Morphological variation and isozyme diversity in Dioscorea alata L . landraces from Vale do Ribeira , Brazil

Traditional growers of the Vale do Ribeira, São Paulo State, grow and make use of several D. alata landraces. This study assessed the genetic diversity of 16 landraces, using isozymatic and morphological markers and comparing them with 19 commercial varieties of D. alata. Their distribution in different levels of organization such as households and communities of the Vale do Ribeira was evaluated. Isozymatic analyses were performed with polyacrylamide (six systems) and starch gels (one system), while the morphological analyses were carried out with 24 traits. Due to the polyploid nature of this species, the isozymatic bands were scored as binary data. Morphological traits were also scored as binary data. Principal coordinates and cluster analyses were conducted for both markers, using for the later the Jaccard ́s similarity coefficient and UPGMA method. The separation of the landraces from the commercial varieties, which showed lower genetic diversity, was reported for both markers. No correlation between genetic and geographical distances was found for both data, which suggests that the observed variability is not spatially structured. Also, no correlation was found between both markers. The analysis of molecular variance (AMOVA) indicated that genetic diversity was mainly found within households for both isozymatic (54%) and morphological (70%) markers. The results obtained for both markers revealed the importance of traditional agriculturists in the Vale do Ribeira in maintaining high diversity for D. alata, even higher than the varieties commercialized in São Paulo State, contributing for the in situ/on farm conservation of this crop.


Introduction
Yam (Dioscorea spp.) is a horticulture crop with expressive world consumption, especially in developing countries.Dioscorea alata L. is probably the cultivated species with the largest world distribution and one of the oldest cultivated yam species (Lebot, 2009).D. alata is the most popular species for cultivation in Brazil, with varieties commonly known as: 'Mimoso', 'Florida' and 'Caipira'.In southern areas of the State of São Paulo (SP), Brazil, traditional farmers practice family farming using the slash and burn system and grow different D. alata landraces, used mostly for their own consumption (Bressan et al., 2005).Some households also grow landraces for local markets, following the same model of subsistence farming, which is characterized by low-energy and intense family labor.
Morphological characterization of D. alata accessions has been based on the list of descriptors recommended by IPGRI/IITA, in Nigeria (IPGRI/IITA, 1997), such as the studies of Cruz et al. (1999), Hasan et al. (2008), Mahalakshmi et al. (2007) and Rhodes and Martin (1972).The isozymatic genetic characterization in D. alata was performed by Lebot et al. (1998) and Twyford (1990).This species was also studied with molecular markers such as RAPD, for the evaluation of resistance to anthracnose (Mignouna et al., 2002a) and genetic diversity studies (Asemota et al., 1996;Cheng and Liu, 1996;Zannou et al., 2009); AFLP, for the study of intraspecific genetic diversity (Egesi et al., 2006;Malapa et al., 2005) and development of a genetic map and analysis of QTL aimed at resistance to anthracnose (Mignouna et al., 2002b).Recently, Arnau et al. (2009) used six microsatellite primers to study the segregation patterns from four different progenies of D. alata.
This study assessed the genetic diversity, through isozyme markers and morphological traits, of commercial varieties and landraces of D. alata from the Vale do Ribeira, verifying how this diversity is distributed at different levels of organization: within households, between households within communities and between communities of the Vale do Ribeira, in the State São Paulo.

Materials and Methods
We evaluated 16 D. alata landraces obtained from the communities that practice traditional farming in the Vale do Ribeira, SP, such as Aroeira, Porto Cubatão, Rio Branco, Mandira, Porto do Meio and Santa Maria in Cananéia (25º00'S; 47º55'W); Sapatu in Eldorado Paulista (24º36'S; 48º22'W); and Icapara, Momuna, Peropava and Arataca in Iguape (24º42'S;47º33'W) (Table 1; Figure 1).In addition, we used two accessions from the germplasm collection of the Agronomic Institute (IAC), Campinas, SP, Brazil, (22º53' S; 47º04' W), and one from Embrapa Cassava and Tropical Fruits, Cruz das Almas, state of Bahia, Brazil, (12º40' S; 39°06' W).To compare with the diversity managed by traditional farmers in the Vale do Ribeira, 16 commercial varieties were also acquired in markets located in Iguape , Santa Rosa de Viterbo (21º28' S; 47º21' W), Piracicaba (22º42' S; 47º38' W) and Matão (21º36' S; 48º22' W), and at a farm in Araras (22º21' S; 47º23' W).From each farm visited in the Vale do Ribeira, a landrace represented by a tuber was collected, except for the households 15, 17 and 21 in the communities 1 Commercial varieties.Morphological evaluations were carried out on individual plants of each variety in the experimental field.The morphological descriptors used in this study were recommended by the International Plant Genetic Resources Institute (IPGRI) and International Institute of Tropical Agriculture (IITA) in Ibadan, Nigeria (IPGRI/IITA, 1997).We have evaluated 24 traits, nine related to leaf characters, six to stem, seven related to the tubers and one related to flowering.These observations were based on phenotypic average of ten measurements for the quantitative descriptors for leaves, stems and tubers in each individual plant.We assessed a total of nine quantitative descriptors and 15 qualitative traits, which were transformed into different classes, attributing codes to each class (Table 2).The morphological evaluations were performed in two stages: the first assessment was carried out six months after transplanting the accessions on the field, as there was sufficient availability of tissues (young leaves, adult leaves and stems), and the second during harvest when the tubers were already formed.In a preliminary step we intended to adapt the protocols for polyacrylimide (discontinuous system) and starch gels (Alfenas et al., 1991;Mignouna et al., 2002c) for D. alata.After optimizing the methodology, we proceeded with the genetic evaluation based on isozymes.
The enzymes were extracted from recently expanded leaves (200 mg) from adult plants (6-8 months).The leaves were ground in liquid nitrogen in 1.5 mL microtubes using a power homogenizer, adding 1 mL of the extraction buffer as described by Mignouna et al. (2002c), and also adding 3% α-mercaptoethanol to minimize damages caused by oxidation.The resulting raw extract was centrifuged at 18,000 g for 20 min at 4ºC.After that, 130 mL of the supernatant were placed in 0.5 mL microtubes, diluted in 150 mL of a solution containing Tris-HCl pH 6.8 and Coomassie blue, the latter component indicating the protein migration in the gel.This amount of extract was enough for four gels, which in its turn allowed the assessment of 38 individuals each.The same control plant was added in each gel as a marker.The voltage was set at 50 V for three hours, and then adjusted for 100 V for the next 13 h, usually running overnight at 4ºC.
Due to the polyploid nature of D. alata, data was genotyped as binary data (presence = 1 and absence = 0), generating a binary matrix.The morphological data was also analyzed by transforming each character state into a code within each descriptor, and these codes were subsequently transformed into binary data.For the isozymatic analyses, the number of bands per isozyme system, the total number of bands and the percentage of polymorphic bands were also registered.
For both markers Jaccard similarity indices were calculated between pairs of individuals based on binary matrices of allozyme and morphological data.Cluster analyses were performed using Jaccard's coefficient and the UPGMA (Unweighted pair group method with an arithmetic mean) clustering method, using the NTSYSpc software version 2.0 (Rohlf, 1992).The groups × stability in the dendrograms were based on estimates of genetic dissimilarities by the procedure of resampling with 1,000 bootstraps using the program BOOD version 2.0 (Coelho, 2001).Principal coordinate analyses were also held for both data, utilizing the software PAST (Hammer et al., 2001).
Binary matrices of allozyme and morphological data were used for the analysis of molecular variance (AMOVA), in or-der to be considered as a summary of molecular and morphological diversity, for a better understanding of the distribution of diversity, according to the hierarchical levels (households and communities).The AMOVA was used only for the materials collected in the fields of traditional farmers in the Vale do Ribeira.For this purpose, we used the squared distances, as described by Excoffier et al. (1992) with the program Arlequin (Schneider et al., 2004).
To evaluate the patterns of spatial variation, we attempted to correlate genetic diversity (isozymatic and morphological) with spatial structure (geographic distances) in order to try to understand the evolutionary dynamics of this crop within traditional agricultural systems in the Vale do Ribeira.The coefficient used was the correlation coefficient (r) through the matrix of genetic distances calculated by the similarity index of Jaccard and geographical distance between the collection sites.Geographical distances were estimated from the original coordinates of the home-gardens and swidden fields using the software TrackMaker version 12.3 (Ferreira Júnior, 2004).The Pearson correlation coefficient (r) was also used to verify the association between the isozymatic and morphological diversity.The significance levels of correlations were checked by the Mantel test (Z), using 1000 random permutations (Mantel, 1967).To obtain the similarity matrixes and to apply the Mantel test the software NTSYSpc version 2.0 was used (Rohlf, 1992).

Results and Discussion
The SOD system showed the highest number of bands (13) followed by PGI and MDH with 12, GOT with 10, G 6 PDH with 9, SKDH with 8 and PGM with 7, in a total of 71 bands for all systems.Furthermore, we observed that 100% of the bands were polymorphic.The combination of different isozymic bands of all enzymes found in both types of gels (starch and polyacrylamide) discriminated the 16 landraces and 13 of the 19 commercial varieties analyzed.
A clear distinction was observed between the landraces (groups I to III, V and VI) and the commercial varieties (group IV), including the accessions obtained from IAC and two landraces from Peropava and Porto Cubatão (Figure 2).The farmer in Peropava reported, during our visit, that he purchased his varieties at CEASA (market supply and distribution center) in the municipality of São Paulo, which explains their proximity to the group of commercial varieties.Nevertheless, the variety from Porto Cubatão was acquired at a market in this community, which led us to the conclusion that it was also from CEASA.Another observation implied from the cluster analysis was that the commercial varieties showed less diversity (0.55 similarity range), while the landraces presented higher diversity (0.71).Similar results were reported by Raji et al. ( 2009) for cassava, where elite cultivars were clearly separated from African landraces evaluated by SSR markers, although less clearly with AFLP markers.Also, corroborating our results, the amount of within-population variation was higher for landraces for both markers, indicating a higher genetic diversity for the landraces.Total similarity ranged from 1.00 to 0.19 in our study indicating, thereby, high diversity among accessions of D. alata.
The presence of possible duplicates, showing the same isozyme phenotypes, was also observed in Figure 2.This occurred among the commercial varieties 23 and 24 from Piracicaba, 8 and 9 from Iguape, 11, 12 and 15 from Santa Rosa de Viterbo and Fazenda Pastre in Araras, and among individuals 13, 14 and 17 from the municipalities of Santa Rosa de Viterbo, Araras and Piracicaba (Table 1).These materials are commercialized in local markets or supermarkets in the State of São Paulo through CEASA, which explains its higher similarity and the possibility of being clones.
The scatter graph obtained from the principal coordinates analysis (Figure 3) showed the separation of landraces from the municipalities of Iguape, Cananéia and Sapatu, in the Vale do Ribeira, from the commercial varieties, which showed greater genetic similarity among each other and were grouped on the left, along with the main axis of the principal coordinate 1, which explained 30% of total variation.The variety originated from Embrapa, in Cruz das Almas, was closer to the landraces from Cananéia according to the dendrogram (Figure 2).The scatter graph also showed the highest variability observed among landraces, with the group II, which is composed of landraces from Cananéia and Iguape, being more distant from the other landraces (Figure 3).
The data obtained from this study show the high isozymic diversity among landraces of the Vale do Ribeira, indicating that traditional farmers maintain and manage high levels of genetic diversity for this species in their home gardens or swidden fields.The communities that cultivate landraces with lesser levels of diversity among themselves are Porto do Meio and Santa Maria, which are geographically closer.A clear separation between the municipalities of the Vale do Ribeira was not observed in this study, as occurred for another studied species, D. trifida (Bressan et al., 2005).
We attempted to correlate the genetic distances and geographic distances of the landraces of the Vale do Ribeira, but they were nil (r = -0.11),although statistically significant (p = 0.01), suggesting that there is no correlation between these two parameters.The lack of correlation between genetic and geographical distances indicates that there must be a lot of exchange of landraces among farmers, which promotes gene flow among households and even among communities, and that the distribution of landraces is linked to anthropogenic factors, especially considering that this species is the most commercialized in the State of São Paulo.
Based on data from 71 polymorphic bands using molecular variance analysis (AMOVA) considering only the 16 landraces cultivated by traditional farmers in the Vale do Ribeira, we found that most of the isozymatic variability is distributed within households (54%), thus emphasizing the importance of the basic evolutionary units (households) in the distribution of genetic diversity (Table 3).Similar results with isozyme markers were observed by Sambatti et al. (2000) for cassava (Manihot esculenta) cultivated by traditional farmers in the Vale do Ribeira, and Faraldo et al. (2000), who found higher genetic diversity in cassava within different re-  In Brazil, this is the first report on isozyme diversity among landraces of D. alata grown by traditional farmers of the Vale do Ribeira.D. alata was studied using isozymes by Twyford (1990), identifying materials originated from tissue culture for the species D. alata, D. cayenensis/rotundata, D. esculenta and D. bulbifera.Another study using four polymorphic enzyme systems (MDH, PGI, SKDH, 6PGD [6-phosphogluconate dehydrogenase]) revealed 66 isozyme phenotypes in 269 cultivars of D. alata (Lebot et al., 1998), each uniquely characterized by the presence or absence of 27 electromorphs.The authors found that identical phenotypes were distributed in different geographical areas and vice versa, with different phenotypes within the same geographic zone.
From the 24 morphological characters evaluated, eight were monomorphic, ie, the 35 accessions showed opposite leaves with only one leaflet, absence of spines and presence of wings on the stem, anticlockwise twining direction, octagonal shape stem, presence of tubers in 100% of the accessions and brown tuber skin colour.Most of these characters define the species D. alata.For the other traits, most of the accessions showed green stems (91%), with 5.7% (two individuals) showing purplish green and 2.9% (one individual) brownish green stems.Stem diameters were, in the majority (69%) between 0.4-0.6 cm, followed by < 0.4 cm stems (31%).Most accessions (97%) had sagittate leaf shape and green coloured petioles (89%), followed by purple (8.6%), with only one individual (2.9%) showing brownish green petioles.The majority (80%) had petiole length measuring from 50 to 100 mm.Regarding the distance between the insertion of the leaf petiole and the upper tip of the leaf (L3), the majority (66%) had a < 20 mm distance, while the others (34%) ranged from 20-40 mm.Still, the majority (86%) had a distance of 100-150 mm between the insertion of the leaf petiole and the lower tip of the leaf (L2).With regard to the larger portion of width (W1), the majority (77%) showed a width of < 100 mm, while the width between the leaf lobes (W2) ranged from < 60 mm (46%) and 60-100 mm (54%).Concerning the flowering aspects, few accessions (8.6%) flowered under the experimental conditions in Piracicaba, SP.Seed production was not observed for the accessions that showed flowering.
Regarding the tubers, most accessions had several tubers (86%) of elongated shape (86%), with white coloured flesh (89%), followed by three individuals with purple flesh (8.6%) and only one with a yellow flesh (2.9%).The three accessions showing purple (27, 28 and 30) and yellow flesh (26) were landraces from Cananéia municipality.Most accessions (77%) had < 200 mm tuber length, while the width of the tubers ranged from 70-120 mm (49%), < 70 mm (37%) and > 400 mm (14%).Rhodes and Martin (1972) used 100 characters to study the variation in 30 D. alata cultivars, concluding that only 28% of the characters used were useful for the classification of the cultivars.In our study, of the 24 traits used, 67% were discriminatory.Hasan et al. (2008) used 47 morpho-logical traits to assess 70 D. alata accessions collected throughout Malaysia.The results obtained from the principal component analysis (PCA) indicated that the characters that most contributed to the the species variability were those related to shape, size and flesh colour of underground tubers; shape and colour of aerial tubers; position, shape, size and vein colour of the leaves; petiole colour; shoot growth rate; and number of days for shoots to germinate.Mahalakshmi et al. (2007) reported the use of morphological trais to obtain a core collection for D. alata, evaluating a total of 772 accessions.The study identified 72 accessions to be part of the core collection using 77 qualitative and 11 quantitative traits.
A similarity range from 0.48 to 1.0 was observed in the cluster analysis of 35 D. alata accessions (16 landraces and 19 commercial varieties) (Figure 4), indicating that the set of landraces and commercial varieties evaluated present high variation in morphological characters, although the lesser magnitude than the variation observed for isozymes (Figure 2).This analysis classified the accessions in three groups (Figure 4), with group I containing most of the landraces from Iguape and the accession from Cruz das Almas, Embrapa (25), group II including all commercial varieties, and landraces from Cananéia (Rio Branco, Porto do Meio, Mandira, Santa Maria and Aroeira), Eldorado (Sapatu), and Iguape (Momuna and Peropava), while group III consisted of three landraces from Cananéia (Aroeira, Porto Cubatão and Rio Branco).Group III differed from the others mainly because of the purple colour of the petiole, purple flesh tuber colour, and the irregular shape of the tubers.Group I, comprised by landraces from Iguape and one accession from Cruz das Almas, differed from the others mainly due to the larger leaf sizes (W3, W2 and L2).Similarly to what was observed with isozymes, the commercial varieties obtained in markets, as well as all accessions from the IAC, were all clustered in group II, apart of the landraces from Cananéia and Eldorado, which stood in a differentiated sub-group.The exception was observed for the landraces 3, from Momuna, and 6, from Peropava, both from Iguape, although the latter was acquired in CEASA, and therefore it was expected to show more similarity to the commercial varieties.
Duplicates were observed, with the similarity coefficient equal to 1, meaning that these individuals had the same morphological phenotype, such as landrace 6 from Peropava community and the commercial varieties 7 and 10, acquired in markets of Iguape municipality, again confirming the proximity of the Peropava accession with the commercial varieties.Duplicates were also found between the commercial varieties 11 and 13 from Santa Rosa de Viterbo, 22, from IAC, and 23, from Piracicaba, and between the landraces 33, from Santa Maria, in Cananéia, and 34 and 35, from Sapatu, Eldorado municipality.Unfortunately, the possible duplicates found using isozyme markers were not coincident with duplicates found using morphological characters.
On the scatter graph obtained from the principal coordinate analysis (Figure 5), with the first two coordinates accounting for 45% of total variability, the three groups ob-served in the dendrogram were clearly formed, with the first coordinate separating group I, on the left, from group II on the right of this parameter, and the second coordinate separating group III from the others.
The correlation between genetic distances and geographical distance was zero (r = 0.03), suggesting that morphological diversity is not structured in space, probably due to the existence of an exchange of materials or mixture of varieties among farmers.Lebot et al. (1998), studying 269 cultivars of D. alata to determine genetic relationships among D. alata originating from the South Pacific, Asia, Africa and the Caribbean with four polymorphic enzyme systems (MDH, PGI, SKDH, 6PGD) and 25 morpho-agronomic descriptors, did not find an agreement between patterns of morphological variation, geographic origins and isozyme polymorphism either.
The molecular variance analysis (AMOVA) based on morphological markers (24 characters), showed that 70% of the morphological diversity were found within households (Table 3), indicating the management of traditional farmers in the selection and maintenance of different materials within the evolutionary units (households).This finding confirms the results obtained by Sambatti et al. (2000) with cassava and Veasey et al. (2007) with sweet potato cultivated by traditional farmers in the Vale do Ribeira.Sambatti et al. (2000) demonstrated that most of the variability, both morphological and isozymic, for the cassava crop was distributed within households.According to Peroni and Martins (2000), the maintenance of diversity and its amplification at the garden level are consequences of the interaction between farmer management and life history components of the crops.In this study we found a large number of local names ("cará guaçu" [guaçu yam], "cará aipim" [cassava yam], "cará branco" [white yam], "cará são João" [Saint John's yam], "cará São João branco" [white Saint John's yam], "cará São João roxo" [purple Saint John's yam], "cará bolo" [cake yam], "cará coco" [coconut yam] and "cará roxo" [purple yam]), that are used by farmers, sometimes hiding levels of diversity not recognized by them.However, the name "cará guaçu" was attributed to landraces 1 and 5 (both from group I), and the name "cará aipim" was attributed to landraces 2 and 4 (both also from Group I).The name "cará branco" was assigned to accessions 3 and 35 (both from Group II) (Figure 4), along with all commercial varieties of white flesh.
The correlation coefficient between morphological and isozymic genetic distances was positive (r = 0.27) but not   significant (p = 0.99), suggesting that there is no correlation between these two parameters.Therefore, traditional farmers are faced with some morphological changes that do not reflect the isozymic differences in D. alata.The lack of correlation between molecular and morphological genetic distances is due to poor genomic coverage achieved with the isozyme loci.However, Telles et al. (2001) showed in their research on natural populations of a Brazilian fruit tree know as "cagaiteira" (Eugenia dysenterica), a positive (r = 0.289) and significant (p < 0.05) correlation between the isozymic and morphological matrices of genetic distances.

Conclusions
Both isozymatic and morphological markers revealed the existence of large genetic diversity among the landraces cultivated by traditional farmers in the Vale do Ribeira, with an even larger diversity for the isozymes.Moreover, both markers separated the commercial varieties, which showed less genetic diversity when compared to the landraces.Most of the diversity in D. alata, which was not spatially structured, was found to be within households of traditional farmers in the Vale do Ribeira.This is due to the management practices and material exchange among farmers, thus emphasizing the important role they have played in the in situ/on farm conservation of D. alata.

Figure 2 -
Figure 2 -Cluster analysis using the UPGMA method for the 35 Dioscorea alata accessions based on 71 isozymic bands, according to the similarity index of Jaccard.

Figure 4 -
Figure 4 -Cluster analysis using the UPGMA method for the 35 Dioscorea alata accessions based on 24 morphological characters, according to the similarity index of Jaccard.

Table 1 -
List of Dioscorea alata accessions evaluated, with their identification, origin, common names and geographical coordinates.

Table 2 -
Descriptors used for morphological assessment of Dioscorea alata accessions.

Table 3 -
Analysis of molecular variance (AMOVA) for allozyme and morphological data assessed on 16 D. alata landraces from the Vale do Ribeira.