Rolling circle reverse transcription enables high fidelity nanopore sequencing of small RNA

Abstract

Small RNAs (sRNAs) are an important group of non-coding RNAs that have great potential as diagnostic and prognostic biomarkers for treatment of a wide variety of diseases. The portability and affordability of nanopore sequencing technology makes it ideal for point of care and low resource settings. Currently sRNAs can't be reliably sequenced on the nanopore platform due to the short size of sRNAs and high error rate of the nanopore sequencer. Here, we developed a highly efficient nanopore-based sequencing strategy for sRNAs (SR-Cat-Seq) in which sRNAs are ligated to an adapter, circularized, and undergo rolling circle reverse transcription to generate concatemeric cDNA. After sequencing, the resulting tandem repeat sequences within the individual cDNA can be aligned to generate highly accurate consensus sequences. We compared our sequencing strategy with other sRNA sequencing methods on a short-read sequencing platform and demonstrated that SR-Cat-Seq can obtain low bias and highly accurate sRNA transcriptomes. Therefore, our method could enable nanopore sequencing for sRNA-based diagnostics and other applications.

Fig 1. Comparison of rolling circle reverse transcription ability by various reverse transcriptases.

(A) Agarose gel electrophoresis was applied to visualize the large cDNA products. Each enzyme was evaluated with both a linear (L) and a circular (C) RNA template. The same reactions products were also resolved by denaturing PAGE gel electrophoresis to visualize the small cDNA products with the (B) a linear RNA template and (C) a circular RNA template.

Fig 2. Small RNA library preparation workflow for nanopore sequencing (SR-Cat-Seq).

(A) Schematic illustration of SR-Cat-Seq workflow. (B) Component requirement for SR-Cat-Seq. (C) Quantification of the rolling circle reverse transcription reactions with or without RNA input.

Fig 3. Quantification of small RNA adapter ligation step and circularization step in the SR-Cat-Seq workflow.

Three sets of reactions were performed in parallel, with representative capillary electrophoresis traces shown. One set without the adapter (column 1, No Adapter), one set without the ligase (column 2, No Circularization) and one set including both (column 3, Full Workflow). (A) Reactions were sampled after adapter ligation. (B) Reactions were sampled after circularization. Note that the circular products migrate faster through the capillary, making their size appear smaller. (C) Reactions were sampled after exoribonuclease digestion. Note that free FAM dye or very short oligos migrate slower through the capillary due to low charge, which causes them to appear larger than their actual size.

Fig 4. High sequencing accuracy and low bias of SR-Cat-Seq workflow.

(A) The number of repeats identified within individual concatemeric cDNA read was plotted against the read length. (B) Consensus sequences assembled from the repeats were mapped to the reference sequence. The accuracy was determined by calculating the percent identity to the reference. The number of reads in each accuracy bin were plotted. (C) Comparison of sequencing bias of four different small RNA sequencing workflows (TruSeq, Low-Bias, SR-Cat-Amp and SR-Cat). Normalized value of each miRNA in miRXplore was plotted for each workflow.

Fig 5. Sequencing human brain small RNA with SR-Cat-Seq workflow.

(A) Comparison of the relative abundance of the different ncRNA categories with four different small RNA sequencing workflows (TruSeq, Low-Bias, SR-Cat-Amp and SR-Cat). (B) Left: Comparison of the number of unique miRNA species detected by four workflows. Right: Venn diagram showing overlap of miRNA species detected among the four workflows. C) Correlations of read counts of miRNA in human brain RNA among the four workflows. The diagonal shows the correlation of two technical replicates of each method. R² values are plotted in the upper-left corner of each plot. D) Correlation of miRNA read counts from each workflow compared to qPCR cycle threshold values from human brain RNA. R² values and the percentage of miRNAs detected with rpm > 2 by each sequencing method are plotted in the upper-left corner of each plot.