The RNA degradome: a precious resource for deciphering RNA processing and regulation codes in plants

ABSTRACT

The plant RNA degradome was defined as an aggregate of the RNA fragments degraded from various biochemical pathways, such as RNA turnover, maturation and quality surveillance. In recent years, the degradome sequencing (degradome-seq) libraries became a rich storehouse for researchers to study on RNA processing and regulation. Here, we provided a brief overview of the uses of degradome-seq data in plant RNA biology, especially on non-coding RNA processing and small RNA-guided target cleavages. Some novel applications in RNA research area, such as in vivo mapping of the endoribonucleolytic cleavage sites, identification of conserved motifs at the 5ʹ ends of the uncapped RNA fragments, and searching for the protein-binding regions on the transcripts, were also mentioned. More importantly, we proposed a model for the biologists to deduce the contributions of transcriptional and/or post-transcriptional regulation to gene differential expression based on degradome-seq data. Finally, we hope that the degradome-based analytical methods could be widely applied for the studies on RNA biology in eukaryotes.

KEYWORDS:

• Plant RNA degradome
• RNA processing
• small RNA (sRNA)-guided target cleavages
• gene differential expression (GDE)
• transcriptional
• post-transcriptional

Introduction

RNA degradation is a constant reaction in the living cells. Although the total products from RNA decay are uniformly defined as RNA degradome, their biogenesis pathways differed greatly [1]. In addition to random degradation during transcript turnover, RNA fragments could be generated through the following three biological processes in plants. First, under the surveillance of the RNA quality-control system, those transcripts harbouring premature termination codons can enter into the nonsense-mediated decay pathway, resulting in massive instable fragments [2]. Second, maturation of the small RNAs (sRNAs) from their precursors will produce traceable intermediates in most cases. Third, an important portion of the RNA degradome was constituted by the cleavage remnants originated from sRNA-guided target slicing. From this point of view, the RNA degradome is a rich deposit of transient fragments containing abundant information of RNA-related events such as processing and regulation.

The advent of degradome sequencing (degradome-seq) technology, also called global mapping of uncapped and cleaved transcripts (GMUCT) [3] or parallel analysis of RNA ends (PARE) [4], has unprecedentedly advanced the researchers’ ability of deciphering the genetic and the regulatory codes from the degraded RNAs [5]. Our previous studies showed the power of degradome-seq data in high-throughput identification of microRNA (miRNA) – target pairs in plants, by a forward [6] or reverse [7] approach. Indeed, construction of the miRNA-mediated regulatory networks based on degradome-seq data has its superiority in reliability compared to computational predictions. On the other hand, the degradome-seq data could be utilized for tracking the processing signals from miRNA maturation [8,9], thus providing a novel approach for identification or re-examination of the miRNA genes. Importantly, other than the miRNAs, such kind of application might be extended to the sRNAs, such as small interfering RNAs (siRNAs).

In the following sections, the uses of degradome-seq in monitoring non-coding RNA (ncRNA) processing and identification of sRNA – target interactions will be summarized according to the recent research progresses. Besides, based on our recent analysis, three distinct distribution patterns of degradome signatures were observed on the differentially expressed sRNA targets. As a result, a plausible model was proposed to dissect the respective contributions of transcriptional and post-transcriptional regulation to gene differential expression (GDE).

Tracking ncRNA processing signals based on degradome sequencing data

MiRNA processing

According to the current protocol, only the polyadenylated RNA fragments will be included in degradome libraries [3–5]. The rationale of degradome-based analysis of miRNA processing comes from the fact that most of the miRNA genes are transcribed by RNA polymerase (Pol) II [10,11], and the two-step endonucleolytic processing of miRNA precursors will result in polyadenylated intermediates [8,9]. In this regard, the degradome-seq data could facilitate reliable identification or re-examination of the miRNA genes in both animals and plants [9,12]. Recently, our group developed two bioinformatics tools for miRNA discovery. One software package, miRNA Digger, starts from the available genomic sequences [13], while the other, PmiRDiscVali, starts from the transcriptome sequencing (RNA-seq) data [14]. Notably, both tools emphasize the importance of degradome-seq data in novel miRNA detection.

In plants, the Dicer-like 1-mediated two-step cropping usually starts from the bases of the hairpin-structured miRNA precursors, which is termed ‘base-to-loop’ processing. However, another processing mode, i.e. ‘loop-to-base’ cropping, has been observed for some exceptional miRNA families, such as MIR319 [15]. In our previous study, the novel utility of degradome-seq data in judging the processing modes (‘base-to-loop’ or ‘loop-to-base’) of the miRNA precursors was proposed [12].

Processing of other types of sRNAs

In both animals and plants, a portion of the long non-coding RNAs (lncRNAs) were suggested to be transcribed by RNA Pol II and possess polyadenylated tails [16,17]. Based on this fact, degradome-seq data might be used to detect the processing signals from certain lncRNAs. Indeed, supported by the degradome-based processing signals, numerous sRNA loci have been discovered on the lncRNAs in both Arabidopsis [18,19] and rice [20]. Accordingly, some related software packages, such as NATpipe [21], have been developed to perform transcriptome-wide search for sRNA loci based on their processing signals.

Deciphering the regulatory codes of gene expression

Transcriptome-wide identification of sRNA – target pairs

Degradome-seq was developed by combining modified 5ʹ rapid amplification of cDNA ends with the next-generation sequencing technology [3–5]. It is an efficient way to validate the miRNA- or siRNA-guided target cleavages in vivo [22–24]. To this end, several bioinformatics tools have been developed for degradome-based identification of the sRNA targets, such as CleaveLand4 [25], SeqTar [26], sPARTA [27] and PAREsnip2 [28]. Among these, PAREsnip2 was indicated to be one of the powerful toolkits with relatively high time and memory efficiency [28]. For several plant species, degradome-evidenced interactions between miRNAs or trans-acting small interfering RNAs (ta-siRNAs) and their targets have been made available in the public databases, such as DPMIND [29] and tasiRNAdb [30]. Interestingly, some novel sRNA species, such as tRNA-derived RNA fragments (tRFs) [31] and transposable element-associated miRNAs [32], were also shown to be capable of performing target cleavages based on degradome-seq data analyses. Besides, it was reported that miRNA-mediated regulation was occasionally observed in the nuclei of plant cells [33,34], which expanded the applications of degradome-seq in RNA biology. In other words, the degradome-seq data could be used to investigate miRNA-guided cleavages on the nascent nuclear transcripts in plants [35]. Taken together, the above studies greatly accelerated the research progress of sRNA-mediated gene regulation in plants.

A degradome-based model for gene differential expression (GDE)

GDE might result from transcriptional (e.g. transcription factors or epigenetic modifications), post-transcriptional (e.g. sRNA-guided target cleavages), or their joint regulatory effects. In our recent studies on the Dendrobium genus plants, the RNA-seq data sets from Dendrobium officinale (Dof), Dendrobium huoshanense (Dhu) and Dendrobium williamsonii (Dwi) were combinatorially used to assemble the reference transcriptome (NCBI SRA accession ID: PRJNA577972). Then, differentially expressed transcripts were identified based on pair-wise comparison. For example, the levels of the transcripts DN29273_c0_g3_i1, DN25995_c0_g3_i4 and DN33037_c2_g2_i2 were observed to be much higher in Dof than those in Dhu (the upper panels of Figure 1(a–c)). Interestingly, these transcripts exhibit quite distinct distribution patterns of the degradome signals (the lower panels of Figure 1(a–c); SRA accession ID: PRJNA577182). According to our previously proposed rules [7], cleavage signals were identified on DN29273_c0_g3_i1 and DN25995_c0_g3_i4 (indicated by the green arrows in the lower panels of Figure 1(a,b)). Furthermore, target site prediction by psRNATarget [36] indicated that the above identified slicing signals were potentially resulted from the sRNA-guided target cleavages (Figure 1(e,f)). Indeed, these sRNAs could be cloned by sRNA-seq (SRA accession ID: PRJNA576944).

Figure 1. A degradome-based model for gene differential expression (GDE) under transcriptional and/or post-transcriptional regulation. Based on the RNA-seq data, three transcripts (DN29273_c0_g3_i1, DN25995_c0_g3_i4 and DN33037_c2_g2_i2) are expressed much higher in Dendrobium officinale (Dof) than in Dendrobium huoshanense (Dhu). Referring to the y axes in the upper panels of (a), (b) and (c), the levels of these transcripts were measured by RPKM (reads per kilobase per million), which were calculated from three biological replicates of RNA-seq experiments (indicated by the x axes). The lower panels of (a), (b) and (c) show the degradome signatures mapped onto the three transcripts. The degradome signal intensities were measured by RPM (reads per million; see the y axes). According to our previously proposed rules [7], the degradome signals potentially resulting from small RNA (sRNA)-guided cleavages were identified and marked by green arrows. (d) By comparison between the degradome profiles and the expression patterns of the transcripts in the two Dendrobium species, the contributions of the transcriptional and the post-transcriptional regulation to the observed GDE were deduced, which were divided into three categories. Left panel: referring to (a), the higher level of DN29273_c0_g3_i1 in Dof might be attributed to the weaker signals of sRNA-guided cleavages. In this case, the post-transcriptional regulation might contribute greater than the transcriptional regulation to GDE. Middel panel: referring to (b), the cleavage signal intensities correlate well with the transcript levels in the two Dendrobium species. Hence, we deduced that the two types of regulation might contribute equally to GDE. Or, in addition to the transcriptional regulation, the post-transcriptional regulation could serve as a compensatory way for transcript level control. Right panel: referring to (c), GDE was observed for the transcript DN33037_c2_g2_i2 in the two Dendrobium species, whereas no obvious cleavage signal was detected in both plants. Thus, the observed GDE should not be resulted from the post-transcriptional regulation, at least not through a sRNA-mediated pathway. (e) Two sRNAs in Dof and two sRNAs in Dhu were predicted to have highly complementary binding sites on the transcript DN29273_c0_g3_i1. (f) MicroRNA172 was predicted to target DN25995_c0_g3_i4 in both Dof and Dhu. For both (e) and (f), the red dashed lines indicate the cleavage sites on the target transcripts, which were evidenced by the degradome signatures. The sequence information of the sRNAs and their binding sites are shown in each figure panels.

Next, the relationship between the cleavage intensities and the transcript levels was examined for each transcript of the two Dendrobium species. Specifically, the cleavage signals on the transcript DN29273_c0_g3_i1 were stronger in Dhu, and its expression level was relatively low in Dhu compared to Dof (Figure 1(a)). We proposed that as a post-transcriptional regulatory pathway, the sRNA-guided cleavages on this transcript might be one of the major contributors to the final output of gene expression (the left panel of Figure 1(d)). For DN25995_c0_g3_i4, the cleavage signal intensities correlate well with the expression levels of this transcript in the two Dendrobium species (Figure 1(b)), indicating that the sRNA regulatory intensities depend on the transcript abundances. In this regard, transcriptional regulation might have its essential role in gene expression, while sRNA-mediated regulation might act as a compensatory system for transcript level control (the middle panel of Figure 1(d)). Finally, weak degradome signals were observed to be randomly distributed on DN33037_c2_g2_i2, although the transcript was differentially expressed between Dof and Dhu (Figure 1(c)). From this point of view, we deduced that transcriptional regulation might be one of the dominant contributors to the observed GDE, and the sRNA-mediated post-transcriptional regulation was likely to be ineffective (the right panel of Figure 1(d)). Summarily, we proposed that the contribution of sRNA-guided target cleavages to GDE could be partially inferred from the degradome signatures.

Materials and methods

RNA sample preparation, library construction and sequencing

The stems of the three-year-old Dendrobium seedlings were collected for RNA sample preparation. Three biological replicates were included for each of the three Dendrobium species respectively. Total RNAs were extracted by using Trizol reagent (Invitrogen, CA, USA) following the manufacturer’s instruction. The total RNA quantity and purity were assayed by using Bioanalyzer 2100 and RNA 1000 Nano LabChip Kit (Agilent, CA, USA) with RIN value higher than 7.0.

For RNA-seq, poly(A)-tailed transcripts were enriched from 5 μg total RNAs by using poly(T) oligo-attached magnetic beads with two purification rounds. Then, the RNAs were fragmented by using divalent cations under elevated temperature. The RNA fragments were reverse-transcribed to create the final cDNA library according to the protocol of the mRNASeq sample preparation kit (Illumina, San Diego, USA). Paired-end sequencing was performed by using illumine Hiseq4000 (LC Sciences, USA).

For sRNA-seq, 1 μg total RNAs were used to construct a sequencing library based on the protocol of TruSeq Small RNA Sample Prep Kit (illumina, USA). The small RNAs were separated from the total RNAs by polyacrylamide gel electrophoresis, and were ligated to the 5ʹ and the 3ʹ adapters by T4 RNA ligase. By using superscript II reverse transcriptase (Promega, USA), the adapter-ligated RNAs were transcribed to single-stranded cDNAs, which were then used as the PCR templates to produce double-stranded cDNAs. The cDNA libraries were obtained by 16% TBE gel electrophoresis. Finally, sRNA-seq was performed by LC-Bio (Hangzhou, China) on illumina Hiseq2500.

For degradome-seq, 20 μg total RNAs were used to construct a degradome-seq library. At least 150 ng poly(A)-tailed RNAs enriched from the total RNAs were subject to annealing with the biotinylated random primers. Strapavidin capture, 5ʹ adaptor ligation to the RNA fragments containing 5ʹ-monophosphates, reverse transcription and PCR amplification were performed step-by-step for library construction. Finally, degradome-seq was performed by LC-Bio (Hangzhou, China) by using illumina Hiseq2500.

Sequencing data pre-treatment

For the RNA-seq data, Cutadapt (http://code.google.com/p/cutadapt/) and the in-house Perl scripts were used to remove those reads with adaptor contamination, low quality bases and undetermined bases. Then, the sequence quality was verified by using FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/).

For the sRNA-seq data, the raw reads were pre-treated by using Fastx-toolkit (http://hannonlab.cshl.edu/fastx_toolkit/), in order to remove the sequencing adapters and the low-quality reads. Reads longer than 35 nt and shorter than 15 nt were discarded.

For the degradome-seq data, the raw reads were also pre-treated by using Fastx-toolkit, in order to remove the sequencing adapters and the low-quality reads.

For both the sRNA-seq and the degradome-seq data, the expression level (RPM, reads per million) of one sequence in a data set was calculated by dividing the raw count of this sequence by the total raw counts of the sequences belonging to this data set, and then multiplied by 10⁶.

De novo assembly and differentially expressed gene analysis

De novo transcriptome assembly was performed with Trinity [37], the RNA-seq data sets (three biological replicates for each species) from Dendrobium officinale, Dendrobium huoshanense and Dendrobium williamsonii were combinatorially used to assemble the reference transcriptome.

Salmon [38] was used to calculate the expression level (RPKM, reads per kilobase per million) for each transcript [39]. The differentially expressed transcripts were selected with log2 (fold change) >1 or log2 (fold change) <-1, and with statistical significance (p value < 0.05) by the R package edgeR [40].

Concluding remarks and further perspectives

In the above sections, we focused on the applications of degradome-seq data in elucidating the mechanisms underlying sRNA maturation and sRNA-guided target gene regulation. However, recent studies also showed some novel uses of this kind of data in the RNA research area, such as in vivo mapping of the endoribonucleolytic cleavage sites [41], identification of conserved motifs at the 5ʹ ends of the uncapped RNA fragments [42], and searching for the regions associated with the stacked ribosomes (or other RNA-binding proteins) on the transcripts [5,43]. Besides, we noticed that nearly all of the degradome-seq data published with the previous reports was originated from polyadenylated RNA fragments. If a degradome library without poly(A) selection could be constructed, it will be intriguing to compare it with the poly(A)-selected one. Based on the analytical subtraction, it will be possible to detect the processing and the regulatory signals from the non-polyadenylated transcriptome. Although the cleavage targets of miRNAs have been rarely identified in animals [44], other animal sRNA species such as siRNAs and PIWI-interacting RNAs (piRNAs) could perform target cleavages [45,46], pointing to the value of degradome-seq data in validating sRNA – target interactions in animals. Additionally, some other novel applications of degradome-seq data, such as tracking the ncRNA processing signals [9], might also be extended to the animal RNA biology.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was funded by the National Natural Science Foundation of China [81730108], [81973635] and [31970637].