Maize (Zea mays L.) is a major cereal crop in sub-Saharan Africa (SSA), where it covers more than 40 million ha of arable land. It is the most important crop for food security, income, and livelihoods for several million smallholder farmers across SSA, especially in eastern and southern Africa where nearly 85% of the maize produced is used as food (Shiferaw et al., 2011). Maize production in SSA was approximately 70 million metric tons in 2020 (FAOSTAT, 2021) and is largely produced by smallholder farmers. Despite its wide cultivation, the average maize yield in SSA is approximately 2.0 t ha^-1, which is far below the global average of approximately 5.8 t ha^-1 (Erenstein et al., 2022). The low maize yield is attributed to several factors including the frequent occurrence of drought, poor soil fertility, inadequate use of inputs such as improved seed and fertilizers, the impact of pests and diseases (Prasanna et al., 2021), and parasitic weeds (Menkir et al., 2012). Maize diseases of major economic importance in SSA include fungal (Asea et al., 2002; Menkir and Ayodele, 2005; Vivek et al., 2010; Sserumaga et al., 2020) and viral diseases (van Rensburg et al., 1991; Kyetere et al., 1999). Many of the pests and diseases of maize in SSA have become endemic to the region but there have been cases of new transboundary pests and diseases in recent years, affecting the food security and livelihoods of several million resource-constrained smallholder farmers (Prasanna et al., 2020). An example of a transboundary disease occurrence in Africa was the emergence of maize lethal necrosis (MLN) disease in SSA (Wangai et al., 2012; Mahuku et al., 2015a), a disease that was first reported in the Americas in the 1970s (Niblett and Claflin, 1978; Uyemoto et al., 1980).

Maize lethal necrosis was first reported in Kenya in 2011 but has since spread to several other eastern African countries between 2012 and 2018 (Wangai et al., 2012; Mahuku et al., 2015b; Adams et al., 2014; Lukanda et al., 2014; Mudde et al., 2018). This viral disease is caused by the coinfection of maize plants by the maize chlorotic mottle virus (MCMV) and any one of the viruses from the family Potyviridae, such as sugarcane mosaic virus (SCMV), maize dwarf mosaic virus (MDMV), or wheat streak mosaic virus (WSMV) (Redinbaugh and Stewart, 2018; Braidwood et al., 2018; Mwatuni et al., 2020). Recent studies revealed that Johnson grass mosaic virus (Stewart et al., 2017) and maize yellow dwarf virus (Wamaitha et al., 2018) in association with MCMV, cause MLN. The emergence of MLN in eastern Africa is attributed to the entry of MCMV into the region (Prasanna et al., 2020) since the presence of SCMV was reported in maize in East Africa much earlier (Kulkarni, 1973; Louie, 1980).

MLN has seriously affected maize grain yield and production in eastern Africa to varying levels. Based on community surveys in 2013, total maize loss in Kenya due to the MLN outbreak was estimated at 0.5 million metric tons year^-1 or 22% of the average annual production, with a value of approximately USD 180 million (De Groote et al., 2016). In 2018, the total quantity of maize lost in Kenya was estimated to be approximately 0.17 million metric tons equivalent to approximately USD 51 million (De Groote et al., 2021). Strategies such as crop rotation to break the disease cycle, the use of clean seed, and vector control have been proposed to manage MLN in SSA. The most economically viable and environmentally sustainable approach to control and manage MLN is the development of resistant or tolerant maize varieties. The economic value of adopting MLN tolerant hybrids was estimated at USD 195–678 million in Kenya and USD 245–756 million in Ethiopia depending on adoption levels of 25–75% (Marenya et al., 2018), suggesting a considerable benefit to farmers in utilizing MLN resistant varieties.

The International Maize and Wheat Improvement Center (CIMMYT) in collaboration with national partners, initiated screening of its germplasm stock and from other sources for resistance to MLN in 2012. A few sources of MLN resistance were identified, and introgression of resistance into CIMMYT’s elite germplasm was initiated. The key lines that have been used for introgression of MLN resistance alleles into CIMMYT’s mid-altitude adapted maize germplasm are the yellow lines KS23-5 and KS23-6 from Kasetsart University, Thailand (Jones et al., 2018; Prasanna et al., 2020). These two lines were extracted from KS23(S)C5, a population that had undergone five cycles of S1 recurrent selection (Jampatong et al., 2010) and have resistance to maize mosaic virus (Brewbaker, 2009). Through pedigree breeding and the use of doubled haploid (DH) technology, several inbred lines have been developed, screened under MLN disease pressure to identify resistant lines with adaptive traits, and utilized for hybrid development. Some of the elite CIMMYT lines have also been converted into MLN resistant versions using the major quantitative trait locus (QTL) (qMLN06.157) from KS23-6 (Prasanna et al., 2020; Murithi et al., 2021).

Knowledge of the genetic basis of resistance to diseases is important in developing breeding strategies. Diallel studies have been used to investigate the genetics of virus disease resistance in maize (Josephson and Naidu, 1971; Loesch and Zuber, 1972; Naidu and Josephson, 1976; Rosenkranz and Scott, 1987; Mutengwa et al., 2012; Beyene et al., 2017; Nyaga et al., 2020). Beyene et al. (2017) and Nyaga et al. (2020) reported that additive gene action is more important than nonadditive gene action for MLN resistance and identified some MLN tolerant inbred lines. Studies on the genetics of maize virus diseases have focused mainly on combining ability for disease parameters but not reciprocal effects that could impact hybrid development plans and disease resistance improvement strategies. To date, studies on MLN in tropical maize have used early to intermediate maturity germplasm. The upper mid-altitude ecologies of Eastern Africa are a major maize production area, where late-maturity maize is the preferred type. However, MLN disease is increasingly affecting this region. Therefore, incorporating resistance to MLN into late-maturity maize germplasm is crucial for sustaining production in this region. Using two different sources of resistance to MLN, alleles for resistance have been introgressed into CIMMYT’s late-maturity maize germplasm suitable for the upper mid-altitude ecology of Eastern Africa. Limited to no information on the combining ability and quantitative genetic parameters of MLN resistance in adapted late maturing tropical maize germplasm has been reported. The objectives of this study were to (i) estimate the combining ability of MLN resistance and other traits among 14 late-maturity inbred lines and assess the importance of reciprocal effects, and (ii) investigate the quantitative genetic basis of MLN resistance in tropical maize.

To effectively combat the spread of MLN in SSA, breeding for resistant varieties coupled with other strategies like clean seed as part of an integrated disease management strategy is important. The development of MLN resistant parental inbred lines with favorable alleles for key agronomic and adaptive traits is critical for the identification of high-yielding adapted MLN-resistant maize varieties. Our objective was to understand the genetics of resistance to MLN and other key traits in the late maturity maize germplasm, information that can be utilized in designing a breeding strategy for MLN resistance. The lines used in this study differed in terms of their source germplasm (drought tolerance, adaptation, and MLN resistance), and selection history with some lines developed through pedigree breeding and others developed through the DH technique. In the present study, highly significant genotypic differences for MLN disease resistance parameters and agronomic traits under artificial inoculation conditions were revealed. Highly significant genotypic differences were detected for all traits except EPP under rainfed conditions. These findings indicate that there was sufficient genetic variability for most of the traits studied in this set of germplasm, implying that progress from selection and ultimately genetic gain can be made in breeding for MLN resistance and other adaptive traits. Access to diverse germplasm resources has enabled CIMMYT to make progress in the development of MLN-resistant germplasm to combat the disease in Eastern Africa (Prasanna et al., 2020).

The broad-sense heritability for MLN disease resistance parameters was high (H ² = 0.92–0.95) in this study which suggests that a greater proportion of the observed phenotypic variation in this germplasm was due to genetic variability, and that breeding for MLN resistance can lead to considerably high genetic gains from selection for these traits. This conclusion is supported by the moderately high narrow-sense heritability (h ² = 0.52–0.56) for MLN disease resistance parameters which suggested that more than 50% of the genetic control was attributed to additive genetic effects. The implication is that the resistance of this germplasm to MLN can be improved relatively rapidly. There are no published reports on narrow-sense heritability for MLN disease resistance parameters to compare our results with. High broad-sense heritability estimates for virus diseases of maize have been previously reported, e.g., for MLN (Beyene et al., 2017), MCMV (Jones et al., 2018), and SCMV (Pokorny and Porubova, 2006).

The genetics of virus diseases have been investigated in tropical maize using a variety of intermediate maturity maize germplasm (Mutengwa et al., 2012; Nyaligwa et al., 2017; Beyene et al., 2017; Nyaga et al., 2020). In our study, we used maize in the late-maturity category to investigate the combining abilities and reciprocal effects in this germplasm. The results of this study revealed a greater contribution of GCA variance than of SCA variance for all traits under artificial MLN inoculation and for agronomic and disease traits under rainfed conditions. This suggests a preponderance of additive genetic effects in the inheritance of these traits in this germplasm. This result is consistent with findings for MLN in a study under artificial inoculation (Beyene et al., 2017), for maize streak virus (Mutengwa et al., 2012; Kim et al., 1989) and for maize dwarf mosaic virus (Loesch and Zuber, 1972; Rosenkranz and Scott, 1987). Both additive and dominant genes have been reported to condition virus resistance in maize (Naidu and Josephson, 1976; Zambrano et al., 2014). With a predominance of GCA over SCA, early testing may be more effective, and promising hybrids can be identified and selected mainly based on the prediction from GCA effects and the most resistant hybrids can be obtained by crossing the parents with the highest GCA (Baker, 1978; Makumbi et al., 2011). To improve both MLN resistance and grain yield potential, breeders can select lines with significant negative GCA effects for MLN resistance, significant positive GCA effects for grain yield, and significant negative GCA effects for foliar diseases to create new hybrid combinations for testing in the relevant target environments. Performance of hybrids can also be predicted based on the performance of single crosses (e.g. Zuber et al., 1973). Recurrent selection that emphasizes GCA can be an effective strategy to improve MLN resistance in this germplasm, as has been recommended for other virus diseases of maize (Josephson and Naidu, 1971; Kim et al., 1989; Mutengwa et al., 2012). With the implementation of forward breeding for MLN and other diseases such as MSV, the identification of MLN resistant lines that possess other adaptive traits for hybrid development should lead to faster genetic gains.

The consistency observed in the GCA effects across the four time points suggests rating genotypes for MLN resistance on multiple dates may not be necessary. Evaluating genotypes at 35 days (MLN3) and 42 (MLN4) days after inoculation should be sufficient for accurate assessment of resistance. The results revealed that five inbred lines had consistently significant desirable (negative for reduced disease) GCA effects for the four MLN resistance scores and AUDPC. This suggested that these lines have favorable alleles for resistance to MLN and can slow disease progression. Inbred line 6 (CKL176616) which had the largest desirable GCA effects for the four MLN resistance ratings derived its resistance alleles from KS23-5 unlike three of the four lines (CKL181281, CKL181379, and CKL182037) whose resistance was from KS23-6. Interestingly, among the five inbred lines, inbred line CKL175755 is not known to have any pedigree breeding history of resistance to viruses. Two other lines (CKL175798 and CKL176082), also without a background of virus resistance showed significant negative GCA effects for MLN4. Detailed studies (e.g., Jones et al., 2018) under artificial MLN, MCMV, and SCMV inoculation in net houses are needed to confirm the response of these lines as they may offer an additional source of alleles for MLN resistance. To date, only lines KS23-5 and KS23-6 have validated QTLs for MLN resistance (Murithi et al., 2021; Awata et al., 2021), but QTLs for resistance to MCMV, one of the two viruses that cause MLN have been mapped in other lines (Jones et al., 2018). In our study, the inbred line CKDHL120918 had significant positive GCA effects for MLN but this line was reported to have significant negative GCA effects for MLN in an earlier study (Beyene et al., 2017). The differences in GCA effects of this line between the two studies could be due to variation in the diallel method used, as different methods can impact GCA estimates (Fan et al., 2014). We used Method 3 in this study, while Beyene et al. (2017) used Method 4 of Griffing (1956). Four inbred lines (CKL181281, CKL182037, CKL176616, and CKL175798) out of the five that had significant negative GCA effects for all MLN disease resistance parameters and AUDPC also expressed significant positive GCA effects for GY under artificial MLN inoculation. These findings demonstrate progress made in developing lines with favorable alleles for MLN-resistance (reduced disease susceptibility) while also combining beneficial alleles for GY under both MLN-infected and rainfed conditions. Furthermore, two of the four lines (CKL181281 and CKL182037) had significant positive GCA effects for GY and significant negative GCA effects for Turcicum leaf blight under rainfed conditions. These two lines with desirable GCA effects for GY and disease resistance across different conditions have the potential as inbred line testers for the MLN breeding program if the right testing strategies (e.g. Castellanos et al., 1998; Pswarayi and Vivek, 2008) are used to confirm their suitability as testers. Suitable testers must correctly classify and discriminate efficiently among test entries (Rawlings and Thompson, 1962).

The use of reciprocal crosses provides a quantitative method to assess the contribution of maternal effects in the inheritance of a trait. When reciprocal differences are strong, parental inbred performance affects the choice of a female parent in a hybrid (Mann and Pollmer, 1981). In this study, reciprocal and maternal effects were not significant for MLN disease resistance parameters, indicating that the disease parameters recorded were not influenced by maternal effects or cytoplasmic inheritance. This suggests that a line that is resistant to MLN can be used either as a female or male parent in a hybrid combination, although other traits such as seed producibility and pollen production must be considered. According to Roach and Wullf (1987), reciprocal crosses have similar nuclear genetic contributions, and any divergence in the performance of reciprocal pairs is due to a maternal or perhaps a paternal effect. The absence of maternal influence on the MLN disease resistance parameters signifies the predominance of additive gene action for MLN resistance since maternal effects are assumed to result from nonadditive gene action. Furthermore, maternal effects can potentially decrease the accuracy of genetic studies. Both cytoplasmic and nuclear maternal genetic effects may increase the observed genetic variance, but if the trait is fully controlled by maternal factors, they could curtail the response to selection (Roach and Wullf, 1987). Therefore, lack of maternal effects in this study suggests that the response to selection for MLN resistance will be minimally impacted. Our results indicated significant reciprocal effects for several agronomic traits including DTA and GY under MLN inoculation, and DTA, EPP, and GY under rainfed conditions. These findings were consistent with other reports of significant reciprocal effects for GY in maize (Ordás et al., 2008; Kagoda et al., 2011). However, our results contrast with those of Jumbo and Carena (2008) and Machida et al. (2010) who reported no significant effects for GY.

The development of multiple stress tolerant maize germplasm is an objective of many breeding programs in SSA, and this requires the selection of parental lines with suitable breeding values for the target traits. In SSA, stress tolerances including tolerance to foliar diseases (TLB and gray leaf spot), viral diseases (MSV and MLN) and low soil fertility and drought stress are required in certain combinations in hybrids for commercial production. The results showed that several lines had desirable GCA effects for a combination of some of the stresses. These lines should be tested in hybrid combinations under managed abiotic stress conditions mainly drought and low soil fertility following established protocols (e.g. Njeri et al., 2017; Makumbi et al., 2018b). Lines that have the highest breeding values across several stresses can be used as parents in biparental populations for development of new inbred lines. Further improvement of tropical germplasm using the new lines can be achieved with the application of modern tools and techniques such as DH, genomic selection, and marker-assisted selection with improvements in phenotyping methods to increase the rate of genetic gain and develop new multiple stress tolerant inbred lines (Cooper et al., 2014; Cobb et al., 2019; Prasanna et al., 2021; 2022).

The diallel analysis model of Hayman (1954a, b) provides estimates of several quantitative genetic parameters for the traits of interest. This method of analysis has been applied to the diallel analysis of quantitative traits alongside Griffing (1956) method to gain a deeper understanding of the inheritance of a trait beyond what a single method of diallel analysis can provide (e.g. Naidu and Josephson, 1976; Hamid et al., 1982; Betrán et al., 2003; Kagoda et al., 2011). The current study is the first to investigate the quantitative genetic parameters of MLN resistance using Hayman’s model. The results indicated that additive effects had a greater contribution in the inheritance of MLN resistance compared to dominance effects based on Hayman’s method, a result similar to that obtained using Griffing’s method for combining ability analysis. A study on the inheritance of Helminthosporium leaf spot in maize (Hamid et al., 1982) also reported congruence between results from Hayman’s and Griffing’s models for several disease resistance parameters. Our results revealed that the alleles controlling resistance or susceptibility to MLN were not equally distributed among the lines used in this study (H₂ < H₁). This was evident through the distribution of the parental arrays along the regression line on the W_r-V_r graph. The 14 parents were therefore unique with respect to the dominance and/or epistatic effects of the genes they possess (Allard, 1956). The W_r-V_r graphs for MLN3 and MLN4 were similar in the placement of the arrays along the regression line. The group of lines that were among the most resistant to MLN (P2 [CKL181281] and P5 [CKL182037]) were located far from the origin at the upper end of the regression line, which indicated that the alleles conditioning resistance in these lines were mostly recessive. These inbred lines had desirable GCA effects for MLN resistance and AUDPC. A group of four lines (P7, P8, P9, and P10) was located closer to the origin of the regression line, which indicated that this set of lines possessed more dominant alleles. Some of these lines, especially P8 and P9, showed significant negative GCA effects for MLN and AUDPC. This finding is of particular interest and necessitates further investigation to understand the genetic basis of resistance in these lines given that resistance to MLN has been reported to be recessive. The positioning of P8 and P9 with respect to other lines with significant negative GCA effects for MLN and exhibited resistance such as P2, P3, P5 and P6, was surprising. This may suggest that these lines exhibited similar phenotypes through different genetic mechanisms (Luckett, 1989). These lines should be tested under artificial inoculation with individual MCMV and SCMV isolates in a net house for better discrimination in terms of resistance to these viruses.

This study revealed that additive genetic effects contribute significantly to the inheritance of MLN in late maturity germplasm adapted to eastern Africa. Five inbred lines, three of which were derived from introgression of MLN resistance from KS23-5 and KS23-6 exhibited significant desirable GCA effects for MLN resistance and GY under artificial MLN conditions. Three of these lines also showed significant desirable GCA effects for GY under disease and rainfed conditions and have the potential to contribute to the development of multiple stress tolerant hybrids for the target product profile. We identified four inbred lines with desirable GCA effects for MLN resistance despite having no known breeding history of virus resistance. Detailed studies under artificial inoculation with individual viruses (MCMV and SCMV) and MLN should help to decipher the genetic basis of resistance to MLN in this group of lines. Reciprocal effects were of minor importance; therefore, breeding programs can use any MLN resistant inbred line as a female or male in hybrid combinations without a significant effect on the MLN response in the final product. The graphical method of analysis revealed the distribution of the lines in relation to the abundance of recessive or dominant alleles, and this information will be useful for selecting parents for biparental populations and hybrid development.