Transcriptome Characterization of Panax Quinquefolius L. seedlings in Response to Arbuscular Mycorrhizal Fungi using RNA-seq

Panax quinquefolius L. has been considered as an important traditional Chinese medicine with a history of more than 300 years in China. Ginsenoside is the main bioactive component. Our research group has found that the accumulation of ginsenoside could be affected by arbuscular mycorrhizal fungi (AMF). However the underlying mechanism how AMF affected the biosynthesis of ginsenoside in P. quinquefolius is still unclear. In this study, the RNA-seq analysis was used to evaluate the effects of AMF (Rhizophagus intraradices, R. intraradices) on the expression of ginsenoside synthesis related genes in P. quinquefolius root. The results indicated that a symbiotic relationship between R. intraradices and P. quinquefolius was established. RNA-seq achieved approximately 48.62 G reads of all samples. Assembly of all the reads involved in all samples produced 63420 transcripts and 24137 unigenes. Differential expression analysis was performed between the control and AMF group. A total of 111 differentially expressed genes (DEGs) in response to AMF vs control were identied, 78 and 33 transcripts were upregulated and downregulated, respectively. Based on the functional analysis, Gene ontology (GO) analysis revealed that most DEGs were related to stress responses and cellular metabolic processes. The Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis identied transduction, plant hormone signal transduction and terpenoids and polyketides biosynthesis pathways. Furthermore, the expression of glycolysis-related genes and ginsenoside synthesis related genes was largely induced by AMF. In conclusion, our results comprehensively elucidated the molecular mechanism how AMF affected the biosynthesis of ginsenoside in P.quinquefolius by transcriptome proling. this the results showed that a symbiotic relationship between AMF and P. quinquefolius was established. Compared with non-AMF inoculation, a total of 111 differential DEGs in the roots of P. quinquefolius were identied and 78 DEGs were up-regulated with the AMF treatment. The expression level of genes involved in glycolysis and ginsenoside biosynthesis showed signicantly changes during the AMF inoculation. Our data elucidated the mechanism of AMF induced secondary metabolite synthesis of P. quinquefolius and provided new insights for improving the quality of ginsenosides of P. quinquefolius.


Introduction
Panax quinquefolius L., which belongs to the Araliaceae family, has been commonly known as American ginseng and considered as one of the most valuable traditional Chinese medicines with a history of more than 300 years in China (China Pharmacopoeia Committee 2015; Schlag and McIntosh 2013). American ginseng has been recorded as "cool" and listed as superior with a tonic effect in "Chinese Materia Medica" (Schlag and Rydlová et al. 2016). The expression of related genes involved in secondary metabolic pathway was regulated by AMF, resulting in a signi cant increase in the content of active components in medicinal plants (Rydlová et al. 2016; Andrade et al. 2013). For instance, the expression of genes encoding tryptophan decarboxylase (TDC), strictosidine synthase (STR), deacetylvindoline (DAT), putrescine nmethyltransferase (PMT) in the biosynthesis of alkaloid in Catharanthus roseus was signi cantly improved by AMF (Andrade et al. 2013). Zhang et al. (2020) found that AMF could promote the accumulation of avonol-glycosides by regulating the expression of phenylalanine ammonia-lyase (PAL), 4-coumarate CoA ligase (4CL), chalcone synthase (CHS), glucosyltransferase (GT) and rhamnosyltransferase (RT) genes in avonoid biosynthesis in Anoectochilus roxburghii. In genus of Panax, it has been observed that the content of total ginsenoside in Panax ginseng was increased by Glomus intraradices (Tian et al. 2019;Cho et al. 2009). While the effects of AMF on the accumulation of ginsenosides in P. quinquefolius have been poorly understood. Our research group has showed that a positive correlation between the mycorrhizal colonization (Rhizophagus intraradices) and the content of ginsenoside in P. quinquefolius has been successfully obtained based on the eld investigation (Ran et al. 2020). Here, we hypothesized that AMF could improve the biosynthesis of ginsenoside by regulating the expression of ginsenoside synthesis related genes, which might be a new effective strategy to promote the quality of American ginseng. Based on the RNA-Seq data, Mandal et al. (2015) signi ed that compared with non-AMF, AM symbiosis upregulated the transcription of eleven steviol glycosides (SGs) biosynthesis genes. Bruisson et al. (2016) showed that the expression of phenylalanine ammonia-lyase (PAL), stilbene synthase (STS) and ROMT (encoding a resveratrol O-methyltransferase) genes involved in the stilbenoid biosynthesis pathway was up-regulated in Grapevine by inoculation with AMF using the RNA-SEq. However the underlying molecular mechanism that how AMF in uences the secondary metabolites of P. quinquefolius has been rarely reported. Therefore, in this study, the AMF (R.intraradices) and P. quinquefolius were used to establish the AM symbiosis, and the molecular mechanism of AMF improving the accumulation of ginsenoside in P. quinquefolius will be studied using the transcriptome technique, which will provide basic data for the elaboration of the mechanism of AMF regulated secondary metabolites.

Experimental materials and design
The mycorrhizal fungus Rhizophagus intraradices (R.intraradices (No. BGC BJ09)) used in this study was purchased from the Institute of Plant Nutrition and Resources, Beijing Academy of Agricultural and Forestry Sciences, China. The inoculum was composed of spore (approximately 500 spores per gram inoculum), hypha, host root segment and culture medium.
Seeds of P. quinquefolius were sterilized with 10% H 2 O 2 for 30 min, washed with sterilized water for three times, and then germinated at 25°C in an incubator after soaking for 24 h. There were two treatments in the present study, including control and AMF treatment (R. intraradices). For treatments, a mixture of soil and sand (2:1, V:V) was used as growth medium, then autoclaved at 121°C, 0.10 MPa for 2 h to kill indigenous AMF propagules and other microorganisms. The selected soil belongs to the Meadow soils (Semiaqueous soil Order) according to the China soil classi cation system. Approximately 3.50 kg of the medium and 30 g of inoculum were mixed, put in a plastic pot (180×145×155 mm) for the growth of P. quinquefolius. There were 3 replicates of each treatment. The non-AMF pot was added with the same amount of inactivated inoculum. We also added 30 ml of ltered inoculant (0.25mm lter membrane), which was free from mycorrhizal propagules to the non-AM treatment to provide the same microorganism biota. The plants were grown in a day/night temperature of 25/20°C, a relative day/night humidity of 70/65%, a day/night of 16/8 h and a photosynthetic photon ux density (PPFD) at the height of the plant of 100 μmol m −2 s −1 . Light was provided by a uorescent lamp. Pots were repositioned weekly to reduce environmental error. All pots were irrigated with 100 ml Hoagland solution with 1/10 P (0.1 mM) strength. The fresh roots of P. quinquefolius seedlings were harvested at 6 months old. Each plant of P. quinquefolius was divided into branch root and taproot The branch root of each P. quinquefolius was used to detect the colonization of AMF and the corresponding taproot was further divided equally into two parts, an aliquot of the taproot was used to determine the content of ginsenoside, and another aliquot was frozen in liquid nitrogen and stored at -80℃ for further analysis.
The analysis of AMF colonization in roots of P. quinquefolius The AMF colonization in roots of P. quinquefolius was determined according to the method reported by Shu et al. (2016). The branch roots were removed from the FAA (Formalin 5 ml + Acetic Acid 5 ml + Alcohol 90 ml) solution, washed several times with distilled water, and then placed in a test tube.
Subsequently, root samples were added with 10% (w/v) KOH at 99°C for 1.5 h, stained with 0.05% (w/v) trypan blue, and estimated the mycorrhizal colonization using a modi ed line intersect method (Rufykiri et al. 2000;McGonigle et al. 1990). 30 root segments per root sample were observed for the presence of AMF structures with repeated three times. The degree of root infection by AMF was assessed using a light microscope (Olympus-BH-2, Japan). The AMF infection was calculated according to the following formula: frequency of AM fungal colonization = (Infected root segments / All root segments)×100%. The AMF colonization intensity according to the following formula: AMF colonization indensity = (95×n 5 +70×n 4 +30×n 3 +5×n 2 +n1) / All root segments×100%, n 5 represented the number of root segments infected at level 5; n 4 represented the number of root segments infected by level 4; n 3 represented the number of root segments infected at level 3, and so on.

Analysis of the content of ginsenoside
The content of ginsenoside was conducted using the methods reported by Liu et al. (2014) with slightly changes. A total of 0.50 g air-dried P. quinquefolius roots were ground into a powder, with 50 ml methanol for extraction. The extracts were obtained with ultrasound for 40 min, and then left to stand for 10 min. The supernatant was ltered with a 0.45 mm microporous membrane and injected with 10 ul for HPLC analysis. The content of ginsenoside was determined using a Kromasil C 18  Total RNA Extraction, cDNA library construction and RNA Sequencing Total RNA was extracted from control and AMF treated root samples using the RNA prep Pure Plant Kit (Tiangen Biotech Co., Ltd., Beijing, China) based on the manufacturer's manual. DNA was then removed using the RNAclean Kit (Tiangen, Beijing, China). The quality and quantity of total RNA was assessed with the Agilent 2100 Bioanalyzer (Agilent Technologies, CA, USA). The library construction and RNA-Seq assay were performed by the Novogene Biotechnology Corporation (Beijing, China). The cDNA library was constructed using A TruSeq™ RNA Sample Preparation Kit (Illumina, Inc.). First, Poly-(A)-containing mRNA was puri ed from the total RNA using oligo (dT) magnetic beads and Oligotex mRNA kits (Qiagen, Germany), following the manufacturer's instructions. Fragmentation was carried out using divalent cations. Fragmentary RNAs were used as template for rst strand cDNA synthesis by an cDNA preparation kit (Illumina, San Diego, CA, USA).
Second-stranded cDNA was synthesized using RNase H and DNA polymerase I. Then cDNAs were subjected to end-repair, phosphorylation, and ligation to sequencing adapters. Afterward, the products enriched by PCR ampli cation were puri ed through 2% agarose gelelectrophoresis and quanti ed by TBS380 (Picogreen). Finally, cDNA libraries were subsequently sequenced using an Illumina HiSeq™ 2000 platform.
Transcriptome assembly, annotation and function enrichment Before assembly, raw data were ltered to remove reads containing adapter, reads containing ploy-N, and low quality reads to generate highquality clean data. At the same time, Q20, Q30, GC-content, and sequence duplication level of the clean data were calculated. After removing the low quality reads, a denovo strategy was used to assemble the clean reads into distinct contigs using Trinity assembly software (Grabherr et al. 2011). All of the contigs were clustered, and the contigs with the longest sequences were de ned as unigenes. For functional annotation, the unigenes, through a BLASTall algorithmbased program with a threshold of E-value < 10 −10 , were searched against different databases, including NCBI non redundant protein sequences (Nr), NCBI non-redundant nucleotide sequences (Nt), Protein family (Pfam), Protein primogenomic cluster (KOG/COG), Arti cial annotated and reviewed protein sequence database (Swiss-Prot), KEGG Ortholog database (KO), Gene Ontology (GO).

Analysis of differential genes
The expression level of gene was determined using the the reads per kilobase per million mapped reads (RPKM) method (Trapnell et al. 2010) by RSEM software (Dewey and Li 2011). The transcript RPKM values were estimated using RNA-Seq by Expectation Maximization (RSEM) with Bowtie read mapping. Differentially expressed genes (DEGs) analysis in pair-wise comparisons was conducted using the DESeq software (Anders and Huber 2010), and the input data of the differentially expressed genes (DEGs) was based on the read counts. The false discovery rate (FDR) method was applied to correct the threshold of the P values in multiple tests for identifying the differences between two groups. A FDR < 0.001 and an absolute value of log2 (ratio) > 2 were applied as thresholds to identify signi cant differences between two groups (Mortazavi et al. 2008). Furthermore, the DEGs were then analyzed through GO and KEGG pathway enrichment analyses.

Simple sequence repeats (SSRs) detection and primer design
In order to check SSRs in P. quinquefolius, the microsatellite identi cation tool (MISA) was used to identify the SSR. The parameters were used to distinguish di-, tri-, tetra-, penta-and hexa nucleotide motifs with the lowest limit of repeats of 6, 5, 4, 4 and 4 respectively, and Primer 3 software was used to devise primers for each SSR. The design parameters of major primer were set as follows: 100 to 300 nt of PCR products, 18 to 24 nt of primer lengths, 60°C optimal annealing temperature, and 40% to 65% of GC content from (Zeng et al. 2019).
Quantitative real-time RT-PCR A total of 5.0 ug RNA of each treatment root sample was used for rst-stand cDNA synthesis using the RNA prep Pure Plant Kit (Tiangen Biotech Co., Ltd., Beijing, China) based on the manufacture's instructions. The synthesised cDNA was used as a template for quantitative real-time polymerase chain reaction (qRT-PCR). The qRT-PCR reaction system with 20 ml volume of reaction mixture made up of 2 ml of dilute template cDNA, 2 ml of primer pairs, 10 ml of GoTaq ® qPCR Master Mix (Promega, USA), and 6 ml of deionized water. The setting program of qPCR reaction was carried out as follows: denaturation at 94°C for 2 min, followed by 35 cycles of 94°C for 30 s, annealing at 57°C for 1 min, 72°C for 50 s and then extension at 72°C for 1 min (Wang et al. 2016). After the reactions, the speci city was assessed by melt curve and size estimation of the ampli ed product. Each gene was quanti ed using three biological replicates. The relative expression of the selected genes was normalized by using the 2 −ΔΔCt method (Livak and Schmittgen 2001). β-Actin was used as internal standard and was ampli ed with the forward primer 5′-AGGAACCACCGATCCAGACA-3′and reverse primer 5′-GGTGCCCTGAGGTCCTGTT-3′. The gene primers used for RT-PCR are shown in Table S1.

Statistical analysis
The data were subjected to analysis of one-way ANOVA using Statistical Product and Service Solutions 17.0 software (SPSS Institute Inc. Chicago, IL, USA) and differences were compared by Duncan's test with a signi cance level of P < 0.05.

AM colonization of P. quinquefolius roots
The colonization of AMF in the P. quinquefolius roots is shown in Fig. 1. Roots of inoculated P. quinquefolius were extensively colonized with treatment of R. intraradices. Excellent symbiotic relationship between R. intraradices and P. quinquefolius was established with the formation of vesicles and hypha. The frequency of AM fungal colonization and AMF colonization indensity in P. quinquefolius root reached 87.78% and 57.50% under the R. intraradices inoculation respectively (Fig S1). No mycorrhizal colonization was observed in control.

Changes in ginsenoside content under AMF inoculation
The content of ginsenosides, including Rb 1 , Rg 1 , Re and Rb 2 , was determined during the AMF inoculation (Fig. 2). The content of Rb 1 was signi cantly increased by R. intraradices, which was estimated as 26.15% higher (P < 0.05) than that of control respectively. It was observed that the content of Re was also markedly enhanced by 22.38% (P < 0.05) by R. intraradices. No signi cant differences in content of Rg 1 and Rb 2 was detected between non-AMF and AMF colonized plant (P > 0.05).
Sequencing, assembly and splicing A total of 172600168 raw reads in AMF inoculation and 156278548 raw reads in the control were obtained. After ltering out the low-quality reads, there were 170350412 clean reads in the AMF inoculation and 153757750 clean reads in the control. A total clean base number of 25.55 G (AMF) and 23.07 G (the control), was obtained respectively. The quality of clean reads included an error of less than 0.03%, the percentage of Q30 base was 92.21% or higher, and the G/C content was approximately 43%. 267266 transcripts were obtained by Trinity software with average lengths of 1291 bp and N50 lengths of 1854 bp (Table 1 and Fig S2a). Furthermore, 75031 unigenes with a mean length of 1137 bp and N50 length of 1729 bp were assembled. Among these unigenes, 24137 (32.17%) were within the range of 300-500 bp, 22444 (29.91%) were within 500 bp-1 kbp, 16394 (21.85%) were within 1-2 kbp, and 12056 (16.07%) were longer than 2 kbp (Table 2 and Fig S2b). Functional annotation and classi cation In order to obtain comprehensive gene function information, the unigenes were successfully annotated using seven public databases (Nr, Nt, Pfam, KOG/COG, Swiss-prot, KEGG, GO). As shown in Table 3,  KOG annotation showed that a total of 10315 unigenes were classi ed into 26 categories (Fig. 3). Within these genes, the cluster of post translational modi cation represented the largest group (O) (1353, 13.12%), followed by translation, ribosomal structure and biogenesis (J) (1245, 12.07%), general function prediction only (R) (938, 9.26%), RNA processing and modi cation (688, 6.67%) and intracellular tra cking, secretion, and vesicular transport (660, 6.40%). In addition, 121 unigenes (1.17%) were related to secondary metabolites biosynthesis, transport and catabolism and 79 as defense mechanism unigenes (V).
All unigenes were classi ed into three GO categories, including biological process (BP), cellular component (CC), and molecular function (MF) (Fig. 4). For the biological process category, the majority of the GO terms were assigned to cellular processes (2487) and metabolic processes (1098). With the cellular components category, 953 and 2803 were classi ed to cell and cell parts respectively. For molecular function, the assignments were mostly binding (675) and catalytic activity (1013). Transcripts related to GO term binding were abundant in the molecular function category.
KEGG database was used to systematically analyze the gene function associated with genomic information. For AMF inoculation, a total of 13806 transcripts were classi ed into 19 KEGG pathways ( Fig. 6). As shown in Fig. 5, the majority of KEGG assigned transcripts were involved in transduction (1569), a large pool of transcripts were attributed to the area of carbohydrate metabolism biosynthesis (1201), additional transcripts were mapped to the area of lipid metabolism biosynthesis (651) and metabolism of terpenoids and polyketides biosynthesis (318).

Analysis of SSRs
In order to further evaluate the quality of assembly, MISA software was used to perform simple sequence repeat (SSR) analysis based on the selected unigene (length > 1 kb). We successfully screened 4958 SSRs, including mono-, di-, tri-, tetra-, penta-and hexa-nucleotide SSRs (Fig. 6). The di-nucleotide SSR has been regarded as the most abundant, followed by mono-and tri-nucleotide SSRs. Table S2 has showed a series of primer pairs for each SSR that match the primer design parameters to further investigate the potential of these SSRs as genetic markers.

Identi cation and analysis of DEGs
A total of 111 DEGs (78 up-regulated, 33 down-regulated) were identi ed in the AMF vs control comparison (Table 4). To further understand the speci c processes at the transcriptional level as well as the gene functions and interactions, GO and KEGG enrichment analyses were performed in this study. Based on the GO enrichment, DEGs were signi cantly (KS value ≤ 0.05) enriched in 10 terms of biological processes (Table 5). These enriched terms were highly linked to biological processes in the adaptive response to AMF inoculation, such as homoserine metabolic process (GO:0009092), L-methionine biosynthetic process (GO:0071265) and L-methionine biosynthetic process from L-homoserine via cystathionine (GO:0019279). Further genes were identi ed from the DEGs data by KEGG annotations, which might be assigned to nine pathways in the KEGG database. The KEGG pathway analyses revealed that the most signi cant metabolic pathways enriched and the most enriched differential genes were Plant hormone signal transduction and Ribosome (Fig. 7). Additionally, Venn diagrams were constructed for displaying the number of shared and exclusively expressed genes between the AMF and control groups. There were 68541 and 66759 expressed genes were detected from the AMF group and control group, respectively. Among them, 61315 were shared, 7227 were exclusive to the AMF treatment, and 5445 were presented only in the control (Fig. 8).  Identi cation of glycolysis-related DEGs Glycometabolism has been regarded as an important primary metabolic process. A large number of DEGs in P. quinquefolius were homologous with well-known related genes of glycosylsis in model plants. In detail, seven genes encoding for hexokinase (HK), two genes encoding for phosphoglycerate isomerase (PGI), four genes encoding for 6-phosphofructokinase (PFK), eight genes encoding for fructosebisphosphate aldolase (FBA), four genes of triosephosphate isomerase (TPI), 11 genes of glyceraldehyde 3-phosphate dehydrogenase (GAPDH), two genes of phosphoglycerate kinase (PGK), three genes encoding for probable phosphoglycerate mutase (GPM), six enolase (ENO) genes and ten pyruvate kinase (PK) genes were identi ed in P. quinquefolius (Fig. 9). Moreover, the changes of glycolysis-related genes expression with AMF treatment were analyzed. The result showed that most of the genes involved in glycolysis were largely up-regulated during the AMF inoculation, suggesting that AMF could induce the rising of glycolysis in P. quinquefolius.

Identi cation and analysis of DEGs
Genes involved in biosynthesis of ginsenoside with AMF treatment were retrieved from the DEG group according to the KEGG annotation. As shown in Fig. 10, the analysis of the transcriptome data showed that 29 unigenes coded for six enzymes in the mevalonate (MVA) pathway were located in the cytoplasm and that 25 unigenes coded for seven enzymes in the 2-C-methyl-D-erythritol 4-phosphate (MEP) pathway were located in plastid, which synthesized the common precursor isopentenyl diphosphate (IPP) for downstream terpenoid pathways. In MVA pathway, four genes of Acetyl-CoA acetyltransferase (AACT), four hydroxymethylglutaryl-CoA synthase (HMGS) genes, eight genes coded for hydroxymethylglutaryl-CoA reductase (HMGR), one mevalonate kinase (MVK) gene, six genes coded for dioxyphosphate methylhydroxyvalerate kinase (PMK), one diphosphomevalonate decarboxylase (MVD) gene, were identi ed in P. quinquefolius. In MEP pathway, eight encoding genes of 1-deoxy-D-xylulose-5-phosphate synthase (DXS), three 1-deoxy-D-xylulose 5-phosphate reductoisomerase (DXR) encoding genes, one 2-Cmethyl-D-erythritol 4-phosphate cytidylyltransferase (CMS) gene, two genes coded for 2-C-methyl-Derythritol 2,4-cyclodiphosphate synthase (MCS), four genes of HDS and ve IDS genes were identi ed in P. quinquefolius. One unigene was predicted to code for isopentenyl-diphosphate delta-isomerase (IPPI), which catalyzes the isomerization of IPP to dimethylallyl diphosphate. Seven geranylgeranyl diphosphate synthase GPS genes, three squalene e-poxidase (SS) genes and seven squalene synthetase (SE) genes were identi ed in downstream terpenoid pathways (Fig. 10a). For the key enzyme genes involved in MVA pathway (Fig. 10b), the result revealed that a number of genes were up-regulated during AMF inoculation, especially, the expressions of HMGR, GPS and SS genes were induced nearly two-fold with AMF inoculation. The key enzymes of the MEP pathway did not change signi cantly between the AMF and control (Fig. 10c). These data indicated that AMF might promote the ginsenoside biosynthesis and the formation of secondary metabolites of P. quinquefolius mainly activate MVA pathway.

qRT-PCR validation analysis
In order to verify the changes of expression level of several key enzymes in ginsenoside biosynthesis, a qRT-PCR assay was conducted. A total of 9 key enzymes genes (HMGR, PMK, GPS, FPS, DXS, DXR, SS, SE and DS) were chosed to test the RNA-seq data in AMF treatment ( Fig S3). As shown in Fig S3, (Ran et al. 2020). In present work, we found that the a symbiotic relationship between R. intraradices and P. quinquefolius was established, and R. intraradices signi cantly promoted the content of ginsenoside, which laid a foundation to further reveal the effect of AMF on the secondary metabolism in P. quinquefolius. To investigate the mechanism of AMF in uencing the secondary metabolism in P. quinquefolius, transcriptome analyses were performed in this study. Based on the evaluation of all transcripts, a number of key enzyme genes related to the ginsenoside biosynthesis of P. quinquefolius were obtained, which would provide a basis for further research the molecular mechanisms of AMF to promote the ginsenosides content in P. quinquefolius. Zhao et al. (2014) also found that mycorrhizal fungi could induce the biosynthesis of secondary metabolites and hormone balance of Cymbidium hybridum based on the transcriptome analysis. In addition, several DEGs involved in plant hormone signal transduction have been found. We speculate that there may be some signal molecule communication between AMF and P. quinquefolius, which will promote the colonization of AMF. Tsuzuki et al. (2016) indicated that AMF also secreted signaling molecules to the host plants including lipochitooligosaccharides and chitin oligomers. The possible mechanisms of signal recognition, exchange and transmission between P. quinquefolius and AMF need to be further studied. To our knowledge, it was the rst study to investigate the transcriptomic responses of P. quinquefolius to AMF inoculation. Next generation sequencing techniques could be conveniently used for analysis of transcriptome pro ling during plant/fungus interactions (Wang et al. 2009). In summary, RNA-Seq technology not only provides a comprehensive gene expression change for P. quinquefolius in response to AMF inoculation, but also provides a new technical means for further analysis the regulation of secondary metabolism by AMF.
Glycolysis has been considered to be the key way to convert glucose into pyruvate, which provides metabolic intermediates for other pathways and energy for organisms (Yang et al. 2018). In general, the transformation of fructose into pyruvate through glycolysis requires the catalytic reaction of many enzymes, which not only act as catalyst and energy regulator, but also act as a signal transmission device to respond to environmental changes (Khanna et al. 2014). In order to study whether the mycorrhizal fungi could affect the glycolysis of P. quinquefolius, all related enzymes unigenes involved in the glycolysis pathway were identi ed by transcriptomes. The results showed that the genes involved in glycolysis of P. quinquefolius were induced by AMF inoculation and most genes were largely upregulated. Thus, under the treatment of AMF, the expression of genes in glycolysis was accelerated, and the metabolism of P. quinquefolius might be changed.
Ginsenosides are supposed to be the main secondary metabolites in P. quinquefolius and play signi cant impacts on the quality of P. quinquefolius (Liu et al. 2008). In order to investigate the mechanism of AMF regulating ginsenoside synthesis, we have analysised the gene changes in the biosynthesis of triterpenoid saponins pathway in P. quinquefolius at the transcription level. From our investigation, the expression of ginsenoside biosynthesis-related genes was changed and the most abundant up-regulated unigenes were assigned to HMGR and GPS in response to AMF inoculation. HMGR has been considered as an important rate-limiting enzyme in MVA pathway, which affects the biosynthesis of ginsenoside by affecting the production of its precursor IPP and DMAPP (

Conclusion
In this work, the results showed that a symbiotic relationship between AMF and P. quinquefolius was established. Compared with non-AMF inoculation, a total of 111 differential DEGs in the roots of P. quinquefolius were identi ed and 78 DEGs were up-regulated with the AMF treatment. The expression level of genes involved in glycolysis and ginsenoside biosynthesis showed signi cantly changes during the AMF inoculation. Our data elucidated the mechanism of AMF induced secondary metabolite synthesis of P. quinquefolius and provided new insights for improving the quality of ginsenosides of P. quinquefolius.