Nanopore-Based Detection of Viral RNA Modifications

Abstract

The chemical modification of ribonucleotides plays an integral role in the biology of diverse viruses and their eukaryotic host cells. Mapping the precise identity, location, and abundance of modified ribonucleotides remains a key goal of many studies aimed at characterizing the function and importance of a given modification. While mapping of specific RNA modifications through short-read sequencing approaches has powered a wealth of new discoveries in the past decade, this approach is limited by inherent biases and an absence of linkage information. Moreover, in viral contexts, the challenge is increased due to the compact nature of viral genomes giving rise to many overlapping transcript isoforms that cannot be adequately resolved using short-read sequencing approaches. The recent emergence of nanopore sequencing, specifically the ability to directly sequence native RNAs from virus-infected host cells, provides not just a new methodology for mapping modified ribonucleotides but also a new conceptual framework for what can be derived from the resulting sequencing data. In this minireview, we provide a detailed overview of how nanopore direct RNA sequencing works, the computational approaches applied to identify modified ribonucleotides, and the core concepts underlying both. We further highlight recent studies that have applied this approach to interrogating viral biology and finish by discussing key experimental considerations and how we predict that these methodologies will continue to evolve.

Keywords: RNA modifications; m6A; nanopore sequencing; virus.

FIG 1

Overview of the nanopore direct RNA sequencing approach and RNA modification detection strategies. (a and b) DRS libraries comprise RNA-cDNA hybrids that are unwound by a helicase when docked to a membrane-embedded nanopore. (a) The RNA strand is selectively pushed through the nanopore in a 3′-to-5′ direction, disturbing the general flow of ions through the pore. (b) Ammeters placed within the narrowest part of the pore measure changes in ionic flow (current), which can subsequently be partitioned into sections attributed to the DNA adapter, poly(A) tail, and body of the sequenced RNA, here visualized as a squiggle plot (124). (c and d) Putatively modified ribonucleotides are identified through either error rate or signal-level analyses. (c) Error rate analyses identify individual positions within RNAs that show significantly different error profiles between two conditions (e.g., control versus treatment). Here, the proportion of reads containing the correct nucleotide at a given position is shown in colors other than gray, while the proportion of reads containing erroneous bases at the same position is shown in gray. The difference in calculated error rate is shown in gold. (d) Error profiles undergo statistical analyses in the form of 2 × 2, 2 × 3, or 2 × 5 contingency tables that are subsequently corrected for multiple testing. M, match; MM, mismatch. (e) For signal-level analyses, the initial process of resquiggling assigns ionic flow features (signal intensity, dwell time, etc.) to their corresponding nucleotides within individual reads. (f) These features either undergo a comparative analysis (between data sets) or are evaluated using data from a prebuilt model generated by machine and/or deep learning.

FIG 2

Brief overview of machine learning.