liver damage

Why focus on endothelium?

Endothelial cells play a major role in tumour progression:

  • Enhancing motility circuits
  • Enabling proliferation circuits
    • Production of growth factors
    • Production of hormones
  • Enhancing viability circuits
    • Production of survival factors
    • Production of cytokines

Systematic profiling of tumour endothelium is key to discovering new therapies.

The problem

Obtaining samples requires

  • Before and during surgery

    • Patient recruitment

    • Sample preparation

  • After surgery

    • Isolation of endothelial cells by cell sorting

    • mRNA Isolation

    • Hybridization to arrays

  • Scanning of the results

These steps require considerable effort with regard to material and personnel.

We choose something different …

Obtaining sufficient numbers of normal and tumour tissue samples from patients is challenging.

To overcome sample challenges, we leveraged public gene expression databases, focusing on high-quality, comparable datasets:

  • Expression data from tumour and normal endothelial cells,

  • No culturing step between the cell and RNA isolation, and

  • Had a sufficient number of samples to enable in-depth characterization.

We selected dataset GSE51401, with 43 samples from 13 subjects

We leveraged public gene expression databases.

Our workflow combined differential expression, pathway, and network analyses to identify key targets:

  1. Differential gene expression with using a paired samples linear model and two factors (origin and activation)
  2. Term Enrichment Analysis,
  3. Gene Set Variation Analysis,
  4. Weighted Gene Coexpression Analysis,
  5. Mapping of central genes to drugs.

To begin the analysis, we address the question: which genes are differentially expressed?

The Differential Gene Expression Analysis (DGEA, Panel A & B) shows:

  • Distinct gene expression changes found in quiescent and activated TECs.

  • Key downregulated genes confirmed; upregulated genes mostly undetectable in single-cell RNASeq.

  • Many TEC gene changes overlap with NEC, suggesting similar angiogenic activation.

We next ask: which pathways are differentially perturbed?

Gene Set Variation Analysis (Panel C & D) shows:

  • Identification of substantial differences in up- and downregulated pathways between quiescent and activated TECs.

  • Upregulated pathways in both groups were mainly linked to cancer hallmarks: deregulated energetics, evasion of growth suppressors, and genome instability.

  • Most other cancer hallmark pathways, including immune evasion and angiogenesis, were consistently downregulated.

  • Specific gene changes highlighted cell cycle activation, DNA biosynthesis, and altered chemokine signaling, suggesting tumour endothelial anergy.

  • Results indicate pro-angiogenic TEC profiles may rely more on evasion of growth suppressors than sustained proliferation or angiogenesis.

Network analysis - investigating groups of genes associated with origin and activation

  • Low-expression genes were filtered out, yielding a co-expression network of 14,690 genes.

  • Network analysis identified 39 gene modules ranging from 33 to 433 genes each.

  • Network analysis revealed six gene modules linked to tumour origin and angiogenic activation.

  • These modules offer insight into tumour-specific gene regulation and angiogenic state.

  • All selected modules showed high internal stability (median membership > 0.8).

  • These modules provide insight into gene regulation linked to tumour origin and angiogenic state.

Next, we turned to modules associated with origin and activation

  • In module M1, both ENG− and ENG+ TECs showed upregulated eigengenes compared to NEC, with no difference between ENG+ and ENG− cells of the same origin.

  • Module M16 showed minor but significant upregulation of eigengenes between ENG− and ENG+ NEC, while ENG+ TECs had further, but not significantly higher, expression.

  • In M15, ENG+ NEC had higher eigengene expression than ENG− NEC, but both TEC groups were downregulated compared to their respective NECs.

  • Module M18 eigengenes were upregulated in ENG+ TEC and NEC compared to ENG− NEC, but cell origin had only a minor effect.

  • Modules M14 and M8 highlighted that cell origin and angiogenic activation influence eigengene expression, with distinct patterns of up- and downregulation observed.

M1 shows a strong association with tumour endothelium

  • M1 module is tied to cell proliferation and contains many druggable hub genes.

  • It contains 29 hub genes, including several druggable targets (e.g., BIRC5, UBE2T, NEK2, CDKN3, TTK, CCNB1, TOP2A, FEN1, AURKA, RNASEH2A, POLE2).

  • STRING database analysis revealed a dense functional interaction network among the hub genes.

  • Strongest interactions involve BIRC5, TOP2A, and NEK2, which connect to other important nodes like CDKN3, UBE2T, and CCNB1.

  • Many of these hub genes are potential drug targets already proposed for other cancer types.

BIRC4 is a key potential target

  • BIRC5 (survivin) shows high expression in tumour endothelial cells (TECs) in hepatocellular carcinoma (HCC), while remaining low in most normal tissues.

  • It plays a key role in preventing apoptosis and promoting cell proliferation, supporting tumour growth in HCC.

  • Elevated BIRC5 levels are linked to poor prognosis and therapy resistance in various cancers, including HCC.

  • Several therapeutic strategies are under development to specifically inhibit BIRC5 activity in cancer cells.

Conclusion: BIRC5 is a novel, actionable target in HCC tumour endothelium.

Do you want to have a tailored analysis of your data? Contact us today.