Full-length transcriptome profiling of Acanthopanax gracilistylus provides new insight into the kaurenoic acid biosynthesis pathway

Abstract

Acanthopanax gracilistylus is a deciduous plant in the family Araliaceae, which is commonly used in Chinese herbal medicine, as the root bark has functions of nourishing the liver and kidneys, removing dampness and expelling wind, and strengthening the bones and tendons. Kaurenoic acid (KA) is the main effective substance in the root bark of A. gracilistylus with strong anti-inflammatory effects. To elucidate the KA biosynthesis pathway, second-generation (DNA nanoball) and third-generation (Pacific Biosciences) sequencing were performed to analyze the transcriptomes of the A. gracilistylus leaves, roots, and stems. Among the total 505,880 isoforms, 408,954 were annotated by seven major databases. Sixty isoforms with complete open reading frames encoding 11 key enzymes involved in the KA biosynthesis pathway were identified. Correlation analysis between isoform expression and KA content identified a total of eight key genes. Six key enzyme genes involved in KA biosynthesis were validated by real-time quantitative polymerase chain reaction. Based on the sequence analysis, the spatial structure of ent-kaurene oxidase was modeled, which plays roles in the three continuous oxidations steps of KA biosynthesis. This study greatly enriches the transcriptome data of A. gracilistylus and facilitates further analysis of the function and regulation mechanism of key enzymes in the KA biosynthesis pathway.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12298-024-01436-7.

Introduction

Acanthopanax gracilistylus W. W. Smith is a medicinal plant of the genus Acanthopanax Miq. in the family Araliaceae. The dry root bark of A. gracilistylus, known as Acanthopanacis Cortex is enriched with active ingredients such as volatile oils, terpenes, sterols, and organic acids (Zheng et al. 2015) contributing to various pharmacological activities such as hypoglycemic (Lu et al. 2018), anti-inflammatory (Xu et al. 2018), anti-tumor (Wu et al. 2014), anti-aging (Xie et al. 2003), anti-fatigue (Zhang 2012), and liver protection (Kuang 2015) effects.

Terpenoids are the main components in A. gracilistylus. To date, two sesquiterpenoids, 41 diterpenoids, and 23 triterpenes and their glycosides have been isolated from the root bark, stems, and leaves of A. gracilistylus (Yang et al. 2020). Diterpenoids exhibit significant biological activities in the human body, such as anti-tumor, antibacterial, antiviral, anti-inflammatory, and immunosuppressive effects, along with the ability to lower blood pressure and prevent thrombosis (Hai et al. 2015). Among the diterpenoids, kaurenoic acid (KA) is a potent anti-inflammatory agent, with a content of approximately 1% in the root bark of A. gracilistylus and a high content in other plants such as Eleutherococcus trifoliatus, Aralia continentalis, and Smallanthus sonchifolius; therefore, all of these plants are used in traditional Chinese medicine in the treatment of rheumatoid arthritis (Xie et al. 2013; Sithisarn et al. 2009; Choi et al. 2011; Ragasa et al. 2008). KA has various pharmacological effects such as antibacterial activity; relaxation of the vascular smooth muscle; antispasmodic, antiplatelet aggregation, antitumor, antihypertensive, diuretic, and anti-inflammatory effects; and protection against osteoporosis (Cavalcanti et al. 2006; Fernandes et al. 2013; Park et al. 2010; Lyu et al. 2011; Ndom et al. 2010; Ambrosio et al. 2010; Tirapelli et al. 2004; Yang et al. 2002).

KA is the precursor of the synthesis of gibberellin (GA), which is a diterpenoid plant hormone that regulates plant growth and development (Barker et al. 2021). Kasahara et al. demonstrated that GA is mainly synthesized through the methylethritol phosphate (MEP) pathway in Arabidopsis thaliana, based on the detection of ¹³C-labeled precursors (Kasahara et al. 2002). The contribution of the mevalonate (MVA) pathway to GA synthesis was suggested to depend on the influx of isoprenoid into the plastids from the cytosol (Flügge and Gao 2005). KA is formed from geranylgeranyl diphosphate (GGPP) through a series of enzymatic reactions. The cyclization of GGPP is catalyzed in a two-step reaction by ent-copalyl diphosphate synthase (CPS) and ent-kaurene synthase (KS) to synthesize ent-kaurene. Three continuous oxidation reactions of ent-kaurene further generate ent-kaurenol and ent-kaurenal to finally generate KA (Hedden 2020). Therefore, ent-kaurene oxidase (KO) is considered the key regulatory enzyme involved in the biosynthesis of KA (Morrone et al. 2010; Su et al. 2016; Yamamura et al. 2018).

Third-generation sequencing technology is characterized by single-molecule real-time sequencing, enabling the direct sequencing of transcripts up to 10 kb in length without requiring a reference genome; however, this technology remains limited by a high error rate, high cost per base, and low throughput (Rhoads and Au 2015). Second-generation sequencing results in short read lengths and is often not capable of identifying full-length gene isomers. Nevertheless, second-generation sequencing can be used for the error correction of full-length transcripts. Accordingly, the combination of second- and third-generation sequencing can offer detailed results at the single-nucleotide level through integration with high-throughput and high-precision data, and can further improve the identification of gene isomers and abundance estimation (Magi et al. 2017; Ou et al. 2020; Winand et al. 2019). Sequencing analyses of several Chinese herbal medicinal plants, including Casuarina equisetifolia (Ye et al. 2019), Ipomoea nil (Hoshino et al. 2016), and Chrysanthemum morifolium (Liu et al. 2021), have offered a deeper understanding of the color, properties, growth rate, and stress resistance of the plants. To date, KA has been largely studied in crops such as rice (Sakamoto et al. 2004) and wheat (Huang et al. 2012) with completed transcriptome databases; however, there is no corresponding transcriptome study related to the KA biosynthesis pathway in A. gracilistylus. Therefore, in the present study, we combined the results obtained from second- and third-generation sequencing platforms of the leaves, roots, and stems to construct the first full transcriptome of A. gracilistylus. We further focused on the genes involved in the biosynthesis pathway of the active medicinal ingredient (KA) of A. gracilistylus. These results can lay an experimental foundation for the development of genetic engineering strategies to increase the yield of KA in A. gracilistylus.

Materials and methods

Preparation of experimental materials

A. gracilistylus was collected from the herb garden of Anhui University of Chinese Medicine, which was identified by Professor Qingshan Yang. The fresh leaves, roots, and stems of A. gracilistylus were separated, washed with sterilized distilled water several times, and then dried with filter paper. All tissues were quickly frozen in liquid nitrogen after collection.

Extraction and analysis of KA by high-performance liquid chromatography

The content of KA in the leaves, roots, and stems of A. gracilistylus was determined by high-performance liquid chromatography-diode array detection (HPLC–DAD) on an Agilent 1260 instrument (Agilent Co., Foster City, CA, USA). First, 0.4 g of dried powder of each tissue was weighed, added to 40 mL cyclohexane, and mixed well. The solution was then ultrasonicated (150 W, 40 kHz) at room temperature for 50 min, and 20 mL of the extract was evaporated to dryness. The powder was dissolved again in methanol and the volume was fixed in a 2 mL volumetric flask (Xie et al. 2013). KA (HPLC-grade ≥ 98%, National Institutes for Food and Drug Control) was used as the standard. The samples were run on an Agilent HC-C18 chromatographic column (250 × 4.6 mm, 5 μm) with the column temperature maintained at 30 ℃ and the detection wavelength set to 205 nm. The mobile phase was 1% acetic acid water and methanol (10:90) and the elution procedure was equivalent elution (Lee et al. 2020). The flow rate was 1.0 mL/min and the injection volume was 10 μL. The KA content in each tissue was measured three times.

Extraction of RNA from A. gracilistylus

After sterilizing and autoclaving the required utensils, the A. gracilistylus tissues were ground in a mortar with the addition of liquid nitrogen. The powder was mixed and placed in centrifuge tubes and the supernatant was collected after centrifugation. Total RNA was extracted from each tissue type with an RNA kit (Omega Bio-Tek, Norcross, GA, USA) according to the manufacturer’s protocol. For mRNA enrichment, the total RNA was mixed with magnetic beads containing OligodT with a polyA tail. A DNA probe was used to hybridize rRNA and then RNaseH was used to digest the DNA/RNA hybrid chain. Finally, DNaseI was used to digest the DNA probe and obtain the required RNA sample after purification.

Library preparation and database annotation

The DNB-seq platform was used for three repeated sequencing of leaves, roots, and stems of A. gracilistylus, and the PacBio iso-seq platform was used to sequence the full-length transcriptome of the mixture of leaves, roots and stems. The raw reads obtained by transcriptome sequencing on the DNB-seq platform were filtered with SOAPnuke (v1.5.2) software (Cock et al. 2010). Reads containing adapters, an unknown base (N) content > 10%, and of low quality were removed to obtain clean reads. After detection and analysis of full-length transcripts from the PacBio iso-seq platform, full-length non-chimeric sequences were clustered. For each cluster composed of highly similar sequences, we applied the Arrow algorithm of PacBio to correct errors and integrate them into a consistent sequence. We filtered the data according to the sequence quality value after error correction and retained only the consistent sequence with high reliability. The library preparation and sequencing were carried out at Beijing Genomics Institute (Shenzhen, China). Finally, the high-quality sequences obtained after clustering and error correction were combined to remove redundancies enabling isoform identification and annotation to seven major databases (Supplementary Table S1).

Isoform expression and differential expression analysis

Bowtie 2 (v.2.2.5) (Langmead and Salzberg 2012) and RSEM (v1.2.8) (Li and Dewey 2011) software were used to calculate the isoform expression of the three tissues of A. gracilistylus, expressed as fragments per kilobase of exon model per million mapped fragments (FPKM). Based on the Poisson distribution principle, the differentially expressed genes (DEGs) of the tissues were analyzed and subject to functional annotation. DEGs were identified according to a false discovery rate (FDR) ≤ 0.001 and more than two fold difference in expression level between groups (tissue types).

Transcription factor (TF) analysis

The getorf (EMBOSS: 6.5.7.0) tool was used to determine the open reading frame (ORF) of each isoform in A. gracilistylus. The ORFs were then compared to the TF protein domain by hmmsearch (v3.0). Finally, the probability of each isoform being a TF was identified according to TF family characteristics described in PlantTFDB.

Structural analysis of KO in A. gracilistylus

ExPASy (https://web.expasy.org/translate/) was used to determine the ORF sequence of KO in A. gracilistylus. Using MEGA 5.0 (Tamura et al. 2011) and CLUSTALX 1.83 (Jeanmougin et al. 1998) software, the amino acid sequences of KO in A. gracilistylus were compared with KO sequences in other plants recorded in the National Center for Biotechnology Information (NCBI) international database. The conserved active site was searched and the secondary structure of KO in A. gracilistylus was displayed using ESPript 3.0 (https://espript.ibcp.fr/ESPript/cgi-bin/ESPript.cgi). Finally, the tertiary structure of KO in A. gracilistylus was simulated by Swiss-Model (https://swissmodel.expasy.org/) and visualized using Pymol (v2.3.2.0) software (Shan et al. 2020).

Validation of isoform expression of key enzymes

The key enzyme isoforms related to the biosynthesis of KA in A. gracilistylus identified by transcriptome analysis were selected for experimental validation by real-time quantitative polymerase chain reaction (RT-qPCR). The reaction mixture was placed in a 0.2 mL PCR tube comprising 7.5 μL 2 × qPCR Mix, 1.5 μL of 2.5 μM each forward and reverse primer (Supplementary Table S2), 2.0 μL of the reverse transcription product (cDNA), and 4.0 μL nuclease-free water. PCR amplification comprised a pre-denaturation step at 95℃ for 30 s, followed by 40 cycles of denaturation at 95℃ for 15 s and annealing and extension at 60℃ for 30 s. The relative expression of each independent isoform was normalized to the poly-ubiquitinase isoform (isoform_52190), and the relative isoform expression was calculated by the 2^−ΔΔCT method (Livak and Schmittgen 2001).

Statistical analysis

The experimental data are expressed as mean ± standard deviation. GraphPad Prism 8 software was used for plotting and statistical analysis. Single-factor analysis of variance was used to compare the mean values between groups. The OmicShare Tools (https://www.omicshare.com/tools/) was used to analyze the correlation between the expression level of isoforms and KA content.

Results

Determination of KA content in A. gracilistylus

HPLC–DAD demonstrated that KA was present in all three tissues of A. gracilistylus, with a significantly higher content in the roots, followed by the stems, and the leaves had the lowest content overall (Fig. 1, Supplementary Table S3, Supplementary Fig. S1).

Fig. 1.

Determination of KA content in A. gracilistylus. a Morphological characteristics of A. gracilistylus. b Results of KA determination in three tissues of A. gracilistylus (error bars indicate standard error from three biological replicates). **p < 0.01, ****p < 0.0001

Transcriptome sequencing and data assembly

The DNB-seq and PacBio iso-seq platforms were used for transcriptome sequencing of A. gracilistylus. The RNA-sequencing datasets from the three A. gracilistylus tissues were deposited in the NCBI Sequence Read Archive (SRA) database (accession: PRJNA945217). The DNB-seq platform was used to sequence the leaves, roots, and stems of A. gracilistylus separately, with a total of 358,103 genes detected and 229,522 genes co-expressed in the three tissues (Supplementary Fig. S2). The average output of each sample was 69.48 MB (Table 1). The PacBio iso-seq platform was used to detect the mixed samples of the leaves, roots, and stems of A. gracilistylus, resulting in 63.59 GB of polymerase read sequences and 62.70 GB of subreads. After insert processing of the reads, we obtained 715,020 circular consensus sequences (CCS). The average read length of the CCS was 5830 bp and the average read quality was 0.97. After assembling the full-length non-chimeric reads into high-quality sequences and removing redundancies 505,880 transcripts were finally generated. The N50 value was 1552 bp and the N90 value was 586 bp. The minimum length was 197 bp, the maximum length was 28,796 bp, and the GC content was 40.71% (Table 2, Supplementary Fig. S3).

Table 1.

Comparison of clean reads and reference gene sequence obtained in second-generation sequencing

Sample	Total clean reads (M)	Total mapping (%)	Uniquely mapping (%)
Ag_L_1	70.19	86.05	3.85
Ag_L_2	69.45	84.81	3.93
Ag_L_3	69.85	85.55	4.04
Ag_R_1	68.75	84.26	5.10
Ag_R_2	69.75	82.91	5.20
Ag_R_3	70.04	82.60	5.10
Ag_S_1	70.18	85.05	3.84
Ag_S_2	68.20	84.20	4.00
Ag_S_3	68.89	82.34	4.93

Table 2.

Third-generation sequencing results and data quality control

Item	Number	Mean length (bp)
Polymerase reads	820,253	77,529.66
Subreads	12,794,646	4900.68
Circular consensus sequences (CCS)	715,020	5830
Full-length non-chimeric reads (FLNC)	681,426	948

Functional annotation and expression analysis of isoforms

To predict the potential functions of the isoforms identified in A. gracilistylus, the 505,880 isoforms were annotated by seven functional databases, including the Non-redundant (NR; 385,189 isoforms), Nucleotide (NT; 331,924 isoforms), Swissprot (291,516 isoforms), Kyoto Encyclopedia of Genes and Genomes (KEGG; 296,226 isoforms), Eukaryotic Orthologous Groups of proteins (KOG; 294,088 isoforms), Pfam (227,250 isoforms), and Gene Ontology (GO; 224,994 isoforms) databases. Among them, 103,418 isoforms were jointly annotated by the seven databases, accounting for 20.44% of the total, and 408,954 isoforms were annotated by any one of the seven major databases, accounting for 80.84% of all isoforms (Table 3).

Table 3.

Annotation results of the transcripts with seven functional databases

Database	Annotated number	Proportion (%)
NR	385,189	76.14
NT	331,924	65.61
Swissprot	291,516	57.63
KEGG	296,226	58.56
KOG	294,088	58.13
Pfam	227,250	44.92
GO	224,994	44.48
Overall	408,954	80.84

The expression levels of 358,103 isoforms in the nine samples of the leaves, roots, and stems of A. gracilistylus are shown in Fig. 2a, with higher expression detected in the stems than in the leaf and root tissues (Fig. 2b).

Fig. 2.

Expression of isoforms in the three tissues of A. gracilistylus. a Distribution of the number of isoforms with different expression levels in the three tissues. b Boxplots of isoforms expressed in three tissues. L, leaf; R, root; S, stem

Annotations and functional classification of isoforms

The 385,189 A. gracilistylus isoforms annotated in the NR database were further analyzed, demonstrating the highest similarity with Daucus carota var. sativa. (48.21%), followed by Vitis vinifera, Sesamum indicum, and Helianthus annuus (Supplementary Fig. S4). This comparison indicated that D. carota has the closest homologous relationship with A. gracilistylus among the plants in the known gene database. The KOG database annotated 395,108 isoforms, which were divided into 25 categories. These were mainly annotated to 62,042 entries in “general function prediction only,” 47,848 entries in “signal transduction mechanisms,” and 37,403 entries in “posttranslational modification, protein turnover, chaperones” (Fig. 3a).

Fig. 3.

Annotations and functional classification of isoforms in A. gracilistylus. a Gene functional annotation in the KOG database. b Gene functional annotation in the GO database

The isoforms annotated in the NR database were further compared to the GO database, resulting in the annotation of 224,994 isoforms, which were divided into three GO categories: biological process, cellular component, and molecular function. The isoforms were mainly clustered in the biological processes “cellular process” (89,896 isoforms), “metabolic process” (85,703 isoforms), and “biological regulation” (28,321 isoforms). The isoforms mainly aggregated in the cellular components “cell” (90,525 isoforms), “cell part” (88,823 isoforms), and “membrane” (73,223 isoforms). The isoforms were mainly concentrated in the molecular functions “binding” (111,422 isoforms), “catalytic activity” (102,904 isoforms), and “transporter activity” (10,917 isoforms) (Fig. 3b).

Identification of genes involved in KA biosynthesis by KEGG analysis

A total of 296,226 isoforms in the transcriptome of A. gracilistylus were annotated in the KEGG database, including five categories (Fig. 4a): cellular processes (13,758 isoforms), environmental information processing (17,155 isoforms), genetic information processing (58,019 isoforms), metabolism (171,421 isoforms), and organic systems (11,520 isoforms), involving a total of 132 metabolic pathways.

Fig. 4.

Annotation of A. gracilistylus isoforms in the KEGG database. a Five categories of biosynthesis pathways and the number of related isoforms. b The biosynthesis pathway of terpenoids and polyketones and the number of related isoforms

A total of 1728 isoforms were annotated to the terpenoid backbone biosynthesis (ko00900, 1142 isoforms) and diterpenoid biosynthesis (ko00904, 586 isoforms) pathways, which are two metabolic pathways of KA synthesis (Fig. 4b). Based on the screening conditions of Q value < 0.05 and FPKM > 1 of isoforms, eight, one, and two isoforms were associated with the 3-hydroxy-3-methyl-glutaryl-coenzyme A reductase (HMGR), mevalonate kinase (MK), and isopentenyl diphosphate isomerase (IDI) enzymes in the MVA pathway, respectively. In the MEP pathway, there were nine, six, and four isoforms encoding the enzymes 1-deoxy-d-xylulose-5-phosphate synthase (DXS), 1-deoxy-d-xylulose 5-phosphate reductoisomerase (DXR), and 2C-methyl-d-erythritol 2,4-cyclodiphosphate synthase (MCS), respectively. In addition, there were four, forteen, three, one, and eight isoforms encoding geranyl diphosphate synthase (GPPS), GGPP synthase (GGPPS), CPS, KS, and KO enzymes, respectively, in the pathway involved in the conversion of isopentenyl diphosphate (IPP) to KA (Supplementary Table S4). The relative expression levels of the isoforms of these key enzymes in each tissue are displayed in the form of a heat map in Fig. 5a. Most of these isoforms were highly expressed in the leaves and roots, whereas the isoforms related to KO were highly expressed in the roots of A. gracilistylus and had lower expression in the leaves and stems.

Fig. 5.

Heatmap of isoforms related to KA biosynthesis in A. gracilistylus. a The expression level of key enzyme isoforms in various tissues of the KA biosynthesis pathway. b Correlation analysis between the content of KA and the expression of isoforms involved in the KA biosynthesis pathway. L, leaf; R, root; S, stem. *0.01 < p < 0.05, **0.001 < p < 0.01, ***p ≤ 0.001

The correlation analysis showed that the levels of two isoforms of HMGR (isoform_13399, isoform_292476), one isoform of MK (isoform_33932), one isoform of IDI (isoform_58522), two isoforms of CPS (isoform_144483, isoform_305763), and two isoforms of KO (isoform_18099, isoform_445625) were significantly correlated with the content of KA (Fig. 5b). The correlation coefficients between the expression levels of these isoforms and the content of KA were all greater than or equal to 0.8. These results demonstrated a strong correlation between KA content and gene expression levels, indicating that these isoforms may be the key contributors to the biosynthesis of KA.

Analysis of DEGs

In the comparison of the leaves and roots, 101,473 DEGs were identified, 47,363 of which were up-regulated and 54,110 of which were down-regulated in the roots (Supplementary Table S5). In the comparison of the stems and leaves, 69,335 DEGs were identified, 29,476 of which were up-regulated and 39,859 of which were down-regulated in the leaves (Supplementary Table S6). In the comparison of the roots and stems, 51,195 DEGs were identified, 31,986 of which were up-regulated and 19,209 of which were down-regulated in the stems (Fig. 6a, Supplementary Table S7). We identified 12,048 common DEGs among the three comparisons (Fig. 6b).

Fig. 6.

Quantity and enrichment analysis of DEGs. a Up-regulated and down-regulated DEGs in different tissues. b Venn diagram of DEGs in different comparison groups. c Enrichment of KEGG pathways for DEGs in the leaves compared to the roots. d Enrichment of KEGG pathways for DEGs in the stems compared to the leaves. e Enrichment of KEGG pathways for DEGs in the roots compared to the stems

The KEGG database was used to annotate the biological functions of DEGs. The 40,542 DEGs identified between the leaves and roots were annotated to 135 metabolic pathways, mainly enriched in carbon metabolism, plant hormone signal transduction, and plant–pathogen interaction (Fig. 6c). The 27,957 DEGs identified between the stems and leaves were annotated to 135 metabolic pathways, mainly enriched in carbon metabolism, biosynthesis of amino acids, and plant hormone signal transduction (Fig. 6d). The 21,880 DEGs identified between the roots and stems were also annotated to 135 metabolic pathways, mainly enriched in carbon metabolism, ribosome, and plant hormone signal transduction (Fig. 6e).

A total of 652 DEGs in the leaf-to-root comparison were annotated to the KA biosynthesis pathway, with 358 of the DEGs specifically up-regulated in the roots. A total of 386 DEGs in the stem-to-leaf comparison were involved in the KA biosynthesis pathway, with 160 DEGs specifically up-regulated in the leaves. A total of 373 DEGs in the root-to-stem comparison were involved in the KA biosynthesis pathway, with 180 DEG specifically up-regulated in the stems (Table 4).

Table 4.

Number of specific up-regulated isoforms between each tissue

Metabolic pathway name	Pathway number	Number of up-regulated isoforms
		Leaf vs Root	Stem vs Leaf	Root vs Stem
Terpenoid skeleton biosynthesis	ko00900	166	105	129
Diterpenoid biosynthesis	ko00904	192	55	51

Structural characterization of KO from A. gracilistylus

KO is the key enzyme in the KA synthesis pathway, which catalyzes the three consecutive oxidation reactions of ent-kaurene to finally generate KA. The ORF length of isoform_18099 in the transcriptome of A. gracilistylus was 1536 bp, encoding KO with 511 amino acids. Sequence alignment analysis of KO in A. gracilistylus showed 74.76% sequence homology with KO in Daucus carota var. sativa (DcKO, XP_017253618.1), 73.19% sequence homology with KO in Trachyspermum ammi (TaKO, AUZ98413.1), 69.25% sequence homology with KO in Camellia sinensis (CsKO, AUD40399.1), and 68.22% sequence homology with KO in Nicotiana tomentosiformis (NtKO, XP_009592896.1). Therefore, the KO gene in A. gracilistylus shows high consistency with that of other plants. The secondary structure analysis of KO in A. gracilistylus showed that the protein mainly consists of 17 α-helices and nine β-sheets (Fig. 7a, Supplementary Fig. S5).

Fig. 7.

Sequence alignment and protein structure model of KO in A. gracilistylus. a Sequence alignment and secondary structure of KO in A. gracilistylus. The substrate-binding domain ETLR, redox site PERF, and heme-binding domain FGGGKRVCAG are labeled with green, blue, and black boxes, respectively. b Cartoon model of the structure of KO in A. gracilistylus; α-helices and β-sheets are represented in cyan and red, respectively. c Active site of KO in A. gracilistylus. The redox sites, substrate-binding domains, and heme-binding domains are indicated as yellow, green, and red spheres, respectively

The three-dimensional structure of KO in A. gracilistylus was predicted using CYP76AH1 (SMTL ID: 5ylw.1.A) from Salvia miltiorrhiza as a model, which showed 33% (L60-N529) sequence similarity to the KO sequence of A. gracilistylus (Fig. 7b). KO in A. gracilistylus possesses the redox site PERF of the eukaryotic CYP450 enzyme, the substrate-binding domain ETLR, and the heme-binding domain FGGGKRVCAG (Fig. 7c).

TFs involved in KA biosynthesis

Based on the A. gracilistylus transcriptome data, 11,528 TFs were identified belonging to 58 TF families (Fig. 8a). The most abundant TF families were MYB (1252 isoforms), AP2-EREBP (1170 isoforms), and WRKY (799 isoforms). According to the functional classification of these TFs by the KEGG database, 64 TFs were found to participate in the metabolism of terpenes and polyketones, all of which belong to the forkhead-associated domain protein (FHA) family. The 64 TFs were screened using FPKM > 1 as the screening condition, and seven highly expressed TFs were obtained. Protein–protein interaction (PPI) network analysis was performed between the seven TFs with FPKM > 1 and key enzymes in the KA biosynthesis pathway. The PPI network contained 13 nodes and 61 edges, showing that DXR and DXS have high interactions with FHA (Fig. 8b).

Fig. 8.

Identification of TFs and the interaction network with key enzymes in the KA biosynthesis pathway of A. gracilistylus. a Classification of TF families to which isoforms belong. b Network interactions between key enzymes in the KA biosynthesis pathway and the FHA family. The size of the circle and the thickness of the line represent the strength of the interaction between proteins

Verification of key enzyme isoform expression levels by RT-qPCR

Six key enzyme isoforms involved in KA biosynthesis with large differences in expression levels between tissues detected by RNA-sequencing were selected for RT-qPCR validation in the three tissues: isoform_53458 (GGPPS), isoform_76283 (GGPPS), isoform_30303 (GPPS), isoform_33932 (MK), isoform_305763 (CPS), and isoform_7162 (CPS). Their expression levels in the leaves, roots, and stems of A. gracilistylus were further analyzed by RT-qPCR. Isoform_53458 and isoform_30303 had the highest expression levels in the leaves; isoform_76283, isoform_305763, and isoform_7162 had the highest expression levels in the roots; whereas isoform_33932 had higher expression in the stems but lower expression in the leaves and roots. These results were consistent with the trend of expression levels of corresponding isoforms obtained in the transcriptome data of A. gracilistylus (Fig. 9).

Fig. 9.

Gene expression analysis of six isoforms related to KA biosynthesis in A. gracilistylus (error bars indicate standard error from three biological replicates). n = 3. L, leaf; R, root; S, stem. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns, p > 0.05

Discussion

At present, second-generation sequencing is widely used to study gene expression, gene function and structure, alternative splicing, and for the prediction of new transcripts (Hu et al. 2021). However, in a single reaction, second-generation sequencing can measure only hundreds of bases, whereas third-generation sequencing can measure thousands of bases (van Dijk et al. 2018). The emergence of third-generation sequencing avoids the introduction error of PCR amplification and overcame the problem of incomplete information of species without a reference genome (Li et al. 2022). In this study, both the DNB-seq and PacBio iso-seq platforms were used to sequence the transcriptome of the leaves, roots, and stems of A. gracilistylus. A total of 358,103 genes were detected using the DNB-seq platform and the average comparison rate of the genes set was 84.20%. In addition, 505,880 isoforms were obtained through the PacBio iso-seq platform, with an average length of 948 bp, N50 length of 1552 bp, and N90 length of 586 bp. Compared with traditional transcriptome research, the transcriptome database of A. gracilistylus obtained in this study has higher reliability and better assembly quality.

Correlation analysis showed that the expression levels of eight genes encoding HMGR, MK, DXR, IDI, CPS, and KO were highly correlated with the content of KA in the tissues of A. gracilistylus. Except for isoform_292476 and isoform_33932, the expression levels of the other six genes were higher in roots than in the other tissues. This could explain the higher content of KA in roots than in the leaves and stems of A. gracilistylus. Therefore, these results provide molecular-level support that the roots are the most important medicinal parts of A. gracilistylus.

Terpenoids are biosynthesized in plants through the MVA and MEP pathways. HMGR is the rate-limiting enzyme of the MVA pathway (Lichtenthaler 2000), and DXS and DXR are the key enzymes of the MEP pathway (Jomaa et al. 1999). Yang et al. inhibited the activities of HMGR and DXR in S. miltiorrhiza with highly specific inhibitors and found that cell growth in the hairy root of S. miltiorrhiza was mainly influenced by the MVA pathway, while tanshinone was mainly synthesized by the MEP pathway (Yang et al. 2012). In the KA biosynthesis pathway of A. gracilistylus, we found eight isoforms related to HMGR and six isoforms related to DXR. Among them, isoform_13399 and isoform_292476 of HMGR were highly correlated with KA content, with correlation coefficients of 0.87 and 0.89, respectively. The isoform_20225 of DXR was significantly correlated with KA content, with a correlation coefficient of 0.70.

MK catalyzes MVA to generate mevalonate-5-phosphate, which is the key enzyme of the MVA pathway. Niu et al. cloned an MK gene (SaMK) from Santalum album. Experiments using several elicitors showed that the expression of SaMK was up-regulated by stimulation with methyl jasmonate (Niu et al. 2021). Lluch et al. analyzed the functional promoter of the MK gene in A. thaliana and found that the regulatory elements between the MK 5′-flanking regions at positions −194 and −295 were crucial for high-level expression of the gene (Lluch et al. 2000).

The conversion of isopentenyl diphosphate to dimethylallyl diphosphate (DMAPP) catalyzed by IDI is the initial step of isoprenoid biosynthesis. DMAPP loses its inorganic pyrophosphate to form isoprene and then forms short-chain isoprenoid with IPP. Chen et al. found that overexpression of IDI in Eucommia ulmoides increased the accumulation of trans-polyisoprene (Chen et al. 2012). In a study of carotenoid-deficient mutants in the fruits of Solanum lycopersicum, Pankratov et al. showed that the lack of IDI1 reduced the concentration of carotenoids in the cotyledons, fruits, and flowers, but did not decrease the concentration of carotenoids in the mature leaves (Pankratov et al. 2016). These findings highlighted the role of IDI in plastids in optimizing the ratio of IPP and DMAPP as precursors of different downstream isoprenoid pathways. The current findings further indicate that increasing the expression level of IDI in plastids may promote the synthesis of KA in A. gracilistylus.

Misra et al. identified two forms of CPS (ApCPS1 and ApCPS2) in a study of the ent-labdane-related diterpene biosynthesis pathway of Andrographis paniculata (Misra et al. 2015). Their findings suggested that ApCPS1 may play a role in primary metabolism, possibly by providing a precursor for the biosynthesis of the plant hormone GA. ApCPS2 was suggested to participate in the tissue-specific accumulation of ent-labdane-related diterpene-specific metabolites. There are two CPS genes in rice. OsCPS1 is involved in the biosynthesis of GA, while OsCPS2 is involved in the biosynthesis of the plant antitoxins phytocassanes A-E and oryzalexins A-F. Using RT-qPCR analysis, Toyomasu et al. showed that the transcripts of OsCPS1 were mainly located in the vascular bundle tissues, while OsCPS2 transcripts were mainly located in epidermal cells. The biological functions of these two CPS genes are different, with OsCPS1 and OsCPS2 required for rice growth and defense, respectively, resulting in different positions of their transcripts in rice (Toyomasu et al. 2015).

KO is a key enzyme involved in three consecutive oxidations to obtain the final product during the synthesis of KA (Hu et al. 2014). Mutations in rice seeds encoding KO genes attenuated the enzymatic activity of the reaction converting ent-kaurene to KA, leading to reduced KA synthesis (Zhang et al. 2020). Itoh et al. isolated five KO-like (KOL) genes from the genome of the semi-dwarf cultivar Tan-Ginnbozu (d35Tan-Ginnbozu) of rice (Itoh et al. 2004). Sequence analysis and complementary experiments showed that OsKOL2 corresponded to D35, while the homozygous cells with D35 allele deletion showed a serious dwarf phenotype. KO in A. gracilistylus belongs to the P450 lineage of the cytochrome superfamily and has a series of characteristic domains: ETLR (E376-R379) is the substrate-binding domain, PERF (P431-F434) is the redox site of eukaryotic CYP450, and FGGGKRVCAG (F450-G459) is the CYP450 cysteine (C) heme ferro ligand-binding site. This was consistent with the reported characteristic domains in Carya cathayensis (Liang et al. 2020), Pyrus pyrifolia (Li et al. 2010), and Scoparia dulcis (Yamamura et al. 2018), indicating that KO is highly conserved in vascular plants.

In summary, the discovery of candidate isoforms related to the biosynthesis pathways of KA is meaningful for improving the effective components of A. gracilistylus through gene regulation. Among these candidate isoforms, HMGR (isoform_13399, isoform_292476), IDI (isoform_58522), and KO (isoform_18099) showed the highest correlation coefficients between the FPKM value and the relative content of KA. This means that high expression of these four genes may promote KA synthesis. The FHA TF family has a typical FHA domain, which is the only phosphoprotein binding domain with strict specificity for phosphothreonine (Durocher et al. 1999). FHA is closely related to stress adaptation and hormone transduction during plant growth and development (Wang 2023). The PPI network showed that DXR and DXS had high interactions with FHA, indicating that DXR and DXS may also play a role in plant stress resistance.

Conclusion

In this study, the full-length transcriptome database of the leaves, roots, and stems of A. gracilistylus was constructed by combining the DNB-seq and PacBio iso-seq techniques for the first time. In total, 505,880 full-length isoforms were obtained and 408,954 of the isoforms were successfully annotated in seven databases. Sixty isoforms encoding 11 key enzymes involved in the biosynthesis pathway of KA in A. gracilistylus were identified. The isoforms encoding the key enzymes HMGR (isoform_13399, isoform_292476), IDI (isoform_58522), and KO (isoform_18099) might play key roles in the accumulation of KA in A. gracilistylus. Moreover, putative TFs involved in KA biosynthesis in A. gracilistylus were discovered. In conclusion, this study will facilitate further research on the A. gracilistylus functional genome and provides important insights into the mechanism of KA biosynthesis in A. gracilistylus.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Project design: QY and JW. Experiments and data analysis: BH, TS, JX, XZ, JZ and RH. Manuscript preparation: BH. Manuscript revision: QY and JW. Sample preparation: BH. All authors read and approved the final manuscript.

Funding

This work was supported by the Investigation and Development of Commonly Used Traditional Chinese Medicine Resources in the Dabie Mountain Area (Grant Number 2021HZ035, RH2200001421), the Natural Science Foundation of Anhui Province of China (Grant Number 2008085MH268), and the Natural Science Research Grant of Higher Education of Anhui Province (Grant Number 2022AH050474).

Data availability

The RNA-seq datasets from three A. gracilistylus tissues were deposited in the NCBI Sequence Read Archive (SRA) database (Accession: PRJNA945217).

Declarations

Conflict of interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.