The NCBI database¹ was used to search for genetic information. The SMART database² was used to analyze amino acid sequences. The BioXM 2.6 software was used to predict the molecular weight and isoelectric point of the gene. ClustalX 1.83 was used for multiple alignments. Neighbor-joining phylogenetic trees were generated using the MEGA 5.1 program.

The expression vector pET29a-Gm6PGDH1 was constructed using the Gateway Technology with a One Step Cloning Kit (Vazyme Biotech, Nanjing, China) according to the protocol of the manufacturer. The sequences of the primers are provided in Supplementary Table 1. The pET29a-Gm6PGDH1 plasmid was transformed into the Escherichia coli Rosetta (DE3) strain, and the expression of Gm6PGDH1 was induced with 0.5 mM IPTG to generate the putative recombinants. Then, the His-tagged Gm6PGDH1 proteins were extracted and purified under native conditions using Ni Focurose 6FF (IMAC), and the products were detected by 10% Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE).

The measurement of Gm6PGDH1 enzyme activity was based on a previous method, with some changes (Šindelář et al., 1999). The recombinants were ultrasonic, broken and centrifuged at 4°C, and the supernatant was enzyme crude extracts. The total reaction volume of the assay was 2 ml containing a 1.8-ml reaction buffer (50 mM Triethanolamine, pH 7.5, 5 mM MgCl₂, 0.18 mM NADP⁺, 1 mM 6PG) and 0.2 ml enzyme crude extracts. Absorbance changes at 340 nm were monitored using a visible spectrophotometer (722N, Shanghai Precision Science Instrument, Shanghai, China). The absorbance value was observed and recorded once every 10 s, and the reaction time was 2 min. A total of three replicates were included in this experiment.

Soybean Williams 82 plants were used in this study. Soybean seeds normally were germinated and grown in a growth chamber under controlled photoperiod and temperature (16 h light; 25°C/8 h dark; 20°C). Five-day-old seedlings (removal of cotyledons) were transferred to a vermiculite medium containing a modified Hoagland solution for 11 days, with 1 mM KH₂PO₄ (1 mM Pi, Pi-sufficient) or 2.5 μM KH₂PO₄ (2.5 μM Pi, Pi-deficient) as the P supply. On the 10th day, the Pi-deficient group was re-supplied with 1 mM Pi for 1 day and was called R1d. The nutrient solution was refreshed every 3 days. Soybean tissues were collected at specific times. All the samples were stored at −80°C prior to RNA extraction.

Three biological replicates, each comprising three individual plants, were used for quantitative real-time PCR (RT–qPCR). Total RNA was extracted using the RNA Prep Pure Plant Kit (TIANGEN, Beijing, China). RT–qPCR analysis was performed as described previously (Li et al., 2020). The relative level of expression was calculated using the formula 2^–△Ct or 2^–△△Ct. Actin (GLYMA_08G146500) or Actin2-8 (AT3G18780) was used as an endogenous control to normalize the samples. All the primers used for RT-qPCR are listed in Supplementary Tables 2, 3.

The coding region of Gm6PGDH1 was amplified with specific primers (Supplementary Table 1), and cloned into the pJIT166-GFP vector to construct the pJIT166-Gm6PGDH1-GFP vector. The fusion (pJIT166-Gm6PGDH1-GFP) and control (pJIT166-GFP) constructs were transformed into Arabidopsis mesophyll protoplast using the polyethylene glycol method (Yoo et al., 2007). The fluorescence of GFP was imaged with a Zeiss LSM780 confocal microscope (Carl Zeiss, SAS, Jena, Germany).

The fused plasmid vector (pCAMBIA3301-Gm6PGDH1) was transformed into wild-type Arabidopsis (Col-0) plants using the Agrobacterium-mediated floral-dip procedure (Clough and Bent, 1998). Transgenic plants were selected on a Basta (20 mgL^–1) medium, and the selected plants were identified by PCR and RT-qPCR (Supplementary Figures 6A,B). Homozygous T3 or T4 seeds and wild-type seeds, which were harvested in the same environment, were used for research.

To confirm the function of Gm6PGDH1 in soybean, the full-length Gm6PGDH1 coding sequence was inserted after the CaMV35S promoter, resulting in pCAMBIA3301-Gm6PGDH1 constructs with a GFP label (the original GUS label was replaced by the GFP label). The sequences of the primers are provided in Supplementary Table 1. Then, the overexpression vector and the empty vector were separately introduced into Agrobacterium rhizogenes K599, which was used to infect soybean (Williams 82) seedlings in order to obtain composite soybean plants with transgenic hairy roots (Attila et al., 2007). When transgenic hairy roots grew approximately 8–10 cm long, a small part was harvested for RT-qPCR identification (Supplementary Figure 1). The primary roots of the transgenic composite plants were cut off, and plants were transferred to Hoagland vermiculite medium containing 1 mM Pi or 2.5 mM Pi for 14 days after 2 days of growth with Hoagland nutrient solution. Each transgenic root represented an independent transgenic line; for each Pi treatment, six independent transgenic lines were included. After 14 days of growth, leaves and hairy roots were harvested separately to determine dry weight and total P concentration. Several hairy roots were taken for further RT-qPCR analysis. An independent transgenic line originating from a composite soybean plant with a transgenic hairy root was considered as a semi-biological replicate. A total of three replicates were included in this experiment.

The fused plasmid vector (pCAMBIA3301-Gm6PGDH1) was transformed into soybean (Williams 82) using the Agrobacterium-mediated transformation method (Li et al., 2017b). Positive transgenic plants were screened in a greenhouse by leaf herbicide painting and PCR-based amplification of genomic DNA. T3-generation plants of homozygous independent transgenic lines were used for research.

A 329-bp DNA fragment of the Gm6PGDH1 gene was PCR-amplified using gene-specific primers (Supplementary Table 1) and cloned into the BPMV RNA2 silencing vector to construct BPMV-Gm6PGDH1 vectors (Zhang et al., 2010). As described earlier, DNA-based BPMV constructs were inoculated in V1 stage soybean seedlings by mechanical friction (Pflieger et al., 2015). The empty BPMV vector was used as a control. Two weeks post-virus inoculation, the leaf edge of some of the plants was wrinkled and deformed (Supplementary Figure 5A), indicating successful infection. Then, small leaves of the plants with infection symptoms were harvested to isolate the RNA for verification of gene silencing by RT-qPCR (Supplementary Figure 5B). The infected plants were transferred to vermiculite medium containing 2.5 μM Pi Hoagland nutrient solutions for 7 days. Each infected plant represented one independent gene silencing line. One independent infected plant was considered as one biological replication. A total of three replicates were included in this experiment.

Five-day-old soy bean plants were grown in a Hoagland medium with two Pi levels (1 mM KH₂PO₄ and 2.5 μM KH₂PO₄) for 30 days. The nutrient solution was refreshed every 3 days. The chlorophyll content of the second newly unfolded leaves at the top was assayed with SPAD-502Plus (Konicaminolta, Japan).

For the plant total P concentration assay, fresh plant samples were heated at 75°C until completely dry, and then ground into powder separately. Approximately 0.05 g of the plant dry samples were weighed and digested with H₂SO₄ and H₂O₂. After cooling, the digested samples were diluted to 100 ml with distilled water. Then, the concentration of P in the solution was determined with 710 ICP-OES (Agilent Technologies, Santa Clara, CA, United States). Each experiment was repeated three times.

For the soluble Pi concentration assay, Tissue Inorganic Phosphorus Content Detection Kit (Solarbio, Beijing, China) was used. Each sample consisted of five seedlings, which were ground into fine powder in liquid nitrogen, repeated three times.

For transgenic soybean root analysis, an Expression 21000XL (Epson, Nagano, Japan) scanner with a root analysis system WinRHIZO 2020 (Regent, Canada) was used. Transgenic Arabidopsis photography and root system analysis were performed as described previously (Li et al., 2020).

Trypan blue was used for the analysis of root activity. The root tips (2 cm) of soybean seedlings were dyed with 0.1% (W/V) trypan blue for 3 min in a boiling water bath, washed with distilled water for 10 min, and the observed under an OLYMPUS CX31 stereo microscope with a SPOT-RT digital camera (Olympus, Melville, NY, United States) attached to a computer.

For ROS accumulation analysis, staining with Diaminobenzidine (DAB) of stress-treated seedlings was performed following the reported protocol (Daudi and O’Brien, 2012). H₂O₂ content was determined according to the method described previously, with slight modifications (Yang et al., 2019b). For the soybean plants, 0.2 g of the root tips were taken as a sample, and for Arabidopsis, five seedlings were taken as a sample. After overnight reaction, the absorbance was read at 390 nm, and H₂O₂ content was determined using a standard curve. Each experiment was repeated three times.

For activity of ROS-scavenging enzymes analysis, soybean root tips (0.2 g) were ground with a 1.6 ml ice-cold 50 mM phosphate-buffered saline (PBS) (pH 7.8). The homogenate was centrifuged at 16,000 × g for 20 min at 4°C, and the supernatant was used for the assays. The activities of SOD, POD, and APX were determined according to established protocols (Luo et al., 2020). Nicotinamide adenine dinucleotide phosphate (NADPH) content was determined by following the method described previously, with slight modifications (Matsumura and Miyachi, 1980; Wei et al., 2019). Homogenate plant tissue samples in 2 ml 0.1 M NaOH solution under ice bath conditions were then centrifuged at 10,000 × g for 20 min at 25°C, and 50 μl of the supernatant of the sampling solution was added with 300 μl of a mixture (0.1 M NaOH, 40 mM EDTA, 4.2 mM MTT, 16.6 mM PMS), followed by 450 μl of 0.1 mM NaCl, and the solution was kept for 5 min at 37°C. Then, the solution was mixed with 50 μl G6P dehydrogenase (0.5 KU/ml) and incubated at 37°C for 40 min. The reaction was terminated with 200 μl of saturated NaCl and centrifuged at 10,000 × g for 10 min at 25°C. The supernatant was discarded, and the precipitate was dissolved with 1 ml 95% ethanol (v/v). After the precipitation was completely dissolved, OD value was measured at 570 nm. Five Arabidopsis seedlings were taken as a sample, and three biological replicates were set in each sample. The NADPH content was expressed as OD₅₇₀/plant FW.

Statistical analyses were performed using Microsoft Excel 2010 and the SAS System for Windows V10. Significant differences were evaluated by two-tailed Student’s t-test or one-way ANOVA and Duncan’s test. All test differences at P < 0.05 were significant. All the error bars were SD value.

A 6PGDH gene (GLYMA_08G254500) was obtained through BLAST searching against the soybean genome utilizing the reported Arabidopsis 6PGDH family protein gene PGD2 as query (Hölscher et al., 2016), and it was named Gm6PGDH1. Gm6PGDH1 was located on Gm8 (22503089-22506054) of soybean, and its full-length coding sequence (CDS) had 486 amino acids with a calculated molecular mass of 53.57 kD and an isoelectric point of 6.4. Amino acid sequence analysis showed that Gm6PGDH1 had a C-terminal 6GPD domain and an N-terminal NAD-binding domain (Figure 1A). Phylogenetic analysis was performed to determine the evolutionary relationship between Gm6PGDH1 and the other plant 6PGDH proteins, among which Gm6PGDH1 was most closely related to At6PGDH2 (Figure 1B). The results of amino acid sequence alignment showed that Gm6PGDH1 was highly conserved, and that the homology between Gm6PGDH1 and At6PGDH1, At6PGDH2, At6PGDH3, Os6PGDH1, and Os6PGDH2 was 76.35, 90.12, 76.76, 85.65, and 70.95%, respectively (Figure 1C).

To further test whether Gm6PGDH1 exhibited an enzymatic activity, the prokaryotic expression vector of Gm6PGDH1 was constructed. The recombinant protein generated by the E. coli Rosetta (DE3) strain expressing plasmid of Histagged Gm6PGDH1 was purified with a Ni-NTA column, and recombinant proteins produced by DE3 strain were detected by SDS-PAGE. A Gm6PGDH1-His fusion protein with an expected size of 54.39 kDa (consisting of target gene and histidine marker) was identified after Coomassie brilliant blue staining (Figure 1D). Meanwhile, the recombinant Gm6PGDH1 protein was assayed for its kinetic properties relative to the substrate NADP⁺. The recombinant proteins were ultrasonic-broken and centrifuged at 4°C, and the supernatant was taken as enzyme crude extracts. Then, the reaction buffer was added into the crude enzyme extracts of Gm6PGDH1-His, and the crude enzyme extracts of an empty carrier were used as the control. The relative activity of bacterial protein was calculated by the slope of the curve, and the results showed that the relative enzyme activity of the prokaryotic-induced Gm6PGDH1 fusion protein was significantly higher than that of the empty carrier (Figure 1E).

To investigate the transcript levels of Gm6PGDH1 in specific tissues, the total RNA was extracted from root, stem, leaf, flower, pod, or seed of the soybean plants at first trifoliate (V1), full bloom (R2), and full seed (R6) stages. The RT-qPCR analysis showed that Gm6PGDH1 was expressed in all the tissues examined, with the highest expression in root at the R2 stage and lowest in leaf at the R6 stage (Figure 2A). The expression patterns of Gm6PGDH1 in the root was altered significantly that the largest transcription was observed at the R2 stage. A similar expression pattern was observed in the leaf (Figure 2A). Then, RNA samples from the root and leaf were used to evaluate the expression patterns of Gm6PGDH1 to Pi starvation by RT-qPCR. The results showed that the expression levels of Gm6PGDH1 significantly increased over time under low Pi conditions. After 10 days of treatment with Pi starvation, the expression level of Gm6PGDH1 in the root and leaf both rose to the maximum value, but the expression level of Gm6PGDH1 in the root was higher than that in the leaf (Figure 2B). Furthermore, the relative levels of Gm6PGDH1 in both the root and the leaf dropped rapidly after recovery of Pi for a day (Figure 2B). These results suggested that Gm6PGDH1 was involved in transcriptional response during soybean Pi starvation adaptions.

Previous literature suggested that there were two forms of 6PGDH in plants: cytosolic 6PGDH and plastidic 6PGDH, and that the N-terminal of plastid 6PGDH amino acid sequence had one more signal peptide sequence compared with cytosolic 6PGDH (Tanksley and Kuehn, 1985; Spielbauer et al., 2013). The domain analysis of Gm6PGDH1 showed no signal peptide at the N-terminal (Figure 1A), so it may be cytosolic 6PGDH. To verify the subcellular localization of Gm6PGDH1, a Gm6PGDH1-green fluorescent protein (GFP) fusion construct and a control construct containing only GFP were generated. Both constructs were transformed into the mesophyll protoplasts of Arabidopsis and observed under a confocal laser scanning microscope. The free GFP alone was present throughout the whole cell, whereas Gm6PGDH1-GFP was specifically located in the cytosol, which demonstrated that Gm6PGDH1 was cytosolic 6PGDH (Figure 2C).

Composite plants with transgenic hairy root attached to wild-type (WT) shoot can be obtained by hairy root transformation, which provides a fast and effective way for functional analysis of genes expressed in soybean root (Zhang et al., 2014; Xue et al., 2017). To produce composite soybean plants consisting of a wild-type shoot with transgenic roots, Agrobacterium rhizogenes should be injected into the cotyledonary node and/or the proximal part of the hypocotyl, which would lead to the development of hairy roots at the infection site. In this case, the chimeric plants have a strong root connection based on direct outgrowth from the vasculature (Attila et al., 2007). By RT-qPCR analysis, the expression level of Gm6PGDH1 in transgenic hairy roots was about twofold over empty vector control hairy roots (Supplementary Figure 1). Under Pi-deficient conditions, the control plants showed more obvious symptoms of P deficiency than the Gm6PGDH1-overexprssing composite plants, such as fewer buds and deep color leaves (Figure 3A). Meanwhile, the dry weight and total P concentration of the roots of the composite transgenic plants were significantly higher than those of the control plants under Pi-deficient conditions, while there was no significant difference under Pi-sufficient conditions (Figures 3C,E). These results suggested that the expression of Gm6PGDH1 in soybean root may affect the P efficiency of plant root by regulating root system development under Pi-deficient conditions. In this experiment, the overexpression of Gm6PGDH1 did not lead to changes in dry weight and total P concentration of the composite transgenic plants leaf, regardless under Pi-deficient or Pi-sufficient condition (Figures 3B,D), that part of the reason may be the shorter processing time.

As an important category of genes involved in the Pi starvation rescue system, PAPs play an important role in plant Pi acquisition (Raghothama, 1999; Tomscha et al., 2004; Liang et al., 2010). Previous studies have shown that the transcript level of 66% GmPAP genes were induced or increased by Pi-deprivation treatment in soybean (Li et al., 2012). In order to further understand the regulatory role of Gm6PGDH1 in Pi starvation, the transcription of five GmPAP genes was analyzed in hairy roots of transgenic composite plants. In the hairy roots of the control plants, Pi starvation significantly induced the expression of the five GmPAP genes (Supplementary Figure 2), which was similar to previous reports (Li et al., 2012, 2017a; Kong et al., 2018). In composite plant hairy roots with overexpressing Gm6PGDH1, however, no significant induction of GmPAP14 and GmPAP21 expression was detected (Supplementary Figure 2). In addition, the induction ratio of the GmPAP3, GmPAP15, and GmPAP17 transcripts in the composite plants hairy roots overexpressing Gm6PGDH1 was reduced by 50–80% compared with the control plants under Pi-deficient conditions (Supplementary Figure 2). There was no significant difference in the expression levels of these genes between composite plants hairy roots with overexpressing Gm6PGDH1 and the control under Pi-sufficient conditions (Supplementary Figure 2). These results suggested that the overexpression of Gm6PGDH1 attenuates the response to Pi starvation of several Pi-starvation inducible GmPAP genes in hairy roots of transgenic composite plants.

Then, we generated three stable overexpressing soybean lines (OX-1, OX-2, and OX-3) to better investigate the role of Gm6PGDH1 response and adaptation to Pi starvation (Supplementary Figures 3A,B). The T3 progeny of 5-day-old transgenic soybean lines and WT seedlings was exposed to a nutrient solution containing a high level of Pi (1 mM Pi, Pi sufficient) and a low level of Pi (2.5 μM Pi, Pi deficient) for 30 days (removal of cotyledons). The three Gm6PGDH1-overexpressing lines showed no significant growth difference compared with WT when grown with the Pi-sufficient solution (Figure 4). However, the Gm6PGDH1-overexpressing lines showed better growth than the WT plants when grown with the Pi-deficient solution (Figure 4A). There was a significant difference in plant height and dry weight between WT and Gm6PGDH1-overexpressing plants under Pi-deficient condition (Figures 4B,C). At the same time, the total P concentrations of the Gm6PGDH1-overexpressing lines were higher than in the WT under Pi-deficient conditions (Figure 4D). Since Pi starvation will lead to a relative increase in chlorophyll content (Devaiah et al., 2009), we monitored the chlorophyll content of all the plants and found that the chlorophyll content of transgenic plants under Pi-deficient conditions was significantly lower than that of WT (Figure 4E). These results showed that the overexpression of Gm6PGDH1 increased the resistance of soybean to Pi starvation.

Since Pi starvation could cause changes in plant root architecture (Péret et al., 2011), we further observed the root development of transgenic soybean under Pi starvation. Under Pi-sufficient conditions, no significant differences in root architecture were observed between the WT and Gm6PGDH1-overexpressing lines (Figures 4F,G). However, the total root length in the Gm6PGDH1-overexpressing lines was significantly longer than that in WT under Pi-deficient conditions (Figure 4F). The number of root tips in Gm6PGDH1-overexpressing lines was significantly higher than that in WT under Pi-deficient conditions (Figure 4G). In addition, trypan blue can enter dead cells and appear blue (Zhang et al., 2011), so we tried to use it to explore the difference in root activity between transgenic soybean lines and WT. Under Pi-deficient conditions, the staining degree of soybean root tip (more accurately the meristematic zone) increased compared with that under P-sufficient conditions, indicating that the number of dead cells in the root tip increased (Supplementary Figure 4). However, we found that the staining degree of transgenic lines was lighter than that of WT, indicating higher root activity of Gm6PGDH1-overexpressing lines than WT under Pi-deficient conditions (Supplementary Figure 4). These results suggested that the improved growth of the Gm6PGDH1-overexpressing lines under Pi-deficient conditions may be partly due to better root system development.

As the OPPP is widely involved in plant redox balance, and 6PGDH is its rate-limiting enzyme (Balzergue et al., 2017), we naturally thought that the overexpression of Gm6PGDH1 might affect the redox system of transgenic soybeans (Lehmann et al., 2009). The T3 progeny of 5-day-old Gm6PGDH1-overexpressing transgenic soybean and WT seedlings were exposed to a nutrient solution containing a high level of Pi (1 mM Pi, Pi sufficient) or a low level of Pi (2.5 μM Pi, Pi deficient) for 10 days (removal of cotyledons). As ROS is a major factor causing oxidative stress, we examined the accumulation of one major ROS, H₂O₂. The root tips of all the treated soybean materials were dyed with DAB. As shown in Figure 5A, Pi starvation accentuates the staining of upper root in the expansion and elongation zones, while the Gm6PGDH1-overexpressing lines were stained to a lighter extent compared with that of WT under Pi-deficient conditions, implying that less ROS was produced in the transgenic soybean lines. Quantitative measurements further demonstrated that H₂O₂ contents in the root tips of the Gm6PGDH1-overexpressing lines were remarkably lower than that of WT (Figure 5B). These results indicated that transgenic lines had better oxidative stress tolerance in a low Pi environment. The important role of ROS-scavenging enzymes in ROS homeostasis prompted us to assess the enzyme activities of SOD, POD, and APX in the root tips of soybean (Gill and Tuteja, 2010). After 10 days of Pi starvation, it was noticeable that the three enzyme activities of the Gm6PGDH1-overexpressing lines were significantly higher than those in the WT (Figures 5C–E). No similar difference was found under Pi-sufficient conditions.

To further elucidate the mechanisms underlying the regulation of antioxidant system by Gm6PGDH1, RT-qPCR was performed to monitor the expression of six oxidative stress-related genes identified by previous transcriptome to be associated with Pi starvation (Zeng et al., 2016). Under Pi-sufficient conditions, the expression levels of these genes showed no significant difference between Gm6PGDH1-overexpressing lines and WT, except that the expression levels of Gm03g36620 and Gm09g02590 were increased in some transgenic soybean lines compared with the WT (Figures 6A–F). At the same time, under Pi-deficient conditions, the transcription level of Gm06g11720 and Gm08g05680 increased (Figures 6A,B), while the transcription level of Gm03g36620, Gm09g02590, Gm09g02650, and Gm15g13491 decreased (Figures 6C–F), which was similar to the results of previous studies (Zeng et al., 2016). Interestingly, the expression of these genes in response to Pi starvation was attenuated in transgenic lines (Figures 6A–F). For example, the expression of Gm06g11720 under Pi-deficient conditions was 2.32 times than that under Pi-sufficient conditions in WT, while the expression of Gm06g11720 under Pi-deficient conditions was 2.14, 1.49, and 2.28 times more than that under Pi-sufficient conditions in OX-1, OX-2, and OE-3, respectively (Figure 6A).

To further characterize the role of Gm6PGDH1 in soybean response to Pi starvation, we attempted to knock down Gm6PGDH1 by virus-induced gene silencing (VIGS). VIGS mediated by Beanpod mottle virus (BPMV) has been successfully performed to study the function of soybean genes involved in defense and other processes, such as GmMPK4s, GmWRKY58, and GmWRKY76 (Liu et al., 2011; Yang et al., 2016). Two weeks post virus inoculation, some plants showed fold deformation of leaf margin (Supplementary Figure 5A), and RT–qPCR analysis showed that the plants inoculated with BPMV-Gm6PGDH1 vectors displayed a reduction of approximately 60% in the transcript levels of Gm6PGDH1 when compared with control plants (Supplementary Figure 5B). The identified plants were subjected to Pi starvation for 7 days (Supplementary Figure 5C). There was no significant difference in the dry weight and total P concentration of the leaves and roots between BPMV-Gm6PGDH1-silenced plants and control plants (Supplementary Figures 5D,E). These results showed that the VIGS of Gm6PGDH1 did not affect dry weight and total P concentration of soybean leaves and roots under low Pi conditions.

In order to further verify the role of Gm6PGDH1 in plant resistance to Pi starvation, Gm6PGDH1-overexpressing Arabidopsis plants were constructed. Seeds of T3 transgenic Arabidopsis and the WT were vertically cultured in a plant gel medium with high Pi (1 mM Pi) or low Pi (62.5 μM Pi) for 14 days. Under high Pi conditions, there was no significant difference in the growth status between the transgenic Arabidopsis lines and WT (Figures 7A,B). However, under low Pi conditions, the transgenic Arabidopsis lines grew much better than WT (Figure 7A). The fresh weight of the transgenic Arabidopsis lines was greater than that of WT under low Pi conditions (Figure 7B). At the same time, we found that the short-term low Pi supply significantly inhibited the growth of Arabidopsis, especially the root growth (Figure 7A), indicating that the length of primary and lateral roots length of plants grown with low Pi was shorter (Figures 7A,C,D). However, the primary and lateral roots of the transgenic Arabidopsis lines were significantly longer than those of the WT after 14 days of treatment with low Pi (Figures 7C,D). Furthermore, the soluble Pi concentration of the transgenic plants was significantly higher than that of WT under low Pi conditions (Figure 7E). These results indicated that the overexpression of Gm6PGDH1 significantly promoted root system development and Pi accumulation in transgenic Arabidopsis under low Pi conditions.

Then, the ROS levels in the Gm6PGDH1-overexpressing Arabidopsis lines and WT were explored. We treated 7-day-old WT and transgenic Arabidopsis lines in an agar medium with high Pi (1 mM Pi) and low Pi (0 mM Pi) for 36h, then used DAB to stain the root tips of all the plants. Under low Pi conditions, the color of plant root was heavier, reflecting the elevated level of ROS (Figure 7F). However, the slighter coloration of DAB staining was observed in transgenic Arabidopsis lines root tip compared with WT under low Pi conditions (Figure 7F). Meanwhile, quantitative measurements showed that the content of H₂O₂ in the transgenic Arabidopsis lines was markedly lower than that in the WT under low Pi conditions (Figure 7G). Furthermore, we measured the NADPH content of the plants, since the OPPP contributes to the production of NADPH to support antioxidant enzyme activity (Valderrama et al., 2006; Lehmann et al., 2009). We found that the content of NADPH in the transgenic Arabidopsis lines was higher than that in WT under low Pi conditions (Figure 7H). These results suggested that the overexpression of Gm6PGDH1 might alleviate the root tip ROS accumulation by improving the content of NADPH under low Pi conditions.