INTRODUCTION

Sorghum [Sorghum bicolor (L.) Moench] is the fifth most important cereal grown worldwide and it is of particular interest in Korea for its health benefits (Awika & Rooney, 2004, Chung et al., 2011). In Korea, waxy endosperm genotypes are preferred (Cho et al., 2015) where it is steamed with rice, and used as an ingredient in rice cakes, and as a starch source in popular beverages. Other uses of the grain include tea, pop sorghum and bread (Ko et al., 2012). As a consequence of these uses, there has been a steady increase in grain sorghum demand in Korea. Given the limited area planted to sorghum and the low average yield (~ 1.6 ton.ha^-1) (Food and Agriculture Organization of the United Nations, 2015), adopting enhanced breeding strategies is fundamental to improve yield potential.

Korean sorghum production has traditionally been small- scale and subsistence, but there is a steady and slow shift to commercial production. There are several benefits to sorghum production in the country. Because sorghum can grow in paddy soils, the crop is a great alternative to rice production (Jung et al., 2013). While this may displace some rice production area, the rotation of sorghum crop could help attenuate the undesired economic and social effects of rice overproduction observed in recent years. Moreover, increase in meat consumption has demanded higher amounts of forage that is currently imported (~ 1.1 million tons annually) (Kim, 2014). Developing a sustainable sorghum production system would be the first step to widen the availability of grain and forage sorghums. To accomplish this requires the utilization of germplasm with high performance and excellent stability.

Pure line varieties and landraces are currently dominant in sorghum production in Korea. However, they tend to be less productive and less stable than hybrids (Duvick, 2005, Duvick, 1999, Gowda et al., 2010, Haussmann et al., 2000, Mühleisen et al., 2013, Oettler et al., 2005). Since the discovery of cytoplasmic male-sterility in sorghum (Stephens & Holland, 1954) the ability of efficiently combining topperforming and complementary parental lines has resulted in an improved agronomic performance in many regions of the world. Although utilizing sorghum hybrids appears rather advantageous (Haussmann et al., 2000, Kenga et al., 2004, Li & Li, 1998, Miller & Kebede, 1984), such technology is still not utilized in Korea; explaining to a great extent, the low average yields. Further, grain yields of waxy-endosperm sorghum genotypes tend to be lower than non-waxy types (Jones & Sieglinger, 1952, Tovar et al., 1977). Thus, the development of waxy grain sorghum hybrids adapted to Korea is important. In China, exploiting the diversity between parental landrace lines improved heterosis and productivity of sorghum hybrids (Li & Li, 1998).

Selecting appropriate parents and determining their performance in hybrid combinations is essential to building an effective hybrid breeding program. Combining ability analyses provide information about the superiority of parental lines by estimating general combining ability (GCA) and the performance of hybrids by calculating specific combining ability (SCA) (Cruz et al., 1997). Moreover, the relative importance of GCA vs. SCA indicates the gene action involved in the traits of interest. Efficient hybrid breeding also requires the development of complementary parental pools to maximize heterosis (Smith & Primomo et al., 2010). Several studies on GCA and SCA have been conducted for sorghum grain yield in diverse genetic backgrounds (Chikuta et al., 2017, Hovny et al., 2005, Makanda et al., 2010). Parents with positive GCA for grain yield and negative GCA for days to heading are considered good general combiners (Can et al., 1997). A wide range in better parent heterosis for grain yield has been observed (9 to 97%) that is dependent on the environment and genotypes under evaluation (Mahdy et al., 2011). Ultimately, the value of heterosis depends on the gene action, GCA, and SCA of the parental lines (Hochholdinger & Hoecker 2007).

Biometric information on grain sorghum lines and landraces in Korea is limited at best. Further, the available germplasm appears to have low yield potential and limited diversity suggesting that exotic germplasm is needed. To assess the utility of improving grain yield through the development of hybrids with wider genetic diversity, we evaluated two distinct sets of waxy sorghum hybrids derived from Korean landraces and US lines (Texas A&M University) in multi-environment trials. The main objectives of this study were to: (i) verify if US x Korean and US x US hybrids increase grain yield in Korea; (ii) assess the performance of waxy hybrids in Korea and the US; (iii) estimate general combining ability and specific combining ability for the lines studied; and (iv) identify superior lines and hybrids for future use in breeding.

MATERIAL AND METHODS

Plant material

Thirty hybrids were evaluated in two different mating scheme designs. In the first layout (Experiment I), two elite male-sterile lines from the Texas A&M Sorghum Breeding Program (A03017 and ATx630) were crossed to six pollinator parents from the South Korean Upland Crop Breeding Program (Donganme, Hwanggeumchal, Jungmo4002, Miryang11, Nampungchal, and Sodamchal) to produce twelve F₁ hybrids. In the second design (Experiment II), nine male-sterile lines (A10595, A11029, A11030, A11031, A11032, A11266, A11267, ATx631, and ATxARG-1) were crossed to two pollinator lines (RTx436 and RTx437) (Miller & Dusek et al., 1992, Rooney & Miller et al., 2003) giving rise to eighteen F1 hybrids. In this case, both female and male parents were from the Texas A&M Sorghum Breeding Program.

Field experiments

All hybrids were grown in Miryang, South Korea from June to October of 2016 and in College Station, Texas, USA from March to July of 2016 (Table 1 for site’s descriptors). The experiments were conducted in a randomized complete block design with three replications at Miryang and two replications at College Station. At the Miryang site, seed of each entry was hand-planted in two-row plots of 4.00 m length at 0.60 m inter-row and 0.20 m intra-row spacing. In College Station, two-row 5.50 m long plots spaced 0.76 m apart were planted at 0.07 m intra-row spacing using a plot planter. The expected population densities for Miryang and College Station were ~ 88,000 and ~186, 000 plants ha^-1, respectively. Standard agronomic practices for fertilization, weed and insect control were followed according to the recommendations of each location. In brief, elemental nitrogen, phosphorus, and potassium were supplied as an N-P-K formulation before planting at a rate of 411 kg ha^-1(22-17-17) in Miryang and 168 kg ha^-1 (11-37-0) in College Station. Pre-emergence herbicides were applied at recommended rates to control broadleaf and grass weeds. Hand weeding was done as needed during crop growth to keep the fields weed free.

Data collection

Data was collected per plot basis in both locations for the following traits. Days until flowering were recorded as the number of days from planting to when half of the plants in a plot reached mid-anthesis. Plant height was measured at maturity as the distance between the base-ground to the top of the panicle in meters as an average for the whole plot. Panicles were hand-harvested near the physiological maturity stage; grain was threshed using plot threshers and grain yield estimated in ton.ha^-1after adjusting all samples to 13% moisture.

Data analysis

The first step was to perform preliminary tests to verify violations of the assumptions for analysis of variance (ANOVA) and mixed models; all assumptions were met. Next, single-environment ANOVAs were conducted for each trait in both Experiments I and II according to the Line x Tester model (1): response = blocks + genotypes + lines + testers + lines x testers + error (Singh and Chaudhary, 1979). In this case, all terms in the model except the error term were considered fixed. A second model (2): response = blocks + genotype + error was fitted, and in this case, all terms were assumed to be random and normally distributed. Using this model, within-environment repeatability was estimated as (3) , where and are the genotypic and error variances, respectively, and r corresponds to the number of blocks. General combining ability (GCA) and specific combining ability (SCA) effects were estimated as described by Singh and Chaudhary (1979).

Due to the unbalanced number of replications between environments, a mixed model framework (Restricted Maximum Likelihood - REML) was used to fit a multienvironment model and to test the effects of G x E. In this case; two distinct models were fit. Model (4) was fit as follows: response = environment + genotype + environment x genotype + block within environment + error. Lastly, model (5): response = environment + line + tester + line x tester + environment x line + environment x tester + environment x line x tester + block within environment + error was fit to test the line and tester effects. Underlined terms were assumed random and normally distributed. From model (4), overall repeatability was calculated as follows (6)

H 2 = σ g 2 σ g 2 + σ e 2 x e l + σ e 2 l r , where σ g 2 and σ e 2 are the genetic

G x E, and error variances. l and r corresponds to the number of environments and blocks, respectively. Mean comparison were done by using Fisher’s Unprotected LSD or Tukey’s HSD depending on the model.

RESULTS AND DISCUSSION

In both experiments, the average performance of lines and hybrids for grain yield, flowering time and plant height were higher in College Station (USA - US) than in Miryang (Korea - SK) (Figure 1). The effects of each environment on the average population performance were significant for days to flowering and plant height. Average grain yield for the hybrids was ~ 700 kg.ha^-1 higher in College Station than in Miryang for Experiment I (US x SK hybrids), while an average increase of ~ 170 kg.ha^-1 was observed for hybrids of Experiment II (US x US hybrids) in College Station compared to Miryang. However, these advantages were not significant (Wald Statistic_{95% CI}). In this study, the experiments were designed following the sorghum production systems predominant in each location. Accordingly, the population densities used in the tests at College Station (USA) were twice as large as the ones adopted in Miryang (Korea). This fact, associated with the longer growing season in College Station (USA) is likely explanations for the yield differences between the environments (Bond and Army et al., 1964, Miller and Quin by et al., 1968).

Across the environments, the hybrids from Experiment I out performed their respective pollinator lines with one exception (Table 1); in Korea, "Donganme" had numerically higher yield than its derivative hybrids. More than 37% of average yield increase was observed for the hybrids compared to their respective parental lines across both experiment and locations. Various hybrids showed statistically significant yield increase relative to their male progenitor which ranged from 2% to 127% in Korea. These results indicate there is a clear advantage to using waxyhybrids in Miryang (Korea). Compared to their parental lines, hybrids from Experiment I (SK x US hybrids) had a higher yield advantage in College Station (USA) than in Miryang. This greater yield advantage of hybrids could be associated with the poor performance of Korean inbred lines in the US conditions which inflated the hybrid performance relative to their parental lines. Cause the Korean parent lines used as parent line were adapted to local conditions and resulted in an increase in grain yield. In Korea, grain yield ranged from 2.1 ton.ha^-1 for the inbred line “Sodamchal” to 5.8 ton.ha^-1 for the hybrid “ATx630/ Jungmo4002”, revealing a potential yield boost of ~ 2.8x for the best hybrid. In College Station, yield variability ranged from 2.8 ton.ha^-1 for the inbred line “Jungmo 4002” to 6.3 ton.ha^-1 for the hybrid “ATx630/Sodamchal”.

In Experiment II, all hybrids outperformed their parental pollinator lines with an average relative performance advantage of 92% in College Station (USA), and 81% in Miryang (Korea) (Table 1-1). Overall, this result demonstrated that heterosis was higher in Experiment II than in Experiment I. The female and male lines in Experiment II belong to distinct heterotic groups out of the Texas A&M Sorghum Breeding Program which has been extensively tested to maximize hybrid vigor. Meanwhile, it is the first time that hybrids derived from the SK and US lines presented herein were tested for agronomic performance. Therefore, it is reasonable to hypothesize that the selected sets of parental lines from Korea (SK) and the US do not belong to fully complementary heterotic pools. The SK lines evaluated in this study have been historically selected to withstand major abiotic and biotic stresses present in the Miryang region. Therefore, due to background effects of SK lines on SK x US hybrids, they were assumed to show a positive response to such stresses. Because these stresses can overshadow yield potential, we expected the overall grain yield performance of SK x US to be higher than US x US hybrids in Miryang. However, this assumption was not true (Figure 1). Due to the same reasons, we assumed that, on average, hybrids derived strictly from US lines would out-yield SK x US hybrids when grown in College Station. However, this assumption was also false. Table 1-2

The levels of hybrid vigor observed in both locations were similar, but shifts in ranks of hybrid performance occurred. For example, the hybrid “ATx630/Sodamchal” had the highest yield at 6.3 ton ha^-1 in College Station while its grain yield in Miryang was the lowest among all hybrids at 3.1.ton ha^-1(Table 1-1). The hybrid “ATx630/ Jungmo4002” showed opposite results. In Experiment II, the hybrid “A11030/RTx437” had the highest yield at 7.1 ton.ha^-1 in Miryang while the hybrid “A10595/RTx437” had the highest yield in College Station at 6.4 ton.ha^-1. Such shifts in grain yield performance explain the significant GxE effects found for grain yield in both experiments. While hybrid using Texas A&M AgriLife germplasm significantly improved sorghum grain yield in Miryang, the results also suggest that line and hybrid development should be done in Korea due to the inconsistency in genotype performance between US and Korea.

Variation was observed for days to flowering and plant height in both locations and both experiments (Figure 1). As discussed previously, genotypes flowered later and were taller in College Station for both experiments. For Experiment I, inbred lines flowered later and were shorter than hybrids in both locations on an average basis. For Experiment II, parental lines flowered earlier than hybrids in Miryang. In College Station, however, inbreds flowered later than hybrids. In both locations, hybrids were taller than inbreds. Most of the top-performing hybrids had acceptable maturity and height. Therefore, such materials would fit well in production systems that use either hand or machine harvesting. the traits between days to flowering, plant height and grain yield had correlation for two distinct experiments and two locations. In College Station, genotypes with higher yield tended to flower earlier and to be taller for both experiments (Table 2). Similar relationships were observed in Miryang except that there was no correlation between days to flowering and yield in Experiment II. Height was consistently correlated with yield except for Experiment I in College Station. The lack of correlation in this environment was likely due to the inconsistency of yield for the Korean parental lines which are not well adapted to the Texas environment.

Repeatability (H²) estimates within locations were high for all traits in both experiments ranging from 0.79 to 0.99 (Table 3). The high repeatability for grain yield and other traits suggests that actual heritability estimates for these traits should be high. For Experiment I, days to flowering had the highest H² value in Miryang and the lowest in College Station. In the case of Experiment II, plant height had the highest H² value in both locations. Grain yield was highly repeatable within locations, and it had the highest H² values across environments for both experiments. Unexpectedly, the H² estimate for flowering time across locations in Experiment II was equal to zero. Negative variance was obtained for the genotype unit in Experiment II indicating that genotypes were less variable than someone would expect from the contributions of the sub-units of which the genotypes were composed (i. e., replicates within location, location, g x e). This could occur if the sub-units, for example, locations, were negatively correlated. Besides this fact, our results suggest that (i) the phenotyping techniques used to measure the traits herein were relatively consistent between the two countries; (ii) there was genetic variability for the genotypes evaluated; and (iii) for initial test cross evaluations in experimental conditions similar to this study, two replications should be sufficient to obtain reliable results.

In the Line x Tester analysis of Experiment I, GCA for lines was significant for days to flowering in College Station and Miryang, and for plant height in College Station. The GCA for tester was significant only for plant height in College Station while the SCA was significant for all traits in Miryang (Table 4). These findings indicate that the relative importance of additive to non-additive gene effects vary depending on the trait and location. For instance, additive effects were the primary effects for days to flowering in College Station while both additive and non-additive effects were important in Miryang. The relative importance of additive and non-additive effects in controlling plant height and grain yield was also inconsistent across locations. In Experiment II, GCA for testers was significant for days to flowering and plant height in College Station and Miryang. The GCA for lines was significant for days to flowering in College Station and plant height in Miryang. In agreement with the results of experiment I, the relative importance of additive to non-additive gene effects varied depending on the trait, location, and genotype. SCA was significant for grain yield in Miryang, and non-addictive effects were important in affecting grain yield in Miryang.

According to GCA estimates (Table 5), three lines from Experiment I (“Donganme”, “Sodamchal”, and “Jungmo4002”) had high and positive GCA values for grain yield in College Station. In Miryang, the lines “Jungmo4002”, “Donganme” and “Hwanggeumchal” had the top GCA estimates. Thus across both environments “Donganme” and “Jungmo4002” are consistently top performers. Among the two seed parents, “A03017” was the superior parent in both locations and it appears to be a good source of additive allele effects to reduce plant height in both locations, which is desirable to minimize lodging and to improve harvest index. In Experiment II, the female lines “A11267”, “ATx631”, and “A10595” were the best general combiners for grain yield in College Station while the lines “A11030”, “A11032”, and “A11267” were the best general combiners for grain yield in Miryang. The male tester line “RTx437” was better than RTx436 for grain yield and had desirable negative GCA values for days to flowering across locations. Within this group of germplasm, there are likely lines that would also combine well with the Korean germplasm.

In Experiment I, SCA effects were highly significant for all traits in Miryang and not significant for any trait in College Station (Table 4). For Experiment II, SCA effects were non-significant for all traits in both locations, except plant height in College station and grain yield in Miryang (for a significance level of 0.01). (Table 4). In Miryang, “ATx630/Jungmo4002” had the highest SCA value for grain yield followed by “AT630/Donganme” and “A03017 /Nampungchal” (Table 6). The hybrid that recorded high SCA effects involved either one of the parents with good GCA for the trait being considered (“A03017”, “Jungmo 4002” and “Donganme”). The parents that were the best general combiners did not always produce the best hybrid. for example A03017/Jungmo4002 had relative low SCA as 8.0. As suggested by several authors (Cruz et al., 1997, Knega et al., 2004, Laosuwan & Atkins 1977, Makanda et al., 2010, Pedersen et al., 1982) and evidenced in the Texas A&M Breeding program, general combining ability is typically used to identify good parents while specific combining ability is used to identify optimum hybrid combinations. Thus, high hybrid would tend to having with favorable SCA estimate and involving at least one of the parents with high GCA. In this experiment Although the number of crosses was relatively small and the parental lines were from pools with broad genetic diversity, the manifestation of SCA was valuable in this study. Further evaluations of hybrids using SK and US lines should allow the optimization of heterotic pools and development of superior breeding populations.

CONCLUSION

The evaluation of two sets of grain sorghum hybrids under production environments in the U.S and Korea indicated that hybrid vigor is manifested between Texas derived seed parents and Korean pure line cultivars used as pollinators. Compared to their parental lines, hybrids had a higher average yield advantage in CS than SK. Most waxy hybrids outperformed their parental pollinator lines in CS and SK especially, “sodamchal” with waxy endosperm had the highest performance with A03017 in Miryang. A comparison of GCA effects for grain yield showed that the positive GCA obtained by A03017, Jungmo4002 and Donganme in SK. The high positive SCA estimates for grain yield were obtained by ATx630/Jungmo4002 and ATx630/Donganme. Consequently, it is feasible to develop grain waxy sorghum hybrids adapted to Korean production systems quickly and that long-term development would very likely lead to greater yield and quality increases. In this development, general combining ability appears to be the primary factor in genetic effects and that selection and evaluation should be primarily in Korea. Finally, a combination of Korean derived and standard grain sorghum from U.S. sources provide an acceptable germplasm base for developing a Korean based sorghum hybrid breeding program.

적 요