Evaluating the genetic potential of population CNA-11

Low temperatures are a serious problem for irrigated rice in southern Brazil, mainly in the State of Rio Grande do Sul. Current cultivars are sensitive to low temperatures in their vegetative and reproductive phases. Such sensitivity, together with climatic conditions during early planting and the reproductive phase, causes losses in yield and increases costs.

Table 8. Realized genetic gain (GR %) and means of yield (g per plot) of the five and ten best families selected in the S_0:2 generation. Their responses to selection in the S_0:3 generation are shown according to the different numbers of recombination cycles carried out with the CNA-5 population of irrigated rice at Goianira, Brazil, 1998/99 and 1999/00 cropping seasons.

a. Where a is the intercept of the linear regression equation.
b. Where b is the linear regression coefficient.
c. Level of significance, according to the t test.

Table 9. Realized genetic gain (GR %) and means of yield (g per plot) of the five and ten best families selected in the S_0:2 generation. Their responses to selection in the S_0:3 generation are shown according to the different numbers of recombination cycles carried out with the CNA-5 population of irrigated rice. The means are averaged for two sites (Lambari and Goianira), Brazil, in the 1998/99 and 1999/00 cropping seasons.

a. Where a is the intercept of the linear regression equation.
b. Where b is the linear regression coefficient.
c. Level of significance, according to the t test.

In terms of genetic improvement, two strategies can be adapted to reduce the effects of low temperatures. The first would be to obtain short-cycle cultivars that would escape cold temperatures during their most critical period, which is the reproductive phase. The second strategy would be to obtain cultivars with genetic tolerance of cold temperatures.

This strategy may involve thinking about composing a population that possesses, within its constitution, sources for increased yield and cold tolerance, and working on it through the recurrent selection methodology. This would permit generating and managing variability with the objective of extracting recombinant lines that combined the two desired traits: increased yield and cold tolerance.

The objective of the study described below is to evaluate, by estimating various genetic parameters, the genetic potential of population CNA-11 in terms of yield, cold tolerance and other phenotypic traits of interest (Lopes, 2002).

The irrigated rice population CNA-11, synthesized by Embrapa Arroz e Feijão, is composed of various genotypes that were used as sources of genes for yield; resistance to the root weevil, blast, Helminthosporium spp., cold tolerance and iron toxicity; early maturity; and grain quality (Rangel et al., 2000b). As sources of genetic male sterility, plants from population CNA-1 were used that contained alleles for yield, early maturity, grain quality, resistance to blast and tolerance of iron toxicity (Rangel et al., 2000a). The source of male sterility used to synthesize population CNA-1 was ‘IR36’ (Singh and Ikehashi, 1981).

The original population CNA-11/0/2 segregated at 50% of fertile plants (Msms) and 50% of male-sterile plants (msms). To select individual fertile plants, the population was planted at Pelotas and Santa Vitória do Palmar. Both these cities are located in the southern region of the State of Rio Grande do Sul. The harvested seeds formed S_0:1 families. This phase of the project was under the responsibility of Embrapa Clima Temperado (Pelotas) and of the Instituto Rio Grandense do Arroz (IRGA; Santa Vitória do Palmar). The seeds of the families were planted in seedling boxes and transplanted to the final site 25 days afterwards, at a density of 0.30 × 0.20 m. In the field, the plants were managed according to the general recommendations for irrigated rice farming (EPAGRI, 1997). At maturity, 258 fertile plants were selected from the two sites according to the following criteria: spikelet sterility, grain type, plant height and days to 80% flowering. The S_0:1 seeds were advanced to S_0:2 in Formoso do Araguaia, State of Tocantins (central Brazil), to increase the quantity of seeds for trials to evaluate phenotypic traits. The seeds from plants of each family were harvested in bulk to constitute the S_0:2 families.

Of the 258 families, only 140 were included in the trials, with half of these being from the selection in Pelotas and the other half from the selection in Santa Vitória do Palmar. The 140 S_0:2 families, together with four checks (BR-IRGA 410, IRGA 418, IRGA 420 and INIA Tacuari), were evaluated field trials at Cachoeirinha and Santa Vitória do Palmar, both of Rio Grande do Sul. The experimental design was triple lattice (12 × 12), and experimental plots were six 3 m long rows, spaced at 0.2 m. The useful area consisted of the four central rows, 2.5 m long, and totalling 2 m². Planting was deliberately done outside the recommended period with the idea of placing the plants under low-temperature stress during the reproductive phase.

Ten phenotypic traits were evaluated in the field and laboratory: grain yield (GY), plant height (PH), days to 80% flowering (FLR80), number of grains per panicle (G/PAN), spikelet sterility (SSTE), grain discoloration (GD), 1000-grain weight (GW1000), whole-grain yield (WGY), grain length (GL) and white belly index for grains (WBI).

Individual and cluster analyses of variance were carried out at both sites, considering the effects of families random and the effects of sites as fixed.

For the cluster analysis, the homogeneity of variances of experimental errors was tested by Hartley’s F-max test [1950, cited by Cruz and Regazzi (1997)].

The data for the traits GD and WBI were transformed by the square root, and for the trait GL by the neperian logarithm, following the interpretation of the analyses of residues and the regression of the original data. We used the Statistical Analysis System package (SAS Institute, 1985).

Based on the cluster analysis of variance at the two sites, the following genetic parameters were determined: variance of the effective error (), mean phenotypic variance among the families (), genotypic variance among the families (), coefficient of genetic variation (CV_g), index of variation (b), heritability (h²), and the phenotypic (r_f), environmental (r_e) and genotypic (r_g) correlations between GY and other traits.

The genetic gain expected by direct selection was also calculated, using the following formula:

where:

Gsd_X is the genetic gain expected by direct selection in the principal trait X

ds_X is the selection differential for the principal trait X,

where ds_X is X_FAM(S) - X_FAM(O) [the mean of selected families (S) minus the mean of the families evaluated or observed (O)]

is the heritability of the principal trait X

The genetic gain expected by indirect selection was estimated by the formula:

where:

Gsi_X(Y) is the genetic gain expected in the principal trait X when selection is practised for the secondary trait Y

ds_X(Y) is the indirect selection differential for the principal trait X, which is obtained in relation to the mean of the trait in those individuals whose superiority was shown based on the other trait Y on which direct selection was practised

is the heritability of the principal trait X

For more information on selection indices, see Geraldi (Chapter 3, this volume).

Genetic gain by selection can also be expressed as a percentage of the original mean of the families:

or even per year, considering that, in this case, the recurrent selection cycle is completed in 2 years:

The effects of families and of the interaction families × sites were significant for all variables, indicating the broad phenotypic variability of the genotypes being studied and also the differing performance from site to site. If, however, the effect of families indicates the possibility of obtaining genetic gains by selection for all traits, the significance of the interaction families × sites may hinder selection, because the best families in one site are not necessarily the best ones in the other site.

For the trait GY, the mean of the families (614 g per plot) corresponded to only 58% of the mean of the checks (1054 g per plot) (Table 10). Variation ranged from 244 to 1289 g per plot. These data show that the families evaluated presented relatively low means of GY, but broad genetic variability, considering that the five best families presented GY as equal to or higher than the mean of the checks.

The traits SSTE and G/PAN are correlated and should be analysed together. The S_0:2 families presented a higher mean for SSTE (38.8%) and a lower mean for G/PAN (54 grains per panicle) than did the checks, which presented 26.1% and 72 grains per panicle, respectively (Table 10). The means of the families for the trait SSTE ranged from 21.2% to 73.0%, with the best-performing check at 17.5% (cv. INIA Tacuari). For the trait G/PAN, the best families also performed less than did the best check. These two components could be regarded as the primary reasons explaining the families’ low GY mean.

The families showed a higher mean for GW1000 (25.6 g), compared with the mean of the checks (24.1 g). However, although the families showed a lower WBI, averaging 2.2, the mean score was higher than that of the checks, which presented a mean of 1.2 (Table 10).

Table 10. Estimates of the genetic parameters of 140 S_0:2 families derived from the irrigated rice population CNA-11 and four checks for the traits grain yield (GY), plant height (PH), days to 80% flowering (FLR80), spikelet sterility (SSTE), number of grains per panicle (G/PAN), grain discoloration (GD), 1000-grain weight (GW1000), whole-grain yield (WGY), grain length (GL) and white belly index (WBI). The estimates were obtained from experiments conducted in Cachoeirinha (FAM1) and Santa Vitória do Palmar (FAM2), Brazil, 1999/00 cropping season.

a. Analyses of variance carried out with data transformed by the square root.
b. Analyses of variance carried out with data transformed by the neperian logarithm.
c. Means adjusted by the minimum squares method, using the LSMEANS command of the SAS package (SAS Institute, 1985)

No significant differences were observed among the means of the S_0:2 families and checks for traits PH, FLR80, GD, WGY and GL. This shows that the evaluated families presented appropriate plant height and days to 80% flowering, and that grain size and yield met the requirements of the Brazilian rice market. The behaviour of these variables could have been the result of efficient selections carried out during the synthesis and evaluation of the original populations in Pelotas and Santa Vitória do Palmar.

The interactions families × sites and checks × sites presented different results for five of the ten traits evaluated. While the first was significant for all traits, the second was not for FLR80, SSTE, G/PAN, WGY and GL. This means that, for these variables, the four checks had the same performance at the two sites.

It should be noted that the number of families is much higher than the number of checks and that the families are formed by a set of genotypes with genetic divergence at many loci and with greater heterozygosity than the cultivars, which could contribute to the significance of the interaction with the environment. In any case, given the importance of the interaction, plant breeders have the responsibility of

(1) evaluating its magnitude and significance
(2) quantifying its effects on improvement techniques and strategies for disseminating technology, and
(3) providing aids that enable the adoption of procedures for their minimization (Cruz and Regazzi, 1997).

The estimates of the genotypic variances and of the heritabilities (h²) were high, confirming that most of the phenotypic variability is due to variation in the genotype (Table 10). The values of heritability fluctuated between 64.9% (G/PAN) and 95.4% (FLR80). This also confers good selection precision by phenotype, as the square root of heritability indicates precision of selection (Falconer, 1987).

For GY, h² was 94.2%, which agrees with values cited in similar works. For example, Cordeiro (2001) found heritability values that varied from 85.6% to 97.2% when evaluating families from population CNA-5 over two generations (S_0:2 and S_0:3) and with different numbers of crosses. Rangel et al. (1998) found lower values, the range varying from 49.9% to 74.0%. Morais (1992) found the values of h² between families ranged from 27.8% to 68.7%.

The coefficient of heritability is a property of the population under study and reflects the degree of genetic variability among the evaluation units. The principal function of heritability is to predict. It expresses the reliability of the phenotypic value as an estimator of the genotypic value (Falconer, 1987). In this case, and like the first cycle of selection in population CNA-11, broad genetic variability among the families was seen. When selection and recombination of the superior families occur, genotypic variance declines (Geraldi and Souza, 2000) and, consequently, heritability is reduced.

The coefficient of genetic variation (CV_g) and the index of variation b also indicate the possibility of obtaining genetic gains by selection. The highest values of CVg were for the traits GY (29.8%), SSTE (16.6%), G/PAN (13.7%), GD (16.1%) and WBI (15.0%) (Table 10). Rodriguez et al. (1998) found similar values for CV_g for the traits GY (20.5%), leaf blast (11.7%), plant height (13.2%), number of spikelets per panicle (11.1%) and percentage of full grains (9.3%). The high values for CV_g indicate that the population being evaluated presents high genetic variability and that a possibility exists of obtaining significant gains for the traits cited by selecting the best S_0:1 plants, according to the evaluation of the S_0:2 families.

The index of variation b quantifies the proportion of the genetic variability with regard to the environment (Vencovsky, 1987). Values equal to or higher than the unit indicate a favourable situation for selection. In this study, values for b were higher than the unit for GY (1.7), PH (1.3) and FLR80 (1.8). Combining the information on the CV_g and b values, we can infer that all the traits evaluated, except GW1000 and WGY, present possibilities of expressing genetic progress through selection. However, considering the values of h₂, which were also high for GW1000 and WGY, we can conclude that all the traits can be improved in population CNA-11.

Grain yield is a significant trait in any plant improvement programme. As a result, the study of correlations between GY and other traits can be useful for selection (Table 11).

The variables GD, GW1000, GL and WBI did not present significant correlations with GY. Hence, they are not included in the selection methods discussed later in this chapter. Variables PH, FLR80, SSTE, G/PAN and WGY showed significant phenotypic and genotypic correlations with GY, the first three being negative and the last two positive (Table 11). This means that the families with higher GY have plants that are shorter, with fewer days to 80% flowering, fewer sterile spikelets, a larger number of grains per panicle and a higher whole-grain yield. These results agree with those of Rangel and Zimmerman (1998) and Rodriguez et al. (1998). In the three studies mentioned, the genotypic correlations were always negative between GY and the traits PH and FLR80.

Based on the results shown in Table 11 and with the possibility of carrying out indirect selection for GY, the best alternatives would be to select through PH, SSTE and G/PAN, which present the highest genotypic correlations. The relationship between SSTE and G/PAN still needs to be taken into account be cause of their high negative correlation (r_g = - 754**, data not presented). That is, within a given panicle, where the values for one trait increase, those of the other decrease. Hence, plant breeders can select between the two traits (SSTE and G/PAN), but use only one as a selection criterion, for both practical and statistical reasons.

Table 11. Phenotypic, environmental and genotypic correlations between the trait grain yield and other evaluated traits of the irrigated rice population CNA-11.

a. t_(0.01;139) = 0.217; t_(0.05;139) = 0.165, according to the t test.

To estimate gains by selection, only the variables that presented significant genotypic correlations with GY were considered. These were PH, FLR80, SSTE, G/PAN and WGY.

For GY, gain through direct selection from the 10 best families for this trait was estimated as 404 g per plot (65.7% of the original mean) (Table 12). In this case, the mean of the 10 selected families, which was 1043 g per plot, was similar to that of the checks (1054 g per plot). This shows that, even in the first selection cycle, families were obtained with a productive potential similar to or better than that of the check cultivars; the best family produced 1289 g per plot. Indirectly, gains were obtained for other traits, accumulating a total gain of 106.6%. The variables with reduced means were PH (-8.2 cm), SSTE (-5.3 percentage points) and FLR80 (-2.5 days), whereas G/PAN (7.2 grains per panicle) and WGY (0.7 percentage points) had increased means (Table 12).

The same performance was observed when the 50 best families were selected, but with the consequent reduction in total gain to 54.5%. The directions of changes in the means for traits as a result of selection perfectly accord with that desired by rice breeders, who seek plants with higher whole-grain yield, and shorter plants, fewer days to 80% flowering and reduced sterility of spikelets.

Table 12. Estimates of genetic gains expected by direct selection for the traits grain yield and spikelet sterility, and by indirect selection for other traits. Values are expressed in absolute (Gs) and in percentage of the original mean (Gs%). Traits are grain yield (GY; g/plot), plant height (PH; cm), days to 80% flowering (FLR80; days), spikelet sterility (SSTE; %), number of grains per panicle (G/PAN) and whole-grain yield (WGY; %). Two intensities of selection were used: i = 7.14%, and i = 35.7%.

a. Efficiency of selection for the trait GY, compared with the Gs% obtained for direct selection for GY.
b. Efficiency of selection for total gain, compared with the total Gs% obtained for direct selection for GY.

One selection criterion for cold-tolerant plants was reduced sterility of spikelets, considering that this trait is easily visualized in the field and correlates with this abiotic stress (Yoshida, 1981). In the recurrent selection programme for cold tolerance at Embrapa Arroz e Feijão (Rangel et al., 2000b), fertile plants were selected from a base population cultivated in cold sites, using the criterion of spikelet sterility. Because selection is done with individual plants, directly measuring GY is not yet possible. This trait is quantified in the evaluation of the S_0:2 families. Hence, SSTE is expected to be a good selection criterion - as confirmed by the high genotypic correlation with GY (r_g = -0.693**) (Table 11).

Taking into account these aspects, this study simulated direct selection for SSTE at two intensities of selection (7.15% and 35.7%). The indirect effects for the other five traits under study were also estimated. It should be noted that, in this case, there is no interest in replacing direct selection for GY with indirect selection through SSTE. The aim is to confirm the consequences for GY in terms of using SSTE as a selection criterion and, thus, infer the precision or efficiency of selecting from individual plants.

On selecting the 10 best families, using as criterion reduced SSTE, family means dropped to -9.0 versus -5.3 for the original mean (Table 12). Indirectly, genetic gains were obtained for GY, PH and G/PAN, but to a lesser degree than with direct selection for GY. For example, the GY variable itself showed a gain of 212 g per plot (Table 12), which represents only 52.7% of that obtained in the former situation. Changes in the means for FLR80 and WGY were insignificant (0.14 and -0.02, respectively).

Reducing the intensity of selection for SSTE to 50 families caused a notable improvement in the performance means of the six traits. In contrast, direct selection for GY corresponded to 58.8% of the GY obtained through direct selection for this trait. Total gain changed from 70.4% for the 10 families to 86.6% in this case (Table 12).

In this study, the mean for SSTE of S_0:2 families was 38.8% (Table 10), ranging from 21.2% and 73.0%. Three factors, one environmental and two genetic, contributed to the broad variability of this trait. From the genetic viewpoint, the presence of the male-sterility gene should be taken into account, as it is still segregating in S_0:2 families at the expected frequency of 0.17, considering that Msms plants were selected in the S_0:0 generation. These underwent one self-pollination phase and msms plants were eliminated from the S_0:1 generation. The S_0:2 generation had the following proportions of zygote types: 0.50 MsMs, 0.33 Msms, and 0.17 msms.

The second aspect is the genetic incompatibility involved in the cross between the indica and japonica genotypes that had comprised the parental materials for populations CNA-1 and CNA-11 (Rangel et al., 2000b).

Considering the aspects already discussed and the data obtained with direct and indirect selection for six traits through SSTE - which, at the intensity of selecting 50 families, showed an 86.6% efficiency (Table 12) - the viability of using this criterion is confirmed for selecting for cold tolerance in irrigated rice. The decisive issue for success in the improvement programme for cold tolerance is the regularity and uniformity at which low temperatures occur, so that the selected plants have superior genetic attributes for the trait and are not merely escaping cold stress through climatic variations.

Despite the genetic gain for cold tolerance in the population on being recombined, the data are insufficient to permit a definitive conclusion on the subject. Yet information had been obtained from the best families evaluated in Cachoeirinha and Santa Vitória do Palmar in the 1999/00 cropping season. The families themselves had derived from plants evaluated under cold stress in Pelotas and Santa Vitória do Palmar in the 1998/99 cropping season. To measure genetic gain for cold tolerance, an experiment should be carried out to compare the original population with the improved, using an efficient methodology to evaluate reaction to cold.

With the results so far obtained, we can conclude that population CNA-11 presents high potential for improvement, with broad genetic variability in all ten traits evaluated.