Selection for hypocotyl diameter results in genetic gain in common bean plant architecture

Studies highlight the hypocotyl diameter (HD) as an effective indicator of plant architecture (PA). Here, we estimated the genetic gain based on HD to improve PA. Twenty populations of cycles zero (C0) and one (C1), both in the F4 generation, were evaluated for PA, grain yield (GY) and HD. Plants with thickest HD in C0 were intercrossed in a circulant diallel mating design. In cycle C1, an estimated genetic gain of 4.93% was achieved for PA and 4.95% for HD. The populations with the highest probability of breeding lines with a thicker HD belong to cycle C1, and this selection strategy did not alter the GY of the populations of this cycle. Thus, indirect selection based on HD is promising for breeding for common bean PA by recurrent mass selection.


INTRODUCTION
The cultivation of common bean (Phaseolus vulgaris L.) has aroused the interest of large producers.Currently, aside from increased grain productivity, disease resistance and grain commercial quality, one of the main objectives of bean breeding programs is the development of lines with upright plant architecture (Silva and Wander 2015).By the use of common bean cultivars with more upright plant architecture, the grain loss caused by mechanical harvesting can be largely reduced (Pires et al. 2014), which is the reason why plant architecture was included among the main target traits of common bean breeding.
Common bean plant architecture is generally evaluated on a score scale (Collicchio et al. 1997).The trait is complex and depends on others such as growth habit, number and angle of branches, number and length of internodes, plant height, pod distribution, and hypocotyl diameter (Santos andVencovsky 1986, Teixeira et al. 1999).Therefore, to ensure a precise and accurate evaluation of plant architecture of common bean based on a score scale, experienced raters are required.Moreover, score scales have generally been used in evaluations at the plot level, but restrictions were observed in evaluations at the individual plant level (Silva et al. 2013a, Silva et al. 2013b).Some authors (Acquaah et al. 1991, Moura et al. 2013) emphasized hypocotyl RSR Anjos et al. diameter as an effective indicator of plant architecture, with the possibility of using this trait in indirect selection for more upright-growing common bean plants.Moura et al. (2013) described the causality of the effect of hypocotyl diameter on plant architecture in an evaluation of the cause-effect relation by path analysis.They concluded that one of the main determinants of the plant architecture score was the hypocotyl diameter.In addition, Silva et al. (2013b) reported a predominance of additive gene effects involved in the control of the trait hypocotyl diameter.Silva et al. (2013a) found a higher heritability estimate for hypocotyl diameter than for score of common bean plant architecture.
For the breeding of quantitative traits in common bean, recurrent selection has been the most indicated strategy (Ramalho et al. 2005), based on the directed mating design described by Ramalho et al. (2012).In this design, recombination occurs in steps and commonly, the best families of the populations are used.For high-heritability traits such as hypocotyl diameter, recombination with individual plants can be performed, which is called recurrent mass selection.The advantages of mass selection are a reduction in the time required to complete one cycle of recurrent selection (Ramalho et al. 2012) and a decrease in the number of treatments evaluated.Therefore, recurrent mass selection requires less experimental area and reduces costs.
Thus, the purpose of this study was to estimate the genetic gain for common bean plant architecture in one cycle of recurrent mass selection for hypocotyl diameter.

MATERIAL AND METHODS
From crosses among 14 common bean lines (Table 1), established by Silva et al. (2013b) in a partial diallel mating design (6 x 8), the 20 most promising populations were selected, considering the general and specific combining ability for the traits plant architecture scores, hypocotyl diameter and grain yield.These populations constituted cycle zero (C 0 -base population) of the recurrent selection program (Table 2).In the F 2 generation, seeds of each of the 20 C 0 cycle populations were sown in the field in the dry growing season of 2011, in plots with five 4-m rows.The F 2 plants were harvested at physiological maturity and, by means of a digital caliper, the hypocotyl diameter of approximately 200 plants of each population was measured 1 cm below the cotyledon node (Figure 1A).
The four F 2 plants with largest hypocotyl diameter of each C 0 cycle population were selected for recombination, so that the recombination unit consisted of four F 3 plants.The 20 populations of cycle C 0 were recombined using a circulant diallel mating design, strategy in which each parent (population) participated in two mating (Ramalho et al. 2012), resulting in 20 cycle-1 (C 1 ) populations (Table 2).The 20 cycle-C 0 and 20 cycle-C 1 populations were advanced in bulk to the F 4 generation, when they were evaluated in the same experiment, together with nine controls (BRS Valente, BRS Campeiro, BRSMG Madrepérola, Pérola, BRSMG Talismã, CNFC 9437, A805, A170, and A525) in the dry growing season of 2013.The experiment was carried out in Coimbra (lat 20º 49' S, long 42º 45' W and alt 720 m asl), a county in the state of Minas Gerais, Brazil.A randomized block design was used, with three replications and experimental plots consisting of four 3-m rows, spaced 0.5 m apart, in which 12 seeds m -1 were sown.The populations were evaluated for plant architecture, hypocotyl diameter, grain yield per plant and grain yield per hectare.The plant architecture was evaluated at the plot level on a 1 -5 score scale adapted from Collicchio et al. (1997), where 1 is assigned to completely prostrate plants and 5 to upright plants.The second row of each plot was harvested separately to assess individual plants for hypocotyl diameter (in mm) and grain yield per plant (in g).The hypocotyl diameter was measured 1 cm below the cotyledon node with a digital caliper (Figure 1A).The three remaining rows were harvested to assess grain yield (kg ha -1 ).Data of plant architecture scores, grain yield per hectare and mean hypocotyl diameter were subjected to analysis of variance.

RSR Anjos et al.
For the statistical analyses, software Genes (Cruz 2013) was used.
To quantify the efficiency of recurrent mass selection for hypocotyl diameter, the gains of one selection cycle (C 0 to C 1 ) were estimated, apart from the prediction of the potential of the C 0 and C 1 populations to breed superior lines.To estimate the genetic gain (GG) of the traits plant architecture, hypocotyl diameter and grain yield, the mean population data of cycles C 0 and C 1 were used.The GG was estimated based on the means of the populations of cycle C 1 (μ̂C  The methodology proposed by Jinks and Pooni (1976) was used to predict the potential of each population of cycles C 0 and C 1 to breed superior lines.In this case, the data of individual plants were used.This methodology estimates the probability of breeding lines that are superior to a control line by a certain percentage.This probability is calculated by the standardized variable Z and corresponds to the area on the right of a given value on the abscissa of the standardized normal distribution.The variable Z for each population in generation F 4 was estimated based on the mean of the control line, increased by 20% (L ̅ ), on the mean of population F 4 (F ̅ 4 ), the phenotypic variance of population F 4 (σ̂ 2 F4 ), the environmental variance estimated with the controls, (σ̂ 2 E ) and on the additive genetic variance (σ̂ 2 A ) present in the F 4 generation, calculated by σ̂ 2 A = 1.143σ̂ 2 F4 -0.143σ̂ 2 E .Thus, the expression Z = ( L ̅ -F ̅ 4 ) / 1.143σ̂ 2 F4 -0.143σ̂ 2 E was used to estimate the standardized variable Z of each F 4 population (Cruz et al. 2012).The lines used as control for hypocotyl diameter and grain yield per plant were A525 and cultivar Pérola, respectively.
The populations were classified according to the probability of breeding superior lines for the two traits hypocotyl diameter and grain yield per plant, separately as well as simultaneously.For the classification of the populations with regard to the probability of developing superior lines for these two traits simultaneously, the probabilities were standardized and summed, according to the selection index proposed by Mendes et al. (2009).

RESULTS
The highest coefficient of experimental variation (CVe) was 12.08% (Table 3), indicating high precision in the evaluation of the traits plant architecture, hypocotyl diameter and grain yield.Similar values were reported by Silva et al. (2013a), Silva et al. (2013b) and Oliveira et al. (2015).There was a significant effect (p ≤ 0.01) of treatments and its partitioning (populations, C 0 -cycle populations, C 1 -cycle populations, and controls) on plant architecture, hypocotyl diameter and grain yield, indicating variability among the populations of both cycles (Table 3).The estimates of the genetic correlation coefficients among the evaluated traits were 0.73 (plant architecture and hypocotyl diameter), -0.59 (plant architecture and grain yield) and -0.20 (hypocotyl diameter and grain yield).
The contrasts involving the population means of the cycles C 0 and C 1 (POP C 0 vs. POP C 1 ) were significant for plant architecture and hypocotyl diameter and non-significant for grain yield (Table 3).Thus, in relation to cycle C 0 , the population means of cycle C 1 of plant architecture scores and hypocotyl diameter were higher.These results indicate that, in the mean, the populations of cycle C 1 had a more upright plant architecture (Figure 1B), larger hypocotyl diameter and their yields were the same as those of the cycle C 0 populations.
The genetic gains obtained for plant architecture scores and hypocotyl diameter were, respectively, 4.93% and 4.95% (Table 3).These results indicate the efficacy of hypocotyl diameter in indirect selection to improve common bean plant architecture, since to obtain the cycle C 1 -populations, mass selection based exclusively on hypocotyl diameter was applied.
For grain yield, there was a reduction of 2.06% from cycle C 0 to C 1 (Table 3).However, the contrast between the mean values of C 0 and C 1 populations (POP C 0 vs. POP C 1 ) showed no significant effect for this trait (Table 3).These results indicate that indirect selection for plant architecture based on hypocotyl diameter did not affect the grain yield means.
The probability values of developing superior lines from cycle C 0 and C 1 populations, based on the methodology of Jinks and Pooni (1976) are shown in Table 4.For the trait hypocotyl diameter, of the 10 populations (25%) with highest probabilities of developing a superior line (PSL), using line A525 as control, eight populations were of cycle C 1 and only two of cycle C 0 .It is worth emphasizing that the two populations of cycle C 0 ranked ninth and tenth.Of the 10 populations with lowest potential for the development of superior lines, i.e., with lowest PSL, only one population was of cycle C 1 and nine were of cycle C 0 (Table 4).These results, associated to the genetic gains obtained for plant architecture scores (4.93%) and hypocotyl diameter (4.95%) (Table 3), confirmed that indirect selection based on hypocotyl diameter in the recurrent mass selection mating design effectively improved the architecture of common bean plants.
For grain yield per plant, six of the ten populations with highest probability of developing superior lines were of cycle C 1 and four of cycle C 0 (Table 4).This result, associated to the non-significance of the contrast between the populations C 0 and C 1 for grain yield (Table 3), confirmed that indirect selection based on the hypocotyl diameter did not alter the potential of cycle C 1 populations for the development of lines with high yield.
Considering the traits hypocotyl diameter and grain yield simultaneously by the selection index (Table 4), it was observed that of the ten most promising populations, eight belonged to cycle C 1 .In this way, indirect selection for hypocotyl diameter, aside from allowing an improvement of the populations with regard to the potential of developing lines with a more upright architecture, had no influence on the potential of the populations for the development of lines with higher yield.

Indirect selection for hypocotyl diameter is promising for breeding of common bean plant architecture
In this study, the indirect gain obtained for plant architecture in one cycle of recurrent mass selection for hypocotyl diameter was 4.93% (Table 3).Pires et al. (2014) reported a mean gain of 1.62% per cycle of recurrent mass selection (gain of 4.87% in three cycles) for plant architecture, where the most upright plants for recombination were selected visually and progenies of the first and last selection cycle considered (C 5 and C 8 ) were used to estimate the selection gain.According to Silva et al. (2013a), the visually evaluated scores of common bean plant architecture had a lower heritability estimate (0.60) than hypocotyl diameter (0.81).Thus, the selection of upright plants based on hypocotyl diameter is promising in breeding for common bean plant architecture.
The gain obtained by indirect mass selection (Table 3) shows the causality of the effect of hypocotyl diameter on plant architecture, with a genetic correlation coefficient among these characters of 0.73.This causality was also described by Moura et al. (2013) in an evaluation of the cause-effect relation by path analysis of 22 morphological and agronomic traits in relation to scores of common bean plant architecture.These authors concluded that the main determinants of 2.17  Success with recurrent phenotypic selection in common bean was reported in some studies, for example, Amaro et al. (2007) estimated a genetic progress of 6.4% for resistance to angular leaf spot (Pseudocercospora griseola) by recombining the most resistant common bean plants.In another study, Silva et al. (2007) crossed the plants on which floral buds grew first, and achieved gains of 2.2% per year in reducing the number of days to flowering.In these studies, the gain with recurrent phenotypic selection was estimated from the means of progenies derived from each selection cycle.
According to Silva et al. (2009), there is a negative and low correlation between the traits plant architecture and grain yield in common bean.However, in our study, although the populations of cycle C 1 had plants with a more upright architecture than cycle C 0 , the populations of the two cycles did not differ in mean grain yield (Table 3).It should be mentioned that to obtain the cycle-C 1 populations, selection was based exclusively on hypocotyl diameter, i.e, this selection strategy did not affect grain yield.
Aside from estimating the genetic gain based on the means of the F 4 populations, the methodology of Jinks and Pooni (1976) was used to determine the potential of these populations for the development of superior lines.This methodology considers both the mean and the variance to quantify the potential of the populations.For hypocotyl diameter, considering the 10 populations with the highest and lowest potential for the development of superior lines, respectively, the results based on the methodology of Jinks and Pooni (1976) indicated that the C 1 populations were superior to C 0 (Table 4).These results, associated with the gains obtained for plant architecture scores (4.93%) and hypocotyl diameter (4.95%) (Table 3), confirmed that indirect selection based on hypocotyl diameter in the recurrent mass selection system effectively improved plant architecture.With respect to grain yield, the results indicated that indirect selection based on hypocotyl diameter did not alter the potential of the cycle-C 1 populations in relation to cycle C 0 , for the development of superior lines.
According to Ramalho et al. (2012), the populations in recurrent selection breeding programs of autogamous plants are not in Hardy-Weinberg equilibrium, since the crosses in each recombination cycle are directed and, in addition, the genotypic frequencies vary with the increasing inbreeding level of the generations.In this sense, the authors recommended that genetic progress should be estimated based on the performance of the lines obtained in each recombination cycle, since the means of traits with genetic control affected by dominance deviation would be altered by these variations in genotypic frequencies.However, the predominance of additive effects involved in the genetic control of hypocotyl diameter and scores of common bean plant architecture (Silva et al. 2013b, Oliveira et al. 2015) justify that the gain estimates for these traits were based on the mean of segregating populations, using the F 4 generation in this case.
The strategy of gain estimation based on the evaluation of segregating populations also allows the use of the methodology of Jinks and Pooni (1976) to quantify the potential of these populations for the development of superior which is based on the mean and variance within the populations.Thus, for hypocotyl diameter and plant architecture score, any inbreeding generation could be used, due to the predominance of additive effects in their genetic control.However, for grain yield, where the genetic control is predominated by dominance effects (Silva et al. 2013b, Vale et al. 2015), the F 4 generation would be more adequate for the methodology of Jinks and Pooni (1976).The reason is that this generation allows a considerable reduction in dominance effects in the control of the target trait for the prediction of the potential of segregating populations.
In breeding, selection will only be effective if genetic variability is available in the target population (Alliprandini and Vello 2004).Thus, the gains estimated based on the evaluation of segregating populations and prediction of the potential of these populations by the methodology of Jinks and Pooni (1976) are particularly interesting in breeding programs, especially in those using recurrent selection.The reason is that the breeder can decide about continuing the intercrossing of the initial populations of the program or to include parents to increase the variability for one or more traits of interest represented with low variability, which would otherwise result in irrelevant gains.

Simultaneous breeding strategy for plant architecture and grain yield in common bean
The use of bean cultivars with a more upright plant architecture does not only reduce losses by mechanical harvesting, but also decreases crop damages caused by cultural practices, reduces disease incidence, e.g., of white mold, and allows the production of grain with optimized quality (Ramalho et al. 1998, Teixeira et al. 1999, Pires et al. 2014).Thus, common bean breeding programs seek to develop cultivars that associate high grain yields with a more upright plant

Figure 1 .
Figure 1. A. Illustration of hypocotyl diameter evaluation with a digital caliper.B. Common bean plants in cycle one (C 1 ) of the recurrent mass selection program for hypocotyl diameter. 1 14

Table 2 .
Genealogy of the 20 populations of cycle zero (C 0 ) and the 20 of cycle one (C 1 ) of the recurrent mass selection program by hypocotyl diameter of common bean 1 Lines description see Table1.

Table 3 .
Summary of analysis of variance of the populations (POP) of the cycles zero and one (POP C 0 and POP C 1 ) evaluated for plant architecture score (PA), hypocotyl diameter (HD) and grain yield (GY). Means of PA, HD and GY of POP C 0 POP C 1 , and the respective genetic gain (GG) **, * Significant at 1 and 5% probability, respectively, by the F test; 1 Hypocotyl diameter in mm; 2 Grain yield in kg ha -1 .