NanoViromics: long-read sequencing of dsRNA for plant virus and viroid rapid detection

Abstract

There is a global need for identifying viral pathogens, as well as for providing certified clean plant materials, in order to limit the spread of viral diseases. A key component of management programs for viral-like diseases is having a diagnostic tool that is quick, reliable, inexpensive, and easy to use. We have developed and validated a dsRNA-based nanopore sequencing protocol as a reliable method for detecting viruses and viroids in grapevines. We compared our method, which we term direct-cDNA sequencing from dsRNA (dsRNAcD), to direct RNA sequencing from rRNA-depleted total RNA (rdTotalRNA), and found that it provided more viral reads from infected samples. Indeed, dsRNAcD was able to detect all of the viruses and viroids detected using Illumina MiSeq sequencing (dsRNA-MiSeq). Furthermore, dsRNAcD sequencing was also able to detect low-abundance viruses that rdTotalRNA sequencing failed to detect. Additionally, rdTotalRNA sequencing resulted in a false-positive viroid identification due to the misannotation of a host-driven read. Two taxonomic classification workflows, DIAMOND & MEGAN (DIA & MEG) and Centrifuge & Recentrifuge (Cent & Rec), were also evaluated for quick and accurate read classification. Although the results from both workflows were similar, we identified pros and cons for both workflows. Our study shows that dsRNAcD sequencing and the proposed data analysis workflows are suitable for consistent detection of viruses and viroids, particularly in grapevines where mixed viral infections are common.

Keywords: Illumina Miseq sequencing; dsRNA; grapevine viruses; nanopore sequencing; plant pathology; total RNAs; viroid; virus detection.

Figure 1

Library preparation workflows for grapevine virus and viroid detection. Direct RNA sequencing (top): Initially, rRNAs were depleted from total RNA and then polyadenylated. Next, different custom reverse transcription adapters were ligated to polyadenylated transcripts, and after reverse transcription (RT), the sequencing adapter was ligated, and a pooled library was loaded on a R9.4.1 flow cell. Direct cDNA sequencing (left-down): Following dsRNA extraction, double-stranded cDNA was synthesized using random primers followed by Klenow polymerase and E. coli DNA ligase to create second strand cDNA. A commercial barcode was ligated to each sample after double-stranded cDNAs were end-prepared. The pooled library was made from 23 different barcoded samples, and after ligating the sequencing adapter, the pooled library was primed and loaded on another R9.4.1 flow cell. (The figure was designed by BioRender.com).

Figure 2

Data analysis workflows for grapevine virus and viroid detection through nanopore sequencing. Preprocessing of raw data (left): The raw data were acquired through MinKNOW software, and then base-calling, filtering based on the quality score (<7), demultiplexing, trimming, the removal of host reads and clustering were performed, followed by error correction and polishing, using various software and software packages. Centrifuge and Recentrifuge strategy (Upper right): Preprocessed reads were taxonomically classified using Centrifuge, followed by a comprehensive analysis by Recentrifuge, which provided a visualization. DIAMOND+MEGAN strategy (Lower right): In addition, the same reads from the previous step were aligned against annotated protein sequences (NCBI-nr) using DIAMOND, and subsequently, MEGAN 6 was used to bin the sequences based on their taxonomy and function profiles. To further curate the taxonomic results, BLASTn and BLASTx were also run.

Figure 3

Comparative analysis by Recentrifuge software of the Centrifuge results. Screenshots of the Recentrifuge HTML interface for different samples and sequencing technologies are shown. The virus superkingdom is drawn in the center, with a hierarchical pie chart. Depending on the confidence level associated with the taxonomic classification, the background color varies for each taxon. (For the results from the other 20 samples, see available in Zenodo: 10.5281/zenodo.7764376.)

Figure 4

The number of assigned reads associated with grapevine viruses and viroids in different datasets. PvEV1: Phaseolus vulgaris endornavirus 1; PvEV2: Phaseolus vulgaris endornavirus 2; PvEV3: Phaseolus vulgaris endornavirus 3; GLRaV2: grapevine leafroll-associated virus 2; GLRaV3: grapevine leafroll-associated virus 3; GRSPaV: grapevine rupestris stem pitting-associated virus; GVB: grapevine virus B; GSV1: grapevine Syrah virus 1; GRGV: grapevine red globe virus; GPGV: grapevine pinot gris virus; GYSVd: grapevine yellow speckle viroid; GYSVd1: grapevine yellow speckle viroid 1; HSVd: hop stunt viroid; CEVd: citrus exocortis viroid. (The heat map table was generated in https://www.datawrapper.de).

Figure 5

Comparative visualization of DIAMOND results by MEGAN software. Multiple datasets (1. Nanopore dsRNAcD sequencing; 2. Nanopore rdTotalRNA sequencing; and 3. dsRNA-MiSeq sequencing) were uploaded to MEGAN and the relative abundance of reads in each sample was extracted.

Figure 6

Assignment of viral reads to taxa using different thresholds in the rdTotalRNA dataset from the BV-12-16 J sample. (A) Taxonomic binning was done using the default parameters in the meganization option of the MEGAN software. (B) Two thresholds, “Min Support Percent” and “Min Support,” were turned off (=0) when meganization was performed. The two numbers following the virus name (red color) consist of the NCBI taxonomy ID followed by a number that indicates the number of assigned reads.

Figure 7

A heatmap illustrating the percentage of viral genome coverage and depth for three different grapevine samples. In each cell, the associated number represents the percentage of genome coverage, followed by the depth of coverage in parentheses. Colors that are darker indicate higher genome coverage percentages.