Seeds of P. notoginseng were collected from mature plants, sown in the wells of seedling-raising plates, and then incubated in a greenhouse under controlled conditions (20–30°C, day length of 12 h). The soil in the seedling-raising plates consisted of a mixture of six parts field soil, two parts perlite and two parts sand (pH: 6.53; electrical conductivity: 280 μS cm⁻¹; 1.6% organic matter; nutrient contents: 105.62 mg kg⁻¹ available N, 195.10 mg kg⁻¹ available P and 124.24 mg kg⁻¹ available K). The plants were initially irrigated once with 50% Hoagland's solution (Hoagland and Arnon, 1950) and watered twice a week with fresh water over the course of the experiment. After 5 months of growth, 8–10 cm-high seedlings were used for subsequent experiments.

Soil samples were collected in October 2014 from a Sanqi ginseng field in Yanshan County (104.33°E, 23.54°N). Ten bulk soil samples were collected from fields in which P. notoginseng was consecutively cultivated for 1, 2, or 3 years. Rg₁ was extracted from the soil with MeOH:H₂O (80:20) and quantified via HPLC-ESI-MS as described by Yang et al. (2015). The effect of the autotoxic compound Rg₁ on the biomass of P. notoginseng was investigated according to its concentration in cultivated soil. Briefly, the ginsenoside Rg₁ (purity ≥98%, Guizhou Dida Biological Technology Co.) was dissolved in methanol and then diluted in distilled water to final concentrations of 0.1, 0.5, 1.0, and 5.0 mg L⁻¹. Distilled water containing the same concentration of methanol (1.0%) was used as a control. Seedlings of P. notoginseng were carefully collected from the soil and washed three times with sterile water. A 100-mL aliquot of Rg₁ solution was added to each sterile glass bottle (250 mL). Ten seedlings were placed in each bottle, and each treatment included six replicates. The seedlings were incubated in a programmable illuminated incubator under a light/dark (L/D) cycle of 12 h/12 h and a temperature cycle of 25°C/20°C. The conductivity of the hydroponic seedling solution was measured using a conductivity meter (FE30, Mettler Toledo Int., Inc., Switzerland) at 12-h intervals. After incubation for 96 h, the wilt ratio of seedlings was calculated as wilt ratio (%) = 100 × (wilted seedlings/total seedlings), and the reduction in biomass was calculated as reduced biomass = fresh weight before treatment – fresh weight after treatment.

Changes in the vigor of the root cells were monitored using a modified capillary root model, as described by Yang et al. (2015). Briefly, the tip of each fibrous root was incubated with Rg₁ solution (1.0 mg L⁻¹) or sterilized water containing 1.0% methanol at room temperature (24°C). The effect of Rg₁ on root cell vigor was observed by staining with fluorescein diacetate (FDA) (5 mg L⁻¹) and the red fluorescent dye propidium iodide (PI) (5 mg L⁻¹) as described in a previous report (Fan et al., 2013). Staining was visualized using a Leica SP5 confocal laser-scanning microscope (excitation at 488 nm and emission above 630 nm; Leica, Wetzlar, Germany) after incubation for 0.5, 1.0, 2.0, or 3.0 h. The experiment was repeated three times, and each replicate included 10 roots.

Seedling roots were treated with Rg₁ solution at a concentration of 1.0 mg L⁻¹ or with sterilized water containing 1.0% methanol. After incubation for 0, 0.5, 1.0, 3.0, 4.0, 6.0, 12.0, or 24.0 h, the root tips were excised with a sterile razor blade and prepared for viewing under a transmission electron microscope (TEM) as detailed by Zhu et al. (2007). The experiment was repeated three times, and each replicate included 10 roots.

Seedling roots were treated with Rg₁ (1.0 mg L⁻¹) or sterilized water containing 1.0% methanol for 0, 3, 12, 24, or 48 h. The fibrous roots were then rapidly harvested from the seedlings at each time point, immediately frozen in liquid nitrogen, and stored at −80°C until RNA extraction. Each treatment included three independent replicates. Total RNA was extracted from fibrous roots using the TRIzol method. Because reference genome was unavailable for P. notoginseng, a transcriptome assembly library was constructed as a reference library by mixing equal amounts of RNA from the above 15 samples. The libraries were sequenced on the Illumina HiSeq™ 2000 platform by Gene Denovo Co. (Guangzhou, China). Clean reads were assembled de novo using the Trinity programme (version: r20140413p1). Unigenes were then annotated employing the NCBI NR, NT, Swiss-Prot protein, KEGG, GO and COG databases.

A gene expression analysis was performed in two sequential steps. First, all the clean reads were mapped to the assembled sequences using Bowtie 2–2.2.3 to calculate the read counts for each transcript (Langmead and Salzberg, 2012). The transcript abundance for each gene was then measured and normalized as fragments per kilobase of exon per million fragments mapped (FPKM) values (Mortazavi et al., 2008). The differentially expressed genes (DEGs) between 0 and 3 h, 0 and 12 h, 0 and 24 h, and 0 and 48 h of Rg₁ treatment were restricted based on a false discovery rate (FDR) ≤ 0.05 and an absolute value of the log₂Ratio ≥ 2. A gene expression cluster analysis of DEGs was performed using the STEM algorithms (Ernst and Bar-Joseph, 2006). Hierarchical clustering of the DEGs related to the response to ROS, ROS-scavenging enzymes, cell wall decomposition, the cell membrane, xenobiotic detoxification, and protein kinases was performed using the FPKM expression values with the pheatmap programme in the R programme environment (version 2.9.0). The transcriptome datasets can be retrieved from the NCBI SRA database under Project ID PRJNA338825.

To test the effects of Rg₁ and exogenous antioxidants [ascorbate (ASC) and gentiobiose] on growth, glass bottles (250 mL) containing 100 g of coarse silica sand (sterilized at 160°C for 3 h) were supplemented with Rg₁ alone (1.0 mg L⁻¹), ASC alone (0, 0.1, 1.0, 10.0, or 50.0 mg L⁻¹), gentiobiose alone (0, 0.1, 1.0, 10.0, or 50.0 mg L⁻¹), or a combination of Rg₁ and antioxidants. The seeds were surface-sterilized with 1% sodium hypochlorite for 5 min and washed three times with sterile water. Ten seeds were sown in the silica sand in each bottle, and 10 mL of the respective treatment solution was added. All the treatments were incubated in a programmable illuminated incubator as describe above. After incubation for 30 days, the germination rate and plant biomass were measured. The experiments were performed three times with six replicates. The resultant data were analyzed via one-way analysis of variance followed by post-hoc Duncan's test (p < 0.05).

To test the effect of Rg₁ and exogenous antioxidants in relation to oxygen damage and enzymatic activities, seedling roots were exposed to Rg₁ (1.0 mg L⁻¹), either alone or with ASC (1.0 mg L⁻¹) or gentiobiose (10.0 mg L⁻¹). After incubation at room temperature (24°C) for 0, 3, 12, 24, or 48 h, the accumulation of ROS (superoxide and H₂O₂) in the roots was observed. Superoxide was stained with 0.5 mg mL⁻¹ nitrotetrazolium blue chloride (NBT) in 10 mM potassium phosphate buffer (pH 7.8), as described by Dutilleul et al. (2003), and then observed with a compound microscope (Leica DM2000, Wetzlar, Germany). The H₂O₂ content was measured spectrophotometrically at 410 nm according to the procedure described by Shi et al. (2015).

Reactive oxygen metabolism-related antioxidants and antioxidant enzymes in the roots were measured at 0, 3, 12, 24, or 48 h. The enzymatic antioxidant activities of superoxide dismutase (SOD), catalase (CAT), peroxidase (POD), ASC peroxidase (APX), glutathione S-transferase (GST), glutathione reductase (GR), dehydroascorbate reductase (DHAR), and monodehydroascorbate reductase (MDHAR) were quantified according to the procedures described by Wu et al. (2017). For each enzyme assay, 100 mg of ground root tissue was homogenized in 1 mL of 50 mM sodium phosphate buffer (pH 7.0) containing 0.2 mM ethylenediaminetetraacetic acid (EDTA) and 1% (w/v) polyvinylpyrrolidone (PVP) on ice, and the homogenates were then centrifuged at 12,000 × g and 4°C for 20 min. The supernatants were subsequently used for the determination of enzyme activities. For CAT, POD, and SOD, absorbance was determined with a spectrophotometer (Spectronic Instruments, Rochester, NY, USA) at 240, 470, and 560 nm, respectively. For DHAR and APX, absorbance was measured using a spectrophotometer at 265 and 290 nm, respectively. For GST, GR, and MDHAR, absorbance was assayed using a spectrophotometer (Spectronic Instruments, Rochester, NY, USA) at 340 nm.

The non-enzymatic antioxidant contents of ASC, dehydroascorbate (DHA), glutathione (GSH), oxidized GSH (GSSG), and malondialdehyde (MDA) were measured. Frozen ground root powder (0.1 g) was homogenized with 5 mL of 5% ice-cold trichloroacetic acid (TCA), and the mixture was then centrifuged at 8,000 × g and 4°C for 10 min. The supernatant was immediately used for analysis. The contents of ASC, DHA, GSH and GSSG in the root samples were determined according the method described by Chen et al. (2011). The MDA contents in root samples were quantified according to the method described by Xu et al. (2016). All enzyme activities and non-enzymatic antioxidant contents in the roots were measured in three biological replicates. Each time point of the Rg₁, Rg₁+ASC, and Rg₁+gentiobiose treatments was compared with each other, and the different time points of each treatment were compared with time 0 h.

The data were analyzed via one-way analysis of variance and compared using post-hoc Duncan's test (p < 0.05) with PASW Statistics 18 (SPSS Inc.). The results of the cluster analysis of gene expression were evaluated using heatmap.2 in the “gplots” package in R (version 3.1.2).

The mean and range of Rg₁ concentrations were 0.70 mg kg⁻¹ and 0.21~1.32 mg kg⁻¹, respectively, in the 1-year consecutively cultivated bulk soil, 1.09 mg kg⁻¹ and 0.83~1.54 mg kg⁻¹ in the 2-year consecutively cultivated bulk soil, and 1.41 mg kg⁻¹ and 1.03~2.01 mg kg⁻¹ in the 3-year consecutively cultivated bulk soil (Table S1). Thus, Rg₁ concentrations of 0.1, 0.5, 1.0, and 5.0 mg L⁻¹ were selected to test its autotoxicity during seedling growth. Compared with the control, the seedling wilt ratio (Figure 1A) and the reduction in biomass (Figure 1B) were increased by Rg₁ treatment in a dose-dependent manner. The greatest values for the wilt ratio and the reduction in biomass were obtained when the seedlings were treated with Rg₁ at a concentration of 1.0 mg L⁻¹. Hence, this concentration was chosen for further analyses.

To identify the progression of Rg₁-induced cell death in roots, we stained the roots with FDA-PI. FDA stains living cells green, and PI stains dead cells red. As shown in Figure 2, most of the apical and subapical root cells in the control treatment were sporadically stained with FDA (Figures 2A–D). However, in roots treated with 1.0 mg L⁻¹ Rg₁, PI-stained dead cells began to distinctly appear 0.5 h after treatment and were significantly increased after prolonged treatment. After 3 h of incubation, almost all root cells were strongly stained with PI (Figures 2E–H). When the roots of seedlings were immersed in water amended with different concentrations of Rg₁, the conductivity of the hydroponic solution significantly increased in a dose-dependent manner (Figure S1).

The effect of Rg₁ on root cell morphology was further observed via TEM. Cells from the apical meristem of untreated roots exhibited few small vacuoles with large, round nuclei and intact organelles (Figure 3A). However, the addition of 1.0 mg L⁻¹ Rg₁ to the roots led to many abnormalities. After treatment from 3 h (Figure 3B) to 24 h (Figure 3C), the cell wall began to thicken, the shape of the cells became distinctly distorted, and the cells began to shrink. As the treatment time increased, the plasma membrane retracted from the cell wall, causing plasmolysis (Figure 3D), and the cytoplasm became concentrated (Figures 3E,F). Furthermore, cellular structures broke down, including digestion of the cell wall (Figures 3G–I) and disappearance of various organelles, and vacuoles occupied most of the space in the cells (Figures 3J–L).

To obtain the reference P. notoginseng transcriptome for the lateral roots, an RNA-Seq library was constructed using RNA from all root samples. A total of 48.95 G nucleotides with a Q₂₀ percentage of 94.39% were generated. The Trinity package assembled 10,0125 unigenes, with a mean size of 631.82 bp (Table S2). A total of 100,125 unigenes were successfully annotated in at least one of the NR, NT, Swiss-Prot, KEGG, GO, and COG databases, and 16,601 unigenes (16.58%) were annotated in all six databases (Figure S2).

To identify the DEGs in response to Rg₁, pairwise comparisons using 0 h as the control and 3, 12, 24, and 48 h as the treatments were performed. After 3, 12, 24, or 48 h of exposure to Rg₁, we identified 5, 38, 1105, and 35 DEGs, respectively, compared with the 0-h treatment (Table S3). All DEGs were separated into three significant clusters (p < 0.001) based on similarities in the expression profiles (Figure 4). Some DEGs involved in plant growth and development, including oxidative phosphorylation, ribosomes, and RNA transport, were significantly down-regulated after expose to Rg₁ for 12 or 24 h (Table S4).

Based on the cellular data, we further analyzed the transcript profiles of genes involved in ROS metabolism and cell wall- and cell membrane-related processes (Figure 5; Table S5). ROS metabolism-related genes, including 10 ROS response genes and 15 ROS scavenging enzymes genes (i.e., SOD, CAT, APX, POD, MDHAR, and GST), were significantly up-regulated by Rg₁ treatment for 12 or 24 h. For example, one GST gene (unigene 0072579), which encodes a type of antioxidative enzyme involved in the ASC-GSH cycle, was significantly up-regulated 100- and 114-fold by Rg₁ treatment for 12 and 24 h, respectively, compared with the level detected at 0 h. Cell wall decomposition-related genes, including two glycoside hydrolase genes, two beta-1,4-glucanase genes, three beta-glucosidase genes, two xyloglucan endotransglucosylase/hydrolase genes, and one chitinase gene, were significantly up-regulated in the roots after exposure to Rg₁ for 12 or 24 h. For example, one glycoside hydrolase gene (unigene 0014244) involved in cell wall degradation was up-regulated 1.94-, 2.21-, and 3.67-fold by Rg₁ treatment for 3, 12, and 24 h, respectively. In addition, 10 membrane protein-related genes were significantly up-regulated, and many genes encoding proteins involved in xenobiotic detoxification (i.e., cytochrome P450, ABC transporters, and other transporters), protein kinases, and transcription factors were also up-regulated (Figure 5; Table S5).

The transcriptomic results demonstrated that Rg₁ might disturb ROS metabolism in roots and cause cell death. To test this hypothesis, we further studied the effect of Rg₁, with or without the exogenous antioxidants ASC and gentiobiose, on root growth and ROS metabolism. Ascorbate is an antioxidant involved in the ASC-GSH cycle (Gill and Tuteja, 2010). Gentiobiose (6-O-β-D-glucopyranosyl-D-glucose) is reported to up-regulate the synthesis of GSH and activate the ASC-GSH cycle (Takahashi et al., 2014). Root biomass was reduced by 63.20% when seedlings were grown on media amended with Rg₁ at a concentration of 1.0 mg L⁻¹ compared with the control (Figure 6). However, the toxicity of Rg₁ to root growth was significantly alleviated to the control levels by the addition of ASC at concentrations of 1.0, 10.0, and 50.0 mg L⁻¹ (Figure 6A) or gentiobiose at a concentration of 10.0 or 50.0 mg L⁻¹ (Figure 6B). Treatment with ASC or gentiobiose alone at concentrations from 0.1 to 50 mg L⁻¹ did not show significant toxicity to plants, as indicated by the root biomass (Figure S3).

To elucidate the effect of Rg₁ treatment on ROS metabolism over time in roots, the status of antioxidants and antioxidant enzymes was measured. O2·- significantly accumulated in the roots after exposure to Rg₁ for 3 and 12 h (Figure 7A). However, the level of O2·- in the roots was reduced by the addition of the antioxidants ASC and gentiobiose (Figure 7A). Significant accumulation of H₂O₂ was measured after 3 h of Rg₁ treatment, and this level was decreased after 12 h of treatment (Figure 7B). Roots that were simultaneously treated with Rg₁ and either ASC (1.0 mg L⁻¹) or gentiobiose (10.0 mg L⁻¹) showed significantly reduced H₂O₂ accumulation compared with the plants treated with Rg₁ alone. The lowest H₂O₂ content was observed after gentiobiose treatment for 3 h (Figure 7B).

Under Rg₁ stress conditions, some root antioxidant enzymes were also significantly affected. Root SOD activity was significantly induced by Rg₁ treatment, and significantly higher SOD activity was observed in roots after exposure to Rg₁ plus gentiobiose treatment for 12, 24, and 48 h and after exposure to Rg₁ plus ASC treatment for 24 h (Figure 8A). CAT and POD activities were slightly increased, albeit not significantly, by treatment with Rg₁ and the antioxidants (Figures 8B,C). The antioxidants and enzymes involved in the ASC-GSH cycle were also affected by Rg₁ treatment. The ASC/DHA ratio and APX activity in roots were significantly decreased after exposure to Rg₁ for 3 h (Figures 9A,B; Figures S4A,B). Treatment with either exogenous ASC or gentiobiose resulted in resumption of APX activity, increased the ASC/DHA ratio and even stimulated the activity of DHAR (Figure 9C). However, MDHAR activity was not significantly affected by treatment with Rg₁ or the antioxidants (Figure 9D). The GSH/GSSG ratio also decreased after exposure to Rg₁ for 12 h (Figure 10A; Figures S4C,D) but significantly increased after the addition of gentiobiose (Figure 10A). Root GR and GST activities were also significantly increased by the exogenous addition of gentiobiose (Figures 10B,C).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.