Background

Gastrodia elata is a rootless and leafless achlorophyllous orchid that grows in a symbiotic relationship with two compatible mycorrhizal fungi, Mycena spp. and Armillaria mellea, during seed germination and vegetative growth, respectively [1–3]. The seeds of G. elata are tiny and do not possess an endosperm, and these seeds germinate only when adequate nutrition is obtained through the digestion of the specific fungi, Mycena spp., which invades the embryonic cells of these seeds [4–9]. Currently, four fungi species, including Mycena anoectochila, M. dendrobii (Fig. 1a), M. orchidicola, and M. osmundicola, isolated from different species of orchids [10], promote the germination of G. elata seeds to form protocorms and further develop into vegetative propagation corms (Fig. 1b) [11–16]. Once vegetative propagation corms have been established from seed germination, G. elata undergoes vegetative growth through an established symbiotic association with the compatible mycorrhizal fungi, A. mellea (Fig. 1c), to yield juvenile tubers (Fig. 1d) [1, 6, 17]. The vegetative propagation corms of G. elata obtain nutrition and energy from A. mellea to develop into tubers, and the growth conditions of tubers are positively and closely associated with the hyphal development of this fungi [1, 9, 18, 19]. The hyphae of A. mellea develop well in the cortical layers of G. elata tubers [1, 20-22]; however, the cells in the pith of tubers digest the invaded hyphae to obtain nutrition and energy [1].

Fig. 1
figure 1

Materials in the study. Fungus of Mycena dendrobii (a), vegetative propagation corm (b), Armillaria mellea (c), and juvenile tuber (d) of Gastrodia elata. Scale bars = 1 cm

Full size image

Since ancient times, Gastrodia elata has been used as a Chinese medicine for the cure of various conditions, including for its analgesic, antiepileptic, neuroprotective, anticonvulsant, and sedative effects against vertigo, general paralysis, and tetanus [23–26]. Numerous functional components isolated from G. elata have been characterized, such as gastrodin (4-hydroxymethylphenyl-β-D-glucopyranoside) and the aglycone gastrodigenin (4-hydroxybenzyl alcohol) as a primary active ingredient [27–29], and the other related components, including 4-hydroxybenzaldehyde, vanillyl alcohol, and vanillin, also show potential anticonvulsant activity [30–36]. Moreover, other constituents in G. elata, including gastrodioside, 4-hydroxybenzyl methyl ether, 4-hydroxybenzoic acid, parishin, β-Sitosterol, bis (4-hydroxybenzyl) sulfide, N6-(4-hydroxybenzyl) adenine riboside, dauricine, citric acid, palmitate, and succinic acid, have been reported [37–39].

Gastrodin, a simple glycoside comprising glucose and 4-hydroxybenzyl alcohol (the precursor of gastrodin), is the major phenolic compound of G. elata, and pharmacological tests have shown that this compound exhibits tranquilization, anti-inflammation, analgesia, cortical neuron protection, memory improvement, sedative, anticonvulsant, free radical scavenging, neuroprotective effect, anesthetic, and antioxidant effects [40–42]. Gastrodin was identified, characterized, and artificially synthesized at the end of 1970s [43]. Furthermore, gastrodin biosynthesis markedly increases from the growth stage of vegetative propagation corms to that of juvenile tubers, which have no flower buds [44]. In general, gastrodin production is derived from 4-hydroxybenzyl alcohol through a one-step glycosylation with different glucose donors. Therefore, one of key enzymes of gastrodin biosynthesis is glucosyltransferase, a large family identified in various plants [45]. Glycosylation is typically the final step in the biosynthesis of secondary plant compounds, resulting in the formation of a large number of glucosides [45–48]. The glycosylation might increase solubility or decrease volatility compared with non-glycosylated molecules [47]. Toluenes are general components in plants that serve as the precursors of plant secondary compounds [49–52]. Toluene is considered as a precursor of gastrodin [53]. The derivation of the metabolic pathway of 4-hydroxybenzyl alcohol derived from toluene is largely unknown in G. elata and other plants. However, the catalytic pathway from toluene to 4-hydroxybenzyl alcohol has been reported to involve hydroxylation through monooxygenase of cytochrome P450 (CYP450) [54, 55], a member of a large enzyme family in plants that catalyzes most of the oxidation steps in plant secondary metabolism [56–58]. The molecular basis of gastrodin biosynthesis remains largely unknown. In the present study, the comparative transcriptome analysis among A. mellea, the vegetative propagation corms and juvenile tubers of gastrodia was conducted using deep sequencing to reveal the gene regulation of gastrodin biosynthesis in G. elata.

Results and discussion

The transcriptome sequencing of NGS

De novo transcriptome sequencing has been used in various functional genomics studies, and is particularly suitable for gene expression profiling in non-model organisms without genomic sequences. The next generation sequencing (NGS) technology not only provides a comparative expressed sequence tag (EST) analysis for gene discovery on a genome-wide scale in non-model plants but also an efficient process for transcriptome sequencing and characterization. NGS platforms, such as the Illumina/Solexa Genome Analyzer and the Roche 454 GS FLX, have been widely used in recent years for the high-throughput sequencing of various organisms [59, 60]. Using these techniques for de novo transcriptome sequencing, EST databases have been successfully obtained for several medicinal herbs, including American ginseng [48], Salvia miltiorrhiza [61], sweet wormwood [62], Euphorbia fischeriana [63], Taxus [64], and other crops, such as chili pepper [65], maize [66], Curcuma longa [67], chestnut [68], Eucalyptus tree [69], olive [70], Camellia sinensis [71], sweet potato [72], Arabidopsis [73], and Phalaenopsis [74]. The Illumina platform is beneficial and useful for gene discovery because this technique can obtain deeper coverage and higher accuracy than Roche 454 sequencing technology [74]. Hence, the Illumina system was used in the present study to clarify the differential gene expression of different life stages of G. elata.

Sequencing and de novo assembly

A total of 21,045,338 (2x75 bases, 51 % GC), 18,436,794 (2x75 bases, 44 % GC) and 18,253,900 (2x75 bases, 45 % GC) high-quality paired-end (PE) reads were generated from Illumina HiSeq2000 platform, and approximately 3 giga bases of sequence data were obtained for each of A. mellea, vegetative propagation corm, and juvenile tuber of G. elata (Table 1). These short sequence reads have been deposited in NCBI under GEO accession number GSE73633. High-quality bases (above Q20) were more than 97 % for all three samples, indicating an excellent quality (Q20 means 1 error per 100 sequenced bases), while another high quality indicator was the aboundance peak of average sequence quality per read located around Q38 (less than 0.0158 % of error rate) for all three samples. The high-quality PE reads were used for de novo transcriptome shotgun assembly (TSA) to build transcript isoforms based on paired-end information. From a total of 161,517 assembled transcript isoforms (≥200 bases), 134,441 transcripts were selected as the representive unigenes with longest length for all loci (i.e., genes). The mapping rates of these high quality reads from all three samples against the total transcripts were all above 86 % (Table 2). The final N50 lengths of 1592 and 1184 bases, and the total lengths of 137,618,051 and 92,177,843 bases were calculated for the transcripts and the unigenes, respectively (Table 2) (Fig. 2). N50 statistics are widely used to assess the quality of the assembly, and the higher the N50 value representing the assembly the better the quality [67]. Compared with other plant transcriptome sequencing of de novo TSAs [61, 67, 69, 71, 72, 75–78], the N50 value obtained in the present study was above the average and adequate for further analysis (Table 3).

Table 1 Basic Statistics of RNA-Seq generated from Armillariella mellea, vegetative propagation corm and juvenile tuber of Gastrodia elata encoding by Illumina HiSeq2000 platform
Full size table
Table 2 Summary of the de novo transcriptome shotgun assembly from all Illumina sequences
Full size table
Fig. 2
figure 2

Summary distribution of the lengths of the 134,441 unigenes from combing three samples of raw reads (>200 nt, mean length = 686 nt, N50 = 1184 nt, Min = 201 nt, Max = 19,660 nt)

Full size image
Table 3 Comparative analysis of plant transcriptome N50 values
Full size table

Functional annotation and Gene Ontology classification

All unigenes were annotated according to the sequence similarity search against NCBI non-redundant protein sequence (nr) database using BLASTX algorithm. A total of 59,932 unique sequences were annotated, accounting for 44.58 % of the total unigenes (Table 4). Gene Ontology (GO) assignment were performed for the functional categorization of the annotated unigenes. A total of 11,645 unigenes were mapped to GO terms, accounting for 8.66 % of the unigenes (Table 4). Because multiple GO terms can be assigned to the same unigene [79], totally 58,488 GO terms were assigned in the present study. The GO annotation showed that these unigenes represent diverse functionalities and are involved in various metabolic pathways. In A. mellea, 9101, 7484, and 5694 GO terms, respectively, represent molecular function, biological process and cellular component categories [See Additional file 1: Figure S1]. In the molecular function category, the terms integral to “binding” (GO:0005488) and “catalytic activity” (GO:0003824) were shown as the most frequently occurring, constituting 19.23 % (4285) and 18.56 % (4136) of the level 2 GO terms, respectively. “Metabolic process” (GO:0008152) and “biological regulation” (GO:0065007) were the most frequently occurring under the biological process category, representing 9.50 % (2116) and 6.97 % (1552) of the level 2 GO terms, respectively. In the cellular component category, “cell part” (GO:0044464) was the most frequently occurring, representing 19.20 % (4277) of the total level 2 GO terms.

Table 4 Summary statistics of unigenes with functional annotations for all combined assembly and for each sample
Full size table

In vegetative propagation corm of G. elata, 17,371, 9875, and 6915 GO terms, respectively, represent molecular function, biological process, and cellular component categories [See Additional file 1: Figure S1]. In the molecular function category, the terms integral to “binding” (GO:0005488) and “catalytic activity” (GO:0003824) occurred most frequently, representing 24.84 % (8484) and 20.71 % (7076) of the total level 2 GO terms, respectively. In the biological process category, “metabolic process” (GO:0008152) was the most frequently occurring, representing 11.95 % (4083) of the total level 2 GO terms. In the cellular component category, “cell part” (GO:0044464) was the most frequently observed, representing 15.69 % (5359) of the total level 2 GO terms.

In the juvenile tuber of G. elata, 10,259, 5927, and 3753 GO terms were shown for molecular function, biological process and cellular component categories, respectively [See Additional file 1: Figure S1]. In the molecular function category, the terms integral to “binding” (GO:0005488) and “catalytic activity” (GO:0003824) were most frequently observed, representing 27.18 % (5420) and 19.53 % (3895) of the total level 2 GO terms, respectively. Metabolic process (GO:0008152) was the most frequently observed under the biological process category, representing 12.37 % (2466) of the total level 2 GO terms. In the cellular component category, “cell part” (GO:0044464) was the most frequently observed, representing 14.8 % (2951) of the total level 2 GO terms.

Differential expression analysis between A. mellea and juvenile tuber of Gastrodia elata

The comparative analysis of the transcriptomes of the A. mellea and G. elata (symbiosis with A. mellea) juvenile tubers was conducted based on the combined transcriptome assembly of all three samples. Among the total 134,441 unigenes, 49,890 and 71,722 unigenes were aligned with reads from A. mellea and G. elata juvenile tubers, respectively [See Additional file 2: Figure S2]. Among these unigenes, only 5547 unigenes were identified in both samples of deep sequencing data. To evaluate differential gene expression, the absolute value of the log2-FC (fold changes) ≥ 1, the q-values < 0.05 and the TMM-normalized FPKM > 0.3 were used as the criteria to determine the significance of gene expression differences [80]. A total of 292 differentially expressed unigenes were revealed in the transcriptome comparison, of which sixty-nine unigenes were significantly up-regulated (log2-FC ≥ 1, FPKM > 0.3, q-values < 0.05) in A. mellea [See Additional file 3: Table S1], and 223 unigenes in G. elata juvenile tubers were expressed at significantly higher levels (log2-FC ≤ -1, FPKM > 0.3, q-values < 0.05) (Fig. 3a) [See Additional file 4: Table S2]. Of 292 the differentially expressed unigenes, only 106 unigenes can be assigned with KEGG Ortholog identifiers (KOids) or enzyme commission (EC) numbers corresponding to biological pathways for cellular functions and molecular interactions after KEGG analysis. Among these, twenty-five up-regulated unigenes from A. mellea and G. elata juvenile tuber were assigned KOids or EC numbers corresponding to twenty-eight isogroups involved in sixteen different pathways [See Additional file 5: Table S3]; and eighty-one down-regulated unigenes were corresponding to 134 isogroups involved in sixty different pathways [See Additional file 6: Table S4].

Fig. 3
figure 3

Differentially expressed unigenes between a Armillaria mellea and juvenile tuber of Gastrodia elata (under the criteria: the absolute value of the log2-FC ≥ 1, the q-values < 0.05 and the TMM-normalized FPKM > 0.3), and b vegetative propagation corm and juvenile tuber (under the criteria: the absolute value of the log2-FC ≥ 1, the q-values < 0.05 and the TMM-normalized FPKM > 10). Numbers of up- and down-regulated unigenes were shoen in boxes

Full size image

Differential expression analysis between the juvenile tuber and vegetative propagation corm of Gastrodia elata

The comparative analysis of transcriptomes of the juvenile tubers (symbiosis with A. mellea) and vegetative propagation corms (asymbiosis with A. mellea) was conducted also based on the combined transcriptome assembly of all three samples. Among the total 134,441 unigenes, 71,722 and 82,712 unigenes were aligned with reads from juvenile tubers and vegetative propagation corms, respectively [See Additional file 7: Figure S3]. Among these, 63,317 unigenes were identified in both samples, and 22,942 differentially expressed unigenes were revealed in the transcriptome comparison using the same criteria (the absolute value of the log2-FC ≥ 1, the q-values < 0.05 and the TMM-normalized FPKM > 0.3). 7383 and 15,559 unigenes were expressed at higher levels in G. elata juvenile tubers and vegetative propagation corms, respectively (data not shown). To focus on highly and differentially expressed unigenes by modifying the threshold of TMM-normalized FPKM larger than ten, 703 highly and differentially expressed unigenes were revealed in the transcriptome comparison. Among which 298 unigenes were significantly up-regulated (log2-FC ≥ 1, q-values < 0.05, TMM-normalized FPKM > 10) in juvenile tubers [See Additional file 8: Table S5], and 405 unigenes in vegetative propagation corms were expressed at significantly higher levels (log2-FC ≤ -1, q-values < 0.05, TMM-normalized FPKM > 10) (Fig. 3b) [See Additional file 9: Table S6]. Of 703 the differentially expressed genes, only 244 unigenes can be assigned to KOids or EC numbers corresponding to to biological pathways for cellular functions and molecular interactions after KEGG analysis. Among these, 112 up-regulated (log2-FC ≥ 1, q-value < 0.05, TMM-normalized FPKM > 10) unigenes from juvenile tubers compared with vegetative propagation corms were assigned to KOids or EC numbers corresponding to 159 isogroups involved in seventy-eight different pathways [See Additional file 10: Table S7]; and 132 down-regulated (log2-FC ≤ -1, q-value < 0.05, TMM-normalized FPKM > 10) unigenes were assigned to KOids or EC numbers corresponding to 168 isogroups involved in eighty different pathways [See Additional file 11: Table S8].

Kusano (1911) first reported that G. elata existed in a mycorrhizal relationship with the wood-rotting pathogen A. mellea; however, this relationship was uncharacterized [18]. Until 1980, Zhang and Li showed that G. elata digests the invasive hyphae of A. mellea as the source of nutrition [1]. Lan et al. (1986) also confirmed that A. mellea was used as the source of nutrition for G. elata through the observation of labeled materials from A. mellea in the transverse section of G. elata. The labeled materials appeared in mitochondria, the endoplasmic reticulum and vacuoles of G. elata cortical cells [20]. When the hyphae of A. mellea are disconnected between wood (source of nutrition for A. mellea) and G. elata, the growth of G. elata terminates and this organism dies; therefore, the role of A. mellea for G. elata was considered as the food for survival [21]. According to the differential gene expression in G. elata in response to A. mellea symbiosis, unigene TRINITY_DN70668_c0_g1 is significantly induced [See Additional file 8: Table S5], as high as ~7 folds, and this gene was annotated as a gastrodianin (i.e., gastrodia antifungal proteins, GAFPs) gene, which digests the cell wall of A. mellea [81]. This result suggested that G. elata digests the invasive hyphae of A. mellea as a source of nutrition according to Zhang and Li (1980) [1]. In the present study, the low-level gene expression of the gastrodianin biosynthetic gene was detected in vegetative propagation corms (i.e., primary corms), which differentiated from protocorms, suggesting that the symbiotic relationship between A. mellea and G. elata only can be developed during the vegetative propagation corms of G. elata [82]. According to previous reports, there are two copies of gastrodianin biosynthetic genes in G. elata [81], and these two genes (unigenes TRINITY_DN70668_c0_g1 and TRINITY_DN48867_c0_g1) were also identified through deep sequencing data in the present study, only unigene TRINITY_DN70668_c0_g1 was significantly induced in the juvenile tubers of G. elata in response to A. mellea symbiosis [See Additional file 8: Table S5]. The result was consistent with the previous gene expression study of the gastrodianin biosynthetic gene in G. elata [81].

Identification and validation of candidate genes involved in gastrodin biosynthesis

The mechanism and related genes in the gastrodin biosynthesis pathway are currently unknown. To the best of our knowledge, gastrodin (4-hydroxymethylphenyl-β-D-glucopyranoside) is a simple glycoside comprising glucose and 4-hydroxybenzyl alcohol [27]. The last biosynthesis enzyme of gastrodin is glucosyltransferase [45]. Gastrodins are synthesized from 4-hydroxybenzyl alcohol with UDP-glucose via glucosylation catalyzed through glucosyltransferase. The precursor of gastrodin, 4-hydroxybenzyl alcohol, is catalyzed through cresols degradation (toluene degradation) from toluene through two steps of hydroxylation via monooxygenase (CYP450) [54, 55]. Therefore, both monooxygenase and glucosyltransferase are considered two key enzymes for gastrodin biosynthesis [53]. According to the chemical structure of gastrodin and gastrodin precursors, analyzed in previous reports, the putative gastrodin biosynthetic pathway is shown in Fig. 4. However, both monooxygenase (CYP450) and glucosyltransferase belong to a large enzyme families involved in different biosynthesis pathways in various plants [45, 54, 55]. Moreover, 4-hydroxybenzyl alcohol and gastrodin were also detected in Anoectochilus formosanus [83]. To determine the candidate genes involved in gastrodin biosynthesis, the comparative analysis of the transcriptomes between vegetative propagation corm and juvenile tuber of G. elata was conducted. A total of thirty and forty-six unigenes, respectively, were annotated to monooxygenase and glucosyltransferase among the 63,317 unigenes expressed both in the juvenile tubers and the vegetative propagation corms of G. elata. Under the criteria of log2-FC ≥ 1, TMM-normalized FPKM > 3 and q-values < 0.05, four putative monooxygenases were selected. Among them, unigene TRINITY_DN54282_c0_g1 was the most abundance in FPKMs and differentially expressed one (~2.4 times higher in the juvenile tubers). Under the same criteria, three putative glucosyltransferases were selected and the unigene TRINITY_DN50323_c0_g1 was the highest differentially expressed one (~3.2 times higher in the juvenile tubers).

Fig. 4
figure 4

Putative gastrodin biosynthetic pathway in Gastrodia elata

Full size image

To validate the differential expression of gastrodin biosynthesis-related unigenes TRINITY_DN54282_c0_g1 and TRINITY_DN50323_c0_g1, we investigated the expression of these genes in different life stages between the vegetative propagation corms and juvenile tubers of G. elata using semi-quantitative RT-PCR and real-time PCR (quantitative RT-PCR, qRT-PCR), and compared the results with fold-changes of FPKM (RNA-Seq). The specific primers for semi-quantitative RT-PCR and qRT-PCR were designed. In semi-quantitative RT-PCR, expression profiling revealed the differential expression of both monooxygenases and glycosyltransferases between the vegetative propagation corms and juvenile tubers of Gastrodia elata (Fig. 5). Both monooxygenases and glycosyltransferase genes were up-regulated in the juvenile tubers and were considered as gastrodin biosynthetic-related genes, as gastrodin production markedly increases from the growth stage of vegetative propagation corms to that of juvenile tubers [44]. In addition, the differential expression of the two genes was also validated through qRT-PCR analysis as shown in Fig. 6. The expression levels of the unigene TRINITY_DN54282_c0_g1 were up-regulated up to 6.5 times in juvenile tubers and 2.4 times for qRT-PCR and RNA-Seq, respectively. Similarly, qRT-PCR and RNA-Seq showed that the expression levels for unigene TRINITY_DN50323_c0_g1 were up-regulated approximately 3.6 and 3.2 times, respectively, in juvenile tubers. The results of both semi-quantitative RT-PCR (Fig. 5) and qRT-PCR analysis (Fig. 6) were consistent with the RNA-Seq data.

Fig. 5
figure 5

Semi-quantitative RT-PCR profile of gasstrodin related candidate genes unigene TRINITY_DN54282_c0_g1 (monooxygenase) and TRINITY_DN50323_c0_g1 (glycosyltransferase) in different life stage of vegetative propagation corm (a) and juvenile tuber (b) of Gastrodia elata tissues with ubiquitin as the internal control. RT+ and RT- represent amplifications with and without reverse transcriptase

Full size image
Fig. 6
figure 6

Quantitative real-time PCR (Q-PCR) validations of (a) unigene TRINITY_DN50323_c0_g1 and (b) unigene TRINITY_DN54282_c0_g1 of RNAseq results (TMM-normalized FPKM fold changes). Comparison of differential expression values between the juvenile tuber (white bar) and vegetative propagation corm (grey bar) of Gastrodia elata determined by qRT-PCR and RNAseq

Full size image

Monooxygenases belong to cytochrome P450 proteins, the largest family of plant proteins, which catalyze most of the oxidation steps in plant secondary metabolism [57, 58]. The comparison of the chemical structures of 4-hydroxybenzyl alcohol and the precursor toluene (Fig. 4), revealed two steps catalyzed through monooxygenase, the conversion of toluene to 4-hydroxytoluene and the conversion of 4-hydroxytoluene to 4-hydroxybenzyl alcohol. Therefore, unigene TRINITY_DN54282_c0_g1 (EC: 1.14.13.-) was considered as a toluene monooxygenase gene, consistent with the KEGG pathway annotation (Fig. 7). Generally, glycosylation is the last step in the biological biosynthesis of secondary metabolism because sugar conjunction results in both the increased water solubility and stability of the compounds [84–86]. In G. elata, glycosyltransferase catalyzes the last step of the gastrodin biosynthesis pathway, which converts 4-hydroxybenzyl alcohol to gastrodin [45] (Fig. 4). Therefore, unigene TRINITY_DN50323_c0_g1 (EC: 2.4.1.12) was considered as a glycosyltransferase gene, consistent with the KEGG pathway annotation (Fig. 8). In short, both unigenes TRINITY_DN54282_c0_g1 and TRINITY_DN50323_c0_g1 might be key enzyme genes that, respectively, participate in the hydroxylation (Fig. 7) and glucosylation (Fig. 8) of gastrodin (Fig. 4).

Fig. 7
figure 7

Hydroxylation by monooxygenase (EC:1.14.13.-) in toluene degradation. The putative enzymes are in red

Full size image
Fig. 8
figure 8

Glucosylation by beta-1,4-glucosyltransferase (EC:2.4.1.12) in starch and sucrose metabolism. The putative enzyme is in red

Full size image

Cloning of gastrodin biosynthesis related genes

The full-length cDNA sequences of monooxygenases (unigene TRINITY_DN54282_c0_g1) and glycosyltransferases (unigene TRINITY_DN50323_c0_g1) from G. elata were further isolated through RACE analysis. The nucleotide sequence of the full-length monooxygenase cDNA has an open reading frame (ORF) of 1476 nucleotides spanning from the first initiation codon (ATG) to the termination codon (TGA), an in-frame stop codon located 12 nt upstream from the initiation codon and an out-of-frame ATG located upstream of the main ORF. The complete ORF encodes a protein of 491 amino acids with a predicted molecular mass of 55.8 kDa (Fig. 9a). In addition, the nucleotide sequence of the full-length glucosyltransferase cDNA has an ORF of 1635 nucleotides spanning from the first initiation codon ATG to the termination codon TGA, an in-frame stop codon located 12 nt upstream from the initiation codon and an out-of-frame ATG located upstream of the main ORF. The complete ORF encodes a protein of 544 amino acids with a predicted protein of 63.1 kDa (Fig. 9b).

Fig. 9
figure 9

a Schematic representation of the mRNA transcripts of (a) unigene TRINITY_DN54282_c0_g1 (monooxygenase) with an out-of-frame AUG and an in-frame stop codon upstream the start codon; b unigene TRINITY_DN50323_c0_g1 (glycosyltransferase) with an out-of-frame AUG and an in-frame stop codon upstream the start codon

Full size image

Notably, both monooxygenases (unigene TRINITY_DN54282_c0_g1) and glycosyltransferases (unigene TRINITY_DN50323_c0_g1) induced in response to fungi symbiosis possess an out-of-frame upstream ATG and an in-frame stop codon in the main ORF within the 5’UTR. In mammals, upstream ATGs/upstream ORFs significantly reduce protein expression levels through a reduction of the translation efficiency [87, 88] or mRNA decay [89–91]. Upstream AUGs/upstream ORFs in the 5′ UTR efficiently disrupt the translation of the downstream coding sequence, thereby reducing the translation efficiency of the main coding region [87, 88]. According to deep sequencing, semi-quantitative RT-PCR, and real-time RT-PCR, both monooxygenases (unigene TRINITY_DN54282_c0_g1) and glycosyltransferase (unigene TRINITY_DN50323_c0_g1) genes were co-expressed in vegetative propagation corms. Therefore, the upstream ATGs of these two genes might result in the low concentration of gastrodin in vegetative propagation corms. The induction of gene expression in response to fungi symbiosis might increase the translation efficiency or mRNA stability of the two key enzymes to increase gastrodin production in juvenile tubers. In response to stress, the translation repression of upstream ATGs/upstream ORFs could be significantly reduced [92–94] or mRNA stability could be increased [95, 96]. In response to fungi invasion, the repression reduction of the translation efficiency of both monooxygenases (unigene TRINITY_DN54282_c0_g1) and glycosyltransferase (unigene TRINITY_DN50323_c0_g1) genes in G. elata might also increase the accumulation of gastrodin in juvenile tubers, as fungal infection could be a biotic stress to G. elata.

Conclusions

The molecular basis of gastrodin biosynthesis in G. elata was clarified based on de novo transcriptome sequencing in the present study. Two putative monooxygenase (unigene TRINITY_DN54282_c0_g1) and glycosyltransferase (unigene TRINITY_DN50323_c0_g1) genes associated with the gastrodin biosynthesis pathway were identified. The genes of the two key enzymes involved in gastrodin biosynthesis might be applied as the target genes for plant gene transformation in future studies to obtain transgenic plants or microbial hosts with gastrodin production. Moreover, this transcriptome dataset also provides important information to accelerate future gene expression and functional genomics studies in G. elata.

Methods

Materials

Plant materials used in this study were got from the Chinese medical farm. The voucher specimens were deposited at the herbarium of the Taiwan Endemic Species Research Institute (TAIE) and the voucher numbers are Hsu 17054 and 17055. The vegetative propagation corms and juvenile tubers of G. elata were, respectively, harvested 1 and 12 months after sowing at the Chinese medical farm (Hakusan City, Jilin Province, China) (Fig. 1b and d). A. mellea was isolated from the juvenile tuber of G. elata and identified based on ITS sequence of nuclear ribosomal DNA (nrDNA) (data not shown). The fungi were cultured on PDA medium according to Mishra and Dubey (1994) for further molecular studies (Fig. 1c) [97].

RNA isolation, cDNA library preparation, deep sequencing and de novo assembly

The vegetative propagation corms of G. elata, juvenile tuber of G. elata and A. mellea was separately harvested and extracted total RNA using Trizol® Reagent (Invitrogen, Carlsbad, CA, USA) and followed by the RNeasy Plant Mini Kit (Qiagen, Hilden, Germany). Purified RNA was quantified at OD260 using a ND-1000 spectrophotometer (NanoDrop Technology, San Diego, CA, USA) and qualitated using an Agilent Bioanalyzer 2100 with the RNA 6000 nano labchip kit (Agilent Technology, Santa Clara, CA, USA). The RNA Integrity Number (RIN) was identified 8.7, 9.1, and 8.0 for vegetative propagation corm, juvenile tuber, and A. mellea, respectively. Each of the RNA libraries was separately constructed using TruSeq RNA Library Preparation Kit v2 (Illumina Inc., San Diego, CA) with standard Illumina protocols. After quality-control using an Agilent Bioanalyzer 2100, each library was paired-end deep sequenced by Illumina HiSeq 2000 sequencer (Illumina Inc., San Diego, CA, USA) according to the manufacturer’s instructions. The quality of the produced fastq sequences was assessed using FastQC program (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/), see Additional files 12, 13 and 14: Figures S4, S5 and S6. Before the assembling stage, the reads were processed to trim adaptor sequences and low-quality ends via Trimmomatic version 0.32 [98] with parameters “ILLUMINACLIP: TruSeq3-PE.fa:2:30:10 SLIDINGWINDOW:4:5 LEADING:5 TRAILING:5 MINLEN:25”. A single de novo transcriptome assembly was generated from high-quality short reads of all three samples using the Trinity software (https://github.com/trinityrnaseq/trinityrnaseq/wiki/) [99]. Trinity v2.1.0 (release 20140413p1) was employed with the default k-mer of 25 and minimum assembled contig length of 200. Candidate coding regions within transcript sequences were identified using TransDecoder (http://transdecoder.github.io/).

Transcript abundance estimation

Quantification of transcripts was estimated using the RNA-Seq by Expectation-Maximization (RSEM) software version 1.2.23 [100], which was bundled with the Trinity software distribution. The RSEM protocal uses the Bowtie software (http://bowtie-bio.sourceforge.net/index.shtml) [101] to align trimmed reads from each sample seperatively to the assembled transcripts, and then computes transcript abundance, estimates the number of aligned fragments corresponding to each transcript, including normalized expression values as FPKM for paried-end reads. In addition, RSEM computes ‘gene-level’ expression values using the Trinity component as a proxy for the gene. For comparing expression levels of different transcripts or genes across samples, a Trinity-bundled script invokes the EdgeR package to perform an additional TMM scaling normalization that aims to estimate differences in total RNA production across all samples [102, 103]. Transcripts with zero TMM-normalized FPKM values for all three samples were removed from the assembly and not counted into the total transcript number, and the rest with longest length per component (i.e., gene locus) defined by Trinity were interpreted to represent “unigenes” for downstream analysis. The data presented in this publication have been deposited in NCBI’s Gene Expression Omnibus [104] and are accessible through GEO Series accession number GSE73633 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE73633).

Gene functional annotation, GO classification and KEGG pathway analysis

The functional annotation of the unigenes was conducted using the BLASTX algorithm of DIAMOND program [105] with E-values < 1.00E-5 and enabled ‘sensitive mode’ against NCBI non-redundant protein sequence database (nr, October 2015), which consisted of all non-redundant peptide sequences from GenBank CDS translations, RefSeq Proteins, PDB (Protein Data Bank) database, SwissProt database, PIR (Protein Information Resource) database, and PRF (Protein Research Foundation) database. Gene Ontology (GO) annotations of unigenes was performed by coverting the GeneBank identifiers (gi) of hits from the BLASTX results into UniProt IDs through the IDmapping data files [106] downloaded from The Universal Protein Resource (UniProt; http://www.uniprot.org/), then the corresponding GO terms were retrieved via the UniProt IDs using home-made Perl scripts. KEGG (http://www.kegg.jp/) [107] pathway annotation of unigenes was performed using the BLASTX algorithm of DIAMOND program with E-values < 1.00E-5 and enabled ‘sensitive mode’ against KEGG gene peptide database (October, 2015), and then the corresponding KO identifiers, EC numbers and pathway categories were parsed using home-made Perl scripts.

Comparative analysis of differentially expressed genes

The number of unigenes per sample was counted if the corresponding TMM-normalized FPKM values of unigenes above zero. For differential expression analysis, each pairwise comparison was performed from the TMM-normalized FPKM values using the R package limma (Linear Models for Microarray and RNA-Seq Data) [108] with upper-quantile normalization. Zero TMM-normalized FPKM values were replaced to 0.001 from either sample to avoid problems associated with zero value. A number of statistics were calculated, including log2-FC (fold changes) between the two samples, the p-values, and associated q-values (FDR-corrected p-values). Differentially expressed unigenes were selected using the criteria that the absolute value of the log2-FC ≥ 1, the q-values < 0.05, and different TMM-normalized FPKM values under variant stringency. Genes with a q-value < 0.05 were considered to be differentially expressed [109].

Semi-quantitative RT-PCR, real-time RT-PCR analysis and SMART-RACE cDNA amplification

The cDNA products were diluted 20-fold with deionized water prior to use as a template in semi-quantitative RT-PCR and real-time PCR [110]. Semi-quantitative RT-PCR reactions were performed in 20 μl reactions containing gene-specific primers [See Additional file 15: Table S9] and the ubiquitin gene primer as an internal control [108, 111]. Additional reaction components were 1X Red Taq Mastermix (RBC Bioscience, Taipei, Taiwan), 1 mM MgCl2 and 1 μM of the specific primers. Following PCR amplification, 5 μl of the PCR products were separated on a 1 % TAE agarose gel containing ethidium bromide, and the bands were photographed under UV light using gel documentation system alpha imager EC (Alpha Innotech, Japan). Real-time RT-PCR was performed using the Power SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA, USA) and a 7900HT Fast Real-Time PCR System (Applied Biosystems, Foster City, CA, USA) according to the manufacturer’s instructions. SDS2.2.2 software (Applied Biosystems, Foster City, CA, USA) was used for the comparative ΔCt analysis, and the ubiquitin gene served as an internal control. The relative gene expression was calculated using the 2-ΔΔCt method [112]. In SMART-RACE cDNA amplification, the 5′ and 3′-RACE (5′ and 3′-rapid amplification of cDNA ends) was performed using the SMART-RACE cDNA amplification kit (Clontech, Palo Alto, CA, USA) according to the manufacturer’s instructions. All primers used in the present study are listed in Additional file 15: Table S9.

Availability of supporting data

The data set supporting the results of this article are available in the NCBI GEO repository, with the accession numbers GSE73633 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE73633). All the supporting data are included as additional files.