The P sojae-susceptible soybean cultivar “Dongnong 50” and the resistant cultivar “Suinong10” were used. Seeds of the resistant cultivar “Suinong 10” were grown in an incubator at 25°C, 70% relative humidity under a 16 h light/8 h dark cycle, and the seedlings at stage V1 (the first true leaves were about to unfold; Fehr et al., 1971) were inoculated with P. sojae zoospores according to the methods described by Ward et al. (1979) and Yang et al. (1996) with minor modifications. Seeds of the susceptible cultivar “Dongnong 50” were obtained from the Key Laboratory of Soybean Biology in the Chinese Ministry of Education, Harbin, and this cultivar was used for genetic transformation experiments.

The online database NCBI (National Center for Biotechnology Information) was used for finding the sequences of high homology genes with GmPSMD. DNAMAN software¹ was used for the sequence multiple alignments, and a phylogenetic analysis of GmPSMD was carried out using MEGA software. The GmPSMD protein structure was predicted using Phyre2².

Total RNA was isolated using Trizol reagent and reverse-transcribed based on the manufacturer’s instructions (Invitrogen, China). cDNA was synthesized from 1 μg of total RNA using a Super Script first-strand cDNA synthesis system (Takara, Dalian, China). qRT-PCR was performed on a LightCycler96 instrument (Roche, Switzerland) using a real-time PCR kit (ToYoBo, Japan). The soybean GmEF1β housekeeping gene (GenBank accession no. NM_001248778) was used as an internal control to normalize all data (Supplementary Table S1). The relative transcript level of target genes was calculated using the 2^–ΔΔCT method. Three biological repeats for each line were performed in each experiment.

Yeast two-hybrid assays were performed using the Frozen-EZ Yeast Transformation II kit (The Epigenetics Company, United States). The full-length GmPSMD was inserted into the pGADT7 expression vector, and GmPIB1 was inserted into the pGBKT7 expression vector. Fusion plasmids pGADT7-GmPSMD and pGBKT7-GmPIB1 were transformed into yeast strain Y₂HGold (Takara Bio, Japan) to identify protein–protein interactions. pGBKT7-53 and pGADT7-SV40 plasmids were used as a positive control, while pGBKT7-Lam and pGADT7-SV40 plasmids were used as the negative control in yeast cells.

To construct GmPIB1-ccluc and GmPSMD-nluc, the coding sequences of GmPIB1 and GmPSMD were cloned into the plant expression vectors pCAMBIA1300-ccluc and pCAMBIA1300-nluc using gene-specific primers, respectively (Supplementary Table S1). The recombinant plasmids were transferred into Agrobacterium tumefaciens GV3101, and transformed Agrobacterium strains resistant to kanamycin and rifampicin were selected and cultured in YEP medium (containing 0.5% Yeast, 1% Peptone, 0.5% NaCl) at 28°C until the OD₆₀₀ reached 0.8–1.2. Agrobacterium cells were collected by centrifugation at 4,000 rpm for 10 min, resuspended in infection liquid (containing 10 mM MgCl₂, 10 mm MES, 150 μmol acetosyringone, pH 5.6) and placed at room temperature (25°C) for 2–3 h. The two recombinant plasmids were mixed in equal volumes and injected into Nicotiana benthamiana. Fluorescence signals were observed after dark culture for 2 days.

The open reading frames of GmPSMD and GmPIB1 were, respectively, fused to the GST-tag of vector PGEX-4T-1 and the 6 × His-tag of vector pET29b (+) (Novagen, Germany) and then transformed into Transetta (DE3) Escherichia coli cells (TransGen Biotech, China). GST-tagged and His-tagged proteins were induced with 0.5 mM isopropyl-β-D-thiogalactoside (IPTG) at 37°C for 4 h. GmPIB1-His and GmPSMD-GST fusion proteins were purified using a GST-Sefinose kit (Sangon, China) or a His-bind Purification Kit (Merck Millipore) at 4°C and subsequently detected by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotting using anti-His and anti-GST antibodies (Abmart, United States), respectively.

The GST fusion protein was fused with 50 μL GST beads (Sangon, China) at 4°C for 1 h then washed three times with 1 × phosphate-buffered saline (PBS) containing 1% Triton-100 before adding the purified fusion protein with His-tag and incubating at 4°C for 4 h. The washing step was repeated, and 40 μL 5 × loading buffer was added for immunoblotting detection.

To construct the pCAMBIA3301-GmPIB1 and pCAMBIA3301-GmPSMD overexpression vector, the coding sequence of GmPIB1 and GmPSMD was cloned into the plant expression vector pCAMBIA3301 with a C-terminal 4 × Myc fusion sequence, respectively. For RNA interference performance, the specific GmPSMD cDNA fragment was amplified and inserted into the vector PFGC5941 (Kerschen et al., 2004). “Dongnong 50” was used to generate transgenic soybean hairy roots by A. rhizogenes-mediated transformation following the instructions described by Paz et al. (2004). Transgenic soybean hairy roots were preliminarily detected by PCR (Primers shown in Supplementary Table S1), and then the level of gene overexpression or silencing were detected by qRT-PCR and those overexpressing GmPIB1 and GmPSMD were identified by immunoblotting with an anti-Myc antibody (Abmart, United States).

Resistance to P. sojae was assessed based on methods described by Ward et al. (1979) with minor modifications. GmPSMD-transformed “Dongnong 50” soybean hairy roots were cultured for about 2 weeks and were placed on a tray with clean and moist gauze. All hairy roots were slightly scratched at the same position with a sterile scalpel, and zoospores of P. sojae race 1 were used to inoculate the wounds. Empty vector (EV) soybean hairy roots were used as controls. Disease symptoms were recorded by photography. The P. sojae biomass was analyzed based on the accumulation of P. sojae TEF1 (GenBank accession no. EU079791), PSEL1 (GenBank accession no. CF840149), and PSEL2 (GenBank accession no. CF839332) in the transgenic soybean hairy roots. The pathogen response assays were performed on three biological replicates, each with three technical replicates.

Total ROS production was determined according to the instructions of Reactive Oxygen Species Assay Kit (Beyotime Institute of Biotechnology, China). Fluorescence was detected at 530 nm emission wavelength and 485 nm excitation wavelength using a Microplate Reader (Bio-TEK, United States; Qian et al., 2009). For the enzyme assays, 0.2 g of soybean transgenic hairy roots was ground with 25 mM HEPES buffer (pH 7.8) containing 0.2 mM EDTA, 2 mM ascorbate and 2% PVP. The homogenate was centrifuged at 4°C for 15 min at 12,000 × g and the resulting supernatant was used for the determination of the enzymatic activity following the method described by Cao et al. (2009). The superoxide dismutase (SOD) activity was measured as the increase in the absorbance at 560 nm according to instructions of Superoxide Dismutase Assay Kit (COMIN, Institute of Biotechnology, Suzhou of China). The catalase (CAT) activity was measured as the decline in the absorbance at 240 nm due to the decrease of extinction according to the instructions of Superoxide Dismutase Assay Kit (COMIN, Institute of Biotechnology, Suzhou of China). The guaiacol peroxidase (GPX) activity was measured as the increase in the absorbance at 340 nm due to guaiacol oxidation according to the instructions of Superoxide Dismutase Assay Kit (COMIN, Institute of Biotechnology, Suzhou of China). The peroxidase (POD) activity was measured as the increase in the absorbance at 470 nm according to the instructions of Superoxide Dismutase Assay Kit (COMIN, Institute of Biotechnology, Suzhou of China). The detection of total ROS and antioxidant enzyme activity assays were performed on three biological replicates, each with three technical replicates.

Hydrogen peroxide (H₂O₂) accumulation was determined according to the method of Velikova et al. (2000). Superoxide anion (O₂^–) accumulation in GmPSMD-transformed and EV soybean hairy roots was determined as follows: soybean hairy roots were inoculated with P. sojae to induce production of O₂^–, and samples were taken at 0, 12, and 24 h. Then the samples were thoroughly ground and vortexed, and 500 μL of 10 mM phosphate buffer (PH = 7.8) was added. After centrifuging at 4°C and 5,000 rpm for 10 min, the supernatant was added with 700 μL phosphate buffer and 100 μL of 10 mM hydroxylamine hydrochloride at 25°C for 20 min. Finally, 0.5 mL of P-aminobenzenesulfonic acid and 0.5 mL of naphthylamine were added and reacted at 25°C for 20 min to read the absorbance of the supernatant at 530 nm.

The 26S proteasome activity assays were performed as previously described by Kisselev and Goldberg (2005). Briefly, GmPSMD and GmPIB1 transgenic soybean hairy roots were inoculated with P. sojae and the samples were taken at 0, 12, and 24 h. The proteasome activity buffer 50 Mm Tris-HCl (pH = 7.5), 250 mM sucrose, 5 mM MgCl₂, 0.5 mM EDTA, 2 mM ATP, 1 mM DTT, 1% of the total volume of PSMF and Roche inhibitor were prepared and added to the sample in proportion. The homogenate was centrifuged at 4°C for 15 min at 10,000 × g and the resulting supernatant was used for the determination of 26S proteasome activity. The supernatant, buffer and fluorescent substrate (proteasome substrate Ø, fluorogenic) were added to a 96-well plate (BD Flacon) at a ratio of 10:137.5:2.5. Fluorescence was detected at 460 nm emission wavelength and 380 nm excitation wave length at 25°C. The changes of fluorescence units in 10 min were recorded.

To explore the proteins interacting with GmPIB1, the transgenic soybean hairy roots overexpressing GmPIB1 were obtained using Agrobacterium rhizogenes-mediated transformation. Proteins interacting with GmPIB1 were screened by immunoprecipitation-mass spectrometry and verified by yeast two-hybrid library screening. pCAMBIA3301-GmPIB1-myc recombinant vector was successfully constructed (Supplementary Figure S1A) and the immunoblotting analysis confirmed successful production of GmPIB1 in transgenic soybean hairy roots (Supplementary Figure S1B). Then the immunoprecipitation was conducted and the specific fluorescent signal was observed in positive hairy roots (Figure 1A), indicating that protein activity of GmPIB1 was good. Specific stripes for co-immunoprecipitating proteins were observed in the experimental group, showing that the precipitation complex contained not only the target protein GmPIB1, but also proteins that may interact with GmPIB1. Several stripes observed in the experimental group and not in the control by silver staining (Figure 1B) were removed for mass spectrometry. completed by Wuhan GenecreateBiological Engineering Company. A total of 392 proteins that might interact with GmPIB1 were retrievaled and analyzed by Proteinpilot software, which were mainly related to energy metabolism, gene expression regulation, and transportion. Among the 392 proteins, 20 candidate genes were selected based on the identification score of mass spectrometry data (Supplementary Table S2). To verify interactions between proteins encoded by the candidate genes and GmPIB1, the full-length GmPIB1 gene was inserted into vector pGBKT7 and the candidate genes into vector pGADT7. Recombinant vectors were transformed into yeast strain Y₂H Gold and titrated on SD/-Leu/-Trp and SD/-Ade/-His/-Leu/-Trp plates to verify protein–protein interactions. Only yeast co-transformed with pGADT7-284 (GmPSMD) and pGBKT7-GmPIB1 grew on SD/-Ade/-His/-Leu/-Trp plates (Supplementary Figure S2). Moreover, GmPSMD had no self-activating activity and the yeast two-hybrid results show that GmPSMD and GmPIB1 interact in yeast (Figure 1C).

We further verified the interaction between GmPIB1 and GmPSMD using pull-down assays in vitro and luciferase complementation assays in vivo. The expression levels of the fusion proteins GmPIB1-HIS and GmPSMD-GST were very high at 1, 2, and 4 h after IPTG induction (Figure 2A), indicating that the proteins were well purified and could be used for pull-down assays (Figure 2B). As shown by immunoblotting analysis, GmPSMD could interact with full-length GmPIB1 in vitro (Figure 2C). We then infiltrated Agrobacterium containing the recombinant plasmids pCAMBIA1300-GmPIB1-ccluc and pCAMBIA1300-GmPSMD-nluc into Nicotiana benthamiana leaves and observed uimioluminiscence at the infiltration areas after dark incubation for 2 days, and the results showed that GmPIB1 and GmPSMD can interact in vivo (Figure 2D).

GmPSMD is located on chromosome 6 with a full length of 1,803 bp and a coding sequence of 1,269 bp; the encoded 423 amino acids contain a PCI domain belonging to the 26S proteasome 19S regulatory subunit (Figure 3A). Phylogenetic tree and alignment analyses revealed that PSMD is divided into four sub-families, A, B, C, and D, among which soybean GmPSMD and Mucuna pruriens MpPSMD, Lupinus angustifolius LaPSMD, Abrus precatorius ApPSMD, Vigna radiata. VrPSMD, Phaseolus vulgaris PvPSMD, Vigna angularis VaPSMD, and Cicer arietinum CaPSMD belong to subfamily A.

The amino acids of PSMD proteins from species belonging to the A subfamily in the phylogenetic tree were selected and analyzed by multi-column alignment using DNAMAN (Figure 3B). The sequence similarity of GmPSMD with MpPSMD, LaPSMD, ApPSMD, VrPSMD, VaPSMD, CaPSMD, and PvPSMD was 70.91, 94.55, 95.97, 95.97, 96.21, 94.55, and 95.02%, respectively (Figure 3C). GmPSMD had the highest similarity of 96.21% with VaPSMD and the lowest similarity of 70.91% with MpPSMD. The tertiary structure of GmPSMD was predicted using the website http://swissmodel.expasy.org. The protein consisted mainly of random coil, α-helix, β-sheet and extended long chains. There were 20 alpha helices and four beta folds (Figure 3D).

To investigate whether GmPSMD is involved in the stress response induced by P. sojae, qRT-PCR was used to examine the transcript levels of GmPSMD in soybean cultivar “Suinong 10” infected with P. sojae. The relative expression of GmPSMD increased after inoculation with P. sojae race 1 and reached the highest level after 36 h (Figure 4A). This indicated that GmPSMD may be involved in the defense response against P. sojae.

To further explore the defense response of GmPSMD against P. sojae. GmPSMD-OE and GmPSMD-RNAi transgenic soybean hairy roots were obtained and the level of overexpression and silencing of GmPSMD transgenic soybean hairy roots were analyzed by qRT-PCR. The relative expression of GmPSMD in GmPSMD-OE transgenic soybean hairy roots was remarkably higher than that in control, while it showed lower level than control in GmPSMD-RNAi transgenic soybean hairy roots (Figure 4B). Characterization of disease resistance in GmPSMD transgenic soybean roots of the susceptible cultivar “Dongnong 50” revealed that the GmPSMD-OE soybean hairy roots turned slightly brown, while those transformed with pCAMBIA3301 empty vector and GmPSMD-RNAi showed rotting and browning of the inoculated parts (Figure 4C), and the lesion areas of GmPSMD-OE transgenic soybean hairy roots were significantly smaller than that in the control, while the GmPSMD-RNAi soybean hairy roots were significantly larger than that in control (Figure 4D).

Moreover, the relative biomass of P. sojae based on transcript levels of the P. sojae TEF1 (GenBank accession no. EU079791), PSEL1 (GenBank accession no. CF840149), and PSEL2 (GenBank accession no. CF839332) in infected soybean hairy roots after 2 days of incubation with zoospores of P. sojae was significantly (^∗∗P < 0.01) lower in GmPSMD-OE lines than in EV hairy roots. Conversely, the biomass of P. sojae in GmPSMD-RNAi roots was higher than that in the control (Figures 5A–C). These results indicated that overexpression of GmPSMD can improve the resistance of soybean hairy roots to P. sojae, while GmPSMD-RNAi transgenic soybean hairy roots showed increased susceptibility to P. sojae.

ROS are key signaling molecules in the interaction between plants and pathogens under stress (Mittler et al., 2004; Shigeoka and Maruta, 2014). To determine if GmPSMD affects the resistance of soybean hairy roots to P. sojae by affecting the production of ROS, the relative ROS levels in GmPSMD-RNAi, GmPSMD-OE, and EV transgenic soybean hairy roots were analyzed at 0, 12, and 24 h after inoculation with P. sojae. Accumulation of ROS in GmPSMD-RNAi hairy roots reached significant levels after 24 h compared with that in control, while ROS accumulation in GmPSMD-OE hairy roots was lower than that in control (Figure 6A).

ROS include hydrogen peroxide (H₂O₂) and superoxide anions (O₂^–). Therefore, the contents of H₂O₂ and O₂^– were further analyzed after inoculation with P. sojae in EV, GmPSMD-OE, and GmPSMD-RNAi transgenic soybean hairy roots. The relative H₂O₂ levels in all three transgenic soybean hairy roots increased gradually with the extension of inoculation time. However, the accumulation of H₂O₂ in GmPSMD-OE transgenic soybean hairy roots was lower than that in the control, while H₂O₂ accumulation in transgenic soybean hairy roots of GmPSMD-RNAi was higher than that in the control (Figure 6B). Levels of O₂^– in transgenic soybean hairy roots were not remarkably different from those of the control (Figure 6C). SOD, POD, CAT and GPX as the main antioxidant enzymes participated in the scavenging of reactive oxygen species and alleviated the damage of membrance system (Zhang et al., 2008). The antioxidant enzymes activity of SOD, POD, CAT, and GPX were analyzed after inoculation with P. sojae in EV, GmPSMD-OE, and GmPSMD-RNAi transgenic soybean hairy roots. The antioxidant enzymes activity increased gradually with the extension of inoculation time. However, the antioxidant enzymes activity in GmPSMD-OE transgenic soybean hairy roots was higher than that in control, while the antioxidant enzymes activity in transgenic soybean hairy roots of GmPSMD-RNAi was lower than that in the control (Figure 7).

To verify if GmPSMD affects 26S proteasome activity of transgenic soybean hairy roots and explore its role in response to P. sojae infection. The 26S proteasome activity was measured in GmPSMD-OE, GmPSMD-RNAi, GmPIB1-OE, and EV transgenic soybean hairy roots inoculated with P. sojae. The results showed that the 26S proteasome activity in hairy roots of GmPSMD-OE and GmPIB1-OE was remarkably higher than that in control, while the 26S proteasome activity in GmPSMD-RNAi soybean hairy roots was remarkably lower than that in the control after infection with P. sojae (Figure 8A). These results indicate that GmPSMD could increase the 26S proteasome activity in transgenic soybean hairy roots during P. sojae infection. We also found that the expression levels of GmPIB1 in GmPSMD-OE transgenic soybean hairy roots and GmPSMD in GmPIB1-OE transgenic soybean hairy roots increased, respectively, which proved that they were a pair of positive regulatory factors (Figures 8B,C).