HETEROTIC RESPONSES OF TROPICAL ELITE MAIZE ACCESSIONS FROM LATIN AMERICA WITH BRAZILIAN TESTERS

Little emphasis has been placed on identifying new sources of tropical maize germplasm that can be used in breeding programs. Additional information on the performance and heterotic classification of tropical germplasm is needed. This study was conducted to identify elite maize accessions from Latin America that could contribute to increase the level of heterosis with the best heterotic patterns of Brazil. Seventy-two elite accessions from the Latin American Maize Project (LAMP) were crossed with the testers BR 105 (flint kernel Suwan background) and BR 106 (dent kernel Tuxpeño background). The 72 crosses plus 9 checks were evaluated using a simple 9 × 9 lattice in four locations that represent one tropical region. The combined analysis showed highly significant differences among treatments for ear weight and the interaction treatment × location was also significant. With both testers better crosses for ear weight were identified in relation to the double cross BR 201 (commercial check). With the tester BR 105, the high parent heterosis ranged from –28% to 26%, the accessions SE 032 and PE 001 were selected for further work. With the tester BR 106, the high parent heterosis ranged from –35% to 17% and the accessions PE 011 and Pasco 14 were selected for further work. The selected accessions will be improved through recurrent selection schemes to increase yield and improve agronomic traits. Afterwards they can be exploited, within and between heterotic groups, to produce highly productive hybrid combinations, or used per se as improved maize varieties.


INTRODUCTION
The international scientific community has recently called attention to the need for more efficient conservation and utilization of plant genetic resources.
Information about heterotic patterns, essential to maximize the use of genetic resources in breeding programs, has been increasing in recent years for tropical maize.Maize germplasm introductions from Mexico have shown good adaptation in Brazil.Their inclusion in breeding programs have resulted in commercial varieties that are good sources for extracting inbred lines in private or public research institutions (Naspolini et al., 1981;Gama et al., 1982;Santos et al., 1994).The Latin American Maize Project (LAMP) provided an excellent oportunity to select the best tropical accessions for participating countries (Salhuana et al., 1992), and now each country can utilize selected accessions as new gene pools for developing improved varieties and hybrids.
The objectives of this study were to: (i) identify among tropical elite maize accessions from the LAMP those that could contribute to increased levels of heterosis with the best heterotic pattern from Brazil; and (ii) incorporate the selected accessions into breeding programs.

MATERIAL AND METHODS
The Latin American Maize Project (LAMP) was an international project funded by Pioneer Hi-Bred International, Inc., to systematically evaluate maize genetic diversity for use in present and future breeding programs.The twelve countries involved evaluated more than 12,000 landrace accessions from the Americas (Salhuana et al., 1992).Because of differences in environments and growing seasons among the twelve countries the project was divided into five homologous areas (HA) according to altitude and latitude.Brazil was included in HA 1 which included tropical regions located between 0 o to 23 o N or S latitudes with altitudes below 1,200 meters above sea level (masl).The project was also divided into five stages of evaluation.Brazil evaluated 1,340 and 352 accessions, in the first and second stages, respectiveley.After the second stage the best 5% were selected for the third stage.The same protocol was used in the other countries, varying only the number of evaluated accessions.In the third stage, seeds of the best 5% were interchanged among countries with the same HA in order to make testcrosses.In this stage in Brazil, testcrosses using the testers BR 105 and BR 106 were made with five accessions from Bolivia, seven from Guatemala, fifteen from Mexico, two from Paraguay, fourteen from Peru, five from Venezuela, seven from the United States, and seventeen from Brazil.
The crosses were made in two isolated fields with different planting dates for accessions and testers.Accessions were detasseled and thus used as females.In 1991, the 72 crosses plus 9 checks were planted in a 9x9 simple lattice design at Sete Lagoas-MG (latitude 19 o 47'45"S and longitude 44 o 14'48"W), Goiânia-GO (latitude 16 o 40'43"S and longitude 49 o 15'14"W), Propriá-SE (latitude 10 o 12'40"S and longitude 36 o 50'25"W) and Janaúba-MG (latitude 15 o 48'09"S and 43 o 18'32"W).The common check in each experiment was the tester (BR 105 or BR 106) which was repeatedly interplanted in each incomplete lattice block.The means for these testers were considered as the superior parents for calculating heterosis estimates.
All experiments used fertilization and cultural practices of conventional maize farming systems.Data were recorded at four locations for plant height (cm), ear height (cm), number of broken stalks (B) and root lodging (R), ear number per plot, and yield measured as ear weight (Y) in t ha -1 .Data for 50% male and female flowering were only observed in Sete Lagoas-MG and Propriá-SE.Yield data were adjusted to 14.5% moisture based on grain moisture samples taken on the same day of harvest.Data for broken stalks and root lodging (B+R) were transformed to . Prolificacy was calculated using an ear index (EI, ear number per plot over final stand).
Analyses of variance were carried out for each location according to Cochran & Cox (1957).Adjusted treatments means were used for the combined analysis over the four locations.This analysis was done based on a randomized complete block design since lattice efficiency of each experiment was low.Location was considered as a random model effect and treatments were considered as fixed effects.Adjusted mean yield values were used for estimating heterosis in relation to the superior parent.LSD was calculated as t 5% r TxL MS / ) ( 2 .

RESULTS AND DISCUSSION
Analysis of variance of the means from the crosses with the tester BR 105, combined across four locations (TABLE 1), showed highly significant differences among treatments for all traits (P≤0.01).For treatments x locations interaction, significant differences were found (P≤0.01) for yield (YI), ear height (EH) and square root of broken stalks + root lodging (B+R), but no significant differences were detected for the plant height (PH) and ear index (IE).The crosses with the tester BR 106 (TABLE 2) also showed significant differences among treatments (P≤0.01) for all traits, but for treatments x locations interaction significant differences were only found for YI(P≤0.05) and B+R (P≤0.01).The four environments where the treatments were evaluated fall could be Scientia Agricola, v.58, n.4, p.767-775, out./dez.2001 classified as HA 1 despite the distance of 1,500 miles between the two furthest locations.Within HA 1, Propriá-SE and Janaúba-MG could represent one subregion, while Sete Lagoas-MG and Goiânia-GO could be included in another subregion.Within these subregions there are large differences in soil type, altitude, and climatic conditions.The first subregion has hot days and high night temperatures with irregularly distributed rainfall, while the second has more moderate climatic conditions and lower night temperatures.Thus, a treatments x locations interaction should be expected for traits that are affected by environment.Even for analysis by grouping subregions, significative differences were found due to treatments x locations interactions for traits Y and B+R (data not shown).Similar results have been shown in tropical regions within environments that are considered more uniform with adapted, improved maize populations (Naspolini et al., 1981;Gama et al., 1982;Santos et al., 1994).Large climatic variability is a problem in tropical regions.For this reason, it is usually recommended to select genotypes for specific environmental conditions to avoid losses in time and to more efficiently use limited financial resources.
TABLE 3 shows mean values for all traits for BR 105 crosses across the four locations, along with the high-parent heterosis (HPH) estimates for yield (Y).For Y means of crosses ranged from 4.4 t ha -1 to 7.7 t ha -1 , while these means for the checks ranged from 5.7 t ha -1 to 8.0 t ha -1 .The best cross (SE 032 x BR 105) produced 7.7 t ha -1 while the double cross commercial check BR 201 produced 6.8 t ha -1 .There was large variability for PH (231 to 267 cm) and EH (119 to 162 cm).Among this group of elite accessions from Latin America, the lowest means for PH and EH were observed from crosses with accessions from Guatemala.For root lodging (B+R), the crosses with accessions from Mexico and Peru gave highest means for PH and EH while crosses with Bolivian accessions showed a trend to having better tolerance to broken stalks and root lodging.Prolificacy, measured by ear index, can serve as an indicator of the adaptability level of the elite accessions from other countries of Latin America.The crosses with lower EI values were from Mexico and showed the poorest nicking of female and male flowering while making crosses TABLE 3).Although the data for male and female flowering were taken in two locations, results indicate that the groups of accessions from Guatemala and from Mexico were the earliest and latest, respectively, confirming the level of adaptability.
The means combined over four environments with tester BR 106, the checks and the high-parent heterosis estimates for Y are shown in TABLE 4. The range of variation for yield was from 4.1 t ha -1 to 7.4 t ha -1 while the commercial check BR 201 produced 6.6 t ha -1 .For the other traits the crosses with BR 106 showed similar trends as the crosses with BR 105, but means were lower with tester BR 106 due to having a lower mean values for these traits or due to the elite accessions having a dent endosperm.Mean values obtained in this study were indicative that there are accessions with desirable performance to be introduced into breeding programs.However, in these agronomically unimproved accessions it is impossible to select for all traits simultaneously, because there will be little gain from selection for any single trait.The best alternative, according to Eberhart et al. (1995), would be to select initially for the principal trait (yield) since most of the other traits have high heritabilities.
Because yield is the primary agronomic trait of interest, and since these elite accessions have never been improved in their native countries, heterosis will be   (Gerrish, 1983;Mungoma & Pollak, 1988;Pollak et al. 1991;Vasal et al. 1993).
According to Pandey & Gardner (1992) the populations that were developed from broad genetic resources without regard to heterotic pattern, generally show low heterosis, but in some specific instances high values can be observed.Populations like these generally show poor performance with almost no heterosis, but can perform very well in a specific environment.For this reason, it is important to maintain the elite accession per se for its use in forming two or more breeding populations to which recurrent selection could be applied.Such improved populations can serve as useful resources for development of new varieties and will be used far more often for developing inbreds or hybrids.
In this study, some elite maize accessions were judged to have performed well in crosses with testers BR 105 and BR 106 (TABLES 3 and 4) because they showed higher yields than a commercial check BR 201.From a practical point of view and considering the limited financial resources for maize breeding programs in public institutions, among the elite accessions the following were selected for further work in Brazil: SE 032 and PE 001 with the tester BR 105, and PE 011 and Pasco 014 with tester BR 106.It is important to emphasize that with more financial resources it would be possible to select elite accessions for each specific ecological region, since significative treatments x locations interactions occurred.These accessions, belonging to two different groups, could enhance heterosis and could be used to develop new varieties for small farmers.By initially improving the elite accessions through recurrent selection, allelic frequencies of desirabele genes will be increased; consequently, accessions will have higher yield with better agronomic traits.The probability of identifying superior lines will be increased (Hallauer and Miranda Filho, 1981) and the two divergent populations will show high cross performance with each other.

TABLE 1 -
Mean squares of the combined analysis of variance for trials with tester BR 105 over four environments for yield (Y), plant height (PH), ear height (EH), ear index (EI) and square root of broken stalks + root lodging + 1 (B+R).

TABLE 2 -
Mean squares of the combined analysis of variance for trials with tester BR 106 over four environments for yield (Y), plant height (PH), ear height (EH), ear index (EI) and square root of broken stalks + root lodging+ 1 (B+R).

TABLE 3 -
Mean values over four environments considering maize elite accessions x tester BR 105 and high parent heterosis estimates for yield in (HPH%).

TABLE 4 -
Mean values over four environments considering maize elite accessions x tester BR 106 and high -parent heterosis estimates for yield in (HPH%).