Ginkgo biloba plants from city of Pizhou, China, were grown in the Medicinal Botanical Garden of China Pharmaceutical University. All the tissues of G. biloba were collected in June 2015. After cleaning with distilled water, different tissues of G. biloba plants, including the main root, fibrous root, leaf, young stem and old stem, were collected. For the further analysis, the main root periderm (R1), the main root cortex and phloem (R2), the main root xylem (R3), the old stem periderm (S1), the old stem cortex and phloem (S2) and the old stem xylem (S3) were separated with a scalpel and tweezers under an anatomical lens. All the tissues were then immediately frozen in liquid nitrogen and stored in -80°C until used.

Chromatographic methanol was obtained from Hanbon Tech Co. (Jiangsu Province, China). Acetonitrile of HPLC grade was purchased from TEDIA Company (Fairfield, OH, United States). Formic acid (99%, HPLC grade) was obtained from ROE (Newark, NC, United States). Deionized water was further purified with a Milli-Q water purification system (Millipore, Milford, MA, United States). Forty-two reference standards were purchased from the National Institute for the Control of Pharmaceutical and Biological Products (Beijing, China). Hesperidin and andrographolide were used as internal standards. All standards are listed in Table 1.

The collected samples were homogenized in liquid nitrogen, and the powder from the samples was vacuum-lyophilized for 72 h. After weighing, the powder samples (100 mg/sample) were extracted with 70% aqueous methanol (8 mL) by vortexing for 3 min, followed by ultrasonication at room temperature for 60 min. Afterward, the resultant mixture was centrifuged at 13000 rpm for 10 min. The collected supernatant and 50-fold diluted supernatant were used for quantitation. All samples were filtered through 0.22 μm membrane filters before UHPLC-QQQ-MS/MS analysis.

The samples were analyzed using a Shimadzu Nexera UHPLC with 8050 QQQ-MS/MS (Shimadzu, Japan). The method of metabolites analysis was performed as described previously with some modifications (Liu et al., 2014, 2015). The chromatographic separation was conducted with an Agilent Zorbax Extend C18 system equipped with a Narrow-bore column (2.1 mm× 150 mm, 5 μ, 80 A), with an injection volume of 2 μL. The sample vials were maintained at 4°C in a thermostatically controlled column compartment. The mobile phase consisted of 0.1% formic acid water (A) and acetonitrile (B). The mobile phases were programed with the gradient as follows: 0–5 min, 8–10% B; 5–15 min, 10–30% B; 15–20 min, 40–49% B; 20–23 min, 49% B; 23–25 min, 49–60% B; 25–26 min, 60–80% B; 26–28 min, 80% B; 28–30 min, 80–8% B; and 30–32 min, 8% B. The flow rate was 0.5 mL/min. Forty metabolites in G. biloba plants was quantitated in this work. A stock solution of 40 standard reference compounds was prepared by dissolving accurately weighted portions of the standards in methanol. Six different mixture solutions of standard concentration were then obtained by precisely mixing the 40 stock solutions with methanol. A calibration curve was constructed with at least six appropriate concentrations in triplicate. The regression equation and linear range of each metabolite are listed in Supplementary Table S1.

The ion source parameters of the mass spectrometer were as follows: nebulizing gas flow rate, 3 L/min; heating gas flow rate, 10 L/min; drying gas flow rate, 10 L/min; interface temperature, 300°C; desolvation temperature, 250°C; and heat block temperature, 500°C.

The raw data from the UHPLC-QQQ-MS/MS was quantitatively analyzed by the multiple reactions monitoring mode. The selection of precursor ion, product ion, collision energy (CE), and Q1/Q3 pre-bias were automatically optimized. All the parameters are listed in Supplementary Table S2. A heat map of the forty metabolites in G. biloba plants was constructed with the software MultiExperiment Viewer 4.9.0. For the statistical analysis, unsupervised principal component analysis (PCA) was used to further analyzed the metabolomic data using the software SIMCA-P 11.5 (Umetrics, Umea, Sweden). The amounts of forty reference compounds in each of the five samples (main root, fibrous root, leaf, young stem, and old stem) formed a data matrix with fifteen rows and forty columns, which was used for PCA analysis, after normalization. The first three principal components were extracted.

Roots and stems of G. biloba were fixed in formalin-aceto-alcohol (FAA, 90 mL of 70% alcohol, 5 mL of 38% formaldehyde and 5 mL of glacial acetic acid) for 48 h. The fixed samples were sequentially dehydrated in 50, 70, 85, 95, and 100% ethanol. The tissues were then cleared, infiltrated and embedded in paraffin wax according to the methods of Ye et al. (2014). The sections of each tissue were made with an HM340E microtome (Microm International, Walldorf, Germany) and stained with safranine and Fast Green. The sections were observed with an Olympus BX60 microscope equipped with Sensys 1400E cooled charge-coupled device (CCD) camera.

Total RNA was extracted from various tissues of G. biloba using a plant RNA isolation kit (BioTeke, Beijing, China) following the manufacturer’s instructions. All RNA samples were digested with DNase I (RNase-free) prior to use. Aliquots of 1 μg of total RNA were employed in the reverse transcriptase reaction using random hexamer primers for the synthesis of first-strand cDNA. The amplification reactions of RT-Q-PCR were performed on a LightCycler^® 96 Real-Time PCR System (Roche) with gene-specific primers, and the SYBR ExScript RT-PCR kit (Takara, Shiga, Japan) was used to confirm changes in gene expression. The thermal cycle conditions used were 5 min at 95°C followed by 40 cycles of amplification (15 s at 95°C, 15 s at 60°C and 30 s at 72°C). A melt curve analysis following each RT-Q-PCR was performed to assess product specificity. Expression patterns of ginkgolide biosynthesis key genes were analyzed in three biological replicates. Quantification of the target gene expression was carried out with comparative CT method (Livak and Schmittgen, 2001). The primers used in RT-Q-PCR are listed in Supplementary Table S3.

The leaf, young stem, old stem, main root, and fibrous root of G. biloba were analyzed by UHPLC-QQQ-MS/MS (Figures 2, 3A). Forty compounds including terpene lactones, flavonoids and phenolic acids were identified by comparing the retention time and mass spectra with standards (Supplementary Table S2). Figure 2 shows the chromatogram of forty mixed standards and internal standards (hesperidin and andrographolide). The content of these compounds was calculated according to the calibration curves with the internal standard method (Supplementary Table S1). PCA was used to further analyze the distribution and levels of metabolites in different tissues of G. biloba. As shown in Figure 3B, the results showed that the scores displayed in the PCA plot could be obviously divided into five relative clusters: leaf, young stem, old stem, main root, and fibrous root. There were significant differences in the occurrence and content of 40 compounds among different tissues of G. biloba.

The heat map (Figure 4) and Table 1 shows the content of 40 detected compounds in different tissues of G. biloba. There are significant differences in the contents of flavonoids, terpene lactones and phenolic acids among different tissues of G. biloba. Most of the flavonoids were mainly distributed in aerial parts especially in the leaves and old stems. Only myricetin, (-)-epicatechin gallate (ECG) and quercetin were generally distributed in the various tissues. Most of phenolic acids were generally distributed among the various tissues of G. biloba. Only protocatechuic acid, chlorogenic acid, and caffeic acid (CA) were mainly distributed in the young stems of G. biloba. Terpene lactones, which include ginkgolides and bilobalide, were widely distributed in various tissues of G. biloba, and the content in the leaves and roots was much higher. The total content of terpene lactones decreases in the order of the leaf, fibrous root, main root, old stem, and young stem.

Among the metabonomics data, the terpene lactones including GA, GB, GC, ginkgolide J (GJ), and BA were further analyzed (Figure 5A). The content of GA, GB, GC, and BA were much higher than that of GJ in various tissues of G. biloba. The content of GA was the highest in the fibrous root, while other ginkgolides were highest in the main root. The distribution of bilobalide was unique, being highest in the leaf.

To reveal the relationship of terpene lactone distribution and biosynthesis, the expression patterns of five key genes in the ginkgolide biosynthetic pathway were analyzed by RT-Q-PCR. The results showed that GbLPS is mainly expressed in the fibrous root (Figure 5B). In addition, GbDXS2 was highly expressed in fibrous roots, which showed nearly 260-fold higher expression compared to other tissues (Figure 5C). The expression patterns of GbDXR2, GbIDS2, and GbGGPPS were similar to that of GbDXS2 (Figure 5D). The combined the metabonomics and expression analysis results show that the fibrous root is the main biosynthesis site of ginkgolides.

The main root was divided into three parts the main root periderm (R1), the main root cortex and phloem (R2) and the main root xylem (R3) for further metabolic analysis and expression analyses. The content of TTLs among R1, R2 and R3 showed significant differences (Figure 6A and Table 2). GA, GB, and BL, the combined content of which accounted for >85% of TTLs, exhibited the highest content in R1. In addition, the content of GB in R1 is 2.5-fold and 3.4-fold higher than that of R2 and R3, respectively, while the distribution of GC and GJ were different from that of the above mentioned TTLs. GC and GJ displayed the highest content in R2 (Figure 6B).

Real-time quantitative polymerase chain reaction was used to detect expression of the key gene of the ginkgolide biosynthetic pathway. The results showed that GbLPS is specifically expressed in R1, which implies that the site of ginkgolides biosynthesis should be in R1 (Figure 6C). The expression of GbDXS2, GbDXR2 and both GbIDS2 and GbGGPPS was also detected, respectively, in R1, R2, and R3. In addition, the expression levels of these four genes were significantly higher in R1 (Figure 6D).

The stem was divided into three parts, the old stem periderm (S1), the old stem cortex and phloem (S2) and the old stem xylem (S3) for further metabolic analysis by UHPLC-QQQ-MS/MS (Figure 7A). The distribution of TTLs in the stem was significantly different from that in the root. The GA, GB, GC, GJ, and BL in the stem had the highest content in S2. The total content of TTLs decreased in the order of S2, S1, and S3, with S2 showing a 4-fold and 6.5-fold higher total TTL content than S1 and S3, respectively (Figure 7B and Table 2). The most abundant TTL in S2 was GB, followed by GC, BL, GA, and GJ.

Previous studies showed that the synthesis of ginkgolides was mainly in the root of G. biloba, but it is not clear in which part of the root ginkgolides are synthesized. To further analyze the synthesis site of ginkgolides, G. biloba plants were divided into five parts (including the main root, fibrous root, leaf, young stem, and old stem) for quantitative analysis. The metabolic profiling results showed that the content of TTLs followed the order of the leaf > fibrous root > main root > old stem > young stem. The expression analysis showed that the key genes in the ginkgolide biosynthetic pathway were highly expressed in the fibrous root, while the expression levels of these genes were very low or even undetectable in the mainroot and other aerial portions. Our results demonstrated that the fibrous root is the important biosynthesis site of ginkgolides.

We further analyzed the metabolites and gene expression in different parts of main root. The main root of G. biloba was divided into three parts, including the main root periderm (R1), the main root cortex and phloem (R2) and the main root xylem (R3). The results of the metabolic analysis showed that GA and GB have the highest content in R1. The content of GA and GB was in the order of R1 > R2 > R3. GC and GJ displayed the highest content in R2. The expression analysis results showed that the ginkgolide biosynthetic pathway key gene GbLPS is specifically expressed in the main root periderm (R1). GbDXS2, GbDXR2, GbIDS2 and GbGGPPS were highly expressed in R1, with little expression in R2 and R3. Previous studies showed that GA is the first compound in the ginkgolide biosynthetic pathway. GB, GC, and GJ are the derivatives of GA by successive additions of a hydroxyl group (Cartayrade et al., 1997; Schafer, 2015). Combining the above results has demonstrated that the main root periderm (R1) is also an important ginkgolides synthesis site of G. biloba. The results were similar with those in Salvia miltiorrhiza Bunge, a widely used herb in traditional Chinese medicine. The key genes of the tanshinone biosynthetic pathway are mainly expressed in the root periderm of Salvia miltiorrhiza and tanshinone production is also mainly carried out in the root periderm (Xu et al., 2015).

The key genes of the ginkgolide biosynthetic pathway have only slight or even no expression in the stem of G. biloba. When combining our results with those the previous studies, we know that there is no ginkgolide biosynthesis in the stem of G. biloba (Cartayrade et al., 1997; Neau et al., 1997; Kim et al., 2012). Cartayrade et al. (1997) reported that after ¹⁴CO₂ labeling experiments, the labeled ginkgolides were first detected in roots and subsequently in the stems and leaves. GbLPS promoter-driven GUS expression in transgenic Arabidopsis alluded that the translocation of ginkgolides occurs through the phloem (Kim et al., 2012).

Our experimental results showed that the content of ginkgolides and bilobalide in the cortex and phloem of old stem (S2) is much higher than that in the old stem periderm (S1) and old stem xylem (S3) in G. biloba. These results imply that ginkgolides and bilobalide can be transported from the roots to the leaves via S2. Several secondary compounds, such as iridoid glycosides, pyrrolizidine alkaloids and glucosinolates, can also be transported in the phloem (Robert and Shmuel, 2009). In Senecio vernalis, synthesis of the backbone structure of the pyrrolizidine alkaloid (PA) is restricted to the roots. PAs can move from the roots to shoots via the phloem (Moll et al., 2002). Here, our results first indicate that ginkgolides in G. biloba are produced in the fibrous root and root periderm of the main root, and then move to leaves though the cortex and phloem of the old stem (S2) (Figure 8). In addition, leaves seem to act as a storage location of ginkgolides and bilobalide only (Figure 8).