A Panel of Synapse-Related Genes as a Biomarker for Gliomas

Xiangwen Ji, Hongwei Zhang, Qinghua CuiFrontiers in Neuroscience 2020 · DOI

TL;DR

Researchers have identified a panel of 17 synapse-related genes that can be used to create a "synapse score." This score effectively predicts glioma prognosis, diagnosis, and grade, outperforming current biomarkers like IDH mutation and 1p/19q codeletion. This offers a promising new tool for monitoring and understanding gliomas.

What Problem Does This Paper Address?

Gliomas, the most common primary brain cancers, pose a significant challenge due to their diverse prognoses and the limitations of existing biomarkers. While IDH mutation and 1p/19q codeletion are valuable, they only apply to a subset of patients, leaving a critical need for more comprehensive and effective biomarkers. Recent research has highlighted the intriguing phenomenon of "neuro-glioma synapses," where neurons and glioma cells communicate electrochemically, and even "pseudo-synapses" formed by metastatic tumor cells. These interactions have been implicated in tumor progression and metastasis, suggesting that genes involved in synaptic function could hold diagnostic and prognostic power. This paper aims to explore this hypothesis by developing and validating a synapse-based biomarker for gliomas.

Methods

The study employed a multi-step approach to identify and validate the synapse-related gene panel:

  1. Gene Ontology (GO) Term Curation: They began by selecting GO terms related to synapse, neuron, neurotransmitter transport, glutamate receptor, and cell junction.
  2. Single Sample Gene Set Enrichment Analysis (ssGSEA): This technique was used to calculate an "enrichment score" (ES) for each GO term in individual samples, allowing for the quantification of gene set activity per sample without requiring group information.
  3. Cox Regression Analysis: This statistical method was used to evaluate the survival prediction performance of the GO terms and individual genes by calculating hazard ratios (HRs) and p-values.
  4. Gene Set Trimming: To refine the gene set, they iteratively assessed the prognostic performance of gene sets of increasing size, selecting the optimal set based on combined false discovery rates (FDRs).
  5. Permutation Experiment: To rigorously validate the selected gene set, they performed a permutation experiment, randomly selecting 17-gene sets from the entire genome and comparing their prognostic ability to the curated synapse-related set.
  6. Synapse Score Calculation: The ESs from ssGSEA for the final 17-gene panel were defined as the "synapse score" for each sample.
  7. Statistical Analysis: Kaplan-Meier curves, Cox proportional hazards regression, Wilcoxon rank sum tests, and Receiver Operating Characteristic (ROC) curve analyses were used to assess survival, grading, and diagnostic abilities.

Datasets

The study utilized publicly available gene expression datasets from three major sources:

  • The Cancer Genome Atlas (TCGA): Including both lower-grade glioma (LGG) and glioblastoma multiforme (GBM) data.
  • The Chinese Glioma Genome Atlas (CGGA): Comprising both microarray and RNAseq data.
  • The Gene Expression Omnibus (GEO): Including several glioma microarray datasets (GSE4290, GSE16011, GSE50161, GSE52009, GSE54004, GSE61374, and GSE107850).

These datasets provided a substantial number of samples (over 3,000 in total) for robust validation.

Key Insights and Ideas

  • Synaptic communication as a potential driver: The study leverages the emerging understanding of how neurons and glioma cells interact via synapses, proposing that genes involved in these interactions could be key to understanding glioma biology.
  • "Synapse Score" as a novel metric: Moving beyond individual gene expression, the development of a quantitative "synapse score" provides a holistic measure of synaptic activity within a tumor sample.
  • Unexpected prognostic association: Counterintuitively, higher synapse scores were associated with better overall survival, contrary to the initial hypothesis that synaptic interactions might promote tumor progression. This highlights the complexity of biological processes and the need for empirical validation.
  • Broader diagnostic potential: The synapse score not only predicts prognosis but also shows promise in distinguishing between glioma and normal brain tissue, suggesting a potential diagnostic application.

Key Results

  • A panel of 17 synapse-related genes was identified as a strong predictor of glioma outcomes.
  • The synapse score demonstrated excellent predictive ability for prognosis, diagnosis, and grading across multiple independent datasets (TCGA, CGGA, and GEO).
  • Patients with higher synapse scores exhibited significantly longer overall survival.
  • Higher WHO grades of gliomas were associated with significantly lower synapse scores.
  • The synapse score showed independent and superior predictive performance compared to established biomarkers like IDH mutation and 1p/19q codeletion. For instance, in multivariate Cox regression, the synapse score remained a significant independent predictor of survival.
  • ROC analyses indicated high diagnostic accuracy, with AUCs reaching up to 0.99 in some datasets for distinguishing glioma from normal brain tissue.

Limitations and Future Work

  • Mechanism remains unclear: While the study establishes the predictive power of the synapse score, the underlying biological mechanisms driving the association between higher synapse scores and better prognosis are not fully elucidated. The authors suggest that this might be analogous to how IDH mutation and 1p/19q codeletion, while promoting progression, are also associated with better prognosis due to factors like chemoradiotherapy sensitivity.
  • Heuristic algorithm for gene set selection: The method used to find the optimal 17-gene set was heuristic, meaning it may not guarantee the absolute best possible combination. The authors acknowledge that other gene sets and machine learning methods might yield even better results.
  • Potential for batch effects: While efforts were made to account for batch effects, the authors note that some terms showed different directions in different batches, suggesting that batch effects and sample differences could influence results.
  • Future work: The authors suggest further exploration of the mechanisms behind the observed associations, the development of machine learning models using synapse-related genes, and validation in larger, more diverse cohorts. They also point out that the identified synapse-related genes could be useful in predicting prognosis, grading, and diagnosis in both brain tumors and normal brain tissues.

Summary updated

Abstract

Gliomas are the most common primary brain cancers. In recent years, <i>IDH</i> mutation and 1p/19q codeletion have been suggested as biomarkers for the diagnosis, treatment, and prognosis of gliomas. However, these biomarkers are only effective for a part of glioma patients, and thus more biomarkers are still emergently needed. Recently, an electrochemical communication between normal neurons and glioma cells by neuro-glioma synapse has been reported. Moreover, it was discovered that breast-to-brain metastasis tumor cells have pseudo synapses with neurons, and these synapses were indicated to promote tumor progression and metastasis. Based on the above observations, we first curated a panel of 17 synapse-related genes and then proposed a metric, synapse score to quantify the "stemness" for each sample of 12 glioma gene expression datasets from TCGA, CGGA, and GEO. Strikingly, synapse score showed excellent predictive ability for the prognosis, diagnosis, and grading of gliomas. Moreover, being compared with the two established biomarkers, <i>IDH</i> mutation and 1p/19q codeletion, synapse score demonstrated independent and better predictive performance. In conclusion, this study proposed a quantitative method, synapse score, as an efficient biomarker for monitoring gliomas.
BibTeX
@inproceedings{ji2020panel,
  author = {Xiangwen Ji and Hongwei Zhang and Qinghua Cui},
  title = {{A Panel of Synapse-Related Genes as a Biomarker for Gliomas}},
  booktitle = {Frontiers in Neuroscience},
  year = {2020},
  doi = {10.3389/fnins.2020.00822},
}
Figure