Comprehensive Transcriptome Profiling Reveals Long Noncoding RNA Expression and Alternative Splicing Regulation during Fruit Development and Ripening in Kiwifruit (Actinidia chinensis)

Genomic and transcriptomic data on kiwifruit (Actinidia chinensis) in public databases are very limited despite its nutritional and economic value. Previously, we have constructed and sequenced nine fruit RNA-Seq libraries of A. chinensis “Hongyang” at immature, mature, and postharvest ripening stages of fruit development, and generated over 66.2 million paired-end and 24.4 million single-end reads. From this dataset, here we have identified 7051 long noncoding RNAs (lncRNAs), 29,327 alternative splicing (AS) events and 2980 novel protein-coding genes that were not annotated in the draft genome of “Hongyang.” AS events were demonstrated in genes involved in the synthesis of nutritional metabolites in fruit, such as ascorbic acids, carotenoids, anthocyanins, and chlorophylls, and also in genes in the ethylene signaling pathway, which plays an indispensable role in fruit ripening. Additionally, transcriptome profiles and the contents of sugars, organic and main amino acids were compared between immature, mature, and postharvest ripening stages in kiwifruits. A total of 5931 differentially expressed genes were identified, including those associated with the metabolism of sugar, organic acid, and main amino acids. The data generated in this study provide a foundation for further studies of fruit development and ripening in kiwifruit, and identify candidate genes and regulatory elements that could serve as targets for improving important agronomic traits through marker assisted breeding and biotechnology.


INTRODUCTION
The genus Actinidia, commonly known as kiwifruit, is widely distributed throughout most of East Asia, and consists of 57 species of climbing plants (Ferguson and Huang, 2007). This genus is considered native to China, since most of the species occur in the southwest of the country. Over the past three decades, kiwifruit has become an economically important fruit crop due to its nutritional quality and unique flavor, and its cultivation worldwide has substantially increased (Ferguson and Huang, 2007;Zhang L. et al., 2010). The annual production of kiwifruit was 3.26 million metric tons in 2013 (http://faostat3.fao.org/). Despite its importance and increasing production, the international kiwifruit industry relies mainly on a few naturally selected cultivars derived from two intraspecific taxa, A. chinensis and A. deliciosa (Chat et al., 2004;Zhang L. et al., 2010).
The flesh of the majority of kiwifruit cultivars is either green or yellow at harvest (Huang and Ferguson, 2003;Wang et al., 2003;Montefiori et al., 2005;Crowhurst et al., 2008). The first commercial red-flesh cultivar is A. chinensis "Hongyang." It is characterized by a medium fruit size (average weight per fruit is 77.6 g), a total acid content of 0.49%, and a soluble solid concentration of 19.6%. Importantly, "Hongyang" also has a high content of vitamin C (136 mg 100 g −1 FW) and anthocyanin (2.99 mg 100 g −1 FW; Wang et al., 2003;Montefiori et al., 2005). Kiwifruit contains sugars such as glucose, fructose, and sucrose, and organic acids such as citric, quinic, malic, and ascorbic acids, and main amino acids such as glutamine, arginine, and aspartate (Redgwell and MacRae, 1992;Capitani et al., 2010). The combination of sugars, organic acids and free amino acids represents the major factor contributing to kiwifruit flavor (Sorrequieta et al., 2010). In general, the total sugar content in "Hongyang" is higher than that of other A. chinensis cultivars; while the concentration of the three main organic acids (citric acid, quinic acid and malic acid) is similar (Wang et al., 2003;Nishiyama et al., 2008). This sugar/acid ratio is believed to contribute to the excellent flavor characteristics for "Hongyang" fruit. Carotenoids, chlorophylls, and anthocyanins in kiwifruit represent forms of dietary antioxidants. Anthocyanins are the main pigments in "Hongyang" fruit inner pericarp, chlorophylls and lutein are the main pigments in outer pericarp, all of them contributing to its overall appearance and attractiveness (Montefiori et al., 2005;Nishiyama et al., 2008).
Sweetness is one of the most important quality traits for kiwifruit cultivation and breeding. Sugars (sucrose, monosaccharides, and polyols) are important molecules in plants and function as a source of energy, building blocks for cell walls, and as osmotic and regulatory molecules (Smeekens et al., 2010). Enzymes associated with sugar metabolism include ADP-glucose pyrophosphorylase (AGPase), sucrose phosphate synthase (SPS), invertase (INV), amylase, sucrose synthase (SUS), fructokinase (FK), and hexokinase (HK; Deluc et al., 2007). Sugar transporters are essential proteins for the transport and allocation of sugars from source to sink cells (Kühn and Grof, 2010). Many sucrose transporters have been characterized in Arabidopsis, grape, and rice (Sauer and Stolz, 1994;Davies et al., 1999). Despite the progress made in identifying genes encoding sugar transporters, little is known about the transcriptional regulation of these genes. While some of the genes involved in sugar metabolism in kiwifruit have been identified (Nardozza et al., 2013), further studies on the regulation of sugar metabolism, especially in fruit tissue, are needed to develop approaches to regulate and improve fruit quality.
Genetic studies and variety breeding in kiwifruit are complex and time-consuming due to its high level of heterozygosity, and long juvenility period. High-throughput sequencing has now become a powerful tool for studying the transcriptome of species with and without sequenced genomes (Wang et al., 2013;Chen et al., 2014;Wu et al., 2014). Crowhurst et al. (2008) generated a collection of 132,577 expressed sequence tags (ESTs) in four Actinidia species (A. chinensis, A. deliciosa, A. arguta, and A. eriantha). Recent efforts using next-generation sequencing data have produced a draft genome sequence of a heterozygous kiwifruit "Hongyang." In addition, an Illumina HiSeq 2000 sequencing platform has been used to generate transcriptomic data from three stages of fruit development in order to facilitate gene prediction and annotation . Recently, the transcript profiles of kiwifruit were constructed and analyzed by Li et al. (2015), with a focus on the secondary metabolism including phytohormones, sugars, starch, anthocyanin, and Lascorbic acid.
At the RNA level, alternative splicing (AS) of pre-mRNAs represents a major mechanism by which the complexity of the transcriptome and proteome is increased (Modrek and Lee, 2002;McGuire et al., 2008;Bartlett et al., 2009;Tang et al., 2013), while long noncoding RNAs (lncRNAs) constitute a crucial regulatory module in diverse gene-silencing pathways (Bardou et al., 2014;Liu et al., 2015). Li et al. (2015) analyzed exon number, transcripts sizes, start sites of the novel transcripts and listed the extend 5 ′ or 3 ′ sites of alternative splicing events in different developmental stages of kiwifruit. However, the sequence and annotation of novel transcripts and other splicing events (exon skipping and intro retention) were not listed. In the present study, by employing sequence data derived from the nine different RNA-Seq libraries, we identified 7051 lncRNAs, 29,327 alternative splicing events and 2980 novel genes that were not annotated in the draft genome of "Hongyang" . Differential gene expression was also characterized between fruits at 20, 120, and 127 (7 days postharvest) days after pollination (DAP) to gain further insight into the genetic regulation of fruit development. Both up-and down-regulated genes were identified at each stage of fruit development. Since sugars are such an essential aspect of fruit flavor in kiwifruit, qRT-PCR was used to specifically study the expression level of genes putatively associated with sugar metabolism.

Plant Material
A. chinensis "Hongyang" was grown in the experimental station of Sichuan Academy of Natural Resources, Sichuan Province, China. Fruit samples were collected from five 5-year-old plant, at 20 days after pollination (DAP; beginning of cell division), 120 DAP (onset of fruit mature, Brix 6.5-7.5), and 127 DAP (onset of postharvest ripening, Brix 9-10). The sampled tissues were immediately frozen in liquid nitrogen and stored at −80 • C for RNA-seq analysis. Two to three biological replicates were collected at each sampling. The physical or physiological parameters of tissues/organs of kiwifruit "Hongyang" were described in Huang et al. (2013).

RNA Extraction, Transcriptome, and Gene Expression Profile Sequencing
Total fruit RNA was isolated using Trizol reagent, treated with DNase I and further purified with RNA clean kit (Promega, USA). RNA quality and quantity were checked with an Agilent 2100 Bioanalyzer RNA Nanochip (Agilent, Santa Clara, CA) and NanoDrop ND-1000 Spectrophotometer (NanoDrop, Wilmington, DE), respectively. RNA samples were pooled from equal amounts of RNA from five independent individuals. The samples were sent to Beijing Genomics Institute-Shenzhen (BGI, Shenzhen, China) for RNA-Seq library construction and sequencing using an Illumina HiSeq 2000 sequencing platform following manufacture's protocols (Illumina Inc, USA). Raw RNA-Seq reads have been deposited in the NCBI sequence read archive (SRA) under the accession number SRA065642 . In addition, the templates of RT-PCR or qRT-PCR were synthesized by using HiFiScript Reverse Transcription kit CW2582 (CWBIO, China), with oligo dT and random primers.

RNA-Seq Read Processing, Assembly, and Transcript Construction
The adaptor and low quality bases were trimmed from raw sequencing reads using Trimmomatic (Bolger et al., 2014), and trimmed reads shorter than 40 bp were discarded. The resulting high quality reads were aligned to a ribosome RNA database (Quast et al., 2013) using Bowtie (Langmead et al., 2009) allowing up to three mismatches, and mapped reads were discarded in the subsequent analyses. The remaining cleaned reads were aligned to kiwifruit genome sequences using Tophat2 ) allowing two mismatches for paired-end reads, and one mismatch for single-end reads. Only reads with perfect matches to the genome were used for reference-guided de novo assemblies using Cufflinks (Trapnell et al., 2010). The assembled transcripts from three different fruit developmental stages were merged together with a kiwifruit gene model using Cuffmerge, which was provided in the Cufflinks package.

Identification of lncRNAs, Novel Genes, and as Events
The assembled transcripts were translated into proteins using ESTScan (Iseli et al., 1999), and the longest protein for each transcript was kept and compared against the Arabidopsis thaliana protein and UniProt (TrEMBL and SwissProt) databases using the BLAST program with an E-value cutoff 1e-4. The blast results were used to assess the coding potential of each assembled transcript using Coding Potential Calculator (CPC; Kong et al., 2007).
To identify lncRNAs, transcripts originally obtained from the kiwifruit predicted gene models or those shorter than 200 bp were first discarded. The remaining transcripts with a CPC score < 0 and an ORF length < 300 bp were identified as lncRNAs. Novel protein-coding genes were identified using the following criteria: (1) they had to be located in intergenic regions and their distance to the closest predicted gene models should be >500 bp, (2) their ORF length should be longer than 300 bp, and (3) their CPC scores should be >0.
AS events were identified from the assembled transcripts using ASTALAVISTA (Foissac and Sammeth, 2007). Different categories of AS events were identified and counted using an in house Perl script (Sammeth et al., 2008).

Differential Gene Expression
The number of clean reads that mapped to each kiwifruit gene model was calculated, and then normalized into fragments per kb exon model per million mapped fragments (FPKM). To identify differentially expressed genes during the fruit development, raw counts of RNA-seq expression data were first transformed using the get Variance Stabilized Data function in the DESeq package (Anders and Huber, 2010). The variance-stabilizing transformed expression data were then fed to the LIMMA package (Smyth, 2004), and F-tests were performed. Raw P-values were adjusted for multiple testing using the Benjamini-Hochberg procedure (Benjamini and Hochberg, 1995). Genes with a ratio between the maximum and minimum expression levels ≥2 and an adjusted P < 0.01 were identified as differentially expressed genes during fruit development.

RT-PCR and qRT-PCR
Primers were designed to amplify lncRNAs, AS events and new genes. PrimeSTART HS DNA polymerase (Takara, China) was used for lncRNAs and the amplification of new genes. The realtime reverse transcription, quantitative PCR (qRT-PCR) was carried out in a total volume of 20 µl, containing 10 µl of SoFast EvaGreen (Bio-Rad, USA), 0.4 µM of each primer, 6 µl of 1:50 diluted cDNA and 3.2 µl ddH 2 O. Thermal cycling consisted of a hold at 95 • C for 30 s, followed by 40 cycles of 95 • C for 5 s, and 60 • C for 20 s. After amplification, samples were kept at 95 • C for 15 s and 60 • C for 1 min. The temperature was then gradually raised, by 0.5 • C every 10 s, to perform a melt-curve analysis. Each sample was amplified in triplicate, and all PCR reactions were performed on the StepOne Real-time PCR System (AB applied Biosystem, USA). The Ct method was employed with Actin (Achn107181) and 18S rRNA (NCBI Accession: AB253775) as endogenous controls . Primers used for RT-PCR and qRT-PCR were listed (Table S1).

Amino Acids, Organic Acids, and Sugars Analysis
Kiwifruit samples at three development stages (20, 120, and 127 DAP) were used for the analysis of amino acids, organic acids, and sugars. Each stage contains three biological replicates. The detection of amino acids was according to Javelle et al. (2003). Organic acids were prepared from three stages samples each containing five fruits. Fruits were put into a homogenizer for homogenizing, and then 40 mL ddH 2 O be added to 10 g of pulp at 4 • C for 4 h. Twenty microliter filtering supernatant was used for liquid chromatograph (Model LC-6AD; Shimadzu, Tokyo, Japan) analysis. HPLC separations were performed using a Diamonsil C18 reverse-phase column (250 mm × 4.6 mm × 5 µm) purchased from USA. The mobile phases were used 10 mm H 2 SO 4 (pH = 2.6), at a flow rate of 0.5 mL/min. The column temperature was set at 30 • C. Sugars were detected by HPLC-ELSD, using an carbohydrate analysis column (3.9 × 300 mm, 10 µm) at 25 • C. The mobile phase was composed of water and acetonitrile (23:77) with a flow rate of 1 mL/min (Nishiyama et al., 2008). Standard of Glucose, fructose, sucrose, citric acid, malic acid were purchased from Sigma (St. Louis, MO, USA).

RESULTS AND DISCUSSION
Sequencing and Overview of the RNA-Seq Dataset RNA-Seq libraries were constructed from three independent biological replicates of whole fruit tissues collected at 20, 120, and 127 DAP. A total of 81.1 million and 25.5 million reads were generated from paired-end (PE) and single-end (SE) libraries, respectively, using an Illumina HiSeq 2000 sequencing platform. Removal of low quality, adaptor sequence, and rRNA contaminated reads resulted in a final number of 66.2 million PE and 24.4 million SE reads, respectively, among which ∼87.5% PE reads and ∼90.6% SE reads were mapped to the kiwifruit genome (Table S2). Among the mapped reads, ∼95% of PE and 89% of SE reads were uniquely aligned.
Approximately 79.1% PE, and 73.7% SE reads were mapped to the kiwifruit genome with perfect matches. These mapped reads were used for reference-guided de novo assembly and assembled into 56,313, 64,892, and 57,656 transcript isoforms in fruits at 20, 120, and 127 DAP, respectively. This corresponded to 49,766, 56,456, and 48,176 gene models, respectively. All of the assembled isoforms, together with the full length transcripts from kiwifruit draft genome , were merged to remove redundancies. This resulted in 54,425 gene models and 105,632 isoforms, among which 39,040 gene models were from kiwifruit predicted. There were 12,185 isoforms that were not expressed (FPKM ≈ 0) at any of the three fruit developmental stages. The majority of them (11,255; 92.4%) were from kiwifruit predicted gene models. The remaining 930 trace-expressed isoforms were not included in any further analyses. In summary, a comprehensive set of 104,702 transcripts, corresponding to 54,422 gene models, was obtained from the three stages of kiwifruit development.

Kiwifruit lncRNAs
A total of 7051 potential lncRNAs were identified in the comprehensive set of kiwifruit transcripts, among which 1511 had CPC scores between 0 and −1 and thus scored as "weak noncoding" while the other 5540 had CPC scores < −1 and thus scored as "strong noncoding" ( Table S3A). The lncRNAs were placed into different groups based on the anatomical properties of their gene loci (Rinn and Chang, 2012). A subset of 6009 lncRNAs located in intergenic regions, with a distance >500 bp to the closest kiwifruit genes, were classified as intergenic large intervening noncoding RNAs (lincRNAs). Another 597 lncRNAs were categorized as overlapping lncRNAs since they were located in protein-coding gene regions. A total of 169 lncRNAs were categorized as antisense lncRNAs since they had more than a 50 bp overlap with their corresponding sense transcript. Finally, 881 were classified as intronic lncRNAs as they resided completely within an intron in protein-coding genes ( Table 1). Importantly, this category system was not applicable to all the identified lncRNAs. A total of 347 lncRNAs were put into an "other group" since they did not confirm to any of the above categories. Additionally, in some cases, the same lncRNAs could be placed in different categories. As shown in the Venn diagram in Figure 1A, most of the antisense lncRNAs (117 out of 169) were also categorized as overlapping lncRNAs since they overlapped with kiwifruit protein-coding genes. Approximately 92% of the intronic lncRNA were also located within intergenic regions. Detailed information on each of the categorized lncRNAs including their genome location, transcript length, and expression profile during fruit development, are provided (Tables S3B-F).
The intronic and antisense lncRNAs appeared as pairs with their corresponding coding or noncoding transcripts. A total of 1736 and 319 of the intronic and antisense lncRNAs existed as transcript pairs, respectively (Tables S3G-I). A majority (94.7%) of the sense-antisense pairs consisted of one noncoding and one coding transcript, indicating a potential predominant cisregulating role for these antisense lncRNAs ( Table S3H). The  other 17 were noncoding-noncoding pairs and further studies are needed to discern their function (Table S3I). Seven lncRNAs were randomly selected to verification by RT-PCR and Sanger sequencing, and six of them were found to be clearly expressed in the three fruit stages ( Figure S2A). LncRNAs are thought to have a wide range of functions in the regulation of gene expression in higher plants (Bardou et al., 2014;Liu et al., 2015). Interestingly, three intronic lncRNAs were identified that overlapped with one neutral invertase (Achn178991) and two sugar transporter (Achn017471 and Achn319221) genes involved in sugar and organic acid metabolism (Table S3G).

Novel Protein-Coding Genes
A total of 2980 novel potential protein-coding genes, containing a total of 4813 transcript isoforms, were identified in this study of kiwifruit (Table S4A). These protein-coding genes tended to have more exons and longer sequences than the identified lncRNAs. Nearly 84% (2498) of these novel genes were transcribed with multi-exons and more than 67% had primary transcript sequences longer than 1000 bp, and CDS sequences of ∼30% of these novel genes were longer than 1000 bp (Figure 2A). Approximately 89% of the transcripts could be functionally annotated, indicating that the sequences obtained for these novel protein-coding genes were of high quality. Four of the new genes, MYB domain protein (Myb17), Flavanone 3-hydroxylase (F3H), F-box family protein (F-box), and SQUAMOSA promoter-binding-like (SPL), which were randomly selected, were expressed in all three of the examined fruit stages. Sanger sequencing of the expressed genes were the same as the predicted sequences ( Figure S2B). Interestingly, genes coding for ascorbic acid biosynthetic enzymes, including L-ascorbate oxidase (AO) and pectinesterase (PME), as well as glutamyl-tRNA reductase (GluTR) genes related to chlorophyll synthesis, and genes coding for anthocyanin biosynthesis, including chalcone isomerase (CHI), flavanone 3-hydroxylase (F3H) were identified among the novel genes (Table S4B).

Alternative Splicing (AS) Events
By removing 11,255 non-expressed isoforms from the kiwifruit predicted genes, a total of 29,327 AS events were identified from transcript isoforms that were all expressed during kiwifruit fruit development ( Table 3). These AS events occurred in 11,868 intron-containing genes, and represented 40,109 different transcripts (Table S5A). These data indicate that 26% of the expressed gene loci (36.7% of intron-containing gene loci), or 42.9% of expressed transcripts, were subject to alternative splicing. These results are comparable to the level of alternative splicing reported in other plant species (Filichkin et al., 2010;Zhang G. et al., 2010;Marquez et al., 2012;Shen et al., 2014; Table 3). The result is consistent with the previous study (Li et al., 2015). While, the total number of exon skipping category (5960) is more than Li et al. reported (2015). Interestingly, in contrast to the AS categories reported in other plant species (Reddy et al., 2013), the intron retention category was not the predominant AS event type in kiwifruit. Identification of AS events was confirmed using individual RNA-Seq data at various stages in different biological replicates. These data confirmed that intron retention was still not the predominant category of AS (data not shown). The percentage of different AS types, especially intron retention events, is likely highly correlated with intron length (Reddy, 2007). Most introns in kiwifruit were ∼100 bp in length ( Figure 2B). However, the average intron lengths in predicted kiwifruit genes and transcript isoforms were 1188 and 1082 bp, respectively, which were much larger than the average intron length in Arabidopsis (∼170 bp) and rice (∼430 bp; Reddy, 2007). Therefore, the lower percentage of intron retention events in kiwifruit may be due to the longer intron length. Interestingly, dramatic changes in the number of AS events in different fruit developmental stages were also observed. In general, the number of AS events during fruit development increased from 20 to 127 DAP. At 20 DAP and 120 DAP, intron retention was the second most abundant type of AS event, representing 17.5 and 21% of the total AS events, respectively. Although the percentage of intron retention events was decreased slightly from 21% at 120 DAP to 17.4% at 127 DAP, the percentage of exon skipping events increased greatly from 11.8 to 24.5%, which was accompanied by a concomitant decrease in the percentage of alternative 3 ′ acceptor and alternative 5 ′ donor AS events (  (0.4%), respectively ( Table 3). These numbers and percentages are consistent with those found in other plant species (Reddy et al., 2013). Five of the AS events, Auxin-response factor gene Achn271111 (TCONS_00051294, TCONS_00051295, TCONS_000512947); b:UDP-glycosyltransferase gene Achn017071 (TCONS_00020411, TCONS_00020412), were randomly selected to verification by RT-PCR and Sanger sequencing ( Figure S3).

Differential Gene Expression during Fruit Development
To analyze genes related to fruit development, gene expression levels were estimated by counting the number of aligned reads to each kiwifruit gene region and converting them to FPKM values (Table S6A). While high correlation coefficients among biological replicates (>0.85) were observed in general, one of the 20 DAP biological replicates was not included in the differential expression analysis due to a low correlation coefficient with the other two biological replicates (Table S7). Significant differences in both up-and down-regulated genes were identified in each stage of fruit development by comparing expression levels with the previous time point (Figure S4; Table S6B).
To verify the results obtained from the RNA-Seq data, qRT-PCR was performed on a subset of 10 genes using gene-specific primers ( Table S1). The 10 transcripts were selected due to their significantly different levels of expression at 20, 120, and 127 DAP as determined by their FPKM values. The results of the qRT-PCR indicated that eight of the ten selected genes exhibited expression levels similar to the levels indicated by their FPKM values (Figure 3; Table S6C).
A total of 5931 differentially expressed genes were identified during fruit development in kiwifruit. More specifically, 4395 genes were differentially expressed in fruits at 20 DAP and 120 DAP; 1739 of which were up-regulated and 2657 of which were down-regulated in fruits at 120 DAP compared to fruits at 20 DAP (Table S6A). Only two, 1-aminocyclopropane-1-carboxylate oxidase (ACC, Achn213921) and kiwellin (Achn022471), of the 10 most highly up-regulated genes have been previously characterized. ACC plays an important role in the biosynthesis of the plant hormone ethylene, which in turn plays a major role in regulating fruit ripening (MacDiarmid and Gardner, 1993). Kiwellin, an allergenic protein formerly isolated from green kiwifruit, exhibits IgE binding capacity (Tuppo et al., 2008). Eighteen of the combined top 10 up-and top 10 down-regulated genes had unclear functions. A seed maturation protein gene (Achn286381) was highly expressed in mature fruit (120 DAP), while two aspartic proteinase nepenthesin genes (Achn141951 and Achn146881) exhibited lower expression levels in mature fruit compared to immature fruit (Table S6D).
A comparison of fruits at 127 and 120 DAP also revealed significant variations in expression. A total of 4091 genes, comprised of 1670 up-regulated and 2421 down-regulated genes, were identified in fruits at 127 DAP. The 10 most up-regulated and 10 most down-regulated genes are listed (Table S6D). Three amylase genes (Achn001191, Achn322221, and Achn141771) were among the top 10 most highly up-regulated genes at the postharvest stage (127 DAP) of fruit development. The upregulation of amylase genes is consistent with the results reported by Richardson et al. (2011). Limited information is available for 17 of the 20 most highly differentially expressed genes ( Table S6D).

Candidate Genes Related to Kiwifruit Sugar and Organic Acid Metabolism
The flavor of kiwifruit is highly dependent on the balance between soluble sugars and non-volatile organic acids (Nishiyama et al., 2008). Sweetness is the most important quality trait for kiwifruit as it influences overall fruit flavor (sugar/acid balance, perception of volatiles), and determines consumer acceptability (Nardozza et al., 2013). Most breeding programmes, however, have a negative impact on this trait. "Hongyang, " in comparison to other A. chinensis varieties, can accumulate more sugar levels (Wang et al., 2003;Nishiyama et al., 2008). Therefore, the expression of genes putatively related to kiwifruit sugar synthesis and accumulation and organic acid metabolism were analyzed (Figure 4; Table S8). Ten AGPase (Glucose-1-phosphate adenylyltransferase) genes, which play a key role in regulating starch biosynthesis, were identified in kiwifruit, and four of them were differentially expressed in different stages of fruit development. Four starch synthase (SS) genes were expressed in all three stages of fruit development, and two of them were more highly expressed in mature (120 DAP) fruit. The expression of two starch branching enzyme genes (SBE) was also detected in all three stages of fruit development, but no significant differences were observed in their expression in any of the stages of fruit maturity. Nineteen β-amylase transcripts were detected, 10 of which exhibited significant differences in expression. Seven β-amylase genes had the highest level of expression in the postharvest stage (127 DAP) of fruit maturity. The qRT-PCR data also indicated that BAM9 (Achn387071) exhibited the highest level in mature (127 DAP) fruit (Figure 5). Starch content is relatively higher in mature fruit and then begins to decrease during postharvest ripening. The content of glucose increased from 0.40 to 6.42 g/100 g fresh weight in three development stages ( Table S9). Because of starch degradation, the amount of glucose, fructose and sucrose was higher in the postharvest stage.
Invertase (INV) is an enzyme that catalyzes the hydrolysis (breakdown) of sucrose. Twenty expressed invertase genes were identified in fruit tissues of kiwifruit, five of which exhibited differential expression. In rice, the gene vacuolar invertase 3 (OsINV3) related to endospermal starch accumulation by regulated the ratio of hexose/sucrose. The coordinated expression of OsINV3 is considered to play an important role in acquiring sink strength for the start of elongation in all types of caryopses (Ishimaru et al., 2005). In kiwifruit, the INV3 homolog (Achn353961) is highly expressed in immature fruit (20 DAP) and may be related to fruit starch accumulation ( Figure 5; Table S8). LIN5 is an invertase that plays a key role in regulating sugar content, fruit development, fertility, and hormonal levels in tomato (Fridman et al., 2004). In kiwifruit, the Lin5-like homolog (Achn120291) may also play a key role in regulating soluble sugar levels ( Table S8). Sucrose synthase (SUS) catalyzes the reversible conversion of sucrose and a nucleoside diphosphate into nucleoside diphosphateglucose and fructose. SUS is involved in the synthesis of UDP-glucose and ADP-glucose in Arabidopsis, which are compounds that are linked to cellulose and starch biosynthesis, respectively (Barratt et al., 2009). A total of nine SUS genes were found to be expressed in fruit tissues at all stages of development, two of which, SUS2 (Achn240251) and SUSA (Achn024141), exhibited differential expression. SUS2 was highly expressed in immature fruit tissues and gradually decreased in subsequent stages of fruit development. In contrast, the expression of SUSA gradually increased as fruit developed from immature to the postharvest ripening stage (Figure 5; Table 5).
SPS is an important component of the plant sucrose biosynthesis pathway. Plant growth and productivity have been correlated with SPS activity in crop plants like maize and rice (Sharma et al., 2010;Okamura et al., 2011). SPS activity has also been correlated with sucrose accumulation in sugarcane stems (Grof et al., 2006). A total of six SPS genes were found to be expressed in kiwifruit, two of which exhibited higher levels of expression relative to the other SPS genes. qRT-PCR data indicated that SPSA (Achn218701) and SPS1 (Achn383801) exhibit a low expression level in immature fruit, and a higher expression level in mature (SPSA) and postharvest ripening (SPS1) stages ( Figure 5). The RNA-Seq data, however, differ somewhat from the qRT-PCR data for the SPSA gene ( Table 5). The content of sucrose increased from 0.23 to 3.50 g/100 g fresh weight in three development stages (Table S9), which may be attributed to SPSA and SPS1 expression.
Sucrose transporters play a central role, as they orchestrate sucrose allocation both intracellularly and at the whole plant level. Sugars are translocated in plants via sugar transporters, which are involved not only in long-distance sugar transport via the loading and unloading of phloem cells, but also in sugar allocation into source and sink cells (Anders and Huber, 2010;Kühn and Grof, 2010). A total of 13 sucrose transporter (SUC) genes were found to be expressed in fruit tissues of kiwifruit, two of which were expressed at a higher level relative to the other sugar transporter genes. A total of 61 sugar transporter, 12 hexose transporter (HT), and five glucose transporter (GT) genes were found to be expressed in the examined fruit tissues, 23, two, and one of which had higher levels of expression, respectively, relative to the other genes in the same families. The qRT-PCR data indicated that sugar transporter (Achn310541) exhibited its highest expression level in mature fruit and therefore may potentially play a role in starch accumulation. Sucrose metabolism, including the activity of sucrose synthases, SPSs, fructokinases, hexokinases, starch synthases, sucrose transporters, glucose transporters, hexose transporters, and UDP-galactose transporters, however, play a more essential role in sugar accumulation in kiwifruit.
Citrate and malate, products of the TCA cycle, are the main organic acids that, combined with sugars, play a key role in fruit flavor. The content of malic acid was 0.36, 2.31, and 2.38 g/100 g fresh weight, and the content of citric acid was 4.02, 9.33, and 9.74 g/100 g fresh weight, respectively, in three development stages (Table S9). Citrate synthase (CS) is the key enzyme in citrate biosynthesis. A total of 17 CS genes were found to be expressed in fruit, however only Achn367241 showed a distinct pattern of expression (Figure 4). It was highly expressed in postharvest ripened fruit relative to the other CS genes. Genes encoding malic enzymes (ME), which synthesize pyruvate by decarboxylating malate, were also analyzed. A total of 15 ME genes were found to be expressed in fruit, and five of them were more highly expressed relative to the other ME genes. A total of 16 malate dehydrogenase genes were also found to be expressed in fruit, however only Achn221601 showed a distinct pattern of expression (Figure 4).

Candidate Genes Related to Kiwifruit Glutamine, Aspartate, and Arginine Metabolism
The combination of sugars, organic acids and free amino acids results taste of fruit (Sorrequieta et al., 2010). We detected the content of 17 amino acids and γ-aminobutyric acid at three development stages (20, 120, and 127 DAP; Table S9). Glutamine, arginine, and aspartate are the major amino acids in kiwifruits, which is similar to the results reported by Redgwell and MacRae (1992). The content of arginine form 1.25 to 1.42 (mg/g Dry weight) is not significant difference in three development stages. Comparably, the content of aspartate (1.50, 0.86, 0.83 mg/g Dry weight) and glutamine (4.52, 1.20, 1.03 mg/g Dry weight) are decreased from 20 DAP to 127 DAP ( Table S9). The key genes putatively related to kiwifruit arginine, aspartate, and glutamine metabolism were analyzed. Six glutamate synthase (GOGAT) genes, six glutamate dehydrogenase genes (GDH), and four glutamate decarboxylase (GAD) genes were expressed in all three stages of fruit development (Table S10). But one of GDH (Achn063781) was slightly higher expressed in immature (20 DAP) fruit. In A. thaliana, AtGOGAT T-DNA insertion line showed a reduction of glutamate and biomass under normal CO 2 condition. According to the results, AtGOGAT played importance of ammonium assimilation in roots (Kojima et al., 2014).
GAD and GDH catalyze glutamate to γ-aminobutyric acid and α-ketoglutarate respectively, as metabolic nexuses to regulate carbon and nitrogen metabolism (Fait et al., 2011). In Panax ginseng C. A. Meyer, the expression of PgGAD gene was enhanced under various abiotic stresses (Lee et al., 2010). Three argininosuccinate synthase (ASS) genes and two argininosuccinate lyase (ASL) genes were expressed in among three stages of fruit development, but the ASS gene (Achn280681) was decreased in 127 DAP. In addition, the ASL gene (Achn041381) was highly expressed in mature fruit and postharvest stage (Table S10). In rice, normal root elongation requires arginine produced by OsASS and OsASL (Xia et al., 2014). Aspartate transaminase (AspAT) is an important enzyme in amino acid metabolism which catalyzes the interconversion of aspartate and α-ketoglutarate to oxaloacetate and glutamate (Brauc et al., 2011). Eleven AspAT genes were found and expressed in all three stages of fruit development (Table S10). In A. thaliana, overexpression the cytosolic AtAspAT influences amino acid metabolism and defense responses against Botrytis cinerea infection (Brauc et al., 2011). Those genes (GAD, GDH, GOGAT, ASS, ASL, AspAT) involved in glutamine, arginine, and aspartate metabolism may take part in stress response or development in kiwifruit.

CONCLUSIONS
Kiwifruit is a highly heterozygous vine plant with limited genomic resources. "Hongyang, " the cultivar chosen for the current transcriptome study, is widely planted in China and represents an excellent reference for genetic studies in kiwifruit. The major achievement of the current study was, through sequence assembly, annotation, expression analysis, and identification and characterization of lncRNAs, AS events, and novel protein-coding genes, to provide a list of potential candidate genes that could serve as targets for genetic improvement of kiwifruit. The study also serves as a resource for the expression patterns of the candidate genes. Collectively, the data in this study represents a valuable resource for further studies of fruit development in kiwifruit.

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fpls.2016. 00335 Table S1 | Primers used for RT-PCR and qRT-PCR analysis.   Table S4 | (A) Newly identified, potential protein-coding genes in the genome of A. chinensis cv. "Hongyang"; (B) Newly identified, novel genes involved in vitamin C, carotenoid, chlorophyll, and flavonoid metabolic pathways and ethylene signaling pathways. Table S5 | (A) Splice variants in A. chinensis cv. "Hongyang"; (B) Splice variants of vitamin C metabolic pathway genes; (C) Splice variants of chlorophyll, and flavonoid metabolic pathway genes; (D) Splice variants of ethylene signaling pathway genes.
Table S6 | (A) Gene expression profiles in A. chinensis cv. "Hongyang" at three stages of fruit development and ripening; (B) Significantly differentially expressed genes of A. chinensis cv. "Hongyang" in three stages of fruit development and ripening; (C) FPKM values of genes randomly selected for realtime-PCR validation; (D) Top 10 differentially expressed genes in three stages of fruit development and ripening.