Relationships among oil content, protein content and seed size in soybeans

Relationships Among Oil Content, Protein Content and Seed Size in Soybeans

Mario Marega Filho¹, Deonisio Destro1^*, Lilian Azevedo Miranda², Wilma Aparecida Spinosa³, Mercedes Concórdia Carrão-Panizzi⁴andRicardo Montalván⁵

¹Departamento de Agronomia da Universidade Estadual de Londrina (UEL), Caixa Postal 6001, CEP 86051-990 Londrina-PR, Brazil,

²Instituto Superior de Ciências Agrárias, Biológicas e Geociências da Universidade Paranaense (UNIPAR), Caixa Postal 224, CEP 87502-210, Umuarama-PR, Brazil;

³Fundação Educacional do Município de Assis (FEMA), Av. São Cristóvão, 635, 19800-000, Assis-SP, Brazil;

⁴EMBRAPA/Soja, Caixa Postal 231, CEP 86001-970, Londrina, PR, Brazil;

⁵Departamento de Agronomia da Universidade do Oeste do Paraná (UNIOESTE), Caixa Postal 96, 85960-000, Marechal Cândido Rondom-PR, Brazil

ABSTRACT

During 1995/96 and 1996/97, experiments were carried out at Londrina State University, aiming at quantifying the oil and protein contents in two groups of soybean genotypes; estimating the phenotypic, genotypic and environmental correlations existent among oil, protein content and seed size, and identifying genotypes for direct human consumption with high protein content. The evaluated characters were Weight of a Hundred Seeds (WHS), expressed in grams/100 seeds, Oil Content (OC) and Protein Content (PC), expressed in %. In the experiment carried out in the field, OC ranged from 12 to 20.37 % and PC from 35.66 to 41.75% while in the experiment carried out in the greenhouse OC ranged from 12.26 to 21.79 % and PC from 32.95 to 41.56 % . The correlations between oil and protein were negative and significant. The relationship among WHS with OC and PC was low and higly affected by the time effect. Due to their high protein content and stability to oil and protein contents, there were distinction among the treatments carried out in the field (GA23 and GA20), and those carried out in the greenhouse (PI408251, Waseda, B6F4 (L-3 less), PI423909 and Tambagura).

Key words: Soybean, Glycine max, Seed composition, Seed size, Oil content, Protein content

INTRODUCTION

The importance of soybean [Glycine max (L.) Merrill] worldwide is due mainly to its high oil and protein contents. Soybean breeders have looked for genotypes with high oil and protein contents along with high grain yield. Knowledge of the genetics concerning interesting traits of one specie is very important for breeding improvement. According to Vencovsky & Barriga (1992), the examination of the nature and extent of relationships among traits is important. A breeding program, commonly, does not look for the genetic improvement of isolated traits, but for the genetic improvement of a set of traits, since it is interesting for the breeder to know how the intervention in one trait can cause alteration in others.

Many attempts have been made through breeding to elevate grain yield. However, other traits related to seed composition have been little studied. The negative correlation existent between oil and protein contents as well as the negative correlation between protein and yield is a problem in obtaining cultivars with high yield and high protein contents. In the so called food-type soybean, the big seeds are used in the production of, for example, tofu and miso and the small seeds are used in the production of sprouts and natto. Knowledge of the relationship between seed composition and seed size would be, therefore, useful in developing these type of cultivars.

The objectives of this study were: (1) to quantify the oil and protein content in two groups of soybean genotypes, (2) to estimate the phenotypic, genotypic and environmental correlation among oil, protein content and seed size in these genotypes, and, (3) identify genotypes with high protein content for direct human consumption.

MATERIAL AND METHODS

Two experiments were carried out at Londrina State University (Londrina, Paraná State, Brazil). One in the field, at School Farm, and another in the greenhouse during the agricultural years of 1995/96 and 1996/97. The experiment carried out in the field used 25 lines (F₉ and F₁₀ generations) from topcrosses between food-type soybean and two grain-type soybean testers, Doko and FT-2, plus the FT-Monsanto cultivar, totalling 28 treatments. The Genetic Dept of ESALQ/USP provided the populations that originated these lines (Table 1).,

In the first year, sowing was carried out on Nov. 07, 1995 and in the second year on Nov. 05, 1996. Treatments were arranged in a randomized complete block experimental design, with three replications. Each plot was made up of a 5.0-m single row. Spacing among plots was 1.0-m. One hundred seeds per plot were used. The seedlings were thinned at V_c stage (Vegetative cotiledonar stage of the scale of Fehr & Cavines, 1977), leaving 50 seedlings per plot. The soil type of the experiment area is "Structured Terra Roxa" (HAPLUDULT). Normal spraying with insecticides was used to control caterpillars and stink bugs, as required. Immediately before sowing, seeds were inoculated with Bradyrhizobium japonicum. Approximately 15 days after plants had reached harvesting maturity (R₈ stage of the scale of Fehr & Caviness, 1977) they were manually harvested and mechanically threshed.

The experiment carried out in the greenhouse (Dept of Agronomy, Londrina State University) was composed of 39 genotypes, among lines and cultivars of food-type soybean (Table 2), most of these provided by EMBRAPA/CNPSoja. Sowings took place on Dec. 06, 1995 and Nov. 04, 1996, respectively.

The experimental design was the same as used in the field experiment, but a three-plant vase constituted each plot.

Evaluated characters:

Weight of a hundred seeds (WHS): The seed size was given by weighting a sample of 100 seeds, expressed in grams.

Oil content (OC): The oil content was determinated gravimetrically after extraction using petroleum ether, in a Soxhlet instrument, technique 920.85 (AOAC, 1990), expressed in dry matter %.

Protein content (PC): The protein content was obtained by total nitrogen determination, according to the micro Kjeldahl method, technique 920.87 (AOAC, 1990), using a 6.25 conversion factor and expressed in dry matter % as well.

Variance analysis and its component parts: Statistical analysis were made using the software Genes (Cruz, 1997). The variance analysis was carried out using traditional proceedings. OC and PC data were transformed by Arc Sen which is, according Gomez & Gomez, 1984, the most appropriate transformation for data expressed in %. Individual and combined analysis of variance (ANOVA) were performed according to the system proposed by Cruz (1997) and Cruz & Regazzi (1997), respectively, considering that genotype was a fixed effect and environment (year) and genotype x years effects were random effects. Genetic variances were estimated among treatments (), among years (), among treatments x years () and error ().

Comparison among means: Mean comparisons were carried out using criteria proposed by Scott & Knott (1974).

Correlations estimation: The phenotypic (r_P), genotypic (r_G) and environmental (error) (r_E) correlations were calculated using the following formula:

Where: , and ; correspond to phenotypic, genotypic and environmental covariance among the characters x and y; and , , , , and corresponds to phenotypic, genotypic and environmental variances for characters x and y. The significance of the phenotypic correlation was checked by a t-test, with n-2 degrees of freedom.

Parting of treatments x years interaction: The parting of treatments x years (TxY) interaction was made according to the system proposed by Banzatto & Kronka (1989).

RESULTS AND DISCUSSION

Correlations among characters:Table 3 shows the estimated phenotypic, genotypic and environmental correlation. It is important to emphasize at this point that the estimated genetic correlation from combined analysis offers a better evaluation of the real relationship between characters, since it uses a genetic variance isolated from year variance and from TxY interaction variance. This doesn’t happen in individual analysis where the genetic variance includes these two other variances.

In the field experiment, for the combined analysis, the r_F between WHS and OC, WHS and PC, and OC and PC were 0.69**, -0.42* and -0.59**, respectively. The rG between WHS and OC was 0.82, and between WHS and PC –0.84. Combined analysis showed a high, positive and significant r_P between WHS and OC, a medium-high and highly significant r_P between OC and PC, however negative (-0.59**). The negative r_P between WHS and PC was relatively low, although significant (-0.42*).

Looking at the correlation estimated individually (year by year), it was observed that r_P and r_G between OC and WHS had a significant fall in 1996, compared to 1995. The correlation between PC and OC changed from one year to other as well. Genotypic correlation became a bit more pronounced from 1995 to 1996 (-0.56 to -0.62), and r_P fell (-0.43* to -0.28).

In general based on, the correlation resulting from individual analysis as well as results from combined analysis, it was found that r_G presented higher values with the same orientation as those of phenotypic correlations. These results agreed with results obtained by Johnson et al. (1955), Anand & Torrie (1963), Kwon & Torrie (1964), Smith & Weber (1968), Byth et al. (1969), Freire Filho (1988) and Zimback (1992) and indicated that the phenotypic expression from association between characters was reduced due to environmental effects. Among the r_F values, all of them differed from zero by the t-test, however, the correlation between WHS and PC differed at 5% level, and between WHS and OC and between WHS and PC at 1% level.

Regarding the relationship between WHS and OC, and WHS and PC, various researchers obtained similar results. Tinius et al. (1993) reported phenotypic values for seed size and oil content in three subpopulations as 0.87, 0.94 and 0.97 and for seed size and protein content in two soybean subpopulations as -0.97 and -0.86. Pantalone et al. (1996) evaluating a F₂ generation estimated a genotypic correlation of 0.21±0.16 for seed size and oil content. However, Miranda (1994) obtained negative estimates for this correlation (-0.24 and -0.29) while Mayor & Soto (1985) obtained positive estimates for seed size and protein content, with correlation values at 0.11 (winter) and 0.21 (summer). The literature shows a large variation in correlation magnitudes among these characters due to differences among environments and evaluated populations.

In the greenhouse, combined analysis showed a genetic correlation between WHS and OC, WHS and PC, and OC and PC of 0.24, 0.04 and -0.65, respectively. Phenotypic correlation values were 0.20, 0.02 and -0.45**, respectively. This showed that the correlation between WHS and OC, and WHS and PC tended to disappear, while the correlation between OC and PC reached medium and negative values (r_P = - 0.45**, r_G = -0.65).

The combined analysis presented a low and highly significant phenotypic correlation between OC and PC (r_P = -0.45**), a low and not significant r_P between WHS and OC (r_P = 0.20) and a null r_P (not significant) between WHS and PC (r_P = 0.02). This suggested that although different populations were used under field and greenhouse conditions, the greenhouse conditions (more stable) tended to reduce the correlations among WHS, OC and PC. The genetic correlations was slightly higher than phenotypic correlation between WHS and OC (r_G = 0.24), and WHS and PC (r_G = 0.04). However, the genetic correlation between OC and PC was higher than phenotypic in magnitude (r_G = - 0.65).

By examining the relationships among characters, under greenhouse conditions, during evaluated years, we could verify that the correlations between WHS with OC and PC was, in general, very low, specially that between WHS and OC. The correlations between WHS and PC in spite of remaining practically null on both years, was negative in 1996. The correlation between OC and PC in 1995 had low values and in 1996 presented medium values, indicating that this relationship was highly affected by years. These results were to those obtained by Johnson et al. (1955), Kwon & Torrie (1964), Tinius et al. (1993) and Pantalone et al. (1996).

Variance analysis: Individual analysis showed highly significant differences among treatments on both years for WHS and OC, while PC showed this result only for 1995 (Table 4). The mean square (MS) among treatments were similar for both studied years considering the character WHS (112 e 129) and very different to OC and PC, showing that the variability of WHS was not so affected by years (Table 4). The same thing did not happen to PC, which was expensively reduced in 1996 with no significant differences among treatments. Individual analysis at greenhouse showed highly significant differences among treatments on both studied years for all studied characters (Table 4). The relationship between the biggest and the smallest MS of error from the field individual analysis were 1.00, 3.08 and 1.35 to WHS, OC and PC, respectively, neither exceeding the relation 4:1 established by Box (1954) nor the relation 7:1 suggested by Banzatto & Kronka (1989) for the performance of combined analysis. It meant that the relationship between extreme MS of error indicated the homogeneity of environments, and, therefore indicated that data could be submitted to combined analysis.

The behavior of treatments, years and TxY interaction were significantly different as shown by the combined analysis, except for PC that did not show significant differences among treatments (Table 4). MS magnitudes emphasized that the variability for years was many times superior than that for the effect of interaction, reaching the extreme of being 32 times higher for OC. The high variance for the TxY interaction showed differences among genotypes behavior on both evaluated years to all characters.

For the greenhouse experiment, the relationship among the biggest and the smallest MS of error were 1.94, 1.72 and 1.43, respectively for WHS, OC and PC, indicating that the environments were homogeneous and could be submitted to combined analysis. Combined analysis showed significant differences among treatments and TxY interaction in the three characters. Years showed significant differences for PC, but not for WHS and OC. The significant TxY interaction and the high value of this interaction showed that the genotypes behaved differently on both studied years in all three characters (Table 4).

According to the Scott & Knott (1974) grouping system, the field experiment showed, in the combined analysis, that treatments were more clearly distinguished in the WHS character (Table 5).

The average in the 28 treatments for WHS was 19.14 g. The maximum value was 30.35 g/100 seeds to GA10, which differed significantly from the others according to the Scott & Knott (1974) grouping test at 5% of significance level. The minimum observed value was 8.58 g/100 seeds to GA22, which showed no difference from GA20 and GA23 (Table 5). The combined analysis showed that the average for OC was 16.62%. The maximum obtained value for this character was 20.37% for GA14 which showed no significant difference from the other 17 treatments. The minimum value was 12% obtained by GA22 which did not differ significantly from GA21 and GA24. For PC, the average of genotypes was 38.64% and the minimum value was presented by FT-2 (35.66 %), not differing from the other eight treatments. GA17 showed the biggest PC (41.75 %) not differing significantly from the other 18 treatments (Table 5). Variation coefficients (VC) for combined analysis were 5.56, 5.72, and 3.92% for WHS, OC and PC, respectively (Table 5). Marschalek (1995) obtained VC varying from 2.45 to 3.74% for OC which were 2.49 to 3.39% by Soldini (1993). Marschalek (1995) obtained VC estimates of 6.04% for PC and 9.58% for WHS. Pulcinelli (1992) reported values of 3.8 and 4.4% for VC of PC. Thus, the precision of this experiment, evaluated by the obtained VC values, could be considered satisfactory.

In the greenhouse experiments, average for the 39 treatments for WHS was 25.37 g / 100 seeds. The maximum value was obtained by Tambagura with 58.73 g/100 seeds, differing from other genotypes significantly (at level of 5% of significance). The minimum value for the WHS characteristic was observed in PI408251 with 10.78 g/100 seeds, which did not differ significantly from Natto and PI205085. The mean of treatments for OC was 17.80% and the maximum value of 21.79% was reached by the F83-8207AB treatment, which did not differ significantly from F82-7843, Delsta, Sel. Stwart and F82-5813.

The lower OC was presented by the PI205085 treatment, which differed significantly from the rest. The average PC for the 39 genotypes was 36.75% whereas the maximum observed value was 41.56% for PI408251, which did not differ from the other nine treatments. The lower observed value for PC was 32.95%, obtained by Paranagoiana, which did not differ significantly, by the Scott & Knott, 1974, grouping test from the other 11 treatments (Table 6). The VC values observed in the combined analysis were 7.85, 6.11 and 3.38% for WHS, OC and PC, respectively. These values showed that the experimental accuracy was satisfactory.

Treatments x years interaction: Table 4 showed that TxY interaction was significant to all three characters studied on both field and greenhouse experiments, suggesting that the genotypes behavior was influenced systematically by year.

To know more about this interaction, two partitions of MS were made: one looking for the behavior of treatments within years and other searching for the behavior of years within each treatment. The behavior of years within treatments indicated the stability of characters within each treatment due to environmental differences in the years. Table 7 shows the behavior of years within treatments in the field experiment. The environmental variation of years presented either higher or lower importance to treatment performance depending on the considered character. Thus, from 28 studied genotypes, 22 did not suffer interaction within years for PC, 16 for WHS and 13 for OC. However, the treatments that were affected by this interaction, in most cases, were highly significant. For the genotypes carried out in the greenhouse (Table 8) environmental variation of years influenced only a small part of the treatments. Thus, from the 39 studied genotypes, 25 did not suffer years interaction for PC, 35 for WHS and 32 for OC. This showed that under greenhouse conditions, as expected, a bigger environmental homogeneity could be obtained for years. From the 39 studied genotypes in the greenhouse experiment, 22 were not affected by the years in the three assessed characters. They were the following genotypes: BR-27, F82-5722A, F82-5769, F82-5813, F82-7843, F83-8017, F83-8185, TMV, Ivaí, Delsta, Soja Feira 86-13, Easycook, PI423909, BR 92-15360, BR 92-22106, Late Giant, Waseda, PI408251, B6F4 (L-3 less), Kunitz 1, Kunitz 2 e Natto. All other treatments presented diferences in at least one of the assessed characters, and, among them seven genotypes [Majos, Faz. Progresso, F83-8207AB, F83-7977, Tambagura, B5F5 (L-2 less) and Paranagoiana] did not suffer the time effect as much as the oil and protein contents.

CONCLUSIONS

1) The population carried out in the field presented an average oil content of 16.62 %, ranging from 12 (GA22) to 20.37 % (GA14), and an average protein content of 38.64 %, ranging from 35.66 (FT-2) to 41.75 % (GA17). On the other hand, the experiment carried out in the greenhouse presented an average oil content of 17.80 %, ranging from 12.26 (PI205085) to 21.79 % (F83-8207AB), and an average protein content of 36.75 %, ranging from 32.95 (Paranagoiana) to 41.56 % (PI408251).

2) There was a high and negative correlation between oil and protein contents.

3) At the field experiment, the combined analysis showed a high and highly significant relationship between WHS and OC (r_P = 0.69**), whereas the relationship was low, negative and significant (r_P = -0.42*) between WHS and PC.

4) In the greenhouse experiment, the combined analysis showed a low and not significant relationship between WHS and OC (r_P = 0.20), and, between WHS and PC (r_P = 0.02). The individual analysis showed that the relationships among these characters, although low, were highly affected by the time effect.

5) Due to their high protein contents and stability to oil and protein contents, there were distinction among the treatments carried out in the field (GA23 and GA20) and those carried out in the greenhouse (PI408251, Waseda, B6F4 (L-3 less), PI423909 and Tambagura).

ACKNOWLEDGEMENTS

The authors thank CNPq and CAPES for financial support.

RESUMO

Durante 1995/96 e 1996/97, foram conduzidos experimentos na Universidade Estadual de Londrina, visando: quantificar os teores de óleo e proteína em dois grupos de genótipos de soja tipo alimento; estimar as correlações fenotípicas e genotípicas existentes entre os teores de óleo, proteína e tamanho das sementes; e, identificar genótipos para consumo humano de forma direta, com elevado teor de proteína. Foram avaliados os caracteres Peso de Cem Sementes (PCS, expresso em gramas / 100 sementes), Teor de Óleo (TO), e Teor de Proteína (TP), expressos em %. Na população conduzida a campo, a característica TO variou de 12 a 20,4 %, e TP de 35,7 a 41,8 %. A população conduzida em casa de vegetação apresentou uma variação de 12,3 a 21,8 % para TO, e de 33 a 41,6 % para TP. As correlações entre TO e TP foram negativas e significativas. A relação do PCS com TO e TP mostrou-se baixa e muito afetada pelos efeitos dos anos. Devido aos seus altos teores de proteína e estabilidade para os teores de óleo e proteína, destacaram-se, dentre os tratamentos conduzidos a campo, GA23 e GA20; e, dentre os tratamentos conduzidos em casa de vegetação, PI408251, Waseda, B6F4 (L-3 less), PI423909 e Tambagura.

Received: April 26, 1999;

Revised: August 13, 1999;

Accepted: April 24, 2000.

• Anand, S.C. & Torrie, J.H. (1963), Heritability of yield and other traits and interrelationship among traits in the F₃ an F₄ generations of three soybean crosses. Crop Sci., 3, 508-511.
• AOAC. Association of Official Analytical Chemists (1990), Official Methods of Anallysis, Virginia.
• Banzatto, D.A. & Kronka, S.N. (1989), Experimentaçăo agrícola. FUNEP, Jaboticabal, pp. 213-236.
• Box, G.E.P. (1954), Some theorems on quadratic forms applied in the study of analysis of variance problems, I. Ann. Math. Stat., 25, 290-302.
• Byth, D.E.; Weber, C.R. & Caldwell, B.E. (1969), Correlated truncation selection for yield in soybeans. Crop Sci., 9, 699-702.
• Cruz, C.D. (1997), Programa Genes; aplicativo computacional em genética e estatística. UFV, Viçosa, 442 pp.
• Cruz, C.D. & Regazzi, A.J. (1997), Modelos biométricos aplicados ao melhoramento genético. UFV, Viçosa, 390 pp.
• Fehr, W.R. & Caviness, C.E. (1977), Stages of soybean development. Ames, Yowa. Yowa State University, Cooperative Extension Service. 11 pp (Special Report, 80).
• Freire Filho, F.R. (1988), Análise genética de um dialelo entre genótipos precoces de soja (Glycine max (L.) Merrill.). Doc. Thesis, Escola Superior de Agricultura "Luiz de Queiroz" / USP, Piracicaba, Brazil.
• Gomez, K.A. & Gomez, A.A. (1984), Statistical procedures for agricultural research. John Wiley & Sons, New York, pp. 306.
• Johnson, H.W.; Robinson, H.F. & Comstock, R.E. (1955), Genotipic and phenotipic correlations in soybeans and their implications in selection. Agron. J., 47, 477-483.
• Kwon, S.H. & Torrie, J.H. (1964), Heritability of and interrelationships among traits of two soybean populations. Crop Sci., 4, 196-198.
• Marschalek, R. (1995), Correlaçőes genéticas e fenotípicas entre produçăo de grăos, teor de proteína e teor de óleo em soja, em vários ambientes. M.Sc. Dissertation, Escola Superior de Agricultura "Luiz de Queiroz" / USP, Piracicaba, Brazil.
• Mayor, Z.F. & Soto, J.A. (1985), Influencia de diferentes factores en el contenido de proteínas del grano de soya (Glycine max). Cięncias de La Agricultura, 25, 51-57.
• Miranda, M.A.C. (1994), Seleçăo recorrente divergente para peso de sementes e porcentagem de óleo em soja com uso da machoesterilidade genética.1994. Doc. Thesis, Escola Superior de Agricultura "Luiz de Queiroz" / USP, Piracicaba, Brazil.
• Pantalone, V.R.; Burton, J.W. & Carter-Jr, T.E. (1996), Soybean fibrous root heritability and genotypic correlations with agronomic and seed quality traits. Crop Sci., 36, 1120-1125.
• Pulcinelli, C.E. (1992), Herança do teor de proteína em soja. M.Sc. Dissertation, Escola Superior de Agricultura "Luiz de Queiroz" / USP, Piracicaba, Brazil.
• Scott, A.J. & Knott, M. (1974), A cluster analysis method for grouping means in the analysis of variance. Biometrics, 30, 507-512.
• Smith, R.R. & Weber, C.R. (1968), Mass selection by specific gravity for protein and oil in soybean populations. Crop Sci., 8, 373-377.
• Soldini, D.O. (1993), Interaçăo genótipos x locais e correlaçőes entre caracteres com ęnfase na produtividade de óleo de soja. M.Sc. Dissertation, Escola Superior de Agricultura "Luiz de Queiroz" / USP, Piracicaba, Brazil.
• Tinius, C.N.; Burton, J.W. & Carter-Jr, T.E. (1993), Recurrent selection for seed size in soybean: III. Indirect effects on seed composition. Crop Sci., 33, 959-962.
• Venkovsky, R. & Barriga, P. (1992), Genética Biométrica no Fitomelhoramento. Sociedade Brasileira de Genética, Ribeirăo Preto, pp.335-434.
• Zimback, L. (1992), Cruzamentos em cadeia entre genótipos adaptados e exóticos de soja com ęnfase na produçăo de óleo. Doc. Thesis, Escola Superior de Agricultura "Luiz de Queiroz" / USP, Piracicaba, Brazil.

^*

Author for correspondence

Publication Dates

• Publication in this collection
  25 Oct 2001
• Date of issue
  Mar 2001

History

• Accepted
  24 Apr 2000
• Received
  26 Apr 1999
• Reviewed
  13 Aug 1999