Time-Course Transcriptome Profiling of a Poxvirus Using Long-Read Full-Length Assay

Abstract

Viral transcriptomes that are determined using first- and second-generation sequencing techniques are incomplete. Due to the short read length, these methods are inefficient or fail to distinguish between transcript isoforms, polycistronic RNAs, and transcriptional overlaps and readthroughs. Additionally, these approaches are insensitive for the identification of splice and transcriptional start sites (TSSs) and, in most cases, transcriptional end sites (TESs), especially in transcript isoforms with varying transcript ends, and in multi-spliced transcripts. Long-read sequencing is able to read full-length nucleic acids and can therefore be used to assemble complete transcriptome atlases. Although vaccinia virus (VACV) does not produce spliced RNAs, its transcriptome has a high diversity of TSSs and TESs, and a high degree of polycistronism that leads to enormous complexity. We applied single-molecule, real-time, and nanopore-based sequencing methods to investigate the time-lapse transcriptome patterns of VACV gene expression.

Keywords: gene expression; long-read sequencing; nanopore sequencing; transcriptome profiling; vaccinia virus.

Figure 1

Read counts and lengths of uninfected and infected samples at each time point. (a) Fractions of reads mapped to host (C. sabaeus) and VACV genomes; dramatically reduced read counts of the host and an increased read count of the virus are observable with viral life cycle progression in MinION data, whereas similar but much smaller changes were detected in the Sequel dataset. (b) Mapped read lengths of the host and virus from MinION and Sequel sequencing from the uninfected sample and post-infection (p.i.) samples; the panel shows the effect of library preparation for PacBio sequencing, yielding longer reads with greater abundance than those from MinION sequencing.

Figure 2

Genome-wide dynamics of VACV TSSs and TESs. (a) Blue dashes represent TSSs on the forward strand, while red dashes represent TSSs on the reverse strand. (b) Blue dashes represent TESs on the forward strand, while red dashes represent TESs on the reverse strand. (a,b) Yellow rectangles represent the ORFs.

Figure 3

Expression profiles of the most abundant dynamic VACV transcripts and transcript isoforms. Rt values of the viral transcripts show distinct expression profiles in the five kinetic clusters. (a) The expression pattern of the viral transcript clusters. Five distinct VACV transcript clusters represented by a heatmap matrix. Each row represents changes in relative expression levels of an RNA. Red rectangles indicate high relative expression values and black rectangles indicate low relative expression values. (b) K-means clustering analysis was used to characterize the temporal expression of the five clusters obtained by hierarchical clustering. Relative Rt mean values of clusters are delineated in different time points of infection. All transcripts were identified using the LoRTIA pipeline. We note here that full-length sequencing provides a unique technology that is able to detect individual transcripts. Techniques lacking this ability (qPCR, short-read sequencing) cannot be used for the validation of the obtained results due to the high complexity of transcriptional overlaps in vaccinia virus.

Figure 4

Dynamic VACV transcriptome. (a) Time-dependent changes of TSS isoforms of the A8R–A9L complex transcript and the most abundant TES isoforms of the A9L gene. The earliest time point when the LoRTIA pipeline identified the c-A8R–A9L complex RNA, which spans two oppositely oriented genes at this genomic region, was 3 h p.i. This transcript was also detected at 4 and 6 h p.i., but disappeared later. The A9L shorter 3′-UTR isoform was expressed at 6 h p.i., while its longer variant was detected at 8 h p.i.; (b) Expression levels of various transcripts and transcript isoforms within the M1L–K2L region, including transcriptional end site (TES) isoforms, mono- and bicistronic variants, and novel putative protein-coding genes. The characteristic feature of genes located in this genomic locus is that they produce longer 3′-UTR isoforms at later time points of infection; transcript structures and counts were determined using the LoRTIA software suite. See Table 1 for terminology.

Figure 5

Expression dynamics VACV transcripts. (a) Kinetics of mono-, bi-, and polycistronic transcripts. The B25.5R–B29R genomic region contains eight ORFs of which no LoRTIA transcripts were detected at 1 h p.i. A monocistronic (B25.5R) and a pentacistronic transcript (B25R–B28R) were expressed at 2–4 h p.i. interval. At later time points, the pentacistronic RNA and its longer overlapping partner ((a), hexacistronic transcript) are expressed; however, the monocistronic transcript disappeared. (b) Expression dynamics of the replication-associated RNAs. Several transcripts were detected at the two ends of the VACV genome, around the replication origin of the virus. These transcripts appeared at 4 h p.i., following the initiation of the DNA replication, which suggests that they might have a regulatory role in this process. See Table 1 for terminology.

Figure 6

Categories of the expression curves of the vaccinia virus transcripts containing the same ORFs. Based on their dynamics, VACV genes can be grouped into five distinct clusters. The limitation of this approach is that only the abundant LoRTIA transcripts can be used for the analysis, while those lower-abundance overlapping transcripts that are not identified by the LoRTIA software are missing from the calculation. Other techniques (e.g., short-read sequencing, real-time RT-PCR) are not suitable for distinguishing between the overlapping transcripts and transcript isoforms; therefore, data generated by these techniques are not comparable with our results.