Sequencing-based quantitative mapping of the cellular small RNA landscape

Jennifer Hu, Daniel Yim, Sabrina M. Huber, Jo Marie Bacusmo, Duanduan Ma, Michael S. DeMott, Stuart S. Levine, Valérie de Crécy‐Lagard, Peter C. Dedon, Bo CaobioRxiv 2019 · DOI

TL;DR

Researchers have developed AQRNA-seq, a novel RNA sequencing method that overcomes significant biases in current techniques, enabling precise and absolute quantification of all small RNA species in a sample. This breakthrough allows for a more accurate understanding of RNA regulation and biology, particularly for challenging RNA types like tRNAs.

What Problem Does This Paper Address?

Accurately quantifying the abundance of different RNA molecules within a biological sample is crucial for understanding gene expression and regulation. However, existing next-generation RNA sequencing (RNA-seq) methods suffer from substantial biases during library preparation, especially when dealing with small RNAs. These biases, primarily introduced during linker ligation and amplification steps, can lead to inaccurate read counts that vary by orders of magnitude depending on the RNA's sequence and structure. This makes it difficult to reliably measure the true abundance of small RNAs, particularly highly structured and modified molecules like tRNAs, hindering our understanding of their roles in cellular processes.

Methods

The core of this work is the development of the AQRNA-seq method, which systematically addresses the biases in traditional RNA-seq. Key innovations include:

  • Optimized Linker Ligation: AQRNA-seq employs a two-step linker ligation strategy. Linker 1 is ligated to the 3' end of RNA with randomized nucleotides at its 5' end to improve efficiency across different sequences. Linker 2 is ligated to the 5' end of the cDNA after reverse transcription, ensuring that even truncated cDNA fragments are captured.
  • AlkB Demethylation: A non-essential but informative step using AlkB enzyme is included to remove certain methyl modifications that can block reverse transcriptase, thus reducing premature cDNA truncation and providing insights into modification sites.
  • RecJ Exonuclease Treatment: This enzyme is used to efficiently remove excess, unligated linkers, eliminating the need for cumbersome purification steps like HPLC.
  • Custom Data Mining Workflow: A specialized bioinformatics pipeline is developed to process the sequencing data, carefully handling multiply-mapped reads and ensuring accurate quantification.

Datasets

The AQRNA-seq method was validated using several datasets:

  • Synthetic RNA Oligonucleotides: Five synthetic oligos of varying lengths (25-80 nt) were used to demonstrate a linear correlation between sequencing read count and input RNA abundance.
  • Miltenyi miRXplore Universal Reference: A library of 963 equimolar synthetic microRNAs with diverse lengths and sequences was used to assess sequence-dependent biases.
  • E. coli K12 tRNA Pool: Small RNAs (<200 nt) from E. coli were used to compare AQRNA-seq results with established methods like 2D gel electrophoresis and Northern blotting.
  • Mycobacterium bovis BCG: Total small RNA was isolated from M. bovis BCG under various stress conditions (nutrient deprivation) to study tRNA dynamics.

Key Insights and Ideas

  • Bias Mitigation is Key: The paper strongly emphasizes that sequence-dependent biases are the primary bottleneck in accurate small RNA quantification by RNA-seq. AQRNA-seq's success hinges on meticulously addressing these biases at multiple steps.
  • tRNAs are a Tough Nut to Crack: The study highlights the significant challenges in sequencing tRNAs due to their complex structures and extensive modifications, which interfere with standard library preparation. AQRNA-seq's ability to handle these molecules is a major advancement.
  • Beyond Abundance: Deeper Insights: AQRNA-seq doesn't just quantify RNA abundance; it also provides single-nucleotide resolution for mapping RNA modifications and identifying fragmentation sites, offering a much richer picture of RNA biology.
  • Absolute Quantification is Achievable: The development of a method that demonstrates a direct, linear correlation between read count and RNA abundance is a significant step towards true absolute quantification, moving beyond relative expression levels.

Key Results

  • Linear Correlation: AQRNA-seq demonstrated a direct, linear correlation between sequencing read count and input RNA abundance for synthetic oligos (r² = 0.92-1).
  • Reduced Bias: When tested with a diverse miRNA library, AQRNA-seq achieved ~75% of miRNAs within a two-fold variation of expected abundance, significantly outperforming existing commercial kits (which typically achieve only 40-50%). Jackpot and dropout sequences were minimal (<3%).
  • tRNA Validation: AQRNA-seq results for E. coli tRNA pools showed strong agreement (r² = 0.81) with orthogonal methods like 2D gel electrophoresis and Northern blotting.
  • Bacterial Stress Response: Application to M. bovis BCG revealed significant insights into tRNA dynamics during stress, including:
  • Up to 80-fold variation in tRNA isoacceptor copy numbers.
  • Site-specific tRNA fragmentation patterns induced by stress.
  • Quantitative maps of ribonucleoside modifications.

Limitations and Future Work

  • AlkB Step: While informative, the AlkB demethylation step is noted as non-essential for basic RNA capture, suggesting that its inclusion is for deeper analysis of modifications.
  • Generalizability: While the paper states AQRNA-seq is broadly applicable to any form of RNA, the primary focus and validation were on small RNAs (miRNAs and tRNAs). Further validation on longer RNAs like mRNA and rRNA is mentioned as a future direction, with the addition of a Covaris fragmentation step suggested for longer molecules.
  • Modification Mapping: The paper suggests that modification mapping could aid in the discovery of previously unannotated or unlocalized modifications in poorly characterized species, indicating a promising avenue for future research.
  • Data Mining Complexity: The paper acknowledges the need for careful library curation and data mining, especially for closely related sequences, implying that ongoing refinement of these algorithms will be beneficial.

Summary updated

Abstract

Abstract Current next-generation RNA sequencing methods cannot provide accurate quantification of the population of small RNAs within a sample due to strong sequence-dependent biases in capture, ligation, and amplification during library preparation. We report the development of an RNA sequencing method – AQRNA-seq – that minimizes biases and enables absolute quantification of all small RNA species in a sample mixture. Validation of AQRNA-seq library preparation and data mining algorithms using a 963-member microRNA reference library, RNA oligonucleotide standards of varying lengths, and northern blots demonstrated a direct, linear correlation between sequencing read count and RNA abundance. Application of AQRNA-seq to bacterial tRNA pools, a traditionally hard-to-sequence class of RNAs, revealed 80-fold variation in tRNA isoacceptor copy numbers, patterns of site-specific tRNA fragmentation caused by stress, and quantitative maps of ribonucleoside modifications, all in a single AQRNA-seq experiment. AQRNA-seq thus provides a means to quantitatively map the small RNA landscape in cells and tissues.
BibTeX
@misc{hu2019sequencing,
  author = {Jennifer Hu and Daniel Yim and Sabrina M. Huber and Jo Marie Bacusmo and Duanduan Ma and Michael S. DeMott and Stuart S. Levine and Valérie de Crécy‐Lagard and Peter C. Dedon and Bo Cao},
  title = {{Sequencing-based quantitative mapping of the cellular small RNA landscape}},
  year = {2019},
  doi = {10.1101/841130},
  howpublished = {bioRxiv},
}
Figure