TL;DR
This paper introduces a high-throughput, long-read single-cell RNA sequencing (scRNA-seq) approach for ovarian cancer, enabling the comprehensive detection of isoforms and genomic alterations at the single-cell level. This method reveals previously uncharacterized isoforms, clarifies gene expression quantification, and uncovers a novel gene fusion, paving the way for more precise oncology and personalized medicine.
What Problem Does This Paper Address?
Understanding the intricate landscape of cancer, particularly its genetic and transcriptomic variations, is crucial for developing effective treatments. Traditional methods often struggle to capture the full spectrum of these alterations, especially at the single-cell level. Specifically, short-read sequencing, while high-throughput, often misses full-length transcript isoforms, leading to an incomplete picture of gene expression. Conversely, long-read sequencing, capable of capturing full-length isoforms, has historically been lower throughput and more challenging to integrate with cell typing. This paper addresses the need for a high-throughput, single-cell method that can simultaneously resolve both transcript isoforms and genomic alterations, providing a more complete genotype-phenotype map of cancer cells.
Methods
The authors employed PacBio HiFi long-read sequencing on ovarian cancer patient samples. To maximize throughput and data quality, they implemented several key strategies:
- Template-switch oligo (TSO) artifact removal: Biotin enrichment was used to reduce artifacts that can inflate read counts.
- cDNA concatenation: Multiple cDNA molecules were linked together per sequencing read to increase effective read length.
- High-depth sequencing: PacBio Sequel II platform was used to achieve a sequencing depth of 12,000 reads per cell.
For isoform classification, they utilized SQANTI, a tool for characterizing long-read transcripts. They also developed a pipeline for identifying gene fusions and mutations. For cell type identification, they leveraged existing marker gene lists and compared clustering results from both short-read and long-read data.
Datasets
The study utilized clinical samples from three high-grade serous ovarian cancer (HGSOC) patients, including omental metastases and matching tumor-free distal omentum tissues.
Key Insights and Ideas
- Isoform complexity is underestimated: The study found that a significant portion (20%) of what's typically considered "protein-coding gene expression" is actually derived from non-coding isoforms, leading to an overestimation of protein levels when using gene-level expression.
- Mesothelial cells transform into CAFs: Evidence suggests that in the metastatic environment, mesothelial cells undergo a transition into cancer-associated fibroblasts (CAFs), potentially driven by the TGF-β/miR-29/Collagen axis.
- Novel gene fusion identified: A previously undetected IGF2BP2::TESPA1 gene fusion was discovered in cancer cells, which was misclassified as high TESPA1 expression in short-read data. This highlights the limitations of short-read sequencing for fusion detection.
- Long-read scRNA-seq is a powerful tool: The technology proved capable of capturing cell type-specific isoforms, genomic alterations (mutations and fusions), and even inferring cell state transitions, demonstrating its potential for personalized medicine.
Key Results
- 152,000 isoforms identified: Of these, over 52,000 were novel.
- 20% overestimation of protein-coding gene expression: Due to the prevalence of non-coding isoforms.
- Identification of a novel IGF2BP2::TESPA1 fusion: This fusion was specific to cancer cells in one patient and was missed by short-read sequencing.
- Accurate mutation calling: Germline and somatic mutations were identified and validated.
- Cell type identification comparable to short-read data: Despite differences in sequencing depth and read length, long-read data enabled robust cell type classification.
Limitations and Future Work
- Low-abundance transcript capture: While high-throughput, the sequencing depth was not sufficient to fully characterize very low-abundance transcripts, such as the T cell receptor repertoire. The authors suggest further increasing sequencing depth or depleting mitochondrial/ribosomal RNA as avenues for future work.
- OAS1 patient-specific expression: Differences in OAS1 isoform expression in one patient suggest potential implications for enzymatic activity, but its impact on ovarian cancer requires further investigation.
- Broader application of the technology: The authors envision long-read scRNA-seq becoming increasingly relevant in oncology and personalized medicine, suggesting its application to other cancer types and for neoantigen detection.
Summary updated