The primary pest affecting maize in Europe is the Mediterranean corn borer (MCB) Sesamia nonagrioides (Lef.). Stem corn borer larvae feed on the stem pith of maize producing tunnels that cause yield losses of approximately 30%, equivalent to worldwide losses of 311.3 million tons every year (Meissle et al., 2010). Tunnels produced in the stalk pith interfere with assimilate movement of nutrients toward the developing ear and increase lodging rate (López et al., 2001). One approach to pest control involves the cultivation of resistant and/or tolerant varieties, particularly in regions where the cultivation of transgenic maize is prohibited or heavily regulated, such as in Europe (Farinós et al., 2004). Genetic improvement of pest resistance relies on attaining varieties that experience reduced damage produced by insects. However, the most resistant varieties could not be the most productive in terms of grain yield. Even more, certain materials have shown a negative correlation between resistance to stem borer attack and grain yield (Sandoya et al., 2008; Butrón et al., 2012). In those cases, selection of varieties that are able to produce higher grain yields under high insect pressure could be a better alternative (Butrón et al., 1999; Ordas et al., 2012). However, at a particular site, the pest infestation level is influenced by multiple factors, exhibiting erratic behavior across years, and differences among varieties for grain yield could be highly affected by the infestation rate. Therefore, it is important to assess varieties’ performance across different scenarios concerning insect attack intensity in order to select varieties with high stable yield.

In order to optimize the development of varieties with reduced damage by insect attack and/or high yield across different scenarios, it is imperative to comprehend the inheritance of the traits involved and to study the regions of the genome or quantitative trait loci (QTLs) associated with resistance using different mapping populations in order to advance in the complete QTLome knowledge for these highly relevant traits (Salvi and Tuberosa, 2015). Accordingly, we can devise the optimal selection method (whether genomic or phenotypic), determine the feasibility of managing any QTL through marker-assisted selection (MAS), or even investigate genes involved in the inheritance of insect resistance. In the specific case of resistance to corn borers, QTLs have been widely sought using biparental populations and linkage-based QTL mapping (Ordas et al., 2009; Samayoa et al., 2014; Samayoa et al., 2015a, 2017; Jiménez-Galindo et al., 2017). However, the use of high-density molecular markers is now widespread due to its low cost, and genome-wide association studies (GWAS) are being conducted for a wide range of traits, including maize resistance to insects (Butrón et al., 2018). GWAS for resistance to the MCB have been conducted using diversity panels of inbred lines (Samayoa et al., 2015b) and multiparent advanced generation intercross (MAGIC) populations (Jiménez-Galindo et al., 2019).

Although many QTLs for resistance to stem corn borers have been found, and molecular markers significantly associated to resistance have been identified, MAS (Flint-Garcia et al., 2003; Samayoa et al., 2019) and genomic selection (Gesteiro et al., 2023) have been scarcely used to improve maize resistance to insects. External validations of detected QTL to assess QTL reliability are advisable before implementing MAS. QTL reliability can be affected by a significant QTL × environment interaction but also by low phenotyping precision.

Phenotyping genetically wide panels of inbred lines or RILs is challenging due to the difficulty of controlling the high experimental error associated with large trials; mostly when evaluations are often done using incomplete block or augmented designs in which most genotypes are assessed only once in each trial. Moreover, assessing plant traits related to performance against pest attack involves managing pest populations that could show variability for survival, voracity, or any other trait that would affect plant performance. Therefore, the high experimental errors associated with these evaluations could greatly interfere with QTL identification.

In addition, reliable additive-effect QTL could not be useful to discriminate among heterozygous genotypes if heterotic effects can counterbalance additive effects. Therefore, QTL validation should be done using inbred lines as well as hybrids to really ascertain the practical value of detected QTLs. Hence, the objective of this study was to determine the reliability and practical usefulness of QTL for resistance and yield reported by Jiménez-Galindo et al. (2019). Identification of QTL was based on a GWAS approach applied to a maize MAGIC population evaluated using augmented designs in a 2-year evaluation. For validation purposes, we selected inbred line sets at the tails of distributions for resistance and yield in the whole MAGIC to maintain genetic variability for both traits. These inbreds were evaluated per se and crossed to a tester (A638) using lattice designs. The results obtained in this study will contribute to expanding the maize QTLome knowledge for traits highly relevant for maize breeding as the they are related to pest tolerance and resistance.

The experimental design used in Jiménez-Galindo et al., 2019 was an augmented design, originally conceived to evaluate a large number of genotypes with the objective of discarding those that presented poor performances. In this study, we have selected a reduced set of inbred lines (IVS) at the tails of distributions for resistance and yield in the whole MAGIC population in order to maintain trait variability but limiting the number of individuals. This allowed us to improve the phenotyping reliability as the new set could be evaluated using an α-lattice design with two replications instead of an augmented design. The α-lattice experimental design provides enhanced control over experimental variability among replications and blocks, improving phenotyping precision and repeatability and increasing the genotypic/phenotypic variation ratio. As a result of the improved experimental design and more accurate phenotyping, we observed a gain in heritability values in both validations sets, inbred lines and hybrids, compared to heritability estimates in the whole MAGIG population (Jiménez et al., 2019). The increased inheritance estimates for tunnel length and yield in the IVS seemed to be a consequence of higher observed genetic variability; meanwhile, observed genetic variance was not increased in the HVS but the higher number of evaluations reduced the variability of hybrid means. As expected, owing to the reduced importance of additive genetic effects in a heterozygous background, we found that hybrids presented half or less than a half of the genotypic variation of the per se inbreds agreeing with (Michel et al., 2022). The reduction of genetic variance in the HVS compared to IVS was significant only for tunnel length because additive effects are the only significant genetic effects for this trait and, consequently, dominance and/or epistasis effects would not contribute to increase genotypic variability among hybrids as it could happen for yield (Butrón et al., 1999, Cartea et al., 1999).

The QTL validation approach used in the current study greatly differed from approaches used in other works, as those studies usually restringed genetic variation to a particular genomic region (single QTL) using near-isogenic inbred lines (NIL) or heterogeneous inbred families (HIF) (Huo et al., 2016; Jiménez-Galindo et al., 2018; Zhao et al., 2022; Yang et al., 2017). Evaluation of NIL and HIF genotypes is very valuable to advance in the isolation and cloning of the genes involved in the inheritance of a particular trait, but is not useful to validate the genetic weight of a particular QTL in a more real scenario in which variation for the trait is not limited. Conversely, results derived from the current validations could provide information of the usefulness of QTL and genomic predictions, both based on the evaluation of the whole MAGIC population, for implementing marker-assisted or/and genomic selection. In addition, the validation of additive-effect QTL detected by Jiménez-Galindo et al. (2019) has been done in homozygous and heterozygous backgrounds, which would allow us to assess the value of these QTLs for improved hybrid development. For validation in a heterozygous background, IVS inbreds were crossed to the Reid inbred line A638. It has been previously shown that using testers from complementary heterotic groups to the heterotic group or groups of the evaluated inbreds is superior for detecting/validating additive-effect QTLs than testers more genetically related to the inbreds (Frascaroli et al., 2009).

In the present study, almost three-fourths of QTLs found as associated to tunnel length by Jiménez-Galindo et al. (2019) were validated. However, the p-values of current associations were high, excluding the suitability of these QTLs for performing MAS but emphasizing that genomic predictions based on genotype–phenotype associations in the whole MAGIC population could be valuable for genomic selection. QTLs in the region 6.05–6.07 were validated in both sets, and this genomic region has also been associated with resistance to stalk tunneling in different mapping populations and environments. In addition, this region encompasses several QTLs for cell wall digestibility (Ralph et al., 2004) and, consequently, deserves to be deeply studied to identify the genes behind those associations. Ten out of 16 (approximately ⅗) QTLs for grain yield found in the whole MAGIC population were validated. The QTL validation ratio was inferior for yield than for tunnel length as expected because dominance and/or epistatic effects are equally or more important than additive effects for yield and could mask additive effects (Butrón et al., 2009). Interestingly, the additive-effect QTL on the long arm of chromosome 8 detected by Jiménez-Galindo et al. (2019) could not be validated in any validation set and, in the same chromosome arm, several authors have found significant QTLs for mid-parent heterosis for yield (Schön et al., 2010.; Tang et al., 2010; Larièpe et al., 2012; Samayoa et al., 2017). Again, p-values of associations found in the current study between markers linked to QTLs detected by Jiménez-Galindo et al. (2019) and grain yield were high, suggesting that genomic prediction based on the evaluation of the whole MAGIC population could be suitable to improve yield. Gesteiro et al. (2023) have already shown that genotypic selection for yield surpassed phenotypic selection, although genotypic and phenotypic selection for reducing tunnel lengths could be preferred to improve yield and resistance simultaneously.

Nine out of 21 and 6 out of 17 QTLs identified for tunnel length and grain yield, respectively, by Jiménez-Galindo et al. (2019) were validated across both validation sets, suggesting a moderate genetic correlation between per se and testcross performance of inbred lines. However, Zhang et al. (2011), although having estimated moderate to high genetic correlations between per se and testcross evaluations for plant and ear heights, showed that few QTLs were detected in both genetic backgrounds. In any case, the validation of a moderate proportion of QTL for tunnel length in both validation sets suggested that genomic predictions for tunnel length based on inbred line evaluations could be useful to develop improved hybrids (Ma et al., 2022). Michel et al. (2022) already estimated promising genomic value predictive abilities between r = 0.49 and r = 0.55 for models predicting maize hybrid traits that were trained with per se data.

In this study, we have used a novel validation QTL approach together with improved experimental designs that led to higher heritability estimates for both traits compared to the whole MAGIC population. A substantial proportion of QTLs associated with tunnel length and grain yield were validated across both sets. The current validations offer insights for guiding the implementation of marker-assisted or genomic selection strategies. Moreover, the study indicates the potential value of utilizing genomic predictions based on inbred line assessments to enhance the development of superior hybrid varieties.

Phenotyping genetically-wide panels of inbred lines or RILs is challenging due to the difficulty of controlling the high experimental error associated with large trials. Therefore, the high experimental errors associated with these evaluations could greatly interfere with QTL identification making necessary to validate those QTL before implementing any maker-assisted or genomic selection program. In such scenario, a novel validation approach has been used that allows to ascertain the reliability of QTL as well as its practical value to contribute to hybrid improvement.