Comparative performance of hybrid generations reveals the potential application of F2 hybrids in upland cotton

The utilization of heterosis has greatly improved the productivity of cotton worldwide. However, a major constraint for the large-scale promotion of F1 hybrid cotton is artificial emasculation and pollination. This study proposed the potential utilization of F2 hybrids to improve upland cotton production through a comparative evaluation of hybrid generations. Eight upland cotton varieties were analyzed and crosses were made according to NCII incomplete diallel cross-breeding design in two cotton belts of China. Variance analysis revealed significant differences in agronomic, yield, and fiber quality in both generations and environments. The broad-sense heritability of agronomic and yield traits was relatively higher than quality traits. Furthermore, the narrow-sense heritability of some traits was higher in F2 than in the F1 generation in both cotton belts. Overall, parental lines Zhong901, ZB, L28, and Z98 were observed with maximum combining ability while combinations with strong special combining ability were ZB × DT, L28 × Z98, and ZB × 851. The yield traits heterosis was predominant in both generations. However, the level of heterosis was altered with trait, hybrid combination, generation, and environment. Interestingly, L28 × Z98 performed outstandingly in Anyang. Its lint yield (LY) was 24.2% higher in F1 and 11.6% in F2 than that of the control Ruiza 816. The performance of SJ48 × Z98 was excellent in Aral which showed 36.5% higher LY in F1 and 10.9% in F2 than control CCRI 49. Further results revealed most hybrid combinations had shown a low level of heterosis for agronomic and fiber quality traits in both generations. Comparatively, ZB × DT and L28 × Z98 showed hybrid vigor for multiple traits in both generations and cotton belts. It is feasible to screen strong heterosis hybrid combinations with fine fiber in early generations. In the two environments, the correlation of some traits showed the same trend, and the correlation degree of Anyang site was higher than that of Aral site, and the correlation of some traits showed the opposite trend. According to the performance of strong heterosis hybrid combinations in different environments, the plant type, yield and fiber traits associated with them can be improved according to the correlation. Through comparative analysis of variance, combining ability, and heterosis in F1 and F2 hybrids in different cotton belts, this study proposed the potential utilization of F2 hybrids to improve upland cotton productivity in China.


Background
Heterosis is a phenomenon by which hybrid progenies show superior performance compared to theirs in the aspect of vegetative growth, reproductive growth, and stress tolerance (Shahzad et al. 2019a). Hybrids have widely been used to improve the crop yield of agronomic and horticultural crops including rice (Li et al. 2016), maize (Yu et al. 2021), tomato Yu et al., 2020), kohlrabi (Singh et al. 2019). The utilization of heterosis increased the 10%∼20% yield of hybrid rice (Oryza sativa) more than conventional cultivars (Luo et al. 2013). The soybean hybrids produced 15%∼25% more yield compared with conventional varieties (Wang et al. 2002). At present, about 80%∼90% of vegetable varieties are hybrids. Even countries such as the Netherlands, the United States, and Israel have more developed hybrid vegetable seed industries. Utilizing heterogeneity is an extremely important genetic improvement technique to boost yield, quality, and resistance to diseases, insects, and pests. Global warming is a major threat to sustainable yield in recent years. Therefore, heterosis has the important practical significance in meeting market demand, improving economic efficiency, and ensuring food security.
Cotton is a major economic crop that has not only a renewable natural textile fiber source but also owns an ample amount of vegetable oil resources (Chen et al. 2007). Approximately 90% of the world cotton yield comes from upland cotton (Gossypium hirsutum L.) while Egyptian cotton (Gossypium barbadense L.) produces only 3% fiber (Fang et al. 2017). The upland cotton has shown significant heterosis for yield traits and altered across various traits, stages, and environments (Schnable et al. 2013). Moreover, hybrid cotton could be more adaptable and stable in varying environments (Shahzad et al. 2019b). Cotton hybrids have been devolved through the utilization of heterosis in China and planted in the main cotton provinces such as Hubei, Hunan, and Jiangxi. The area of hybrid cotton planted was about 70% of the total cotton grown in these provinces (Xing et al. 2017). Heterosis has become a crucial way to increase cotton yield and improve fiber quality. Selecting and promoting hybrid cotton with strong heterosis have a meaningful impact on cotton production in China. However, artificial emasculation seed production is the main way to utilize cotton heterosis. Due to the high cost of seed production, the utilization of F 1 heterosis is largely restricted to vast hybrid commercialization. To mitigate this challenge, the promotion and application of hybrid cotton increased rapidly with the expansion of cotton planting area in Xinjiang, and people gradually shift their attention to using the F 2 generation of cotton hybrids.
Many cotton breeders have already proposed the utilization of F 2 cotton hybrids to reduce the cost of seed production and to meet the demands of cotton growers in diverse ecological environments. A large number of research findings showed that F 2 hybrids still have certain competitive advantages over inbred parents (Meng et al. 2019;Chen et al. 2021). Combining ability is an important index to determine the transmission ability of excellent characters, to correctly evaluate the advantages and disadvantages of combinations, and to select excellent parents and hybrid combinations to boost the efficiency of any breeding program (Wang et al. 2012;Liu et al. 2019;Shi et al. 2021). In this study, eight upland cotton varieties were selected as experimental materials and crosses were made according to the NC II incomplete diallel cross (5 × 3) breeding design. The performance, combining ability, and heterosis were analyzed in F 1 and F 2 hybrids for multiple traits and locations. The main objective of our study is to compare F 1 and F 2 hybrids and combine them with the breeding practice of strong hybrid cotton to select the best combination of heterosis and provide a reference for the feasibility of parental selection and utilization of F 2 heterosis in China.

Experimental materials and field design
The field tests were conducted from 2020 to 2021. All 15 F 1 hybrid combinations used in this study were produced by adopting North Carolina mating design II by crossing five upland cotton inbred lines as the female parents with three different inbred lines as the male parent with different nuclear backgrounds which have been reported in our previous studies (Li et al. 2019;Shahzad et al. 2019b). Specifically, the inbred lines Zhong 901 (P1), ZB (P2), SJ48 (P3), L28 (P4), and K8 (P5) were used Keywords: Upland cotton, F 2 generation, Combining ability, Heterosis, Heritability  (Table 1). In 2021, eight parents, 15 F 1 and F 2 hybrid combinations were planted in two different cotton belts of China, i.e., in Anyang which is located in Henan, and in Aral which is located in Xinjiang (40°55′N, 81°28′E). Ruiza 816 and CCRI 49 were used as the control varieties in Anyang and Aral, respectively. All experimental materials were planted in a randomized complete block design, with 3 replicates. In Anyang, each material was planted in four rows without mulching, and in Aral adopts film mulching, and each material was planted under one film with six rows. Each block was 9.6 m 2 , and guard rows were set up around. The density was set according to the different ecological environment types, as 45 000 plants per hectare in Anyang, and 150 000 plants per hectare in Aral. Seeds were sown in late April in sequential years and the crop management practices followed the local recommendations.

Investigation and methods of phenotypic traits
In mid-September, the plant height (PH), the height of first fruit branch (HFFB), length of first fruit branch (LFFB), the second fruit branch length (SFBL), fruit branch number (FBN), and boll number (BN) for each plant were investigated. When more than 90% of bolls had opened, one fully-opened boll was randomly selected from each of 50 individual plants and weighed to estimate boll weight (BW). The weight of seed cotton per plot was used to calculate seed cotton yield (SCY) and lint yield (LY) per hectare, and the lint percentage (LP, the ratio of the fiber weight on the seed cotton to the weight of the seed cotton

Variance analysis of F 1 , F 2 hybrids, and their inbred parents in different cotton belts
The variance analysis was performed for 15 F 1 , and F 2 hybrids, and their eight inbred parents. The variance was extremely significant (P < 0.01) for the majority of traits in different cotton belts (Tables 2 and 3). All agronomic, yield, and fiber quality traits except FU and FE showed significant differences in F 1 generation among the combinations in Anyang (Table 2). Similarly, the differences among the combinations of F 2 generation reached significant or extremely significant for all traits which indicated that the differences in these traits were mainly caused by genetic variation. The male inbred lines had a nonsignificant variance in LFFB and SFBL in both F 1 and F 2 generations. In contrast, male inbred lines demonstrated significant variance for the majority of yield and fiber quality traits in both generations. The male variance was extremely significant and improved in F 2 generation for SCY, LY, BN, BW, LP, and FL. Furthermore, male inbred lines exhibited significant differences for PH, SFBL, FE, and FS only in F 2 as compared with the F 1 generation. The female inbred lines had a significant variance in eight traits of the F 1 generation, while the difference was significant only in seven traits of the F 2 generation. The female × male interaction variance was significant for the majority of traits in both generations except HFFB, FE, and MIC. Table 3 summarized the analysis of variance results for all traits in Aral. All combinations in F 1 displayed extremely significant differences in agronomic and yield traits, whereas FL and FS-related fiber quality traits had significant differences. Similarly, F 2 only showed extremely significant differences in agronomic and yield traits other than BN. The variance was significant among male parents for most of the traits in the F 1 generation specifically in LFFB, SFBL, SCY, LY, BN, LP, FL, FS, and MIC. The variance for female inbred was inconsistent between the two generations for 80% of agronomic and some yield traits. For example, LP had a highly significant difference in the F 1 generation . In contrast, PH, HFFB, SFBL, and FE had significant differences in F 2 . Interestingly, the female × male interaction variance was extremely significant for PH, HFFB, LFFB, FBN, and BW in F 1 and F 2 .

Heritability analysis of F 1 , F 2 hybrids, and their eight inbred parents in different cotton belts
Heritability estimates the ratio of genetic variance to phenotypic variance. The broad-sense heritability (H 2 ) and narrow-sense heritability (h 2 ) were determined for all traits. Heritability analysis with Anyang was detailed in Table 2. According to the results, the percentage of H 2 was strong for the majority of traits in both hybrid generations. In particular, LFFB, SFBL, FBN, SCY, LY, BW, LP, and FL stated that H 2 is greater than 70% in both hybrid generations. Conversely, FU had a lower percentage of H 2 relative to other traits in both hybrid generations. Further results determined that h 2 was strong and above 50% HFFB, SCY, LY, BW, LP, and FL in both F 1 and F 2 hybrids. LFFB and FU had very low h 2 in both generations than all other traits. Heritability analysis for Aral detected that H 2 for PH, LFFB, SFBL, and BW was great than 65% in both F 1 and F 2 . The heritability of fiber traits in Aral was relatively lower than in Anyang. Specifically, FU and FE had low h 2 , which was less than 20% in F 1 and F 2 . Interestingly, the h 2 of some traits in the F 2 generation was higher than that in the F 1 generation. For instance, SCY, BW, LP in Anyang and LY in Aral had higher h 2 , in F 2 with more than 50%. These findings put forth a clue that it is significant to select these traits in the F 2 generation. The traits with lower heritability can easily be affected by the environment. Hence, these traits can be improved through longer screening cycles during breeding measures.

General combining ability analysis of inbred parents in different cotton belts
General combining ability analysis is useful to screen superior inbred parents for specific or a set of traits. Based on the results of combining ability analysis, the GCA of the parental line was different and altered with generation, trait, and environment (

Special combining ability analysis of F 1 and F 2 hybrids in different cotton belts
The SCA revealed the performance of a cross and provide an opportunity for the utilization of heterosis in crop breeding. The SCA of all combinations was altered with traits and environments (Tables 5 and 6). In Anyang, it was observed that combinations 6, 8, 9, 10, 12, 13, and 14 have five or more than five traits with SCA values greater than zero in both generations. Among these, combinations 9, 10, and 12 all had positive SCA for SCY and LY in both F 1 and F 2 hybrids. Besides, combination 9 had also shown positive SCA for LFFB, BN, BW, FE, and MIC, and the SCA values of combination 10 were positive for BW, LP, FL, and FS. Meanwhile, combination 12 had better performances with positive SCA for LFFB, SFBL, BN, BW, FL, and FS (Table 5). The SCA analysis results in Aral were shown in Table 6. Among all combinations, only 2, 8, and 11 of SCY and LY were detected to be positive for SCA in both F 1 and F 2 . In particular, combination 2 had shown higher SCA for nine traits including SCY, LY, BW, LP, FL, FU, FE, FS, and MIC. Interestingly, the SCA of this combination was improved in F 2 for FL, FU, FE, and FS. It was observed that Combination 8 showed better performance as well as positive SCA for SCY, LY, BW, LP, FU, and FE. Combination 11 exhibited positive SCA in eight traits including PH, HFFB, FBN, SCY, LY, BW, FE, and MIC. Besides these, combinations 3, 6, 9, 14, and 15, had positive SCA for most of the traits in F 2 as compared to F 1 . These combinations most probably can be selected in the F 2 breeding generation to improve these traits in Aral. Overall, analysis results revealed that combinations 9 and 2 had improved performance in F 2 in both cotton belts which emphasizes the selection of these combinations in earlier generations would be effective for the future breeding program.

The screening of hybrids with excellent heterosis in multiple traits
In this study, the level of MPH, BPH, CH, and HD for different traits, hybrid combinations, and in different cotton belts were analyzed. The analysis results revealed that the level of heterosis altered with the trait, hybrid combination, generation, and environment (Additional file 1). The majority of combinations in Anyang had shown the highest heterosis for yield traits as compared to agronomic and fiber quality traits. For instance, in the F 1 generation, combination 12 exhibited the highest MPH (45.9%) for LY, and the LY of combination 6 had the highest BPH at 36.3%. Moreover, the highest CH was 28.4% which had shown by combination 7 for BN. Most combinations of HD were positive for yield traits but negative for agronomic and fiber quality traits. It may be because of the negative MPH, BPH, and CH in agronomic and fiber quality traits. Among F 2 generation, the LY of combinations 5, 1, and 9 witnessed the highest MPH (24.0%), BPH (20.9%), and CH (11.6%) values, respectively. Intriguingly, combination 9 had outstanding MPH, BPH, and CH in multiple traits as compared to other combinations. The analyzed results in Aral showed F 1 had the highest CH for LY (36.5%). This was exhibited by combination 8. However, combination 2 had the highest MPH (21.9%) and BPH (19.7%) for SCY among others. Besides this, a positive HD was measured for most yield traits among all hybrid combinations. While agronomic and fiber quality traits had negative HD in most hybrid combinations. The results revealed hybrid combinations had shown positive MPH, BPH, and CH for yield traits in the F 2 generation. Interestingly, combinations 2 and 9 had shown outstanding heterosis in multiple traits in Aral (Additional file 1). The overall analysis determined that combination 9 had the best hybrid vigor in both generations and cotton belts. Therefore, it can be considered an outstanding hybrid for both cotton zones. Subsequently, this study further screened the top eight hybrid combinations with superior performance in multiple traits. The results revealed CH, MPH, and BPH in selected hybrids were altered with generation and cotton belts (Fig. 1, Additional file 2). It was determined that more than 6 combinations had better CH, MPH, and BPH in both generations and cotton belts. However, some combinations had superior CH, MPH, and BPH in both generations but one cotton belt. In this regard, combination 12 had similar performance in Anyang while combination 2 and combination 9 in Aral. Besides this, some combinations exhibited strong vigor in both cotton belts but only in one generation. Such as combination  2 and combination 7 had shown better CH, MPH, and BPH in F 1 . Combination 9 displayed better CH, MPH, and BPH in F 2 . Comparatively, combinations 2 and 9 showed excellent performance in multiple traits for both generations and cotton belts. These encouraging results evaluate the potential of F 2 hybrids to improve cotton productivity in China.

Correlations among various traits in two different cotton belts
The relationship between traits is a dynamic factor in the selection of plant breeding materials. The correlation analysis between agronomic, yield, and fiber quality traits in Anyang was summarized in Fig. 2A. A significant positive correlation was observed among yield (SCY, LY) and yield components (BN, BW, LP). All fiber quality traits except FU showed a negative correlation with yield traits. A significant positive correlation between FU and yield was detected. The correlation between yield and agronomic traits was either non-significant negative or positive. Similar results were observed among most of the fiber quality and agronomic traits. Most fiber quality traits including FL, FE, and FS had a positive correlation with each other. However, MIC had a strong negative correlation with FL and FS but a positive correlation with FU. The correlations were undulating among agronomic traits. For instance, PH had a significant negative correlation with LFFB and MIC but had a significant positive with FBN and FS. A significant positive correlation was observed between SFBL and LFFB.
The correlation analysis in Aral revealed SCY had a significant positive correlation with LY and BN whereas LY was positively correlated with BN and BW (Fig. 2B). The correlation of BW was significantly positive for FE and MIC. LP showed a significant negative correlation with FL and FS. In contrast, it had a significant positive correlation with MIC. Among fiber quality traits, FL had a significant positive correlation with FE and FS. MIC had a significant negative relationship with FL and FS. The agronomic traits had shown diverse correlations but few were significant. For instance, PH had an extremely positive correlation with HFFB, FBN, and FL. And PH negatively correlated with LP and MIC. Moreover, SFBL positively correlated with FS. FBN positively correlated with FL and FE. Overall, analysis results propose that agronomic, yield, and fiber quality traits can be improved independently in both cotton belts.

Discussion
Cotton plays a critical role in textile industry development, employment opportunity, and foreign exchange earnings. Genotypes with higher yield and fine fiber are desired in upland cotton. This synchronized improvement of multiple traits in upland cotton demands more crossing, assessment, screening, and useful resources. The utilization of heterosis is the most suitable method to achieve such vast breeding aims. Worldwide, difficulties in producing F 1 hybrid seeds have restricted the commercial use of heterosis in cotton. However, this study compared the performance, combining ability, and heritability in both F 1 and F 2 generations in two cotton belts. Further the potential utilization of F 2 hybrids was screened and discussed to improve cotton production in China.
Parental selection has critical importance in hybrid cotton breeding. However, the identification of potential parents is a laborious job. In the utilization of heterosis, selected parental materials should have superior performance, physiology, combining ability, and heritability. GCA refers to the average performance of a parental line in hybrid offspring and mainly anticipates the role of heritable additive genes contribution (Liu et al. 2019;Shang et al. 2012). Therefore, statistics of GCA determined the selection of parental lines in the future breeding program. Previous studies have already shown that parents with high GCA can be well exploited through heterosis to produce superior hybrids (Hassan et al. 2000;Lukonge et al. 2008). In our study, GCAs in the majority of parental lines were positive but the values altered with generation. Moreover, yield traits were detected with higher GCA and fiber traits with lower. Previous researches in F 1 and F 2 hybrids stated similar statistics for combining ability in upland cotton (Tang et al. 1993;Khan et al. 2009). Among all inbred parental lines used in this study, P4 and P7 had the best GCA for multiple traits in F 1 and F 2 generations, and in both cotton belts (Table 4). These inbred lines' superior performance in multiple traits, generations, and environments proposed their utilization in the further breeding program to develop elite hybrids. Interestingly, our results showed that the GCA of P4 was improved for LFFB, FBN, LY, BW, LP, and MIC in F 2 as compared with F 1 in both environments. The abrupt increase may be the result of heterogeneous material with different effects in F 2 which probably lead to good adaptation in different environments. The estimate of heritability defines the range of genotypic and phenotypic variances. Therefore, it reveals the potential of parents to be selected and exploited to develop high-yielding genotypes. High heritability and GCA increased the probability of selecting hybrid offspring with good performance in early generations (Sun et al. 1994;Jia et al. 2017). Our results displayed that the majority of yield traits had strong H 2 and h 2 among different generations and environments (Tables 2 and 3). The traits with high heritability indexes showed are less vulnerable to diverse environments. Thus, simple selection in early generations would be an effective strategy to improve these traits (Soomro et al. 2010). In cotton breeding, GCA and heritability analysis provide a foundation to screen highly dominant materials (Li et al. 2010a). However, combined performance across multiple generations and ecological zones could be an efficient method to identify elite breeding populations.
Estimates of SCA reflect the average performance of a hybrid combination and are mainly produced by the action of dominant or epistatic gene interaction. This non-additive gene action mediates the mechanism of heterosis in upland cotton (Ahuja and Dhayal 2007;Shahzad et al. 2020). Thus, estimates of SCA provide an opportunity to screen potential hybrid combinations in a particular generation or environment (Soomro et al. 2012;Khan et al. 2015). Our study revealed that the magnitude of SCA varied with the traits, generations, and environments. Interestingly, combinations 9 and 2 had shown positive SCA effects in multiple traits in both F 1 and F 2 generation in two cotton belts, but for combination 3 and 15 in the two environments, the SCA of F 1 and F 2 showed opposite results in multiple traits related to yield, quality, and agronomic traits. This is consistent with previous studies on cotton F 1 hybrid combination with strong performance, and the dominance in F 2 was not necessarily well (Shang et al. 2012) (Tables 5 and 6). In particular, yield and yield components were identified with higher SCA effects in these hybrids. Such promising results proposed that superior combinations may be utilized as F 2 hybrids to increase yield or as an elite population in advance breeding experiments. Besides this, those F 2 hybrid combinations with superior performance in a specific cotton belt would most likely be utilized to improve cotton productivity in such zone. Previous research stated that GCA and SCA were independent and higher GCA does not essentially interlink with higher SCA. Therefore, more emphasises should be on SCA effects rather than the GCA effects of inbred parents during the process of hybrid selection (Yang et al. 2009;Peng et al. 2015;Canavar et al. 2011). Correlation between traits plays a vital role in plant material selection (Liu et al. 2008). Our results showed a negative correlation between yield and quality characters in both cotton belts (Fig. 2). These results were consistent with those previously reported by different researchers (He et al. 2009;Li et al. 2010b). These results enabled improvement in yield-related traits independent of fiber quality traits. Moreover, some agronomic traits showed a significant positive correlation with yield and quality traits in this study. Therefore, these agronomic traits should also be taken into consideration in the breeding of hybrids across mechanical harvest cotton zones. Apart from this, how to improve fiber quality is still an important research topic in hybrid cotton breeding.
The utilization of heterosis improved the productivity of crops. Utilization of heterosis is one of the key ways to improve stagnant yield in upland cotton. However, the major challenge is the difficulty of producing F 1 seed through manual emasculation and pollination (Wu et al. 2004) which caused the high cost of production and seed impurity. To mitigate this challenge, the commercial use of F 2 hybrids is proposed by many researchers (Li et al. 2000;Iqbal et al. 2015). The upland cotton belongs to allotetraploid, its F 2 segregation is not severe as in diploid rice and maize (Chen et al. 2020). Additionally, cotton has a long harvest period, and the plant architecture, growth stages, and agronomic traits may not have a direct impact on the yield and fiber quality of F 2 generations (Wang et al. 2011;Kong et al. 2017). These unique cotton characteristics provide an opportunity for the utilization of F 2 hybrids to improve productivity. In this study, some combinations of F 2 hybrid generation performed well in multiple traits. For instance, combination 9 had shown excellent performance in multiple traits in both cotton belts ( Fig. 1; Additional file 1). It illustrated that combinations with strong vigor performed well in F 2 (Liu et al. 2007;Zhang et al. 2018). Moreover, heterogeneity in F 2 most like enabled wider environment adaptation as compared with F 1 and inbred parents. Commercialization of elite F 2 hybrids not only reduced production costs but also increased yields and promotes hybrid cotton.

Conclusions
In this study, we systematically evaluated the potential breeding applications of F 2 hybrids by comprehensive comparative analysis of their field performance on yield, quality, and plant architecture-related traits. The combining ability variance and heritability of traits significantly differed across multiple traits in two generations and both environments, suggesting that it is meaningful to select and breed hybrid F 2 generations in upland cotton. The GCA of parents P4 (L28) and P7 (Z98), and the F 1 and F 2 generations of hybrid combination ZB × DT and combination L28 × Z98 in both environments were all outstanding in many traits such as yield, quality, and plant architecture. Therefore, it is feasible to breed cotton F 2 with potential for production and application by synthetically evaluating the yield, quality, plant architecture traits, and environmental adaptability of hybrid cotton F 2 through strict parent selection and in multi-plot experiments for several years.