Soybean ecotype Williams 82 was used for the phenotypic and functional analysis in this study. Healthy soybean seeds were chlorinated and planted into vermiculite after 3-day germination. Furthermore, 7-day seedlings were inoculated with Bradyrhizobium diazoefficiens USDA110 and grown in a growth chamber under a 16 h light:8 h dark cycle at 28°C. The seedlings were watered with nitrogen-free nutrient solution (containing 2.5 mM K₂SO₄, 2 mM MgSO₄•7H₂O, 1 mM KH₂PO₄, 0.15 mM FeCl₂, 1.5 mM CaSO₄2•H₂O, 46 μM H₃BO₃, 9.1 μM MnCl₂•4H₂O, 0.75 μM ZnSO₄, 0.5 μM CuSO₄, 0.11 μM MoO₃, 9.4 × 10^–2 μM CoCl₂•6H₂O).

The soybean plants were harvested at 28 days post inoculation (dpi) for measuring the fresh weight, dry weight, plant height, and N content of shoot. Nodules were separately harvested for measuring nitrogenase activity, nodule number, and fresh weight. For quantitative real-time PCR (qRT-PCR) analysis, nodules were harvested at 10, 17, 21, 28, and 35 dpi. All tissues were frozen in liquid nitrogen and stored at –80°C for further mRNA and protein analyses.

Total RNA was extracted using the RNAprep Pure Plant Kit (Tiangen, Beijing, China), and cDNA was synthesized using a PrimeScript™ RT reagent Kit with gDNA Eraser (Takara, Otsu, Shiga Prefecture, Japan). SYBR Premix EX Taq™ (Takara, Otsu, Shiga Prefecture, Japan) was used for qRT-PCR analysis by a CFX96™ real-time system (Bio-Rad, Berkeley, CA, United States). The qRT-PCR reaction conditions were: 95°C 30 s, 95°C 10 s, 56°C 15 s, and 72°C 10 s. The fold change in the expression of each sample was standardized using GmActin11 gene and analyzed by the 2^–ΔΔCT method (Livak and Schmittgen, 2001). Lowercase letters represent statistically significant differences (p < 0.05) according to Tukey’s HSD test; asterisks indicate statistically significant differences according to Student’s t-test (two-tailed) (**p < 0.01). All experiments were repeated at least three times. The primers of qRT-PCR are listed in Supplementary Table 1.

A GmHSP17.1 protein was cloned into pET-28a (+) vector and introduced into the Escherichia coli strain, BL21 (DE3) (EMD Chemicals Inc., Gibbstown, NJ, United States). GmHSP17.1 proteins were induced by 0.5 mM isopropyl-β-d-thiogalactoside for 4 h at 28°C and purified by Ni-Agarose Resin (Lot 01376/10531, CWBIO, China) according to the manufacturer’s instructions. Green fluorescent proteins (GFPs) were purified in the same way as GmHSP17.1 and used as negative controls in this experiment. The chaperone activity of GmHSP17.1 was performed by measuring the chemically induced aggregation of insulin (from bovine pancreas, Sigma-Aldrich, St. Louis, MO, United States) and thermal aggregation of malate dehydrogenase (MDH; from porcine heart, Sigma-Aldrich) according to previous protocols (Yang et al., 2021).

Promoter of GmHSP17.1, 2.4-kb length upstream of ATG, was cloned into pBI121 vector to generate pHSP17.1::β-glucuronidase (GUS) construct. pHSP17.1::GUS was transformed into Agrobacterium rhizogenes strain K599 for further hairy roots transformation, as described previously (Kim et al., 2013). The nodules on the transgenic hairy root were harvested at 10, 17, 21, and 28 dpi for GUS staining. GUS staining was performed as described previously (Luo et al., 2013; Zhong et al., 2013).

The full-length open reading frame (ORF) of GmHSP17.1 was cloned into pCamE-GFP vector between SalI and BamHI enzyme sites for an overexpression analysis. For the RNAi constructs, about 200 bp fragment specific to GmHSP17.1 was inserted between BamHI and KpnI, SpeI and SacI, respectively, in pTCK-303-GUS vector, as described previously (Du et al., 2016; Wang et al., 2020). Empty vectors alone were used as negative controls. Then, the vectors were transformed into hairy roots through Agrobacterium rhizogenes strain K599, as described previously (Kim et al., 2013). GFP fluorescence signal and GUS staining were used to identify the positive transgenic hairy roots.

For the observation of infection cells, three nodules randomly selected at 28 dpi were fixed in formaldehyde-acetic acid solution for 24 h at 4°C with three independent experiments. After embedding in paraffin, 5 μm sections were prepared using a microtome (RM2016). After dewaxing, the nodule sections were stained with 0.1% Toluidine Blue and images of three sections for each nodule were captured with a scanner (Pannoramic DESK, P-MIDI, P250). The percentage of area of infection cells to total cells in a nodule section and surface area of 100 infection cells were calculated with Image-Pro Plus 6.0 software.

The coding DNA sequence (CDS) of GmHSP17.1 was cloned into the pET-28a-Avi (+) vector to generate His-GmHSP17.1-Avi fusion protein, which contains an additional Avi-tag at the C-terminal end (Du et al., 2015). Then, the constructed vector was transformed into the E. coli strain, BL21 which was pre-transformed with BirA for biotinylation (Tirat et al., 2006). Fusion protein of GmHSP17.1 was induced by 0.5 mM IPTG for 4 h at 28°C and purified by affinity chromatography using streptavidin agarose resin (Thermo Fisher Scientific, Waltham, MA, United States). GFP was used as a negative control. The total protein of soybean nodule was extracted with Plant Protein Extraction Reagent (CWBIO, Beijing, China) and incubated with purified GmHSP17.1 protein at 4°C. The targeted proteins were separated by affinity chromatography, and then, purified three times with buffer solution. The purified proteins were isolated by electrophoresis with 12% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gel and were further analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) (Supplementary Table 2).

The CDSs of GmHSP17.1 or GmRIP1 was cloned into 326-GFP vector to generate GmHSP17.1-GFP of GmRIP1-GFP, respectively, for subcellular localization. GFP fluorescence was captured using a confocal microscope (Fvi10, OLYMPUS, Tokyo, Japan). The CDSs of GmHSP17.1 and GmRIP1 were cloned into the p326YFP^N and p326YFP^C, respectively, to generate GmHSP17.1-YFP^N, GmHSP17.1-YFP^C, GmRIP1-YFP^C, and GmRIP1-YFP^N for BiFC assay. The constructed vectors were transformed into Arabidopsis protoplasts, according to the methods described previously (Yoo et al., 2007). Yellow fluorescent protein (YFP) fluorescence was observed using a confocal microscope (Fvi10, OLYMPUS, Tokyo, Japan).

The interaction between GmHSP17.1 and GmRIP1 was verified by a yeast two-hybrid experiment using the Matchmaker Gold Yeast Two-Hybrid System according to the instructions (Clontech, 630489, Mountain View, CA, United States). GmHSP17.1 was cloned into pGBKT7 as bait (pGBKT7-GmHSP17.1), while the CDS of GmRIP1 was cloned into pGADT7 as prey (pGADT7-GmRIP1). The pGBKT7-GmHSP17.1 and pGADT7-GmRIP1 were co-transformed into yeast-two hybrid (Y2H) Gold cell, and then incubated on SD/-Leu-Trp plates at 30°C for 3 days. The positive clones were then transferred to SD/-His-Ade-Leu-Trp plates and SD/-His-Ade-Leu-Trp + X-α-Gal + AbA (Takara, Otsu, Shiga Prefecture, Japan) and incubated at 30°C for 5 days to confirm the interaction. In addition, pGBKT7-53 and pGADT7-T were used as positive controls, and pGBKT7-Lam and pGADT7-T as negative controls. The primers used in the Y2H assay are listed in Supplementary Table 1.

The CDS of GmRIPP1 was cloned into pET-28a (+) vector to generate His-tag fusion protein (His-GmRIP1), and the CDS of GmHSP17.1 was cloned into the pET-28a-Avi (+) vector to generate His-GmHSP17.1-Avi fusion protein. These constructs were transformed into the E. coli strain BL21 for protein induction. His-GmHSP17.1-Avi was purified using a streptavidin agarose resin, and then incubated with the total proteins isolated from BL21 expressing His-GmRIP1. Finally, the western blot was performed with anti-His monoclonal antibody (Invitrogen, Carlsbad, CA, United States) (Du et al., 2010). The primers used in pull-down assay are listed in Supplementary Table 1.

Nitrogenase activity was measured by acetylene reduction assay according to the protocol described previously (Oh et al., 2001). The dried nodule samples were nitrate-boiled and the N content was measured using Semimicro-Kjeldahl determination method in a nitrogen analyzer as described previously (Wang et al., 2020).

Peroxidase activity of nodules was measured using a Peroxidase Activity Detection Kit (Solarbio, China) according to the manufacturer’s instructions. The content of reactive oxygen species (ROS) of nodules was measured using a Plant ROS Kit (Chenglin, Beijing, China), according to the manufacturer’s instructions.

Statistical analyses were performed using GraphPad Prism 7 software.

An sHSP, GmHSP17.1, was identified in the proteome of mature nodules of soybean, and the expression profile in various organs showed that the transcript abundance of GmHSP17.1 in nodules was about 2.7 times higher than in flowers, followed by roots, stems, and leaves (Figure 1A). Next, the transcript accumulation of GmHSP17.1 was determined via qRT-PCR in soybean roots inoculated with Bradyrhizobium diazoefficiens USDA110, we found that GmHSP17.1 was induced more strongly in nodules. The transcript of GmHSP17.1 increased gradually during the nodule development until 28 dpi and decreased at 35 dpi (Figure 1B). GmHSP17.1 gene (Glyma.06g157800) was located on chromosome 6, and a gene model analysis revealed that GmHSP17.1 had only one exon and no intron, with a length of 832 bp for the predicted mature transcript, 106 bp for the 5′ untranslated region (UTR), 273 bp for the 3′ UTR, and an ORF of 453 bp (Figure 2A). The ORF of GmHSP17.1 encoded a predicted protein of 150 amino acid residues with a conserved α-crystallin domain, a defined domain for sHSPs family in plant species (Figures 2B,C). Furthermore, phylogenetic tree showed that GmHSP17.1 belonged to the CI subfamily of sHSPs (Figure 2D), which were reported to be localized in cytoplasm (Waters, 2013). sHSPs are known to be molecular chaperones in living cells (Haslbeck and Vierling, 2015; Carra et al., 2017). The molecular chaperone activity of GmHSP17.1 was confirmed by insulin and MDH as conventional substrates. The results indicated that GmHSP17.1 efficiently prevented chemically induced aggregation of insulin by DTT and thermal aggregation of MDH at 45°C while no holdase activity was detected in the presence of GFP (Supplementary Figures 1A,B).

To further analyze the expression pattern of GmHSP17.1 in soybean nodules, 2,400 bp promoter sequence upstream of start codon was fused with the reporter gene GUS, and the resulting construct GmHSP17.1pro::GUS was then used to generate transgenic composite hairy roots and nodules. Histochemical staining analysis indicated that GmHSP17.1 was preferentially expressed during nodule development, consistent with the qRT-PCR results (Figure 3A). To further understand the subcellular localization of GmHSP17.1, GmHSP17.1 was fused with reporter gene GFP (GmHSP17.1-GFP) driven by the CaMV 35S promoter. GmHSP17.1-GFP was then transfected into Arabidopsis protoplasts, and strong GFP expression signal was captured in the cytoplasm, in agreement with GFP control (Figure 3B). In the meantime, the subcellular localization of GmHSP17.1 was also verified in the protoplasts of soybean transgenic hairy roots. A strong GFP fluorescence signal was also observed in the cytoplasm of root protoplasts, as well as cytosolic protein GFP used as a marker (Figure 3C). Taken together, these data showed that GmHSP17.1 was preferentially expressed in nodules, indicating an important role in nodule development in soybean.

To investigate the function of GmHSP17.1 in nodule development and BNF capacity in soybean, we performed phenotypic analyses of transgenic composite soybean plants either overexpressing or suppressing of GmHSP17.1 (Figures 4A,B). The success of transformation in transgenic hairy roots was determined by qRT-PCR, and we found that the expression of GmHSP17.1 in overexpression lines was 4-fold of the expression in control lines, while in suppression lines, the expression of GmHSP17.1 was reduced by 57.7% (Figure 4C). The nodule number, fresh weight, and nodule size increased by 38.8, 56.6, and 12.5% in overexpression lines, while decreased by 52.5, 65.6 and 27.6% in suppression lines, respectively, in comparison with control lines (Figures 4D,F). Consistently, the expression of leghemoglobin gene GmLbc3 was also increased in overexpression lines and decreased in the RNAi lines (Figure 4G). As a result, nitrogen fixation efficiency was markedly affected due to increased and decreased nitrogenase activity in GmHSP17.1 overexpressed and RNAi nodules, respectively (Figure 4H). Furthermore, the infection cells of nodules were examined by toluidine blue staining and we found the GmHSP17.1 overexpressed nodules displayed larger infection cells and infection areas, while RNAi nodules showed smaller infection cells and infection zone, compared with that of control lines (Figure 5).

In addition, soybean plant growth was significantly influenced due to an altered expression of GmHSP17.1 in transgenic composite plants (Supplementary Figure 2A). An overexpression of GmHSP17.1 resulted in the increase of 48.4, 33.2, 33.8, and 8.2% in plant height, shoot fresh- and dry weight, and N content; on the other hand, the suppression of GmHSP17.1 showed inhibition of plant height, shoot fresh- and dry weight, and N content by 34.3, 24.4, 26.0, and 27.3%, respectively, compared with that of control lines (Supplementary Figures 2B–E). Taken together, these results indicate that GmHSP17.1 affect the nodule development and nitrogen fixation, and subsequently along with the plant growth.

To further illuminate the molecular mechanism of GmHSP17.1 in nodules, interacting proteins of GmHSP17.1 were isolated by LC-MS/MS (Supplementary Figure 3A). A peroxidase protein, homology of a rhizobium-induced peroxidase (Rip1) identified in M. truncatula (Goormachtig et al., 1995; Ramu et al., 2002), designated GmRIP1 was chosen as the first candidate of GmHSP17.1 (Supplementary Table 2). The phylogenetic analysis showed that GmRIP1 belonged to class III peroxidases which include all secretory plant-specific peroxidases (Supplementary Figure 3B). In the meantime, we found that the peroxidase activity of nodules was increased during nodule development (Supplementary Figure 3C). These data indicated that peroxidases may have important roles in nodules. Next, the interaction between GmHSP17.1 and GmRIP1 was further confirmed by an independent Y2H assay. BD-GmHSP17.1 and AD-GmRIP1 or BD-GmRIP1 and AD-GmHSP17.1 constructs were co-transformed into Y2H cells, respectively, and positive colonies were selected on SD/-Trp-Leu-His-Ade + X-α-gal + AbA medium (Figure 6A). To verify the interaction between GmHSP17.1 and GmRIP1 in vivo, we performed the bimolecular fluorescence complementation (BiFC) analysis in Arabidopsis protoplasts. A strong YFP fluorescence signal was detected in the cytoplasm of Arabidopsis protoplasts expressing GmHSP17.1-YFP^N and GmRIP1-YFP^C or GmHSP17.1-YFP^C and GmRIP1-YFP^N, whereas no YFP fluorescence was observed in the negative control combinations GmHSP17.1-YFP^N/YFP^C and GmRIP1-YFP^N/YFP^C (Figure 6B). In addition, the interaction between GmHSP17.1 and GmRIP1 was also confirmed by pull-down assay using recombinant purified proteins in E. coli (Figure 6C). Taken together, we conclude that GmHSP17.1 directly interacts with GmRIP1 in vivo and in vitro.

Peroxidases are important enzymes, acting as antioxidants, in plants that involves in the production and scavenging of ROS, such as superoxide radicals and H₂O₂ (Wang et al., 2015). In the nodulation process, ROS was produced in response to rhizobium infection and could be detected later in infection thread and infection zones of nodules (D’Haeze et al., 2003). Subcellular localization of GmRIP1 was determined in Arabidopsis protoplasts. GmRIP1-GFP fusion protein driven by the CaMV 35S promoter was transiently expressed in Arabidopsis protoplasts, and the expression of GmRIP1-GFP was detected in the cytoplasm (Figure 7A). Next, expression profiles of GmRIP1 in various tissues of soybean were analyzed by qRT-PCR, and the data showed that GmRIP1 was highly expressed in nodules and roots than in stem, leaf, and flower (Figure 7B). Expression of GmRIP1 in different developmental stages of nodules was determined, and we found that GmRIP1 was gradually increased during nodule growth (Figure 7C). All these data indicated that GmRIP1 played a key role in the nodulation in soybean.

To further confirm the interaction effect between GmHSP17.1 and GmRIP1, peroxidase activity of GmRIP1 was measured in GmHSP17.1 overexpression and suppression lines, and we found that peroxidase activity increased by 63.3% in overexpression lines and decreased by 35.1% in suppression lines (Figure 8A). As a result, the content of ROS was significantly affected by the altered peroxidase activity. The content of ROS was greatly decreased by 7.8% in GmHSP17.1 overexpression lines and increased by 25.9% in suppression lines (Figure 8B). Taken together, we conclude that GmHSP17.1 conferred nodule development and nitrogen fixation partially through the regulating peroxidase activity of GmRIP1.