Stromal protein CCN family contributes to the poor prognosis in lower-grade gioma by modulating immunity, matrix, stemness, and metabolism

Zhihui Liu, Jiasheng Wu, Hang Ji, Hongtao Zhao, Fang Wang, Jiawei Dong, Jiheng Zhang, Nan Wang, Xiuwei Yan, Kaikai Wang, Shaoshan HuFrontiers in Molecular Biosciences 2022 · DOI

TL;DR

Researchers have developed a "CCNScore" based on the expression of stromal proteins (CCN1 and CCN4) that accurately predicts the prognosis of lower-grade glioma (LGG). A high score identifies a subset of tumors that mimic the aggressive behavior of glioblastoma, characterized by high inflammation, metabolic shifts, and a stiffened extracellular matrix.

What Problem Does This Paper Address?

Lower-grade gliomas (LGGs) are a common form of brain cancer that, despite being "lower grade," often progress to more lethal forms or recur after treatment. While immunotherapy has revolutionized other cancers, it has largely met its "Waterloo" in gliomas.

The authors suggest the secret to this resistance lies in the tumor microenvironment (TME)—the complex neighborhood of cells and structural proteins surrounding the tumor. Specifically, they focused on the CCN family of stromal proteins, which are known to regulate the extracellular matrix (ECM). While we know these proteins are important in biological signaling, their specific role in driving the malignancy and immune-evasion of LGG has remained a "black box."

Methods

The study employed a robust bioinformatics pipeline to bridge the gap between gene expression and clinical outcomes:

  • Risk Modeling: Using RNA-seq data, the team applied LASSO (Least Absolute Shrinkage and Selection Operator) regression to identify which CCN family members were the most significant predictors of survival. This resulted in a simplified "CCNScore."
  • Multi-omics Integration: They didn't just look at RNA; they integrated data on DNA methylation (epigenetics), copy number variations (genomics), and single-cell sequencing to see how these proteins interact with immune cells and the brain's physical structure.
  • Functional Analysis: Techniques like ssGSEA (single-sample Gene Set Enrichment Analysis) were used to map the CCNScore to specific biological "hallmarks," such as metabolism and stemness.

Datasets

The authors utilized several high-profile public repositories to ensure their findings were reproducible across different populations:

  • TCGA (The Cancer Genome Atlas): Used as the primary training and discovery cohort.
  • CGGA (Chinese Glioma Genome Atlas): Used as an external validation set to confirm the score's accuracy.
  • GTEx: Provided data on normal brain tissue for baseline comparison.
  • PanCanAtlas & MEXPRESS: Used to explore deeper molecular underpinnings like methylation and mutation patterns.
  • TISCH: A single-cell database used to pinpoint exactly which cells (e.g., macrophages vs. tumor cells) were expressing these proteins.

Key Insights and Ideas

  • The "GBM-like" Transition: The most striking finding is that LGGs with a high CCNScore look genetically and behaviorally like Glioblastoma (GBM)—the most aggressive form of brain cancer. They frequently harbor mutations in EGFR, PTEN, and NF1, which are classic GBM markers.
  • The Matrix as a Shield: High CCNScore tumors showed significantly higher levels of extracellular matrix remodeling. A "stiffer" matrix isn't just a structural change; it promotes hypoxia (low oxygen) and protects the tumor from both the immune system and chemotherapy.
  • Stemness and Metabolism: High CCNScore correlates with higher "stemness" (the ability of cancer cells to self-renew) and a major shift in metabolism. These tumors down-regulate energy metabolism while up-regulating the synthesis of building blocks like nucleotides, essentially "retooling" the cell for rapid growth.
  • Immune "Hot" but Exhausted: Unlike the low-score group, which was "immune quiet," high-score tumors were teeming with immune cells. However, these were mostly immunosuppressive cells (like M2 macrophages and MDSCs) that help the tumor hide from the body's defenses.

Key Results

  • The Formula: The final risk model was distilled into a simple equation: CCNScore = (0.06 * CCN1) + (0.86 * CCN4). CCN4 emerged as the dominant independent risk factor.
  • Prognostic Power: The CCNScore consistently achieved an AUC (Area Under the Curve) > 0.65, reaching as high as 0.857 for predicting 1-year and 3-year survival. This outperformed many traditional clinical traits.
  • Subtype Differentiation: The score successfully distinguished between IDH-mutant (generally better prognosis) and IDH-wildtype (worse prognosis) tumors, providing a molecular shortcut for clinicians to assess risk.

Limitations and Future Work

While the study is computationally exhaustive, the authors acknowledge a few hurdles:

  1. Correlation vs. Causation: The study is primarily based on retrospective data analysis. While the correlations are strong, "wet lab" experiments (in vitro and in vivo) are needed to prove that blocking CCN4 actually slows tumor growth.
  2. Functional Class Scoring: The authors used the FCS approach to quantify pathways, which treats genes independently and may miss complex interactions between protein networks.
  3. Therapeutic Targeting: While the paper identifies CCN proteins as promising targets, developing drugs that can effectively cross the blood-brain barrier to inhibit these specific stromal proteins remains a significant engineering challenge for the future.

Summary updated

Abstract

<b>Background:</b> The CCN family of stromal proteins is involved in the regulation of many important biological functions. However, the role of dysregulated CCN proteins in lower-grade glioma (LGG) remain less understand. <b>Methods:</b> The clinical significance of the CCN proteins was explored based on RNA-seq profiles from multiple cohorts. A CCNScore was constructed using LASSO regression analysis. The PanCanAtlas data and MEXPRESS database were employed to elucidate molecular underpinnings. <b>Results:</b> The expression of CCN4 was associated with poor prognosis in LGG. The CCNScore (CCN1 = 0.06, CCN4 = 0.86) showed implication in prognosis prediction, subtype assessment and therapy selection. The gene mutation pattern of the high-CCNScore group was similar with glioblastoma, including EGFR, PTEN, and NF1 mutation frequently. Besides, the high-CCNScore group was comprised of samples mainly classic-like and mesenchymal-like, had lower methylation levels, higher stemness, higher inflammation, higher levels of extracellular matrix remodel and dysfunction of metabolic pathways. On the other hand, the low-CCNScore group consisted mainly of IDH-mutation LGG, and was characterized by TP53, CIC, and ATRX gene mutations, hyper-methylation status, lower stemness, lower proliferation, immune quietness and low extracellular matrix stiffness. <b>Conclusion:</b> In summary, these results outlined the role of CCN family in LGG and provided a potential and promising therapeutic target.
BibTeX
@inproceedings{liu2022stromal,
  author = {Zhihui Liu and Jiasheng Wu and Hang Ji and Hongtao Zhao and Fang Wang and Jiawei Dong and Jiheng Zhang and Nan Wang and Xiuwei Yan and Kaikai Wang and Shaoshan Hu},
  title = {{Stromal protein CCN family contributes to the poor prognosis in lower-grade gioma by modulating immunity, matrix, stemness, and metabolism}},
  booktitle = {Frontiers in Molecular Biosciences},
  year = {2022},
  doi = {10.3389/fmolb.2022.1027236},
}
Figure