INTRODUCTION

The onion (Allium cepa L.) is one of the more important vegetable crops and is grown all over the world. Brazil produces approximately 1,165,402 t of onions on 54,179 properties, with an average yield of 29.6 t ha^-1. Santa Catarina is the leading onion-producing state, with an output of 528.439 t and an average yield of 29.3 t ha^-1, grown mostly by subsistence farmers (^{IBGE, 2020}). Production is concentrated mainly in the region of Alto Vale do Itajaí, with 8,282 farms in Santa Catarina alone according to current estimates (^{EPAGRI/CEPA, 2019}). Onion farmers are not only searching for higher-yielding cultivars but also for genotypes with differentiated chemical properties that could make onions a competitive product with high added (^{Porta et al., 2014}). In addition to health benefits, the onion has the ability to improve or accentuate the taste of other foods (^{Mallor et al., 2011}). So, low-pungency onions have been gaining popularity (^{Mallor and Sales, 2012}; ^{Coca et al., 2013}; ^{Kim et al., 2017}; ^{Yoo et al., 2019}, ²⁰²⁰). This onion type can be used for direct consumption without any processing. In the United States for example, they are consumed as “fresh fruit” or directly in salads.

In Brazil, low-pungency cultivars are not yet available. Thus, the development of low-pungency onions without losses in desirable characteristics such as bulb yield breeding programs should seek sources of and propose strategies to exploit genetic variability (^{Denre et al., 2011}; ^{Cramer et al., 2021}; ^{Lo Scalzo et al., 2021}). The onion is a sulfur-demanding crop (S). This element is an important constituent of several amino acids, e.g., cystine, methionine, cysteine and tryptophan, as well as a precursor of volatile sulfur compounds. The content of S-based compounds determines the pungency level of onions (^{Tóth et al., 2018}). In onion fields, this element is frequently replaced by fertilization (in view of the high extraction of the nutrient) (^{Santos et al., 2019}). Thus, taking sulfur fertilization into account can reveal important interaction effects on variables of interest in genetic studies. The onion pungency level can be determined indirectly by measuring the pyruvic acid content (^{Wall and Corgan, 1992}).

Because of the onion's reproductive system, breeding can be based on: i) population improvement where genetic and genotypic frequencies are determined (^{Falconer and Mackay, 1996}) and selection for an increase in the frequency of favorable alleles is applied; and ii) hybrid breeding where the beneficial effects of heterosis are exploited. The hybrid-based approach is preferred with the efficiency of genetic-cytoplasmic male sterility systems in which crossing male-sterile with maintainer lines is a mandatory step. The main advantages of hybrid onion production are guaranteeing genetic consistency in the commercial product (bulbs) and homogeneity in seed germination.

The initial step in exploring the effects of heterosis is classifying populations into different heterotic groups. A heterotic group contains related or unrelated genotypes, from the same germplasm or not, with a similar combining ability and a heterotic response when crossed with genotypes from other genetically distinct groups (^{Melchinger and Gumber, 1998}). The more divergent the parents are, the higher the resulting genetic variability in the segregating population and the more likely the reassortment of alleles in new favorable combinations is because of the combination of contrasting characteristics in the parents. The progenies resulting from directed hybridizations have the complementarity of favorable alleles and open the way for selection in genetic improvement programs (^{Faria et al., 2019}; ^{Khandagale and Gawande, 2019}; ^{Pevicharova et al., 2020}).The objective was to characterize heterotic groups among onion genotypes to produce hybrids that combine low pungency with an unaltered high bulb yield.

MATERIAL AND METHODS

RESULTS AND DISCUSSION

The idea of improving a crop based on ideotypes is not new (^{Donald, 1968}). The onion ideotype designed for this study should have the ability to produce above-reference levels of bulb yield, alongside lower levels of pyruvic acid (low pungency) in the same genetic background. A multivariate analysis has the advantage of taking the combined variation of numerous traits into consideration, in particular when these variables are related in a relevant dependency structure (^{Hair et al., 2009}). In this context, a significant effect of all treatment factors (P<0.05) was detected with the multivariate analysis of variance (MANOVA) when considering pyruvic acid (PA), sulfur (S) and bulb yield (BY), simultaneously (Tab. 2).

For maximum exploitation of heterosis, parents must be chosen according to the germplasm origin from unrelated groups, commonly called heterotic groups (^{Melchinger and Gumber, 1998}). Superior progenies in relation to the parents can be achieved with a cross between plants from different groups for the target trait. A study on genetic divergence proposed the analysis of morphological, agronomic and biochemical descriptors using multivariate methods, resulting in the indication of 13 groups from 64 evaluated onion accessions (^{Buzar et al., 2007}). Among these groups, varieties can be indicated for crosses and exploitation of heterosis.

Previous studies have emphasized the problem of sulfur applications for the pungency level (or pyruvic acid content) in onions. Some argue that the application of extra sulfur to soils with a high sulfur content does not increase pungency or related compounds in short-day onions (^{Lee et al., 2009}). Others have confirmed that the pungency level and sugar content of onions cannot be significantly altered by the soil type or by applying extra sulfur to soil with sufficiently high S levels (^{Hamilton et al., 1998}). However, it has been found that sulfur applications significantly increase pungency in chive (Allium fistulosum L.), with mean pyruvic acid contents of 2.34 and 6.76 µmol g^-1 for all cultivars grown at 0.01 and 4.00 mmol L^-1 SO₄ ^2-, respectively (^{Liu et al., 2009}). Brazilian genotypes must be genetically characterized with regard to the specific traits of the ideotype. The conditions of soil, management and climate are environment-specific, which would make a generalization of the crop performance unfeasible. This study addressed the identification of heterotic groups based on bulb pungency and yield (Tab. 2).

The significant interaction effect indicated the differentiated performance of the genotypes in response to each fertilization level. In other words, a qualitative description of the interaction is required. Since the effect of greatest interest is on the genotypes, variations in environments, with and without fertilization, can be evaluated separately. This analysis may suggest the formation of different heterotic groups, depending on the environment in which the genotypes will be cultivated. The first canonical variable captured 80.03 and 73.33% of the accumulated variance without and with fertilization, respectively (Tab. 2). For a satisfactory interpretation of the variability in the genotypes, the first two canonical variables must represent 80% of the total variation in an analyzed data set (^{Cruz et al., 2012}). Thus, the contribution of agronomic traits can be identified, and the respective heterotic groups established satisfactorily with two dimensions (Fig. 1).

Figure 1. Multivariate dispersion of 26 onion genotypes according to canonical scores for pyruvic acid (PA), sulfur (S) and bulb yield (BY): a) treatment without fertilization of any genotype; b) treatment without fertilization of male-sterile (LA) and maintainer (LB) lines only; c) treatment with fertilization of all genotypes; d) treatment with fertilization of male-sterile (LA) and maintainer (LB) lines only. 

The canonical discriminant analysis visualized scores for each genotype according to the controlled fertilization levels. Also, by considering the germplasm source, the male-sterile (LA) and maintainer (LB) lines were separated (Fig. 1B and D) from the other genotypes to indicate crosses between heterotic groups that contain the male-sterile cytoplasm and the respective maintainers. According to the visual inspection, three and four groups were formed in the treatments without fertilization (Fig. 1A and B); while, the treatments with fertilization (Fig. 1C and D) only formed two groups.

For the characterization of the genotypes, this result indicated a buffering effect from the sulfur fertilization since a lower number of groups was established in the environment where sulfur was supplied. This was also evident when the linear combinations of the canonical variables were analyzed. In the treatment without fertilization, the first canonical variable revealed that the function that most effectively differentiated the onion genotypes was Can1 = 0.469 × PA + 51.994 × S + 0.016 x BY. In the fertilization treatment, the reference function consisted of Can1 = 0.043 × PA + 36.745 × S - 0.168 × BY. This shows that the contribution of the variables was comparatively higher in the treatment without fertilization. When the element was supplied, the bulb yield was not significant for discriminating the accessions.

The potential genotypes for hybridization (male-sterile and maintainer lines) and clustering in the treatment without fertilization formed four heterotic groups (Fig. 1B). However, the discriminant analysis formed groups based on a visual inspection, contrast-based mean comparison tests and standardized canonical coefficients, which were essential for comparing possible differences in performance between the genotypes and indicated which traits contributed more to the discrimination (Tab. 3). Since the objective for establishing the heterotic groups was to form onion hybrids, comparisons were made between the male-sterile × maintainer lines (LA × LB). All male-sterile lines were compared with the maintainer lines LB19, LB20 and LB24 (contrasts C1 to C18), except for line LA16, which, because of its greater distance, was compared with all maintainer lines (contrasts C19 to C24), except LB24, which belonged to the same group.

Contrast C1 (LA11 × LB19) indicated significant differences between the mean vectors (Tab. 3). PA (0.04) and BY (1.54) contributed to the differences between the lines. Although the pungency levels of both lines were relatively lower than those of the other genotypes (7.42 and 7.03 µmol g^-1), BY was contrasting since line 11 produced three times the value of line 19. In other words, these genotypes belong to different heterotic groups, and if crossed, the beneficial effects of heterosis for pungency (negative selection) and bulb yield (positive selection) can be exploited.

Similarly, the comparison between lines LA11 × LB20 was significant (contrast C2), whereas pungency (-0.22) did not contribute to the observed difference (Tab. 3). In this case, the mean vector with an expressive contribution was S (2.28). Maintainer line LB20 had a sulfur content of 0.32 mg g^-1, while the male-sterile line LA11 had 0.23 mg g^-1. The genotype performance differed in relation to sulfur uptake even in the treatment without fertilization, reinforcing the importance of understanding these relationships and establishing effective planning for plant selection.

Line LA11 also had a significant effect when compared to maintainer line LB24 (contrast C3). However, the traits that contributed most to discrimination were PA (0.54) and S (1.90). In this case, the pungency level of maintainer line LB24 was higher than that of the others (mean of 13.86 µmol g^-1), while that of male-sterile line LA11 was half of this value (mean of 7.42 µmol g^-1). In addition, the maintainer line had one of the lowest bulb yields of all evaluated genotypes.

The genetic bases that explain heterosis are based on the existence of heterozygous loci (^{Falconer and Mackay, 1996}). The more distant and homozygous the populations involved in the cross are, the greater the expectations for the exploitation of hybrid vigor is (resulting in greater heterozygosity in the F₁ population). Since the onion is affected by inbreeding depression, varieties maintain homozygous recessive alleles that are hidden by heterozygosity. Thus, in some crosses, hybrids with lower values of pyruvic acid can be detected.

The male-sterile lines LA12 and LA13 had the same behavior pattern as LA11 when crossed with the maintainer lines (contrasts C4 to C9). The contribution of bulb yield was notable in the discrimination of treatments when lines LB19 and LB20 were involved and in the pungency trait when lines LB19 and LB24 were involved. The male-sterile lines LA11, LA12 and LA13 probably not only belonged to the same heterotic group but were also improved in the same population, constituting only different selections of the same open-pollination variety, for example.

In all comparisons involving line LA14, trait PA contributed significantly to treatment discrimination (contrasts C10 to C12) (Tab. 3). Line LA14 had a 5.0 µmol g^-1 higher pyruvic acid content than lines LB19 and LB20. For contrast C10, all traits contributed to treatment discrimination, mainly bulb yield, with the highest effective contribution (1.59). On the other hand, the only comparison with no significant difference was contrast C14 (P=0.1534). Although located in different heterotic groups, the multivariate scores of lines LA15 and LB20 were closely related, which may explain the non-significance of this comparison. This reinforces the importance of making comparisons based on probability values, rather than only on decisions based on visual inspection. In the comparisons C13 to C15, line LA15 had the lowest pungency content of all male-sterile lines, both in the treatment without (6.83 µmol g^-1) and in the treatment with fertilization (5.80 µmol g^-1). In other words, this line was not responsive to this environmental factor, which makes it very desirable for breeding programs for the selection of low-pungency genotypes.

The contrasts involving the male-sterile line LA16 were all significant (C19 to C24). Despite the differentiation of the lines by pungency and sulfur in the bulbs, they were close in relation to the trait bulb yield (Tab. 3). In addition, in contrast to LA15, line LA16 had the highest pyruvic acid content (14.08 µmol g^-1) of all studied genotypes.

The National Onion Labs Inc. (NOL) in the United States certifies onion genotypes according to the pyruvic acid content (following the same protocol as in this study), with levels between 0 and 5.5 µmol g^-1 for low to medium-pungency onions. The germplasm of this study was comparatively more pungent than the above levels. A study carried out in Chapadão do Lageado (Alto Vale do Itajaí-SC), where the same analysis method was used, reported a variation in pungency from 4.84 to 7.61 µmol g^-1 (^{Schunemann et al., 2006}). A cultivar common to the germplasm of both studies was Bola Precoce, for which a lower pyruvic acid content (5.80 µmol g^-1) was detected in our study (10.16 µmol g^-1). Possibly, the local conditions and sulfur previously available in the soil prior to fertilization may have significantly increased the pungency of the onion genotypes tested in Ituporanga-SC.

Generating new genotypes to meet current consumer demands is one of the major challenges for breeding, more so, when no information is available for the trait of interest within the accessions available in the program. Consequently, the characterization must be initiated. However, considering more than one trait simultaneously and thus the existing covariation between the response variables that make up the ideotype can significantly improve efficiency and reduce the time invested for genotype selection. In addition, it would make no sense to add a different characteristic to a crop (low pungency) without maintaining the expected levels of bulb yield (in this case). This study established, within the available genetic variation in the EPAGRI onion germplasm, that four heterotic groups can be recombined in the future to select the best genetic background. Some of the suggested hybridizations are i) crosses of low-pungency and contrasting genotypes for the trait bulb yield (LA11 × LB19) (Tab. 3). ii) crosses of genotypes contrasting for the trait pungency, but similar in bulb yield (LA11 × LB24), iii) crosses of male-sterile lines with greater pungency than the maintainer line (LA14 × LB19) and iv) crosses of not very pungent male-sterile lines that are stable regardless of the cultivation system (LA15 × LB19).

CONCLUSIONS

This study established, within the available genetic variation in the EPAGRI onion germplasm, that four heterotic groups can be recombined in the future to select the best genetic background.