FIELD EVALUATION OF COMMON BEAN FOR REACTION TO WEB BLIGHT AND HIGH TEMPERATURE 1

During the past 30 years, common bean ( Phaseolus vulgaris L.) production in Cen- tral America has expanded into the humid, lowlands (Gourdji Aldana de León, 2010). This allows for production during non-traditional growing seasons, result- ing in a greater and more stable supply of beans in the region. In the humid tropics, seed yield and seed quality of beans are threatened by web blight caused by Rhizoctonia solani Kühn (teleomorph: cucumeris 2008). High temperatures can disrupt reproductive development and reduce seed yield of heat-sensitive bean genotypes (Porch and Jahn, 2001). This study aimed to identify the response of breeding lines to web blight using a spatial model. A ﬁeld trial was conducted at the Isabela Substation of the University of Puerto Rico Agricultural Experiment Station (18.47°, - 67.05°) at 128 m above sea level. The soil at the ﬁeld site is an Oxisol (very-ﬁne, kaolinitic isohyperthermic Typic Hapludox) (Muñoz et al., 2018). The trial was planted on 13 June 2017. An augmented design with two replications was used. The treatments were 479 F 7 common bean-breeding lines from Zamorano University that were selected from the third cycle of recurrent selection for resistance to web blight. The pink bean PR0401-259 and the black bean MHN 322-49 were resistant check lines and the white bean cultivar ‘Morales’ and the small red bean cultivar ‘Amadeus 77’ were susceptible checks in the ﬁeld trial. The four checks were randomized within sub-blocks containing 20 rows. The experimental units were single l-m rows spaced at 0.76 m between rows. Approximately 30 in were inoculated with a mycelial suspension of Rhizoctonia solani Kühn using the methods described by Take- gami et al. (2004). The isolate which was originally collected at the Isabela Substation from bean plants with web blight symptoms and characterized in the subgroup AG-1- 1E in a different study (González et al., 2012). The isolate, stored in sterile sand, was recovered, multiplied and used to inoculate bean plants at 45 days after planting. The disease reactions on the leaves were evaluated at 34 d after inoculation using the CIAT 1 to 9 scale (van Schoonhaven and Pastor-Corrales, 1987). Experimental units (1-m row lengths) were harvested to estimate seed yield. Damaged seed was removed each experimental unit to estimate percentage (%) of damaged seed.

The data was analyzed using a model for an incomplete block design (IBD) with the following response variables: web blight scores at 34 days, percentage of damaged seeds, and seed yield. The model included a fixed line effect and random replication and blocks within repetition effects. Since the blocks consisted of long strips, several models with spatial correlation were also considered to account for a possible spatial correlation between the responses, and, hence, to attain smaller standard errors to compare the means. The best spatial model (exponential correlation) was selected using the Akaike and Bayesian Information Criteria (AIC and BIC) as fitted using the GLIMMIX and Variogram procedures in SAS v. 9.4 software (SAS Institute, Cary, NC).

Web blight scores at 34 days after inoculation
A comparison of the square root of the average of the squared standard errors of the means shows that the spatial correlation model (exponential) produced a value of 0.582, whereas the model that considers the design structure IBD had a value of 0.614. Since the objective was to compare means of bean lines, the spatial model is more appropriate, because it yields means with greater precision.
In the field map the observations of the web blight scores at 34 days vary spatially, which is why the model that considered this spatial correlation fits better than the model that considered blocks and repetitions. The high, medium and low values of the web blight scores at 34 days tended to be grouped in space. It should be noted, however, that each observation comes from a different bean line. The spatial correlation models consider the correlation between residuals, and the map of these residuals also show spatial trends. Therefore, there seems to be a spatial association for the variable score of web blight at 34 days (percentage of severity of the disease). Results from the semivariogram model suggest that experimental units separated by distances < 10 meters are spatially correlated (Figure 1). At distances > 10 meters, the fitted semivariance begins to flatten. Therefore, plots that are spaced > 10 m apart could be considered independent for practical purposes.
The IBD model and the spatial correlation (exponential) model were considered for percentage of disease severity of web blight at 34 days after inoculation. The means based on the corresponding model (LSMEANS) were ranked from lowest to highest web blight score. Figure 2 shows a strong correlation (r=0.95) between rankings of means from the IBD and the exponential correlation models.

Percentage of damaged seeds
For this variable, the assumption of variance homogeneity was not met. Therefore, a logarithmic transformation of percentage of damaged seeds plus one was used. The variogram and the AIC and BIC criteria showed that the IBD model was better than models considering spatial correlation; hence the results for this variable are based on the IBD model. There seems to be no spatial correlation for this variable.
A list of the bean lines (treatments) was created including ranking of the average percentage of damaged seeds and average web blight scores (severity of disease) at 34 days after inoculation that are in the first quartile; that is, the 120 best bean lines with respect to web blight scores were matched with the 120 best bean lines with respect to percentage of damaged seeds and the intersection identified. A plot of the rankings of these two characteristics show bean lines with the best overall performance in the lower right quadrant (Figure 3). Eleven bean lines are among the 25% having the lowest web   blight scores at 34 days after inoculation and among the 25% of the bean lines having the lowest percentage of seed damage (Table 1).

Seed yield
The LSMEANS using the best spatial correlation model (exponential) for seed yield (kg/ha) was ranked from highest to lowest. A list was prepared containing the rankings of average seed yield and ranks of average web blight scores at 34 DAI that were in the first quartile, that is, the 120 best bean lines with respect to web blight scores at 34 DAI and the 120 best bean lines with respect to seed yield (Figure 4). There were 34 bean lines that appear in both lists. Both rankings were plotted with the lower right quadrant showing the 34 best bean lines with respect to both characteristics (Table 2).
Bean lines combining superior seed yield and low levels of leaf and seed damage were identified. Lines 55 (MHC 3-4-31) and 129  were the only lines that were simultaneously in the top quartile for all three traits. Line 398 (MHC 3-32-11) line 399 (MHC 3-32-16) and line 403 (MHC 3-32-24), which are in the desirable quartile for web blight scores and seed yield in Puerto Rico also had had resistant scores (3.0) to web blight in trials conducted at Zamorano University in Honduras (data not shown.).
The pink bean line PR0401-259 was a good check for resistance to web blight and tolerance to heat and 'Morales' was a good susceptible check for both traits. The black bean line MHN 332-49 had low levels of web blight symptoms in the leaves but produced low seed yield due to sensitivity to heat. The small red cultivar 'Amadeus 77' yielded under