CROSSING AMONG MAIZE POPULATIONS FOR RESISTANCE TO FALL ARMYWORM

Among the insects infecting the maize (Zea mays L.) crop in Brazil, the fall armyworm (Spodoptera frugiperda Smith, 1797, Lepdoptera: Noctuidae) is considered one of the most important because it causes the highest damage to yield. Genetic resistance to the fall armyworm has be an effective control strategy. The main objective of this work was to evaluate new germplasm sources for resistance to the fall armyworm, the key pest for the maize crop in Brazil. A partial diallel design between 20 varieties of Brazilian germplasm and nine exotic and semi-exotic varieties of different origin was used. The 180 crosses and 29 parental varieties along with two commercial checks were evaluated in three locations in the State of São Paulo State (Brasil). Fall armyworm resistance (FAWR) under artificial and natural infestations, grain yield (GY), and plant height (PH) were analyzed. The populations CMS14C and MIRT, and hybrid São José x MIRT showed the highest resistance, with values of 1.8, 1.7 and 1.4, respectively. Populations PMI9401 and PR91B, and the hybrid CMS14C x (B97xITU) had best yields, with 4893, 3858 and 5677 kg ha, respectively. Heterosis ranged from –28% to 47% for FAWR and from –21% to 125% for GY, with mean values of –0,43% and 31%, respectively. Genotype by environment interaction was not significant for FAWR. The effects of varieties and heterosis were significant for all traits, showing that both additive and dominance effects may be important as sources of variation. For FAWR, only specific heterosis presented significance, suggesting strong genetic divergence between specific pairs of parental populations. Brasilian populations PMI9302 and São José, and the exotic population PR91B presented high performance per se, and also in croses for FAWR and GY. Crosses PMI9401 x (Cuba110 x EsalqPB1) and São José x MIRT presented high specific heterosis effects for both characters. These populations can be useful to be introgressed in maize breeding programs.


INTRODUCTION
Among the insects infecting the maize crop in Brazil, the fall armyworm (Spodoptera frugiperda Smith, 1797) is considered one of the most important because it causes the highest economic damage.Larvae feed mainly on leaves, reducing the photosynthetic area of the plant and affecting indirectly the grain production.Losses The genetic potential of different materials are usually studied in controlled crosses; in this sense, the diallel mating scheme has been useful and largely used to estimate the potential value of genotypes per se, their combining ability and heterosis effects (Hallauer & Miranda Filho, 1995).Miranda Filho & Geraldi (1984) suggested the partial diallel cross between two fixed sets of varieties or lines, as an adaptation of the complete diallel model of Gardner & Eberhart (1966).Geraldi & Miranda Filho (1988) concluded that the adapted model of Gardner & Eberhart (1966) is more informative than the partial diallel adapted from Griffing's (1956) model.The objectives of the present work were to evaluate maize germplasm from Brazil and exotic or semi-exotic germplasm for their agronomic traits and resistance to fall armyworm; to estimate genetic parameters for agronomic traits and resistance to fall armyworm using the partial diallel scheme (Miranda Filho & Geraldi, 1984); and to select populations with high level of resistance and desirable agronomic traits to be indicated for their use in breeding programs.

Germplasm
The identification of populations used in this work is shown in Table 1, divided in two groups: Group I, comprising 20 Brazilian populations previously introduced in the NAP (Núcleo de Apoio à Pesquisa) Project; and Group II, represented by nine exotic and semi-exotic populations of diverse origins.The 20 populations of Group I were selected for fall armyworm resistance, from 113 populations of the NAP Project, under artificial infestation (data not shown).
Exotic populations were introduced on the basis of their pattern of resistance to S. frugiperda or other corn borer species in their sites of origin.Some of the exotics (MIRT and PR91B) were used directly and others were previously crossed with local populations to develop semi-exotics better adapted to local conditions.Two commercial hybrids (MASTER: resistant; and P3041: susceptible) were used as checks, which were chosen as the extremes for resistant pattern from a set of 12 cultivars previously evaluated under artificial and natural infestation (data not shown).
The use of resistant cultivars has been suggested by several authors (Sprague & Dahms, 1972;Wiseman & Widstrom, 1992) as an efficient and feasible method to control armyworm infestation, either by decreasing the population of insects or by increasing the efficiency of insecticide application, thus reducing the amount of chemical products and also reducing the negative impact on environment.
In spite of the great genetic variability of the maize germplasm in Brazil, little is known in relation to the genetic potential of different materials as sources of resistance to insect infestations.The characterization of germplasm and the identification of new sources of resistance to be used in the development of new commercial varieties become more important when faced to the presence or the increase of insects attacks.
Breeding for resistance to fall armyworm has had a slow development until the advent of techniques for artificial infestation that allowed uniform evaluations of large number of plants in the field.Such techniques were developed by Ortega et al. (1980) and modified by Wiseman & Widstrom (1980).Artificial infestation has the advantage to provide more uniform levels of attack, with approximately the same number of larvae per plant.The evaluation under such conditions are really necessary when the high natural infestation does not occur in the experiment.On the other hand, there are environments where the population of insects remain at a high level every year, so that natural infestation is sufficient to assure evaluation with the desired precision.In the evaluation under field conditions, visual scales are used, which take into account the type of damage (size and number of lesions) and the damaged tissues (leaves and/ or whorl) (Carvalho, 1970;Davis et al., 1992).
The genetic base of the resistance to fall armyworm has been demonstrated to be of quantitative nature and the variation is due mainly to additive genetic effects although the presence of dominant gene action cannot be ruled out (Williams et al., 1989;Widstrom et al., 1992;Widstrom et al., 1993;Guimarães e Viana, 1994e Williams et al., 1995).The identification of new sources of resistance must be a continuous process in any breeding program for resistance to insects aiming the introgression of new genes into the genetic base of the commercial cultivars.The incorporation of new genotypes, either local or exotic, in the evaluations increase the chances for identification of mechanisms and/or genes for resistance that were not previously available.In Brazil, several studies have been conducted for the evaluation of resistance to fall armyworm, either under natural infestation (Fornasieri Filho et al., 1980;Lara et al., 1984;Marques et al., 1988) or under artificial infestation (Viana & Potenza, 1991;Viana & Guimarães, 1994;Nishikawa, 1999).

Evaluated traits
The resistance to the fall armyworm (RFA) was evaluated in EXP [1] under artificial infestation and in EXP [2] under natural infestation.Artificial infestation in the field was made by application of young larvae mixed with corn cob grits using a manual dispenser "bazooka" (Wiseman et al., 1980).Infestations were made in eight plants per plot in the stages V6 to V8, with approximately 15 larvae per plant.In EXP [1] visual evaluation of the damage was at 7 and 14 days after infestation.In EXP [2] the first evaluation was 7 days after the stages V6-V8, following another evaluation one week later.In both experiments a scale of notes (0: non-damaged plant to 5: plant whorl completely destroyed) was used (Carvalho, 1970).Data for statistical analysis was represented by the average of two measurements, as suggested by Davis et al. (1992).Grain yield (GY: total grain weight of the plot, in kg ha -1 ) and plant height (PH: mean of ten plants per plot, in cm) were also included for analysis of EXP [2] and EXP [3].

Statistical analysis
Preliminary analysis of variance for each experiment and combined analysis over experiments were performed for all traits.The analysis of variance of the partial diallel tables combined over experiments followed the model (Miranda Filho & Geraldi, 1984;Belluci, 1994): where Y ijk is the mean of the cross between i th population (Group I) and j th population (Group II) in the k th experiment; µ is the mean of the two groups; α= 1, α = -1 and α = 0 for parental varieties of groups I and II and crosses, respectively; d is a measure of the difference between group means; v i and v j are the effects of populations for Groups I and II, respectively; l k is the fixed effect of experiments; θ = 0 for parental populations (Y ii and Y jj ) and θ = 1 for crosses (Y ij ); h is the average heterosis over experiments; h i and h j are the effects of heterosis of populations for Groups I and II, respectively; s ij is the specific heterosis of the cross i x j; e ijk is the error term associated to the observed mean (Y ii , Y jj or Y ij ) in the k th experiment; all other terms in the model refer to the interaction of the respective effects with experiments (environments).

RESULTS AND DISCUSSION
Observed means are presented for three traits: RFA, combined for EXP.[1] and EXP.[2]; GY for EXP.

Preliminary analyses of variance
The preliminary analysis of variance, combined over experiments (data not shown), revealed differences between the hybrid checks for the variable RFA, indicating the effectiveness of the conditions of infestation for the discrimination between genotypes.Also, the nonsignificance of the interaction checks x experiments indicated that the reaction of the hybrid checks was relatively stable, even though the level of incidence of S. frugiperda was different between experiments.The non significance of the treatment x experiment interaction also reinforce the stable response of genotypes under the two different conditions of infestation.
For the variable GY significance was detected for both treatments and treatment x experiment interaction, indicating results should be interpreted for each experiment representing different environments.A similar pattern of significance was observed for the variable PH, but the mean square for treatments was 6.6 times greater than for the interaction treatment x experiment; although significant, the later showed a lower level of variation, as indicated by the F test (F = 1.4).For that reason, all parameters for PH were calculated with the mean over experiments.

Analysis of the diallel tables
Table 3 shows the analyses of variance for the three variables for the partial diallel table, represented by observed means of parental populations and their crosses.Differences were observed between groups for RFA (P < 0.01) and PH (P < 0.05).For RFA the difference was for advantage of Group II (lower mean) with an estimated parameter of d = 0.162.This result is explained by the fact that the NAP populations (Group I) were chosen for their pattern of resistance to S. frugiperda but, except for CMS14C, CMS23 and São José, all other were never submitted to selection for resistance.On the other hand, the exotic populations (Group II) were released for their high pattern of resistance to S. frugiperda and other Lepdoptera species.
Heterosis showed significance (P < 0.01) for all traits.For RFA the significant components of heterosis were population heterosis in Group I and specific heterosis; h = -1.04%)indicated non unidirectional dominance effects (Vencovsky & Barriga, 1992), which can be explained by the presence of different mechanisms of resistance, under the control of different genetic systems.In fact, the heterosis effects in crosses varied from negative (-28%) to positive (47%) in the crosses N14 x E3 and N3 x E3, respectively (Table 4).For GY and PH all the heterosis components showed significance.
The significance for GY heterosis variation is explained by the wide range of estimates, which varied from -20.0% (N12 x E6) to 138.3% (N14 x E5) and from -23.0% (N12 x E6) to 114.1% (N14 x E3) in EXP 2 and EXP 3, respectively (Table 5).Mid-parent heterosis higher than 20% was represented by 64% and 68% of the crosses in EXP 2 and EXP 3, respectively.Hallauer & Miranda Filho (1995) reported heterosis of 19.5% on the average of crosses involving 611 different parental varieties.High heterosis suggests a high concentration of homozygotes, as expected in races or populations continuously maintained through small population size (Miranda Filho, 1999).However, the lack of adaptation also may result in a low performance of the parental population and, if the other parent allows the recovery of an acceptable pattern of adaptation, the final result may be the expression of a high heterosis (Regitano Neto & Miranda Filho, 1999).Heterosis for PH is of lesser interest for a breeding programs and it has shown low expresion in population crosses (Hallauer & Miranda Filho, 1995).Observations (data not shown) indicated heterosis varying from -11.8% to 26.4% and only 16% of the crosses showed mid-parent heterosis higher than 10%.For GY the highest GCA effects in EXP 2 were observed for N20 (WP12), N3 (CMS14C) and N17 (PMI9401) in Group I and for E4 (B97 x ITU), E8 (Cuba 110 x ESALQ PB1), E2 (PR91B) and E3 (B95 x ITU) in Group II.In EXP 3 the highest GCA effects were for N17 (PMI9401), N5 (CMS454), N18 (São José) and N20 (WP12) in Group I and E8 (Cuba110 x ESALQPB1), E4 (B97 x ITU), E2 (PR91B) and E7 (Cuba113 x ESALQPB1) in Group II.

Selection of populations
Heterosis in variety crosses can be used directly in the first generation of the cross between two parental varieties.The parental populations of heterotic crosses also can be used for the development of inbred lines to be used in hybrid crosses.On the other hand, the synthesis of composites by intercrossing n parental varieties is a mean to retain heterosis in the population, because (n-1)/n of the heterosis component remain as part of the expected mean of the new population (Miranda Filho & Vencovsky, 1984).
the present work, expressive heterosis effects were detected for both RFA and GY, which can be utilized by selecting the outstanding crosses.The effects of varieties or populations also were detected in both groups of populations and selection based on this parameter leads to the identification of the most appropriate populations for intrapopulation recurrent selection.
The correlation coefficients between RFA and GY, calculated from measurements of the hybrid check, planted every fifteen rows in the experiments, were non significant: r = -11%, P = 0.46 under artificial infestation; and r = -1,4%, P = 0.92 under natural infestation.For this reason, selection of populations based only on GY is not recommended and, both traits (RFA and GY) must be taken into account.
Populations N4 (CMS23), N15 (PMI9302) and N18 (São José) from Group I, and E1 (MIRT) and E2 (PR91B), from Group II, showed the best performance for RFA in crosses.The good resistance level of crosses between these two selected groups is explained by the fact that N4, N15 and N18 exhibited the best estimates of g i in Group I and E1 and E2 showed the best level of resistance per se (v j ) in Group II.
For GY an analysis for each experiment (environment) was necessary because the existence of genotype x experiment interaction.In EXP 2, N4 did not presented a good performance in relation to GCA effect (g i ), while N15 and N18 showed good levels for GCA.In Interactions of effects with experiments were nonsignificant for RFA and AP (except populations I x experiments); for GY all the effects presented interaction with experiments.
Dominance effects can be an important source of variation for all traits.The relative contribution of each source of variation to the total sum of squares shows that dominance effects are differentiated among the studied traits.For RFA 32% of the total sum of squares are due to the variation of populations while 68% are attributed to the total heterosis; specific heterosis is responsible for 84% of the total variation of heterosis.For similar situations, Miranda Filho & Vencovsky (1984) considered that even if the additive and dominance effects are confounded in the effects of populations (v i and v j ), it is reasonable to suppose that the dominance effects are more expressive at the interpopulation (population cross) level.The detection of dominance effects in the control of the resistance to S. frugiperda was already reported by Williams et al. (1978), Widstrom et al. (1993), Guimarães &Viana (1994), andWilliams et al. (1995).For GY and PH, the proportion of the variation due to populations and heterosis were 46.9 : 53.1 and 77.8 : 22.2, indicating that the dominance at the interpopulation level is more expressive for GY than for PH.Similar results were reported by Miranda Filho & Vencovsky (1984) and Santos et al. (1994).Low heterotic response for PH has also been reported by many authors (Castro et al. 1968;Barriga e Vencovsky, 1973;Souza, Jr., 1981;Miranda Filho & Vencovsky, 1984).
The estimates of the effects (µ, d, v i , v j , h , h i , h j ; except s ij ) in the partial diallel model are presented in Table 6 for RFA and PH and in Table 7 for GY for both experiments.The effect of general combining ability (GCA) was estimated by g i = 1 2 v i + h i and g j = 1 2 v j + h j (Geraldi & Miranda Filho, 1988).Outstanding populations in Group I for GCA (negative g i ) of RFA were N4 (CMS23), N15 (PMI9302) and N18 (Tuxpeño Amarelo).CMS23 is a population that had undergone selection at EMBRAPA under artificial selection and was released for its good level of resistance to S. frugiperda.Results observed in the present work allowed the identification of populations, as mentioned above, to be used as source of resistance at levels similar to CMS23.
In Group II, two populations from CIMMYT (E1 -MIRT and E2-PR91B) exhibited the best values for resistance (negative v j ).The use of v j instead of g j (Table 4) for discriminating the best populations is explained by the non significance of population heterosis effects.MIRT (Multiple Insect Resistance for Tropical Regions) was selected at CIMMYT (Smith et al., 1989) and released for its high resistance to several species of the family Lepdoptera (Ostrinia nubilalis, Diatraea saccharalis, D. grandiosella, and S. frugiperda).PR91B is a population selected at CIMMYT (Colombia) for resistance to S. frugiperda.Group 2, E1 showed a low performance for GCA while E2 were among the best ones for this effect.In EXP 3, N18 and E2 were oustanding for GCA, while N4 and E1 were among the poorest ones.Finally, when considering the general performance for GY in crosses, N4 and E1 should be discarded among the populations selected for RFA.Therefore, N15, N18 and E2 are recommended as the most promising for incorporation in breeding programs, as indicated by their good level of resistance to S. frugiperda and acceptable agronomic traits.For plant height, N15 and E2 showed negative values for GCA and should be indicated as source of genotypes to lower plant architecture.
Selection of specific crosses also can be done from the presented results.High specific heterosis for both RFA and GY were shown by crosses N17 (PMI9401) x E8 (Cuba 110 x ESALQ PB1) and N18 (São José) x E1 (MIRT).Total heterosis in percent of mid-parent in these crosses were 9% and 27% for RLC and 26% and 24% for GY, respectively, the indicated crosses for their properties can be used directly for the exploitation of heterosis in reciprocal recurrent selection programs, or as base for the development of inbred lines to be used in crosses.

Table 1 -
Designation and origin of populations from the NAP Project (Group I) and exotic and semi-exotic populations (Group II).

Table 2 -
Observed means for resistance to fall armyworm (RFA), grain yield (GY) and plant height (PH) for populations and crosses.

Table 3 -
Mean squares and their significance in the analyses of diallel tables for the variables RFA, GY and PH combined over experiments.*, *, ns -significance levels for the F test: P < 0.01, P < 0.05 and non significant, respectively. 1Mean squares multiplied by 10 -5 *

Table 4 -
Estimates of total heterosis (h ij : upper valuer) and specific heterosis (s ij : lower values) for RFA in population crosses.

Table 5 -
Estimates of total heterosis (h ij ) for GY in two experiments(EXP 2: upper values; and EXP 3: lower values).
h ij : expressed in percent of mid-parent.