HIT-scISOseq: High-throughput and High-accuracy Single-cell Full-length Isoform Sequencing for Corneal Epithelium

Yingfeng Zheng, Zhichao Chen, Zhuoxing Shi, Kun-Hua Hu, Jiayong Zhong, Chunxiao Wang, Wen Shi, Ying Chen, Shuang Xie, Feng Luo, Xiaochen Bo, Chong Tang, Yizhi Liu, Chuan‐Le XiaobioRxiv 2020 · DOI

TL;DR

This paper introduces HIT-scISOseq, a novel method that significantly boosts the throughput and accuracy of single-cell full-length isoform sequencing. By overcoming limitations of existing long-read technologies, HIT-scISOseq enables a more comprehensive understanding of cellular transcriptomes, revealing novel isoforms and complex splicing events previously inaccessible.

What Problem Does This Paper Address?

Single-cell RNA sequencing (scRNA-seq) has revolutionized our ability to study cellular heterogeneity. However, traditional short-read scRNA-seq technologies are limited in their ability to resolve complex transcript isoforms, alternative splicing, and chimeric transcripts. While long-read sequencing technologies offer the potential to capture full-length transcripts, they have historically suffered from low throughput and high error rates, making them expensive and difficult to scale for large single-cell studies. This paper addresses the critical need for a high-throughput, high-accuracy, and cost-effective method for single-cell full-length isoform sequencing.

Methods

The core innovation of HIT-scISOseq lies in its two-phase approach:

  • Capture Phase: To mitigate artifacts common in 10x Genomics library preparation (specifically, template-switching oligonucleotide or TSO artifacts), the researchers developed a biotinylated PCR primer strategy. This primer captures desirable full-length cDNAs, allowing for their enrichment via streptomycin magnetic beads. This effectively reduces TSO artifacts from ~50% to less than 8%.
  • Concatenation Phase: To maximize the utility of PacBio's long-read sequencing capacity, the researchers devised a novel library preparation method. This involves ligating multiple full-length cDNA molecules end-to-tail to create a single, long SMRTbell insert. This "concatenation" approach allows for more efficient sequencing on PacBio's Sequel II platform, yielding significantly more data per SMRT cell.

Additionally, the authors developed a custom computational pipeline, "scISA-Tools," specifically designed to process the linked transcript reads generated by HIT-scISOseq, which are not compatible with standard IsoSeq3 pipelines.

Datasets

The primary dataset used for demonstrating HIT-scISOseq's capabilities comes from cynomolgus monkey corneal epitheliums. The study analyzed approximately 4,000 normal and 8,000 injured corneal epithelial cells. The authors also performed cross-platform validation using Illumina short-read RNA sequencing (NGS) on the same single-cell 10x Genomics cDNA samples.

Key Insights and Ideas

  • Bridging the Gap: HIT-scISOseq effectively bridges the gap between the high throughput of short-read sequencing and the isoform-resolution capabilities of long-read sequencing, offering the best of both worlds.
  • Artifact Reduction is Crucial: The significant reduction in TSO artifacts is a key enabler for improving data quality and downstream analysis accuracy.
  • Concatenation is Key for Throughput: The clever concatenation of multiple transcripts into a single long read is a critical innovation for maximizing data output from expensive long-read sequencing platforms.
  • Unveiling the "Dark Matter" of the Transcriptome: By enabling high-resolution isoform analysis, HIT-scISOseq can reveal novel isoforms and complex splicing events that are often missed by short-read methods, providing a deeper understanding of gene regulation and function.

Key Results

  • Throughput Boost: HIT-scISOseq generates over 10 million high-accuracy full-length isoforms from a single PacBio Sequel II 8M SMRT Cell, representing an 8-fold increase in data output compared to standard single-cell isoform sequencing protocols.
  • Artifact Reduction: The method reduces TSO artifacts from ~50% to <8%.
  • High Accuracy: The method achieves high accuracy, with mapped FLNC reads showing high mappability (>98%) and mapping identity (>97%).
  • Comprehensive Transcriptome Profiling: The study successfully profiled transcriptomes of thousands of corneal epithelial cells, identifying cell-type-specific expression patterns, novel isoforms, and injury-related splicing events.
  • Cross-Platform Concordance: HIT-scISOseq data showed strong concordance with Illumina NGS data in terms of gene expression quantification and cell-type classification.

Limitations and Future Work

  • Concatenation Length Optimization: The authors found that while longer concatenations are possible, they increase the risk of irreversible cDNA damage (nicks), which can impair polymerase performance. They optimized their system for ~5 kb concatenations (2-3 full-length transcripts) to balance length and quality. Future work could focus on further optimizing input quality by reducing DNA nicks and exploring size selection methods for longer concatenated molecules.
  • Broader Applicability: While the study focused on mRNA, the authors suggest that the HIT-scISOseq approach could be adapted to target other RNA species like LncRNA or eccRNA by altering concatenation library composition.
  • Multiplexed Analysis: The high accuracy and throughput of HIT-scISOseq make it suitable for multiplexed single-cell RNA-sequencing of pooled individuals, enabling the study of natural polymorphisms and sample identification.
  • Platform Adaptability: HIT-scISOseq is compatible with commercially available 10x Genomics platforms and is expected to be adaptable to other microwell-based and combinatorial-indexing technologies.

Summary updated

Abstract

Abstract Single-cell isoform sequencing can reveal transcriptomic dynamics in individual cells invisible to bulk- and single-cell RNA analysis based on short-read sequencing. However, current long-read single-cell sequencing technologies have been limited by low throughput and high error rate. Here we introduce HIT-scISOseq for high-throughput single-cell isoform sequencing. This method was made possible by full-length cDNA capture using biotinylated PCR primers, and by our novel library preparation procedure that combines head-to-tail concatemeric full-length cDNAs into a long SMRTbell insert for high-accuracy PacBio sequencing. HIT-scISOseq yields &gt; 10 million high-accuracy full-length isoforms in a single PacBio Sequel II 8M SMRT Cell, providing &gt; 8 times more data output than the standard single-cell isoform PacBio sequencing protocol. We exemplified HIT-scISOseq by first studying transcriptome profiles of 4,000 normal and 8,000 injured corneal epitheliums from cynomolgus monkeys. We constructed dynamic transcriptome landscapes of known and rare cell types, revealed novel isoforms, and identified injury-related splicing and switching events that are previously not accessible with low throughput isoform sequencing. HIT-scISOseq represents a high-throughput, cost-effective, and technically simple method to accelerate the burgeoning field of long-read single-cell transcriptomics.
BibTeX
@misc{zheng2020hit,
  author = {Yingfeng Zheng and Zhichao Chen and Zhuoxing Shi and Kun-Hua Hu and Jiayong Zhong and Chunxiao Wang and Wen Shi and Ying Chen and Shuang Xie and Feng Luo and Xiaochen Bo and Chong Tang and Yizhi Liu and Chuan‐Le Xiao},
  title = {{HIT-scISOseq: High-throughput and High-accuracy Single-cell Full-length Isoform Sequencing for Corneal Epithelium}},
  year = {2020},
  doi = {10.1101/2020.07.27.222349},
  howpublished = {bioRxiv},
}
Figure