2026 Volume 76 Issue 1 Pages 27-47
Following the recent global progress in the establishment of soybean genomic resources, DNA marker technologies have been actively implemented in soybean breeding programs in Japan, thereby enhancing the efficiency of varietal improvement and facilitating the successive release of cultivars developed through the application of DNA markers. In particular, DNA markers developed for the selection of useful traits, such as pod-shattering resistance and bacterial leaf pustule resistance present in foreign germplasms, have facilitated the precise introgression of only the desired alleles from genetically divergent foreign germplasms into Japanese backgrounds. By increasing selection efficiency and shortening breeding cycles, these advances have substantially contributed to the recent improvements in Japanese soybean breeding. The present review consolidates findings on genetic variations identified in Japanese cultivars and breeding materials, which have been investigated for the development of such DNA markers, and from knowledge expected to contribute to future soybean breeding in Japan.
Owing to its high protein and oil content, soybean [Glycine max (L.) Merr.] is the most important legume worldwide for food and industrial materials. It is also an important source for the production of traditional Japanese foods such as tofu, natto, miso, boiled beans, soy sauce, and edamame. In 2021, the demand for soybeans in Japan was approximately 3,540,000 tons, and that for related food use was approximately 1,030,000 tons per year (MAFF 2023). Therefore, almost all soybean produce, approximately 252,000 tons produced in Japan, is used for food, accounting for 24% of the soybean demand. Considering food security and the global protein crisis, the Ministry of Agriculture, Forestry, and Fisheries (MAFF) of Japan set an ambitious government target for domestic soybean production: 390,000 tons by 2030 in its Basic Plan for Food, Agriculture and Rural Areas formulated in 2025 (MAFF 2025). Assuming the current yield levels (161 kg/10a; average yield from 2019 to 2023), increasing soybean production by approximately 130,000 tons within the next five years would require an expansion of the current cultivation area (approximately 154,700 ha, MAFF 2023) by approximately 86,000 ha, which is extremely challenging. Therefore, enhancing cultivation management practices and improving soybean yield potential to increase production per unit area are essential. Furthermore, effectively incorporating seed quality traits and compositional factors identified in previous studies into domestic soybean breeding programs is crucial. Genetic control of seed appearance, quality, and composition leads to improvement in seed grade under the official agricultural inspection and adds value in terms of taste and functionality. These positive effects can lead to higher market prices and, ultimately, enhanced profitability for producers, thereby encouraging increased motivation for planting. This review summarizes the contributions of DNA markers and genomic information to addressing these challenges through breeding, and explores their potential future applications.
The most popular cross-breeding methods in Japan are mass breeding (Fig. 1). Artificial hybridization is carried out with the number of crosses ranging from approximately 30 to 50 per year. MAS is sometime applied to F1 seeds to verify successful cross-hybridization. The size of the F2–F3 populations is approximately 2,000 plants per cross. In single-seed descent, the multiple-seed procedure is often employed: one to two pods are harvested from each plant, and the pods are bulked together and threshed. Individual selection is generally conducted in the F3 or F4 generation, with field selection based on traits such as lodging resistance, plant architecture, maturity, and number of pods. After harvest, promising plants are further selected based on seed size, external seed quality, incidence of defective seeds, and protein content. Pedigree selections start from the F4 or F5 generation, and the total number of lines across all cross combinations in the F5 generation ranges from 500 to 2,000. MAS is applied at these stages: individual selection from the population, mainly the F4 generation, and first year of line cultivation, mainly the F5 generation. In early generations such as F2 and F3, typically 1,000 to 2,000 individuals are cultivated per cross combination. Because of the large population size, both sampling and genotyping are labor-intensive and inefficient; consequently, MAS is rarely implemented at this stage.
The conventional soybean breeding process and MAS in Japanese public institutions.
From the F6 generation onward, five plants are harvested individually from each selected line resulting in five lines that comprise the next generation’s line group. Yield trials are also conducted in parallel with line selection. The preliminary yield trials are conducted from the F5 to F7 generation, whereas advanced yield trials and multiple regional tests are initiated in the F8 generation and conducted in subsequent generations. Applications for plant variety registration and the production of breeder’s seed are most often carried out from the F10 generation onward. For plant variety registration, approximately one variety is commonly developed every two to three years at each breeding location.
Traditional MAS methods primarily rely on polyacrylamide or agarose gel electrophoresis. However, as described above, the need to process a large number of samples annually has led to the increasing adoption of more efficient technologies such as capillary electrophoresis systems and DNA sequencers. In particular, fragment analysis using DNA sequencing enables the simultaneous identification of multiple PCR products in a single reaction using primers labeled with four distinct fluorescent dyes, which significantly enhances throughput and efficiency (Kato et al. 2017, Sayama et al. 2011, Watanabe et al. 2018). However, the use of such advanced technologies involves substantial operational costs, particularly related to equipment maintenance and the employment of specialized technicians. In addition, given the limited manpower available to breeders, performing MAS in addition to core breeding activities is often impractical, both in terms of time and labor. In this context, the division of labor in MAS should become increasingly common, with specialized teams or units handling genotyping tasks separately from field breeding.
Japanese soybeans have for a long time been subjected to selection for use as food and exhibit different characteristics from overseas soybeans. Seed characteristics for food processing, particularly for tofu, natto, and boiled beans, are important targets in soybean breeding in Japan, and seed quality and high protein content have been examined thoroughly (Escamilla et al. 2019, Taira 1990). In contrast, the focus for soybean breeding in other countries, such as the United States and Canada (hereafter, North America), and Brazil, has been improving its oil content and yield (Morrison et al. 2000, Tamagno et al. 2022, Umburanas et al. 2022). The yield gap between Japan and major soybean-producing countries such as the United States and Brazil began to widen around the 1990s, and in recent years has exceeded a twofold difference (Fig. 2). Boehm Jr. et al. (2019) has been pointed out that the increase in soybean yield in the United States has been largely driven by genetic improvement, and that this yield gain has come at the expense of seed protein content.
Trends in soybean yields in the United States, Brazil, and Japan (FAOSTAT).
Although the scale of breeding differs among countries, the fundamental breeding processes are largely similar to those used in Japan. Nevertheless, the three major points of divergence are outlined below.
The first is the duration of cultivar development. As noted earlier, in Japan, it generally takes more than 10 years from the initial cross to cultivar release, whereas in North America, the process is typically completed in approximately 7 years (Vieira and Chen 2021). This acceleration is achieved by advancing the F2 and F3 generations in warm off-season nurseries located in the Southern Hemisphere, such as Chile and Puerto Rico.
The second point is the process of breeder seed production. In Japan, from approximately the F5 generation onward, repeated line selection is conducted over multiple years to increase genetic uniformity. Conversely, in North America, the process is relatively streamlined: promising plants are first selected from the population in yield trials. Seeds from each selected plant are then multiplied in bulk for each line while confirming that the resulting line is genetically uniform. The resulting line-level seeds are sown, and the genetic uniformity of each family is re-evaluated. If no uniformity issues are detected, all the families are harvested in bulk to produce breeder seeds. Given Japan’s stringent variety registration system, the bulk method for breeder seed production is not easily applicable. However, the North American system allows for a smoother transition from breeding to seed production, and, ultimately, to seed delivery for growers.
The third point concerns the yield evaluation methods. In North America, mechanical harvesting using plot combines has long been a standard practice, whereas in Japan, yield testing has traditionally been dependent on manual harvesting. Therefore, the methodological difference between yield assessment in breeding facilities and actual farmers’ field harvesting practices may have contributed to the yield gap observed between the Japanese and North American cultivars. In Japan, mechanical harvest-based yield evaluation has historically lagged, largely because of the predominance of paddy-field conversion farmlands and limited field size. However, with the urgent need to develop high-yield cultivars, mechanized yield testing has recently been introduced. Indeed, cultivars such as ‘Sorahibiki’ (Shimamura et al. 2025), ‘Soramizuki’ (Kato et al. 2023), ‘Soratakaku’ (Sayama et al. 2025) and ‘Soraminori’ (Oki et al. 2025) have been released, demonstrating the effectiveness of this approach. MAS remains challenging for key agronomic quantitative traits such as yield, protein content, and lodging resistance. Consequently, breeders must continue to rely on field-based evaluations, visually assess individual soybean plants for these important traits, and make the appropriate selection decisions. Thus, it is crucial to distinguish between traits that are suitable for MAS and those that are more effectively addressed through phenotypic selection, such as yield evaluation using mechanical harvesting.
The first soybean reference sequence for the US soybean cultivar ‘Williams 82’ was released in 2010. In the latest version, approximately 53,000 genes were predicted across roughly 978 million base pairs (Schmutz et al. 2010). Since then, high-quality reference genomes have been successively published from various countries, including the US cultivar ‘Lee’, a Chinese wild soybean (Valliyodan et al. 2019), the Chinese cultivar ‘Zhonghuang 13’ and wild soybeans (Shen et al. 2019), 23 cultivated and 3 wild soybeans (Liu et al. 2020), the Korean cultivar ‘Hwangkeum’ (Kim et al. 2021), the Chinese cultivar ‘Jidou 17’ (Yi et al. 2022), the Chinese vegetable soybean cultivar ‘Zhenong 6’ (Liu et al. 2022), telomere-to-telomere gap-less genome assemblies of the Chinese cultivar ‘Yundou1’, and wild soybean ‘Yesheng71’ (Jia et al. 2024). Recently, Yano et al. (2025) constructed a nanopore-based genome reference for seven Japanese and four overseas soybean cultivars. Notably, a high-quality reference genome with 58,646 genes was constructed for the cultivar ‘Enrei’, which serves as the reference standard for domestically produced soybeans. Although the availability of precise genomic information has advanced significantly, the number of genes with clearly defined functions remains limited. Functional annotations have been assigned to numerous soybean genes; however, most of them are based on predictions derived from homologous genes in other organisms and model plant species. Agronomically and industrially important traits such as yield and seed quality are regulated by complex interactions among numerous genes, further limiting the number of genes whose functions have been definitively characterized.
Agriculturally important genes are difficult to directly identify; even so, their approximate chromosomal locations can be inferred by mapping techniques using DNA markers and quantitative trait locus (QTL) analysis. These approaches allow for MAS of desirable traits without requiring full elucidation of the underlying genes. SoyBase (https://legacy.soybase.org/), developed by USDA, is the world’s largest integrated database for soybean breeding information. It provides searchable chromosomal location data for 7,246 loci and QTLs associated with 909 traits, thereby considerably facilitating the design of precise DNA markers. High-quality DNA markers capable of reliably detecting the presence or absence of beneficial traits have significantly contributed to selection efficiency and shortening of breeding cycles. In recent years, DNA markers have played a crucial role in the targeted improvement of soybean cultivars in Japan; however, in many cases, their use in cultivar development has not been thoroughly documented, and comprehensive reports on the application of MAS in Japanese soybean breeding remain scarce. The following sections present the identified genes, as summarized in Supplemental Table 1, and underlying mechanisms for each trait as well as the DNA markers and cultivars developed through their application in MAS.
Soybeans are cultivated across the Japanese archipelago, which spans a wide range of latitudes. Therefore, selecting cultivars that have adapted to the specific photoperiod and temperature conditions of each region is essential. In northern Japan, where snowfall occurs, the ability of a cultivar to mature before the onset of snow is a particularly critical trait. Moreover, the duration of the vegetative growth phase prior to floral induction significantly influences regional adaptability as well as plant architecture, biomass accumulation, and ultimately yield. Since the 2000s, the availability of genomic information has enabled elucidation of the molecular mechanisms underlying approximately 18 major genes related to flowering and maturity.
Among these, the E1 gene is known to have the most substantial effect on flowering time. It encodes a transcription factor with a B3 domain and is considered to be a legume-specific gene (Xia et al. 2012), absent in model plants such as rice and Arabidopsis. The expression of E1 is regulated by the ELF gene (Lu et al. 2017, classical J locus) and the LHY gene (Dong et al. 2021), which in turn control the downstream florigen genes GmFT2a (Kong et al. 2010) and GmFT5a (Takeshima et al. 2016). Xia et al. (2012) reported multiple alleles at the E1 locus, including e1-nl (a complete deletion resulting in the loss of photoperiod sensitivity), e1-fs (a frameshift mutation), and e1-as (a missense mutation with reduced function). Cultivars lacking all four dominant alleles—E1, E2, E3, and E4—have been identified in Hokkaido. Additionally, homologs of E1, namely E1La (Xu et al. 2015) and E1Lb (Zhu et al. 2019), participate in the regulation of florigen gene expression.
The E2 gene corresponds to the GIGANTEA gene, a signal transduction factor in the circadian clock (Watanabe et al. 2011). Loss-of-function mutations in E2 result in early flowering, and such alleles are commonly found in cultivars from Honshu. The E3 and E4 genes encode PHYA3 (Watanabe et al. 2009) and PHYA2 (Liu et al. 2008), respectively. Loss-of-function mutations lead to early flowering. Cultivars adapted to high-latitude regions, such as Tohoku and Hokkaido, often carry mutations in E4, whereas loss-of-function alleles of E3 are frequently observed in Honshu cultivars; in addition, DNA markers for E2, E3 and E4 have been developed (Tsubokura et al. 2014).
Tof11 and Tof12 encode pseudo-response regulators (PRRs) that mediate signal transduction in response to environmental light cues (Lu et al. 2020). Most Japanese soybean cultivars possess a non-functional Tof12 allele and a functional Tof11 allele. In contrast, some foreign cultivars carry non-functional Tof11 alleles, which affect flowering time and shorten the reproductive growth phase.
Florigen genes GmFT2a and GmFT5a promote floral induction under short-day conditions. The maturity gene E9 is FT2a, and its recessive allele delays flowering because of its lower transcript abundance, which is caused by allele-specific transcriptional repression owing to the insertion of SORE-1 (Zhao et al. 2016). Regardless of the day length, the e9 allele delayed flowering. Hokkaido cultivars carry this mutation.
The J gene (ELF gene) is an allele found in cultivars adapted to low-latitude regions that delay flowering under short-day conditions (Lu et al. 2017). Under these conditions, the J gene suppresses E1 expression, allowing for the expression of florigen genes and floral induction. When the J gene loses its function, E1 expression is no longer suppressed under short-day conditions, resulting in delayed flowering. The LUX gene, which forms a complex with the J gene, has been identified as a component of the Evening Complex in soybean (Bu et al. 2021).
As noted above, flowering and maturity are controlled by multiple genes; thus, materials exhibiting similar phenotypes may nonetheless possess different genetic compositions. However, the genes described above allow a substantial degree of predictive inference, and this information will facilitate detailed genotyping of breeding materials in the future. Consequently, if crossing combinations that are unlikely to exhibit substantial segregation for maturity in their progeny can be pre-selected, efficient yield evaluations can be conducted while minimizing the confounding effects of maturity variation.
Soybeans exhibit three distinct growth habits: determinate types, which cease stem elongation upon floral induction; indeterminate types, which continue stem elongation and leaf formation after flowering; and semi-determinate types, which show intermediate stem termination between determinate and indeterminate types (Bernard 1972). Determinate growth is common in Japanese cultivars, whereas indeterminate and semi-determinate growth are more prevalent in overseas varieties. The genetic background and environment determine whether indeterminate phenotypes have a higher yield than determinate phenotypes (Cober and Morrison 2010, Kato et al. 2015). Most soybean cultivars developed in high-latitude countries have an indeterminate growth habit; however, this is not found in any of the current commercial Japanese cultivars. ‘OAC Dorado’, a Canadian cultivar with an indeterminate growth habit, produced significantly greater yield than Hokkaido’s leading cultivar ‘Yukihomare’ in fields in Hokkaido, indicating that it may be possible to breed a high-yielding cultivar with an indeterminate growth habit (Yamaguchi et al. 2019b).
Growth habit is regulated by the interaction between the Dt1 gene, which encodes the TFL1 protein that suppresses floral meristem formation (Tian et al. 2010), and the Dt2 gene, which encodes a MADS-box transcription factor that binds to the promoter region of Dt1 and acts as a repressor (Ping et al. 2014). Combinations of these alleles determine the growth habit: dt1dt2 and dt1Dt2 result in determinate growth, whereas Dt1Dt2 and Dt1dt2 lead to semi-determinate and indeterminate growth habits, respectively. The MAS for the Dt1 allele is conducted in breeding programs mainly in Hokkaido.
Soybeans develop inflorescence meristems on each node, including the shoot apex. Raceme phenotype is an important trait associated with the number of floral buds and pods. qTRL18-1, a QTL associated with terminal raceme length, was detected in the proximal region of the Dt2 locus (Yamaguchi et al. 2014a). Another QTL, qTRL11-1, was detected only in the qTRL18-1 background, indicating an interaction between qTRL18-1 and qTRL11-1. ‘Tokei 1122’, which has a long terminal raceme with qTRL18-1 and qTRL11-1, was bred to improve the yield of low-branching cultivars under dense planting conditions. In the yield tests under dense planting conditions (33 plants m–2), the seed yield of ‘Tokei 1122’ was 34.8 kg/a, whereas ‘Toyoharuka’, a low-branching cultivar bred in Hokkaido, yielded 29.6 kg/a, representing an approximately 18% increase (Kitabatake et al. 2020).
Lodging resistance is an important trait in soybean breeding because lodging significantly reduces cultivation management operations, mechanical harvest efficiency, light-use efficiency, soybean yield, and quality when plants fall over or lean excessively in the field due to typhoons and storms with heavy rain and wind (Saitoh et al. 2012). In Japan, poor drainage in converted paddy fields often lead to lodging after frequent rainfall and strong winds, including typhoons. Breeding for lodging resistance helps ensure crop growth stability under adverse weather conditions, which is increasingly important under climate variability.
Most of the investigated QTLs for lodging resistance in Japanese cultivars is located in the proximal region of the maturity or growth habit locus, Dt1. qLS19-1 was identified using a population derived from a cross between ‘Toyoharuka’ (a lodging-tolerant cultivar) and ‘Toyomusume’ bred in Hokkaido (Yamaguchi et al. 2014b). The Toyoharuka allele at qLS19-1 increases the number of primary lateral roots and causes high pushing resistance, resulting in a lower ratio of the pushing resistance moment (Kitabatake et al. 2019). Eight QTLs associated with seed yield were identified using a population derived from a cross between ‘Toyoharuka’ and ‘Toyomusume’ (Yamaguchi et al. 2021a). Six breeding lines pyramiding favorable alleles at the seven QTLs associated with seed yield and qLS19-1 were developed from the same population, and their seed yields tended to exceed those of their parental cultivars (Yamaguchi et al. 2021b). ‘Toyomadoka’ is a lodging tolerant cultivar bred in Hokkaido, and has ‘Toyoharuka’ in its pedigree (Kobayashi et al. 2020). The marker analysis revealed that ‘Toyomadoka’ has the ‘Toyoharuka’ allele at qLS19-1, indicating that MAS for the ‘Toyoharuka’ allele at qLS19-1 will be effective for improving lodging tolerance. Backcross breeding introducing the ‘Toyoharuka’ allele at qLS19-1 to easily lodging cultivars is in progress, and the effects of qLS19-1 will be validated in various genetic backgrounds in future.
A major QTL, designated qSI13-1 on chromosome 13, was founded from a Japanese germplasm ‘Y2’ exhibiting significantly shorter internode lengths compared to modern Japanese cultivars (Oki et al. 2018). This QTL confers shorter plant stature and internode length without affecting the flowering time, node number, or seed yield. Advanced breeding lines developed through backcrossing using major cultivars such as ‘Fukuyutaka’ as recurrent parents have been developed, and their use in future breeding programs, as well as high-yield cultivation practices, is being considered. A robust lodging tolerance QTL, qLT13-1, was identified on chromosome 13, explaining 20% of phenotypic variance in lodging angle across different environments (Hishinuma et al. 2025). US cultivars-derived alleles at this locus consistently improve lodging tolerance under multiple conditions. Recently, an increased copy number of two gibberellin 2-oxidase 8 genes (GmGA2ox8A and GmGA2ox8B) in a close genomic region on chromosome 13 was identified as the gene responsible for reduced trailing growth and shoot length (Wang et al. 2021a). An increased copy number reduces gibberellin activity, resulting in shorter plant height and contributing to lodging resistance. Another gene, PH13, which regulates main stem length, was isolated from different genomic regions on chromosome 13 (Qin et al. 2023). This gene encodes a suppressor of the phyA-105 (SPA) protein, which suppresses photomorphogenesis as part of a complex with COP1, and truncated PH13 with reduced interaction with COP1, resulting in the accumulation of STF and reduced plant height. These genes found on chromosome 13 are likely to be associated with the lodging resistance QTL mentioned above, so it is necessary to verify the selection effect on lodging using DNA markers designed around these genes.
Pod-shattering resistance is a critical agronomic trait for enhancing soybean yield because it prevents seed loss resulting from natural pod dehiscence, particularly under conditions of delayed harvest (Funatsuki et al. 2008, Yamada et al. 2009). This trait is especially beneficial for mechanical harvesting as it minimizes seed loss during operations and contributes to yield stability (Ndeke et al. 2024, Yamada et al. 2017). Of these, pdh1 is the most widely utilized in major soybean-producing countries such as the United States and Canada (Funatsuki et al. 2014). The PDH1 gene encodes a dirigent-like protein involved in the regulation of pod wall torsional stress, which leads to pod opening (Funatsuki et al. 2014). Although this trait provides a clear advantage by reducing seed loss at harvest, it has been underutilized in Japan, where many cultivars still lack effective pod-shattering resistance, in contrast to the high prevalence of this trait in US cultivars. Therefore, using DNA markers for pdh1, new pod-shattering resistant varieties were developed through backcrossing with elite Japanese cultivars, ‘Enrei’, ‘Kotoyutaka’, ‘Sachiyutaka’, and ‘Fukuyutaka’ as recurrent parents. As a result, the pod-shattering resistant lines ‘Enrei no Sora’ (Yamada et al. 2013), ‘Kotoyutaka A1 gou’ (Yamada et al. 2013), ‘Sachiyutaka A1 gou’ (Hajika et al. 2016), and ‘Fukuyutaka A1 gou’ (Hajika et al. 2019) were developed. Although ‘Enrei no Sora’ showed a slightly later maturity compared to ‘Enrei’, no significant differences in agronomic traits or processing qualities were observed in the other varieties. In a yield comparison trial using combine harvesting between ‘Fukuyutaka’ and its pod-shattering resistant line ‘Fukuyutaka A1 gou’, the yield of ‘Fukuyutaka’ was 127.4 kg/10a, whereas ‘Fukuyutaka A1 gou’ yielded 183.8 kg/10a, representing an approximately 44% increase (Hajika et al. 2019). This yield advantage is attributed to reduced seed loss due to shattering, as the loss of ‘Fukuyutaka’ (based on small-plot sampling) was 91.3 kg/10a, compared to only 23.2 kg/10a for ‘Fukuyutaka A1 gou’, clearly demonstrating the effect of pdh1 in reducing pod shattering. As of 2024, the cultivation areas of ‘Enrei no Sora’, ‘Kotoyutaka A1’, ‘Sachiyutaka A1’, and ‘Fukuyutaka A1’ were 3,408 ha, 2,326 ha, 1,514 ha, and 4,381 ha, respectively, and further expansion of their cultivation is expected in the future.
Phenotypic selection for pod-shattering resistance has been conducted since 1975 in Hokkaido, and some cultivars bred in Hokkaido have pdh1, including ‘Kariyutaka’ (Tanaka et al. 1993), ‘Hayahikari’ (Yumoto et al. 2000), ‘Yukihomare’ (Tanaka et al. 2003), ‘Yukihomare R’ (Suzuki et al. 2017), ‘Toyomizuki’ (Yamaguchi et al. 2023), and ‘Toyomadoka’ (Kobayashi et al. 2020). Particularly, MAS for pdh1 was conducted in the F4 generation during the breeding process of ‘Toyomadoka’.
Two other genes have been identified in relation to pod-shattering resistance in soybeans: SHAT1-5, which encodes an NAC transcription factor (Dong et al. 2014), and Sh1, which encodes a C2H2-like zinc finger transcription factor (Li et al. 2024). The functional allele of Sh1 promotes pod shattering by repressing the expression of SHAT1-5, which leads to reduced secondary wall thickness in the fiber cap cells of the abscission layer of pods.
However, 19.8% yield loss due to pod shattering has been reported, even for varieties with pdh1 and SHAT1-5 resistant alleles (Takahashi et al. 2023, Yamada et al. 2017). Generally, delays in harvesting and environmental conditions, such as high temperature and low humidity, increase the risk of pod shattering (Ndeke et al. 2024). The soybean cultivation area per farming corporation is expanding, and with ongoing climate change, the development of new genetic controls to improve pod-shattering resistance is highly anticipated.
Increasing soybean seed size is a critical trait in breeding because of cultural preferences and market demand in Japan. Larger seeds are particularly valued by food manufacturers for traditional food products such as tofu and boiled soybeans. Because seed size can be readily evaluated visually in practical breeding programs in Japan, MAS has rarely been employed, even though seed size is a key trait in domestic soybean breeding. Soybean seed size is a typical quantitative trait controlled by multiple loci with small effects, and genes controlling maturity and oil and protein content also influence seed size (Duan et al. 2023). However, only a few causal genes of natural variations related to seed size have been identified. Zhang et al. (2024) summarized the molecular regulatory network controlling seed size in soybeans, and more than 30 genes have been reported to regulate seed size. GA3ox1 encodes gibberellin 3β-hydroxylase, and reductions in individual seed size are observed in lines harboring mutations in this gene, despite the concomitant increase in seed yield driven by higher seed number (Hu et al. 2022). GmSW17 (Seed Width 17) on chromosome 17 controls seed width and weight, and encodes ubiquitin-specific protease affecting cell expansion and cell division (Liang et al. 2024). GmKIX8-1 encodes a protein containing a KIX domain that suppresses cell proliferation, and loss-of-function mutant plants show an increased seed size (Nguyen et al. 2021). Differences in the copy number of the CT-core microsatellite motif in the 5ʹ-untranslated region of WRKY15a are associated with gene expressions of this gene and seed sizes (Gu et al. 2017). To facilitate seed size control and enhance breeding efficiency in the future and enable the accurate prediction of seed size distributions in segregating progenies based on parental allele combinations, a comprehensive catalog of genes controlling seed size across a wide range of breeding materials, including varieties, breeding lines, and genetic resources will have to be compiled.
Soybean seeds accumulate approximately 40% protein, 20% lipids, and 35% carbohydrates (Cheftel et al. 1985) and are utilized as a raw material in the food and feed industries. Approximately 70% of soybean seed protein comprises two major storage proteins: glycinin (11S globulin) and β-conglycinin (7S globulin) (Hill and Breidenbach 1974, Thanh and Shibasaki 1978). A trade-off exists between glycinin and β-conglycinin contents, influencing food processing properties (Yang et al. 2016). ‘Yumeminori’ (Takahashi et al. 2004) and ‘Nagomimaru’ (Hajika 2009), which lack 7S α and αʹ subunits and show increased glycinin content, and ‘Nanahomare’, which lacks most glycinin subunits and accumulates β-conglycinin (Yagasaki et al. 2013) have been developed. Experimental lines like ‘QF2’ lack both β-conglycinin and glycinin but compensate with free amino acids (Takahashi et al. 2003). Selection of these soybean cultivars lacking seed storage proteins has relied on sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) analysis. However, selections on SDS-PAGE are labor-intensive and sometimes fail to separate glycinin subunits clearly leading to phenotyping errors due to visual judgment. Therefore, the development and application of DNA markers for these storage proteins offer a more reliable and efficient approach for stable evaluation and large-scale selection of breeding lines.
Glycinin is composed of five subunits: A1aB2, A1bB1b, A2B1a (Group I), A3B4, and A5A4B3 (Group II) (Nielsen 1985), and is encoded by five genes: Gy1 (A1aB2), Gy2 (A2B1a), Gy3 (A1bB1b), Gy4 (A5A4B3), and Gy5 (A3B4) (Nielsen et al. 1989). In addition, the presence of two pseudogenes, Gy6 and Gy8, and one gene with weak expression activity, Gy7, has been confirmed (Li and Zhang 2011). On the other hand, 7S globulin consists of three subunits: α, αʹ, and β, and were reported to be encoded by a family of 15 genes, CG-1 to CG-15. CG-1 (αʹ-subunit), CG-2 (α-subunit), CG-3 (α-subunit), and CG-4 (β-subunit) have been identified, whereas whether genes CG-5 to CG-15, which have been reported as genes, are functional is unclear (Harada et al. 1989, Singh et al. 2015). In addition, the mutation that lacks all 7S globulin subunits found in wild soybean ‘QT2’ is controlled by the dominant Scg-1 mutation (Hajika et al. 1998). Scg-1 is located on chromosome 20, and post-transcriptional silencing occurs owing to the inverted repeat sequence of the α subunit gene, suppressing the expressions of the β-conglycinin genes (Tsubokura et al. 2012). DNA markers for the selection of these genotypes have been developed (Sayama et al. 2015), and their implementation in practical breeding programs is in progress.
Protein and oil contents are negatively correlated due to competition for carbon skeletons, and increasing seed number reduces assimilate concentration per seed, creating a negative correlation between protein content and yield—major obstacles for breeding high-yield, high-protein soybeans. Guo et al. (2022) summarized QTL information for soybean seed protein content in SoyBase. In total, 249 QTLs were detected on all 20 chromosomes; however, stable and consistent QTL with considerable effects is limited. According to Patil et al. (2017), the Soybean Genetics Committee has officially confirmed cqPro-15 on chromosome 15 and cqPro-20 on chromosome 20 as stable QTLs in many populations. Causal genes for natural variations in protein content have been identified on chromosomes 15 and 20. GmSWEET10a, located on chromosome 15, encodes a sucrose transporter (Wang et al. 2020). When a mutation occurs in the promoter region, the transcription level increases, leading to increased transport of sucrose and hexose, which are carbon sources for lipids, to the seeds, thereby increasing lipid content and seed size and decreasing protein content. POWR1, located on chromosome 20, encodes a protein that regulates lipid metabolism and nutrient transport (Goettel et al. 2022). Varieties with a transposon inserted into the CCT domain show altered downstream gene expression, resulting in increased lipid content and seed size and in decreased protein content. POWR1 may have been utilized in the development of high-protein soybean cultivars in Japan. ‘Saikai 20’ (PI 423949) is known as a high-protein breeding line with a high-protein allele of POWR1 (Fliege et al. 2022), and has been used as a valuable parent in crosses, leading to the development of other promising high-protein lines and varieties such as ‘Tomutan’. These varieties were developed through selection based on the protein content measured using near-infrared spectroscopy. Although their sequences of POWR1 have not been compared yet, pedigree information suggests that they may have inherited the high-protein allele from ‘Saikai 20’. As mentioned above, the development of high-protein cultivars has thus far been dependent on phenotypic evaluation using near-infrared spectroscopy; however, these findings are expected to contribute to MAS of high-protein varieties in the future.
Soybean oil contains approximately 13% α-linolenic acid (18:3), 55% linoleic acid (18:2), 18% oleic acid (18:1), 4% stearic acid (18:0), and 10% palmitic acid (16:0) (Clemente and Cahoon 2009). To improve the oxidative stability of soybean oil and its quality during high-temperature cooking, mutational breeding has been conducted to increase the proportions of oleic acid (18:1), a monounsaturated fatty acid. Oleic acid is converted to linoleic acid (18:2) by delta-12 fatty acid desaturase 2 (FAD2), which is encoded by two genes: GmFAD2-1a and GmFAD2-1b (Anai et al. 2008). In double mutants combining GmFAD2-1a and GmFAD2-1b, the oleic acid content exceeds 80% of total fatty acids, whereas polyunsaturated fatty acids are reduced to below 10% (Hoshino et al. 2010).
The biosynthesis of α-linolenic acid (18:3) is controlled by three genes encoding omega-3 fatty acid desaturases (GmFAD3-1a, GmFAD3-1b, and GmFAD3-2a) (Anai et al. 2005). In triple mutant lines combining loss-of-function alleles of these genes, the α-linolenic acid content in seeds is reduced to the 1% level of total fatty acids (Hoshino et al. 2014).
The double mutant line of GmFAD2-1a and GmFAD2-1b was developed as the soybean cultivar ‘Sadai HO1 gou’ in 2018 (Anai 2020). This cultivar exhibits five times higher oxidative stability compared with ‘Fukuyutaka’, and its extremely low linoleic acid content results in almost no formation of n-hexanal with undesirable green or rancid odor via lipoxygenase activity.
Soybean saponins are a group of glycosides composed of oleanane-type triterpenoid aglycones with various sugar moieties and more than 50 kinds of saponin molecules have been identified (Tsukamoto 2012). Group A saponins are responsible for undesirable bitter or astringent aftertastes (Okubo et al. 1992). In contrast, DDMP saponins and their derivatives (e.g., group B and E saponins) derived from soyasapogenol B exhibit various health-promoting properties such as the prevention of dietary hypercholesterolemia and suppression of colon cancer cell proliferation (Yano et al. 2017). Therefore, genetical controls of saponin composition could improve both the taste and functionality of soybean-derived food products, ultimately enhancing the value of soybean cultivars.
To date, five genes: Sg-1 (Sayama et al. 2012), Sg-3 (Yano et al. 2018), Sg-4 (Takagi et al. 2018), Sg-5 (Yano et al. 2017), and Sg-9 (Sundaramoorthy et al. 2019), have been identified, and the use of the mutant allele of Sg-1 has been the greatest contributor to soybean breeding in Japan. The Sg-1 gene encodes a UDP-sugar-dependent glycosyltransferase that mediates glycosylation of the terminal sugar at the C-22 position of group A saponins (Sayama et al. 2012). In the codominant alleles Sg-1a and Sg-1b, the glycosyltransferases UGT73F4 and UGT73F2 are functionally expressed, leading to the production of group A saponins with acetylated xylose (Aa-type) and acetylated glucose (Ab-type), respectively, as terminal sugars. The key functional distinction between Sg-1a and Sg-1b is the amino acid at position 138, which is Ser in Sg-1a and Gly in Sg-1b; this difference is critical for defining their sugar donor specificities. Conversely, sg-10–a and sg-10–b are loss-of-function alleles of Sg-1a and Sg-1b, respectively, resulting in the production of A0-type saponins that lack acetylated terminal sugars. Because acetylated sugars are considered responsible for imparting undesirable bitterness and astringency, processed soybean products such as soymilk and tofu made from A0-type varieties have been evaluated as having superior flavor quality. Based on this characteristic, varieties with A0-type group A saponin, such as ‘Kinusayaka’ (Kato et al. 2007) and ‘Sumisayaka’ were developed and registered in 2008 and 2024, respectively. During the development of these cultivars, MAS using SSR markers located near Sg-1 was partially employed in conjunction with phenotypic evaluation based on thin-layer chromatography. These cultivars are mainly used for soymilk production and were cultivated on approximately 94 ha in the Tohoku region and 155 ha in the western region of Japan by 2023.
Boiled seed hardness is an important factor in the processing of soybean food products, not only for cooked beans but also for miso and natto, which necessitate bean softening by steaming or boiling prior to fermentation (Motoki et al. 2013, Yoshioka et al. 2009). ‘Hyoukei-kuro 3’ is a Japanese soybean variety known for its soft boiled seed texture after steaming. Hirata et al. (2014) have identified qHbs3-1 and qHbs6-1 associated with boiled seed hardness on chromosomes 3 and 6 using RILs developed from a cross between ‘Natto-shoryu’ (hard) and ‘Hyoukei-kuro 3’ (soft). Subsequent progress in positional cloning and related studies led to the identification of pectin methylesterase (PME) as the causal gene of qHbs3-1 (Toda et al. 2015). PME de-esterifies pectin in cell wall by converting methoxyl to carboxyl groups, releasing methanol and protons (H+). Carboxyl groups cross-link with Ca2+, increasing cell wall rigidity and boiled seed hardness. The ‘Hyoukei-kuro 3’ allele introduces a premature stop codon, producing a truncated PME lacking activity, likely preventing hardness during boiling. Additionally, F3ʹH, the gene underlying the T locus, has been identified as the causal gene for qHbs6-1 (Toda et al. 2022). Ca2+ and Mg2+ contents of varieties with F3ʹH active allele (e.g., ‘Hyoukei-kuro 3’) were higher in pods and lower in seeds than those with inactive allele (e.g., ‘Natto-shoryu’), suggesting that F3ʹH inhibits the transport of Ca2+ and Mg2+ from pods to seeds as well as increase in heat-labile pectin caused by fewer cross-linked divalent ions, and softens boiled beans.
Effective amplification refractory mutation system (ARMS) markers for PME were developed by Toda et al. (2020). Although there are no reported cases where it has been directly introduced into leading varieties through backcrossing, it is utilized in population selection and early generation line selection, particularly for breeding varieties suitable for natto and boiled beans. In contrast, F3ʹH has been reported to exhibit pleiotropic effects, including suppression of Ca2+ and Mg2+ translocation, as well as influence on pubescence color, seed coat color, and chilling tolerance (Cober et al. 1998, Yamaguchi et al. 2021c). Among these, seed coat color is particularly problematic; when combined with the I locus in the IITT genotype, the yellow seed coat becomes dark, resulting in an undesirable seed appearance. Therefore, in contrast with PME, F3ʹH is not actively utilized in breeding.
Lipoxygenases (LOXs) in soybean seeds are enzymes that catalyze the peroxidation of polyunsaturated fatty acids and the generation of volatile compounds, such as n-hexanal, which cause undesirable flavors in soybean products. The absence of LOX1, LOX2, or LOX3 in seeds is controlled by the single recessive alleles, lx1 (Hildebrand and Hymowitz 1981), lx2 (Davies and Nielsen 1986), and lx3 (Kitamura et al. 1983), respectively. Hajika et al. (1991) developed a mutant lacking all lipoxygenases by gamma-ray irradiation and the first lipoxygenase-less variety, named ‘Ichihime’, was developed using the progeny of this mutant (Hajika et al. 2002, 2009). Several lipoxygenase-less Japanese cultivars including ‘L-star’ and ‘Suzusayaka’ have been developed.
Previously described, lipoxygenase-deficient varieties have primarily been selected through phenotyping based on SDS-PAGE analysis. In addition, in segregating populations, the selection of deficient individuals has relied largely on a simple and rapid screening method based on the bleaching activities of these isozymes developed by Suda et al. (1995) owing to financial and labor constraints. Consequently, the development and application of DNA markers remain limited. However, because all lipoxygenase-deficient traits exhibit recessive inheritance, distinguishing between wild-type and heterozygous genotypes based solely on phenotypes is impossible. Therefore, MAS is particularly effective during backcrossing, when elite cultivars are used as recurrent parents. The genes responsible for LOX1, LOX2, and LOX3 have been isolated and are located on chromosomes 13, 13, and 15. Notably, Lx1 and Lx2 are positioned in close proximity and approximately 3 kb apart. In Japanese breeding programs, the mutant alleles lx1, lx2, and lx3 are derived from ‘PI 408521’ (premature stop codon caused by a 74bp deletion), ‘PI 86023’ (missense mutation substituting Gln for His-532), and ‘PI 205085’ (premature stop codon caused by a single guanine deletion) (Lenis et al. 2010, Wang et al. 1994), respectively. Given that the mutation sites are known, identifying the SSR markers located near these regions that are suitable for MAS will be essential to facilitating efficient and accurate breeding.
The soybean cyst nematode (Heterodera glycines Ichinohe) is one of the most damaging soybean pests. In Japan, races 1, 3, and 5 of SCN have been reported. Three major loci, rhg1, rhg2, and Rhg4, are well known to confer resistance to soybean cyst nematodes. The rhg1a allele derived from ‘Peking’ encodes SNAP18, a soluble NSF attachment protein (Liu et al. 2017), whereas Rhg4 encodes SHMT08, a serine hydroxymethyltransferase (Liu et al. 2012). The interaction between these loci modulates resistance strength. The rhg1 locus exhibits copy number variation (CNV). The rhg1b allele from ‘PI 88788’, with 10 copies, confers strong resistance independently (Cook et al. 2012). However, resistance breakdown is a concern, prompting interest in the rhg1a and rhg2 combination found in ‘PI 90763’ (Basnet et al. 2022). The strongest candidate gene for rhg2 is GmSNAP11 (Lakhssassi et al. 2017). Most race 3 resistance alleles in Japanese cultivar are derived from ‘Gedenshirazu’, a landrace from the Akita prefecture. In contrast, ‘PI 84751’ has been used as a genetic source of resistance to races 1 and 3. The genotype of ‘Gedenshirazu’ is rhg4/rhg1-g/rhg2-g, whereas that of ‘PI 84751’ is Rhg4/rhg1-s/rhg2-s (Suzuki et al. 2012). Rhg4, rhg1-s, and rhg2-g or rhg2-s are necessary for race 1 resistance, and either rhg1-g and rhg2-g or rhg2-s, or rhg1-s and rhg2-g or rhg2-s is necessary for race 3 resistance. Whole-genome sequencing of ‘Suzuhime’ (PI 494182) derived from a cross between ‘PI 84751’ and ‘Koganejiro’, revealed three candidate genes for race 1 resistance: GmSNAP18 (rhg1-s), GmSNAP11(rhg2-s), and GmSHMT08 (Rhg4) (St-Amour et al. 2020). The GmSNAP02 is associated with race 2 resistance in ‘PI 90763’ and ‘PI 437654’, and confers a unique mode of resistance to SCN through loss-of-function mutations that implicate GmSNAP02 as a nematode virulence target (Usovsky et al. 2023). Resistance to races 1 and 3 from ‘PI 84751’ was introduced into ‘Yukihomare’, a leading variety in Hokkaido, by MAS using DNA markers for rhg1-s and Rhg4 and ‘Yukihomare R’ was developed (Suzuki et al. 2017). ‘Suzumaru R’, a suitable variety for natto, was developed by recurrent back-crossing with the donor parent ‘Chukou 1901’, using MAS with DNA markers linked to Rhg4, rhg1-s, and rhg2-g to introduce resistance to races 1 and 3 into the genetic background of ‘Suzumaru’ (Kurosaki et al. 2017). The production areas of ‘Yukihomare R’ and ‘Suzumaru R’ were 8,459 and 1,383 ha in 2023, respectively. They account for 21.6% of soybean production area in Hokkaido.
Soil-borne Oomycete Phytophthora sojae causes one of the most serious diseases in Japan. Although various control measures for this disease have been proposed (Dorrance 2018, Sugimoto 2013), the most cost-effective and easily adoptable strategy is the cultivation of resistant cultivars. Generally, resistance to PSR was classified into complete (qualitative) resistance through Rps genes and field (quantitative) resistance (Jia and Kurle 2007, Otolakoski and Huzar-Novakowiski 2024). When the pathogen infects host plants, Phytophthora sojae emits effector proteins encoded by avirulence (Avr) genes. In soybean varieties with complete resistance, the corresponding Rps gene recognizes these effectors and triggers a defensive response. In contrast, if the soybean variety is susceptible, or if there is a mismatch between the pathogen’s Avr gene and the host’s Rps gene, disease develops (Hou et al. 2023). More than 40 Rps genes have been reported (Lin et al. 2022). Rps1d and Rps1k are the most effective resistance genes among the 14 Rps genes examined using 109 P. sojae isolates from 14 regions in Japan (Moriwaki 2010). Although resistance is believed to be conferred by nucleotide-binding site leucine-rich repeat (NBS-LRR)-type proteins, many causal genes remain unidentified due to the highly repetitive structure of these loci. However, pan-genome analysis and complementation tests confirmed that Rps11 is an NBS-LRR gene (Wang et al. 2021b). Stacking multiple Rps genes is considered the most effective strategy for durable resistance (Sahoo et al. 2021). In Japan, soybean varieties such as ‘Waseshiroge’ (RpsWA) (Sugimoto et al. 2011) and ‘Tosan 231’ (Matsuoka et al. 2021) have been reported to possess potentially novel race-specific resistance. These cultivars exhibit resistance to domestic pathotypes of PSR and thus hold promise for future use in resistance gene stacking strategies.
Although race-specific resistance by certain Rps genes has been effective for many years, field resistance is also an important complement to Rps-mediated complete resistance. It is broader than resistance dependent on a single Rps gene, and field resistance does not negatively affect yield under conditions where PSR infection does not occur (Dorrance 2018). In addition, because partial resistance does not exert strong selective pressure on the pathogen population, it is considered more durable. de Ronne et al. (2022) reported a novel major QTL that markedly enhanced the resistance to a broad range of pathotypes of P. sojae on chromosome 15. ‘Conrad’ is a representative cultivar with field resistance, and more than 35 QTLs or genomic regions have been reported in different biparental populations (Chandra et al. 2022). Sugimoto (2013) reported a major domestic variety, ‘Fukuyutaka’, possesses field resistance. Although the genetic basis of resistance in the cultivars is still under investigation and no resistant varieties have yet been developed using DNA markers, further progress is highly anticipated.
Soybean bacterial pustule, caused by Xanthomonas axonopodis pv. glycines, is a serious foliar soybean disease that occurs predominantly in warm and humid climates (Sinclair 2015). The disease results in both yield and quality losses; infection can reduce seed number and size, leading to yield losses of 60% in severe cases (Kim et al. 2022, Sinclair 2015). Effective control of this disease involves the use of resistant varieties, and the resistance gene, rxp, has been successfully used in breeding programs to confer strong resistance (Hartwig and Lehman 1951). The Rxp gene encodes a transcription factor and the resistance allele, rxp, harbors a mutation that results in a truncated protein (Taguchi-Shiobara et al. 2024). This resistance allele is widely used in North American varieties, but rarely in Japanese varieties. Given the increasing incidence of disease in central and northern Japan owing to rising temperatures and frequent extreme weather events, the introduction of resistance genes in Japanese varieties is essential for the future stability of soybean production.
‘Suzuotome 2 gou’ is a newly developed soybean variety in which resistance to bacterial pustule has been successfully introduced. The recurrent parent variety, ‘Suzuotome’, had been cultivated primarily in Kyushu area, and is characterized by small seed size and suitability for natto processing. One of the major limitations of ‘Suzuotome’ is its high susceptibility to bacterial pustule, which frequently causes premature defoliation and yield reduction in the Kyushu region. To overcome this challenge, ‘Suzuotome 2 gou’ was developed through backcrosses four times using MAS with SSR markers flanking the rxp gene using a resistant variety, ‘Suzukaren’, as a donor parent. ‘Suzuotome 2 gou’ retains the agronomic and quality characteristics of ‘Suzuotome’, including early maturity, small seed size, and natto processing suitability, while the resistance and yield in ‘Suzuotome 2 gou’ are superior.
SMV causes mosaic discoloration and deformation of leaves and mottling of the seed coat, resulting in reduced seed yield and quality, respectively. This virus is transmitted both by aphids and through infected seeds, making chemical control with pesticides alone insufficient for its eradication (Hill 2015). Six strains (A, A2, B, C, D, and E) have been reported based on their pathogenicity in Japanese test varieties. Geographical differences were observed in the occurrence of each strain. For example, strains A and B are distributed throughout Japan, whereas strains C and D are mainly found in Kanto, Hokuriku, and the middle and southern parts of the Tohoku region (Hashimoto and Nagasawa 1987, Takahashi et al. 1980). Although the strain A2 has been reported in some regions of western Japan, the strain E, which causes severe symptoms including necrosis, has rarely been observed (Saruta et al. 2005). Most Japanese soybean varieties are resistant to strains A and B; however, many varieties are susceptible to strains C and D. Therefore, resistance to strains C and D should be incorporated into breeding programs that target regions where these strains are prevalent.
Canadian soybean variety ‘Harosoy’ has been identified as a resistant donor to SMV strains C and D (Takahashi et al. 1980). ‘Dewamusume’ is the first variety developed using this resistance allele (Ishikawa et al. 1979). Since then, successive resistant lines have been developed, primarily through mass selection based on large-scale inoculation using a spray gun. However, Kato et al. (2016) demonstrated that resistance to strains C and D, derived from the cultivar ‘Harosoy’, is conferred by Rsv3, which encodes a CC-NBS-LRR protein (Tran et al. 2018). Utilizing flanking SSR markers as selection tools, Kato et al. (2016, 2017) conducted recurrent backcrossing with MAS and successfully developed promising lines, such as ‘Tohoku 169’ and ‘Tohoku 173’.
Two other major loci, Rsv1 and Rsv4, are also associated with SMV resistance. Rsv1 has 10 known alleles, one of which (Rsv1-y) is recognized as a separate locus, Rsv5 (Klepadlo et al. 2017). The Rsv1 region forms a gene cluster and its causal gene has not been definitively identified, although CC-NBS-LRR-type proteins have been implicated (Wu et al. 2019). Rsv4 functions via a distinct mechanism involving a nuclease that infiltrates the SMV replication complex and degrades double-stranded RNA, thereby suppressing viral replication (Ishibashi et al. 2019). Although ‘Peking’, a genetic resource serving as the Rsv4 donor, had not been often utilized previously owing to its inferior agronomic traits, the development of DNA markers has facilitated its incorporation through MAS. In general, NBS-LRR proteins suppress disease development through a strain-specific hypersensitive response. However, because viral mutations may overcome this resistance (Kobayashi et al. 2014), utilizing multiple resistance genes in domestic breeding programs is favorable. ‘Peking’ has black, small, and flattened seeds, and thus the Rsv4 gene has rarely been utilized in Japanese breeding programs where seed appearance quality is critical. However, identification of the Rsv4 gene has enabled the development of effective DNA markers, leading to increased breeding use in recent years. ‘Haregokoro’ was developed by introducing resistance to SMV conferred by ‘Peking’ as well as those to peanut stunt virus (PSV) and southern bean mosaic virus (SBMV), into the leading cultivar ‘Sachiyutaka A1 gou’ by recurrent back crossing with MAS. It was adopted as a recommended cultivar in Okayama prefecture in 2023.
Peanut stunt virus (PSV) causes yield reduction and seed quality deterioration owing to mottling in southwest Japan. Genetic analysis of resistance to two PSV isolates (PSV-K and PSV-T) revealed that resistance in cultivars such as ‘Hyuga’, ‘Harosoy’, and ‘Tsurunotamago 1’ is governed by a single dominant gene located at the same locus (Saruta et al. 2012). This resistance gene, designated Rpsv1, was mapped near the SSR marker Satt435 on chromosome 7 using RILs of ‘Hyuga’ × ‘Enrei’.
The soybean dwarf virus (SbDV) affects soybean production, particularly in Hokkaido and Aomori prefectures. Genetic studies have identified the Indonesian cultivar ‘Wilis’ as a source of resistance, leading to the discovery of a major resistance gene, Rsdv1 (Uchibori et al. 2009). This gene was fine-mapped to a 44-kb region between the SSR markers Sat_11 and Sct_13 on chromosome 5 (Yamashita et al. 2013). Near-isogenic lines (Rsdv1-NILs) incorporating Rsdv1 showed strong resistance to SbDV under both greenhouse and field conditions. In addition, Kato et al. (2017) developed seven elite soybean lines, including ‘Tohoku 174’, by introducing the Rsdv1 into ‘Ohsuzu’, a major cultivar in Aomori, through recurrent backcrossing. These lines exhibited resistance under field conditions in Hokkaido, which suggests that Rsdv1 was effective not only in the genetic background of varieties in Hokkaido but also in those adapted to the Honshu region.
The Raso1 from cultivar ‘Adams’ confers antibiosis resistance to the foxglove aphid (Aulacorthum solani), a major vector of SbDV while ‘Adams’ suffers SbDV infection to a limited extent. Fine-mapping localized Raso1 to a 63-kb region on chromosome 3, which included a nucleotide-binding site-leucine-rich repeat (NBS-LRR) gene and two additional candidate genes (Ohnishi et al. 2012). Introduction of Raso1 into the susceptible cultivar ‘Toyomusume’ resulted in strong aphid resistance across all backcross lines, but was insufficient to confer SbDV tolerance.
The common cutworm (Spodoptera litura Fabricius) is a major lepidopteran soybean pest that causes serious damage to leaves and pods, significantly affecting plant growth and yield (Komatsu et al. 2010, Yadav et al. 2023). Resistance to lepidopteran insects has been classified into two categories: antixenosis and antibiosis, and different resistance genes are involved in each mechanism (Rector et al. 1999, 2000). Several resistance genetic resources against herbivorous, ‘IAC100’, ‘Soden-daizu’ (PI 229358), ‘Miyako white’ (PI 227687) and ‘Kosamame’ (PI 171451) have been identified and used in breeding programs in the United States (Cooper and Hammond 1999, Hatchett et al. 1976, Lambert and Kilen 1984). In Japan, a natto cultivar ‘Suzukaren’ with moderate resistance to CCW was developed from the progeny derived from the cross of ‘Suzuotome’ and resistant Brazilian variety ‘IAC 100’ (Takahashi et al. 2013).
In the selection of resistant lines, field-based screening may vary in accuracy depending on the year and environmental conditions. Therefore, the development and application of DNA markers are being advanced to enable more stable and efficient selection. Two QTLs located on chromosome 7, CCW-1 and CCW-2, were identified through antibiosis evaluation using a population derived from a cross between ‘Fukuyutaka’ and a resistant variety in Japan, ‘Himeshirazu’, and have been incorporated into soybean breeding efforts in Japan (Komatsu et al. 2005, 2010). Their effects were validated by comparing ‘Fukuyutaka’ and their near-isogenic lines ‘NIL-CCW1 + CCW2’ into which two resistance alleles derived from ‘Himeshirazu’ had been introduced, and the insect growth indexes (defined as pupal weight divided by the duration of the sixth instar) of ‘NIL-CCW1 + CCW2’ were found to be 50% lower than those of ‘Fukuyutaka’ (Komatsu et al. 2005). ‘Fukuminori’ was the variety with resistance to CCW selected from ‘NIL-CCW1 + CCW2’ lines, and its important agricultural characteristics, e.g., maturity, yield, protein content and the processability for tofu, were almost the same as those of ‘Fukuyutaka’ (Takahashi et al. 2017). However, the resistance level of ‘Fukuminori’ is clearly inferior to that of ‘Himeshirazu’ (Takahashi et al. 2017), and thus more effort for pyramiding the CCW resistance genes will be needed for significant reductions in pesticide application or yield improvement.
Research is increasingly focused on identifying the genetic factors associated with antixenosis resistance. Notably, the QTLs qRslx1 and qRslx2 on chromosomes 7 and 12, respectively, were derived from a cross between ‘Fukuyutaka’ and ‘Himeshirazu’ (Oki et al. 2012). However, because the position of qRslx1 was similar to that of CCW-1 and the resistant alleles of qRslx1 and qRslx2 were derived from Himeshirazu and Fukuyutaka, respectively, these QTLs have not been effectively utilized for pyramiding in combination with CCW-1 and CCW-2. An antixenosis resistance QTL from wild soybean (G. soja) collected in Kumamoto prefecture was identified at almost the same position as CCW-2 on chromosome 7 (Oki et al. 2019). In contrast, two antixenosis resistance QTLs, qRslx3 and qRslx4, have been discovered on chromosomes 7 and 2, respectively, in wild soybeans collected in Hiroshima prefecture (Oki et al. 2017). The resistant alleles of these QTLs are derived from wild soybeans and are located on chromosomes that differ from those of CCW-1 and CCW-2. The incorporation of these QTLs into breeding programs is expected to contribute to the development of cultivars with enhanced resistance to common cutworms.
Waterlogging stress in soybeans can occur at any growth stage; however, early stage waterlogging, particularly during germination, has a disproportionately negative effect on yield. This is particularly critical in Japan, where soybean production in converted paddy fields often exceeds that in upland areas. Sayama et al. (2009) identified QTLs on chromosomes 8 and 12 that significantly affected germination and seedling rates, respectively. Soil hypoxia, a major consequence of waterlogging due to limited oxygen diffusion, remains a primary challenge. QTL analysis of root development under hypoxic conditions using RILs from ‘Tachinagaha’ and the tolerant landrace ‘Iyodaizu’ confirmed stable QTLs for root length (Qrld-12) and surface area (Qrsad-12) on chromosome 12 over multiple years (Nguyen et al. 2017).
A selection method for flooding tolerance at the flowering stage was developed, where plants were subjected to flooding in the field. ‘Shokukei-32’, a breeding line from the Hokkaido Research Organization with superior waterlogging tolerance, was developed using this method (Kousaka et al. 2013). The qFTA2–1, a QTL associated with flooding tolerance at the flowering stage, was detected on chromosome 2 using a population derived from a cross between ‘Shokukei-32’ and the flooding-sensitive variety ‘Toyoharuka’, and is effective specifically in the cross with ‘Toyoharuka’ (Kurosaki et al. 2025).
Chilling tolerance of soybeans is a critical trait for stable production in Hokkaido, the northernmost island of Japan. After the onset of flowering, yellow soybean seeds are damaged by low temperatures, leading to seed coat deterioration, including pigmentation and cracking around the hilum (Morrison et al. 1998, Srinivasan and Arihara 1994, Takahashi 1997, Takahashi and Asanuma 1996). This phenomenon is known as cold-induced seed coat discoloration (CD). CD is most likely caused by inhibition of CHS silencing by low temperatures (Kasai et al. 2009). A CD-tolerant yellow hilum cultivar, ‘Toyoharuka’, possesses a polymorphic GmIRCHS structure, which has been designated as the Ic (Kasai et al. 2009, Yamaguchi et al. 2015). Ic, a novel allele of the I locus, inhibits pigmentation of the hilum and entire seed coat. The Ic allele is a major gene for CD tolerance, and a DNA marker distinguishing between I and Ic alleles, named the “Ic marker”, has been shown to be effective for the selection of CD-tolerant soybean plants (Ohnishi et al. 2011). The Ic allele was highly effective in a field with severe cold weather damage (Yamaguchi et al. 2019a).
Chilling temperatures result in the appearance of cracked seeds in soybean crops, particularly in those grown in eastern and northern Hokkaido. The coats of the cracked seeds are severely split on the dorsal side, and the cotyledons are exposed and frequently separated. A stable QTL associated with seed cracking (SC) tolerance was identified in the proximal region of the I locus, and the IcIc allele promoted a stronger SC tolerance (Yamaguchi et al. 2015). Proanthocyanidin accumulation on the dorsal side of the seed coat, controlled by the I locus, leads to cracked seeds; and that homozygous IcIc alleles confer SC tolerance (Senda et al. 2018).
Histological and texture analyses of the seed coat revealed that the ability to maintain hardness and flexibility under low temperature contributes to SC tolerance in ‘Toyomizuki’, an SC-tolerant cultivar with the II allele (Yamaguchi et al. 2023). A QTL associated with SC tolerance in ‘Toyomizuki’, was detected on the chromosome 8, but the distance between this QTL and I locus was estimated to be 2–3 Mb. The marker analysis revealed that ‘Toyomadoka’ has the ‘Toyomizuki’ allele at this QTL, indicating that MAS for the ‘Toyomizuki’ allele will be effective (Kobayashi et al. 2020).
Tawny-pubescent soybeans, controlled by the dominant T allele at the T locus encoding flavonoid 3ʹ-hydroxylase, show superior chilling tolerance compared with gray-pubescent soybeans with the recessive t allele (Funatsuki and Ohnishi 2009). Chilling tolerance conferred by T is caused by the increasing antioxidant activity of 3ʹ,4ʹ-dihydroxylated flavonol derivatives (Toda et al. 2011). In RILs derived from the chilling-tolerant cultivar ‘Hayahikari’ and chilling-sensitive cultivar ‘Toyomusume’, a QTL detected on chromosomes 6 was directly associated with the T locus (Funatsuki et al. 2005).
The pubescence color gene (T) is associated with the suppression of low-temperature-induced seed coat deterioration (Takahashi 1997, Takahashi and Asanuma 1996). T is also effective in the suppression of SC in the Ic background, indicating that gene pyramiding of Ic and T contributes to high SC tolerance (Yamaguchi et al. 2021c). The combination of I and T alleles darkens the yellow seed coat, leading to a dirty seed appearance; however, seeds harvested from near-isogenic lines with the IcIc/TT and IcIc/tt alleles looked similar to each other and did not have a dull appearance. Therefore, the problem arising from the introduction of T into yellow hilum cultivars may be overcome by using the Ic allele.
The advent of next-generation sequencing (NGS) technologies has drastically transformed SNP analysis by enabling rapid and comprehensive decoding of whole-genome sequences. For traits such as yield, which are controlled by complex interactions among numerous genes, GS utilizing genome-wide SNP data has been applied across various crop species. For soybeans, efforts to develop GS-based breeding strategies to enhance the efficiency of selection for yield and other agronomically important traits in other countries are underway. Although no published studies that clearly document the commercial use of GS in soybean breeding by private companies in the United States or other countries have been identified, several cases have been reported at university and public research institutions in the United States; for example, Smallwood et al. (2019) applied GS models to predict yield, protein content, oil content, and fatty acid composition. Although yield prediction proved to be challenging, the models achieved high prediction accuracy for oleic and linolenic acid contents. Bandillo et al. (2023) compared genomic and phenotypic selection for yield improvement. Their results demonstrated that genomic selection outperformed phenotypic selection at high selection stringency (10–20%). Miller et al. (2023) developed GS models for yield, protein, and oil content using a training population from the soybean breeding program at the University of Georgia, which targets commercial cultivar development. Similar to the findings of Smallwood et al. (2019), the yield was difficult to predict; however, moderate prediction accuracy was achieved for protein and oil content.
The SoyaGen project is a collaborative research initiative launched in Canada to accelerate soybean genetic improvement through genomics-derived solutions funded by Genome Canada and other national and provincial partners (Belzile et al. 2022). The SoyaGen project brings together academic researchers, breeders, and industry stakeholders. The main objectives include (1) the development of genomic prediction models for agronomic traits such as yield, maturity, disease resistance, and seed composition; (2) the identification and deployment of disease resistance loci, particularly for Phytophthora root rot and soybean cyst nematodes; and (3) the integration of GS into public and private breeding programs. Notably, the project demonstrated that GS can outperform traditional phenotypic selection under certain conditions, and that GS-based parental selection in actual breeding pipelines led to higher frequencies of elite progeny using high-throughput genotyping platforms and extensive historical breeding data (Jean et al. 2021). SoyaGen also developed decision-support tools and MAS strategies that have already been adopted by breeding programs in Canada. This project represents one of the first real-world implementations of GS in soybeans, offering a model for genomics-enabled breeding that can be adapted to other crops and regions.
The Strategic Innovation Promotion Program Phase 3 (SIP3) has launched a national project aimed at establishing a robust breeding platform and advanced cultivation technologies to support the stable and efficient production of high-quality soybean protein in Japan (Secretariat of Science, Technology and Innovation Policy, Cabinet Office 2023). In response to the growing global demand for plant-based proteins, driven by food security concerns and environmental sustainability, Japan has positioned soybean, a key source of plant protein, as a strategic crop. Conventional Japanese soybean breeding programs face several issues such as the limited development of high-yield cultivars with adaptation to food processing and consumer needs, vulnerability to recent extreme climate change, and deterioration in cultivation management due to labor shortages. To overcome these issues, SIP3 seeks to establish an integrated breeding and cultivation framework that leverages cutting-edge genomic and digital technologies, fostering innovation from breeding to on-farm implementation through industry-academia collaboration. As enhancing soybean yield in Japan has become an urgent priority to increase self-sufficiency and strengthen national food security, an integrated analysis-based breeding platform based on genomic information will be established to enable public research institutions and companies to breed more efficiently.
As exemplified by the Canadian breeding model, constructing a new breeding platform capable of selecting optimal crossing combinations is anticipated to contribute substantially to breeding efficiency. For example, when attempting to develop high-yielding cultivars by crossing varieties that differ in their composition of flowering-related genes, segregation of those genes in the hybrid progeny may cause undesirable variations in traits such as maturity, plant height, and biomass. This noise interferes with the accumulation of yield-enhancing alleles. Such issues are particularly problematic when incorporating yield-related genes from overseas high-yielding cultivars that are genetically divergent from the Japanese varieties. Conversely, eliminating crossing combinations that lack agriculturally essential traits, such as pod shattering and disease resistance, or selecting combinations in which such traits do not segregate in the progeny are advantageous. Achieving this necessitates cataloging the allele compositions of flowering-related and other agriculturally essential genes in the breeding materials under consideration. Advances in NGS technology are expected to play a key role in supporting this process. In future, if the allelic compositions of a wide range of breeding materials are cataloged and predictive tools for segregation in hybrid progeny are developed, it will be possible to eliminate crossing combinations that result in the segregation of flowering-related genes or the absence of essential agronomic traits. This will significantly enhance the overall efficiency of breeding programs in Japan.
A.K. drafted the entire manuscript. N.Y. and S.K. wrote the sections of manuscript related to the breeding in the Hokkaido and Honshu region, respectively. All authors read and approved the final version.
This work was supported by grants from cross-ministerial Strategic Innovation Promotion Program (SIP) “Building a Resilient and Nourishing Food Chain for a Sustainable Future” (funding agency: Bio-oriented Technology Research Advancement Institution) (Grant Number JPJ012287 to S.K., N.Y. and A.K.). We would like to express our sincere gratitude to Dr. Kyoko Takagi for kindly reviewing part of the manuscript and providing valuable comments.
