Nanopore long and direct RNA-seq of coronaviruses

Sequence analyses of RNA virus genomes remain challenging due to the exceptional genetic plasticity of these viruses. Although standard (short-read) sequencing technologies allow to readily determine consensus sequences for these 'quasispecies', it is far more difficult to reconstruct large numbers of full-length haplotypes of (i) RNA virus genomes and (ii) subgenome-length (sg) RNAs comprised of noncontiguous genome regions that may be present in these virus populations. Here, we used a full-length, direct RNA sequencing (DRS) approach without any amplification step utilizing nanopore technology to characterize viral RNAs produced in cells infected with a human coronavirus representing one of the largest RNA virus genomes known to date. We also show how supplementary short-read sequencing (Illumina) can be used to reduce the error rate of nanopore sequencing.Nanopore long-read data:For nanopore sequencing, 1 µg of RNA in 9 µl was carried into the library preparation with the Oxford nanopore direct RNA sequencing protocol (SQK-RNA001). All steps were followed according to the manufacturer's specifications. The library was then loaded on an R9.4 flow cell and sequenced on a MinION device (Oxford Nanopore Technologies). The sequencing run was terminated after 48 h. We sequenced two samples (WT and SL2).The raw signal data was basecalled using Albacore (v2.2.7, available through the Oxford nanopore community forum). Besides the basecalled data (fastq), we also deposited the raw signal files (fast5) at the OSF (https://doi.org/10.17605/OSF.IO/UP7B4). For the WT sample, the MinION was started 4 times and for the SL2 sample, it was started 2 times. We combined all signal files of WT and SL2, respectively, for basecalling.Illumina short-read data:Illumina short-read sequencing was performed using the TruSeq RNA v2 kit to obtain RNA from species with polyA-tails and without any strand information. The three samples (WT, SL2_SARS, SL2_BCoV) selected for this study were prepared on a HiSeq 2500 lane and sequenced with 51 cycles. After demultiplexing, 23.2, 22.0, and 23.8 million single-end reads were obtained for the WT and the two SL2 samples, respectively. The raw sequencing data can be found here (fastq).