Editing a gibberellin receptor gene improves yield and nitrogen fixation in soybean

ABSTRACT

Soybean is an important source of oil, protein, and feed. However, its yield is far below that of major cereal crops. The green revolution increased the yield of cereal crops partially through high‐density planting of lodging‐resistant semi‐dwarf varieties, but required more nitrogen fertilizers, posing an environmental threat. Genes that can improve nitrogen use efficiency need to be integrated into semi‐dwarf varieties to avoid the overuse of fertilizers without the loss of dwarfism. Unlike cereal crops, soybean can assimilate atmospheric nitrogen through symbiotic bacteria. Here, we created new alleles of GmGID1‐2 (Glycine max GIBBERELLIN INSENSITIVE DWARF 1‐2) using clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR‐associated nuclease 9 (Cas9) editing, which improved soybean architecture, yield, seed oil content, and nitrogen fixation, by regulation of important pathways and known genes related to branching, lipid metabolism, and nodule symbiosis. GmGID1‐2 knockout reduced plant height, and increased stem diameter and strength, number of branches, nodes on the primary stem, pods, and seeds per plant, leading to an increase in seed weight per plant and yield in soybean. The nodule number, nodule weight, nitrogenase activity, and nitrogen content were also improved in GmGID1‐2 knockout soybean lines, which is novel compared with the semi‐dwarf genes in cereal crops. No loss‐of‐function allele for GmGID1‐2 was identified in soybean germplasm and the edited GmGID1‐2s are superior to the natural alleles, suggesting the GmGID1‐2 knockout mutants generated in this study are valuable genetic resources to further improve soybean yield and seed oil content in future breeding programs. This study illustrates the pleiotropic functions of the GID1 knockout alleles with positive effects on plant architecture, yield, and nitrogen fixation in soybean, which provides a promising strategy toward sustainable agriculture.

Knockout of the soybean gibberellin receptor gene GmGID1‐2 reduced plant height; strengthened stems; increased the number of branches, nodes, pods, and seeds; and improved yield, seed oil content and nitrogen fixation.

INTRODUCTION

Soybean (Glycine max) is one of the most important oil crops worldwide and is a main source of vegetable oil and protein (Graham and Vance, 2003). Soybean yield is far below that of cereal crops; the worldwide average yield of soybean, wheat, rice, and maize was 2.72, 3.45, 4.63, and 5.71 t/ha, respectively, from 2013 to 2022 (FAO, https://www.fao.org/statistics/en/) (Micronesia, 2008). The green revolution greatly increased the yield of major cereal crops, such as rice and wheat (Evenson and Gollin, 2003), partially due to the high‐density planting of semi‐dwarf varieties, which are more lodging‐resistant but require more nitrogen fertilizer to increase the tiller number and, thus, yield (Li et al., 2003; Liao et al., 2019; Wu et al., 2020). To achieve the green revolution, mutant alleles at semi‐dwarf 1 (sd1) in rice (Sasaki et al., 2002; Spielmeyer et al., 2002) and reduced height 1 (Rht1) in wheat (Peng et al., 1999) were used to reduce plant height by enhancing DELLA protein activity and accumulation, which repressed plant growth and inhibited nitrogen use efficiency (Li et al., 2018a; Swarbreck et al., 2019). A new green revolution should integrate key genes for nitrogen use efficiency into semi‐dwarf varieties to avoid fertilizer overuse (Li et al., 2018a; Liu et al., 2021b). In soybean, plant height should not be excessively reduced (Liu et al., 2020) to carry a certain number of pods and seeds. Also, differing from cereal crops, which require large amounts of inorganic nitrogen fertilizers to achieve high yields, soybean, as a legume, can assimilate atmospheric nitrogen through symbiotic bacteria to provide more than 70% (Li et al., 2023) of the necessary nitrogen (Herridge et al., 2008). Previous studies have suggested that DELLA represses plant growth (Xu et al., 2014), while positively regulating rhizobia infection (Mcadam et al., 2018) and nodule symbiosis (Ferguson et al., 2011; Fonouni‐Farde et al., 2016; Jin et al., 2016) in Pisum sativum and Medicago truncatula. Genes that can improve both plant architecture and nitrogen fixation need to be explored to achieve a sustainable green revolution in soybean. The gibberellin (GA) receptor GID1 (GIBBERELLIN INSENSITIVE DWARF 1) regulates DELLA degradation in the presence of GA (Mcginnis et al., 2003; Sasaki et al., 2003; Ueguchi‐Tanaka et al., 2007). We therefore propose that the improvement of plant architecture, yield, and nitrogen fixation can be achieved by editing GID1 to promote DELLA accumulation in soybean.

To test this hypothesis, in this work, transgenic soybean lines overexpressing (OE) GmGID1‐2 were obtained, and GmGID1‐2 OE increased plant height. We then generated the GmGID1‐2 knockout (KO) mutants using clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR‐associated nuclease 9 (Cas9) editing. This resulted in a reduced plant height, improved soybean architecture, yield, seed oil content and nitrogen fixation, which proved our hypothesis and provides a promising strategy toward sustainable agriculture. GmGID1‐2‐mediated regulation of plant height, branching, seed oil content, and nitrogen fixation was further investigated using molecular and transcriptomic analyses. This is the first report on the functions of GID1 in soybean. Compared with the previous studies on GID1s in other plants (Ueguchi‐Tanaka et al., 2005; Griffiths et al., 2006; Cheng et al., 2019; Illouz‐Eliaz et al., 2019), our results discovered new roles of GmGID1‐2 in regulation of nitrogen fixation and seed oil content in addition to plant architecture and yield, and illustrates the pleiotropic functions of the GID1 ^KO alleles with positive effects on plant architecture, yield, and nitrogen fixation in soybean, which broadens our understanding of the function of GID1 and provides a predominant gene for soybean improvement.

RESULTS

GmGID1‐2 regulates plant height by modulating DELLA in soybean

There are five GmGID1 genes (GmGID1‐1, GmGID1‐3, GmGID1‐4, and GmGID1‐5) in soybean, and GmGID1‐2 was more abundant than the other four genes in most soybean tissues (Figure S1), indicating its potential important role in soybean. The co‐localization of GmGID1‐2 and the nuclear marker H2B‐mCherry showed that GmGID1‐2 was located in the nucleus (Figure S2). To analyze the function of GmGID1‐2, we generated transgenic soybean lines OE or KO GmGID1‐2 (GmGID1‐2 ^OE and GmGID1‐2 ^KO) in Williams 82 background. The relative expression levels of GmGID1‐2 in GmGID1‐2 ^OE‐1, GmGID1‐2 ^OE‐2, and GmGID1‐2 ^OE‐3 lines were 106‐, 125‐, and 153‐fold of that in Williams 82 (Figure S3A), while the 1‐bp insertion and 11‐bp deletion resulted in early termination of GmGID1‐2 translation in GmGID1‐2 ^KO‐1 and GmGID1‐2 ^KO‐2 lines (Figure S3B–D). GmGID1‐2 ^OE lines were significantly taller and thinner while GmGID1‐2 ^KO lines were shorter and thicker than Williams 82 (Figure 1A–C). The longitudinal sections of soybean hypocotyls (Figure 1D, E) showed that the average cell length significantly increased and the cell width significantly decreased in GmGID1‐2 ^OE lines compared with Williams 82 (Figure 1F, G), in contrast to that in GmGID1‐2 ^KO lines (Figure 1H, I). These results suggest that GmGID1‐2 modulates plant height and stem diameter by regulating cell length and width.

Figure 1.

GmGID1‐2 (Glycine max GIBBERELLIN INSENSITIVE DWARF 1‐2) regulates plant height and cell size by interacting with DELLA in soybean

(A–C) Phenotype (A), plant height (B), and stem diameter (C) of transgenic soybean lines (in the Williams 82 background) and wild‐type (WT) (Williams 82) at V1 stage (when the first trifoliolate leaf was fully unfolded) in the greenhouse. Phenotype (A), scale bar, 5 cm. Plant height (B) and stem diameter (C), data are means ± SE (n = 9 plants per line). Different lowercase letters above bars indicate significant differences determined by one‐way analysis of variance, followed by Duncan's multiple range test at an α level of 0.05. (D, E) Longitudinal sections of soybean hypocotyls as shown in (A). Scale bar, 50 μm. (F–I) Length (F, H) and width (G, I) of hypocotyl cells, as shown in (D, E). Data are means ± SE (n = 30 cells). Statistical significances were determined using two‐tailed Student's t‐tests. ****P < 0.0001, compared with Williams 82. (J) Yeast two‐hybrid (Y2H) assays showing the interactions of GmGID1‐2 with DELLAs (DELLA1 and DELLA3 are shown here, DELLA2, DELLA4, DELLA5, DELLA6, DELLA7, and DELLA8 are shown in Figure S4) in the absence or presence of Gibberellic acid (GA₃) (100 μmol/L). DDO/X‐α‐Gal, double dropout medium (without Leu/Trp). QDO/X‐α‐Gal quadruple dropout medium (without Ade/Leu/Trp/His). Negative control, BD‐Lam + AD‐T. Positive control, BD‐53 + AD‐T. (K–M) Luciferase complementation imaging (K, L) and co‐immunoprecipitation (Co‐IP) (M) assays confirming that GmGID1‐2 interacts with DELLA1 and DELLA3. The color scale in (K, L) represents the fluorescence signal intensity. (N) DELLA protein abundance in the shoot apical meristem (SAM) of soybean plants at V1 stage determined using Western blot. Actin, loading control.

It is suggested that GID1 interacted with DELLA, a growth repressor, to regulate plant height (Ueguchi‐Tanaka et al., 2005). To verify the interactions between soybean DELLA proteins and GmGID1‐2, we identified all eight DELLA proteins in soybean and performed yeast two‐hybrid (Y2H) experiments. GmGID1‐2 strongly interacted with GmDELLA1 and GmDELLA3 (Figure 1J) compared with the other DELLA proteins (Figure S4), and the interactions depended on GA₃. The interaction of GmGID1‐2 with DELLA1 and DELLA3 was further validated by luciferase (LUC) complementation (Figure 1K, L) and co‐immunoprecipitation (Co‐IP) assays (Figure 1M). In previous studies, GID1 KO resulted in more DELLA protein accumulation in rice and Arabidopsis, while OE lines showed the opposite pattern (Ueguchi‐Tanaka et al., 2005; Griffiths et al., 2006). Therefore, we examined the DELLA protein levels in GmGID1‐2 ^OE and GmGID1‐2 ^KO lines using Western blot (Figure 1N). DELLA proteins were reduced in GmGID1‐2 ^OE lines but increased in the GmGID1‐2 ^KO lines compared with those in Williams 82 (Figures 1N, S5). These results support that GmGID1‐2 affects plant height by regulating DELLA protein levels in soybean.

Previous studies have shown that GID1 represses DELLA accumulation (Ueguchi‐Tanaka et al., 2007), decreasing GA20ox expression (Zentella et al., 2007) and thus reducing GA synthesis (Yamaguchi, 2008). We verified the negative feedback regulation (Figure S6A) of the GA biosynthesis (Zentella et al., 2007; Fukazawa et al., 2014, 2017) in this study. DELLA protein level (Figures 1N, S5), relative GmGA20ox expression (Figure S6B, C), and the total content of active GA (Figure S6D), including GA₁, GA₃, and GA₄ (Figure S6E–G) that are the most common active forms of GA, were decreased in the GmGID1‐2 ^OE lines but significantly increased in the GmGID1‐2 ^KO lines, as expected.

GmGID1‐2 KO improves plant architecture, yield, and seed oil content in soybean

We investigated the effect of GmGID1‐2 on important agronomic traits and yield in soybean at maturity. Compared with Williams 82, GmGID1‐2 ^KO lines showed reduced plant height (Figures 2A, B, S7A) and seed size (Figure S8), but increased number of branches, nodes, pods, and seeds per plant (Figures 2C–F, S7B–E). Thus, GmGID1‐2 ^KO lines showed improved seed weight per plant (Figures 2G, S7F). GmGID1‐2 ^KO lines had more three‐ and four‐seed pods (Figures 2H–J, S7G), higher maximum pod number per clump and increased average seed number per pod than Williams 82 (Figure S9). Moreover, the stem diameter and strength of the GmGID1‐2 ^KO lines were higher than those of Williams 82 (Figure 2K–M), suggesting that GmGID1‐2 KO could also improve the lodging resistance of soybean plants through stronger stalks. When grown in the field, the yield of GmGID1‐2 ^KO lines increased 9.69%–17.16% compared to that of Williams 82 (Figure 2N–P). GmGID1‐2 ^KO lines also had higher content of seed oil (Figures 2Q, R, S10A, B) and lower content of seed protein than Williams 82 (Figures 2S, T, S10C, D), while GmGID1‐2 ^OE lines had lower content of seed oil and higher content of seed protein (Figure S11). The plant height of GmGID1‐2 ^OE lines was too high (Figure S12A), and they could not grow upright normally. Although the number of branches and nodes of GmGID1‐2 ^OE lines did not change significantly (Figure S12B, C), the other yield‐related traits, such as pod number, seed number, and seed weight per plant, decreased compared to Williams 82 (Figure S12D–F). Overall, GmGID1‐2 KO improved plant architecture (stronger and thicker stems, more branches, nodes, pods and seeds), yield, and seed oil content in soybean. We also found that GmGID1‐2 did not affect the seed germination (Figure S13) and had little effect on the flowering time (Figure S14) of soybean, suggesting that GmGID1‐2 could be a valuable gene for improving soybean yield with little negative effect on other important agronomic traits.

Figure 2.

GmGID1‐2 knockout improves plant architecture, yield, and seed oil content in soybean

(A) Plant architecture of field‐grown soybean. Scale bar, 10 cm. (B–G) Plant height (B), number of branches per plant (C), number of nodes on main stem (D), number of pods per plant (E), number of seeds per plant (F), and seed weight per plant (G) of soybean grown in the field at Hefei in the years 2023 and 2024, respectively. Data are means ± SE (n = 18 plants per line, there are three replications within each year, with six plants per line within each repeat). (H) Representative image showing the frequencies of four‐, three‐, two‐, and one‐seed pods on individual soybean plant grown in the field. Scale bar, 2 cm. (I, J) Number of four‐, three‐, two‐, and one‐seed pods on individual soybean plant grown in the field at Hefei in the years 2023 (I) and 2024 (J), respectively. Data are means ± SE (n = 18). (K) Stem of field‐grown soybean. Scale bar, 1 cm. (L, M) Stem diameter (L) and stem strength (M) of soybean plants grown in the field at Hefei in 2023. Data are means ± SE (n = 18). (N) Representative image displaying 20 plants per soybean line grown in the field (Hefei, 2023). Scale bar, 10 cm. (O, P) Plot yield of soybean grown in the field at Hefei in the years 2023 (O) and 2024 (P). Data are means ± SE (n = 3 replicates per year). (Q–T) Seed oil (Q, R) and protein (S, T) content of soybean grown in the field at Hefei in the years 2023 and 2024, respectively. Data are means ± SE (n = 18). In (B–G, I, J, L, M and O–T), statistical significances were determined using two‐tailed Student's t‐tests. NS, not significant, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, compared to Williams 82.

GmGID1‐2 affects the transcript abundance of genes involved in branching and lipid metabolism

To investigate how GmGID1‐2 affects the branch number in soybean, we collected the lateral buds (Figure 3A) for RNA sequencing (RNA‐seq) analysis, which could provide us a genome‐wide transcriptomic change in pathways as well as known and novel genes involved in branching mediated by GmGID1‐2. A total of 1,706 up‐regulated and 573 down‐regulated differentially expressed genes (DEGs) were identified between Williams 82 and GmGID1‐2 ^KO lines (Figure 3B). Among these DEGs, Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) terms related to plant hormone signal transduction, zeatin biosynthesis, cell wall biogenesis, response to auxin, lignin catabolic process, and GA‐mediated signaling pathway were significantly enriched (Figure 3C). These results are consistent with the previous report that plant hormones regulate lateral bud growth (Dong et al., 2023). We further performed reverse‐transcription quantitative polymerase chain reaction (RT‐qPCR) to examine the expression patterns of soybean genes homologous to known branching‐related genes, such as negative regulators TB1 (TEOSINTE BRANCHED1) which encodes a TB1, CYCLOIDEA, and PROLIFERATING CELL FACTOR (TCP) domain transcription factor (Doebley et al., 1997; Guo et al., 2013); DWARF 14 (D14) which encodes a receptor for the branching‐repressing phytohormone strigolactone (SL) (Yao et al., 2016; Liu et al., 2022), and positive regulators DWARF 53 (D53), which encodes a repressor of SL signaling (Jiang et al., 2013; Zhou et al., 2013); NGR5 (NITROGEN‐MEDIATED TILLER GROWTH RESPONSE 5), which encodes an APETALA2‐domain (AP2) transcription factor (Wu et al., 2020). Soybean genes homozygous to the known negative regulators of branch numbers, including GmTB1a, GmTB1b, and GmD14, were significantly down‐regulated in GmGID1‐2 ^KO lines, whereas soybean genes homozygous to the known positive regulators, including GmD53, GmAP2a, and GmAP2b, were up‐regulated in GmGID1‐2 ^KO lines (Figure 3D–F), which explains why the branch numbers were significantly increased when GmGID1‐2 was knocked out in soybean.

Figure 3.

GmGID1‐2 knockout affects the transcript abundance of genes involved in branching and lipid metabolism

(A) Lateral buds (indicated by the yellow arrow) of soybean plants at 20 d after germination for RNA sequencing (RNA‐seq). Scale bar, 5 cm. (B) Volcano plot showing differentially expressed genes (DEGs) in the lateral buds, as shown in (A), between Williams 82 and GmGID1‐2 ^KO lines. FDR, false discovery rate. FC, fold change. (C) Enriched Kyoto Encyclopedia of Genes and Genomes (KEGG), and Gene Ontology (GO) terms among the DEGs in soybean lateral buds between Williams 82 and GmGID1‐2 ^KO lines. Up‐ and down‐regulated genes are shown in red and blue, respectively. (D–F) Relative expression of GmTB1a and GmTB1b, which are homologous to maize TB1 (D); GmD14 and GmD53 associated with strigolactones (E); and GmAP2a and GmAP2b, which are homologous to rice NGR5 (F), in soybean lateral buds by reverse‐transcription quantitative polymerase chain reaction (RT‐qPCR); messenger RNA (mRNA) abundance values are relative to that of Williams 82 (set as 1), and GmUKN1 was used as the reference gene. Data are means ± SE (n = 3 biological replicates). (G) Soybean seeds at 30 day after flowering (DAF) for RNA‐seq. Scale bar, 1 cm. (H) Volcano plot showing DEGs in developing soybean seeds at 30 DAF, as shown in (G), between Williams 82 and GmGID1‐2 ^KO lines. (I) Enriched KEGG and GO terms among the DEGs in soybean seeds between Williams 82 and GmGID1‐2 ^KO lines. Up‐ and down‐regulated genes are shown in red and blue, respectively. (J, K) Relative expression of GmZF351a, GmZF351b, GmZF392, and GmNFYA, which are known genes affecting soybean oil content, by RT‐qPCR; mRNA abundance values are relative to that of Williams 82 (set as 1), and GmUKN1 was used as the reference gene. Data are means ± SE (n = 3 biological replicates). In (D–F, J, K), statistical significances were determined using two‐tailed Student's t‐tests. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, compared to Williams 82.

We also performed RNA‐seq analysis using the developing soybean seeds at 30 d after flowering (DAF) (Miao et al., 2020) to explore the GmGID1‐2‐mediated regulation of genes and pathways in seeds (Figure 3G–K). There were 732 up‐regulated and 2,868 down‐regulated genes between Williams 82 and GmGID1‐2 ^KO lines (Figure 3H). Among these DEGs, KEGG and GO terms related to carbon fixation, alpha linolenic acid metabolism, linolenic acid metabolism, and so forth, were significantly enriched (Figure 3I). The relative transcript abundance of known positive regulators of oil content, including GmZF351a, GmZF351b, GmZF392, and GmNFYA (Li et al., 2017a; Lu et al., 2021), was significantly increased in GmGID1‐2 ^KO lines (Figure 3J, K). These results indicate that GmGID1‐2 regulates genes and pathways related to lipid metabolism to affect seed oil content in soybean.

GmGID1‐2 KO enhances nodule formation and nitrogen fixation

The above results showed that GmGID1‐2 ^KO lines had a significantly increased number of branches, nodes, pods, seeds, and final yield (Figures 2C–G, O, P, S7B–F), more three‐ and four‐seed pods (Figures 2I, J, S7G), and a higher maximum pod number per clump (Figure S9), suggesting that GmGID1‐2 ^KO lines have a better “source” to supply more nutrients to the sink tissues, such as pods and seeds. As a typical legume crop, the main nitrogen source in soybean comes from symbiotic nitrogen fixation (Li et al., 2023). Therefore, we investigated the effect of GmGID1‐2 on nodule formation and nitrogen fixation. Compared with Williams 82, the number and dry weight of nodules per plant, as well as individual nodule weight increased in GmGID1‐2 ^KO lines but decreased in GmGID1‐2 ^OE lines (Figure 4A–D). Moreover, the nitrogenase activity in nodules (Figure 4E), ureide content in stems (Figure 4F), and total nitrogen content in shoots (Figure 4G) increased on average by 16.69%, 34.99%, and 20.96%, respectively, in GmGID1‐2 ^KO lines compared to Williams 82, suggesting that GmGID1‐2 knockout resulted in an improved nitrogen fixation ability, which could provide more nitrogen from the “source” to the “sink” organs.

Figure 4.

GmGID1‐2 knockout enhances nitrogen fixation in soybean

(A) Phenotype of nodules. Scale bar, 0.5 cm. (B–G) Number of nodules per plant (B), dry weight of nodules per plant (C), dry weight of individual nodule (D), nitrogenase activity (U/g fresh weight) in nodules (E), ureide content (nmol/g dry weight) in stems (F), and nitrogen content in shoots (G) of Williams 82 and transgenic soybean lines. Data are means ± SE (n = 4 plants per line). Different lowercase letters above the bars indicate significant differences determined by one‐way analysis of variance, followed by Duncan's multiple range test at an α level of 0.05, while bars with the same letters are not significantly different.

To unveil the causes of increased number and dry weight of root nodules per plant in GmGID1‐2 ^KO lines, we investigated the effect of GmGID1‐2 on the formation and development of root nodules (Figure 5). After being inoculated by rhizobia, root hairs that grew in waves and formed tight curls were referred as deformed hairs and infection foci, respectively. Our results showed that GmGID1‐2 ^KO lines had more deformed root hairs, infection foci and nodule primordia compared with Williams 82, while GmGID1‐2 ^OE lines showed the opposite (Figure 5A–F), indicating that the increased number of nodules could result from increased rhizobia infections. We also observed the development of nodules at 3, 6, 10, and 14 day after inoculation (DAI). Compared with Williams 82 and GmGID1‐2 ^OE, GmGID1‐2 ^KO lines formed nodules earlier, which could lead to bigger nodules (Figure 5G–I), indicating that the increased nodule dry weight is likely due to earlier nodulation.

Figure 5.

GmGID1‐2 knockout increases rhizobia infection and early nodulation in soybean

(A–C) Phenotype of deformed root hairs (A), infection foci (B) and nodule primordia (C). In (A) and (B), the yellow arrows indicate deformed root hairs and infection foci. Scale bar, 50 μm. In (C), scale bar, 300 μm. At 6 day after inoculation (DAI), 2‐cm root segments below the soybean root‐hypocotyl junction were cut and stained with 1% (w/v) methylene blue. (D–F) Number of deformed root hairs (D), number of infection foci (E), and number of nodule primordia (F) of GmGID1‐2 ^OE, Williams 82 and GmGID1‐2 ^KO lines. Data are means ± SE (n = 9 plants per line). Different lowercase letters above the bars indicate significant differences determined by one‐way analysis of variance, followed by Duncan's multiple range test at an α level of 0.05. (G–I) The phenotypes of nodule development at 3, 6, 10 and 14 DAI. Scale bar, 0.5 cm.

To further explore the reasons for the earlier nodulation in GmGID1‐2 ^KO lines, we investigated the expression of early symbiosis marker genes (Wang et al., 2019, 2022; He et al., 2021; Xu et al., 2021a; Chen et al., 2022b; Yun et al., 2022), including GmNSP1a, GmNINa, and GmENOD40, at the initial rhizobia infection stage. Compared with Williams 82, these genes were significantly up‐regulated in GmGID1‐2 ^KO lines but down‐regulated in GmGID1‐2 ^OE lines (Figure 6A–C), consistent with the observed phenotype at early nodulation stage. To determine how GmGID1‐2 affects nodule symbiosis in soybean, we performed RNA‐seq analysis of roots and nodules at 28 DAI (Wang et al., 2019). A total of 2,374 up‐regulated and 2,855 down‐regulated genes were identified between Williams 82 and GmGID1‐2 ^KO lines (Figure 6D). Among these DEGs, KEGG and GO terms related to isoflavonoid biosynthesis, phenylpropanoid biosynthesis, flavone and flavonol biosynthesis, nitrogen metabolism, glucose import, and hydrogen peroxide catabolic processes were significantly enriched (Figure 6E). These results are consistent with the previous reports that flavonoids (Subramanian et al., 2007; Chen et al., 2023) and peroxides (D'haeze et al., 2003; Minguillón et al., 2022) are necessary signaling factors for nodule symbiosis. Flavonoids also have antioxidant properties (Liu et al., 2024), which therefore might protect nitrogenase from reactive oxygen species (ROS) damage during the nitrogen fixation process (Dalton et al., 1998) in mature nodules. Nodule symbiosis requires sugar to provide energy (Ke et al., 2022), and increased nitrogen assimilation was found with stronger symbiotic nitrogen fixation (Carter and Tegeder, 2016). Next, we investigated the expression patterns of known key genes involved in soybean nodule symbiosis, nitrogen fixation and assimilation. The transcripts of genes that are primarily associated with nodule symbiosis including GmNAC181, GmENOD93, GmNINa, and GmNSP1a (Wang et al., 2019, 2022; He et al., 2021), genes relevant to nitrogen fixation including GmNAS1 and GmNAP1 (Ke et al., 2022), and genes related to nitrogen assimilation including GmGS1 and GmGS1‐2 (Masalkar and Roberts, 2015), were more abundant in the root nodules of GmGID1‐2 ^KO lines than in Williams 82 (Figure 6F). These results suggest that GmGID1‐2 modulates nodule fixation by regulating several important metabolic pathways and known genes associated with nodule symbiosis and nitrogen fixation.

Figure 6.

GmGID1‐2 knockout affects the expression of genes related to nodule symbiosis

(A–C) Relative expression of early symbiosis marker genes, GmNSP1a, GmNINa, and GmENOD40, in soybean roots at 1 day after inoculation (DAI). The expression levels are relative to that in Williams 82 (set as 1), and GmUKN1 was used as the reference gene. Data are means ± SE (n = 3 biological replicates). Different lowercase letters above the bars indicate significant differences determined by one‐way analysis of variance, followed by Duncan's multiple range test at an α level of 0.05, while bars with the same letters are not significantly different. (D) Volcano plot showing differentially expressed genes (DEGs) between Williams 82 and GmGID1‐2 ^KO lines based on RNA sequencing (RNA‐seq) data of roots and nodules at 28 DAI. FDR, false discovery rate. FC, fold change. (E) Enriched Kyoto Encyclopedia of Genes and Genomes (KEGG), and Gene Ontology (GO) terms among the DEGs in soybean root nodules between Williams 82 and GmGID1‐2 ^KO lines. Up‐regulated genes are shown in red, and no enriched KEGG and GO terms were identified among down‐regulated genes. (F) Heatmap of the expression of genes associated with root nodule symbiosis in soybean. The color scale represents Log₂(FPKM + 1). FPKM, fragments per kilobase of transcript per million mapped reads. Data represent means of three biological replicates. Statistical significances were determined using two‐tailed Student's t‐tests. Fold change, ratio of the transcript abundance in GmGID1‐2 ^KO versus Williams 82.

Natural variation in GmGID1‐2 is associated with plant height, seed oil content, and yield in soybean germplasm

To determine if the natural variation in GmGID1‐2 was associated with important agronomic traits in soybean, we analyzed the sequence variation in GmGID1‐2 among soybean germplasm. Among the 264 soybean accessions, there are no other sequence variations but only two single nucleotide polymorphisms (SNPs) leading to synonymous mutations were identified in the coding sequence (CDS) of GmGID1‐2 (Table S1). Only three synonymous SNPs (Table S2) were identified in the CDS of GmGID1‐2 in the SoyOmics (Liu et al., 2023) database (https://ngdc.cncb.ac.cn/soyomics/index), and no sequence variations were found in the CDS of GmGID1‐2 in SoyGVD database (https://yanglab.hzau.edu.cn/SoyGVD/#/), suggesting that there is no natural loss‐of‐function allele for GmGID1‐2. Therefore, we further analyzed the sequence variations in the GmGID1‐2 promoter region (2,000 bp upstream). A total of 17 SNPs with a minor allele frequency (MAF) ≥ 0.05 were identified and used for regional association analysis among 264 soybean accessions (Table S3). The SNP at Gm03:36395036 in the GmGID1‐2 promoter region was significantly associated with plant height (Figure 7A, B). Soybean accessions carrying the Gm03:36395036_C (Pro‐C) allele of GmGID1‐2 had a reduced plant height and seed size (100‐seed weight) than those carrying the Gm03:36395036_T (Pro‐T) allele (Figure 7C, D). To investigate the allelic effect of the SNP at Gm03:36395036 on GmGID1‐2 expression, RT‐qPCR assays were performed. Soybean accessions carrying Pro‐C had lower GmGID1‐2 expression levels than those with Pro‐T (Figure 7E). Furthermore, promoter LUC assays showed that Pro‐C had a lower promoter activity than Pro‐T (Figure 7F, G). These results suggest that the natural Pro‐C allele of GmGID1‐2 is associated with lower GmGID1‐2 expression and reduced plant height, consistent with the observed phenotype of GmGID1‐2 ^KO lines.

Figure 7.

Natural variation in GmGID1‐2 and pleiotropic effects of GmGID1‐2 knockout

(A) Manhattan plot for regional association analysis of plant height in 264 soybean accessions using a mixed linear model. The horizontal line in red represents the significant threshold (P < 1/n, n is the number of markers). The red dot represents the strongest association at Gm03:36395036. (B) Gene structure of GmGID1‐2. Pro‐C and Pro‐T correspond to two promoter types of GmGID1‐2 with C or T at Gm03:36395036, respectively. (C, D) Plant height (C) and 100‐seed weight (D) of soybean accessions with Pro‐C and Pro‐T types, using the data from field experiments at Dangtu (Anhui province) and Xuzhou (Jiangsu province). (E) Expression level of GmGID1‐2 in shoot apical meristems (SAMs) of soybean accessions carrying Pro‐C and Pro‐T at V1 stage (when the first trifoliolate leaf was fully unfolded) by reverse‐transcription quantitative polymerase chain reaction (RT‐qPCR); GmUKN1 was used as the reference gene. (F) Promoter‐luciferase (LUC) image of Pro‐C and Pro‐T by transient expression in tobacco leaves using an in vivo plant imaging system. The LUC reporter gene was driven by Pro‐C or Pro‐T promoter. The color scale represents the fluorescence signal intensity. (G) Promoter activity analysis of Pro‐C and Pro‐T. Data are means ± SE (n = 5 biological replicates). (H) Distribution of Pro‐C and Pro‐T in 264 soybean accessions originated from different ecoregions in China. (I) GmGID1‐2 knockout improves yield, plant architecture, seed oil content, and nitrogen fixation in soybean. In (C–E), statistical significances were determined using Wilcoxon tests. *P < 0.05, ***P < 0.001. In (G), statistical significance was determined using a two‐tailed Student's t‐test. ***P < 0.001.

We used the SoyGVD database to further investigate the effect of natural variation in GmGID1‐2 on soybean agronomic traits. Compared with soybean accessions carrying the Pro‐T type of GmGID1‐2, those with the Pro‐C type of GmGID1‐2 had a lower plant height and protein content, but higher oil content and yield (Figure S15). Thus, the Pro‐C allele of GmGID1‐2, which is the major allele in different ecoregions in China, was superior in the natural soybean population (Figure 7H). When GmGID1‐2 was knocked out in Williams 82, which also had the Pro‐C allele, the yield and seed oil content were further increased by 13.99% and 1.64%, respectively (Figure 7I; Table 1). The plant height was further reduced by 5.89%, while the node number and stem diameter of the primary stem increased by 9.18% and 17.27%, respectively, and the number of branches, pods, seeds per plant increased by 60.00%, 27.60%, and 31.11%, respectively, in GmGID1‐2 ^KO lines compared with Williams 82 (Table 1). These results suggest that the new loss‐of‐function alleles of GmGID1‐2 generated in this study could be used as valuable genetic resources to further improve soybean plant architecture, yield, and seed oil content (Table 1).

Table 1.

Effects of GmGID1‐2 knockout in the Williams 82 (Pro‐C type) background

	Williams 82 background (Pro‐C type)
Trait	Williams 82 (Mean ± SE)	GmGID1‐2 KO (Mean ± SE)	Increase or reduce (%)
Yield (t/ha)	2.93 ± 0.11	3.34 ± 0.11	13.99
Oil content (%)	21.98 ± 0.06	22.34 ± 0.06	1.64
Plant height (cm)	106.97 ± 1.32	100.67 ± 1.08	−5.89
Node number on primary stem	20.59 ± 0.33	22.48 ± 0.20	9.18
Diameter of primary stem (mm)	6.31 ± 0.12	7.40 ± 0.14	17.27
Branch number	2.75 ± 0.11	4.40 ± 0.11	60.00
Pod number per plant	57.28 ± 1.50	73.09 ± 1.16	27.60
Seed number per plant	152.14 ± 3.96	199.47 ± 3.29	31.11
Seed weight per plant (g)	18.46 ± 0.58	23.33 ± 0.43	26.38
100‐weight (g)	15.51 ± 0.21	14.16 ± 0.10	−8.70
Protein content (%)	42.74 ± 0.12	42.04 ± 0.12	−1.64

Data shown in this table are from six replications of the field trials at Hefei in the years 2023 and 2024 (three replications per year), under the planting density of 200,000 plants/ha. For yield, data are means ± SE (n ≥ 6, three plots of 6 m² per line for each year). For other traits, data are means ± SE (n ≥ 36, six plants per line for each repeat). The average values of the GmGID1‐2 ^KO are from two knockout (KO) lines (n = 72).

DISCUSSION

The green revolution successfully boosted cereal crop yields through high‐density planting of semi‐dwarf varieties under a high fertilizer regime. Semi‐dwarfism is conferred by the Rht1 mutation in wheat and the sd1 allele in rice, both leading to DELLA protein accumulation (Liu et al., 2021a). Although DELLA protein accumulation reduces plant height, it also inhibits the absorption of inorganic nitrogen fertilizer in cereal crops (Swarbreck et al., 2019). In cereal crops, which do not have nitrogen‐fixing nodule symbiosis, this cannot be uncoupled by modifying a single gene. Increasing the abundance of another gene, OsGRF4, in the sd1‐containing semi‐dwarf variety is needed to enhance nitrogen use efficiency without loss of dwarfism (Li et al., 2018a). In this study, beneficial semi‐dwarfism with improved plant architecture, yield, seed oil content, and nitrogen fixation are achieved simultaneously in soybean, thus providing a promising strategy for a sustainable green revolution in soybean. Although accumulated DELLA in GmGID1‐2 ^KO lines might inhibit the absorption of inorganic nitrogen fertilizers as seen in rice (Li et al., 2018a), nitrogen fixation provides more than 70% (Li et al., 2023) of the necessary nitrogen (Herridge et al., 2008) in soybean, thus the enhanced nitrogen fixation would likely override the disadvantage of low uptake of inorganic nitrogen fertilizer in GmGID1‐2 ^KO lines.

The GA signaling pathway plays roles in regulation of both plant height (Wang et al., 2017) and nodule symbiosis (Hayashi et al., 2014). DELLA is the central regulator of GA signaling, which represses plant growth (Xu et al., 2014), while positively regulating the rhizobia infection and nodule symbiosis (Fonouni‐Farde et al., 2016; Jin et al., 2016; Mcadam et al., 2018; Velandia et al., 2024). However, if DELLA is overexpressed, excessive DELLA proteins will seriously inhibit plant growth (Ueguchi‐Tanaka et al., 2005; Griffiths et al., 2006). Gibberellin promotes nodule organogenesis, but inhibits the infection of rhizobia by degrading DELLA proteins (Mcadam et al., 2018; Velandia et al., 2024). The GA biosynthesis mutants showed suppressed nodule formation, while application of GA to the roots restored the number of nodules of a GA‐deficient mutant, indicating GA is required for nodulation (Ferguson et al., 2005). GIDs bind to and activate the degradation of DELLA proteins (Mcginnis et al., 2003; Sasaki et al., 2003; Ueguchi‐Tanaka et al., 2007). Based on above knowledge, if the GA biosynthesis genes are knocked out, low GA level would lead to inhibited nodule organogenesis (Chu et al., 2022), and reduced plant height due to DELLA accumulation (Sasaki et al., 2002; Spielmeyer et al., 2002). Therefore, semi‐dwarfism and improved root nodule symbiosis could be simultaneously achieved possibly by either up‐regulation of DELLA (not overexpression), or KO one of the GIDs homologs because the triple mutant with all GIDs absent showed severely reduced growth (Griffiths et al., 2006). In this study, we generated GmGID1‐2 ^KO lines, which only KO one GID gene (GmGID1‐2), but still have four other GID homologs in soybean. This resulted in suitable increased levels of DELLA in shoot apical meristem (SAM) and roots (Figures 1N, S16), therefore leading to reduced plant height and improved nitrogen fixation (Figures 2, 4).

It has been found that GID1 interacts with other proteins to regulate the expression of downstream genes, such as DELLA protein in the GA signaling pathway to regulate plant height and tillering in rice (Ueguchi‐Tanaka et al., 2005), AP2 transcription factor in regulation of tiller number in rice (Wu et al., 2020), and CRY1 in response to blue light in Arabidopsis (Xu et al., 2021b; Yan et al., 2021; Zhong et al., 2021). To investigate the possible regulatory network mediated by GmGID1‐2, we performed RNA‐seq and RT‐qPCR analyses, and found that genes involved in branching, lipid metabolism, and nodule symbiosis were differentially expressed between GmGID1‐2 ^KO lines and Williams 82 (Figures 3D–F, J, K, 6F). Five genes, including GmAP2b, GmZF351a, GmZF392, GmNAC181, and GmENOD93, showed most fold‐changes among the up‐regulated genes when KO GmGID1‐2. Luciferase reporter assays confirmed the transcriptional repression of these five genes by GmGID1‐2 (Figure S17). To test whether GmGID1‐2 functions through DELLA protein, we analyzed the effect of DELLA on the transcriptional activity of the above five genes. The results showed that DELLA only enhanced the transcriptional activity of GmNAC181 but not the four other genes (Figure S18), indicating GmGID1‐2 mediated regulation of plant architecture, seed oil content and nitrogen fixation is likely through interacting with other proteins or transcription factors in addition to its interaction with DELLA (Figure S19). The previous researches in rice showed that the AP2 transcription factor NGR5 is a target of GA–GID1‐mediated proteasomal destruction (Wu et al., 2020), whereas AP2 represses the expression of branching inhibitory genes such as D14 (encoding the receptor of SL) (Yao et al., 2016; Liu et al., 2022), and D14 further ubiquitinates D53, a repressor of SL signaling (Jiang et al., 2013; Zhou et al., 2013), leading to increased branching (Figure S20). In this study, KO of GmGID1‐2 leads to up‐regulated GmAP2 and GmD53, down‐regulated GmD14, and increased number of branches, which is consistent with the GA–SL signaling pathways involved in the regulation of branching in rice (Figure S20). Further experiments should be carried out to elucidate the regulatory network mediated by GmGID1‐2.

GmGID1‐2 ^KO lines have increased oil content but reduced protein content in seeds compared with Williams 82, which is consistent with the oil content usually being negatively correlated with protein content in soybean seeds (Patil et al., 2017; Zhong et al., 2024). This negative correlation is likely due to substrate competition in the synthesis of oil and protein (Figure S21A). The phosphoenolpyruvate in the glycolytic pathway serves as a substrate to form either pyruvate through pyruvate kinase (PK) for oil biosynthesis, or oxaloacetic acid through phosphoenolpyruvate carboxylase (PEPC) for protein biosynthesis (Sugimoto et al., 1989; Deng et al., 2014; Zhao et al., 2018). We therefore examined the activities of PK and PEPC in soybean seeds at 30 DAF, and the results confirmed that the activities of PK for oil biosynthesis in the seeds of GmGID1‐2 ^KO lines are higher than that of Williams 82 (Figure S21B), while the activities of PEPC for protein biosynthesis showed the opposite (Figure S21C), which could result in increased oil content but reduced protein content in the seeds of GmGID1‐2 ^KO lines. It is possible that the enhanced pyruvate pathway to increase oil content overrides the effect of improved nitrogen fixation which is usually associated with improved protein content (Zhong et al., 2024), therefore the seed protein content is reduced even though the nitrogen is increased in the shoots of GmGID1‐2 ^KO lines. In addition, it has been reported that after nitrogen fertilizer treatment, the protein content decreases while the oil content increases in soybean seeds (Purcell et al., 2004), which supports our findings.

In soybean, genes and microRNA (miRNA) involved in GA biosynthesis and catabolism, including GmGA3ox (G. max GA 3 β‐hydroxylase, a GA synthase enzyme), GmGA2ox (G. max GA 2‐oxidase, a GA catabolic enzyme), GmILPA1 (G. max Increased Leaf Petiole Angle1, a subunit of the E3 ubiquitin ligase anaphase promoting complex/cyclosome), GmSPA3a (G. max Suppressor of PHYA 105 3a underlying the reduced internode 1 mutant), and miR166 (G. max microRNA 166), have been found to regulate soybean plant height, branch number, internode length, seed size, and yield (Hu et al., 2022; Zhao et al., 2022; Li et al., 2023; Sun et al., 2023). Our results not only confirmed that modulation of the GA signaling pathway can regulate plant height, number of branches, nodes, pods and seeds, seed size, and yield in soybean, but also showed that GmGID1‐2 KO significantly improved nitrogen fixation, seed oil content, stem diameter and strength. A lower center of gravity (reduced plant height) and thicker stems are important for achieving crop lodging resistance, which is favorable to high‐density planting to achieve high yield per unit area, yet increased number of branches might need more growth space to reach maximum yield potential. To find out which planting density is favorable for GmGID1‐2 ^KO lines, we conducted field trials in 2024, under two different planting densities, 200,000 plants/ha and 300,000 plants/ha. We found that although the number of branches, nodes on the main stem, pods and seeds per plant of the GmGID1‐2 ^KO lines still increased compared with Williams 82 (Figure S22A–E), the increased percentages under 300,000 plants/ha (Figure S22) are less than those under 200,000 plants/ha (Figure 2): an increase of 31.03% versus 60.00% for branch number per plant, 4.00% versus 9.18% for node number on the main stem, 10.99% versus 27.60% for pod number per plant, and 13.16% versus 31.11% for seed number per plant. Again, the 100‐seed weight of the GmGID1‐2 ^KO lines was significantly lower than that of Williams 82 (Figure S22F), leading to no significant difference in the seed weight per plant and yield (Figure S22G, H). The above results indicate that the GmGID1‐2 ^KO soybean lines with lower plant height but more branches and pods are favorable to a normal planting density.

Because GmGID1‐2 encodes a GA receptor (Ueguchi‐Tanaka et al., 2005), we also investigated the responses of soybean plants to exogenous GA₃ application. The plant height of Williams 82 increased significantly after GA₃ application, and GmGID1‐2 ^KO lines had similar responses to exogenous GA₃ application as Williams 82, while GmGID1‐2 ^OE lines did not show significant changes in plant height after GA₃ application (Figure S23). Gibberellin‐dependent degradation of DELLA is the key in plant response to GA, and DELLA is epistatic to GID1 in the GA signaling pathway (Ueguchi‐Tanaka et al., 2005; Griffiths et al., 2006). In GmGID1‐2 ^OE lines with very high expression levels of GmGID1‐2 (Figure S3A), DELLA protein abundance was very low (Figure 1N), and overexpressed GID1‐2 protein could interact with DELLA to deactivate DELLA function (Ariizumi et al., 2008); therefore, not much DELLA was left in GmGID1‐2 ^OE lines to be degraded after GA₃ application, resulting in no significant changes in plant height after exogenous GA₃ application. Williams 82 and GmGID1‐2 ^KO lines had more DELLA protein accumulation than GmGID1‐2 ^OE lines (Figure 1N); therefore, plant growth inhibition was released after GA₃ application through GA‐mediated degradation of DELLA, leading to increased plant height. Although GmGID1‐2 was knocked out in GmGID1‐2 ^KO lines, there were still four other GmGID1 homologs in soybean, which also respond to GA (Figure S24). Therefore, the GmGID1‐2 ^KO lines (not completely loss of GA signaling function) can still respond to GA₃ due to the existence of other GmGID1 homologs in soybean. Unlike the mutants with all GIDs absent, which have no functional GA receptors, and therefore have a GA‐insensitive phenotype, such as the gid1‐1 mutant in rice, which only has one GID1 protein (Ueguchi‐Tanaka et al., 2005), and the gid1a‐1 gid1b‐1 gid1c‐1 triple mutant in Arabidopsis, which has three GID1 proteins (Griffiths et al., 2006). The expression levels of the other four GmGID1s were lower in the GmGID1‐2 ^KO lines but higher in the GmGID1‐2 ^OE lines (Figure S25), which is likely due to the negative feedback regulation (Figure S6): when more DELLA accumulates in GmGID1‐2 ^KO lines (Figures 1N, S16), the content of active GA increases (Figure S6) and the expression of other GID1s decreases (Figure S25), which is consistent with the down‐regulation of GID1s upon GA treatment of soybean (Figure S24) and the previous reports in Arabidopsis (Griffiths et al., 2006; Zentella et al., 2007).

Although previous studies in Arabidopsis, litchi, peach, rice, tomato, and wheat found that GID1 genes modulate plant height, branch number, male fertility, flower and seed development (Ueguchi‐Tanaka et al., 2005; Griffiths et al., 2006; Wu et al., 2011; Cheng et al., 2019; Illouz‐Eliaz et al., 2019; Yan et al., 2021; Zhang et al., 2024b), the function of GID1 has not been reported in soybean. The phylogenetic tree of GID1 proteins from the major legumes and other plants shows that the GID1s from soybean are most closely related to GID1s from the leguminous crop Phaseolus vulgaris, but distant to the GID1s from monocotyledonous crops such as rice, maize and wheat (Figure S26), implying the GID1s in other legumes might have similar functions as GmGID1 in soybean. We only found one report on the natural variation in GID1 gene, which shows that a nonsynonymous single nucleotide mutation in GID1c resulted in dwarfism and increased number of branches in peach (Cheng et al., 2019). We explored the natural variations in GmGID1‐2 among soybean germplasm, and identified a SNP at Gm03:36395036 in the GmGID1‐2 promoter region that was significantly associated with GmGID1‐2 transcript abundance and plant height (Figure 7A–C). The SoyGVD database showed that soybean accessions carrying the Pro‐C allele of GmGID1‐2 had a lower plant height but higher oil content and yield than those with Pro‐T allele (Figure S15). Further analyses showed that the frequency of Pro‐C allele of GmGID1‐2 in the released cultivars is higher than that in soybean landraces (Figure S27), suggesting the frequency of superior allele Pro‐C increased during soybean improvement. We also noticed that the frequencies of this superior allele differ among the soybean ecoregions in China, where ecoregions of II and III have more than 90% of Pro‐C allele, which is higher than the other ecoregions (Figure 7H), indicating the yield of more cultivars in the ecoregions I, IV, V, and VI could be further improved by introducing this superior allele. Therefore, we designed a derived cleaved amplified polymorphic sequence (dCAPS) marker for Gm03:36395036_C/T, which was verified in 20 extreme soybean accessions with contrasting phenotypes (Figure S28). This marker can be used for marker‐assisted selection in soybean breeding programs. In addition, yield can be further improved by GmGID1‐2 KO in soybean varieties with a natural superior GmGID1‐2 allele (Table 1). The +1 bp and −11 bp InDels could be used as markers to assist the selection of the new edited GmGID1‐2 allele. In summary, we showed that improved plant architecture, higher seed oil content and yield, and enhanced nitrogen fixation can be achieved in soybean by genetically editing a single gene, GmGID1‐2 (Figure 4; Table 1). This provides a promising strategy to breed high‐yield and high nitrogen‐fixation varieties, moving soybean further toward a sustainable green revolution.

MATERIALS AND METHODS

Plant materials and growth conditions

All soybean (Glycine max) accessions used in this study were provided by the National Center for Soybean Improvement at Nanjing Agricultural University (NJAU, Nanjing, China). Soybean accession Williams 82 was used for genetic transformation. The growth conditions in the greenhouse at NJAU were as follows: temperature of 28°C/24°C (day/night), photoperiod of 14/10 h (day/night) and about 70% humidity. A randomized complete block design was used for all phenotypic evaluations. In the field, soybean plants were grown in rows, and each row was 2 m long with a 0.5 m space between rows. The field trials for transgenic soybean lines and Williams 82 were conducted at the experimental station of the Anhui Academy of Agricultural Sciences in Hefei, China (latitude 31°79′ N; longitude 117°31′ E), in 2020, 2021, 2023, and 2024. In the years 2020 and 2021, there were four replications and two rows for each line per repeat, with 20 plants per row. In the years 2023 and 2024, yield trials were conducted in plots (2 × 3 m plot with six rows for each line, three replications per year). In 2023, there were 20 plants per row. In 2024, there were two planting densities with 20 and 30 plants per row, respectively. Extra rows of Williams 82 were always grown on the border as protection rows. A total of 264 soybean accessions were grown at the experimental stations of NJAU in Dangtu, China (latitude 31°57′ N; longitude 118°50′ E), and Xuzhou Academy of Agricultural Sciences in Xuzhou, China (latitude 34°29′ N; longitude 117°19′ E), in 2021. There were three replications at each location and one row for each line per repeat, with 20 plants per row. All field experiments were conducted in the summer planting season.

Phenotypic evaluation of agronomic traits

Important agronomic traits, including plant height, stem diameter, the number of branches, nodes, pods, and seeds per plant, seed size, seed weight per plant, and seed oil and protein content, were measured in both greenhouse experiments and field trials. Stem strength and yield per plot were measured in field trials. For greenhouse experiments, at least three plants for each soybean line were used for phenotypic evaluation. For field experiments, to obtain the average value of agronomic traits per plant, at least three plants in the middle of each row were randomly selected for the measurement. After maturity, the number of branches, nodes, pods, and seeds per plant were counted. Plant height was measured using a ruler. The stem diameter and seed size were measured using a vernier caliper. The seed weight was measured using an electronic scale. For yield trials, within each replicate, soybean plants within each 2 × 3 m plot were harvested and dried, and the seeds were weighed to determine the yield for each line. To measure the seed oil and protein content, about 15–20 g of fully matured seeds per plant were randomly selected after drying (water content < 8.5%), and then near‐infrared analysis (NIR) with the calibration model was performed using Grain Analyzer (Hacisalihoglu et al., 2016) InfratecTM1241 (Foss, Hillerød, Denmark). The bending strength of soybean stems was measured at R6 stage (full seed) in the field using a portable plant stalk‐strength meter YYD‐1 (Zhejiang Top Cloud—Agriculture Technology, Hangzhou, China). The strength was recorded when soybean plants were tilted 45° at a height of 20 cm from the ground. Flowering time was recorded as the number of days from emergence to the appearance of the first open flower of soybean plants grown in a greenhouse (Lu et al., 2017). Thirty seeds (six seeds per replicate with five replicates) per line were sown in sterile moist sand. After 3 d, the seed germination rate was calculated as the number of germinated seeds divided by the total number of seeds sown, as a percentage (%).

Phylogenetic analysis

The gene IDs of GID1s from Arabidopsis thaliana, Lotus japonicus, Medicago truncatula, Phaseolus vulgaris, Pisum sativum, Oryza sativa, Triticum aestivum, Zea mays, and Glycine max were obtained from previous studies (Ueguchi‐Tanaka et al., 2005; Griffiths et al., 2006; Yoshida et al., 2018), and the full sequences of GID1 proteins were downloaded from Phytozome (https://phytozome.jgi.doe.gov/pz/portal.html). An unrooted phylogenetic tree was constructed using MEGA 5.0 (Tamura et al., 2011) based on the neighbor‐joining method with 1,000 bootstraps.

Tissue expression pattern analysis

RNA‐seq data from soybean different tissues were downloaded from Phytozome (https://phytozome.jgi.doe.gov/pz/portal.html) and Soybase (https://www.soybase.org). The raw data of fragments per kilobase of transcript per million mapped reads (FPKM) was transformed to Log₂(FPKM + 1) and then displayed as heatmaps using OriginPro version 2024b (https://www.originlab.com).

RNA isolation and RT‐qPCR

Total RNA was isolated using an RNA Pure Plant Kit (TIANGEN Biotech, Beijing, China). The complementary DNA (cDNA) was synthesized using a HiScript II Q RT SuperMix (VAZYME, Nanjing, China), with the following program: 42°C for 2 min, 50°C for 15 min, and 85°C for 5 s. Primers (Table S4) were designed based on the CDS of genes (Table S5) using Primer‐Blast from the National Center for Biotechnology Information (NCBI: https://www.ncbi.nlm.nih.gov/tools/primer-blast/). GmUKN1 was used as the reference gene for RT‐qPCR, which was performed on a LightCycler480 System (Roche, Diagnostics, Switzerland) with ChamQ Universal SYBR qPCR master mix (VAZYME, Nanjing, China), using the following procedure: 95°C for 5 min, followed by 40 cycles at 95°C for 10 s, and 60°C for 10 s. Relative gene expression was analyzed using 2^−ΔCt and 2^−ΔΔCt (Livak and Schmittgen, 2001). Three biological replications were performed for each gene.

Subcellular localization

The construct of 35S:GmGID1‐2‐GFP in the backbone vector pBinGFP4 (Liu et al., 2014; Ma et al., 2017) was used for subcellular localization, and H2B‐mChrry was used as a nucleus marker (Howe et al., 2012). These plasmids were transformed into Agrobacterium tumefaciens strain EHA105 (Hood et al., 1993) and co‐transformed into tobacco (Nicotiana benthamiana) leaves (Rolland, 2018). EHA105 and tobacco were obtained from the National Center for Soybean Improvement at NJAU. Fluorescence was observed with a confocal laser scanning microscope (Zeiss LSM780, Oberkochen, Germany).

Soybean transformation

The 35S:GmGID1‐2 construct in the backbone vector pCAMBIA3301 was used to overexpress GmGID1‐2 (Table S5). GmGID1‐2 was edited using CRISPR/Cas9 technology (Jinek et al., 2012). The single guide RNA (sgRNA) was designed using CRISPR‐P2.0 (http://cbi.hzau.edu.cn/CRISPR2/) (Lei et al., 2014) and cloned into the pGES201 vector (Bai et al., 2020). SWISS‐MODEL (https://swissmodel.expasy.org) was used to predict the 3D structures of the proteins. These plasmids were transferred into A. tumefaciens strain EHA105 and used for Agrobacterium‐mediated transformation of soybean (Li et al., 2017b).

Paraffin sectioning

Fresh hypocotyl tissues of soybean plants were immersed in fixative for 24 h, dehydrated, and embedded in paraffin. Sections with a thickness of 4 μm were cut with a microtome, dewaxed, and stained with plant safranine for 2 h. After decolorization, they were stained with plant fast green for 6–20 s and observed using a light microscope (Nikon E100, Tokyo, Japan).

Hormone treatment

To investigate soybean response to GA, soybean plants at V1 stage (when the first trifoliolate leaf was fully unfolded) were sprayed with enough 100 μmol/L GA₃ solution (+GA) or ddH₂O (−GA) to ensure that each leaf was wet according to a previously published method (Ueguchi‐Tanaka et al., 2005; Griffiths et al., 2006). Nine days after spraying, the plant height was measured using a ruler. The SAM was collected at 0 and 0.5 h after spraying to analyze the expression levels of the GmGID1 genes.

Hormone quantification

The SAM of soybean plants at V1 stage (including Williams 82, equally mixed sample of three GmGID1‐2 ^OE lines, and equally mixed sample of two GmGID1‐2 ^KO lines) were used for GA quantification based on a previously described method (Balcke et al., 2012). About 1 g of fresh SAM samples were ground with liquid nitrogen and added to 10 mL of extraction solution (isopropanol/hydrochloric acid mixture) buffer. Active GA concentrations were measured using high‐performance liquid chromatography–tandem mass spectrometry (HPLC–MS/MS; AGLIENT1290, Santa Clara, CA, USA; SCIEX‐6500Qtrap (MSMS), Boston, MA, USA).

Bradyrhizobium japonicum inoculation, ureide content, nitrogenase activity, and nitrogen content measurements

To investigate the effect of GmGID1‐2 on soybean nitrogen fixation, soybean lines were planted in pots with sterilized vermiculite and grown to V1 stage in the greenhouse. The plants were then inoculated with B. japonicum strain USDA110 (obtained from Deyue Yu at NJAU, optical density at 600 nm (OD₆₀₀) = 0.08). The plants were supplied with 0.58 mmol/L nitrogen solution (Wang et al., 2014, 2021), containing 94.5 mg/L Ca(NO₃)₂, 435 mg/L K₂SO₄, 444 mg/L CaCl₂, 136 mg/L KH₂PO₄, 241 mg/L MgSO₄, 36.7 mg/L Fe‐ethylenediaminetetraacetic acid (EDTA) (Na), 0.83 mg/L KI, 6.2 mg/L H₃BO₃, 16.9 mg/L MnSO₄·H₂O, 8.6 mg/L ZnSO₄·7H₂O, 0.25 mg/L NaMoO₄·2H₂O, 0.025 mg/L CuSO₄·5H₂O, and 0.025 mg/L CoCl₂·6H₂O. At 3, 6, 10, 14 DAI, the formation and development of nodules were observed. At 28 DAI, soybean roots, nodules, leaves, and stems were harvested to determine the fresh and dry weight, number of nodules, nitrogenase activity in nodules (Chen et al., 2022a), ureide content in stems (Shakoor et al., 2023; Zhang et al., 2024a), and total nitrogen content in shoots (Li et al., 2018b; Yang et al., 2022).

Infection foci and nodule primordia observations

To observe the infection thread, 2‐cm root segments below the root‐hypocotyl junction were cut and harvested at 6 DAI and then rinsed briefly in sterile phosphate‐buffered saline to remove vermiculite/perlite particles. The roots were then fixed with ethanol : glacial acetic acid (3:1) for 2 h and washed three times with deionized water. Next, the roots were stained with 0.01% methylene blue for 15 min and washed three times with deionized water, as previously described (Wang et al., 2014). The stained soybean root segments were observed with a super depth of a field microscope (Lecia DVM6, Wetzlar, Germany).

Measurement of PEPC and PK activities

The activities of PEPC and PK in soybean seeds at 30 DAF were determined according to the previously published methods (Zhang et al., 2008; Lepper et al., 2010).

Yeast two‐hybrid assays

The Matchmaker^TM Gold Yeast Two‐Hybrid System (Clontech, Shanghai, China) was used for the Y2H assays (Wang et al., 2015). The full‐length CDSs of GmGID1‐2 and GmDELLAs (GmDELLA1, GmDELLA2, GmDELLA3, GmDELLA4, GmDELLA5, GmDELLA6, GmDELLA7, and GmDELLA8, Table S5) were cloned into pGBKT7 (BD:GmGID1‐2) and pGADT7 (AD:DELLAs) vectors. Yeast two‐hybrid assays were performed on double dropout medium (DDO/X‐α‐Gal without Leu/Trp) and quadruple dropout medium (QDO/X‐α‐Gal without Ade/Leu/Trp/His) in the absence or presence of 100 μm GA₃.

Luciferase complementation imaging (LCI) assays

For LCI assays (Liu et al., 2018), full‐length CDSs of GmGID1‐2, GmDELLA1 and GmDELLA3 (which showed strong interaction with GmGID1‐2 in Y2H) were fused to the pCAMBIA1300‐nLUC and pCAMBIA1300‐cLUC vectors (Chen et al., 2008). These plasmids were transformed into A. tumefaciens strain GV3101 with pSoup‐19 (purchased from ZOMANBIO, Beijing, China) (Koncz, 1986) and co‐transformed into N. benthamiana leaves. Empty plasmids were used as negative controls. Fluorescence was observed using a luminescent imaging system (Tianon 5200, Shanghai, China).

Western blotting and Co‐IP assays

Soybean SAM and root tissues at V1 stage were used for western blotting (WB) assays (Lu et al., 2017). The total protein was extracted using extraction buffer: 50 mmol/L Tris‐HCl (pH 7.5), 0.5 mmol/L EDTA, 150 mmol/L NaCl, 0.1% NP‐40, 1 mmol/L phenylmethylsulfonyl fluoride, and 1× complete protease inhibitor cocktail (Roche, Basel, Switzerland). The protein concentration was determined using a bicinchoninic acid protein concentration assay kit (Biosharp, Beijing, China). The protein was subjected to sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE), transferred to a polyvinylidene fluoride (PVDF) membrane, and detected by treating the membranes with DELLA antibodies (1:2,000, Abmart, Shanghai, China, ZW024893S), Actin antibodies (1:2,000, Abmart, Shanghai, China, M001814), and goat anti‐rabbit mouse immunoglobulin G – horseradish peroxidase (IgG‐HRP) (1:5,000, Abmart, Shanghai, China, M21001). Photographs were taken using a luminescent imaging system (Tianon 5200, Shanghai, China). For image quantification of the DELLA protein content, WB images were analyzed using ImageJ software (https://imagej.nih.gov/ij/).

Co‐immunoprecipitation analysis (Yue et al., 2021) was performed using N. benthamiana. The full‐length CDSs of DELLA1 and DELLA3 were cloned into pTF101‐35S‐HA vectors, and the GmGID1‐2 coding sequence was introduced into the pTF101‐35S‐Flag vector. These plasmids were transformed into A. tumefaciens strain EHA105 and co‐transformed into N. benthamiana leaves. The total protein was extracted from N. benthamiana leaves after incubation in extraction buffer for 48 h. The protein lysis product was incubated with Protein A/G beads for 1–3 h, and the beads were washed three times with a wash buffer that consisted of 50 mmol/L Tris‐HCl (pH 7.5), 200 mmol/L NaCl, 10% glycerol, 0.1% NP‐40, 1 mmol/L EDTA, and 1× complete protease inhibitor cocktail. The immunoprecipitate was separated using SDS–PAGE and transferred to a PVDF membrane. The proteins were detected by treating the membranes with HA‐Tag(26D11) mAb (1:5,000, Abmart, Shanghai, China, M20003), DYKDDDDK‐Tag(3B9) mAb (1:5,000, Abmart, Shanghai, China, M20008), and goat anti‐rabbit mouse IgG‐HRP (1:5,000, Abmart, Shanghai, China, M21001). Photographs were taken using a luminescent imaging system (Tianon 5200, Shanghai, China).

RNA‐seq

The lateral buds at 20 d after germination, developing seeds at 30 DAF, and roots and nodules at 28 DAI were used for RNA‐seq. Total RNAs were extracted using an RNA Extract Kit (TIANGEN Biotech, Beijing, China). The sequencing libraries were generated using Hieff NGS Ultima Dual‐Mode mRNA Library Prep Kit for Illumina (Yeasen Biotechnology, Shanghai, China) following the manufacturer's recommendations and sequenced on an Illumina NovaSeq platform (Illumina, San Diego, CA, USA) at Biomarker Technologies, Beijing, China. After removing the low‐quality reads and reads containing adapters or ambiguous bases (Ns), the remaining clean reads were mapped to the soybean Williams 82 reference genome (Wm82.a2.v1) (https://phytozome-next.jgi.doe.gov) with HISAT (Kim et al., 2015). StringTie (Pertea et al., 2015) was then used to assemble the reads, and the transcriptome was reconstructed for subsequent analysis. Differential expression analysis of the two groups was performed using DESeq2 (Robinson et al., 2010; Love et al., 2014). The P‐values were adjusted using Benjamini and Hochberg's approach for controlling the false discovery rate. Genes with an adjusted P‐value < 0.01 and fold change ≥ 2 were considered differentially expressed. Gene Ontology and KEGG pathway enrichment analyses were performed using the previously described methods (Ashburner et al., 2000).

Association analysis

Regional association analysis (Sosso et al., 2015) was performed using TASSEL 5.0 software (Bradbury et al., 2007). The average values of plant height of 264 soybean accessions across two locations (Dangtu and Xuzhou) and 17 SNPs with MAF ≥ 0.05 within the 2,000 bp promoter region of GmGID1‐2 were used (Table S3). In this study, the threshold for a significant association was determined using a previously published method (Yang et al., 2014), in which P < 1/n (n is the number of markers).

Cross‐validation of association between sequence variation in GmGID1‐2 and important soybean traits

The associations between the SNP at Gm03:36395036 in the promoter region of GmGID1‐2 and important soybean traits (including plant height, seed oil content, seed protein content, and yield) were verified using the soybean germplasm dataset in the SoyGVD (Yang et al., 2023) database (https://yanglab.hzau.edu.cn/SoyGVD/#/).

Derived cleaved amplified polymorphic sequence marker development

A specific primer pair of dCAPS marker were designed for the SNP at Gm03:36395036 in GmGID1‐2 promoter region using dCAPS Finder 2.0 (Neff et al., 2002) (http://helix.wustl.edu/dcaps/). After amplification with a specific primer pair, the PCR products were digested with KpnI restriction enzyme (NEB, Beverly, MA, USA), which could be cut when using the DNA template of Pro‐C type but could not be cut when using the DNA template of Pro‐T type, leading to a difference of 28‐bp in length. Electrophoresis was conducted with 2% agarose gel and photo‐documented using a Gel Imaging Documentation System (Bio‐Rad, Hercules, CA, USA).

Luciferase reporter assays

Two types of GmGID1‐2 promoters (Pro‐C and Pro‐T), and promoters of GmAP2b, GmZF351a, GmZF392, GmENOD93, and GmNAC181, were amplified using genomic DNA as a template and cloned into the pGreenII0800‐LUC vector (Li et al., 2014). Pro‐C:LUC, Pro‐T:LUC, Pro‐GmAP2b:LUC, Pro‐GmZF351a:LUC, Pro‐GmZF392:LUC, Pro‐GmENOD93:LUC and Pro‐GmNAC181:LUC were transformed into A. tumefaciens strain GV3101 with pSoup‐19 (Koncz, 1986). A. tumefaciens strain GV3101 with Pro‐C:LUC or Pro‐T:LUC was transformed into N. benthamiana leaves (Jin et al., 2021). A. tumefaciens strain GV3101 (containing Pro‐GmAP2b:LUC, Pro‐GmZF351a:LUC, Pro‐GmZF392:LUC, Pro‐GmENOD93:LUC, or Pro‐GmNAC181:LUC), was co‐transformed with A. tumefaciens strain EHA105 (containing 35S:GmGID1‐2, 35S:GmDELLA1, or 35S:GmDELLA3) into N. benthamiana leaves. Luciferase activity was observed using a luminescent imaging system (Tianon 5200, China).

Statistical analyses

Differences between two groups were analyzed using the Student's t‐test (Phillips, 1990) and Wilcoxon test (Bauer, 1972) in R software (https://www.r-project.org). Duncan's multiple range test was used to compare the differences among multiple groups in SPSS v.26 (IBM, https://www.ibm.com/products/spss-statistics). The figures were generated using OriginPro version 2024b (https://www.originlab.com).

CONFLICTS OF INTEREST

The authors declare no conflicts of interest.

AUTHOR CONTRIBUTIONS

Y.L. and J.T. conceived and designed the experiments. J.T., S.Y., S.L. and K.Z. performed the experiments. J.T., S.Y., X.Y., K.Z. and T.L. conducted field trials. J.T., X.Y. and T.J. analyzed the data. J.T., S.Y. and X.Y. generated the pictures. Y.L., S.Z. and J.G. contributed key materials and data. Y.L. supervised the study. J.T. and Y.L. wrote and revised the manuscript. All authors read and approved the manuscript.

Supporting information

Additional Supporting Information may be found online in the supporting information tab for this article: http://onlinelibrary.wiley.com/doi/10.1111/jipb.70026/suppinfo

Figure S1. Expression of five GmGID1 genes in soybean different tissues

Figure S2. Subcellular localization of GmGID1‐2

Figure S3. Identification of soybean transgenic lines

Figure S4. Analyses of the interactions between GmGID1‐2 and soybean DELLA proteins

Figure S5. Digital analysis of DELLA protein bands in the shoot apical meristem (SAM) of soybean

Figure S6. GmGID1‐2 knockout increased the total active gibberellin (GA) content

Figure S7. GmGID1‐2 knockout reduced plant height, improved plant architecture and yield‐related traits in greenhouse

Figure S8. GmGID1‐2 overexpression increased the seed size, and GmGID1‐2 knockout reduced the seed size of soybean

Figure S9. GmGID1‐2 knockout increased the maximum number of pods per clump and average seed number per pod

Figure S10. GmGID1‐2 knockout increased the oil content and reduced protein content in soybean seeds

Figure S11. GmGID1‐2 overexpression reduced the oil content and increased protein content in soybean seeds

Figure S12. Effect of GmGID1‐2 overexpression on the plant height and yield‐related traits in soybean

Figure S13. GmGID1‐2 did not affect soybean seed germination

Figure S14. The effect of GmGID1‐2 on the flowering time of soybean

Figure S15. Association of the SNP at Gm03:36395036 in GmGID1‐2 with plant height, oil content, protein content, and yield

Figure S16. Analysis of DELLA protein abundance in soybean roots

Figure S17. GmGID1‐2 inhibited the promoter activities of GmAP2b, GmZF351a, GmZF392, GmNAC181, and GmENOD93

Figure S18. DELLA did not affect the promoter activities of GmAP2b, GmZF351a, GmZF392 and GmENOD93 but enhanced GmNAC181 promoter activity

Figure S19. A proposed model to illustrate how GmGID1‐2 regulates oil content, branch number, nodule symbiosis and plant height in soybean

Figure S20. The known regulators of branching in GA–SL signaling pathways in rice

Figure S21. Enhanced activity of PK in oil biosynthesis pathway could lead to increased seed oil content of GmGID1‐2 knockout lines

Figure S22. Soybean phenotypes under the planting density of 300,000 plants/ha

Figure S23. Effect of exogenous GA₃ application on soybean plant height

Figure S24. Relative expression of GmGID1s in response to gibberellic acid (GA₃)

Figure S25. Effect of GmGID1‐2 on the expression of other GmGID1s in soybean

Figure S26. Phylogenetic analysis of GID1 proteins from soybean and other plants

Figure S27. Allele frequencies of the SNP at Gm03:36395036 in GmGID1‐2 among 264 soybean accessions

Figure S28. The dCAPS marker designed for the SNP at Gm03:36395036 in GmGID1‐2

Table S1. Natural variations in the coding sequence of GmGID1‐2 in 264 soybean accessions

Table S2. Natural variations in the coding sequence of GmGID1‐2 in SoyOmics database

Table S3. The 17 polymorphic single nucleotide polymorphisms (SNPs) in the GmGID1‐2 promoter region with MAF ≥ 0.05 in 264 soybean accessions

Table S4. Primers used in this study

Table S5. Gene IDs in this study

Biographies

Tang, J. , Yang, S. , Li, S. , Yue, X. , Jin, T. , Yang, X. , Zhang, K. , Yang, Q. , Liu, T. , Zhao, S., et al. (2026). Editing a gibberellin receptor gene improves yield and nitrogen fixation in soybean. J. Integr. Plant Biol. 68: 75–95.

Data availability statement

RNA‐seq data can be accessed through the NCBI under BioProject ID PRJNA1106984.