Introduction

Traditional agriculture maintains high levels of species and genetic diversity of cultivated plants (Altieri et al. 1987; Beckerman 1983; Bellon and Brush 1994; Bellon 1996; Lambert 1996; Brush 2000; Zaldivar et al. 2002, 2004). Home gardens are a form of traditional agriculture widely distributed throughout the world (Hodel and Gessler 1999; Lamont et al. 1999). Typically, home gardens are small plots where a variety of species are grown, mostly for consumption by the family unit (Watson and Eyzaguirre 2001). These gardens are rich in species diversity, as many crop species are present and often coexist with their wild relatives (Rico Gray et al. 1990; Vargas 1990; Padoch and de Jong 1991; Lamont et al. 1999; Watson and Eyzaguirre 2001).

The importance of traditional agroecosystems for the maintenance of genetic diversity in cassava has been examined by Boster (1985), Elias et al. (2000a, 2000b, 2001a, 2001b, 2004), Salick et al. (1997), and Sambatti et al. (2001). These authors reported high levels of morphological variation in cassava grown by the Aguaruna Jivaro from Peru (Boster 1985), the Amuesha from Peru (Salick et al. 1997), the Makushi from Guyana (Elias et al. 2000, 2001, 2004), traditional farmers from the Atlantic Coast in Brazil (Sambatti et al. 2001), and Tukano Amerindians and traditional farmers from the Amazon region in Brazil (Elias et al. 2004). They also found that traditional farming practices maintain genetic variation. Moreover, some individuals not only preserve their original genotypes, but also create and incorporate new genotypes into their gardens.

Cassava (Manihot esculenta Crantz) is a root crop native to the Neotropics (Rogers and Fleming 1977; Hershey 1994). Cassava is of great economic and dietary importance in rural areas. In most parts of the tropics, it is produced mainly by small farmers cultivating marginal soils (Cock 1985). It is grown for its starchy tuberous roots, and is a main staple food for many Amerindians and peasants in the Neotropics (Best and Henry 1994).

Using multiple genetic markers, it has been shown that cassava has high levels of genetic diversity (Hussain et al. 1987; Ramirez et al. 1987; Fregene et al. 1994; Hershey 1994; Roa et al. 1997; Chavarriaga-Aguirre et al. 1998, 1999). Most of these studies have focused on accessions of South American origin. In general, there is very little known about the genetic diversity of Mesoamerican cassava (Chavarriaga-Aguirre et al. 1999), particularly in Central America. Zaldivar et al. (2004) used isozyme markers to study genetic diversity of cassava cultivars being grown in home gardens of Chibchan Amerindians from Costa Rica. Their data revealed high levels of genetic variation.

In the present paper, we describe the level of microsatellite variation of cassava grown in home gardens at two Chibchan Amerindian reserves in Costa Rica. In addition, we analyze the distribution of genetic variation between and within Amerindian reserves. We also analyzed a sample of the two most widely grown commercial varieties in Costa Rica for comparative purposes.

Methods

Study populations

Cassava is a monoecious perennial shrub, well adapted to drought, poor soils and high temperatures (Fregene et al. 1994). Cultivated cassava is normally vegetatively propagated by stem cuttings, and traditionally only some genotypes are maintained as clones. All Manihot species, including M. esculenta, are believed to be predominantly outcrossing species, and experience severe inbreeding depression when they are self-pollinated.

Gulick et al. (1983) defined three main areas of diversity for cassava (M. esculenta). Two of these regions are located in South America. The first one includes large areas in Brazil and Paraguay. The second one is found in Venezuela, Colombia, and northern Brazil. The third one is centered in Nicaragua and extends south to Costa Rica and Panama and north into Honduras. This last region is considered a priority area for collection of cassava and has not been studied in detail (Gulick et al. 1983).

This study was conducted at two Chibchan Amerindian reserves in Costa Rica. We collected samples at the Talamanca Reserve, inhabited by Bribris and Cabecares, and the Coto Brus Reserve, inhabited by Guaymis. Bribris and Cabecares are the largest Amerindian groups in the country (Guevara and Chacon 1992; Barrantes 1993). Guaymis migrated from their ancestral territories in Panama about 60 years ago and have settled in various small Reserves, the largest ones being Conte Burica and Coto Brus (Guevara and Chacon 1992; Barrantes 1993; Camacho 1996). Bribris, Cabecares, and Guaymis grow cassava in home gardens around their households (Zaldivar et al. 2002) and it is an important staple crop (Young 1971; Gordon 1983; Vargas 1990; Camacho 1996; Borge and Castillo 1997).

The Talamanca Reserve is located on the Atlantic side of the Talamanca highlands. This reserve has an area of about 54,000 ha (Borge and Castillo 1997). Population size within the reserve is about 8,000 individuals, 80% of whom are Bribris and 20% are Cabecares (Borge and Castillo 1997). Mean annual precipitation is 4,000 mm and mean annual temperature is 25.6°C (Borge and Castillo 1997). Vegetation types include lowland wet forests and lower montane forests. In this study we only visited settlements located in the Lower Talamanca region, near the center of the reserve. The extent of this area is about 9,600 ha. Altitude ranges from sea level to about 300 m. The following settlements were visited: Barrio Escalante, Sepecue, Mojoncito, Coroma, Bajo Cohen, and Uruchico.

The Coto Brus Reserve is located in the Southern Pacific Region (Camacho 1996). This Reserve has an area of 7,500 ha and the population size is about 826 individuals (Camacho 1996). Mean annual rainfall is 3,900 mm and mean temperature is about 25°C (Camacho 1996). Altitude ranges between 700 and 1,700 m. Vegetation types include lowland forests and lower montane rain forests. The settlements visited within the Reserve were the following: Limoncito or La Casona, Brus Mali, Villa Palacios, Betania, and Caño Bravo.

Two commercial varieties grown in Costa Rica were also considered in this study. One commercial variety (Valencia) accounts for more than 95% of the total production of cassava in the country and it is mainly grown for export (W. Quiros, Oficina Nacional de Semillas, Costa Rica, personal communication). The other variety, known as Manyí, is primarily grown for local consumption. Both are introduced varieties. Extensive cassava plantations are found in San Carlos, Alajuela, Costa Rica. Elsewhere in the country, there is little commercial production of cassava.

Field study and collection

Plant material for these analyses was collected during a series of short visits to each Reserve. The visits took place during the dry season (January–April) of 1994, 1995, and 1997. The average length of each visit was about 10 days for a total of 35 field days in each of the two Reserves. A total of 138 home gardens were visited for this study, 55 at Coto Brus and 83 at Talamanca. Cassava was found in 33 and 32 gardens at Coto Brus and Talamanca, respectively. Zaldivar et al. (2002, 2004) provide a detailed description of the methodology used in the field study.

With the consent of the owners of the gardens, we collected stem cuttings from cassava plants growing in them. If we found more than one morphologically different cassava plant in one garden we collected them as separate accessions. Table 1 shows the number of morphologically distinct samples/clones collected in each site. Cuttings were brought to the Universidad de Costa Rica main campus in San Jose where they were rooted in a greenhouse. Once the plants were growing well, we transplanted them to an agricultural field station. Some of the plants collected during the1994 and 1995 field seasons were stolen or died, and therefore were not available for genetic analysis.

Table 1 Number of accessions analyzed from each location and number of households visited in each Reserve
Full size table

Samples of two commercial varieties, Valencia and Manyí, were obtained directly from EARTH University, at Pococí. A total of 13 accessions of Manyí and 19 accessions of Valencia were analyzed from this sample. In addition, we obtained 16 samples from the Valencia variety from commercial plantations in Pital, San Carlos. However, due to mortality only four of the latter were available for analysis. All this material was also transplanted and maintained in an agricultural field station.

DNA extraction

Total genomic DNA was extracted from leaves of M. esculenta using a modification of the protocol described by Lohdi et al. (1994) and Karp et al. (1997). Approximately 250 mg of leaf tissue was ground in liquid nitrogen with a pestle in a 1.5 ml Eppendorf tube. The ground tissue was mixed with 600 μl of 0.1 M Tris–HCl pH 8.0, 1.0 M NaCl, 2% CTAB, 2% PVP, and 1% β mercaptoethanol extraction buffer. The mix was incubated at 60°C for 40 min with agitation every 10 min. The samples were then centrifuged for 10 min at 12,000  ×  g at 10°C. The supernatant solution was placed in another tube, and one volume of chloroform-octanol (24:1) was added, mixed softly, and centrifuged for 10 min. The supernatant was transferred to another tube where 1/10 volume of 3 M NaAC pH 5.2 and 0.6 volume of ice-cold isopropanol was added. The mix was kept at 5°C for 10–20 h. DNA was precipitated with 70% ethanol, air dried, and later resuspended in 100 μl of 1.0 M TE buffer.

Genetic analysis

We studied the levels of genetic variation using 13 microsatellite marker loci developed by Chavarriaga et al. (1998). The sequences of these primers were obtained from Chavarriaga et al. (1998). We used the following loci: GA12, GA13, GA16, GA21, GA57, GA126, GA127, GA131, GA134, GA136, GA140, GA161, and GAGG5. The PCR reactions were performed using a Rapidcycler (Idaho Technology) in a total volume of 15 μl containing 1.5 μl of 10 × PCR buffer (50 μM Tris–HCl pH 8.3, 2,5 mg/ml of BSA and 10 μM MgCl2), 0.2 μl (0.3 units) Taq polymerase (Promega), 1.2 μl 0.32 mM dNTP, 20 ng of template DNA, and 3.0 μl DNA free water. We also added 1.6 μl 0.4 μ M solution of forward and reverse primers. PCR reaction conditions were the following: 94°C for 10 s, followed by 30 cycles of 15 s at 94°C, 40 s at 54–56°C, and 1 min at 72°C. A final extension step of 4 min at 72°C was added after the last cycle. PCR products were visualized on 40 cm polyacrylamide gels using silver staining (Promega). The different alleles for each locus were identified according to their molecular weight, where the most anodal allele was arbitrarily identified as "A", with the remaining alleles identified sequentially. The size of each allele was determined using regression of migration distance against fragment size of a 50bp molecular ladder (Fermentas).

Data analysis

Genetic diversity in each territory was quantified by means of the number of alleles per locus (A), the effective number of alleles per locus (AE), observed heterozygosity (HO), and Nei’s expected heterozygosity (HE) (Nei 1973), for each locus and averaged over all loci. The significance of allele frequency differences among locations was assessed using Fisher exact test (Weir 1996). The effective number of alleles was estimated as the reciprocal of the homozygosity (Hartl and Clark 1989). In addition, genetic differentiation was determined using the infinite allele model Fst (Weir and Cockerham 1984). These analyses were conducted using the program POPGENE 1.31 (Yeh et al. 1999).

Levels of genetic differentiation were estimated using the program Tools for Population Genetic Analyses (TFPGA) and a hierarchical design of the different locations within each Amerindian reserve and the commercial varieties. Theta S and Theta P were determined for the locations within reserves and between the main three groups considered here; namely, Coto Brus, Talamanca and commercial varieties. Theta S is an indicator of differentiation among locations within groups and Theta P indicates differentiation among groups (Weir and Cockerham 1984). The program uses bootstrap sampling to generate 95% confidence intervals using 1000 replications. The degree of relatedness between regions, based on Nei’s genetic distances, was represented in a tree using UPGMA. Bootstrap sampling of loci tested the robustness of clusters of the tree, using 103 simulations.

We also determined the number of distinct multilocus genotypes (G) found among all accessions considered in this study. We recorded the frequency of each multilocus genotype observed in each Amerindian reserve and in each of the cultivated varieties. We used the Shannon index to determine the diversity of multilocus genotypes observed in both Amerindian reserves and the cultivated varieties. Shannon diversity index was calculated using the following equation:

$$ H=-\sum_{i=1}^{S}P_{i}\hbox{ln}(P_{i}) $$

where p i is the relative abundance of each genotype among the total number of accessions genotyped for each site. For a fixed number of multilocus genotypes (G), the maximum Shannon diversity (H max ) is obtained when all genotypes are equally abundant. Comparisons of H with H max can be used to determine how evenly distributed the genotypes are. Lastly, we estimated Shannon Evenness using the equation:

$$ \hbox{J}=\hbox{H}/H_{max} $$

A larger Shannon index does not necessarily mean that both richness and evenness are high. Larger values for Shannon index may result from a substantially greater richness or greater evenness.

Results

A total of 13 microsatellite loci were examined in this study. Sample sizes used to determine allele diversity for the different loci ranged from 27 to 57 plants for the Coto Brus reserve, from 33 to 38 for the Talamanca reserve, and from 27 to 34 for the commercial varieties. Only one locus (GA13) was monomorphic in both Amerindian reserves. In addition, the same locus, along with three other loci (GA12, GA21, and GA136), were monomorphic in the commercial varieties commonly cultivated in this part of Costa Rica (Table 2). Thus, the proportion of polymorphic loci found in each of the Amerindian reserves was higher than that found in the two most widely grown commercial varieties in Costa Rica (P = 92.3 and P = 69.2, respectively).

Table 2 Allele frequency for the 13 loci examined for each population considered in this study
Full size table

Our data revealed differences in allele frequencies between the two reserves; however, the most frequent allele was the same at most loci. Differences were, in general, due to alleles occurring at low frequency in both areas. There was also variation in the number of alleles present at each locus (Table 3). For example, loci GA12, GA126 and GA131 showed an additional allele at Coto Brus, where sample sizes were larger. At locus GA127, the number of alleles present was the same, but their identity was not the same, allele A being only found at Talamanca and allele B only at Coto Brus (commercial varieties also show allele A). Overall, the mean number of alleles per locus and the mean number of effective alleles was similar for the two reserves. The number of alleles per locus found in the commercial varieties was lower than that found in the two reserves (ne = 1.7854, 1.7234 and 1.5572, for the Coto Brus, Talamanca, and commercial cassava, respectively).

Table 3 Summary of genetic variation statistics for all loci for each population considered in this study
Full size table

Genetic diversity was similar among the three groups (Table 4). Common indicators of genetic diversity such as mean expected heterozygosity and Nei’s expected heterozygosity were similar for both Amerindian reserves. These values were slightly lower for commercial varieties. In addition, observed heterozygosity was higher than expected for all groups. The high estimates of observed heterozygosity in commercial varieties were due to a large excess of heterozygotes in loci GA126, GA127, and GA131. This finding was supported by the mean values of Fis and Fit. In this case, both are negative, indicating deviation from Hardy-Weinberg proportions due to excess heterozygotes (Table 5). Hierarchical estimates of genetic differentiation indicate that genetic differentiation between locations within each group was larger than that between groups (Theta S = 0.0775 (S.D. = 0.0195) and Theta P = 0.0204 (S.D. = 0.0084)).

Table 4 Summary of heterozygosity statistics for all loci for each population considered in this study
Full size table
Table 5 Summary of F-Statistics and gene flow for all loci
Full size table

Genetic distance between groups, calculated using Nei’s unbiased estimator of genetic distance, ranged from 0.0167 to 0.0343 (Nei 1978) (Table 6). Cassava commercial varieties and cassava from Coto Brus showed the greatest genetic distance. In addition, values for Wright Fst revealed that most of the genetic variation was found within groups and that there was little genetic differentiation between groups (Fst = 0.0320) (Table 5). Our data also revealed high rates of gene flow among the three groups. When we only considered the two Amerindian reserves, genetic differentiation was lower (Fst = 0.0154) and our estimate of gene flow was higher (Nm = 15.97) than when we included the commercial varieties in the analysis.

Table 6 Nei’s genetic identity (above diagonal) and genetic distance (below diagonal) among the groups examined
Full size table

A more detailed analysis of genetic distances between different sampling locations revealed that the variety Manyi was the most distinct (Table 6 and Fig. 1). Mean genetic distance across all locations was 0.1534 (range 0.1197–0.2269). Our data also revealed that the genetic distance between the Valencia variety and samples collected from different locations within Amerindian reserves was small. Mean genetic distance was 0.0316 (range 0.0111–0.0590). The Valencia variety was very similar to the samples collected at Bajo Cohen, and Sepecue from Talamanca, and to Caño Bravo from Coto Brus. In addition, the multilocus genotypes found in this variety were also found in the Amerindian reserves, particularly at Talamanca. Genetic distances between samples from different locations within Amerindian reserves were generally low (mean Nei’s genetic distance = 0.0305), ranging from 0.0185 (Caño Bravo-Villa Palacios) to 0.1094 (Uruchico-Betania). We found a total of 52 distinct multilocus genotypes among all cassava accessions considered in this study. About 28 multilocus genotypes were found in Coto Brus and 19 in Talamanca (Table 7). Only one mutilocus genotype was common to both Amerindian reserves (Fig. 2). In addition, a total of seven multilocus genotypes were found among the accessions of the two commercial varieties that we studied; four of them among the accessions of the variety Manyí and three among the accessions of the variety Valencia (Table 7). Only one multilocus genotype was found in both the variety Valencia and in seven accessions from Talamanca.

Fig. 1
figure 1

Location of the Amerindian reserves sampled in this study

Full size image
Fig. 2
figure 2

UPGMA tree based on Nei’s genetic distances (Nei 1978) estimated from 12 polymorphic loci of the sampling locations of cassava considered in this study

Full size image
Table 7 Number of accessions of cassava assigned to a multilocus genotype, number of multilocus genotypes determined, Shannon diversity index (H), maximum Shannon diversity (Hmax), and evenness for each Amerindian reserve and commercial variety
Full size table

There was variation in the frequency of each multilocus genotype in each Amerindian Reserve. For example, two genotypes were over-represented among the 61 accessions for which we determine their multilocus genotype in Coto Brus (present in 11 and 13 accessions, respectively). The frequency of the other multilocus genotypes in the reserve ranged between 1 and 3 accessions, with most genotypes (17) observed in only one accession. Similarly, one multilocus genotype was observed in 12 of the 38 accessions in Talamanca. The frequency of the other multilocus genotypes in the reserve ranged between 1 and 4 accessions, with 13 genotypes observed in only one accession. Shannon diversity index revealed higher diversity in Coto Brus, followed by Talamanca and the commercial varieties. However, our data also showed that the reduction in diversity due to over-representation of the dominant multilocus genotypes was similar for both Amerindian reserves (J = 0.85 in both reserves). Both cultivated varieties showed lower Shannon diversity scores (H) and lower evenness (J) (Table 7).

Discussion

We found high levels of microsatellite variation in cassava accessions collected in home gardens of Chibchan Amerindians from Costa Rica. Our data revealed that 12 of the 13 loci examined were polymorphic, and we found a total of 37 alleles in the two Amerindian reserves. Most loci had three or four alleles, and we observed five alleles at one locus. In contrast, only nine loci were polymorphic within the two commercial varieties examined, and we only found a total of 23 alleles. Moreover, only one locus showed more than two alleles in the commercial varieties.

Chavarriaga et al. (1998) showed that the 13 microsatellite loci that we used were polymorphic in a segregating F1 population of 83 individuals. Some of these loci were also tested for a wide sample of cassava from the core collection of CIAT (from 28 to 522 accessions sampled per locus). Overall, Chavarriaga et al. (1998) found high levels of heterozygosity and allelic diversity in their sample. In this study, we found that the observed heterozygosity values for most loci were similar to those obtained by Chavarriaga et al. (1998). For loci GA12, GA 21, and GA126 we found larger allele sizes than those reported by Chavarriaga et al. (1998).

Elias et al. (2004) used 10 of these same markers to characterize genetic variation in sweet and bitter cassava cultivated by the Makushi in central Guyana, and several populations of local varieties from Brazil, as well as a sample from the core collection at CIAT. Overall, our results are comparable to their findings for the sweet varieties, except for locus GA140, where they found greater allele variation. Elias et al. (2004) also found more allele diversity for the bitter varieties and the sample from the core collection at CIAT. The bitter varieties showed greater allele diversity for loci GA21, GA131, GA134, and GA140 than those reported here. The sample from the core collection also had more alleles for loci GA16, GA21, and GA140.

We also found higher levels of heterozygosity than expected for all groups of accessions. This finding is supported by the high negative values of the Fit coefficients, indicating an excess of heterozygotes in relation to expectations under Hardy –Weinberg equilibrium. This finding coincided with the high levels of heterozygosity for sweet cassava reported by Elias et al. (2004). In addition we found a high number of multilocus genotypes, which supports the notion of polyclonality among traditional landraces as proposed by Second et al. (1997), Colombo et al. (1998), Elias et al. (2001, 2004), and Sambatti et al. (2001).

Previous studies using isozyme markers also found high levels of genetic variation for cassava (Hussain et al. 1987; Ramirez et al. 1987; Fregene et al. 1994; Hershey 1994; Chavarriaga et al. 1999; Resende et al. 2000). In a previous study conducted using these same populations, we reported high levels of isozyme variation (Zaldivar et al. 2004). Our data revealed that six of the nine loci examined were polymorphic and the loci found to be polymorphic were the same reported as such by other authors (Ramirez et al. 1987; Fregene et al. 1994).

In our previous work using isozyme markers (Zaldivar et al. 2004), we reported that most of the variation was found within populations. However, we observed significant differentiation between populations (Gst = 0.11). Our findings using microsatellite markers indicate lower levels of genetic differentiation among populations (Fst = 0.03) than those reported using isozyme markers. Genetic differences among populations were primarily due to alleles that are present in relatively low frequencies (Table 2). Therefore, their contribution to genetic variation is small, and their importance in our estimate of genetic differentiation may be under-represented (see Hedrick 1999 for further discussion). Elias et al. (2001) studied the genetic structure of cassava grown in traditional farming systems in Guyana using eight of the same microsatellite marker loci. They reported moderate levels of genetic differentiation among varieties grown in the area (Fst = 0.363), but low genetic differentiation between the populations of seedlings resulting from unmanaged sexual reproduction (Fst = 0.013). In another study including traditional landraces from Guyana and Brazil, they found low differentiation between populations (Fst = 0.0744). But, as Hedrick (1999) points out, the meaning of these estimates should be interpreted with caution due to the high levels of heterozygosity observed in cassava. Moreover, the correct interpretation of F-statistics is even more difficult because cassava is primarily clonally propagated.

Our results indicate that traditional home gardens maintain similar levels of genetic diversity in both Amerindian Reserves, as they shared most alleles, and in many cases, their frequencies were similar. These findings suggest a similar origin or the occurrence of gene exchange between reserves. Guaymi ancestral territories in Bocas del Toro province, Panama (Gordon 1983), are adjacent to the Bribri and Cabecares territories in Costa Rica. This may have facilitated gene exchange in the past. Our findings also indicated that, with the only exceptions of Caño Bravo (Coto Brus) and Uruchico (Talamanca), accessions collected from locations within the same Reserve tend to cluster together. However, genetic distances between accessions collected within the Amerindian Reserves were small, ranging between 0.0081 and 0.1094. Overall, the geographically isolated accessions from Uruchico showed the highest genetic distance with accessions from other Amerindian locations.

Boster (1983) reported that the coexistence of different varieties of a given crop in the same field or in neighboring fields is common in traditional farming systems. Boster (1985) found very high levels of morphological diversity in cassava grown by the Aguaruna Jivaro from Peru, and suggested that the Aguaruna select cultivars based on perceptual distinctiveness, thereby promoting morphological variation in their cultivars. Elias (2000) and Sambatti et al. (2001) proposed that sexual reproduction and recombination may amplify genetic diversity in cassava, and that Amerindians from Guyana and farmers from Brazil tend to incorporate plants with novel genotypes into their collections. Salick et al. (1997) also found high levels of morphological variation in cassava grown by the Amuesha from Peru. Moreover, they found that some farmers within the population purposely breed different varieties and create variable crops (Salick et al. 1997). Elias et al. (2000, 2001) argued that recombination and gene flow play a major role in the dynamics of genetic diversity in traditional farming systems. They reported the occurrence of natural intervarietal crosses in cassava. They also argued that volunteer plants resulting from unmanaged sexual reproduction in traditional farming systems are often incorporated by farmers into the germplasm used for vegetative propagation. They proposed that the incorporation of such seedlings is a frequent event, and that it may explain the excess of heterozygous plants that they observed in the populations considered in their study.

Pujol et al. (2005, 2006) demonstrated how Amerindian agricultural practices favor heterozygosity in cassava. First, Pujol and collaborators demonstrated that, even when Amerindian farmers primarily propagate cassava clonally, the selective retention of seedlings resulting from sexual reproduction was inadvertently favoring heterozygosity. Palikur Amerindians in French Guiana tend to retain large seedlings when weeding their fields. Farmers allowed these seedlings to grow and later used them for cuttings. Because plant size and heterozygosity are correlated, this practice allows maintenance of genotypically diverse and heterozygous stocks. We lack the data to demonstrate that Amerindians in Costa Rica inadvertently select for heterozygous individuals, but we often found fruits, seeds, and seedlings in the homegardens that we visited. Moreover, it was also common to find multiple multilocus genotypes growing in the same household (at least 8 gardens in Coto Brus and 7 in Talamanca).

In general, all measurements of genetic diversity indicate high levels of variation in the accessions collected in both Amerindian reserves. The proportions of polymorphic loci, and the number of alleles present in the reserves, are comparable to those found in Amazonian landraces (Elias et al. 2004). Most of the alleles found in these accessions appear to be different from those reported by Elias et al. (2004). It is difficult to establish equivalence or difference of size of microsatellite alleles between studies, but only three loci (GA16, GA131, and GA134) show similar size for the most frequent allele. However, allele diversity was similar in the two studies. In addition, it is important to point out that the overall variance in allele size for the same markers in Costa Rica and South America was also similar. The high levels of heterozygosity that we found suggest that recombination and gene flow may play an important role in the preservation of genetic diversity. Our findings support the notion that the areas of primary diversity of cassava centered in Nicaragua, and extending south to Costa Rica and Panama and north into Honduras (as proposed by Gulick et al. 1983), should be considered a priority area for collection and conservation.

Our data also reveal that there is diversity in the frequency and abundance of multilocus genotypes between the Amerindian reserves and the commercial varieties considered in this study (Table 7). Overall, Coto Brus showed a higher number of multilocus genotypes (G) than Talamanca and the commercial varieties. However, given the number of genotypes observed in each reserve, Coto Brus and Talamanca showed similar evenness scores. Meanwhile, the number of multilocus genotypes and the evenness scores were lower for both commercial varieties in comparison to those of the Amerindian reserves. Moreover, there is little overlap in the observed multilocus genotypes. These findings indicate that these reserves and their homegardens play an important role in the conservation of genetic diversity of cassava, and most likely other crop plants.

In summary, these findings strongly support the notion that traditional home gardens are a useful tool for in situ conservation of plant genetic resources. Genetic exchange between unrelated clones or varieties may facilitate recombination. This scenario may promote the appearance of novel gene combinations, which are likely to be incorporated by Amerindians into the germplasm of their cassava populations. However, the modernization of farming practices, in response to economic and other pressures, suggest that home gardens may be fragile, and thus a threatened conservation unit.