Analysis of the global transcriptome of longan (Dimocarpus longan Lour.) embryogenic callus using Illumina paired-end sequencing

Background Longan is a tropical/subtropical fruit tree of great economic importance in Southeast Asia. Progress in understanding molecular mechanisms of longan embryogenesis, which is the primary influence on fruit quality and yield, is slowed by lack of transcriptomic and genomic information. Illumina second generation sequencing, which is suitable for generating enormous numbers of transcript sequences that can be used for functional genomic analysis of longan. Results In this study, a longan embryogenic callus (EC) cDNA library was sequenced using an Illumina HiSeq 2000 system. A total of 64,876,258 clean reads comprising 5.84 Gb of nucleotides were assembled into 68,925 unigenes of 448-bp mean length, with unigenes ≥1000 bp accounting for 8.26% of the total. Using BLASTx, 40,634 unigenes were found to have significant similarity with accessions in Nr and Swiss- Prot databases. Of these, 38,845 unigenes were assigned to 43 GO sub-categories and 17,118 unigenes were classified into 25 COG sub-groups. In addition, 17,306 unigenes mapped to 199 KEGG pathways, with the categories of Metabolic pathways, Plant-pathogen interaction, Biosynthesis of secondary metabolites, and Genetic information processing being well represented. Analyses of unigenes ≥1000 bp revealed 328 embryogenesis-related unigenes as well as numerous unigenes expressed in EC associated with functions of reproductive growth, such as flowering, gametophytogenesis, and fertility, and vegetative growth, such as root and shoot growth. Furthermore, 23 unigenes related to embryogenesis and reproductive and vegetative growth were validated by quantitative real time PCR (qPCR) in samples from different stages of longan somatic embryogenesis (SE); their differentially expressions in the various embryogenic cultures indicated their possible roles in longan SE. Conclusions The quantity and variety of expressed EC genes identified in this study is sufficient to serve as a global transcriptome dataset for longan EC and to provide more molecular resources for longan functional genomics.


Background
Longan (Dimocarpus longan Lour.), a tropical/subtropical fruit tree in the family Sapindaceae, is of great economic importance in Southeast Asia. Because the status of embryo development determines seed size, fruit quality, percentage of fruit set, and yield in longan, efforts to improve fruit quality and yield have included studies on regulation of longan embryo development using cytological, molecular, and proteomics approaches [1,2]. Such research has been hampered, however, by the extremely high genetic heterozygosity of longan and early embryo sampling difficulties [3]. Because plant somatic embryogenesis (SE) shows close similarities on morphological and molecular levels to normal zygotic embryogeny [4][5][6][7], the longan SE system has been used as a system for investigating regulation of in vitro and in vivo embryogenesis in longan [8][9][10]. Studies focusing on molecular biology and proteomics of the longan SE system have been conducted using differential display reverse transcription PCR (DDRT-PCR), homology cloning, quantitative realtime PCR (qPCR), two-dimensional electrophoresis, and protein bio-mass spectrometry (MALDI-TOF, Q-TOF), resulting in the isolation and identification of hundreds of related genes and proteins [1].
But little genomic or proteomic information of the above-mentioned studies is available for the longan embryo. As of July 2013, only 652 nucleotide sequences and 66 expressed sequence tags (ESTs) had been deposited in the NCBI GenBank database. Although many key longan genes and proteins have been cloned and identified, molecular resouces of longan are still limited because genomic and transcriptomic information is lacking. Consequently, an accelerated effort to acquire transcriptomes of longan embryogenesis is needed. A few transcriptomic studies of embryogenesis have been conducted in rice [11,12], poplar [13,14], Arabidopsis [15,16], Gossypium hirsutum [17], Solanum tuberosum [18], Elaeis guineensis [19], Brassica napus [20], soybean [21], and maize [22]; these studies were mainly focused on calli or embryogenic calli, and involved techniques such as Illumina sequencing, massively parallel signature sequencing (MPSS), EST analysis, microarray analysis, and suppression subtraction hybridization (SSH). No research has been performed on the longan transcriptome.
To assist in the identification, quantification, and classification of genes expressed in longan embryogenic callus (EC), we generated a global transcriptome from longan EC using high-throughput Illumina RNA sequencing, and analyzed functions, classification, and metabolic pathways of the resulting unigenes using bioinformatics. We then comparatively analyzed expression patterns to reveal 23 selected unigenes participating in longan SE. The resulting assembled and annotated transcriptome should serve as a highly useful resource for the identification of genes involved in longan SE.

Results
Illumina sequencing, de novo assembly, and sequence analysis of the D. longan transcriptome To obtain a global overview of the longan EC transcriptome, we constructed a cDNA library from a longan EC RNA sample. Using an Illumina HiSeq 2000 sequencing system, 64,876,258 clean reads (comprising 5.84 Gb of nucleotide data) were obtained after removing low-quality reads and adaptor sequences. Q20, N, and GC percentages were 95.88%, 0.01%, and 45.54%, respectively (Table 1).
Using the SOAPdenovo assembly program, all highquality reads were assembled into 491,067 contigs longer than 75 bp, with a median length of 138 bp and an N50 of 98 bp. The size distribution of these contigs is shown in Additional file 1. The length of 380,516 contigs (77.49%) ranged from 75 to 100 bp; 15,556 contigs (3.17%) were longer than 500 bp, and the remaining were mainly between 200-499 bp in length.

Protein coding region (CDS) prediction of the D. longan transcriptome
To determine the function of longan embryogenic unigenes, BLASTx alignment (E-value ≤ 1 × 10 -5 ) between unigenes and Nr, Swiss-Prot, KEGG and COG protein databases was carried out, and the results were used to predict unigene transcriptional orientations and coding regions. A total of 41,644 unigenes (20,999 in sense and 20,645 in antisense orientations) were identified in the longan EC library, with 27,281 unigenes remaining unidentified.
For validation and annotation of gene names, CDS, and predicted proteins, all assembled unigenes were first searched against Nr and Swiss-Prot databases using BLASTx. In total, 55.94% (38,555) of the putative protein unigenes showed significant similarity to known plant proteins in the databases. The distribution of unigenes with homologous matches was 200-500 bp (27,881; 72.31%), 600-1000 bp (7,101; 18.42%), 1100-3000 bp (3,466; 8.99%), and >3000 bp (107; 0.28%). Furthermore, With respect to plant growth and developmental functions, analysis of unigenes longer than 1000 bp (most including the entire ORF) showed that at least 328 unigenes of embryogenesis-related genes were expressed in longan EC. Among them, pentatricopeptide repeat-containing protein genes (253 unigenes) were the most dominant group, followed by EMB (Embryo defective) family genes (42 unigenes), and then MEE (Maternal effect embryo arrest) family genes (7 unigenes). Surprisingly, in addition to the embryogenesis-related genes mentioned above, many reproductive growth-related genes were also expressed in EC, including genes related to flowering, meiosis, floral organ development, female and male gametophyte development, embryo sac development, ovule development, endosperm development, pollen tube growth, inflorescence meristem growth, floral organ number control, petal loss, and tapetum formation. Furthermore, some vegetative growth-related genes, such as those related to apical meristem growth, root growth, and mycorrhizal formation, were also expressed in EC (Table 2).

GO functional annotation and classification of the D. longan transcriptome
To functionally categorize D.longan expressed genes, Gene Ontology (GO) terms were assigned to assembled unigenes. Based on BLASTx hits against the Nr database, Blast2GO [23] and WEGO [24] were used to obtain GO annotations and classifications according to molecular function, biological process, and cellular component ontologies.
Based on Nr annotations, 38,845 unigenes were assigned to the three main GO categories and 43 subcategories, which included cellular process, metabolic process, death, development process, cell, organelle, antioxidant activity, catalytic activity, binding, enzyme regulator activity, transcription regulator activity, and translation regulator activity ( Figure 1). Of the three main GO categories, cellular component was the most dominant category (17,417; 44.8%), followed by biological process (11,609; 29.89%) and molecular function (9,819; 25.28%) ( Figure 1).

COG functional annotation and classification of the D. longan transcriptome
The Clusters of Orthologous Groups (COG) database is based on a set of coding proteins with complete genomes and information about systematic evolutionary relationships of bacteria, algae, and eukaryotes. All longan unigenes were searched against the COG database to predict and classify by possible function. Overall, 17,118 (24.84%) unigenes were assigned to 25 COG categories ( Figure 2).
Taken together, the annotated longan unigenes provided valuable information for investigating specific processes, functions, and pathways involved in longan EC development, and allowed identification of novel genes in non-model organisms.
Based on the analyzed qPCR data, all selected unigenes were expressed at varying levels in different embryogenic tissues ( Figure 8). MRH2, NPGR1, PPR1, REF6, NPG1, SWP2, and VLN1 were expressed throughout the different tissue culture developmental stages, although no significantly expression profiles were observed. Expression levels of EMB2, EMB3, FRP, SPK1, VLN2, AAH, and SWP1 were low in TE, and high in HE and CE, while BEL1-LIKE exhibited the highest expression in TE. GHMP2, PPR2, and NPGR2 were highly expressed in ICpEC, GHMP1, EMB1, and SWA1 showed strong expression in HE, moderate expression in EC, and weak expression in ICpEC. EDA7 and SWA2 were highly expressed in EC. These results confirm that differential expressions of these unigenes have potential roles during longan SE.

Discussion
Feasibility of Illumina paired-end sequencing and assembly for non-model species with unsequenced genomes such as longan Understanding the dynamics of plant transcriptomes is helpful for studying the complexity of transcriptional regulation and its impact on phenotype [25]. Transcriptome sequencing is one of the most important tools for gene discovery and expression pattern identification, but traditional EST sequencing based on the Sanger method is time-consuming and costly. Because of their high throughput, accuracy, and low cost, next-generation sequencing technologies, such as Illumina/Solexa, 454, and MPSS, have been used successfully for plant genomic and transcriptomic analyses in many organisms [26][27][28]. In this study, approximately 64 million clean reads (5.84 Gb of nucleotides) were obtained from longan EC using Illumina HiSeq 2000 sequencing and assembled into 68,925 unigenes, more than that reported for plants such as S.indicum [29], T.chinensis [30], C.sinensis [31], G. hirsutum [32] and I.batatas [33] using the same technology. Compared with previous studies, these sequences produced shorter unigenes (mean = 448 bp) than those assembled from Sesamum (629 bp), Taxus (1077 bp), I.batatas (581 bp), Poncirus trifoliata (1000 bp; from MPSS) and Jatropha curcas (916 bp; from 454) [34], but longer than those generated from C.sinensis [31], Fagopyrum (341 bp; 454) [35], and maize (218 bp; 454) [36]. More importantly, 5,696 (8.26%) of the assembled unigenes were longer than 1000 bp. These results demonstrate that Illumina sequencing technology can be an effective tool for gene discovery in non-model organisms. Moreover, 58.95% (40,634) of the putative protein unigenes showed significant similarity to known plant proteins in databases, a higher percentage than that reported for S.indicum (54.03%) [29], I.batatas (46.21%) [33], and Epimedium sagittatum (38.50%) [37]. On the other hand, average unigene length in our study was shorter than that obtained for most plant species, and there were many unassembled reads; difficulties with the de novo transcriptome assembly may be attributed to various factors, such as short sequence fragments, assembly options, genes expressed at low levels, repetitive sequences, alternative splicing, and lack of a reference genome [33].
Studies have shown that plant SE is morphologically and molecularly similar to zygotic embryogenesis [4][5][6][7]. Just as a plant zygotic embryo, the most complex plant organ, epitomizes the entire plant, so too a plant somatic embryo can be considered to represent an entire plant. In our study, we uncovered 68,925 unigenes, the majority of which reflected expression of genes required for plant in vitro embryogenesis, such as Pentatricopeptide repeat proteins (PPR) and Embryo defective (EMB) family genes. A previous study has shown that mutations of different PPRs have distinct impacts on embryo morphogenesis [38]. In our study, 253 PPR unigenes longer than 1000 bp were identified, and two highly abundent PPR unigenes were further confirmed and found to be expressed throughout longan SE. PPR1_Unigene 68247 was highly expressed in ICpEC and HE, while PPR2_Unigene 68600 mRNA was abundant in EC and ICpEC. These results suggest the involvement of these genes during early developmental stages of longan SE. In Arabidopsis, 250 EMB genes have been confirmed to be required for normal embryo development [39], with EMB175 displaying aberrant cell organization and undergoing morphological arrest before the globular-heart transition [38]. In our study, three EMB unigenes longer than 1000 bp-EMB1_Unigene68678, EMB2_Unigene68326, and EMB3_Unigene1123-were verified and strongly expressed in HE and CE. In addition, EMB1_Unigene68678 was also highly expressed in GE and moderately detectable in EC and TE, while EMB2_Unigene68326 and EMB3_Unigene1123 exhibited moderate expression in GE and TE. High expression levels of these selected EMBs in longan HE and CE indicate their possible roles in the development of longan SE.
Surprisingly, however, there were also many unigenes expressed in EC associated with reproductive growth characteristics (such as flowering, gametophytogenesis, and fertility), and vegetative growth (such as root and shoot growth). In our study, 13 reproductive growthrelated genes longer than 1000 bp were confirmed by qPCR during the developmental stages of longan SE. For example, the fertility-related FRP (fringe-related protein), was strongly expressed in HE and CE, and also weakly in TE. Genes related to pollen growth-NPGR (no pollen germination), NPGR (no pollen germination-related), and VLN (Vilin-like)-displayed ubiquitous but weak expression during longan SE. NPGR2_Unigene68058 was highly expressed in ICpEC, and VLN2_Unigene67205 was accumulated in HE and CE, but barely detectable in TE. SWA1/2, essential for gametogenesis in Arabidopsis [40,41], were also differentially expressed during longan SE; while they were both highly expressed in EC, while SWA1_Unigene68185 was also expressed in HE. The GHMP kinase enzyme family, a primary determinant of sexual fate in Caenorhabditis elegans [42], was also detected in our study. GHMP1_Unigene10997 expression was high in HE, and GHMP2_Unigene67027 transcripts accumulated in ICpEC. BEL1-LIKE, is required for cytokinin and auxin signaling during ovule development in Arabidopsis [43], was expressed highly in TE and moderately in HE and CE, but was barely detectable in ICpEC; this suggests it may play a major role in longan SE during late embryonic stages. EDA7, related to embryo sac development, was strongly expressed in EC, TE, and CE. These results all demonstrate that these 13 reproductive growthrelated genes also play roles during longan SE development. Five vegetative growth-related genes were also chosen for qPCR analysis across the six sequential developmental stages of longan SE. These genes included MRH2 (morphogenesis of root hair 2), which is likely involved in polarized growth of root hairs in Arabidopsis [44], REF6 (relative of early flowering 6), which plays divergent roles in the regulation of Arabidopsis flowering [45], and SWP (STRUWWE LPETER), which plays an important role in defining the duration of cell proliferation [46]. All of these genes were expressed at varied levels in different embryogenic tissues, suggesting their wide involvement in various developmental stages during longan SE. In particular, SWP1_Unigene68809 was highly expressed during late stages of longan SE, but was barely detectable in GE. In addition, SPK1 (SPIKE1), required for normal cell shape control and tissue development [47], and AAH (allantoate amidohydrolase), related to nitrogen fixation, were both strongly expressed in HE and CE. The number and variety of expressed genes in EC suggests that EC practically reflects the entire SE profile, and even that of the entire plant. In addition, EC might be considered as a "gene pool" for isolating various plant target genes, such as genes related to flowering, pollen development, root and shoot growth, and plant-pathogen interactions. This EC dataset provides new candidates with possible roles in longan somatic embryogenesis.
In Arabidopsis, the largest number of unannotated signatures was found in callus: 1,655 (6.7%), compared with 884 in inflorescences, 935 in leaves, 1,089 in roots, and 907 in siliques [15]. In our study, we also found many unigenes (704; 4.11%) with unknown function, demonstrating how little is known about the biology of undifferentiated plant cells. Thus, the EC transcriptome can be used to more effectively discover new genes in longan.
Using Solexa sequencing, 27% of identified unigenes in Populus euphratica callus were found to be differentially expressed in response to salt stress; these genes were mainly involved in transport, transcription, cellular communication, and metabolism [14]. During Arabidopsis callus development, 241 genes were found to be upregulated and 373 to be down-regulated. The most highly up-regulated genes encoded an unknown protein (At3g60420) and acireductone dioxygenase (At2g26400), and the most highly down-regulated genes included a DR4 protease inhibitor (At1g73330), two peroxidase genes (At5g17820 and At5g666390), two pEARLI 1 genes (At4g12480 and At4g12470), and two that encoded subtilases (At5g59090 and At5g44530) [16]. In rice callus cells, 16,000 expressed genes were identified using the microarray suite (MAS) 5.0 detection algorithm [48]. These studies demonstrate that a large number of genes involved in various biological and metabolic pathways are expressed in plant EC. In our study, 17,306 (25.11%) unigenes were assigned to 199 KEGG pathways, including Metabolic pathways, Plantpathogen interactions, Biosynthesis of secondary metabolites, and Photosynthesis and Genetic information processing-related pathways. These results lay a foundation for further identification of longan EC-related genes.

Conclusions
In summary, our study generated the first large-scale transcriptome dataset of longan EC. In addition, the types and quantities of genes expressed in longan, as well as their functions, classification, and metabolic pathways, were revealed for the first time. Twenty-three unigenes related to embryogenesis and reproductive and vegetative growth were differentially expressed in various embryogenic cultures, indicating their possible roles in

Longan EC cDNA library construction and Illumina sequencing
For Illumina sequencing, Poly(A) + RNA was isolated from longan EC total RNA using Dynal oligo(dT) 25 beads according to the manufacturer's instructions. Following purification, fragmentation buffer was added to cleave the mRNA into short fragments. First-strand cDNA was synthesized using these short fragments as templates, along with SuperScript III reverse transcriptase and N6 random hexamer primer. Second-strand cDNA was then synthesized using buffer, dNTPs, RNaseH and DNA polymerase I. The resulting double-stranded cDNA was subjected to end-repair using T4 DNA polymerase, DNA polymerase I Klenow fragment, and T4 polynucleotide kinase, and ligated to adapters using T4 DNA ligase. Adaptor-ligated fragments (200 ± 25 bp long) were purified using a QiaQuick PCR extraction kit and eluted with EB buffer. After analysis using agarose gel electrophoresis, suitable fragments were selected as templates for PCR amplification. Sequencing of the resulting longan EC cDNA library was carried out with an Illumina HiSeq 2000 system.

Data filtering and de novo assembly
Following sequencing of the EC cDNA library, deconvolution and quality value calculations were performed on the resulting raw images. Before assembly, high-quality clean reads were generated from the raw reads by removing adapter sequences, duplicated sequences, low-quality reads with ambiguous bases ('N'), and reads with more than 10% of Q-values < 20 bases. All subsequent analyses were based on clean reads. Transcriptome de novo assembly was performed using SOAPdenovo v1.03 (http:// soap. genomics.org.cn). First, high-quality clean reads with a certain length of overlap were combined by SOAPdenovo into longer fragments with no unknown sequences ('N') between them. Contigs were then joined into scaffolds using paired-end information. Finally, paired-end reads were used again for gap filling of scaffolds to obtain sequences with the smallest number of Ns and which could not be extended on either end. The resulting sequences were defined as unigenes. The entire set of reads used for final assembly was submitted to the NCBI Sequence Read Archive under the accession n°SRA050205. BLASTx alignment with a cut-off E-value of 1 × 10 -5 was performed between unigenes and Nr, Swiss-Prot, KEGG, and COG databases, and all plant proteins in the databases were taken into consideration in the search for homology. The best results from the alignment were used to predict unigene coding regions and direction (i.e., 5' to 3'). When results from different databases conflicted, a priority order of Nr > Swiss-Prot > KEGG > COG was followed. Unigenes without homologs in any of the above databases were scanned using ESTScan [51] to predict coding regions and sequence direction.

Gene annotation, classification, and metabolic pathway analysis
To assign putative functions to longan unigenes, various bioinformatics approaches were used for further annotation, classification, and metabolic pathway analysis. First, the unigenes were aligned to Nr and Swiss-Prot protein databases using BLASTx (E-value < 1 × 10 -5 ), retrieving protein functional annotations. To better understand the annotation and distribution of the longan gene functions, Blast2GO [23] was used in conjunction with the Nr annotations to retrieve GO annotations of longan unigenes. WEGO software [24] was then used for GO functional classification of all longan unigenes according to molecular function, biological process, and cellular component ontologies. To predict and classify possible functions, longan unigenes were also compared against the COG database. The biological interpretation of longan unigenes based on the KEGG database was further extended by assigning them to metabolic pathways using BLASTx.

Gene validation and expression analysis by real-time quantitative PCR
Twenty-three unigenes with potential roles in longan SE were chosen for validation using real-time quantitative PCR (qPCR) with gene specific primers designed using DNAMAN 6.0. Relative mRNA levels from each unigene in RNA isolated as described above from six D.longan tissue samples (EC, IcpEC, GE,HE,TE and CE) were quantified with respect to internal standards [2]. All reactions were performed in triplicate in a LightCycler 480 qPCR instrument (Roche Applied Science, Switzerland), with a dissociation curve used to control for primer dimers in the reactions. Abundances of the 23 unigenes were calculated relative to the expression of reference genes DlFSD1a, EF-1a, and eIF-4a. Gene names, primer sequences, product sizes, PCR efficiencies, and annealing temperatures are given in Additional file 3.
Additional file 3: The selected gene names, primer sequences, product sizes, PCR efficiencies, and annealing temperatures.