Plant materials: The annual seedlings of P. notoginseng used for treatments were bought from Miaoxiang Sanqi standard plantation that located in Zhuang-miao Autonomous Prefecture of Wenshan Yunnan Province, China. These seedlings had homogeneity in growth status and biomass. AMF strains: Spores of Glomus intraradices and G. etunicatum mixed in the soil were purchased from the Institute of Root Biology, Wuhan Yangtze University. After morphological identification, the AMF spore soil was mixed into the sterilized soil culture substrate at a ratio of 1:20, and then the seeds of maize and clover were planted in the soil, of which the root systems were employed for spore propagation in semi closed conditions.

Experimental groups: In total, five treatment groups treated with one or two factors of AMF, MeJA and SHAM, including AMF, AMF-MeJA, AMF-SHAM, MeJA and SHAM, and one blank control group (CK) were set up, among which the AMF was a long-term treatment group for five months, and the MeJA and SHAM were short treatments for one month before sampling. The P. notoginseng seedlings were randomly assigned to groups, and each group contained more than 30 plants. The cultivation substrate used in this experiment was composed of nutrient soil and vermiculite (18:1), which were mixed evenly and sterilized with intermittent humid heat (121 °C, 2 h for twice). After drying, the sterilized cultivation substrate was subpackaged into clean pots, and each pot contained 1.5 kg mixed soil. Before planting, the P. notoginseng seedlings described above were rinsed with purified water to remove the attached soil and sprayed with benomyl to inhibit the indigenous strains. Three P. notoginseng seedlings were planted into each pot, and 15-20 g AMF soil containing about 100 spores (G. intraradices:G. etunicatum = 1:1) was placed 1~2 cm around roots. After 5 months, P. notoginseng seedlings were treated three times by irrigating roots with 20 mL of 100 µmol/L MeJA or SHAM every time, in which SHAM is a specific inhibitor of LOX in the biosynthetic pathway of JA.

P. notoginseng has strict requirements for the environmental conditions. The seedlings were planted in a greenhouse at 25 ± 3°C and 60 ± 5% relative humidity, avoiding direct sunlight and water accumulation and maintaining ventilation. After six months, the fresh roots were harvested and used for RNA extraction, transcriptome sequencing and the determination of PNS contents and activities of key enzymes involved in the biosynthesis of saponins and JAs. The remaining samples were stored at -80°C.

The biomass of P. notoginseng, including plant height, total fresh weight and fresh root weight, were measured directly, and the notoginsenoside contents were determined using high performance liquid chromatography (HPLC). Dried sample powder of 0.6 g was placed in 25 mL volumetric bottle, with 70% methanol at constant volume to the scale. Saponins were extracted for 45 min using an ultrasonic extraction method, then the solution was continuously filtered with 0.45 μm filter paper and 0.22µm filter membrane, respectively (Cao et al., 2023a). A binary gradient elution system consisted of acetonitrile (A) and water (B), and the separation was achieved using the program listed in Supplementary Table 1 . The parameters of flow rate, injection volume, detection wavelength and column temperature were 1 mL/min, 20 μL, 203 nm and 30°C, respectively. Linear regression equations of ginsenosides Rg1, Rb1, Rd, Re and notoginsenoside R1 were generated and are listed in Supplementary Table 2 .

The concentrations of five phytohormones including JA, MeJA, jasmonoyl-L-isoleucine (JA-Ile), 12-oxophytodienoic acid (OPDA) and salicylic acid (SA) were determined using an ultra-performance liquid chromatography (UPLC) series SCIEX-6500Q trap mass spectrometer (MS/MS). Approximately 50 mg of fresh root samples were ground to powder with liquid nitrogen, and then added 10 µL of 100 ng/mL corresponding internal standard material and 1 mL mixed extractant of methanol/water/formic acid (15:4:1), followed by centrifugation, concentration and redissolution (Floková et al., 2014; Li et al., 2016). Chromatographic separation of five types of phytohormones was performed using a Waters ACQUITY UPLC HSS T3 C18 column (1.8 µm, 100 mm×2.1 mm). Phase A (water/0.4% acetic acid solution) and phase B (acetonitrile/0.4% acetic acid solution) of LC–MS/MS were utilized to conduct gradient elution at a constant flow rate of 0.35 mL/min at 40°C, and the injection volume was 2 µL (Cai et al., 2014; Niu et al., 2014; Xiao et al., 2018). The elution program was performed according to Supplementary Table 3 . In this experiment, the electrospray ionization (ESI) temperature was set at 550°C, and the mass spectrum voltage was set at 5,500 V in positive ion mode and -4,500V in negative ion mode. The curtain gas (CUR) was set at 35 psi (Pan et al., 2010; Cui et al., 2015; Šimura et al., 2018). Linear regression equations of JA, MeJA, JA-Ile, OPDA and SA were generated and are listed in Supplementary Table 4 .

The activities of key enzymes involved in notoginsenoside biosynthesis, including SS, SE, CYP450 and GT, were determined using plant ELISA kits (SS, SE and CYP450: 96T, Jiangsu Meimian Industrial Co., Ltd, China; GT: 96T, Quanzhou jiubang Biotechnology Co., Ltd, China) (Yin et al., 2017; Wang L. et al., 2021; Zhang et al., 2021). The activities AOC and LOX are two main key enzymes that are responsible for the biosynthesis of JA, and their activities were measured according to the instructions of the plant allene oxide cyclase enzyme activity assay ELISA Kit (96T, Quanzhou Jiubang Biotechnology Co., Ltd, China) and plant lipoxygenase enzyme activity assay EELISA Kit (96T, Quanzhou Jiubang Biotechnology Co., Ltd, China), respectively (Qiu et al., 2020).

Three replicate fresh root samples were randomly selected from each group, stored in liquid nitrogen and then sent to MetWare (Wuhan Metwell Biotechnology Co., Ltd., China) for RNA sequencing (RNA-Seq). When the integrity, purity and concentration of RNA were qualified, double-stranded cDNA could be synthesized using kits and then sequenced using the Illumina HiSeq platform.

The raw data were quality controlled by fastp software (Chen et al., 2018). After sequencing data filtering, GC content distribution checking and sequencing error rate checking, clean reads were acquired for de novo assembly. Original RNA-sequence data were deposited in the National Center for Biotechnology Information (NCBI) with the accession number PRJNA953146. All unigenes were searched against the NCBI protein nonredundant (NR), Karyotic Orthologous Groups (KOG), Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) databases using DIAMOND BLASTX with a threshold of E < 1.0E-5 to identify the proteins whose sequences were most similar to those of the given transcripts to retrieve their functional annotations (Buchfink et al., 2015). Functional annotation of proteins was carried out by searching against the NCBI non-redundant nucleotide (Nt) database using BLASTn algorithms with a threshold of E < 1.0E-5.

The expression level of each transcript was quantified using the RSEM program (Li and Dewey, 2011; Langmead and Salzberg, 2012). DESeq2 and edgeR were used for differential expression analysis of samples, in which |log₂ ^{(Fold Change)}| ≧ 1 and false discovery rate (FDR) < 0.05 were considered to be significantly differentially expressed (Robinson et al., 2010; Love et al., 2014; Varet et al., 2016). DESeq2 and edgeR are implemented as a package for the R statistical environment, and the count matrix and metadata are stored in an S4 class derived from the SummarizedExperiment class of the GenomicRanges package (http://www.bioconductor.org/packages/release/bioc/html/DESeq2.html) and Bioconductor project (http://bioconductor.org/packages/RNAseq123), respectively (Lawrence et al., 2013; Love et al., 2014). The GO and KEGG were employed to analyze the functional enrichment and metabolic pathways of P. notoginseng transcriptomic data, and a corrected P value < 0.05 indicated that the enrichment was more significant. The transcripts per million (TPM) reads of these unigenes were obtained from RNA-seq data, and the log₂ ^(TPM) values of the differential expressed genes (DEGs) were repeatedly calculated based on three biological replicates. The online tool SangerBox 3.0 (http://vip.sangerbox.com/home.html) was used to generate heatmaps of DEGs involved in notoginsenoside (Shen et al., 2022).

A quantitative real-time PCR (qPCR) validation assay was carried out according to the description of Cao et al. (2023a), and 10 target genes and β-actin (reference gene) were used to validate the DEG results of Illumina. All the primers used for qPCR are listed in Supplementary Table 5 , and the qPCR was carried out using TB Green^® Premix Ex Taq™ II (Takara) on a LightCycler^®96 real-time PCR system (Roche, Switzerland). The data were processed using Q=2^-ΔΔCT to obtain the relative expression level of each gene, in which Q means the relative expression level.

All data were processed and analyzed statistically with Microsoft Excel 2010 and SPSS 26.0. Assumptions of normality and homogeneity of variance were tested prior to all statistical tests. The significant differences were tested with one-way analysis of variance (ANOVA), followed by Tukey tests at the level of 0.05. Origin 2021 and R 4.3.2 were used for the data visualization. Standard deviation (SD) were used for error bars, n≧3. The biomass, contents of saponins, activities of key enzymes and concentrations of JAs were presented as the mean ± SD, n≧3. The correlation among DEGs, saponin and JAs’ contents and key enzyme activities were accomplished using the Mantel test with linkET R package (https://github.com/Hy4m/linKET) (Huang, 2021). Pearson analysis were used to determine the correlation of expression levels of DEGs and activities of key enzymes, and the threshold was set as r > 0.6 (Wang H.R. et al., 2023).

As shown in Figure 1 , AMF inoculation, as well as the addition of exogenous MeJA, significantly promoted the growth of P. notoginseng, of which the plant height and the fresh weight of roots and aboveground parts increased by 31.69/21.99, 55.46/23.69 and 71.53/69.69% (AMF/MeJA), respectively, compared with the CK group, directly indicating that the promoting effect of AMF was higher than that of MeJA. The biomass of P. notoginseng was not significantly reinforced under the superimposed treatment of AMF and MeJA. Compared with AMF, the addition of SHAM (AMF-SHAM) significantly decreased the promoting effect of AMF (P < 0.05), and the inhibitory rates to plant height and the fresh weight of roots and aboveground parts were 8.51, 20.88 and 21.55%, respectively.

Compared to CK, the contents of notoginsenoside R1, ginsenoside Rg1, Re, Rb1, Rd and PNSs significantly increased by 1.85, 1.56, 3.47, 1.64, 1.94 and 1.68 times, respectively, under the AMF treatment (P < 0.05). The addition of exogenous MeJA also significantly increased the contents of five monomeric saponins and PNSs. However, the saponin contents of P. notoginseng in the AMF-MeJA and SHAM groups were significantly lower than that of AMF, and the decreasing amplitudes of notoginsenoside R1, ginsenoside Rg1, Re, Rb1, Rd and PNSs were 38.57/73.51, 50.58/21.29, 69.42/69.21, 19.73/39.94, 54.37/67.17 and 36.64/38.62% (AMF-MeJA/SHAM), respectively, suggesting that the biosynthesis and accumulation of notoginsenosides were largely affected by AMF colonization and that the addition of MeJA participated in the relevant regulatory process. In addition, the supplementation with SHAM could significantly increase the content of ginsenoside Rg1, but it was lower than the promoting effect of AMF and MeJA.

The activities of key enzymes SS, SE, CYP450 and GT increased by 2.55-, 1.58-, 1.49- and 1.47-fold, respectively, under AMF inoculation ( Figure 2 ). Similar changes were also observed in the MeJA group, in which the activities of the key enzymes SS, SE, CYP450 and GT increased by 2.32-, 1.54-, 1.51- and 1.34-fold, respectively. In general, the activities of SE, CYP450 and GT in the AMF-MeJA group were significantly lower than those in AMF and MeJA groups, but higher than those in the CK group. The addition of SHAM improved the activities of key enzymes, but it can weaken the roles of AMF in activity promotion. Compared to the AMF group, the activities of the key enzymes SS, SE, CYP450 and GT in SHAM decreased by 51.02, 22.67, 29.02 and 28.13%, respectively ( Figure 2 ).

As shown in Figure 3 , the concentrations of endogenous JAs, including JA, MeJA, JA-Ile and OPDA, were 30.86, 36.56, 12.13 and 42.39 ng/g, respectively, under inoculation with AMF and significantly increased by 3.24-, 4.04-, 1.97-, 2.06- and 1.25-fold compared to the CK group. This result suggested that the colonization of AMF participated in the biosynthesis and accumulation of endogenous JAs in the fibrous roots of P. notoginseng. The addition of exogenous MeJA also caused an increase in endogenous JAs, particularly JA, MeJA and JA-Ile, suggesting that short-time treatment with MeJA disturbed the metabolism of endogenous JAs. The concentrations of JA, JA-Ile and OPDA in AMF-MeJA group were significantly lower than those in the AMF group and close to those in the CK group, speculating that the addition of MeJA inhibited the regulatory effect of AMF on the JAs. However, as a specific inhibitor to JA biosynthesis, the addition of SHAM decreased the concentrations of JA, MeJA, JA-Ile and OPDA, of which the concentrations were significantly lower than those of the AMF and MeJA groups, particularly in MeJA. Compared to CK, the SA concentration decreased in all treatment groups except for AMF.

AOC and LOX are two key enzymes that continuously catalyze the substrates hexadecatrienoic acid and α-linolenic acid in chloroplasts, which is a key step in JA biosynthesis. As shown in Figure 4 , the activities of AOC and LOX in P. notoginseng roots under AMF were greater than those of the other groups. The addition of MeJA increased the activities of AOC and LOX, and a significant difference in LOX activity was observed between the MeJA and CK groups. Compared to the AMF group, the activities of AOC and LOX decreased significantly under AMF-MeJA, suggesting that the addition of exogenous MeJA can reduce the role of AMF in regulating JA biosynthesis. In addition, the short-term treatment with SHAM had no significant effect on AOC and LOX activities.

The molecular mechanisms of notoginsenoside biosynthesis mediated by MeJA under the inoculation of AMF were determined by RNAseq analysis. Six transcriptomes were generated from the roots of groups AMF, MeJA, AMF-MeJA, SHAM, AMF-SHAM and CK, respectively. Eighteen transcriptomes from the six groups with three replicates yielded a total of 1,108,183,168 clean reads from 1,155,716,266 raw reads, and the clean data size of each transcriptome was larger than 8.46 Gb. The values of Q20 and Q30 for the above transcriptomes were in the ranges of 97.42~97.96% and 93.1~94.36%, respectively, and the overall sequencing error rate was 0.03% ( Supplementary Table 6 ). Clean reads of 76.20~81.00% can be mapped to a single locus of the P. notoginseng genome with RSEM software ( Supplementary Table 6 ). Pearson’s correlation coefficient showed that treatment groups of AMF, MeJA, SHAM, AMF-MeJA and AMF-SHAM had a strong correlation (R² ≧ 0.75) between two duplicate samples. However, the correlation between CK and the other groups was relatively low, with a range of 0.53~0.88 ( Supplementary Figure 1 ).

Compared to the CK group, a total of 3,553 DEGs were detected in the AMF group, including 1,461 (41.12%) and 2,092 (58.88%) upregulated and downregulated genes, respectively. Ten DEGs distributed in different families were selected for qRT–PCR validation, and the line correlation of DEGs in qRT–PCR and RNA-seq were compared by normalizing FPKM values with log₂ ^{(Fold_change)}. The results showed that the expression changes detected by RNA-seq were closely related to those determined by qRT–PCR, with a correlation coefficient R² of 0.93824 ( Figure 5A ), which confirmed the accuracy and reliability of the RNA-seq results.

In total, 124,110, 94,304, 77,830, 124,298 and 84,972 unigenes were matched to existing gene models in the NR, KEGG, Pfam, Trembl and SwissProt protein databases, respectively ( Supplementary Table 7 ). However, the match ratios were not as high as expected due to the absence of research on RNA-seq of plant fibrous roots, particularly in P. notoginseng. As illustrated in Figure 5B , 42,512, 21,456 and 5,130 unigenes in 124,110 were homologous to Daucus carota subsp. sativus, Quercus suber L. and Vitis vinifera L., respectively.

The results of GO classification showed that 106,540 annotated unigenes (approximately 41.37%) of P. notoginseng with BLAST matches to known proteins were assigned to 3 main GO categories, including biological process (290,760, 37.27%), cellular component (348,933, 44.73%) and molecular function (140,390, 18.00%). These GO terms were further subdivided into 60 subcategories, and gene-encoding “binding” and “catalytic activity” proteins accounted for a large part of the molecular function category, e.g., key enzymes for saponin synthesis ( Figure 5C ; Supplementary Table 8 ). The results of KEGG analysis showed that 94,304 unigenes (approximately 36.62%) were found to have significant matches in the KEGG pathway database ( Supplementary Table 9 ). DEGs that are involved in metabolic pathways and biosynthesis of secondary metabolites were closely related to treatments such as AMF and MeJA, in which 37 (AMF vs. CK) and 45 (MeJA vs. CK) DEGs were involved in the biosynthesis of terpenoids, respectively, e.g., triterpenoids, sesquiterpenoids and monoterpenoids ( Figure 5D ). According to KEGG annotation, many unigenes in the items, e.g., lipid transport and metabolism, carbohydrate transport and metabolism, energy production and conversion and signal transduction and mechanisms, were found to be closely related to the substance and energy exchange of the AMF-P. notoginseng symbiotic system and signal pathways mediated by plant hormones ( Supplementary Figure 2 ; Supplementary Table 10 ).

Notoginsenoside biosynthesis in roots is regulated by various environmental factors and directly results in change in the expression of key enzyme-encoding genes. In total, 381 DEGs were found in the six transcriptome libraries of AMF, MeJA, AMF-MeJA, SHAM, AMF-SHAM and CK, in which 1, 3, 1, 5, 1, 4, 4, 8, 7, 3, 186 and 147 unigenes were clustered into the families mevalonate kinase (MVK), phophomevalonate diphosphate kinase (PVK), diphosphomevalonate decarboxylase (MVD), 1-deoxy-D-xylulose-5-phosphate synthase (DXS), γ-deoxy-D-xylulose-5-phosphate reductoisomerase (DXR), diphosphocytidyl-2-C-methyl-d-erythritol kinase (ISPE), (E)-4-hydroxy-3-methylbut-2-enyl pyrophosphate reductase (ISPH), geranyl diphosphatesynthase (GDPS), SE, DS, CYP450 and GT, respectively ( Supplementary Table 11 ). The DEGs of CYP450 and GT may be largely responsible for the biosynthesis and accumulation of different types of notoginsenosides, especially in response to AMF colonization and MeJA addition. As shown in Figure 6 , the expression levels of DEGs significantly increased under AMF inoculation and MeJA addition, e.g., DXS, SE, CYP450 and GT. The expression fold changes of PnISPH_2, PnISPH_3, PnCYP450_1, PnCYP450_4, PnCYP450_5, PnCYP450_6, PnCYP450_9, PnCYP450_12, PnCYP450_14, PnCYP450_15, PnGT_1, PnGT_4, PnGT_5, PnGT_6, PnGT_8, PnGT_9 and PnGT_11 in the AMF group were larger than those in the MeJA and AMF-MeJA groups. Compared to AMF, the expression levels of most DEGs listed in the heatmap decreased significantly under AMF-SHAM, suggesting that the addition of SHAM can inhibit the activation of AMF in saponin biosynthesis. A large difference in the expression levels of DEGs, e.g., PnDXS_2, PnISPH_1, PnGDPS_1, PnSE_2, PnDS_1, PnCYP450_5, PnCYP450_6, PnCYP450_7, PnCYP450_8, PnCYP450_10, PnGT_2, PnGT_3, PnGT_4, PnGT_9 and PnGT_10, among MeJA, SHAM and CK showed that JA played a positive promoting role, while SHAM had no obvious effect.

Key enzyme-encoding genes involved in JA biosynthesis responded to the treatments of AMF, MeJA and SHAM, and were differentially expressed ( Supplementary Table 12 ). PnLOX-1 and PnLOX-3 significantly upregulated their expression levels in response to AMF, and the relative expression levels of PnLOX-2 and PnOPR-1 increased under MeJA and AMF-SHAM. In general, the effects of AMF, MeJA and AMF-SHAM on the expression of key enzyme-encoding genes were not exactly the same, thereafter affecting the endogenous concentrations of JAs.

As shown in Figure 7A , there was a high correlation between expression levels of DEGs involved in notoginsenoside synthesis and activities of key enzymes that catalyzes the synthesis of saponins and JAs. Obviously, the contents of five monomeric saponins and PNSs and the concentrations of JAs were closely related to activities of key enzymes SS, SE, CYP450, GT, AOC and LOX ( Figure 7B ). In addition, expression levels of DEGs were also well correlated with contents of saponins and JAs ( Figure 7C). These results fully showed that JAs mediated the synthesis of saponins and responded to the induction of AMF that representing by DEGs. Pearson’s correlation results showed there were generally high correlation among DEGs or key enzyme activities.