Analysis of diallel crosses between six varieties of durum wheat in semi-arid area

The study of morphological genetic determinism characteristics and production of durum wheat (Triticum durum Desf.) through the use of a diallel cross between six varieties, including two Algerian genotypes and four French genotypes was conducted in semi-arid high area of Chellif, Algeria. The analysis of variance for general combining ability (GCA) and specific combining ability (SCA) abilities and reciprocal effects were carried out according to the method of griffing. The results show a preponderance of additive effect for characters: length of the straw, thousand kernel weights and the number of grains per spike. An interesting relation was found for the three characters, between the effect parental heterosis (HI) and parental values (VI). Hybrids Ardente/Nefer, Nefer/Ardente and Ardente/Guem Goum Erkham showed a significant heterosis of 51 to 76% compared to their parents. A negative heterosis of the length of straw was found in crosses: Hedba3/Excalibur: -14%; Guem Goum Erkham/Excalibur: -18%; NE/H3: -19%; Excalibur/Accent: -24%; Hedba3/Guem Goum Erkham -32%. The additivity had a considerable influence on the expression of straw length, number of grains per spike and had also a lesser extent for the thousand grain weight. The preponderance of additive effect in the functioning of the genetic variability of straw length and thousand grain weights was confirmed in F2. Positive relations between parent heterosis, parental values and high ratios of GCA/SCA help in the selection of these crosses.


INTRODUCTION
Algeria continued to import high quantities of wheat in 2011 and this was despite the government's efforts to develop the agricultural sector and especially grain production (National Council for Statistical Information of Algeria).The cultivation of durum wheat in Algeria uses a germoplasm not very productive.The genotypes of high yielding introduced have failed to increase production, which prompted farmers to refuse the introduced varieties (Hazmoune, 2000).Boeuf (1932), Laumont et al. (1961) and Baillaud (1991) indicated the existence of high genetic diversity of Algerian durum wheat.
All efforts of the improvement and introduction of new varieties in Algeria were directed toward the interior zones (Setif Metidja, Chelif) which are more favorable for the cultivation of wheat.It is imperative to focus these efforts towards the arid area which present 40 to 50% the global cultivated area in Algeria (Rachedi, 2003).Increasing yields in unfavorable areas to culture should *Corresponding author.E-mail: boussal_alg@yahoo.fr.Tel: (+213) 559 35 1514.Fax: +213 27 66 4232.
be directed toward the improvement of cultivars that have a specific adaptation to local stresses (Acevedo, 1989).The increase in cereal yields in the world was the result of genetic improvement and better crop management.Brancourt et al. (2003) estimated the gain in yield attributed to genetic improvement between 33 to 63%.The introduced varieties are more productive and less susceptible to disease; they have not been adopted by farmers for their sensitivity to environmental constraints and for their low straw production especially in drought conditions (Rachedi, 2003;Benlaribi et al., 1998).For farmers, the local varieties are still the only way to get a self-sufficiency yield.The selection of wheat varieties in recent decades has been focused to favour the decrease in plant size and the increase of the number of grains per spike and the grain weight per plant (Austin et al., 1980).The development of the wheat yield is determined by the number of grains per spike and the thousand grain weight increased and at the same time, these components results in an effective gain in performance.Grain yield is strongly related to the thousand grain weight (Ward-Law et al., 2000).The height of the straw is a very complex character determined by two sets of factors: some are related to the precocity at heading and the others concern the period of elongation of the internodes (Paquet, 1968).The study of the inheritance of these characters: the length of the straw, thousand kernel weight and the number of grains per spike will help to conduct the selection of new genotypes that may be adapted in the study area (semi arid high Chelif, Algeria).The main objective of the study was to evaluate through the analysis of a diallel cross, the genetic inheritance of the three characters mentioned above.

Plant materials
The tests were based on six genotypes of durum wheat (Triticum durum Desf).These genotypes were: Gum Guem Erkham: local variety cultivated in the region of Tiaret (Algeria), which has a very high straw, a very large grain and a high one thousand grain weights under high water stress (Benlaribi et al., 1998); Hedba03: variety selected at INRA, El-Harrach, Algeria in 1921, cultivated in the Setif region, Algeria.This variety is highly appreciated by farmers and it has a very high straw, a medium grain and a high thousand grain weights under water stress.It is characterized by a long growth cycle (Hemis et al., 2009;Benbelkacem, 1993).
The four others varieties imported from the station of research in Montpellier, France (INSAM) were: Acalou (AC), Ardent (AR), Excalibure (EX) and Nefer (NE), and they are characterized by high efficiency, a very short straw, sensitive to water stress of end cycle.They are precocious and contain the dwarf gene that comes from the Chinese variety Akamoudji (D'amato, 1989).

Field experiment
Tests were carried out in 2005/2006 at the experimental station of Technical Institute of large crops (ITGC) of high Chillef, Algeria.It consists of randomized complete block design with three replications.The surface of the basic plot measured was 3.20 m², Brahim and Mohamed 287 and had six lines spaces of 20 cm between them.The F1 seeds were from the hybridized parents, while the F2 grains were the result of self-fertilization of F1.The analysis of variance for general GCA and SCA combing abilities and reciprocal effects were carried out according to Griffing (1956) method which is a series of reciprocal F1 and their parent's combinations.The F1 seeds from cross diallel was performed manually between six parents, while the grains of the F2 were the result of self-fertilization of the F1.The straw length was measured from the tray tillering to the basis of spike.The thousand grain weight was evaluated by taking just the normal grains of wheat.To estimate the number of grains per spike, we took the average value of the total number of grains of wheat plant by the total number of spikes of the same plant.
The statistical method used in this work is a personal application under Excel (2007) using the Griffing method formulas of Griffing (1956) and Thomas (1987).
The analysis of the results was made by adopting the method 3, model 1 of Griffing (1956): Where, ij is the 1 p (number of varieties); k = 1…b (block number ); l = 1….c,(number of observations per plot micro); u = mean population; g i and g j are the effects of variance (AGC) general combining ability of varieties i and j; s ij is the effect of variance (ASC) specific combining ability as s ij = s ji ; r ij is the genotypic effect reciprocalof r ij = -r ji and e ijk = the effect of the specific observation error.

Analysis of F1
Table 1 shows the mean values of the different characters in parents and their F1 and F2 hybrids obtained from a complete diallel of six varieties.We observed a decrease in average values in F1 hybrids from the average values of the parents in the three characters.The comparison of the average parental thousand grain weight character was 30.96 higher than that of F1 hybrids with 28.17.The same character on the 30 F1 hybrids recorded nine crosses with positive heterosis.If we compare the best parent and F1 hybrids, we see that the following crosses: GE / EX, GE / AC H3/NE, AR / GE and AC / GE gave a positive heterosis between 7 and 26%.The comparison between the mean of parents and F1 hybrids showed a decrease of 11% between the average and the parental hybrids in number of grains per spike whereas the hybrid GE/GE and AC/NE gave a value of number of grains per spike better than the better parent.For the 30 hybrids in Straw length, we found 12 crosses with a positive heterosis which varied between 2 and 17% when they were compared to their average parent.For the same character, the F1 hybrids: GE / EX H3/EX, AC / GE, NE / GE NE/H3, EX / GE and AR/H3 showed negative heterosis ranging from -2 to -15%.The analysis of variance showed a significant effect for the sources of variation in GCA and SCA variances in the length of the straw and the number of grains per spike, while the thousand grain weight revealed no significance (Table 3).The variance ratios GCA / SCA variance were positive: 4.104, 1.313 and 1.118 respectively for thousand grain weight, number of grains per spike and (Table 4).The values of GCA among three characters (Table 5) were higher than the values of SCA.The classification of genotypes based on their GCA is shown in Table 6.The variety (NE) showed better GCA for straw length in F1.The character number of grains per spike showed the lowest GCA for the varieties (GE and EX) in F2.
The ratios of the sum of the squared deviations of GCA on the total sum of squared deviations were 24% for the thousand grain weight, 29% for straw length and 32% number of grains per spike Table 4.

Analysis of F2 hybrids
The study of the average of all varieties by the Newman-Keuls test at 5% shows a great diversity among varieties for the three characters.The comparison between the mean of parents and hybrids showed that F2 hybrids hadan average value high relative to parent average for three characters.The number of grains per spike character gave a high heterosis for the following hybrids: AR / NE = 6%, AR / GE = 51%, NE / AR = 60%.For the straw length character, a negative heterosis was found for the crosses H3/EX: -14%; GE / EX: -18%; NE/H3: -19%; EX / AC: -24% and H3 / GE: -32%.The analysis of variance results of GCA, SCA and of the reciprocal effects, revealed highly significant effects (Table 2).The reciprocal effects were significant for straw length but were not significant for the thousand grain weights and number of grains per spike.The order of the variances of parental GCA was unchanged between the F1 and F2 for the three characters (Table 6).The correlation coefficients between GCA F2 and F1 were 0.563, 0.563 and 0.568 respectively for straw length, thousand grain weight and number of grains per spike.

DISCUSSION
The number of varieties and their choices has not been determined at random, so the conclusions cannot apply to all varieties of the study.The analysis of diallel plane between the six durum wheat varieties led us to the general combining ability of different characters.The decrease in the average value of hybrids compared to parents in the three characters can be explained by the environmental conditions that prevailed (water stress) to the end of the cycle.The obtaining of hybrids with a lower straw length compared to their parents is an interesting phenomenon for possible selection of these genotypes.It is likely that the dwarf gene in imported varieties had emerged, which produced short hybrids; H3/EX crossings, GE / EX NE/H3, EX / AC and H3/GE had a parent abroad.
This shortening of the straw is treated as a negative heterosis character (Paquet, 1968).Hanifi Mekliche et al. (2008) found a negative heterosis ranging between -9.66 and -11.57% for the stem height of durum wheat.The positive relationship was high which was carried out by the GCA variance and a high negative heterosis in the length of the straw of some hybrids may be the consequence of the accumulation of genes or alleles linkage groups not conducive to the expression of this character.The SCA were significant for all the traits with the exception of flag leaf (Houshmand et al., 2009).For Gallais (1974), the best variety is the one that has a good GCA.
The strong AGC was recorded in local varieties at straw length and TKW, and the lowest was found in number of grains per spike in F1 and F2.David et al. (1999) found no significant correlation between the lines and GCA.Ranvir et al. (1982) showed that the variances of GCA were significant in number of grains per spike, TKW and the height of the plant.The SCA has a great influence on number of grains per spike, thousand grain weight and LS characters (Mekliche et al., 2009).Mekliche et al. (1999) obtained a positive heterosis for barley estimated between 20 to 100% depending on the material used and the environment.The crossbreeding achieved in the laboratory between genotypes of different groups of tetraploid wheat suggests an important heterosis value (Battaillo, 2004).The positive relationship between parental heterosis and parental values for three characters indicate the possibility of predicting them to a selection.Significant correlations were observed for all characters except for SCA (Bhullar et al., 1988).Stability and permanence of the additive effect over generations gives it greater importance in selection (Falconet, 1972).According to Gallais (1974) the best variety is the one that has a good GCA.David et al. (1999) found no significant correlation between the values of lines and their GCA.The preponderance of GCA in the functioning of the total genetic variability at straw length and thousand grain weight seems to be total (Table 7).The contribution of the variance of GCA in the expression of the variability of characters remains subject to environmental conditions and could cause changes in the constitution of the genome (Gallais, 1977).The effects of GCA are positively related to the values for the three lines of characters; specific effects are significant and high.A comparative study between hybrid F1, F2 and their parents gave an average correlation coefficient (r ² = 0.30) indicating the additive to directly influence the expression of character for straw length.
The positive correlations can be obtained from the effects of weak interactions between the homologous genes (Demarly, 1977).The preponderance of additive effects in the functioning of the genetic variability straw length and thousand grain weight is confirmed.Correlation coefficients between GCAF1 and parental values are: r ² = 0.67, r ² = 0.24, r ² respectively for straw length and thousand grain weight.

Conclusion
The study of diallel cross of durum wheat showed more favorable elements for the selection of these genotypes under the same conditions.The results show significant variability for the genotypes studied.The presence of a dominant additivity and a negative heterosis of some hybrids for the character straw length is an excellent index for the selection of plants to short straw in F2.The negative heterosis for the character straw length may

Table 1 .
Mean values of parents of F1, F2 hybrids of a diallel cross for TKW, LS and NGE.

Table 2 .
Analysis of variance for F1 and F2 hybrids.

Table 3 .
Estimates of GCA and SCA effects of parents and crosses for different traits in F1 and F2.

Table 5 .
The values of GCA and heterosis in F2 LSD 5%.

Table 6 .
Ranking of genotypes according to their GCA for F1 and F2.