Screening and Comprehensive Analysis of Cancer-Associated tRNA-Derived Fragments

Yiran Zhou, Qinghua Cui, Yuan ZhouFrontiers in Genetics 2022 · DOI

TL;DR

Researchers have systematically identified 1,516 cancer-associated tRNA-derived fragments (ca-tRFs) across seven major cancer types. By integrating clinical data with small RNA expression profiles, the study reveals that these fragments are not mere degradation byproducts but active regulators of oncogenic pathways and immune infiltration that can define distinct patient survival subtypes.

What Problem Does This Paper Address?

For decades, the fragments resulting from the cleavage of transfer RNAs (tRNAs) were dismissed as molecular "noise" or cellular debris. However, recent evidence suggests these tRNA-derived fragments (tRFs) function similarly to microRNAs, silencing genes and influencing cellular behavior.

The challenge is that most research has focused on isolated tRFs in specific cancers. We lacked a comprehensive, "pan-cancer" view of which tRFs actually drive disease progression, correlate with clinical stages, or predict patient survival. This paper seeks to bridge that gap by creating a systematic pipeline to identify high-priority, cancer-associated tRFs (ca-tRFs) and map their functional landscape across multiple malignancies.

Methods

The authors developed a rigorous computational pipeline to filter the vast sea of small RNAs:

  1. The "Two Out of Three" Strategy: To be classified as a "cancer-associated tRF" (ca-tRF), a fragment had to meet at least two of three criteria:
  2. Significant differential expression between tumor and normal tissue.
  3. Correlation with clinical tumor stage.
  4. Correlation with overall patient survival.
  5. Target Prediction (tRFTar): Using their previously developed tool, the authors predicted gene targets by simulating base-pairing energy between tRFs and mRNAs within Argonaute (AGO) protein complexes.
  6. Network Analysis: They used a Random Walk with Restart (RWR) algorithm on protein-protein interaction networks to identify "gene modules" targeted by these tRFs.
  7. Immune Deconvolution: Using the EPIC tool, they estimated the proportion of immune cells (like T-cells) in bulk tumor samples to see if tRF expression linked to the tumor microenvironment.
  8. Consensus Clustering: They applied unsupervised machine learning to group patients into novel subtypes based solely on their tRF expression profiles.

Datasets

The study relied on several high-quality data repositories:

  • TCGA (The Cancer Genome Atlas): Provided clinical data and gene expression for seven cancer types (BRCA, HNSC, KIRC, KIRP, LIHC, LUSC, and THCA).
  • MINTbase v2.0: A specialized database containing tRF expression profiles derived from TCGA small RNA-sequencing data.
  • tRFTar: The authors' own database for AGO-mediated tRF-target interactions.
  • PICKLE 3.0: A database of cross-validated protein-protein interactions used for the network analysis.

Key Insights and Ideas

  • Non-Random Origins: ca-tRFs aren't produced randomly. They primarily originate from specific "hotspots" on parental tRNAs, particularly those carrying Valine (TAC) and Arginine (CCT).
  • The "Super-Biomarker": One specific fragment, tRF-28-RS9NS334L2DB, emerged as a standout. It was significantly upregulated in all seven cancer types studied and correlated with advanced tumor stages in five of them, marking it as a prime candidate for future diagnostic development.
  • Pathway Convergence: While different cancers have different tRF signatures, they often target the same critical "onco-pathways," such as PI3K-Akt signaling, FoxO signaling, and angiogenesis (blood vessel growth).
  • Immune Connection: ca-tRFs showed a significantly stronger correlation with T-cell infiltration than non-associated tRFs. This suggests that tRFs might play a role in how tumors evade the immune system.
  • Independent Prognostic Value: The tRF-based cancer subtypes identified by the authors often provided prognostic information that was independent of existing classification systems (like iCluster), suggesting that tRFs capture a unique layer of biological information.

Key Results

  • 1,516 ca-tRFs were identified across the seven cancer types.
  • Functional Enrichment: Target gene modules were heavily enriched with known cancer-driving genes (validated via the NCG v6.0 database).
  • Survival Distinction: In cancers like Kidney (KIRC) and Liver (LIHC), the tRF-based subtypes showed dramatic differences in survival rates (p < 0.0001). For instance, in LIHC, Subtype 2 represented a low-risk group, while Subtypes 4 and 5 were highly malignant.
  • Sequence Similarity: The authors found that cancer-specific tRFs often share high sequence similarity with "general" (pan-cancer) tRFs, suggesting a conserved functional core with slight variations for tissue specificity.

Limitations and Future Work

The authors acknowledge several hurdles:

  • AGO-Dominance: The study focused exclusively on tRFs that work through Argonaute proteins (microRNA-like mechanism). tRFs are known to interact with many other RNA-binding proteins, which were not captured here.
  • Data Scarcity: There is a lack of independent, large-scale small RNA-seq datasets outside of TCGA, making it difficult to validate these subtypes in entirely different patient cohorts.
  • Experimental Gap: While the computational evidence is strong, "wet lab" experiments (like CRISPR knockouts or luciferase assays) are needed to confirm the specific mRNA targets of the most promising ca-tRFs.

Ultimately, this work moves tRFs from the "junk" pile to the "control room," providing a roadmap for using these small fragments as the next generation of cancer biomarkers and therapeutic targets.

Summary updated

Abstract

tRNA-derived fragments (tRFs) constitute a novel class of small non-coding RNA cleaved from tRNAs. In recent years, researches have shown the regulatory roles of a few tRFs in cancers, illuminating a new direction for tRF-centric cancer researches. Nonetheless, more specific screening of tRFs related to oncogenesis pathways, cancer progression stages and cancer prognosis is continuously demanded to reveal the landscape of the cancer-associated tRFs. In this work, by combining the clinical information recorded in The Cancer Genome Atlas (TCGA) and the tRF expression profiles curated by MINTbase v2.0, we systematically screened 1,516 cancer-associated tRFs (ca-tRFs) across seven cancer types. The ca-tRF set collectively combined the differentially expressed tRFs between cancer samples and control samples, the tRFs significantly correlated with tumor stage and the tRFs significantly correlated with patient survival. By incorporating our previous tRF-target dataset, we found the ca-tRFs tend to target cancer-associated genes and onco-pathways like ATF6-mediated unfolded protein response, angiogenesis, cell cycle process regulation, focal adhesion, PI3K-Akt signaling pathway, cellular senescence and FoxO signaling pathway across multiple cancer types. And cell composition analysis implies that the expressions of ca-tRFs are more likely to be correlated with T-cell infiltration. We also found the ca-tRF expression pattern is informative to prognosis, suggesting plausible tRF-based cancer subtypes. Together, our systematic analysis demonstrates the potentially extensive involvements of tRFs in cancers, and provides a reasonable list of cancer-associated tRFs for further investigations.
BibTeX
@inproceedings{zhou2022screening,
  author = {Yiran Zhou and Qinghua Cui and Yuan Zhou and Yuan Zhou and Yuan Zhou},
  title = {{Screening and Comprehensive Analysis of Cancer-Associated tRNA-Derived Fragments}},
  booktitle = {Frontiers in Genetics},
  year = {2022},
  doi = {10.3389/fgene.2021.747931},
}
Figure