Two wheat cultivars, sourced from the Gene Bank of the National Institute of Agricultural Sciences, Rural Development Administration (RDA), Republic of Korea (ROK), were employed to implement the synergized breeding strategy (SBS). The Korean hard white spring wheat cv. Jokyoung (IT 213249) served as the recurrent parent (Kang et al., 2006), whereas Canadian hard red spring wheat cv. Garnet (IT 205495) was used as the donor parent (McCallum and DePauw, 2008). The target gene of Garnet was Glu-B1i, located on chromosome 1B, which enhances bread quality by encoding the high-molecular-weight glutenin subunit (HMW-GS) 17 + 18 in wheat (Payne et al., 1987; Jang et al., 2017; Guo et al., 2019; Shin et al., 2020; Guzmán et al., 2022). Jokyoung is a leading bread wheat cultivar in Korea that harbors Glu-B1b, which encodes the HMW-GS 7 + 8 (Jang et al., 2017).

The synergized breeding strategy (SBS) incorporated three breeding tools: speed breeding (SB) system, phenotypic selection (PS), and marker-assisted backcrossing (MABC). The SBS scheme is illustrated in Figure 1 . An initial artificial cross between Jokyoung and Garnet was conducted under field conditions. Two rounds of backcrossing with Jokyoung were performed under modified speed breeding combined with speed vernalization (SV+mSB), as described by Cha et al (Cha et al., 2023b). The BC₂F₂ population was evaluated under mSB conditions, which enabled visual phenotypic selection based on the difference in days to heading (DTH) between the two parental lines. BC₂F₃ lines were planted in the field during the spring season, and a second visual phenotypic selection was conducted considering DTH and grain color. Molecular markers were applied to the BC₁F₁, BC₂F₁, BC₂F₂, and BC₂F₄ generations to select lines closely resembling Jokyoug genetically and harboring the Glu-B1i gene from the donor parent. The foreground selection marker, cauBx642, was used in BC₁F₁, BC₂F₁, and BC₂F₂ to identify lines carrying Glu-B1i as heterozygous (BC₁F₁ and BC₂F₁) or homozygous (BC₂F₂). Background selection at BC₂F₄ was aimed at assessing the recurrent genomic background recovery. All BCnF₁ plants were cultivated in 12-cm pots, which provide a larger number of spikes and grains (Cha et al., 2020), while plants from the BC₂F₃ generations were sown in 72-cell trays.

Genomic DNA was extracted from fresh seedling leaves of each plant using the MagMax DNA Multi-Sample Kit (Applied Biosystems, Woburn, MA, USA) and Kingfisher DNA extraction machine (Thermo Fisher Scientific, Waltham, MA, USA) following the manufacturer’s instructions. For foreground selection, the cauBx642 marker was used, as reported by Xu et al (Xu et al., 2008). The polymerase chain reaction (PCR) was carried out under the following conditions: initial denaturation at 95°C for 5 min, followed by 35 cycles of denaturation at 94°C for 30 s, annealing at 59°C for 30 s, and extension at 72°C for 30; and final extension at 72°C for 5 min. PCR products were separated by electrophoresis on a 3% agarose gel at 200 V for 45 min.

For background selection, a total of 731 kompetitive allele-specific polymerase chain reaction (KASP) assays were used to identify the genetic background of the developed backcrossed lines. Among these 731 assays, 155 assays referred to CIMMIYT wheat molecular genetics (Dreisigacker et al., 2016), 70 assays to Rasheed et al (Rasheed et al., 2016), and the rest to the CerealsDB core marker set (Wilkinson et al., 2012). The PCR conditions were as follows: denaturation at 95°C for 15 min; 10 touchdown cycles at 95°C for 20 s, annealing starting at 65°C and decreasing by 1°C per cycle for 25 s; additional 30 annealing cycles at 95°C for 10 s and 57°C for 1 min; and final extension at 72°C for 5 min. Fluorescence detection and data analysis were performed using a QuantStudio3 real-time PCR instrument and QuantStudio Design & Analysis Software v1.5.1 (Applied Biosystems, Carlsbad, CA, USA).

All glasshouse-based and field tests were conducted at the National Institute of Crop Science (NICS), RDA, Miryang, ROK (35°29’32.9” N, 128°44’33.4” E). Agronomic traits, including days to heading (DTH), days to maturity (DTM), culm length (CL), spike length (SPL), number of tillers per plant (TN), number of grains per spike (GN), thousand-grain weight (TGW), test weight (TW), and grain yield (GY), were assessed following the RDA Standard Evaluation Manual for Agricultural Experiments and Research (RDA, 2012).

In the field, the BC₂F₃ lines were planted in late February 2021, whereas the BC₂F₅ lines were sown in early November 2021, representing the spring and winter growing seasons, respectively. BC₂F₅ lines underwent observed yield trials (OYT) in 2021, and BC₂F₆ lines were evaluated in both OYT and advanced yield trials (AYT) in 2022. For the OYT tests, lines were hill-seeded in the field with a spacing of 30 × 12 cm between and within the lines (a seed per hill). For the AYT tests, a rate of 14 kg/10a for each line was sown in the field as drill seeding, with 5 m rows and 30 cm spacing between rows. Fertilizer consisting of 9.1 kg N, 7.4 kg P₂O_5, and 3.9 kg K₂O per 10 a was applied in all field tests.

To analyze flour quality, the OYT and AYT lines were milled using a Brabender Quadurmat Junior mill (CW Brabender Instruments, Inc., South Hackensack, NJ, USA) and a Buhler MLU 202 laboratory mill (Bühler AG, Uzwil, Switzerland), respectively.

HMW-Gs were extracted from the flour and identified using a lab-on-a-chip (Agilent 2100 bioanalyzer & protein 230 kit, Agilent Technologies, Waldbronn, Germany), following the series of processes proposed by Shin et al (Shin et al., 2020).

Flour protein content was determined using LECO FP628 (Laboratory Equipment Co., St. Joseph, Mich., USA) with a nitrogen factor of 5.7. The gluten content and index were measured using a Glutomatic 2200 (Perten Instruments AB, Sweden). The SDS sedimentation value was determined following the method proposed by Axford et al. (1979), with a modified sample amount of 3 g. All experiments, including bread-making quality, were conducted in accordance with the American Association of Cereal Chemists (AACC) methods (AACC, 2010), and the experimental results were standardized based on 14% moisture content.

RStudio (version 1.4.1717; RStudio, PBC, Boston, MA, USA) was used for all statistical analyses. Chi-square tests, analysis of variance, Duncan’s multiple range tests, and t-tests were performed using the ggplot2 and agricolae packages.

SBS was executed according to the scheme shown in Figure 1 . The F₁ seeds were obtained from a cross between cv. Jokyoung and cv. Garnet under field conditions were planted under SV+mSB conditions in October 2019. Jokyoung was subjected to two rounds of backcrossing under SV+mSB conditions, followed by two rounds of foreground selection using the cauBx642 marker to detect the Glu-B1i allele.

The first backcross commenced in December 2019, and BC₁F₁ seeds were planted in February 2020. Of the six BC₁F₁ plants, three that exhibited the heterozygous Glu-B1i allele were selected for the second backcrossing ( Figure 1 , Table 1 ). A total of 31 BC₂F₁ seeds were sown in May 2020. Among these, 13 plants carried the Glu-B1i allele heterozygously and were harvested by the family for the next generation ( Figure 2A , Table 1 ).

BC₂F₂ plants were cultivated in September 2020 under mSB conditions for phenotypic selection (PS) based on days to heading (DTH). As DTH varied between Jokyoung and Garnet under mSB conditions, PS was conducted within the BC₂F₂ populations ( Figure 2C , Supplementary Table S1 ). Of the 31 BC₂F₂ families derived from each BC₂F₁ individual plant, 12 BC₂F₂ families with comparable DTH to Jokyoung and confirmed to carry Glu-B1i through bulked segregant analysis were chosen ( Supplementary Table S1 ). Individuals from each family were assessed using the cauBx642 marker. Among the 692 individuals evaluated, 195 lines harboring homozygous Glu-B1i were selected for subsequent generations ( Figure 2B , Table 1 and Supplementary Table S1 ).

A total of 133 BC₂F₃ lines, yielding a sufficient quantity of seeds for field trials, were planted in the field in February 2021. PS was conducted based on the differences in DTH and grain color between the two parental lines. Consequently, 27 lines showing a heading date similar to Jokyoung (May 6) and white grains were selected ( Figures 2D, E , Supplementary Figure S1 ). Subsequently, the 27 selected BC₃F₄ lines were grown under mSB conditions for background selection.

Of the 731 KASP assays collected, 141 exhibited polymorphisms between the two parents ( Supplementary Table S2 ). Initially, 27 BC₃F₄ lines were assessed using 101 KASP assays. Sixteen lines, showing more than 85% recurrent genome recovery and more than 95% homozygosity, were identified. Among these, eight lines with the highest recurrent genome recovery and homozygosity underwent further testing with an additional 40 KASP assays ( Table 2 , Supplementary Figure S2 ). As a result of testing a total of 141 polymorphic KASP assays, these eight lines showed an average recurrent parent genome recovery of 83.9% (79.4–91.5%). These eight lines were evaluated as observed yield trial (OYT) in November 2021, and NIL-1 demonstrated the highest background genome recovery (91.5%) and similar agronomic traits, including heading date, maturity date, culm length, and other agronomic characteristics, compared to Jokyoung while carrying the introduced Glu-B1i allele from Garnet ( Figure 2F , Supplementary Table S3 , Supplementary Figure S3 , and Supplementary Figure S4 ). Implementing SBS, from the initial crossing to the planting of the BC₂F₆ generation as OYT, took only 3.5 years, whereas it would have taken 7.5 years with conventional breeding methods, consequently reducing the time by 53%.

NIL-1 was designated Milyang56 and compared with the recurrent parent Jokyoung across two tested years (2021 and 2022) and trial methods (OYT and advanced yield trial(AYT)). A two-year OYT test was conducted to assess the annual variation. Overall, no significant differences were observed between Jokyoung and Milyang56 in most major agronomic traits, including DTH, days to maturity (DTM), culm length (CL), number of grains per spike (GN), and thousand-grain weight (TGW), across tested years and trial methods ( Figures 3A, B , Table 3 , Supplementary Tables S4–S6 ). While most agronomic traits exhibited significant differences due to environmental variations, such as the test year and trial method, there were no significant differences between Jokyoung and Milyang56. Although Milyang56 showed a shorter spike length than Jokyoung across both tested years and trial methods, no significant differences were observed in TGW and grain yield between Jokyoung and Milyang56 ( Figure 3C , Table 3 and Supplementary Table S5 ).

When the grains were milled, and the flour was tested, the protein content and SDS sedimentation value of Milyang56 were significantly higher than those of Jokyoung, irrespective of the tested years or trial methods ( Figures 3D, E ). Additionally, the gluten index tended to be higher in Milyang56, although no statistically significant difference was observed in the annual test ( Figure 3F , Supplementary Table S5 and Supplementary Table S6 ). As a result of introgressing the Glu-B1i allele without altering the agronomic traits, the bread loaf volume of Milyang56 increased by 9.9% compared to that of Jokyoung ( Figure 3G ).

Our study demonstrated how a series of SB methods could be combined with other breeding techniques to form an upgraded and realized model within breeding programs. It showed breeders how to choose and apply proper generation acceleration systems according to the breeding purpose and characteristics of the breeding materials. Both the mSB and SV+mSB systems were utilized to develop a new wheat line, “Milyang56”. The mSB system notably reduced the DTH of spring wheat cultivars; however, differences in DTH between different cultivars still existed (Ghosh et al., 2018; Watson et al., 2018). Therefore, phenotypic selection was conducted using this difference point in the BC₂F₂ populations. In contrast, the SV+mSB condition was used in two rounds of backcrossing to exploit its ability to ensure a uniform DTH, enabling artificial crossing between parents with diverse genetic backgrounds (Cha et al., 2022b). The series of SB systems was also naturally combined with field tests, along with breeding purposes and schedules.

Molecular breeding methods, such as MAS and MABC, were also applied under mSB and SV+mSB conditions to enhance both the speed and efficiency of breeding. We demonstrated a real-world case of the SB + PS + molecular breeding model named SBS. Compared to traditional breeding methods under field conditions, SBS reduced breeding time by 53% for developing the BC₂F₆ NIL. Foreground selection was conducted for BC₁F₁ and BC₂F₁, and background selection was conducted for the BC₂F₄ generation. Milyang56 showed 91.5% recurrent parent genome recovery with Glu-B1i introgression from the donor parent despite only two rounds of backcrossing. The theoretical average percentages of recurrent parent genome recovery in the BC₂ and BC₃ generations were 87.5 and 93.8%, respectively. However, these percentages are only achieved with large populations and are usually lower in the smaller population sizes typically used in actual breeding programs (Collard et al., 2005). For example, the average genome recovery was reported to be 92.2% despite both foreground and background selection being conducted in the BC₁F₁ and BC₂F₁ generations of maize (Babu et al., 2005; Hasan et al., 2015). Although only one background selection was conducted in our study, a higher genome recovery than the theoretical average was retained because of the complementary PS.

However, opportunities exist to conduct MABC more efficiently. The accuracy of individual selection was improved when background selection was conducted in the lower generations, BC₂F₁ and BC₂F₂, rather than BC₂F₄. A NIL with a 96.2% recurrent genome recovery was developed as a result of background selection, with 71 assays for BC₂F₁ and six assays for the BC₂F₂ generation in rice (Kang et al., 2019). In practice, most MABC studies have conducted background selection for BC₁F₁, BC₂F_1, and BC₂F₂ (Vishwakarma et al., 2014; Rai et al., 2018; Bellundagi et al., 2022).

Background selection in early generations combined with PS may significantly increase recurrent genome recovery with fewer backcrossings compared to traditional methods. To conduct genotyping in a limited time (1–2 months before harvesting) in a series of SB conditions, breeders must be able to analyze the genotype of a large amount of breeding material simply and quickly. Therefore, we utilized KASP assays, which are highly accurate, inexpensive, and flexible (Yang et al., 2019). Different types of molecular markers can be utilized for both foreground and background selection, according to the breeder’s conditions.

HMW-GS Glu-B1i is known to enhance bread quality by increasing protein content, gluten strength, and gluten extensibility (Guo et al., 2019; Guzmán et al., 2022). However, environmental conditions affect the allelic performance of HMW-GSs in terms of the end-use quality of wheat (Kroupina et al., 2023). In Korea, Glu-B1i has been reported to have higher protein content and SDS sedimentation values than other alleles in 180 wheat varieties carrying various Glu-1 alleles (Cha et al., 2022a). However, important agronomic traits, especially DTH and DTM, were diverse in the varieties analyzed in this study, and no previous study has investigated the effects of Glu-B1i on similar genetic backgrounds in Korea. Therefore, we developed Jokyoung-NIL, Milyang56 and compared the effect of the Glu-B1i allele on end-use quality. Our study confirmed the positive effect of Glu-B1i on bread-making quality, along with previous studies using the SBS method and an extremely shortened breeding period. This method would facilitate the comparison of the effects of specific genes or loci in various environments by quickly developing NILs, and it can be utilized in gene stacking.