GENETIC DIVERGENCE AMONG COTTON GENOTYPES GROWN IN THE MAIN SEASON AND OFF SEASON

The objective of this work was to evaluate the agronomic performance and estimate the genetic divergence of 18 cotton genotypes grown in the main season (sowed in December, 2012) and off season (sowed in January, 2013), considering their agronomic characteristics and resistance to Ramularia leaf spot. A randomized block experimental design was used, with five replications. The characteristics evaluated were plant height, first branch height, position of first fruiting branch, height of first fruiting branch, length between nodes, number of nodes, average number of bolls per plant, average boll weight, area under the disease progress curve (AUDPC) related to the Ramularia leaf spot severity, weight of 100-seed from the plant middle third, fiber percentage, average production per plant, yield and cotton fiber quality. The results were subjected to individual and joint analysis of variance and the genetic divergence was estimated according to multivariate procedures (Mahalanobis' generalized distance and Tocher's optimization method). The dissimilarity matrices were summed to estimate the genetic divergence, considering both growing periods. Genetic variability was found among the genotypes evaluated, in both the main season and off season. The characteristic that most contributed to the genetic divergence in the main season was the production per plant and, in the off season, was the fiber percentage. According to the results of the present work, the crosses between the genotypes BRS-335 and FMT-707; FM-910 and FMT-707; and IMA-08-12427 and FMT-707 are recommended.


INTRODUCTION
Cotton belongs to the Malvaceae family, whose genus Gossypium has more than 50 native species in arid and semi-arid regions of the Americas, Asia, Africa and Australia, of which 45 are diploid (2n=26) and 5 allotetraploids (2n=52) (FANG et al., 2013;TYAGI et al., 2014).
However, only four cotton species are commercially cultivated, of which two are diploid (G. arboreum L. and G. herbaceum L.) and two allotetraploids (G. hirsutum L. and G. barbadense L.). The species G. hirsutum has shown higher levels of genetic variability than the other three species (WENDEL et al., 1992;ABDURAKHMONOV et al., 2008). However, the genetic basis of G. hirsutum, from which derives the existing cultivars, is considered restricted and that variability is not present in these cotton cultivars (IQBAL et al., 2001;MOIANA et al., 2015).
Brazil is the fifth largest producer of cotton, which had a production of 1.275 million Mg of raw cotton in 2012 (FAO, 2012). However, some diseases affect the economic growth of this crop. Ramularia leaf spot (Ramularia areola G. F. Atk.) is considered the cotton main disease, which may compromise more than 36% of the yield and decrease the cotton fiber resistance (GILIO, 2014).
In addition to the resistance to diseases, several economic characteristics must be considered by the plant breeder to obtain a commercial cultivar, such as yield, boll and seed weight, fiber percentage, fiber quality, resistance to pest and plant architecture (CIA; SALGADO, 2005;FARIAS et al., 2008).
The variability between the parent plants has great importance to achieve a constant improvement of these characteristics in a population. The absence of genetic variability of the parent materials used for breeding limits the obtaining of promising lines (VILELA-MORALES;VALOIS, 2000).
The plant variability can be assessed by the quantification of genetic divergence, through agronomic and morphological characters and molecular markers. Regarding the quantitative agronomic characters, the variability can be assessed through dissimilarity measures, such as the Euclidean distance and Mahalanobis' generalized distance (CRUZ;CARNEIRO, 2003;REGAZZI;CARNEIRO, 2012).
The preference for the Mahalanobis' generalized distance is because this method takes into account the residual variances and covariance between the characteristics evaluated, thus needing an experimental design CARNEIRO 2003;REGAZZI;CARNEIRO, 2012).
The use of agronomic characters and multivariate techniques is common for plant breeding of different crops, such as maize (MORO; SILVEIRA; CARGNELUTTI FILHO, 2007, CARDOSO et al., 2009SIMON;KAMADA;MOITEIRO, 2012;MIRANDA, 2003), common bean (ELIAS et al., 2007;BARELLI et al., 2009;CORREA;GONÇALVES, 2012), castor bean (RODRIGUES et al., 2010) and cassava (ZUIN et al., 2009). According to Hallauer and Miranda Filho (1981), in addition to the high genetic divergence, the plant performance related to the desired characteristics in the desired environment is greatly important for choose the plant parents, in order to successfully obtaining superior lines.
The objective of this work was to evaluate the agronomic performance and estimate the genetic divergences of 18 cotton genotypes grown in the main season (sowed in December, 2012) and off season (sowed in January, 2013), considering their agronomic characteristics and resistance to Ramularia leaf spot.

Experimental area
The experiments were conducted in the experimental field of the Mato Grosso State University (UNEMAT), Tangara da Serra campus (14°39'0''S, 57°26'0''W and altitude of 320 m). The region has annual average temperature of 24,4ºC, annual average precipitation of 1,500 mm and relative humidity of 70-80%, with rainfall concentrated in October to March, and a dry season in April to September (DALLACORT et al., 2011).

Experimental design and materials
Two experiments were conducted with different sowing times, December, 2012 (main season) and January, 2013 (off season). A randomized block experimental design was used with 18 treatments (genotypes) (

A -Resistance to Ramularia leaf spot
The Ramularia leaf spot disease occurred naturally. The symptoms were observed 65 days after planting, when the weekly evaluations started, which continued up to the 128 th day after planting. The Ramularia leaf spot severity was evaluated using the diagrammatic scale proposed by Aquino et al. (2008). This scale consists of nine severity levels ranging from 0.05% to 67.20% of area injured by the disease, in which 1= 0.05%, 2= 0.50%, 3= 1.00%, 4= 2.00%, 5%= 4.00, 6= 8.00%, 7= 16.00%, 8= 32.00% and 9= 67.20%. The Ramularia leaf spot severity values were used to calculate the area under the disease progress curve (AUDPC) (SHANER; FINNEY, 1977): , where n is the number of evaluations, N is the Ramularia leaf spot severity and t i+1 -t i is the interval between two evaluations.

B -Plant architecture and production
The agronomic characteristics evaluated were plant height (PH) from the base to the apex (cm); first branch height (FBH) from the base to the first branch (cm); position of first fruiting branch (PFFB) (cm); height of first fruiting branch (HFFB) from the base to the first fruiting branch (cm); length between nodes (LBN) (cm); number of nodes (NN); average number of bolls per plant (ANBP); average boll weight (ABW), average weight of all bolls of six plants (g); 100-seed weight from the plant middle third (100SW), average of 100 seeds of six plants (g); fiber percentage (F%) from 20 bolls from the middle third of each plant; average production per plant (APP), average from the seed cotton production (g) of each plant; and yield (Y), cotton seed production (g) from each plot converted to Kg ha -1 . All variables were assessed in six plants per plot, except the yield.

C -Cotton fiber quality
The fiber characteristics evaluated were length (UHM), uniformity index (UI), resistance

Data analysis
The data of the characteristics evaluated were subjected to individual and joint analysis of variance and when significant, the means of the production characteristics were subjected to the Scott-Knott test at 5% probability. The homogeneity between growing periods was assessed by calculating the maximum F, as described by Cruz, Regazzi and Carneiro (2012). The genotypic coefficient of determination (h 2 ), was estimated by the genotypic and residue mean squares. The genetic divergence was assessed by the method of Mahalanobis' generalized distance (D 2 ) (MAHALANOBIS, 1936). Similar groups were determined by the Tocher's optimization method and the unweighted pair group method with arithmetic mean (UPGMA). The dissimilarity matrices were summed for the Tocher's optimization method and UPGMA to estimate the genetic divergence considering both growing periods. The correlation between the dissimilarity matrices of the main season, off season and joint periods was carried out using the methodology suggested by Mantel (1967), with 10,000 permutations. The validation of the UPGMA grouping was performed using the cophenetic correlation coefficient (CCC) (SOKAL; ROHLF, 1962). Analyses of variance and all multivariate analyzes were performed in the software Genes (CRUZ, 2006) and dendograms were generated using the software Mega-5.2 (TAMURA et al., 2011).

RESULTS AND DISCUSSION
The variance analysis of the experiment conducted in the main season showed significant differences between the genotypes averages for most of the characteristics evaluated ( Table 2).
The experiment conducted in the off season showed significant differences between the genotypes for the characteristics first branch height (FBH), elongation (ELG), yellowness level (+B) and maturity (MR). The coefficient of variation of the characteristics evaluated ranged from 1.14% (MR) to 28.04% for the area under the disease progress curve (AUDPC) in the main season, and from 0.57% for the uniformity index (UI) to 23.12% (AUDPC) in the off season. The coefficient of variation for most variables was relatively low in both growing periods ( Table 2).
Most of the genotypic coefficient of determination was higher in the off season, which ranged from 50.0% to 97.94, while in the main season, it ranged from 30.09% to 96.93%. Thus, the off season may be better for parent selection. For example, the genotypic coefficient of determination was higher for the AUDPC in the off season, indicating that the off season period evaluated is better for selecting this characteristic. This result may be related to the water stress that usually occurs in the off season (ECHER et al., 2010) and to the intensity of the R. areola inoculum, which is greater in the off season due to the various development cycles of the pathogen, denoting that Ramularia leaf spot occurs earlier in the off season.
The fiber reflectance (RD) and count strength index (CSP) were the only characteristics that presented no homogeneity of residual variances, with maximum calculated F of 7.16 and 7.98, respectively. The interaction between genotypes and growing periods (main season and off season) was significant for most of the agronomic characteristics evaluated, except the fiber characteristics ELG, RD, yellowness level (+B), MR and CSP, indicating that the genotype performance is dependent on the growing periods (Table 3). The coefficient of variation in the joint analysis was low for most of the characteristics evaluated, ranging from 1.07% to 25.85% (Table 3). Hoogerheide et al. (2007) evaluated eight cotton genotypes in different regions of the Mato Grosso State (MT), Brazil, and found significant interaction between the genotypes and environment, indicating that the genotype results had no consistency in the different locations.
The variables related to production are shown in Table 4, whose average values formed two groups for the main season, as well as for the off season. The genotypes BRS-335, FM-910, FM-951, 705-FMT, FMT-707, FMT-709, IMA-08-12427 and IMA-09-474 had the highest productions in both growing periods. The genotypes BRS-336, FMT-705 and IMA-08-12427 in the main season and the FMT-709 and BRS-336 in the off season presented the highest fiber percentage (F%).
The genotypes Nuopal-RR and IAC-08-2031 had the greatest average boll weight (ABW) in the off season (Table 4). On the other hand, the ABW of the 18 genotypes differed in only two groups in the main season, in which the the genotypes BRS-336,  and IMA-08-12427 had the lower values.
The average number of bolls per plant (ANBP) ranged from 5.87 to 9.80 in the main season and from 3.34 to 6.15 in the off season. The highest ANPB were presented by the genotype IMA-08-12427 in the main season and the IMA-09-2059 in the of-season (Table 4). Table 2. Individual analysis of variance of the 23 agronomic and fiber characteristics of 18 cotton genotypes grown in the main season and off season.
** significant at 1% by the F test; * significant at 5% by the F test; ns = not significant; h 2 = genotypic coefficient of determination; CV = coefficient of variation; Agronomic characteristics: PH = plant height; FBH = first branch height; PFFB = position of first fruiting branch; HFFB = height of first fruiting branch; LBN = length between nodes; NN = number of nodes; ANBP = average number of bolls per plant; ABW = average boll weight; 100SW = 100-Seed weight in the middle third; F% = fiber percentage; APP = average production per plant and Yield. Fiber technological characteristics: UHM = length; UI = uniformity index; STR = resistance index; ELG = elongation; MIC = micronaire; RD = fiber reflectance; +B = yellowness level; MR = maturity; SFI = short fiber index and CSP = Count Strength Index. Table 3. Joint analysis of variance of the 23 agronomic and fiber characteristics of 18 cotton genotypes grown in the main season and off season.
** significant at 1% by the F test; * significant at 5% by the F test; ns = not significant; CV = coefficient of variation;     Table 4. Characteristics related to production and response to Ramularia leaf spot of 18 cotton genotypes grown in the main season and off season.
Means followed by the same letter in the column do not differ significantly by the Scott-Knott test at 5% probability. Agronomic characteristics: Yield; APP = average production per plant; ANBP = average number of bolls per plant; ABW = average boll weight and F% = fiber percentage; AUDPC = Area under the disease progress curve.  Means followed by the same letter in the column do not differ significantly by the Scott-Knott test at 5% probability. Agronomic characteristics: Yield; APP = average production per plant; ANBP = average number of bolls per plant; ABW = average boll weight and F% = fiber percentage; AUDPC = Area under the disease progress curve.
The characteristic that presented the highest relative contribution, regarding the genetic divergence in the main season, were the average production per plant (APP) (18.15%), followed by the ANBP (12.71%), height of first fruiting branch (HFFB) (11.79%) and F% (8.99%). On the other hand, in the off season, the characteristics that most contributed to the genetic divergence were the F% (17.86%), ANBP (15.27%), number of nodes (NN) (11.15%) and AUDPC (10.76%) ( Table 5). The result of each characteristic was dependent on the period in which the experiment was conducted, thus, experiments in the off season may be more efficient for selecting genotypes with some characteristics, such as resistance to Ramularia leaf spot, since the contribution to the AUDPC divergence was significantly higher in the off season. The same dependence was observed for the agronomic characteristics, for which the main season may be more efficient for selecting superior genotypes regarding some characteristics, such as the APP, ANBP, HFFB and F%.  The grouping analysis by the Tocher's optimization method formed five groups for both growing periods, however, the genotypes contained in each group were different. In the main season, one group of twelve genotypes, one of two genotypes, and three of one genotype were formed. In the off season, one group of eight genotypes, one of six genotypes, one of two genotypes and two of one genotype were formed (Table 6).

Figure 1.
Dendrogram of the dissimilarity patterns established by the unweighted pair group method with arithmetic mean (UPGMA), using quantitative data of 18 cotton genotypes sowed in the main season and off season. (Cophenetic Correlation Coefficient CCC = 0.72, p<0.01, 10,000 permutations).
The two grouping methods, Tocher and UPGMA, presented different results for the main season, off season and when considering both growing periods. Suinaga, Bastos and Rangel (2005) estimated the divergence in cotton accessions and found similarity between both methods, differently from the results found in the present work.
The use of the UPGMA method combined with the Tocher's optimization method, for evaluations of dissimilarity, can ensure a greater certitude in estimating the genetic divergence of genotypes (ELIAS et al., 2007;ZUIN et al., 2009;SIMON;KAMADA;MOITEIRO, 2012  The correlation between the dissimilarity matrices of the main season and off season was not significant, which is another indication of heterogeneity in the response of the genotypes to the different growing periods. The correlation between the joint matrix and the mains season, and joint matrix and off season, showed a significant result (p<0.01) by the Mantel test, with similar amplitudes (0.7282 and 0.7284, respectively) ( Table 7).
The lack of correlation between the main season and off season (Table 7) indicates that the variability achieved in the off season cannot be achieved in the mains season, thus the need for genetic divergence studies in both periods. The evaluation of the divergence in both growing periods is important for lower the possibility of getting misinformation, since the environmental variability have to be considered in these studies. (CARGNELUTTI FILHO et al., 2008).  and Group VII (IMA-09-2059) ( Table 8). Based on the divergence assessed in both growing periods, the interbreeding between the FM-910 or BRS-335 and FMT-707 are recommended. These are not the the most genetically distant genotypes, but they had good performances related to production characteristics, and the genotype FMT-707 was among the most resistant to Ramularia leaf spot.

Main season
Off season Joint periods 0.7282** 0.7284** Off season 0.0639 ns The dendrogram developed by the UPGMA method, considering both growing periods (Figure 3), formed five groups, considering the cutoff point at 80, Group I consisting of six genotypes (BRS-335, Fibermax-966-LL, Nuopal, IMA-07-6035, Fibermax-975-LL, IMA-09-2059), Group II consisting of six genotypes , Group III consisting of one genotype (IMA-08-12427), Group IV consisting of three genotypes (IMA-09-474, IAC-08-2031 and IAC-09-848), and Group V consisting of two genotypes . Based on the dendrogram, which illustrate the pattern dissimilarity by the UPGMA method, crosses between genotypes BRS-335 and FMT-709 or IMA-08-12427 are recommended, since they are divergent and have good performances of production and resistance to Ramularia leaf spot. The Cophenetic Correlation Coefficient (CCC) for the dendrogram, which illustrate the pattern dissimilarity by the UPGMA method, was 0.72 for the main season, 0.69 for the off season and 0.68 for the joint periods (main season and off season) (Figures 1, 2 and 3). These CCC values are considered adequate, since values equal to or higher than 0.56 are considered ideals (VAZ PATTO et al., 2004), indicating that the dendograms obtained by the UPGMA method satisfactorily reproduces the information contained in the dissimilarity matrix.

CONCLUSIONS
The genotypes FMT-707, FMT-705, IAC-09-848 and IMA-08-12427 were the most resistant to Ramularia leaf spot in both growing periods evaluated (main season and off season), and can be used as sources of resistance to this disease in breeding programs.
The genetic variability found among the 18 cotton genotypes used in this study is sufficient to start a cotton breeding program.
The evaluation of cotton genotypes in different periods (main seasons and off season) provided a better assessment of the genetic variability, since the variability assessed in this work in main season was different than the variability accessed in the off season.