RNA sequencing and functional analysis implicate the regulatory role of long non-coding RNAs in tomato fruit ripening

1Although genome-wide transcriptome sequencing has revealed that ~90% of eukaryotic genomes are transcribed, only 1–2% of the genome encodes proteins, suggesting the presence of a large proportion of non-coding RNAs (ncRNAs). The ncRNAs are classified as housekeeping ncRNAs and regulatory ncRNAs. The ‘housekeeping’ ncRNAs include rRNAs, tRNAs, small nuclear RNAs (snRNAs), and small nucleolar RNAs (snoRNAs), whereas the ‘regulatory’ ncRNAs refer to small ncRNAs and long non-coding RNA (lncRNAs). Small ncRNAs, such as microRNAs (miRNAs) and small interfering RNAs (siRNAs), have been well studied in the last decade, as they are essential for post-transcriptional and transcriptional regulation in eukaryotes. In contrast, lncRNAs are less characterized so far. It is generally believed that lncRNAs are >200 nucleotides in length and mainly transcribed by RNA polymerase II (Pol II), which are always capped, polyadenylated and frequently spliced. A few lncRNAs are generated by plant-specific Pol V, capped at the 5′ end and lacking apparent poly(A) tails. These lncRNAs function as a scaffold for the RNA-directed DNA methylation (RdDM) pathway. In addition, Pol III is also involved in production of lncRNAs. With the development of next-generation sequencing, thousands of lncRNAs have been identified in model plants, such as Arabidopsis thaliana, Medicago truncatula, Triticum aestivum, Oryza sativa, and Zea mays. However, the function of lncRNAs has not been fully studied to date. Recently, the regulation mechanism of vernalization in Arabidopsis by COOLAIR (cool-assisted intronic non-coding RNA) and COLDAIR (cold-assisted intronic non-coding RNA), two species of lncRNAs transcribed from Flowering Locus C, has been illustrated. LncRNAs also function as endogenous target mimics (eTMs) for a few miRNAs, which provide a new mechanism for regulation of miRNA activity. LDMAR (long-day-specific male-fertility-associated RNA) plays an essential role in photoperiod-sensitive male sterility of rice.

2As part of the human diet, tomato has been domesticated for a few centuries. Today, tomato is the seventh most important crop species (after maize, rice, wheat, potatoes, soybeans, and cassava) and is the second most consumed vegetables in the world (after potatoes and before onions). Fleshy fruit ripening of tomato is a specific morphological trait that is not present in other model plants, which means that tomato has become a model organism for basic research and applied purposes, in particular as a model system for investigations into the regulation of fruit ripening. However, the number, expression pattern, and characteristics of lncRNAs in tomato are still largely unknown. Therefore, it is necessary and urgent to discover and identify novel lncRNAs and understand the function of lncRNAs in tomato fruit ripening.

3ripening inhibitor (rin) is one of the most famous tomato mutants that completely abolishes the normal ripening process and thus is widely used as an excellent genetic tool in the study of fruit ripening. RIN was clearly elucidated as a master regulator (transcription factor) of fruit ripening to control most major ripening-related processes. In this study, a comprehensive set of lncRNAs from wild-type and rin mutant tomato fruit was identified using paired-end strand-specific RNA sequencing (ssRNA-Seq). In total, 3679 putative lncRNAs were discovered and were found to be distributed in every tomato chromosome; 85.1% of lncRNAs were transcribed from intergenic regions. Tomato lncRNAs are shorter, and harbour fewer exons and less coding potential than the protein-coding genes. Compared with wild-type tomato, a lot of lncRNAs showed significantly differential expression in the rin mutant. Moreover, down-regulation of the expression of some novel intergenic lncRNAs (lncRNA1459 and lncRNA1840) in wild-type tomato fruit induced an obvious delay of fruit ripening. These results strongly suggested that lncRNAs play an important role in the regulation of tomato fruit ripening. The findings provide new insight into the study of fruit ripening.

4Wild-type AC (Solanum lycopersicum cv. Ailsa Craig) and rin (cv. Ailsa Craig, backcross parent) tomato were grown in the greenhouse under standard greenhouse conditions (26 °C under 16h lighting, followed by 8h darkness at 20 °C), with regular additions of fertilizer and supplementary lighting when required. To collect AC fruits, they were tagged at anthesis, and harvested at the immature green (IM), mature green (MG), breaker (BR), pink (PK), and red-ripe (RR) stages, which occurred at means of 37, 42, 46, 51, and 56 days post-anthesis (DPA), respectively. Fruits of the rin mutant were picked at the BR stage. Immediately upon harvesting, the pericarp was manually dissected, frozen in liquid nitrogen, and stored at –80 °C. Wild-type MicroTom (S. lycopersicum cv. MicroTom) were also planted for virus-induced gene silencing (VIGS) in tomato fruits.

5Total RNA was extracted from the fruits of AC and the rin mutant at the BR stage (two biological replicates per genotype combined from 10 fruits each) using DeTRNa reagent (EarthOx, CA, USA) according to the manufacturer’s protocol. The RNA concentration and purity were measured using an NAS-99 spectrophotometer (ATCGene, NJ, USA). The RNA integrity was checked by agarose gel electrophoresis. Genomic DNA was removed from extracted total RNA by DNase treatment. Due to some lncRNAs lacking the poly(A) tail, total RNA was treated to remove rRNA, retaining lncRNA both with and without a poly(A) tail. The quality of the RNA and lack of contaminating rRNA were confirmed using the Agilent 2100 Bioanalyzer. Four strand-specific RNA libraries with an insert size of ~250–500 nucleotides were prepared according to a UTP method, and submitted to the Beijing Genomics Institute (BGI, Shenzhen, China) for 100bp paired-end sequencing on the Illumina HiSeq 2000, at a depth of ~70 million reads per library (for statistics on read counts, see Table 1). The data for this article have been deposited in the National Center for Biotechnology Information (NCBI) Sequence Read Archive (http://www.ncbi.nlm.nih.gov/sra/) under accession number SRP04432.

6Sequencing reads were quality checked and trimmed to remove barcode and adaptor sequences. To rule out rRNAs, all reads were aligned to plant rRNA sequences by the Short Oligonucleotide Analysis Package (SOAP2; http://soap.genomics.org.cn/soapaligner.html). Information on plant rRNAs was extracted from the NCBI Non-Redundant (NR) data set (http://www.ncbi.nlm.nih.gov/). The clean reads from each library were aligned with the tomato reference genome (SGN release version SL2.50; ftp://ftp.sgn.cornell.edu/tomato_genome) using TopHat (version 2.0.8; http://ccb.jhu.edu/software/tophat/index.shtml). Only reads with no more than two mismatches were obtained and used to construct transcripts of each sample separately using Cufflinks (version 2.0.2; http://cole-trapnell-lab.github.io/cufflinks/) based on the tomato genome reference (SL2.50 genome).

7The assembled transcripts were compared with the tomato genome annotated protein sequences (SGN release version ITAG2.4; ftp://ftp.sgn.cornell.edu/tomato_genome) using BlastX. The non-redundant transcripts having significant alignment (P<1.0E-10, identity >90%, coverage >80%) with tomato proteins were excluded. For size selection, perl scripts were used to extract transcripts larger than 200 nucleotides. For the open reading frame (ORF) filter, a perl script was developed to pick up the transcripts that had short ORFs (<100 amino acids). Since a real lncRNA does not have an ORF, a putative ORF of the lncRNA candidate is defined by the longest consecutive codon chain of the lncRNA. Furthermore, to filter the transcripts containing a known protein domain, transcripts were aligned to the Protein database of KEGG (Kyoto Encyclopedia of Genes and Genomes), the NR data set, COGs (NCBI Phylogenetic classification of proteins encoded in complete genomes), and Swiss-Prot (Swiss-Protein database) using BlastX (P<1.0E-5, identity >90%, coverage >80%). Moreover, the resulting transcripts above were uploaded to the Coding Potential Calculator (CPC) to test the protein-coding potential. Only transcripts that did not pass the protein-coding score test were considered for the next step of the analysis. To rule out housekeeping ncRNAs (including tRNAs, snRNAs, and snoRNAs), all resulting transcripts were aligned to housekeeping ncRNA databases, including tRNA and snRNA sequences collected from the NCBI; and snoRNAs from the Plant snoRNAs Database (version 1.2; http://bioinf.scri.sari.ac.uk/cgi-bin/plant_snorna/home). LncRNA candidates that have significant alignment (P<1.0E-10, identity >90%, coverage >80%) with housekeeping lncRNAs were excluded from further analyses. To rule out miRNA precursors, putative lncRNAs were aligned against tomato miRNA precursors from miRBase (Version 21) and from the Tomato Functional Genomics Database (TFGD). LncRNAs that have significant alignment (P<1.0E-10, identity >90%, coverage >80%) with miRNA precursors were excluded. Finally, the remaining transcripts were considered as tomato lncRNAs, and are listed in Supplementary Table S1 available at JXB online.

8A diagram was generated to show the localization and abundance of lncRNAs and protein-coding genes in the tomato genome by the program Circos. Centromere locations were according to the report from the.

9According to the locations relative to the nearest protein-coding genes, the annotated lncRNAs was subdivided into four categories: (i) lncRNAs without any overlap with other protein-coding genes are classified as intergenic lncRNAs (lincRNAs); (ii) lncRNAs totally in the some protein-coding loci are classified as intragenic lncRNAs; (iii) lncRNAs with some overlap with genes on the same strand, are classified as overlap lncRNAs; and (iv) antisense lncRNAs overlapping with exons of a protein-coding transcript on the opposite strand. Perl scripts were developed to classify these four categories.

10In terms of transcript length and exon number, lncRNAs and protein-coding genes were analysed. Transcript length categories were <300, 300–400, 400–500, 500–600, 600–700, 700–800, 800–900, 900–1000, and >1000 nucleotides. Exon number categories were: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, and >10. The proportion of different kinds of lncRNAs and protein-coding transcripts was calculated.

11Tomato lncRNAs were predicted as miRNA targets using the psRNATarget and psRobot (H.J. ). The miRNA eTMs from tomato lncRNAs were predicted according to the rules of a previous study.

12Differentially expressed lncRNAs between AC and rin were identified using the cuffdiff program. The fold changes were calculated via log2 (FPKM rin/FPKM AC). LncRNAs exhibiting a |fold change| ≥1 and adjusted P-values <0.01 were selected as differentially expressed lncRNAs.

13Total RNA was isolated from fruit samples using DeTRNa reagent (EarthOx, CA, USA). The RNA concentration and purity were measured using a NAS-99 spectrophotometer (ATCGene, NJ, USA). The RNA integrity was checked by agarose gel electrophoresis. Genomic DNA was removed from extracted total RNA by DNase treatment. A 2 μg aliquot of total RNA was used for cDNA synthesis using a TransScript One-Step gDNA Removal and cDNA Synthesis SuperMix kit (Trans, Beijing, China) with random primer.

14qRT–PCR was performed using SYBR Green PCR Master Mix with a real-time PCR System CFX96 (Bio-Rad, CA, USA). qRT–PCR conditions were as follows: 95 °C for 10min, followed by 40 cycles of 95 °C for 15 s and 60 °C for 30 s. Fluorescence changes of SYBR Green were monitored automatically in each cycle, and the threshold cycle (Ct) over the background was calculated for each reaction. Samples were normalized using Actin, and the relative expression levels were measured using the 2−ΔΔCt analysis method. Three biological replicates were performed, and the reactions were performed in triplicate for each run. Student’s t-test was used to determine whether the qRT–PCR results were statistically different from two samples (*P<0.05; **P<0.01). Duncan’s multiple range test was used for three samples (P<0.01). Oligonucleotide primers used are listed in Supplementary Table S2 at JXB online.

15VIGS of MicroTom fruit was performed using Tobacco rattle vius (TRV) according to a previous study. lncRNA fragments and RIN fragments of 300–500bp were anlysised using the VIGS tool (http://solgenomics.net/tools/vigs) to avoid off-target silencing and then amplified from tomato cDNA with PCR. A pTRV2-lncRNA or RIN construct was generated by inserting the EcoRI-digested PCR fragment of lncRNA or RIN into the pTRV2 vector. Agrobacterium strain GV3101 containing pTRV1, pTRV2, and pTRV2-lncRNA vectors were grown at 28 °C in LB medium (pH 5.6) containing 10mM MES and 20 μM acetosyringone with kanamycin, gentamycin, and rifampicin antibiotics. After shaking for 12h, cultures were harvested and resuspended in infiltration buffer (10mM MgCl2, 200 μM acetosyringone, 5% sucrose) to a final OD600 of 1.0. Resuspensions of pTRV1 and pTRV2 or pTRV2-lncRNA were mixed at a ratio of 1:1 and left at room temperature for 3h. Agrobacterium was infiltrated into the carpopodium of fruits with a 1ml syringe. Tomato fruits infiltrated with pTRV1 and pTRV2 were used as controls. Each inoculation was carried out three times,and each time six different plants were infiltrated. When the VIGS phenotype was visible, different sections of tomato fruits were collected and stored at –80 °C.

16Poly(A)-enriched [poly(A)+] and poly(A)-depleted [poly(A)–] RNAs were isolated from total RNAs of tomato fruit using an Oligotex mRNA Mini Kit (Qiagen, CA, USA). For RNA digestion, total RNAs from tomato fruits were divided into each of four tubes and were treated as follows. First, the RNAs were incubated for 1h at 37 °C with or without enzymes: tube 1 and tube 2 with buffer only; tube 3 with T4 polynucleotide kinase (New England Biolabs, MA, USA); and tube 4 with 5′ pyrophosphohydrolase (New England Biolabs. After ethanol precipitation, RNAs in tube 1 were incubated for 1h at 37 °C with buffer only, whereas RNAs in the other three tubes were incubated with 5′ to 3′ exoribonuclease, XRN-1 (New England Biolabs), for 1h at 37 °C. The RNAs were extracted before being subjected to RT–PCR.

17cDNA synthesis was performed on total RNAs, poly(A)+, poly(A)–, and different RNA digestion fractions using a TransScript One-Step gDNA Removal and cDNA Synthesis SuperMix kit (Trans, Beijing, China) with random primers. PCR was performed using EasyTaq PCR SuperMix (Trans) with PCR system T-100 (Bio-Rad). PCR conditions for RIN, lncRNA1459, or lncRNA1840 were as follows: 94 °C for 3min, followed by 30 cycles of 94 °C for 20 s, 55 °C for 30 s, and 72 °C for 20 s. All PCR data presented are representative of three independent experiments. Oligonucleotide primers used are listed in Supplementary Table S1 at JXB online. RIN served as a positive control for poly(A)+ RNAs. PCR with genomic DNA was used as a positive control. Reverse transcription which is performed in the absence of reverse transcriptase served as a negative control.

18Paired-end ssRNA-Seq has become a powerful tool for the discovery of lncRNA, but also facilitates identification of transcript orientation. To identify lncRNAs in tomato fruits, paired-end ssRNA-Seq of transcripts from AC and rin fruits at the BR stage was performed in two biological replicates. A total of ~28 million clean reads was obtained (Table 1; Fig.1). A total of 38 159 unique transcripts were assembled from high-throughput RNA-Seq data from AC and rin fruits (Fig. 1).

19To distinguish lncRNA candidates, six sequential stringent filters to the 38,159 transcripts were employed (Fig. 1). First, these transcripts were filtered with tomato coding gene sequences. Almost 48% (18,407) of transcripts were coding genes, and the remaining 52% (19,752) of transcripts might be non-coding, which was consistent with other studies showing that ncRNAs were widely transcribed. It is generally believed that lncRNAs are longer than 200 nucleotides in size and might have a short ORF but not be able to encode polypeptides longer than 100 amino acids. This filter was then applied to the 19,752 transcripts, and 5430 transcripts were recovered (Fig. 1). The transcripts that might encode conserved protein domains were further filtered by comparing them with the four protein databases (KEGG, NR, COGs, and Swiss-Prot), and 4976 transcripts were obtained (Fig. 1). Next, the CPC was used to assess the protein-coding potential in order to eliminate 1252 possible coding transcripts (Fig. 1). After employing four stringent criteria, 3724 transcripts were considered as putative lncRNAs. Because housekeeping ncRNA (tRNAs, snRNAs, and snoRNAs) and miRNA precursors are two specific species of lncRNAs that function differently from other lncRNAs, the putative lncRNAs were next aligned to comprehensive sets of housekeeping ncRNAs and miRNA precursors sequences (for details, see the Materials and methods) to filter out 20 and 25 transcripts, respectively (Fig. 1). Thus, a total set of 3679 transcripts (3981 isoforms) were obtained and defined as tomato lncRNAs (Fig. 1).

20Next the lncRNAs were mapped onto the recently released tomato reference genome. A Circos plot clearly showed that tomato lncRNAs were not evenly distributed across chromosomes (Fig. 2A). Similar to protein-coding genes, lncRNAs have lower densities in the pericentromeric heterochromatin regions than in the euchromatin (Fig. 2A). This result suggested that lncRNAs may share similar features of transcription with the protein-coding genes. In addition, some lncRNAs were transcribed from loci much closer to the telomeres than protein-coding genes. For instance, some lncRNAs were generated from the ends of chromosomes #1 and 3 (Fig. 2A).

21According to the locations relative to the nearest protein-coding genes, lncRNAs were further classified into four types: intergenic, intragenic, overlap, and antisense lncRNAs (Fig. 2B). Whereas 5.2% and 8.6% of the lncRNAs either overlapped with genes or were transcribed from inside genes (most from introns), the majority of lncRNAs (85.1%) were located in intergenic regions (Fig. 2B). This observation is consistent with previous studies, further indicating that the type of lncRNA was termed as long intergenic non-coding RNA. In addition, only a small portion (1.1%) of lncRNAs are antisense of protein-coding genes. This result is unexpected as a previous study suggested that there were many antisense lncRNAs in Arabidopsis. Interestingly, the numbers of the four types of lncRNAs from Watson and Crick strands were similar (Fig. 2B).

22Previous studies have shown that both plant and animal lncRNAs are shorter and harbour fewer exons than protein-coding genes. To determine whether tomato lncRNAs share these features, the distribution of length and exon number of 3679 lncRNAs were analysed compared with all tomato predicted protein-coding transcripts (34,726 genes from the SL2.50 genome). Figure 3A shows that ~78% of lncRNAs ranged in size from 200 to 1000 nucleotides, with only 22% >1000 nucleotides. In contrast, for the protein-coding transcripts, ~50% of them were >1000 nucleotides. Interestingly, most (90%) of the genes encoding tomato lncRNAs only contained one or two exons, while the number of exons for the protein-coding genes ranged from one to ≥10 (Fig. 3B). These results together indicated that, unlike protein-coding genes, the majority of the tomato lncRNAs are relatively shorter and contain fewer exons.

23Interplay between miRNAs and lncRNAs is one of the important functional patterns seen for lncRNAs. LncRNAs could be targeted by miRNAs and could also function as eTMs of miRNAs. To examine whether lncRNAs are bona fide targets for miRNAs, all of the 3679 lncRNAs we checked using psRNATarget and psRobot. Only three miRNA targets were identified (Fig. 4A). The recovery of a small number of miRNA targets probably resulted from the low expression level of lncRNAs that were not detected by RNA-Seq. Of these three miRNA targets, lncRNA504 was the target of sly-miR6024 that was involved in plant immunity (F. ). LncRNA3613 was the target of sly-miR5304 that has been only identified in solanaceous plants. In addition, lncRNA3294 was the target of sly-miR169 that is engaged in drought tolerance of tomato.

24Although eTMs are suggested to be widespread in Arabidopsis, here only three eTMs from lncRNAs were predicted (Fig. 4B). LncRNA246 was the eTM of sly-miR399, a miRNA that plays an important role in regulating phosphate homeostasis (. LncRNA1127 is the eTM of sly-miR390-3p, which accumulates more than sly-miR390-5p. Interestingly, in the case of sly-miR6024, it not only targeted lncRNA504 but also was targeted by the eTM of lncRNA3444 (Fig. 4).

25Because the rin mutant showed a strong non-ripening phenotype compared with AC, it was hypothesized that there might be some novel ripening-related lncRNAs present in rin. Bioinformatics analysis revealed that 3530 of the 3679 tomato lncRNAs were accumulated in both AC and rin (Fig. 5A). Only 23 and 126 lncRNAs were expressed specifically in AC or rin, respectively (Fig. 5A). To identify further ripening-related lncRNAs, the levels of lncRNAs were compared between AC and rin. A total of 677 lncRNAs were significantly differentially expressed between AC and rin. Compared with AC, 490 of 677 lncRNAs were up-regulated in rin, and the other 187 lncRNAs were down-regulated (Fig. 5B). To investigate whether these differentially expressed lncRNAs are engaged in fruit ripening, 10 of them were arbitrarily selected, five from a highly up-regulated group and five from a down-regulated group. The differences in their expression levels observed by RNA-Seq were experimentally validated by qRT–PCR (Fig. 6). In addition, the fold change in the lncRNA expression level of qRT–PCR and RNA-Seq was closely correlated (R 2=0.76, P<0.001) (Supplementary Fig. S1 at JXB online). These results indicated that these lncRNAs were indeed ripening-related lncRNAs in tomato fruits, further suggesting that these lncRNAs are likely to play some roles in fruit ripening.

26Compared with AC fruits, accumulation of intergenic lncRNA1459 and lncRNA1840 was lower in the rin mutant than in AC (Fig. 6A). It was hypothesized that lncRNA1459 and lncRNA1840 may regulate the ripening process of tomato fruits. To test this hypothesis, VIGS was performed to silence lncRNA1459 and lncRNA1840 in Micro-Tom fruits. VIGS of RIN was used as a positive control. Intriguingly, 2 or 3 weeks after infiltration, compared with TRV control fruits (already ripening) (Fig. 7A, D), tomato fruits injected with TRV-RIN, TRV-lncRNA1459, or TRV-lncRNA1840 showed partial ripening, with a ripening section (red) and a non-ripening section (green or yellow) (Fig. 7B–D, E–G). Semi-quantitative PCR analysis suggested that a recombinant virus could spread from carpopodiums to fruits (Supplementary Fig. S2 at JXB online), and then induce the VIGS of lncRNAs or RIN in tomato fruits. For positive VIGS control, the expression of RIN was decreased to 20% in the yellow section of tomato fruits (Fig. 7H). Compared with TRV control fruits, the transcript level of lncRNA1459 in the green sections of TRV-lncRNA1459 fruits was dramatically decreased by 65% (Fig. 7I). On the other hand, the expression of lncRNA1459 in the red sections of TRV-lncRNA1459 fruits was comparable with that in TRV control fruits. Similarly, the level of lncRNA1840 in yellow sections of TRV-lncRNA1840 fruits decreased to 18% compared with TRV control fruits (Fig. 7J). Therefore, the non-ripening phenotype of TRV-lncRNA1459 or TRV-lncRNA1840 fruits clearly resulted from the silencing of lncRNA1459 or lncRNA1840. The result strongly suggested that the two novel lncRNAs were involved in the regulation of tomato fruit ripening and might play an essential role in fruit ripening.

27For further characterization of lncRNA1840 and lncRNA1459, the expression pattern of thse lncRNAs during fruit ripening was first explored. The transcripts of lncRNA1840 accumulated to a high level at the IM stage, and then the expression levels dropped to the lowest at the MG stage. As the fruit became red, lncRNA1840 increased rapidly up to the RR stage (Fig. 8A). However, expression of lncRNA1459 increased during fruit ripening, peaking at the PK stage (Fig. 8B). The transcriptional patterns of both lncRNA1840 and lncRNA1459 indicated that they were ripening related.

28Most lncRNAs are transcribed by Pol II, and polyadenylation is part of the maturation process of Pol II-dependent transcripts. In order to determine whether lncRNA1840 or lncRNA1459 has a poly(A) tail, total RNAs of tomato fruits were separated into poly(A)+ and poly(A)– fractions followed by the detection of lncRNAs by RT–PCR. As shown in Fig. 8C, RIN was detected only from total and poly(A)+ RNA, which is a positive control of poly(A)+ RNA. lncRNA1840 and lncRNA1459 were also detectable from total and poly(A)+ RNAs, but not from poly(A)– RNAs, suggesting these that two lncRNAs had poly(A) tails and might be Pol II-dependent transcripts.

29Normally, 5′ initiating nucleotides of Pol II transcripts have 7-methylguanosine caps. To determine the 5′ end structure of lncRNA1840 and lncRNA1459, various enzymatic treatments of total RNAs were performed. Total RNAs were treated with 5′ pyrophosphohydrolase (RppH), which decaps RNA and converts the 5′ 7-methylguanylate cap or 5′ triphosphate to 5′ monophosphate, or T4 polynucleotide kinase (PNK), which adds a 5′ phosphate group to 5′ hydroxyl RNAs. The RNAs were then digested with XRN-1, a 5′ to 3′ exoribonuclease that acts on RNAs only with a 5′ monophosphate group. Next, random primed RT–PCR was conducted on these treated RNA samples to detect lncRNA1840 or lncRNA1459. The RNA samples treated with PNK and XRN-1 or XRN-1 alone showed a transcript level of lncRNA1840 and lncRNA1459 comparable with RNAs with no treatment, suggesting that lncRNA1840 and lncRNA1459 did not have a 5′ monophosphate group (Fig. 8D). Furthermore, dramatic reduction in the abundance of lncRNA1840 and lncRNA1459 was only found in RNAs treated with RppH followed by XRN-1 (Fig. 8D), and lncRNA1840 and lncRNA1459 were polyadenylated (Fig. 8C), together indicating that lncRNA1840 and lncRNA1459 might have a 5′ 7-methylguanylate cap, but not a 5′ triphosphate group.

30Recent studies revealed that lncRNAs exert a crucial role in various biological processes of plants. Although many lncRNAs have been identified from model plants, such as Arabidopsis, wheat, rice, and maize, much work still remains to be done with tomato. In the present study, a total of 3679 lncRNA loci (3981 isoforms) were identified in tomato, a model plant for study of fruit ripening (Fig. 1). Although the strict criteria pipeline for identification of tomato lncRNAs is similar to that used in previous studies in plants, there are several advantages to the present list of lncRNAs. (i) The lncRNAs should include some tomato lncRNAs that lack polyadenylation as the RNA-Seq data were obtained from RNAs that were depleted only of rRNA, but not of non-polyadenylated RNAs. In line with this, many lncRNAs without poly(A) tails have recently been identified from poly(A)– RNA-Seq in Arabidopsis. (ii) Another ubiquitous limitation for recent lncRNA studies is that strand information from RNA-Seq data was missing. Here, the strand-specific RNA-Seq allowed easy identification of the transcription orientation of lncRNAs, thus providing a useful resource for further functional analysis.

31There are also some limitations to the list of tomato lncRNAs obtained. (i) Due to the inherent limitations of using 100bp paired-end RNA-Seq, it is difficult to obtain the complete sequences for all predicted tomato lncRNAs. (ii) Normally, 1–2% of the genome encodes proteins; however, 48% of the assembled transcripts had significant homology with tomato protein-coding genes from the first filter of strict criteria (Fig. 1). These results indicated that the RNA-Seq was not deep enough to recover tomato lncRNAs fully. (iii) Some previously characterized lncRNAs have the potential to encode peptides longer than 100 amino acids, such as HOTAIR (HOX transcript antisense RNA), XIST (X-inactive specific transcript), and KCNQ1OT1 (KCNQ1 overlapping transcript 1). However these type of lncRNAs could not meet the relatively strict criteria used here for definition as lncRNAs and thus were probably filtered out in the screening.

32In summary, although some tomato lncRNAs might be excluded due to the sequencing limitations and strict bioinformatics criteria, a relatively robust and reliable list of tomato lncRNAs is provided. The list of lncRNAs will probably be very useful for other researchers.

33Transcriptomic sequencing on different varieties of tomato revealed the presence of a large number of ncRNAs. Previous studies have already shown that small RNAs are involved in the regulation of tomato fruit. Differential accumulation of small RNAs during tomato fruit ripening indicated that the regulation mechanism of small RNAz for fruit ripening would be complicated. Furthermore, degradome sequencing of tomato fruit revealed that a number of miRNA targets were genes which were previoulsy already characterized as important factors in tomato fruit ripening. For example, the target of miR156/157 is Colorless Non-Ripening (CNR), an epigenetic mutation of which could inhibit tomato fruit ripening. The target of miR172 is APETALA2 (AP2), which is a negative regulator of tomato fruit ripening. In contrast, ripening-related lncRNAs have not been as comprehensively identified and functionally examined in tomato. The present RNA-Seq data and further qRT–PCR analysis revealed that many lncRNAs were significantly differentially expressed in the rin mutant compared with AC (Figs 5, 6). The result clearly suggested that lncRNAs might be involved in the regulation of tomato fruit ripening.

34Because the transcription machinery is not perfect, there are large numbers of spurious RNAs that might be by-products. Recently many studies discovered thousands of lncRNAs from many plant species; however, they did not suggest that they are functional because ‘lncRNA’ is only a name for transcripts. The current dogma is that a few lncRNAs are functional and most are not; therefore, functional identification and assignment would be important for these lncRNAs. Due to the numbers of ripening-related lncRNAs that were identified in the present study, a rapid and high-throughput method would be necessary for their functional characterization.

35VIGS is a widely used tool to identify gene function in tomato fruit development and ripening because it is an easy, rapid, reliable, and transformation-free method. In addition, partial sequence information of one transcript is sufficient to silence itself by VIGS. Here it was demonstrated that VIGS is a powerful tool to study lncRNAs too. VIGS of lncRNA1459 and lncRNA1840 in tomato fruits resulted in a delay of fruit ripening (Fig. 7), taken together with the expression patterns of these two lncRNAs during fruit ripening (Fig. 8), provides strong evidence that lncRNA1459 and 1840 are functional in fruit ripening regulation.

36Capping RNA is critical for RNA interactions with many nuclear and cytoplasmic proteins, and plays essential roles in RNA stability, splicing, nucleocytoplasmic transport, and translation initiation to regulate RNA accumulation. Also, polyadenylation of RNA is important for nuclear export, translation, and stability of RNAs. End structure analysis of lncRNA1840 and lncRNA1459 indicated that they might be Pol II-dependent transcripts with 5′ caps and 3′ poly(A) tails (Fig. 8C, D), which is very helpful for further functional characterization of these two lncRNAs in the future. Thus this study sheds new light on the regulation of fruit ripening, which might trigger more comprehensive studies on tomato lncRNAs. It would be necessary to investigate further the functional motifs and target genes of lncRNAs in tomato, which would help to elucidate fully the regulatory mechanisms of lncRNAs on fruit ripening.

37Supplementary data are available at JXB online.

38Figure S1. Expression levels as determined by RNA-Seq and qRT–PCR are highly correlated.

39Figure S2. Semi-quantitative PCR detection of recombinant TRV RNA in uninjected tomato fruits with infiltrated carpopodiums.

40Table S1. List of lncRNAs of tomato.

41Table S2. Primers used in the study.