Background

Dicistronic poly(A)-mRNA-tRNA transcripts occur when a protein-coding gene (PCG) and its adjacent tRNA gene are transcribed as a single transcriptional unit. Prior work shows that such co-transcribed mRNA-tRNA molecules can move systemically through the plant to target distant tissues [1]. The transgenic removal of tRNA or tRNA-like structure (TLS) sequences from these dicistronic transcripts renders them immobile, suggesting that movement is controlled by signals within the tRNA structure, sequence, or both [1]. Furthermore, as part of the co-transcribed units, mRNAs are translated to functional proteins in the target tissue, suggesting tRNAs or TLSs can facilitate the movement of these transcripts, potentially expanding their role in non-cell autonomous signaling pathways [1,2]. In field-grown commercial grapevines, the expression patterns of dicistronic tRNA-mRNAs were linked to the geographical location of the vineyards as well as the tissue type analyzed [3], which indicates that the expression patterns of these transcripts may be related to the microclimate, such as temperature or light, in different vineyards. We previously reported the development of a reusable computational workflow to reliably detect dicistronic tRNA-mRNA transcripts from RNA-seq datasets. The dicistronic tRNA-mRNA transcript (DiRT) workflow identified dicistronic tRNA-mRNA transcripts in multiple vascular species but not in non-vascular lineages, suggesting that their co-transcription may be related to phloem transportation of mRNA into distant tissues [3].

Although the majority of nuclear-encoded genes in plants are expressed as monocistronic units, an increasing number of studies report that plants utilize polycistronic and dicistronic transcripts to coordinate complex regulatory networks. For example, plant small nucleolar RNAs (snoRNAs) are frequently co-transcribed with adjacent tRNAs as dicistronic precursors [4]. In Arabidopsis thaliana and rice, these snoRNAs are transcribed from an upstream tRNA promoter to form dicistronic tRNA-snoRNA transcripts, which are subsequently cleaved into independent, functional tRNA and snoRNA [4]. Similarly, microRNAs (miRNAs) have been shown to cluster together in the rice genome and transcribed as a single polycistronic transcript [5]. These findings underscore the genome-wide prevalence of eukaryotic polycistronic transcription. While the biological roles of these co-transcribed units are still emerging, accurately capturing these complex transcripts directly from widely available short-read transcriptomic data remains a bottleneck. The first version of the DiRT pipeline v1 was developed to identify dicistronic tRNAs located upstream or downstream of mRNAs [3], but did not include tRNA-related sequences present in the untranslated regions (UTRs) of protein-coding genes. Indeed, many of the TLSs that share some structural similarities with canonical tRNAs but are non-functional as a translational adaptor are found localized to UTR regions of protein-coding genes [1]. Transgenic studies using TLSs derived from a tRNA in which specific functional domains were deleted were still able to confer long-distance mobility of their co-transcribed mRNA [1,2]. In Arabidopsis, TLS motifs located in the UTR of CK1 play a key role in regulating gene function by acting as a mobility motif for the CK1 transcript [1]. Furthermore, studies of turnip yellow mosaic virus (TYMV) and brome mosaic virus have demonstrated that TLS in the UTR regions of viral genes can mimic endogenous plant tRNAs to trigger systemic mobility through the phloem [6–8]. Together, these findings highlight the biological relevance of TLS and the importance of computational workflows to reliably detect dicistronic UTR-localized TLS sequences, as they may be functionally important in mobile signaling pathways in plants. Here, we address this gap by introducing DiRT pipeline v2, an enhanced version that incorporates the detection of tRNA-related sequences within mRNA UTRs.

Dicistronic tRNA-mRNA transcripts are also known to exhibit low expression levels compared to monocistronic nuclear-encoded protein-coding genes based on RNA-seq data [1,3], indicating the need for internal expression controls to distinguish biologically relevant transcript signals from technical noise or sequencing artifacts. The DiRT v1 pipeline relied on active transcription in the intergenic region between the tRNA and mRNA as indicative of co-transcription and compared the read coverage in this region with the two closest introns in the dicistronic protein-coding genes. tRNA-mRNA combinations were selected for further analysis in the workflow if the read coverage in the intergenic region was significantly higher than the intron-level coverage. However, this strategy excluded analyzing potential dicistronic tRNA-mRNA transcripts from protein-coding genes with one or no introns. As a significant proportion of protein-coding genes in plant species have one or no introns [9,10], excluding these genes may cause significant bias in the analysis of potential dicistronic tRNA-mRNA transcripts.

To overcome this bias, we modified the workflow for genes with no introns. Here, the average sequencing depth of the intergenic region [between the tRNA/TLS and the coding sequence (CDS) forming the given transcript] is compared with the top depth obtained for the first two introns from the closest protein-coding gene. For the analysis of dicistronic protein-coding genes with a single intron, an intron from the next closest flanking gene was used for sequencing depth comparison. After implementing these adjustments to control for minimum expression threshold levels, RNA coverage of each base pair of the intergenic region was estimated. Candidates that show both intergenic region continuous coverage and sequencing depth above background noise were classified as dicistronic tRNA-mRNA transcripts.