Progress in soybean functional genomics over the past decade.

Soybean is one of the most important oilseed and fodder crops. Benefiting from the efforts of soybean breeders and the development of breeding technology, large number of germplasm has been generated over the last 100 years. Nevertheless, soybean breeding needs to be accelerated to meet the needs of a growing world population, to promote sustainable agriculture and to address future environmental changes. The acceleration is highly reliant on the discoveries in gene functional studies. The release of the reference soybean genome in 2010 has significantly facilitated the advance in soybean functional genomics. Here, we review the research progress in soybean omics (genomics, transcriptomics, epigenomics and proteomics), germplasm development (germplasm resources and databases), gene discovery (genes that are responsible for important soybean traits including yield, flowering and maturity, seed quality, stress resistance, nodulation and domestication) and transformation technology during the past decade. At the end, we also briefly discuss current challenges and future directions.

Introduction

Cultivated soybean (Glycine max [L.] Merr.) was domesticated from wild soybean (G. soja Sieb. & Zucc.) in China approximately 5000 years ago, after which was spread worldwide (Carter et al., 2004; Wilson, 2008). Currently, soybean has become one of the most economically important oilseed and biodiesel crops and also serves as a main source of protein and oil for both human food and animal feed (Hartman et al., 2011). Early soybean breeding mainly relied on repeated selection of preferred seeds by farmers from cultivated population. Starting from the early 1900s, artificial hybridization was applied. The first modern soybean cultivar developed by hybridization was released in North American breeding programmes in 1940s (Rincker et al., 2014; Wolfgang and Charles, 2017). Afterwards, artificial hybridization started to be widely incorporated in soybean breeding (Anderson et al., 2019; Li et al., 2001). Artificial hybridization greatly expanded the genetic base of developed lines and significantly improved soybean adaptation and production (Anderson et al., 2019).

With the arising and developing of molecular biology technology, marker‐assisted selection (MAS) was used to speed up the breeding process, particularly in development of disease‐ and insect pest‐resistant cultivars (Li et al., 2020b). Using different kinds of genetic markers, linkage and physical maps were constructed (Chan et al., 2012; Cregan et al., 1999; Marek et al., 2001; Song et al., 2004). By integrating available genetic maps and physical maps, the Consensus Map 4.0 was built (Hyten et al., 2010a; Hyten et al., 2010b). Using these markers, large numbers of quantitative trait loci (QTLs) affecting related traits have been identified in soybean. However, the limited number of molecular markers and their uneven distribution limited the efficiency and accuracy of QTL positioning.

The reference genome of a cultivated accession (Williams 82) was released in 2010 (Schmutz et al., 2010), which brought about the era of soybean functional genomics (Chan et al., 2012; Li et al., 2017a; Wang and Tian, 2015; Xia et al., 2013). Benefit of having the reference genome, research publications on soybean have almost doubled compared with ten years ago, with higher ratios on stress, omics and nodulation (Figure 1). Here, we review the advances in soybean functional genomics and transformation technology during the past decade and discuss the challenges and prospects for future soybean functional genomic studies.

Figure 1

Statistics of publications with term ‘soybean’ from 2009 to 2020. (a) The number of publications with the term ‘soybean’ in each year. (b) The percentage of publications with the term ‘soybean’ and the indicated terms in different research fields.

Progress in omics

Advances in sequencing technologies, particularly in long‐read sequencing, have led to the production of improved assembly genomes (Burton et al., 2013; Yang and Huang, 2018). After the release of the reference genome based on Williams 82, Kim et al. (2010) assembled a genome sequence of a wild soybean (G. soja var. IT182932) using Illumina‐GA and GS‐FLX. Shen et al. de novo assembled a high‐quality genome for cultivar ‘Zhonghuang 13’ (Gmax_ZH13) using single‐molecule real‐time (SMRT) sequencing, optical mapping, chromosome conformation capture sequencing (Hi‐C) and next‐generation sequencing (HiSeq) (Shen et al., 2019; Shen et al., 2018a). Xie et al. (2019) then assembled a high‐quality reference genome for wild soybean W05 in 2019. Valliyodan et al. (2019) de novo assembled references for another two cultivars and one wild soybean using a combination of short‐ and long‐read technologies. In addition to the de novo assembly of reference genomes, pan‐genome also progressed significantly in soybean. In 2014, Li et al. reported the first soybean pan‐genome built by assembly of seven wild soybeans decoded using second‐generation sequencing technology (Li et al., 2014b). Recently, Liu et al. (2020f) constructed a soybean pan‐genome by de novo assembly of 26 representative wild and cultivated soybeans using long‐read sequencing. This assembly produced not only golden‐grade genomes for each accession, but also for the first time reported a graph‐based genome in plants, which provides a promising platform for future in‐depth soybean functional genomic studies (Liu and Tian, 2020; Lyu, 2020; Tao et al., 2020; Willson, 2020).

The reference genome sequences revealed that soybean had undergone at least two whole‐genome duplication (WGD) events within the last 60 million years (Schmutz et al., 2010). It was found that approximately 50% of the paralogs arising from the WGD had undergone subfunctionalization at the expression level in soybean, suggesting that the main consequence of WGD in soybean may be at the regulatory level (Roulin et al., 2013). The diversification of WGD genes was confirmed by the evolutionary and functional analyses several gene families (Li et al., 2016b; Wang et al., 2015d; Yang et al., 2016; Zhang et al., 2017). It was also found that, despite the WGD, the genes in networks with a high level of connectivity were more conserved than those with low connectivity (Kim et al., 2015; Severin et al., 2011). Moreover, WGD genes exhibited a decreased frequency of alternative splicing compared with non‐WGD genes, which was associated with their reduced intron length, exon number and transcriptional level (Shen et al., 2014). WGD also affected the evolution of transposable elements (Du et al., 2010b; Du et al., 2012; Tian et al., 2012).

Soybean population genetics has also greatly advanced in the past ten years due to the availability of a soybean reference genome and the development of next‐generation sequencing technology. The resequencing of dozens of wild and cultivated soybean accessions revealed the consequences of the artificial selection accompanying domestication and showed that genetic diversity was significantly decreased after domestication (Lam et al., 2010; Li et al., 2013b). Integration of selection sweep and genome‐wide association studies (GWAS) on domestication traits using the data from a resequencing of 302 wild soybeans, landraces and improved cultivars disclosed the genetic loci responsible for traits related to soybean domestication, improvement and local breeding (Zhou et al., 2015). Moreover, it was found that naturally occurring introgression was widespread and counteracted genetic bottlenecks during soybean domestication (Wang et al., 2019e). Hundreds of small RNAs had been identified in soybean (Arikit et al., 2014; Golicz et al., 2018; Lin et al., 2020a; Wong et al., 2011; Zhou et al., 2013), some of which showed tissue‐specific or time‐specific transcriptional patterns, indicating their biological relevance (Arikit et al., 2014; Lin et al., 2020a). Population genetic analyses suggested a co‐evolution of MIRNA and miRNA targets during soybean domestication (Liu et al., 2016a; Zhao et al., 2015a).

So far, over 6000 soybean accessions have been genotyped to investigate the footprints of breeding (Chung et al., 2014; Fang et al., 2017; Han et al., 2016; Kajiya‐Kanegae et al., 2021; Liu et al., 2020f; Qi et al., 2021; Qiu et al., 2014; dos Santos et al., 2016; Shimomura et al., 2015; Torkamaneh et al., 2018; Valliyodan et al., 2016; Zhou et al., 2015) (Table 1). The large amount of resequencing data has generated a great number of nucleotide polymorphism (SNP) markers, which greatly facilitate the haplotype map construction (Torkamaneh et al., 2021) and SNP Arrays development (Lee et al., 2015; Song et al., 2020; Wang et al., 2016a), and also increased the efficiency and accuracy of gene/QTL mapping (Baird et al., 2008; Bandillo et al., 2017; Lee et al., 2015; Patil et al., 2016).

Table 1

Whole‐genome sequencing in soybean.

Accession information	Method	Accession number	Reference
Williams 82 (cultivar)	De novo sequencing and assembly	GCA_000004515.3	Schmutz et al. (2010)
14 cultivars; 17 wilds	Re‐sequencing	SRA020131	Lam et al. (2010)
T182932 (wild)	De novo sequencing and assembly	SRA009252	Kim et al. (2010)
10 cultivars; 5 wilds	Re‐sequencing	ERP002622	Chung et al. (2014)
7 wilds	De novo sequencing and assembly	PRJNA195632	Li et al. (2014b)
9 semi‐wilds; Maliaodou (semi‐wild); Lanxi1(wild)	Re‐sequencing; De novo sequencing and assembly	PRJNA227063	Qiu et al. (2014)
W05 (wild)	De novo sequencing and assembly	GCA_000722935.2	Qi et al. (2014)
240 cultivars; 62 wilds	Re‐sequencing	SRP045129	Zhou et al. (2015)
Enrei (cultivar)	Reference‐based assembly	GCA_001269945.2	Shimomura et al. (2015)
28 Brazilian soybean	Re‐sequencing	PRJNA294227	dos Santos et al. (2016)
404 fully domesticated; 36 semi‐domesticated; 72 non‐domesticated	Re‐sequencing		Han et al. (2016)
7 wilds; 43 landraces; 56 cultivars	Re‐sequencing	SRP062245	Valliyodan et al. (2016)
291 landraces; 278 cultivars	Re‐sequencing	PRJCA000205	Fang et al. (2017)
102 cultivars	Re‐sequencing	SRP094720	Torkamaneh et al. (2018)
Zhonghuang 13 (cultivar)	De novo sequencing and assembly	CRA001007	Shen et al. (2018a)
Zhonghuang 13 (cultivar)	De novo sequencing and assembly	CRA001810	Shen et al. (2019)
W05 (wild)	De novo sequencing and assembly	SRP158454	Xie et al. (2019)
1 wild; 2 cultivars	De novo sequencing and assembly	GCA_002907465.1;GCA_002905335.1 & PRJNA48389	Valliyodan et al. (2019)
3 wilds; 9 landraces; 14 cultivars	De novo sequencing and assembly	PRJCA002030	Liu et al. (2020f)
177 landraces; 21 breeding lines	Re‐sequencing	PRJDB7281	Kajiya‐Kanegae et al. (2021)
134 cultivars	Re‐sequencing	SRP062560	Qi et al. (2021)

In addition to the progresses in genomics, studies on soybean transcriptomics, epigenetics and proteomics were also explored. Transcriptome of different tissues from different developmental stages illuminated gene expression profiling at a whole‐genome level (Severin et al., 2010; Wang et al., 2014b). Integration of the gene co‐expression network from RNA‐seq data of 1978 samples with previously reported QTLs identified a candidate gene that may control flowering time in soybean (Shen et al., 2018a). Meanwhile, large numbers of differentially expression genes (DEGs) that may be related to stresses were identified using the transcriptome data from plants subjected to stress treatments, such as drought, flooding, salt or heat (Chen et al., 2016; Liu et al., 2019; Wang et al., 2018a; Xu et al., 2018). These candidate genes provide clues for further functional study. Recently, 1298 publicly available soybean transcriptome datasets were combined to generate a comprehensive atlas of expression, which can be accessed through the website http://venanciogroup.uenf.br/cgi‐bin/gmax_atlas/index.cgi (Machado et al., 2020).

DNA methylation profiling analyses revealed that hypomethylation could affect the expression of neighbouring genes (Song et al., 2013b). Kim et al. (2015) found that CG body‐methylated genes were abundant in duplicated genes that exhibited higher expression level than single copy genes. Moreover, methylation changes proximal to the transposable elements (TEs) were associated with the divergence of expression profiles of duplicated genes (El Baidouri et al., 2018; Hossain et al., 2017). Shen et al. (2018b) investigated the variation of DNA methylation during soybean domestication and improvement and found that differentially methylated regions that are not associated with any genetic variation were enriched in carbohydrate metabolism pathways. It was found that DNA demethylation/methylation also plays critical roles in stress responses, such as continuous cropping (Liang et al., 2019), salinity stress (Song et al., 2012) and cyst nematode infection (Rambani et al., 2020).

Using proteomics approaches, numbers of genes response to various stresses were identified (Khan and Komatsu, 2016; Khatoon et al., 2012; Komatsu et al., 2010; Komatsu et al., 2017; Komatsu et al., 2011; Sobhanian et al., 2010; Wang and Komatsu, 2018). Similarly, the protein changes through different developmental stages were also investigated (Agrawal et al., 2008; Arai et al., 2008). Along with the research progresses in proteomics, several proteomics databases were provided. The Soybean Proteome Database (SPD) stores soybean proteomics data obtained from both gel‐based and gel‐free techniques (Komatsu et al., 2017; Ohyanagi et al., 2012). A user‐intuitive database (http://oilseedproteomics.missouri.edu) stores the expression profile data for proteomics research on soybean and other oilseeds plant (Agrawal et al., 2008).

Germplasm resources and databases

Soybean is rich in germplasm resources that carry large amounts of variations. Over the long history of soybean cultivation, more than 60 000 accessions have been developed (Carter et al., 2004; Wilson, 2008). In China, more than 40 000 soybean accessions were collected from all over the nation and stored in the Chinese Crop Germplasm Centre. In the United States, over 20 000 accessions were collected from around the world and stored at the U.S. Department of Agriculture (Song et al., 2015). Investigations of the population structure and genetic diversity of the core collections suggested that the accessions exhibit geographical patterns (Haupt and Schmid, 2020; Li et al., 2008).

Mutant populations are important in facilitating gene cloning and functional analysis. Through fast neutron, gamma radiation and ethylmethane sulphonate (EMS) mutagenesis, several soybean mutant populations had been constructed (Bolon et al., 2014; Espina et al., 2018; Li et al., 2017e; Tsuda et al., 2015). These populations enabled the identification of some genes responsible for important traits (Anderson et al., 2019). For example, using TILLING, Liu et al. (2012) identified a soybean cyst nematode‐related gene, GmSHMT, at the Rhg4 locus. Dobbels et al. (2017) identified an oil biosynthesis‐related gene (KASI) through forward screening of a soybean fast neutron mutant population. Reinprecht et al. (2009) found that mutations in two Fad3 genes lead to a low level of linolenic acid in the EMS mutant RG10.

Several soybean databases related to soybean genomics, transcriptomics, proteomics and germplasm analyses have been developed. These freely available databases include SoyBase, a database for genetics and genomics from USDA‐ARS (Brown et al., 2020; Grant et al., 2010); SoyTEdb, a database of transposable elements (Du et al., 2010a); SoyNet, a database for co‐functional networks (Kim et al., 2017); SoyProDB: a database for seed proteins (Tavakolan et al., 2013); SoyProLow: a database for low‐abundance proteins (Tavakolan et al., 2014); and SoyKB, a database for translational genomics and molecular breeding (Joshi et al., 2014; Table 2). These databases provide valuable platforms for soybean research.

Table 2

Soybean databases

Database	URL	Description	Reference
Soybean gene expression atlas	http://www.soybase.org/soyseq	a database of soybean 14 tissues specific gene expression	Severin et al. (2010)
Soybean cDNA sequenced	http://digbio.missouri.edu/soybean_atlas	a cDNA database of soybean developmental tissues specifically in root hair and meristem	Libault et al. (2010)
SoyNet	http://www.inetbio.org/soynet	a database for network‐based functional predictions	Kim et al. (2017)
Soybean transcriptome data	http://venanciogroup.uenf.br/cgi‐bin/gmax_atlas/index.cgi	a database of 1,298 publicly available soybean transcriptome	Machado et al. (2020)
Proteomics of oilseeds	http://oilseedproteomics.missouri.edu	expression profile data for proteomics research on soybean and other oilseeds plants	Agrawal et al. (2008)
Soybean Proteome Database SPD	http://proteome.dc.affrc.go.jp/Soybean/	a database of soybean proteomics	Ohyanagi et al. (2012); Komatsu et al. (2017)
SoyBase	http://www.soybase.org	a database of soybean genetics and genomics	Grant et al., 2010; Brown et al. (2020)
SoyTEdb	http://www.soytedb.org	a database of soybean transposable elements	Du et al. (2010a)
SoyProDB	http://bioinformatics.towson.edu/Soybean_Seed_Proteins_2D_Gel_DB/Home.aspx	a database for soybean seed proteins	Tavakolan et al. (2013)
SoyProLow	http://bioinformatics.towson.edu/Soybean_low_abundance_proteins_2D_Gel_DB/Gel1.aspx.	a database for soybean low abundant proteins	Tavakolan et al. (2014)
SoyKB	http://soykb.org	a database of soybean translational genomics and for soybean molecular breeding	Joshi et al. (2014)

Genes responsible for important agronomic traits

In the past decade, great efforts have been made to identify the genetic loci and the genes responsible for different agronomic traits, including yield, seed quality, stress, development and domestication. The efforts led to the identification of hundreds of QTLs (Hacisalihoglu et al., 2018), part of which have been integrated into the Soybase database (https://www.soybase.org/). Despite numerous QTLs were identified, only small number of responsible genes for these QTLs have been cloned, leaving most of them need to be cloned and functionally characterized. Here, we focus on a review of the genes that have been functionally validated (Table S1).

Yield components

Yield is one of the breeding priorities for soybean. Soybean yield is determined mainly by plant architecture, seed weight and size, and seed number per pod. In the past decade, several key genes controlling yield and related traits were cloned.

Plant architecture

Plant architecture is a key trait that significantly affects the yield of soybean. Dt1 and Dt2 are two key genes for plant height and growth habits (Liu et al., 2010; Ping et al., 2014; Tian et al., 2010). Analyses of population genetics suggested that Dt1, which is a homolog of Arabidopsis Terminal Flower 1 (TFL1), underwent artificial selection to create a determinate growth habit at early stages of landrace dissemination (Tian et al., 2010). Dt2 encodes a MADS‐domain factor (Ping et al., 2014) that could bind the promoter of Dt1 and repress its transcription (Liu et al., 2016b). Further functional investigation showed that Dt2 was a direct activator or repressor of the precursors of eight miRNAs whose target genes were associated with meristem maintenance, flowering time, stomatal density, water‐use efficiency (WUE) and/or stress responses (Zhang et al., 2019a). Other genes regulating plant architecture have also been characterized, such as SQUAMOSA‐Promoter Binding Protein‐Like (SPL) and Late Elongated Hypocotyl (LHY). The spl9a/spl9b‐1/spl9c/spl9d homozygous quadruple mutant plants had more branches than WT (Bao et al., 2019). SPL was regulated by miR156 to affect plant architecture in soybean (Sun et al. 2019; Wang and Wang 2015). When miR156b was overexpressed, the transgenic lines increased yield per plant by 46%‐63%, which was resulted from improved long branches number, nodes and pods number, 100‐seed weight (Sun et al., 2019). In soybean, the quadruple mutant of GmLHY displayed reduced plant height and shortened internodes (Cheng et al., 2019). Recently, GmMYB14 were characterized as an important factor to regulate plant architecture, high‐density yield and drought tolerance through the brassinosteroid (BR) pathway in soybean (Chen et al., 2021).

Seed size

In Arabidopsis, P450/CYP78A gene family had been found to control seed size (Fang et al., 2012; Wang et al., 2008). The orthologs of P450/CYP78A in soybean, including GmCYP78A10, GmCYP78A72, GmCYP78A57 and GmCYP78A70, showed conserved function to control seed size (Du et al., 2017; Wang et al., 2015b; Zhao et al., 2016). The genes BIG SEEDS1 (BS1) (Ge et al., 2016), GmKIX8‐1 (Nguyen et al., 2020), GmCIF1 (Tang et al., 2017), GmPSKγ1 (Yu et al., 2019), WRKY15a (Gu et al., 2017) and phosphatase 2C‐1 (PP2C‐1) (Lu et al., 2017b) were also involved in seed development in soybean. Furthermore, SoyWRKY15a and PP2C‐1 were found to have undergone artificial selection during soybean domestication (Gu et al., 2017; Lu et al., 2017b).

Flowering

Soybean is a short‐day and photoperiod sensitive plant (Miranda et al., 2020). Genetic analyses of natural variants have identified 12 maturity loci—E1 to E11 and J—that control flowering time and maturity. In the past decade, significant progress has been made in soybean flowering gene identification and functional characterization (Lin et al., 2020b).

E1 encodes a soybean‐specific transcription factor that contains a plant‐specific B3 domain (Xia et al., 2012). E1 could repress flowering under long‐day (LD) conditions, while its leaky allele (e1‐as) and null alleles (such as e1‐fs and e1‐nl) caused earlier flowering (Xia et al., 2012; Xu et al., 2013a; Yasutaka et al., 2013). E2 encodes a homolog of Arabidopsis GIGANTEA (GI), which is a component of the circadian clock (Watanabe et al., 2011). The functionally dominant E2 delayed flowering time, while the recessive e2 induced flowering (Watanabe et al., 2011). E3 and E4 were found to be functionally redundant and encode GmPHYA3 (Watanabe et al., 2009) and GmPHYA2 (Liu et al., 2008; Yasutaka et al., 2013), respectively, which are homologues of the photoreceptor phytochrome A (PHYA) (Franklin and Quail, 2010). E3 and E4 were involved in the control of flowering under LDs with high red:far‐red (R:FR) quantum ratios and low R:FR ratios, respectively (Cober et al., 1996; Cober and Voldeng, 2001). J is an important locus for the adaptation of soybean to lower latitudes. J was found to encode an EARLY FLOWERING 3 (ELF3) homolog (Fang et al., 2020; Lu et al., 2017a; Yue et al., 2017). A pair of pseudo‐response‐regulators (PRRs; named Tof11 and Tof12) were reported to contribute to changes in flowering and early maturity in soybean (Day et al., 2010; Gong, 2020; Li et al., 2020a; Lu et al., 2020b; Wang et al., 2020c).

Other genes have also been reported to function in regulating flowering in soybean, particularly genes homologous to flowering‐related genes from other species. FLOWERING LOCUS T (FT) is required for flowering and widely conserved among plant species. In soybean, 10 FT/TSF homologs were identified. These FT/TSF homologs had divergence functions, and their natural variation might be associated with soybean adaptation to different environments (Jiang et al., 2019; Kong et al., 2010; Wu et al., 2017). Overexpression of GmFT2a/E9 or GmFT5a promoted flowering in soybean, whereas overexpression of GmFT1a or GmFT4 suppressed flowering (Cai et al., 2018; Cai et al., 2020b; Chen et al., 2020b; Kong et al., 2010; Liu et al., 2018; Nan et al., 2014; Takeshima et al., 2016; Zhai et al., 2014a). FT4 is a candidate gene for the E10 locus (Samanfar et al., 2017). Although FT2a and FT5a showed similar functions in inducing flowering, FT5a was more effective than FT2a in the post‐flowering termination of stem growth (Takeshima et al., 2019). Further investigation showed that several floral genes, such as GmAP1, GmSOC1 and GmLFY, were significantly induced by GmFT2a/E9 and GmFT5a (Nan et al., 2014). Overexpression of GmAP1a, one of the homologs of Arabidopsis APETALA1 (AP1) which functions as a class A gene in the ABCDE model of flowering, resulted in early flowering and reduced plant height compared with the wild type under short‐day (SD) conditions (Chen et al., 2020c). GmFDL19, which is a bZIP transcription factor, interacted with GmFT2a/E9 and GmFT5a. Overexpression of GmFDL19 in soybean resulted in early flowering through the regulation of the expression of GmAP1a (Nan et al., 2014). GmGBP1 functioned as a positive regulator upstream of GmFT2a/E9 and GmFT5a to activate the expression of GmFULs to promote flowering under LDs (Zhao et al., 2018).

In the photoperiod‐controlled flowering pathway, the circadian clock‐regulated gene CONSTANS (CO) is crucial for the induction of the FT gene (Turck et al., 2008). The soybean genes GmCOL1a/GmCOL1b showed high sequence homology to AtCO. The late‐flowering phenotype of the Arabidopsis co‐1 mutant was fully complemented by overexpression of GmCOL1a or 1b, suggesting that they function similarly to AtCO at the protein level (Wu et al., 2014). Further analysis showed that GmCOL1a controlled flowering time by suppressing the expression of GmFT2a/E9 and GmFT5a under LDs. In addition, GmCOL1a/1b was up‐regulated by E1, E2, E3 and E4 (Cao et al., 2015a). It was found that overexpression of GmAGL1, which encodes a MADS‐box protein, promoted flowering, maturity and led to a smaller floral organ (Chi et al., 2017; Zeng et al., 2018). Moreover, several miRNA families were also demonstrated to play important roles in controlling flowering in soybean. Overexpression of MIR156b, which is involved in the PHYA‐mediated photoperiod response pathway, delayed flowering under LDs (Cao et al., 2015b). In Arabidopsis, miR156 regulated the expression of miR172 via SPL9 (Wu et al., 2009). In soybean, miR156b down‐regulated MIR172 and SPL9 and up‐regulated the miR172 target gene GmTOE4a in soybean (Cao et al., 2015b). In addition, E2 could promote the maturation of MIR172a via increasing the expression of DICER‐LIKE 1 and SERRATE homologs (Wang et al., 2016b).

Functional investigations have demonstrated that the above genes are involved in a complicated PHYA‐mediated photoperiod response pathway that regulates flowering time in soybean (Kong et al., 2010; Li et al., 2020a; Lu et al., 2020b; Lu et al., 2017a; Xia et al., 2012; Xu et al., 2015). In Figure 2, we summarized the current understanding of this flowering time regulation under LDs. Briefly, E3 and E4 were up‐regulated Tof11 and Tof12 expression. Then, Tof11 and Tof12 proteins physically could bind to the promoters of the GmLHY genes to suppress their expression, which suppressed their subsequent induction of J. At low levels, J is not able to suppress the transcription of E1. The abundance of E1 decreases the expression of GmFT2a/E9 and GmFT5a, ultimately resulting in delayed flowering and later maturity (Figure 2). Recently, it was found that GmLUX2, an ortholog of the Arabidopsis Evening Complex (EC) component LUX ARRHYTHMO (LUX), could physically interact with GmELF3a/b to regulate the expression of several circadian clock‐associated genes and directly repress E1 expression (Fang et al., 2021a) (Figure 2).

Figure 2

Proposed molecular regulation network of flowering in soybean. Soybean is a short‐day flowering plant. E3 and E4 mediate flowering by responding to the ratios of red (R) and far‐red (FR) light. Under long days (LDs), E3 and E4 induce the expression of E1. GmTof11 and GmTof12 inhibit the expression of GmLHYs by binding to their promoters. GmLHY proteins directly bind to the promoter region of J to induce its expression and bind to the promoter of E1 to suppress its transcription. LUX can physically interact with GmELF3a/b to repress E1 expression. E1 inhibits the expression of the flowering‐inducing factors GmFT2a and GmFT5a. This suppresses the expression of floral identity genes (GmAP1, GmSOC1s, GmLFYs, GmFULs). As a result, flowering is delayed under LDs. Under SDs, the induction of E1 is decreased, which releases the transcriptional suppression of GmFT2a and GmFT5a, promoting flowering. The E2 and miRNA‐dependent flowering regulation modules are also influenced by photoperiod under PHYA mediation (E3 and E4). GmCOL1a/1b and GmTOE4a inhibit flowering via down‐regulation of GmFT2a and GmFT5a expression. GmTOE4a expression is possibly controlled by miR172, which is under the control of miR156b. miR156b may delay flowering by negatively regulating the GmSPLs.

Others

Other genes that control yield‐related developmental traits were also identified. Ln, which encodes a C2H2 zinc finger transcription factor, was found to be a key regulator of seed number per pod (Fang et al., 2013; Jeong et al., 2011; Jeong et al., 2012). GmILPA1, which encodes an APC8‐like protein, was found to control petiole angle (Gao et al., 2017). Chicken Toes‐Like Leaf and Petalody Flower (CTP) was found to be a novel regulator that controlled leaf and flower development in soybean. Mutation of CTP resulted in leaf defects and additional flower parts developing into petals (Zhao et al., 2017). It was found that GmCRY1s could modulate gibberellin metabolism during the response to reduced blue light, affecting yield in a shaded environment (Lyu et al., 2021).

Several genes controlling senescence are also identified. D1 and D2, homologs of STAY‐GREEN (SGR), and GmCHLI, a subunit of magnesium (Mg)‐chelatase, regulated chlorophyll degradation in soybean (Fang et al., 2014; Li et al., 2016a; Slattery et al., 2016). The maternally inherited stay‐green gene cytG was found to encode the chloroplast protein PsbM (Kohzuma et al., 2017). Under blue light, cryptochrome2 (CRY2a) interacts with the soybean basic helix‐loop‐helix transcription activator CIB1 to regulate leaf senescence in soybean (Meng et al., 2013).

Seed composition

Protein, oil and isoflavone content are the main soybean seed quality traits and are determined via interaction between quantitative loci and the environment. Seed storage proteins usually showed negative correlation with oil content (Chaudhary et al., 2015) and seed yield (Bandillo et al., 2015; Chung et al., 2003; Zhang et al., 2018). Besides, the contents of soluble sugar, mineral nutrition and amino acid are also important for seed quality. Hundreds of QTLs that affected soybean seed quality have been identified using both pedigree‐based and association mapping approaches. Some of this information has been integrated into Soybase (https://www.soybase.org/).

Oil content and fatty acid

Triacylglycerols (TAGs) are the main components of soybean seed oil. Lines carrying knockdowns of four TAG lipase‐encoding genes, namely GmSDP1‐1, GmSDP1‐2, GmSDP1‐3 and GmSDP1‐4, have significantly increased seed oil content (Kanai et al., 2019). Overexpression of the phospholipid diacylglycerol acyltransferase GmPDAT increased seed oil and also altered size‐related traits (Liu et al., 2020a; Liu et al., 2020b). A comprehensive assessment of the gene co‐expression networks showed that GA20 oxidase (GA20OX) and nuclear transcription factor Y subunit alpha (NFYA) are two key drivers of seed traits. Overexpression of GA20OX and NFYA enhanced seed size and weight and oil content, respectively (Lu et al., 2016). It was found that oleosin‐encoding GmOLEO1 was predominantly expressed during seed maturation, and overexpression of GmOLEO1 significantly increased seed oil content by 10.6% (Zhang et al., 2019b). Overexpression of GmWRI1a, an APETALA2/ethylene responsive element‐binding protein (AP2/EREBP) encoding gene, under the control of a seed‐specific promoter significantly increased total oil and fatty acid content and also changed fatty acid composition (Chen et al., 2018b). Genetic modification of the fatty acid desaturase FAD significantly changes the composition of polysaturated and polyunsaturated fatty acids (Al Amin et al., 2019; Andreu et al., 2010; Combs and Bilyeu, 2019; Haun et al., 2014; Lakhssassi et al., 2017; Lim et al., 2014; Wagner et al., 2011). Genetic modification of DGAT, which encodes a type 1 diacylglycerol acyltransferase, increased the oil content and decreases the soluble carbohydrate content in soybean seeds (Roesler et al., 2016).

Several transcription factors were also found to be associated with seed quality. Overexpression of transcription factors GmbZIP123, GmMYB73, GmZF351, GmZF392 and GmWRI1b enhanced lipid content (Guo et al., 2020; Li et al., 2017b; Liu et al., 2014b; Song et al., 2013a). Further functional investigation showed that the enhancement of oil content by GmMYB73 was through interaction with GL3 and EGL3 and the suppression of GL2, a negative regulator of oil accumulation (Liu et al., 2014b). GmZF351 was more highly expressed in cultivated soybeans compared with wild soybeans, and this was due to a mutation of its promoter, which had undergone artificial selection during soybean domestication (Li et al., 2017b). GmZF392 physically interacted with GmZF351, and both GmZF392 and GmZF351 could be further induced by GmNFYA (Lu et al., 2021).

Isoflavone

Isoflavone content is highly influenced by the genes involved in chalcone synthesis, including chalcone synthase, chalcone reductase and chalcone isomerase (Dastmalchi et al., 2016). Among 11 chalcone reductase (CHR) proteins of soybean, CHR5 was found to interact with a 2‐hydroxyisoflavanone synthase (IFS) isozyme and most correlated with the distribution patterns of 5‐deoxyisoflavonoids, indicating that it may play an important role in the isoflavone pathway (Mameda et al., 2018). In soybean, some prenyltransferase‐encoding genes exhibited isoflavonoid‐specific patterns (Sukumaran et al., 2018). When the gene encoding carotenoid cleavage dioxygenase 4 (GmCCD4) was knocked out, the loss‐of‐function lines showed increased carotenoid content and yellow flowers (Gao et al., 2021). Several independent studies demonstrated that the MYB gene family played important roles in affecting isoflavone synthesis, which may involve a pathway that consists of 14‐3‐3s and CHS proteins (Chu et al., 2017; Li et al., 2012; Pandey et al., 2014; Yi et al., 2010). Recently, it was reported that selection of a class B heat‐shock factor, HSFB2b, in soybean domestication promoted flavonoid biosynthesis and enhanced salt tolerance (Bian et al., 2020). An analysis combining GWAS and QTLs identified GmMPK1, which encodes a mitogen‐activated protein kinase, as one of the candidate genes responsible for isoflavone content (Wu et al., 2020).

Protein content, amino acid and others

Compared with genes for oil content, fewer genes controlling protein content or amino acid have been functionally identified (Krishnan and Jez, 2018). It was found that MGL, a putative methionine γ‐lyase, may be responsible for the accumulation of S‐methylmethionine in soybean seed (Teshima et al., 2020). Overexpression of the cytosolic isoform of O‐acetylserine sulfhydrylase (OASS) and the plastid ATP sulfurylase isoform 1 improved the cysteine‐rich proteins and sulphur amino acid content in transgenic soybean, independently (Kim et al., 2012b; Kim et al., 2020). Rab5a, a small GTPase‐encoding gene, was reported to be involved in post‐Golgi trafficking of storage proteins in developing soybean cotyledons (Wei et al., 2020b). Recently, it was found that SWEET sugar transporters played important roles in soybean seed quality through effects on the contents of both oil and protein. GmSWEET15 mediated sucrose export from endosperm to early embryo (Wang et al., 2019d). GmSWEET10a and GmSWEET10b determined oil and protein contents and seed size simultaneously in soybean through affecting sugar allocation from seed coat to embryo (Miao et al., 2020; Wang et al., 2020e; Zhang et al., 2020c).

Stress resistance

Abiotic and biotic stresses greatly affect soybean yield and quality (Anderson et al., 2019). Drought and salinity can reduce soybean yield by 40% through negative impacts on growth, nodulation, flowering, seed quality and seed quantity (Anderson et al., 2019; Papiernik et al., 2005; Specht et al., 1999). Stress (including drought, salt, temperature stress, flooding and disease) has been intensely studied in soybean (Kunert et al., 2016; Li et al., 2021; Li et al., 2014a; Phang et al., 2008; Ramesh et al., 2019; Robison et al., 2019; Shu et al., 2020; Whitham et al., 2016; Widyasari et al., 2020).

Drought

Several transcription factors were reported to play important roles in soybean drought tolerance, such as WRKY and NAC gene families. GmWRKY54 increased drought tolerance by activating genes in the abscisic acid and Ca2+ signalling pathways (Wei et al., 2019). Overexpression of WRKY20 from a wild soybean significantly enhanced drought tolerance in Arabidopsis (Luo et al., 2013). GmNAC8 overexpression and knockout transgenic lines exhibit significantly higher and lower drought tolerance, respectively, suggesting that GmNAC8 is a positive regulator of drought tolerance (Yang et al., 2020).

U‐box (PUB) proteins function as E3 ligases in plants. When GmPUB6 was overexpressed in Arabidopsis, the transgenic lines exhibited decreased plant survival under drought stress condition (Wang et al., 2020d). Several reports suggested that circadian clock genes are also involved in drought tolerance, such as the GmLHYs and GmLCL, could affect drought tolerance (Wang et al., 2021a; Yuan et al., 2021). It is reported that many clock genes respond to both flooding and drought, with the expression patterns of some genes shifting in amplitude and phase (Marcolino‐Gomes et al., 2014; Syed et al., 2015).

Other genes that affect drought tolerance in soybean have also been identified. For instance, the genes responsible for pubescence density, Pd1, P1, Ps and Pd2, were also found to be involved in drought tolerance in soybean (Liu et al., 2020c; Pfeiffer et al., 2003). Overexpression of the LOS5/ABA3 gene, which encodes a molybdenum cofactor sulfurase, using a constitutive expression promoter in soybean enhanced drought tolerance and increased seed yield by at least 21% under drought stress (Li et al., 2013a). Drought tolerance in soybean through foreign gene transformation was also achieved, such as the overexpression of rice cystatin oryzacystatin I (OCI), Arabidopsis DREB1A, and the sunflower transcription factor HB4 (Quain et al., 2014; Rakocevic et al., 2018; Ribichich et al., 2020).

Salt

Stresses, such as salinity, osmotic stress, imbalance of ions, ion toxicity and excessive reactive oxygen, have significant effects on the growth of plant (Ruiz‐Lozano et al., 2012). Under high salinity, plants slow down photosynthesis and ramp up sugar catabolism to provide extra energy for survival (Liu et al., 2019). Balancing the ions and reducing the ion toxicity are the keys to enhance the salt resistant. Through forward genetic approaches, two research groups independently found that GmSALT3, which encodes a cation/H+ exchanger, could limit the accumulation of sodium ions (Na+) in shoots and enhance salt tolerance in soybean (Guan et al., 2014; Qi et al., 2014). Further investigation suggested that GmSALT3 functioned through exclusion of sodium ions (Na+) from the leaf via a root‐derived mechanism and exclusion of chloride ions (Cl–) via a shoot‐derived process (Qu et al., 2021). GmCDF1, encoding a cation diffusion facilitator, could negatively regulate salt tolerance by maintaining K+‐Na+ homeostasis in soybean (Zhang et al., 2019c). GmAKT1, a K+ transporter encoding gene, located in the plasma membrane and was recently reported to play an important role in soybean salt resistance by regulating the K+ uptake and Na+/K+ balance (Wang et al., 2021b).

Other genes that have effects on salt tolerance in soybean were also reported. GmNAC109 and SALT INDUCED NAC1 (GmSIN1) were found to promote root growth and increase abiotic stress tolerance through up‐regulation of ABA synthesis‐associated and ROS generation genes (Li et al., 2019c; Yang et al., 2019). The salt‐inducible gene GmbZIP110 encodes a protein that could bind to the promoters of genes with an ACGT motif and impact the expression of many stress‐related genes and enhance salt tolerance (Xu et al., 2016). Overexpression of GmNAC20 or GmNAC11 enhanced salt tolerance in transgenic Arabidopsis plants (Hao et al., 2011). Interestingly, miR172 was also involved in the response to salt stress. Under salinity, plants with hairy roots overexpressing the pre‐miR172a had healthier leaves and larger roots. Further investigation showed that miR172a promoted salt tolerance mainly through cleaving the AP2/EREBP‐type transcription factor SSAC1 gene, releasing its inhibition on THI1, which encodes a positive regulator of salt tolerance (Pan et al., 2016). Overexpression of soybean MIR172c conferred tolerance to both water deficit and salt stress in transgenic Arabidopsis (Li et al., 2017d). The plant homeodomain protein GmPHD6 functioned as a histone code reader and interacted with LHP1 to form a transcriptional activator that regulated genes for salt tolerance. Overexpression of GmPHD6 improved salt tolerance in soybean (Wei et al., 2017). HSFB2b, a B heat‐shock factor encoding gene, improved salt tolerance by promoting flavonoid biosynthesis. Interestingly, HSFB2b had undergone artificial selection during soybean domestication (Bian et al., 2020). In addition, mitogen‐activated protein kinase (MAPK), BURP‐domain proteins and NAD(P)H dehydrogenase (NDH) were also found to be involved in salt tolerance in soybean (He et al., 2015; Im et al., 2012; Wang et al., 2012a).

Genes can affect both drought and salt tolerance were also reported. For instance, overexpression of GmMYB118, GmERF135, GmCDPK3 improved drought and salt tolerance simultaneously (Du et al., 2018; Wang et al., 2019a). Interestingly, it was also found that even different members from the same WRKY family having differential effects on abiotic stress tolerance in soybean (Wang et al., 2015a; Wei et al., 2019; Zhou et al., 2008).

Minerals

Iron deficiency results in stunting and yield loss and is one of the most common and severe nutritional stresses for soybean (Hacisalihoglu et al., 2018). Soybean plants reprogramme metabolism under iron deficiency (Chu et al., 2019). Moreover, the circadian clock was highly responsive to iron deficiency (Li et al., 2019b). Several genes that may be involved in iron deficiency regulation were identified. Silencing of GmRPA3, which encodes subunit 3 of replication protein A, showed reduced iron deficiency chlorosis (IDC) and increased chlorophyll content under iron‐deficient conditions (Atwood et al., 2014). It was found that the physical interaction between two soybean bHLH factors, GmbHLH57 and GmbHLH300, was important for Fe homeostasis (Li et al., 2018a). Genetic variation of the genes encoding a Fe deficiency‐induced transcription factor and a Fe/Zn‐regulated transporter may be responsible for variation of iron use efficiency among different soybean accessions (Liu et al., 2020f; Peiffer et al., 2012).

Several reports have focused on the investigation of phosphorus‐use efficiency (PUE). GmACP1 encodes an acid phosphatase and is a candidate for a PUE locus identified via GWAS. Overexpression of GmACP1 in soybean hairy roots significantly increased PUE (Zhang et al., 2014). Under Pi starvation, the expression of GmPT1, GmPHR25 and CWPs was induced, indicating they may be involved in PUE regulation (Song et al., 2014; Wu et al., 2018; Xue et al., 2017). The functions of genes in the GmALMT family were also investigated, and GmALMT5 was found to enhance utilization of soluble P under P‐limited conditions (Peng et al., 2018). Moreover, ethylene may also affect PUE. Up‐regulating GmETO1, which encodes ethylene‐overproduction protein 1, could significantly improve phosphorus uptake and use efficiency in soybean (Zhang et al., 2020d). A single point mutation in GmHMA3, which encodes a heavy‐metal transporter, may be responsible for differential cadmium (Cd) translocation and accumulation in the seeds of different soybean accessions (Wang et al., 2012b). Further, hydrogen sulphide affected aluminium and nitrogen assimilation (Wang et al., 2019b; Zhang et al., 2020e).

Pathogens

Soybean cyst nematode (SCN) is the most devastating soybean pathogen in the United States, with yield losses ranging from 1.9 to 3.5 million tonnes per year (Liu et al., 2015; Wrather and Koenning, 2006). SCN infection induced a large number of genes related to cell wall modification, stress response, defence and signal transduction (Kandoth et al., 2011; Rambani et al., 2015; Tucker et al., 2011). The proteins potentially related to SCN were collected in the database SCNProDB (Natarajan et al., 2014).

Rhg1 and Rhg4 are two major QTLs/genes conveying SCN resistance (Liu et al., 2012; McHale et al., 2012; Mitchum, 2016). Rhg4 encodes a serine hydroxymethyltransferase (SHMT), which mediated a novel plant resistance mechanism against a pathogen (Liu et al., 2012). In the Rhg1 region, the copy number of three genes, Glyma18g02580, Glyma18g02590 and Glyma18g02610, was associated with SCN resistance (Cook et al., 2014; Cook et al., 2012; Lee et al., 2016). Further investigation demonstrated that the repeated copies of a gene encoding atypical α‐soluble N‐ethylmaleimide‐sensitive factor (NSF) attachment protein (α‐SNAP) were the most likely candidate for conferring resistance to SCN (Liu et al., 2017; Patil et al., 2017). In Rhg1(+) germplasm, analysis of an unusual NSF allele [Rhg1‐associated NSF on chromosome 07; NSF (RAN07)] revealed that NSF (RAN07) exhibited stronger in vitro binding with Rhg1 resistance‐type α‐SNAPs, suggesting that an atypical co‐evolution of the soybean SNARE‐recycling machinery balances the acquisition of an otherwise disruptive housekeeping protein, enabling a valuable disease resistance trait (Bayless et al., 2018).

Resistance to SCN involves salicylic acid. Overexpression of a salicylic acid methyltransferase gene conferred resistance to SCN (Lin et al., 2016; Lin et al., 2013). The concentration of 1‐aminocyclopropane‐1‐carboxylic acid (ACC) and expression of ACC synthase were both higher in SCN‐colonized root pieces and root tips than in other parts of the root (Tucker et al., 2010). Moreover, CLE (encodes an extracellular protein), GmAFS (encodes one member of the soybean terpene synthase gene family), MIR396 and t‐SNAREs also respond to SCN infection (Dong et al., 2020; Guo et al., 2015; Lin et al., 2017; Noon et al., 2019; Noon et al., 2016). Along with SCN infection, variation of DNA methylation in some genomic regions associated with changes in gene expression (Rambani et al., 2020).

A report on the dissection of the QTL for southern root‐knot nematode (RKN) resistance in soybean identified three QTL and mapped the major QTL to a 29.7‐kb region on chromosome 10 (Xu et al., 2013b). In addition, there are several reports on soybean resistance to other pests. For instance, defence against Anticarsia gemmatalis larvae was modulated by solar UV‐B radiation and ethylene (Dillon et al., 2018). Recently, a gene encoding a VQ motif‐containing protein, GmVQ58, was found to enhance soybean resistance to the common cutworm (Li et al., 2020c).

Asian soybean rust (ASR), caused by the obligate biotrophic fungus Phakopsora pachyrhizi, and is one of the most economically important diseases for soybean. Through in vivo assessment by Mach‐Zehnder double‐beam interferometry, Loehrer et al. found that P. pachyrhizi might be able to forcefully invade a wide range of plants through appressorial turgor pressure (Loehrer et al., 2014). Both a UDP‐glucosyl transferase and phenylpropanoid metabolism are essential for the infection of P. pachyrhizi (Beyer et al., 2019; Langenbach et al., 2013), while coumarin could be used as a natural fungicide against ASR (Beyer et al., 2019; Langenbach et al., 2013). Evolutionary analyses indicated that some disease resistance genes have conserved function between soybean and other species, indicating that genes identified from other species could be used in soybean disease resistance (Ashfield et al., 2014; Okutani et al., 2020; Redditt et al., 2019; Wang et al., 2014a; Wei et al., 2020a). When the gene CcRpp1 (Cajanus cajan Resistance against Phakopsora pachyrhizi 1) from pigeon pea or NHR‐linked genes from Arabidopsis were transferred to soybean, the transgenic lines exhibited resistance to P. pachyrhizi (Kawashima et al., 2016; Langenbach et al., 2016).

Through GWAS and QTL analyses, several loci associated with pattern‐triggered immunity (Valdés‐López et al., 2011), sudden death syndrome resistance (Zhang et al., 2015) and white mould (Zhao et al., 2015b) were identified. Genetic mapping suggested that STAY‐GREEN genes may be involved in sudden death syndrome (Chang et al., 2019), and Rsv4, which encodes an RNase H family protein with dsRNA‐degrading activity, may be responsible for broad‐spectrum mosaic virus resistance (Ishibashi et al., 2019). Overexpression of the HSP40 gene, which encodes a nuclear‐localized, type‐III DnaJ domain‐containing protein, or GmLMM1, which encodes a malectin‐like receptor kinase, regulated cell death and disease resistance (Liu and Whitham, 2013; Wang et al., 2020a). Overexpression of the plasma membrane protein gene GmDR1 generated broad‐spectrum immunity (Ngaki et al., 2021). Interestingly, genes encoding small peptides, such as GmSubPep, encoding a 12‐amino acid peptide (Pearce et al., 2010), and GmPep914, encoding an 8‐amino acid peptide (Yamaguchi et al., 2011), also played important roles in defence against disease. It was recently found that hydroperoxide lyase modulated the defence response and conferred lesion‐mimic phenotype in soybean leaves (Wang et al., 2020f).

Phytophthora root and stem rot, caused by Phytophthora sojae, is another destructive soybean disease. Overexpression of GmERF5, which encodes ethylene response factor 5 (Dong et al., 2015), or GmMYB29A2, which encodes a glyceollin transcription factor (Jahan et al., 2020), significantly enhanced resistance to P. sojae. Moreover, overexpression of some microRNAs, such as miR393, could promote soybean defence against P. sojae (Wong et al., 2014).

Soybean mosaic virus (SMV) is one of the most prevalent viral diseases and could significantly reduce yield losses in soybean. Until now, four dominant SMV resistance loci (Rsv1, Rsv3, Rsv4 and Rsv5) have been genetically identified (Hayes et al., 2000; Jeong et al., 2002; Klepadlo et al., 2017; Yu et al., 1994). Recently, GmST1, which encodes a sulfotransferase, was identified as the responsible gene for conferring the resistance to strains G2 and G3 (Zhao et al., 2021).

R genes specifically activate resistance responses that are effective against diverse pathogens (Zheng et al., 2016). In soybean, Kang et al. (2012) predicted that the genome contains a total of 319 nucleotide‐binding site/leucine‐rich repeat (NBS‐LRR) R genes. The ULP1‐NBS‐LRR protein GmRpp1 confers immunity to P. pachyrhizi (Pedley et al., 2019). Overexpression of the TIR–NBS–LRR R gene GmKR3 enhanced soybean resistance to several strains of soybean mosaic virus (SMV), which is one of the most prevalent viral diseases and could significantly reduce yield losses in soybean (Xun et al., 2019). Silencing of the soybean NDR1 homologs (GmNDR1) showed that they were required for pathogen resistance (Selote et al., 2014). GmMPK4‐ and GmMPK6‐silenced plants displayed strong phenotypes, including induction of PR gene expression and increased SA levels (Liu et al., 2014a; Liu et al., 2011). Fine‐tuning the expression of pathogen avirulence (Avr) effector genes using genome editing impacted the compatibility of plant disease, which provided clues to improve crop disease resistance (Ochola et al., 2020).

Other stressors

Salicylic acid (SA) and abscisic acid (ABA) are two important phytohormones for stress resistance. Knockdown of either the phenylalanine ammonia lyase or isochorismate synthase biosynthesis pathway shuts down SA biosynthesis and abrogates pathogen resistance (Shine et al., 2016). In soybean, it was found that ABA‐Sensitive 1 (GmABAS1), which encodes a 1R‐subtype of MYB, functioned as a transcriptional repressor that enhances ABA sensitivity (Ku et al., 2020). Several transcription factor families were found to be involved in various stress responses. For instance, AP2/ERF‐type transcription factor family members, including GmERF3 and GsERF7, played cardinal roles in regulating resistance to diseases as well as salt and drought (Feng et al., 2020; Zhang et al., 2009), whereas the DREB1/CBF‐type transcription factors function in heat, drought and cold stresses (Kidokoro et al., 2015).

Other genes that response various stresses were also reported. The root‐specific protein kinase‐encoding gene GmWNK1 and the mitogen‐activated protein kinase‐encoding gene GmMPK were found to regulate plant growth and development and, in turn, affect stress responses (Liu et al., 2011; Wang et al., 2010). Malate exudation mediated by Gm Representative, which encodes an expansin, was found to be involved in multiple abiotic stresses (Guo et al., 2011). A chaperone binding protein (BiP) functioned as a negative regulator to attenuate stress‐induced cell death (Reis et al., 2011). Omega‐3 fatty acid desaturase was found to play an important role in cold response in soybean (Román et al., 2012). For competition with weeds, a PIF3‐like protein was identified as a candidate that might manipulate the weed stress response in soybean (Horvath et al., 2015).

Nodulation

Symbiotic nitrogen fixation (SNF) through root nodulation is an important feature of legumes and plays important roles in plant growth. Over the past decade, the genes required for symbiotic nitrogen fixation have been thoroughly investigated, and significant progress has been achieved (Roy et al., 2020).

A comprehensive phylogenomic analysis revealed that multiple losses of some key genes, such as NIN and RPG, were essential for the origin of SNF (Griesmann et al., 2018). A global co‐expression analysis suggested that ancient orthologous and duplication events before the origin of legumes had paved the way for nodule formation and nitrogen fixation (Wu et al., 2019). Large‐scale transcriptome and metabolome investigations revealed a number of genes and metabolic pathways that are induced or suppressed during nodulation (Agtuca et al., 2020; Hayashi et al., 2012b; Libault et al., 2010). For instance, G protein‐encoding genes and a putative beta‐carotene hydroxylase gene (GmBCH1) exhibited significant transcriptional changes in response to rhizobium infection. RNA interference suppression of the genes encoding G protein and GmBCHs severely impaired nitrogen fixation as well as nodule development, suggesting they are positive regulators in nodule formation (Choudhury and Pandey, 2013; Kim et al., 2013). Furthermore, the G protein cycle was regulated by the activity of phosphorylation‐dependent G protein signalling proteins (Choudhury and Pandey, 2015). The G‐box Factor 14‐3‐3 genes, SGF14c and SGF14l, had been shown to function as dimers in soybean nodulation (Radwan et al., 2012).

In soybean, the formation of symbiotic root nodules was highly affected by several host genes, referred to as Rj or rj (Hayashi et al., 2012a). Rj1 and Rj5 encode putative Nod factor receptors (NFRs) (Hayashi et al., 2012a; Indrasumunar et al., 2010; Indrasumunar et al., 2011), Rj2/Rfg1 encodes a Toll‐interleukin receptor/nucleotide‐binding site/leucine‐rich repeat (TIR‐NBS‐LRR) plant resistance (R) protein (Yang et al., 2010), while Rj7 encodes a nodule autoregulation receptor kinase (Hayashi et al., 2012a; Nishimura et al., 2002; Searle et al., 2003). One report suggested that a gene encoding thaumatin‐like protein (TLP), a pathogenesis‐related (PR) protein, might be the candidate for the Rj4 locus (Hayashi et al., 2014; Tang et al., 2016).

It has been suggested that flavonoids act as chemotactic signals to rhizobia under low‐N conditions in legumes (Liu and Murray, 2016; Subramanian et al., 2006). GmMaT, a malonyl‐CoA:flavonoid acyltransferase‐encoding gene, was found to catalyse isoflavone malonylation and affect malonyl isoflavone secretion, which had an effect on soybean nodulation (Ahmad et al., 2021). Flavonoids and/or related compounds such as isoflavones are released from legume roots and attract the Nod factors (NFs), which are secreted by rhizobia to initiate symbiotic nitrogen fixation (Figure 3). Soybean isoflavone synthase (IFS), a key enzyme in the biosynthesis of isoflavones, was induced by Bradyrhizobium japonicum. When the expression of IFS was knock down in soybean hairy root composite plants, these plants showed severely reduced nodulation (Liu and Murray, 2016; Subramanian et al., 2004; Subramanian et al., 2006). Additionally, glycolysis and lipid biosynthesis may also play essential roles in nodule formation (Chen et al., 2020a; Gillman et al., 2014; Lakhssassi et al., 2020; Zhang et al., 2020b).

Figure 3

Proposed molecular network of soybean nodulation. Under low soil N, soybean plants produce flavonoids that trigger the production of bacterial Nod factors (NFR1 and NFR5), which, together with other signals, are perceived by receptors at the plasma membrane of plant epidermal cells. Rj2/Rfg1 encodes a member of the Toll‐interleukin receptor/nucleotide‐binding site/leucine‐rich repeat (TIR‐NBS‐LRR) class of plant resistance (R) proteins that functions in symbiotic root nodules. Nodule overproduction is prevented by AON signalling, in which GmRIC1 and GmRIC2 activate GmNARK. GmNARK induces shoot‐derived cytokinins that, in turn, repress the transcriptional activity of miR172c. miR172c is a key positive regulator of nodulation and promotes the cleavage of the mRNA encoding its target gene Nodule Number Control 1 (NNC1), which directly suppresses the transcription of the early nodulin gene ENOD40. Auxin, brassinosteroids (BRs), strigolactones, cytokinins and abscisic acid (ABA) are reported to function in nodulation in soybean. GmBEHL1, related to the BR signal, is an NNC1‐interacting protein. GmBEHL1 is suggested to be a co‐repressor of NNC1 and to negatively regulate soybean nodulation.

The number of nodules formed on the roots of soybean is systemically controlled by autoregulation of nodulation (AON) (Caetano‐Anolles and Gresshoff, 1991). NTS1/GmNARK was found to be an important sensor that controls nodulation (Searle et al., 2003). Mutation of the NTS1/GmNARK locus led to prolific nodulation (Carroll et al., 1985a; Searle et al., 2003). Another two signalling molecules, Bradyrhizobium‐induced and acting systemically 1 (GmRIC1) and GmRIC2, were involved in the transition of long‐distance signals from root to shoot. Further investigation showed that GmNARK can be activated by GmRIC1 and GmRIC2 (Reid et al., 2012). Overexpression of GmRIC1 and GmRIC2 strongly suppressed the nodulation in a GmNARK‐dependent manner (Reid et al., 2011) (Figure 3). An inoculation‐ and NARK‐dependent gene (GmUFD1a) responds in both the bioassay and intact, inoculated plants, indicating that it might be a novel component of the autoregulation pathway (Reid et al., 2012). The transcription factors GmNF‐YA1a and b were also identified as positive regulators in AON (Schaarschmidt et al., 2013). Overall, GmNIC1 (nitrate‐induced and acting locally) and GmRIC1 played key roles in AON and were relied on the activity of the nodulation autoregulation receptor kinase GmNARK (Reid et al., 2013).

Nitrogen and nitrogen compounds such as nitrate have been reported to negatively control nodulation (Carroll et al., 1985a,b; Day et al., 2010; Lim et al., 2014; Reid et al., 2011; Tanaka et al., 1985). The processes of nitrogen regulation of nodulation are distinct from AON and act immediately. Interestingly, a leucine‐rich receptor‐like kinase, GmNARK, was shared between AON and nitrate‐dependent regulation of nodulation (Ferguson et al., 2019). Under high nitrate, loss‐of‐function GmNARK mutants, nts (nitrate tolerant symbiosis), still exhibited super nodulation (Carroll et al., 1985b). GmNARK perceived the nodulation‐suppressing CLE peptides, such as GmRIC1 and GmRIC2, in the shoot through the AON pathway and sensed the nodulation‐suppressing CLE peptides, such as GmNIC1, in the root through nitrogen regulation of nodulation (Ferguson et al., 2019; Lim et al., 2014; Reid et al., 2011).

The development of nodules is significantly affected by phosphate (Pi). Pi starvation severely inhibited both nodulation and biological N2 fixation (Hernandez et al., 2009), with decreased nodule numbers, nodule size and nitrogenase activity of soybean. Two phosphate transporters, GmPT5 and GmPT7, regulated phosphate transport and in turn affected nodulation in soybean (Chen et al., 2019; Qin et al., 2012). GmPT5 controlled Pi entry from roots to nodules, was critical for maintaining Pi homeostasis in nodules and subsequently regulated nodulation and growth performance (Qin et al., 2012). Overexpression of GmPT7 promoted nodulation and increased plant biomass, shoot nitrogen and phosphorus contents, improving soybean yield by up to 36% (Chen et al., 2019). The proteins phosphate‐transporter 1 (PHT1) and its regulator phosphate–starvation–response 1 (PHR1) worked as a PHR1‐PHT1 module to maintain Pi homeostasis and affected nodule development (Lu et al., 2020a).

Hormones have long been known to control nodule organogenesis (Grunewald et al., 2009). The transcription factors BRI1‐EMS suppressor 1 (BES1)/brassinazole‐resistant 1 (BZR1) played key roles in the brassinosteroids (BRs) signalling pathway (Yan et al., 2018). GmBEHL1 was identified as an Arabidopsis BES1/BZR1 homolog and can interact with NODULE NUMBER CONTROL 1 (NNC1), a transcriptional repressor that mediates soybean nodulation (Wang et al., 2014c; Yan et al., 2018). Knockdown and overexpression of GmBEHL1 increased and decreased the number of nodules, respectively (Yan et al., 2018). The strigolactone (SL) biosynthesis enzymes GmMAX1a and GmMAX4a were apparently regulated by rhizobia infection. GmMAX1a and GmMAX4a knockdown lines exhibited decreased nodule number (Rehman et al., 2018) (Figure 3). GmMAX2 interacted with D14 and KAI to influence the SL and karrikins (KARs) signalling pathways to affect soybean root nodulation (Ahmad et al., 2020). Among the YUCCA (YUC) gene family, GmYUC2a functioned as an important regulator of auxin biosynthesis to modulate nodulation (Wang et al., 2019f). The β‐carotene hydroxylase GmBCH catalysed the conversion of β‐carotene to β‐zeaxanthin, which was related to the ABA synthesis pathway. RNAi‐mediated silencing of GmBCH1/2 impaired nodule development and symbiotic nitrogen fixation (Kim et al., 2013).

It was also found that quite a few miRNA families showed transcript‐level responses to nodulation (Jin et al., 2018; Yan et al., 2015; Yan et al., 2016). For example, miR393j‐3p was significantly up‐regulated during nodule formation, and ectopic expression of miR393j‐3p significantly reduced nodule formation (Yan et al., 2015). The function of miR393 may be through regulation of GmTIR1/GmAFB3‐based auxin perception (Cai et al., 2017). miR172 is another important miRNA that was essential for nodule development. Overexpression of MIR172 or the miRNA‐encoded peptide miPEP172c both resulted in an increase in nodule numbers in transgenic soybean roots (Couzigou et al., 2016; Yan et al., 2013). Further investigation showed that the function of miR172c was through the GmNINa‐miR172c‐NNC1 regulatory module (Wang et al., 2019c; Wang et al., 2014c) (Figure 3). MIR160 promoted auxin activity by suppressing the levels of the ARF10/16/17 transcription factors to direct proper nodule formation and maturation in soybean (Nizampatnam et al., 2015; Turner et al., 2013). The expression levels of miR167 and its target were up‐ and down‐regulated by auxin in soybean. Moreover, miR167 could positively regulate nodule numbers by repressing the target genes GmARF8a and GmARF8b, which were homologous genes of the Arabidopsis auxin response factor AtARF8 (Wang et al., 2015c) (Figure 3). In addition, mis‐expression of miR482, miR1512 and miR1515 increased nodulation (Li et al., 2010). Recently, Ren et al. showed that small RNA fragments (tRFs) derived from rhizobial transfer RNA (tRNA) serve as signal molecules that regulate host nodulation. Three families of rhizobial tRFs (Bj‐tRF001, Bj‐tRF002 and Bj‐tRF003) were confirmed to regulate host genes associated with nodule initiation and development (Ren et al., 2019), which represented a new evidence of a root‐shoot‐root signalling mechanism during nodulation (Shine et al., 2019; Zhang et al., 2020a).

The soybean gene early nodulin 40 (ENOD40) played a pivotal role in nodule organogenesis (Charon et al., 1999; Kumagai et al., 2006; Wan et al., 2007). NNC1 regulated the expression of ENOD40 by binding to the AP2 cis‐elements of ENOD40 promoter, which consequently represses ENOD40 expression and negatively regulated nodulation (Wang et al., 2014c). Several other genes that could affect soybean nodulation were also identified, such as FW2.2‐like 1 (GmFWL1) (Libault et al., 2010), LysM‐type receptor kinase (GmNFR1alpha) (Indrasumunar et al., 2011), ecto‐apyrase (GS52) (Tanaka et al., 2011), ureide transporter (UPS1) (Collier and Tegeder, 2012), symbiotic ammonium transporter 1 (SAT1) (Chiasson et al., 2014), GmEXPB2 (Li et al., 2015a) and VAMP721a and VAMP721d (Gavrin et al., 2016). The transcription of the gene target of rapamycin (GmTOR) and its key downstream effector, GmS6K1, were activated during nodulation. When GmS6K1 was knocked down, nodule development was severely impaired, suggesting an important role for the rapamycin pathway in nodule development (Um et al., 2013). GmVTL1a, which function as a transporter of ferrous iron from the infected root cell cytosol to the symbiosome, moved iron across the symbiosome membrane to bacteria’s and played a crucial role in nitrogen fixation (Brear et al., 2020; Liu et al., 2020d). In addition, nodulation was affected under acidic and drought conditions (Gil‐Quintana et al., 2013; Lin et al., 2012). A very recent study reported that Nodule Number Locus 1 (GmNNL1), which encodes a novel R protein, may trigger immunity and inhibit nodulation (Zhang et al., 2021).

Domestication

Plant domestication is one of the most important aspects contributing to the development of agriculture (Diamond, 2002). In addition to investigations at the population level through the resequencing of germplasm (Lam et al., 2010; Li et al., 2013b; Sedivy et al., 2017; Zhou et al., 2015), some genes responsible for traits linked to soybean domestication were identified.

One important agronomic trait that was targeted by human selection during crop domestication is decreased pod shattering and seed dispersal (Sedivy et al., 2017). SHAT1‐5, a gene encoding a NAC (NAM, ATAF1/2 and CUC2) transcription factor, was found to be a prime domestication gene, and the allele in cultivated soybean improves the thickening of the fibre cap cells and suppresses pod shattering (Dong et al., 2014). Pdh1 is another gene that affects the pod shattering phenotype and showed artificial selection in landraces of Japan, Korea and other South‐East Asian countries. The cultivated alleles promoted torsion of dried pods under low humidity, causing higher pod dehiscence (Funatsuki et al., 2014).

Loss of dormancy is another important domestication trait and relates to both physiology and physical, structural changes (Finch‐Savage and Leubner‐Metzger, 2006). In soybean, Hs1‐1 has long been identified as an important locus corresponding to loss of dormancy through a physical change (Liu et al., 2007). Sun et al. performed a genetic analysis and suggested that Hs1‐1 encodes a calcineurin‐like metallophosphoesterase transmembrane protein (Sun et al., 2015a), while Jang et al. reported that a single nucleotide polymorphism in another gene, a endo‐1,4‐β‐glucanase, in this region may be responsible for Hs1‐1 (Jang et al., 2015). Another key locus responsible for seed dormancy is G, which was found to encode a CAAX amino‐terminal protease protein. Interestingly, G had undergone parallel selection across different plant families (Wang et al., 2018b), which may make it a candidate gene for the acceleration of crop improvement (Lyu, 2018; Rendon‐Anaya and Herrera‐Estrella, 2018; Wei and Huang, 2018).

As a short‐day flowering plant, the spread of soybean cultivation latitudinally requires adaptation to new photoperiods. Several flowering‐related genes in soybean have been selected during human cultivation (Cober et al., 2010; Kim et al., 2012a). GmCRY1a and GmCOL7a, soybean homologs of Arabidopsis CRYPTOCHROME 2 (CRY2) and CONSTANS (CO), respectively, were reported to exhibit strong selection signatures (Li et al., 2013b; Wang et al., 2016c; Wu et al., 2014; Zhang et al., 2008). The major maturity loci E1, E2, E3 and E4, where several flowering‐associated genes are located, have contributed to local adaptation (Kanazawa et al., 2009; Liu et al., 2008; Wang et al., 2016c; Watanabe et al., 2009; Watanabe et al., 2011; Xia et al., 2012; Zhai et al., 2014b). A recent study found that homologous pseudo‐response‐regulator (PRR), Tof11 and Tof12, had undergone strong selection (Lu et al., 2020b). Several yield‐related genes, such as GA20OX, NFYA (Lu et al., 2016), Dt1 (Tian et al., 2010), SoyWRKY15a (Gu et al., 2017) and PP2C‐1 (Lu et al., 2017b), had also undergone selection during soybean domestication.

Transformation technology

Transgenic technology

Soybean transgenic technology is a necessary tool for soybean gene function study. Genetic transformation of soybean has been studied for over two decades; however, the progress remains slow and inefficient, which is why the functional validation in some studies was performed in Arabidopsis instead of soybean. Several transformation systems were developed, including using shoot meristems (Mccabe et al., 1988; Rech et al., 2008), hypocotyls (Aragão et al., 2000; Dan and Reichert, 1999; Wang and Xu, 2008), embryo (Finer and Mcmullen, 1991; Trick and Finer, 1998), immature cotyledons, half‐seed explants (Liu et al., 2004; Paz et al., 2006) and cotyledonary nodes (Li et al., 2017c; Liu et al., 2004; Sato et al., 1993). Considering the operation, reproducibility, copy number of foreign DNA and experimental cost, Agrobacterium‐mediated cotyledonary node (CN) soybean transformation was commonly used nowadays (Hinchee et al., 1988; Paz et al., 2006; Somers et al., 2003). The overall average transformation efficiency was 3.8%–8.7% (Li et al., 2017c; Paz et al., 2006). Recently, Pareddy et al. (2020) enhanced the average transformation efficiency to 18.7%. But it is still lower than that in rice of 23% (Ge et al., 2006; Lin and Zhang, 2005) and maize of over 30% (Ishida et al., 1996; Yang et al., 2006).

The general transformation process includes seed sterilization and germination; Agrobacterium infection; co‐cultivate soybean explants and Agrobacterium; shoot induction; shoot elongation; rooting; and at last transferring the plants to pots containing soil (Figure 4). In these processes, many factors can affect the efficiency. The first effector is soybean genotypes. Song et al. compared transformation efficiency and regeneration rate of twenty soybean varieties and found that transformation efficiency between different varieties differed significantly (0.31%–4.59%) (Song et al., 2013c). Secondly, in the process of Agrobacterium infection which is one of the most important steps, all of Agrobacterium concentrations, soybean explant status, Agrobacterium suspension medium and co‐cultivation time will affect the infection efficiency. Another key process in determining the transformation efficiency is explant regeneration. It was reported that plant hormone plays critical role in inducing the regeneration of explant and its suitable concentration could improve the efficiency (Li et al., 2017c). Chen et al. (2018a) reported that adding L‐glutamine and L‐asparagine together into culture media will increase the transformation efficiency by suppressing the expression of GmPRs.

Figure 4

General procedure of Agrobacterium‐mediated cotyledonary node soybean transformation. (a) Seed sterilization. Selection of soybean varieties dominates the efficiency of transformation affecting on Agrobacterium infection and regeneration capacity (Song et al., 2013c). (b) Seed imbibition. (c) Preparation of Agrobacterium. (d) Agrobacterium infection. It is one of the most important steps, Agrobacterium concentrations, soybean explants, Agrobacterium suspension medium, and co‐cultivation time will affect the infection efficiency. (e) Co‐cultivation. Suitable concentration of plant hormone is necessary to (f) Shoot induction and (g) shoot elongation that will improve the efficiency (Li et al., 2017c). Adding L‐glutamine and L‐asparagine together into culture media will increase the transformation frequency of soybean by suppressed the expression level of GmPRs (Chen et al., 2018a). (h) Rooting. (i) Transplanting. Words in the blue boxes indicate the methods of improving infection efficiency.

Genome editing

Genome editing can introduce precise modifications into the genome to obtain predictable and desired traits, which has been proved to be a powerful approach for functional study and molecular design breeding than the traditional genetics approaches, such as mutagenesis, transgenic RNAi or overexpression (Gao, 2021; Rodriguez‐Leal et al., 2017). Of the different genome editing systems, CRISPR (clustered regularly interspaced short palindromic repeat)/Cas (CRISPR‐associated) system shows high efficiency and has been extensive applied in different species (Feng et al., 2013; Jiang et al., 2013; Miao et al., 2013; Shan et al., 2013; Svitashev et al., 2015; Upadhyay et al., 2013).

The first knock out and DNA homology‐directed recombination (HDR) soybean plant created by CRISPR/Cas9 technology was successfully obtained in 2015 (Li et al., 2015b). In 2016, Du et al. (2016) found that changing the AtU6‐26 promoter into GmU6‐16g‐1 promoter of the CRISPR/Cas9 system could significantly improve the efficiency of targeted mutagenesis in soybean. In soybean, nearly 75% of the genes present in multiple copies and knockout of a single gene usually does not exhibit mutant phenotype. It is important to develop a dedicated CRISPR/Cas9 system that can edit multiple homologous genes simultaneously. By optimizing the steps of vector construction, sgRNA assessment, pooled transformation and sgRNA identification, a CRISPR/Cas9 system that can generate multiplex mutagenesis with higher efficiency was developed (Bai et al., 2020). In nature, beside the alleles caused by loss‐of‐function mutations, large part of the phenotypic variations in agronomic traits are resulted from single nucleotide polymorphism (SNP) variations. Damage of the function of the whole gene using gene‐editing system usually leads to severe phenotype, which may not be optimizable for agronomic trait improvement in production. Therefore, generation of point mutations at specific sites affecting important agronomic traits is of great value in molecular breeding (Mishra et al., 2020). Recently, ‘base editing’ has been developed from CRISPR/Cas9 system, which converts single base into another without requiring DNA double‐strand breaks or a donor template (Komor et al., 2016). Cai et al. successful applied the system to create point mutant of GmFT2a and GmFT4 (Cai et al., 2020a).

Nowadays, CRISPR/Cas9 was widely applied in soybean functional studies (Cai et al., 2018; Du et al., 2016; Li et al., 2019a; Michno et al., 2015; Sun et al., 2015b; Xu et al., 2020). For instance in identification of the genes controlling flowering time, frameshift mutations generated by CRISPR/Cas9 demonstrated that GmFT2a mainly function under short day (SD), whereas GmFT5a had more significant effects under long day (LD) (Cai et al., 2018; Cai et al., 2020b). Similarly, knockout of GmPRR37 by the CRISPR system suggested that it can repress flowering under LD (Wang et al., 2020c). Two CRISPR/Cas9 gene‐editing mutants of Glyma.13G114200 exhibited male sterility phenotype, confirming that it was the casual gene for GmMS1 for male sterility (Fang et al., 2021b; Jiang et al., 2021; Nadeem et al., 2021). CRISPR/Cas9 was also be applied in yield and seed quality‐related trait modifications, such as to alter plant architecture by editing GmLHYs (Cheng et al., 2019) or SPL9 (Bao et al., 2019; Cai et al., 2018; Cheng et al., 2019), to increase seed number per pod by editing Ln (Cai et al., 2021), to reduce beany flavour by knocking out LOXs (Wang et al., 2020b), to increase isoflavone content by editing GmF3H1, GmF3H2 and GmFNSII‐1 simultaneously (Zhang et al., 2020f) and to alter the fatty acid profiling by editing FAD2‐2 (Al Amin et al., 2019). In the future, more application of ‘base editing’ for single gene or for multiple genes simultaneously will greatly promote the functional study and molecular design breeding in soybean.

Challenges and future perspectives

The Green Revolution is one of the most remarkable events in agriculture and greatly increased the production of major crops (Hedden, 2003). However, few improvements in yield have been made for soybean over the past six decades. In order to meet the needs of a growing world population, soybean yield must increase at a faster rate than it is at present (Ainsworth et al., 2012; Ray et al., 2013). There is an urgent need for a soybean ‘Green Revolution’ to breed supper varieties with the ideal plant architecture that are adapted to high‐density planting environment (Liu et al., 2020e). At this point, most soybean production occurs in South America, North America and Asia. In the future, Africa might become another dominant soybean production area. Therefore, there is a need to start studies on genetic improvement and production technologies for an Africa‐adapted soybean. Moreover, there is a need to develop new varieties to meet future environmental changes and to create a more sustainable agricultural system (Bishop et al., 2015; Kumagai et al., 2015; Mourtzinis et al., 2015; Ruiz‐Vera et al., 2013; Tian et al., 2020).

To create a super variety, breeders normally need to stack multiple desirable traits into a single line. However, most important traits are quantitatively controlled and exhibit inherited correlations. Bringing about a ‘Green Revolution’ in soybean may prove to be an exacting task because of the unique plant architecture and the complicated components that determine the final yield in soybean (Liu et al., 2020e). Understanding the inherited mechanism of each trait and the regulatory network among different traits will help us to design the desired crops (Tian et al., 2020). A dissection of the genetic networks underlying 84 agronomical traits has provided insights into the molecular design of soybean (Fang et al., 2017).

De novo domestication of new crops aims to rapidly domesticate wild or semi‐wild plants into agricultural crops with favoured agronomic traits by utilization of combines modern technologies, including genomics, gene editing and synthetic biology (Fernie and Yan, 2019; Khan et al., 2019; Li et al., 2018b; Yu et al., 2021; Zsogon et al., 2018). Wild soybean showed higher protein content, lower oil content and higher stress tolerance, whereas most of the current cultivated soybeans exhibited relatively lower protein content and higher oil content. So far, not many genes related to soybean domestication has been identified, leaving the domestication traits are still poorly understood. For instance, switching plant architecture from twinning (sprawl) habit in wild soybeans to upright habit in cultivated soybeans is one of the most prerequisite domestication trait for soybean; however, a genetics dissection of this important trait is unclear yet. With the rapid development of functional genomics in soybean, an ever‐increasing number of genes related to agronomic traits are being cloned, which may enable us to re‐domesticate the wild soybean into a new crop keeping the characters of higher protein content and higher stress tolerance.

At this time, soybean functional studies and breeding still face some technical challenges. The lack of a stable and high‐efficiency transgenic system is one of the biggest challenges, which makes functional study more time‐consuming than in species with well‐developed transformation systems, such as rice and Arabidopsis. Another challenge is phenotyping. Because soybean is quite sensitive to photoperiod, the phenotypes of a soybean line usually exhibit significant variations in different environments, which make precise phenotyping and functional study more difficult. The recent development of new technologies, such as target base editing (Cai et al., 2020a) and a transient expression system (Xiong et al., 2019), will facilitate soybean functional studies. Moreover, the recently developed graph‐based soybean pan‐genome will both rejuvenate previous omics data and revolutionize functional and evolutionary genomic studies in soybean (Liu et al., 2020f; Liu and Tian, 2020).

Conflict interest

We declare that we have no conflict of interest.

Author contributions

M.Z., S.L., Z.W., Y.Y., Z.Z., Q.L., X.Y., Z.D., F.K., B.L., B.R. and Z.T. drafted the manuscript. Z.T. conceived the article and revised the manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (31801380), the National Key Research & Development Program of China (2017YFD0101305), and the State Key Laboratory of Plant Cell and Chromosome Engineering (PCCE‐KF‐2019‐05).

Table S1 Representative genes related to agronomically important traits in soybean.

Click here for additional data file.