SMRT sequencing of the full-length transcriptome of the white-backed planthopper Sogatella furcifera

Introduction

The white-backed planthopper (WBPH) Sogatella furcifera (Hemiptera: Delphacidae) is an economically destructive rice pest distributed throughout Asia (Ma et al., 2017; Matsumura et al., 2014). As a migratory pest, S. furcifera has a great ability to spread, causing widespread damage to crops (Wang et al., 2012). S. furcifera mainly affects crops in two ways, by sucking phloem sap and causing growth delay (Zhou et al., 2013); in severe cases, it causes ‘hopper burn’ (Lei et al., 2014; Wang et al., 2017). Furthermore, S. furcifera transmits plant viruses such as southern rice black-streaked dwarf virus (SRBSDV), rice ragged stunt virus (RRSV), and rice grassy stunt virus (RGSV) (Chen et al., 2018; He et al., 2016; Zhou et al., 2013). Currently, this pest is controlled by extensive application of chemical insecticides (Jin et al., 2017; Matsumura et al., 2014; Yang et al., 2018), but long-term use of chemical agents can lead to the development of pest resistance, causing re-infestation (Matsumura et al., 2014; Zhou et al., 2013). In addition, the overuse of pesticides can lead to environmental pollution and safety issues. Therefore, a practical approach to pest control is urgent instead of using pesticides. For example, pest control can be carried out by regulating key processes such as growth and development, immune defenses, and reproduction in pests. The target genes regulating these processes should be identified and isolated to develop new approaches to pest control. Therefore, the acquisition of long-fragment gene sequences is necessary for studies on gene functions and to identify genes that can be targeted for pest control applications.

The transcriptome can reflect the number and types of genes in a cell, revealing biochemical processes at the physiological and cellular levels (Alba et al., 2005). The next generation of high-throughput sequencing, also known as second-generation sequencing, has enabled a better understanding of various gene expression and regulation mechanisms (Bao et al., 2013). This method does not require a reference genome, and therefore it is suitable for non-model species (Chao et al., 2018). In 2017, genome and transcriptome analyses of S. furcifera were performed using this method (Wang et al., 2017). De novo assembly and characterization of the S. furcifera transcriptome provided comprehensive data for gene function identification. The RNA-Seq data served as a basis for screening and analyzing candidate genes in S. furcifera for pest control. As a cogent tool for understanding gene expression and identifying splice junctions, second-generation sequencing has become a focus of research; however, it does not provide the full-length sequence of all genes (Djebali et al., 2012; Nagalakshmi et al., 2008; Zhang et al., 2018). A full-length transcript is defined as the entire transcript from the 5′ to 3′ end and containing the poly-A tail. However, owing to the read-length limitation in second-generation sequencing, the transcript obtained by this method is incomplete (Wang et al., 2017). Single-molecule real-time (SMRT) sequencing, a third-generation sequencing technology developed by Pacific Biosciences (PacBio), can effectively overcome this problem (Eid et al., 2009; Korlach et al., 2010). With its long read-length advantage, it can directly read reverse-transcribed full-length complementary DNA (cDNA) and provide high-quality, long, and intact transcripts (Sharon et al., 2013). The average read length is 10–15 kb and the maximum length is up to 70 kb (Stadermann, Weisshaar & Holtgrawe, 2015), which helps the accurate identification of isoforms that share exons (Faist et al., 2009). Furthermore, this method can identify new genes and complement genome annotations, and contribute to the accurate analysis of fusion genes, homologous genes, and superfamily genes or alleles.

PacBio SMRT sequencing has been used for whole-transcriptome profiling of several plants, animals, and microorganisms and numerous high-confidence transcripts have been obtained (Chao et al., 2018; Jia et al., 2018; Zeng et al., 2018). Although the genome of S. furcifera has been partially sequenced (Wang et al., 2017), information about the structure and sequence of transcripts is limited and the genes have not been well annotated. In the present study, the full-length transcriptome of S. furcifera was obtained by SMRT sequencing. Using the procured transcriptome data, we performed full-length sequence prediction, SSR analysis, lncRNA prediction, and gene function annotation. The transcriptome data provide numerous full-length sequences of S. furcifera, which will accelerate genome annotation and enhance our understanding of the genetic structure and diversity of S. furcifera.

Materials & Methods

Results

Non-redundant transcript acquisition

To obtain a series of non-redundant transcripts, we clustered highly similar coding sequences together using CD-HIT with an amino-acid sequence identity threshold of 99%. All 251,109 consensus transcripts were used for the acquisition of non-redundant transcripts, and finally, 156,138 unigenes (mean length of 2,994 bp) were obtained. The length distribution of the unigenes was counted at an interval of every 500–1,000 bp. As shown in Table 4, unigenes longer than 3,000 bp accounted for 46.6%, whereas those shorter than 500 bp only accounted for 8.9% of the total genes. As shown in Table 5, 92.5% of unigenes had only one isoform, while only 7.5% of unigenes had two to ten isoforms.

SSR detection

A total of 100,360 SSR sequences (5,833,286 bp) were obtained, with 74,401 SSRs and 25,959 SSR-containing sequences. The cumulative number of sequences with at least one SSR was 73,671 and the number of SSRs present in compound formation was 27,268. Moreover, SSRs consisting of one to six (mono-, di-, tri-, tetra-, penta-, and hexa-nucleotide) tandem repeats were identified. Mono-repeats (46,035; 62.98%) were the most abundant in the WBPH unigenes, followed by tri-repeats (21,834; 29.87%) and di-repeats (4,453; 6.09%). The frequencies of tetra-, penta-, and hexa-repeat types were only 0.86% (632), 0.14% (99), and 0.05% (39) in the WBPH unigenes, respectively (Fig. 2).

Table 3:

Statistics of CCS results.

(i) CCS: circular consensus sequence. (ii) 5′-primer: number of reads with 5′signals. (iii) 3′- primer: number of reads with 3′signals. (iv) Poly-A: number of reads with Poly-A tails. (v) Full length: full-length reads. (vi) FLNC: full-length non-chimeric reads. (vii) Average FLNC read length: average length of full-length non-chimeric reads. (viii) Consensus reads: number of reads of the non-redundant sequence obtained after clustering.

Sample	CCS	5′-primer	3′-primer	Poly-A	Full length	FLNC	Average FLNC read length	Consensus reads
S. furcifera	630,447	547,929	562,407	503,155	421,026	388,348	2,682	251,109

DOI: 10.7717/peerj.9320/table-3

Table 4:

Summary of the transcriptome assembly.

Percent of transcripts or unigenes (%): percentage of transcripts or unigenes in corresponding length range (bp).

Length range (bp)	Transcripts	Unigenes
<500	15253 (6.1%)	13966 (8.9%)
500–1000	4139 (1.7%)	2304 (1.5%)
1000–2000	62045 (24.7%)	31395 (20.1%)
2000–3000	75229 (30.0%)	35699 (22.9%)
>3000	94443 (37.5%)	72774 (46.6%)
Total	251109	156138

DOI: 10.7717/peerj.9320/table-4

Table 5:

Number of genes corresponding to the transcripts.

1/2/3/4/5/6/7/8/9/10: number of genes containing the same number of transcripts.

Isoform number	1	2	3	4	5	6	7	8	9	10
Unigene number	144,413	5,814	2,123	1,074	647	443	291	221	182	930

DOI: 10.7717/peerj.9320/table-5

TF detection

We analyzed TFs by comparing the transcript sequences to those in the AnimalTFDB 2.0 database, resulting in 1,288 TFs. The numbers of TFs enriched were as follows: zf-C2H2 (425), ZBTB (166), TF_bZIP (71), Homeobox (68), and HMG (60) (Fig. 3).

Prediction of coding sequences

We used verified protein sequences of this species or closely related species for ANGEL training and then ran the ANGEL prediction for the given sequences. Then, 121,395 coding sequences were generated; among them, 48,873 coding sequences containing the start and stop codons were defined as complete open reading frames (ORFs). Subsequently, the number, size, and length distributions of the 5′ and 3′ UTR regions were analyzed. In total, 8,225 transcripts were annotated to the 5′-UTR regions and 2,231 to the 3′-UTR regions. As shown in Fig. 4, CDS lengths ≤ 1,000 bp accounted for 85.5% (103,696), followed by those from 1,000–2,000 bp (14,625; 12.0%) and >2000 bp (3,074; 2.5%).

Identification of lncRNAs

We used four tools to identity unique transcripts without protein coding potential (i.e., lncRNAs). The CNCI tool identified 110,664 lncRNAs, Pfam identified 121,199 lncRNAs, Plek identified 51,434 lncRNAs, and CPC identified 98,911 lncRNAs. In total, 44,324 lncRNA transcripts were predicted by all four methods (Fig. 5).

Discussion

The SMRT sequencing platform produces long reads that can effectively resolve the issue of completed gene structures. The strengths of PacBio RNA sequencing have been widely investigated in various species (Allen et al., 2017; Chao et al., 2018; Jia et al., 2018; Park et al., 2017; Sharon et al., 2013; Tombacz et al., 2017; Zeng et al., 2018). To date, a draft genome sequence of S. furcifera has been published (Wang et al., 2017), but the sequences of most full-length transcripts are still unknown. We mapped SMRT sequencing transcripts to the S. furcifera reference genome (Wang et al., 2017) using GMAP (Wu & Watanabe, 2005). The percentage of mapped transcripts occupied 64.79% and unmapped occupied 35.21% (Table S1). As we mentioned, rice virus SRBSDV, RRSV and RGSV were transmitted by S. furcifera, so there were some of the transcript sequences appear to be of viral origin in our sequencing results.

One of the advantages of SMRT sequencing is that it provides new comprehension of full-length sequences, gene structures, and gene functions. The results of our study indicate that SMRT sequencing is useful for both genome annotation and gene function studies (Dong et al., 2015). The full-length transcriptome of S. furcifera was analyzed using PacBio SMRT sequencing and 15.64 Gbp of clean data was obtained, including 630,447 CCSs, of which 388,348 were identified as FLNC transcripts. In addition, 251,109 consensus transcripts were obtained using transcript cluster analysis of the FLNC reads, which included 29,700 high-quality transcripts. The lengths of the FLNC transcripts were aligned to the sizes of the transcriptome library (Fig. 1B). PacBio CCSs and FLNC reads did not require assembly of short next generation sequencing (NGS) reads. Furthermore, the high competence of PacBio SMRT sequencing to generate full-length transcript sequence may well be connected with its long-read property. In our study, 156,138 unigenes with mean length of 2,994 bp were obtained from the SMRT sequencing platform, which is much better than those assembled using only Illumina sequencing. For example, Li et al. (2016) obtained 51,842 unigenes with a mean length of 746 bp, and Xu et al. (2012) obtained 81,388 unigenes with a mean length of 555 bp.

lncRNAs are a class of non-protein-encoding transcripts of over 200 nt in length that are important to regulate gene expression at various levels. Currently, the roles of lncRNAs in insects, such as Drosophila melanogaster, Agasicles hygrophila, Plutella xylostella, and N. lugens, have been reported. Studies on D. melanogaster have shown that lncRNAs are involved in sex determination (Mulvey et al., 2014), motor behavior and climbing ability (Li et al., 2012), courtship behavior in males (Chen et al., 2011), sleep behavior (Soshnev et al., 2011), and inactivation of X staining (Deng & Meller, 2006; Smith, Allis & Lucchesi, 2001). In recent years, with increasingly intense research on lncRNA functions, the RNA-Seq data of some insects are being re-explored and several insect lncRNAs have been discovered. While studies on the function of lncRNA in fruit flies and honeybees have made crucial progress, in other insects, such as agricultural pests, the study of lncRNA function is impeded by several difficulties and challenges. In our study, 44,324 lncRNAs were predicted by all four methods, and their function in S. furcifera requires further investigation.

Full-length transcript sequence information is crucial for genome annotation and gene function research. However, most methods used to obtain a full-length cDNA clone are time consuming, expensive, and inefficient (Bower & Johnston, 2010; Chen, Wang & Wang, 2016; Schmidt & Mueller, 1999). To date, only a few full-length cDNA sequences have been reported in S. furcifera. Recently, Liang et al. (2018) used second-generation sequencing technology to sequence the transcriptome of S. furcifera and reported that the unigene length was 200–400 bp, accounting for 41.50% of the total number of genes, whereas those longer than 4,000 bp only accounted for 3.37%. Furthermore, 18,416 ORFs were generated based on the unigenes. In addition, the generated ORFs were compared with those in the Nr, Swiss-Prot, GO, COG, and KEGG databases, and 18,415 transcripts were annotated. In the present study, we used the PacBio SMRT sequencing platform to obtain 100,360 SSRs and 121,395 CDSs, of which 48,873 carried complete ORFs. Unigenes longer than 4,000 bp comprised 26.4% of the total number of genes, whereas those shorter than 500 bp accounted for 8.9% (Table 4). Therefore, our results show that SMRT sequencing is a useful and effective tool for acquiring reliable full-length transcripts of S. furcifera. Using SMRT sequencing, 95,495 transcripts were annotated, which will assist future research and the identification of gene functions in S. furcifera.

Conclusions

Through the PacBio SMRT sequencing platform, a large collection of full-length transcripts were obtained. The number and mean length of the unigenes, as well as the number of complete ORFs from SMRT sequencing were much better than those from Illumina sequencing. The obtained transcriptome data can assist further studies on gene function in S. furcifera and help clarify the interaction of S. furcifera in the ecosystem.

Supplemental Information