Agro-morphological characterization of upland rice accessions

Agro-morphological characterization of germplasm accessions is fundamental in order to provide information for plant breeding programs (Lin, 1991). Several researchers reported the use of agro-morphological markers in the characterization and study of rice (Oryza sativa L.) germplasm diversity. Li et al. (2000) obtained consistent results analyzing the correlation between genetic and morphological differentiation in 111 accessions of rice from the Japonica and Indica groups. Yawen et al. (2003) studied the genetic diversity on 5285 accessions of indigenous rice in China and found considerable morphological variation among accessions. In India, Patra and Dhua (2003) analyzed the agro-morphological diversity of upland rice and in Vietnam, Fukuoka et al. (2006) assessed the variability in agronomic characters among landraces of aromatic rice populations. Using 42 agro-morphological characters associated with 39 SSR markers, Bajracharya et al. (2006) examined the genetic relationships among 147 high altitude rice landraces from Jumla, Nepal. Seetharam et al. (2009) characterized 30 rice genotypes comprising landraces, pure lines, somaclones, breeding lines and varieties specifically adapted to costal saline environments using SSR markers and morphological characters. Yibo et al. (2010) investigated morphological and genetic diversity in 11 surviving populations of O. rufipogon Griff. from Hainan Island, China, and reported a clear morphological differentiation among or within populations. This type of characterization has also been used on wild species in Brazil. Rosa et al. (2006) characterized eight populations of O. glumaepatula Steud. collected in different hydrographic basins and Veasey et al. (2008) analyzed the genetic variability among the four wild rice species occurring in South America, reporting agro-morphological variation for O. glumaepatula, followed by O. latifolia. The aim of the present study was to characterize 146 upland rice genotypes using agro-morphological characters and to provide useful information to facilitate the choice of genitors for rice plant breeding programs.


Introduction
Agro-morphological characterization of germplasm accessions is fundamental in order to provide information for plant breeding programs (Lin, 1991).Several researchers reported the use of agro-morphological markers in the characterization and study of rice (Oryza sativa L.) germplasm diversity.Li et al. (2000) obtained consistent results analyzing the correlation between genetic and morphological differentiation in 111 accessions of rice from the Japonica and Indica groups.Yawen et al. (2003) studied the genetic diversity on 5285 accessions of indigenous rice in China and found considerable morphological variation among accessions.In India, Patra and Dhua (2003) analyzed the agro-morphological diversity of upland rice and in Vietnam, Fukuoka et al. (2006) assessed the variability in agronomic characters among landraces of aromatic rice populations.Using 42 agro-morphological characters associated with 39 SSR markers, Bajracharya et al. (2006) examined the genetic relationships among 147 high altitude rice landraces from Jumla, Nepal.Seetharam et al. (2009) characterized 30 rice genotypes comprising landraces, pure lines, somaclones, breeding lines and varieties specifically adapted to costal saline environments using SSR markers and morphological characters.Yibo et al. (2010) investigated morphological and genetic diversity in 11 surviving populations of O. rufipogon Griff.from Hainan Island, China, and reported a clear morphological differentiation among or within populations.
This type of characterization has also been used on wild species in Brazil.Rosa et al. (2006) characterized eight populations of O. glumaepatula Steud.collected in different hydrographic basins and Veasey et al. (2008) analyzed the genetic variability among the four wild rice species occurring in South America, reporting agro-morphological variation for O. glumaepatula, followed by O. latifolia.
The aim of the present study was to characterize 146 upland rice genotypes using agro-morphological characters and to provide useful information to facilitate the choice of genitors for rice plant breeding programs.

Materials and Methods
One hundred and forty-six accessions of upland rice were assessed in this study (Table 1).Three Brazilian cultivars were used as controls: IAC 202, IAC 47 and Bonança.
The experiment was carried out in Recife, state of Pernambuco in northeast Brazil, (08°01' S; 34°56' W -10.3 m altitude), from March to July 2007, during the agricultural year in the region.The mean temperature during the experiment was 23.8 °C and the index annual rainfall was 1.193 mm.
Seeds were germinated in polystyrene trays with a substrate composed of Pinus husk, vermiculite, fine peat, macro and micronutrients.Twenty-eight days after emergence, the most vigorous seedlings of each genotype were transplanted to the field, following the procedure reported by Villela and Furlani Júnior (1996).The statistical experiment design was a randomized block with three replicates.Each plot consisted of a row of six plants spaced 0.15 m apart, with the rows spaced 0.40 m apart.
The accessions were characterizated according to descriptors established by the International Rice Research Institute and the International Board for Plant Genetic Resources (IRRI, 1980).A total of 14 qualitative (Table 2) and 14 quantitative descriptors (Table 3) were used.The qualitative descriptors were: leaf color (LC), leaf pubescence (LP), lodging resistance (LR), internode ABSTRACT: The agro-morphological characterization is fundamental in order to provide information for plant breeding programs.The aim of the present study was to characterize 146 accessions of upland rice (Oryza sativa L.), based on qualitative and quantitative agro-morphological descriptors.The experiment was conducted in Recife, state of Pernambuco, Brazil, using a randomized block design with three replicates.Polymorphism was observed among 12 of 14 qualitative characters evaluated, whereas significant differences (p < 0.05) were observed for 11 of the 14 analysed quantitative traits.Genetic variance was higher than environmental variance and the average inheritability coefficients were above 80 % for all characters, which ensures the predominance of the genetic components in the differences observed among accessions.On the cluster analysis for qualitative traits the accessions were classified in two groups with a total of 18 duplicates, whereas for the quantitative traits three groups were obtained with few subgroups.The principal component analysis for quantitative traits showed great dispersion of the accessions.The most divergent group of accessions included the genotypes Mitsukasane, Mie, Tomoe mochi, Ooba kirishima and Nourin mochi 6, which showed a higher number of spicklets per plant.There is high variability among the rice accessions from the germplasm collection studied, which presents great importance for breeding programs or for genetic studies on this species.Keywords: genetic variability, germplasm, ex-situ conservation, pre-breeding  1 -difficult (less than 25 % of the grains were removed); 2 -intermediate (from 25 % to 50 % of the grains were removed); 3 -easy (more than 50 % of the grains were removed).
After maturation For the quantitative characters the statistical analyses were performed using the F test and in order to group the accessions, the Scott and Knott (1974) test was used.Statistical significance was set at 5 % (p < 0.05) and the GENES software was used for all tests (Cruz, 2006).The software PAST (Hammer et al., 2001) was used for the cluster and principal component analyses.Qualitative and quantitative traits data were analyzed separately.
The qualitative traits were transformed into binary data considering the presence or absence (1/0) of each character state.The Jaccard similarity coefficient was estimated for qualitative traits, while Euclidean distances were obtained for the quantitative traits.UPGMA (Unweighted pair group method with arithmetic mean) was the clustering method used for both qualitative and quantitative data.

Results and Discussion
Polymorphism was found in 12 of the 14 qualitative traits studied; the non-polymorphic traits were the internode color (light green) and panicle type (intermediate).Panicle type (PT) is generally affected by water deficiency before flowering, because this condition changes the angle of the panicle branches and the density of the grains (Bonow et al., 2007).Internode color (IC) often shows variation on color tonality depending on the age of the plants and the soil fertility conditions.Plants usually present a dark green tonality on high soil fertility and a light green one on low soil fertility (Fonseca et al., 2002), which agrees with the soil type used in this study, a franc-clay and low fertility one.
For the traits stigma color (SC) and glumella pubescence (GP), two alternative forms or types were observed.White (65 %) and purple (35 %) genotypes were observed for SC, while presence (76 %) and absence (24 %) of the pubescence were found in GP.Veasey et al. (2008) reported variation within and among populations of O. glumaepatula for stigma color, with three populations 100 % white, four populations 100 % purple, and the others with varying degrees for each color, while white was the only color found on the other South American wild rice species (O.latifolia, O. grandiglumis and O. alta).
The characters showing higher variability were leaf pubescence (74 % pubescent, 20 % glabrous and 6 % intermediate), lodging resistance (37.5 % absent, 37.5 % few lodged and 25 % moderate), presence and distribution of awn per panicle (PDAP) (55 % absent, 23 % present only at the end of the panicle and 22 % present on all panicle), threshability (73 % easy threshability, 21 % intermediate and 6 % difficult), endosperm type (31 % glutinous endosperm, 42 % intermediate and 27 % not glutinous), and grain shape (74 % semi-long, 20 % long, 5 % tightly long and 1 % semi-rounded grain shape).The grain shape character also showed the highest variation in studies conducted in Pakistan by Siddiqui et al. (2007).Although this character is mentioned as qualitative, its evaluation is carried out as a quantitative trait, according to the grain dimentions.The major variation observed was probably due to polygene action and the influence of environmental causes.In relation to PDAP, while the awn is present its length should be influenced by the soil fertilization and plant density (Fonseca et al., 2002).Studies carried by Bisne and Sarawgi (2008) to characterize 32 aromatic rice accessions of Badshah Bhog group from IGKV, Raipur, Chhattisgarh germplasm, found the highest variation among accessions for the traits leaf blade colour, lemma and palea colour, apiculus colour, and lemma and palea pubescence.
Considering the 14 quantitative traits analyzed in this study, differences among accessions (p < 0.05) were found for 11 traits (Table 4).The non-significant traits were flag leaf width, culm diameter and the ratio culm/width of the grain.Veasey et al. (2008) also found significant differences and high variability for 11 populations of O. glumaepatula, mainly for the characters number of tillers, plant height at flowering, leaf length and width, culm length, days to heading, panicle number, panicle height, flag leaf length, spikelet length and awn length.
The coefficients of variance (CV%) showed an optimum experiment precision, with values below 11 % for all characters, except for PPP (CV = 24.2%).The analysis performed by Patra and Dhua (2003) had CV values below 27.1 % for all agronomic traits evaluated, except for the character number of panicles, directly related to grain production, showing a CV value of 30.7.
Results concerning the estimates of variance components among accessions indicate that the variance observed was due to genetic factors, as the values obtained for the b index (CVg/CVe) were above 1 for all characters (Table 5).When the b index is higher than 1, there are favorable conditions for selecting superior genotypes considering the predominance of genetic factors rather than environmental factors (Vencovsky and Barriga, 1992).The traits number of spikelets per panicle, culm length, panicle fertility, flowering cycle and flag leaf length   , 1821.59, 180.30, 90.13, 54.81 and 53.85, respectively (Table 5).The estimate of the average inheritability among accessions was high, over 82 % for all quantitative traits.The better prediction of the environmental component of phenotypic variance, helps the more realiable estimation of both broad and narrow sense heritability, and so the better planning program for cross and self pollinated plants improvement (Fahliani et al., 2010).
The Scott and Knott (1974) (p < 0.05) analysis allowed the formation of several groups among accessions for all characters (not shown data).The flag leaf length character classified the accessions in eight groups.The group with the highest average showed a variation from 58.2 cm (accession 121) to 64.3 cm (accession 147), while the group with the lowest average varied from 25.9 cm (accession 10) to 30.5 cm (accession 90).Eight groups were obtained for the character flag leaf width, varying from 2.52 cm (accession 6) to 2.7 cm (accession 99) for the group with higher scores and from 1.36 cm (accession 91) to 1.63 cm (accession 137) for the group with lower scores.
Finally, character plant production classified the accessions in four groups.The genotypes that presented the higher means for this trait were Oota wase (accession 72), 35.6 g per plant, Bansei tarou (accession 123), 33.8 g per plant, Jouon (accession 71), 31.1 g per plant and Nourin mochi 1 (accession 37), 31.0 g per plant.The group that showed the lower means varied from 4.6 g per plant (accession 35) to 13.0 g per plant (accession 38).These same traits showed variation among 147 rice accessions in studies conducted by Bajracharya et al. (2006).In studies of ecogeografic and genetic diversity based on morphological characters of indigenous rice in China, Yawen et al. (2003) found wide variations for characters related to production, such as length of the panicle (10-36 cm) and grain number per panicle (30-340).
In the cluster analysis for the quantitative traits using the Euclidean distance, three groups were formed (Figure 3).One of the groups (group I) classified the accessions 56, 147 and 81 as more divergent than the others, for presenting higher number of panicles, fertile spikelets and plant production, which are important agronomic characters.The analysis of the descriptors used showed that the number of panicles per plant was the most important character in the group classification.The scatter graph considering the quantitative traits showed that accessions 56 (Ooba kirishima), 81 (Nourin mochi 6) and 147 (IAC 202,control), belonging to the most divergent group in the cluster analysis (group I), were also more divergent for the principal component analysis (PCA), followed by accessions 15 (Mitsukasane), 16 (Mie), 47 (Tomoe mochi) and 149 (Bonança, control) (Figure 4), which belong to a subgroup in group 2 of the cluster analy- sis.These seven accessions had high number of spicklets per plant, while accessions 47 and 147 showed high panicle lengths; accession 147 showed high flag leaf lengths as well.This last accession is the genotype IAC 202, one of the controls used in this study.The first and second components in the PCA represented 83.1 % and 8.4 % of the total variation among accessions.If the initial components accumulate a relatively high percentage of total variation, generally above 80 %, they satisfactorily explain the variability among individuals (Madia et al., 1979).
No duplicates were identified among the studied accessions for quantitative traits in the cluster analysis (Figure 3), which means there is a high diversity among the accessions for these traits.Cluster analysis based on agro-morphological diversity assessment in rice was also reported in several studies, such as Ghalain (2006), Naik et al. (2006), Hien et al. (2007), Sarawgi and Bhisne (2007), Mathure et al. (2010), Li et al. (2010) and Zhang et al. (2010).Seetharam et al. (2009) characterized 30 rice genotypes comprising landraces, pure lines, somaclones, breeding lines and varieties by SSR markers and morphological characters.At Euclidian distances the genotypes were grouped into five clusters and the PCA components explained 38.4 % of total variation.Therefore, all data presented in this study allowed the identification of cultivars with distinct morphological traits that could be used as potential parents to facilitate high resolution QTL mapping and validate candidate genes responsible for quantitatively agronomic characters (Li et al., 2010).

Figure 1 -
Figure 1 -Dendrogram of similarity patterns by the UPGMA method based on Jaccard similarity coeffi cient estimated from 14 qualitative agromorphological traits analyzed in 146 accessions of Japanese upland rice and three cultivars (controls) used in the experiment.

Figure 2 -
Figure 2 -Scatter graph of 146 accessions of Japanese rice upland germplasm collection of the Universidade Federal Rural de Pernambuco -UFRPE and cultivars (controls) used in the experiment from 14 qualitative agro-morphological traits.

Figure 3 -
Figure 3 -Dendrogram of similarity patterns by the UPGMA method based on the average Euclidean Distance estimated from 14 quantitative agromorphological traits analysed in 146 accessions of Japanese rice and and cultivars used in the experiment.

Figure 4 -
Figure 4 -Scatter graph of 146 accessions of Japanese rice upland germplasm collection of the Universidade Federal Rural de Pernambuco -UFRPE and cultivars (controls) used in the experiment from 14 quantitative agro-morphological traits.

Table 1 -
List of accessions of upland rice from Japan from the germplasm collection of Universidade Federal Rural de Pernambuco -UFRPE and the Brazilian cultivars (controls) studied.

Table 2 -
Qualitative descriptors used to evaluate 146 upland rice accessions from Japan and three Brazilian cultivars (controls).

Table 3 -
Quantitative descriptors used to evaluate 146 upland rice accessions from Japan and three Brazilian cultivars (controls).

Table 4 -
CV -coeffi cient of variation. 3*Signifi cant at a 0.05 probability level, by F test.; ns not signifi cant at a 0.05 probability level, by F test.

Table 5 -
Estimates of the variance components, phenotypic and genetic parameters of 14 quantitative traits for 146 upland rice accessions from Japan and three Brazilian cultivars (controls).