Temporal and spatial genetic structure in Vitellaria paradoxa (shea tree) in an agroforestry system in southern Mali

Molecular Ecology (2004) 13, 1231–1240

doi: 10.1111/j.1365-294X.2004.02144.x

Temporal and spatial genetic structure in Vitellaria paradoxa (shea tree) in an agroforestry system in southern Mali

Blackwell Publishing, Ltd.

B O K A R Y A L L A Y E K E L L Y ,* O L I V I E R J . H A R D Y † and J E A N - M A R C B O U V E T ‡ *Institut d’Economie Rurale, Programme Ressources Forestières Centre Régional de la Recherche Agronomique de Sikasso, Sikasso, Mali, †Laboratoire de Génétique Evolutive et d’Ecologie vegétale, Université Libre de Bruxelles, Bruxelles, Belgium, ‡Cirad–Forêt, Campus International de Baillarguet, TA 10C, BP 5035, 34398 Montpellier, cédex, France

Abstract Ten microsatellite loci were used to investigate the impact of human activity on the spatial and temporal genetic structure of Vitellaria paradoxa (Sapotaceae), a parkland tree species in agroforestry systems in southern Mali. Two stands (forest and fallow) and three cohorts (adults, juveniles and natural regeneration) in each stand were studied to: (i) compare their levels of genetic diversity (gene diversity, HE; allelic richness, Rs; and inbreeding, FIS); (ii) assess their genetic differentiation (FST); and (iii) compare their levels of spatial genetic structuring. Gene diversity parameters did not vary substantially among stands or cohorts, and tests for bottleneck events were nonsignificant. The inbreeding coefficients were not significantly different from zero in most cases (FIS = − 0.025 in forest and 0.045 in fallow), suggesting that the species is probably outbreeding. There was a weak decrease in FIS with age, suggesting inbreeding depression. Differentiation of stands within each cohort was weak (FST = 0.026, 0.0005, 0.010 for adults, juveniles and regeneration, respectively), suggesting extensive gene flow. Cohorts within each stand were little differentiated (FST = − 0.001 and 0.001 in forest and fallow, respectively). The spatial genetic structure was more pronounced in fallow than in forest where adults showed no spatial structuring. In conclusion, despite the huge influence of human activity on the life cycle of Vitellaria paradoxa growing in parkland systems, the impact on the pattern of genetic variation at microsatellite loci appears rather limited, possibly due to the buffering effect of extensive gene flow between unmanaged and managed populations. Keywords: forest tree, gene flow, human impact, inbreeding coefficient, kinship coefficient, microsatellite Received 15 September 2003; revision received 12 January 2004; accepted 12 January 2004

Introduction Temporal and spatial genetic structure are influenced by biological aspects of a tree species. Many studies have revealed that factors such as limited seed and pollen dispersal, density, fragmentation, colonization history, isolation into small patches, differential mortality and micro-environmental selection have an impact on the spatial genetic structure of many forest tree species (Levin & Kerster 1974; Epperson 1993; Austerlitz et al. 2000; Petit 2001). Of these factors, the pattern of gene dispersal is most Correspondence: J.M. Bouvet. Fax: 33 4 67 59 37 33; E-mail: [email protected] © 2004 Blackwell Publishing Ltd

significant (Schaal 1980; Hamrick et al. 1993; Ennos 1994). For instance, genetic similarity among progeny spatially clustered near their maternal parent because of limited seed dispersal generates a substantial spatially localized genetic structure (Perry & Knowles 1991; Schnabel et al. 1991; Berg & Hamrick 1994; Geburek & Tripp-Knowles 1994; Chung & Epperson 2000a,b). In contrast, if seeds are widely and independently dispersed, only a weak spatial genetic structure will result (Dewey & Heywood 1988; Loiselle et al. 1995; Chung et al. 1999, 2000). Another factor influencing temporal and spatial genetic structure in tree species may be human activity because this leads to many changes in forest community and ecosystem processes. Several studies have assessed the impact

1232 B . A . K E L L Y , O . H A R D Y and J . - M . B O U V E T of human practices on the genetic diversity and genetic structure of forest tree species (Knowles et al. 1992; Young & Merriam 1994; Aldrich et al. 1998), but very few have focused on tree species growing in parkland systems in sub-Saharan Africa (Hossaert-Mckey et al. 1996; Leonardi et al. 1996; Epperson & Alvarez-Buylla 1997; Kitamura et al. 1997a,b). Measuring and understanding the effects of human practices on the genetic structure of tree populations in parkland systems is essential for management strategies in agroforestry ecosystems. Parkland, defined as a regular, systematic and ordered presence of trees within fields (Sautter 1968, cited by Bagnoud et al. 1995), is the result of a long evolutionary process during which an association between natural elements (trees and shrubs conserved, maintained and enhanced because of their utility) and crops happens within a regularly exploited space. Parkland includes fields in which different types of crops (cereals, cotton, peanuts, sorghum) are produced over several years, and fallow, i.e. when cultivation activity is stopped on parcels of land in order to restore soil fertility. The length of the fallow period (3–4 to 25–30 years) is unique to each farmer, depending upon the land he possesses, the needs of his household and the way he manages his land. On these two elements (field and fallow), there are one or two dominant tree species, as is the case for Vitellaria paradoxa parkland. The latter are the most represented in the soudanian zone, occupying 4.7 × 106 ha in Mali (Boffa 1998), and are classified as ‘oil parkland’. They are characterized by the dominance of V. paradoxa (Sapotaceae) and reflect the presence of stable human populations. V. paradoxa is conserved and maintained in the field by farmers because of its economical importance. Its fruit pulp is consumed by humans and animals, and butter is extracted from the seed kernel for local consumption (oil for cooking and ointment in traditional medicine) and for trade on local and international markets. The species faces a high degree of thinning, selection and natural mortality, leading to a noticeable reduction in density. Because the field/fallow cycle may contrast with natural forest, especially for the long-term evolutionary dynamics of the tree species, the impact of human practices on the genetic diversity and genetic structure of V. paradoxa raises many questions. Genetic markers and spatial statistics are efficient tools to assess the impact of evolutionary forces on plant genetic structure (Loiselle et al. 1995; Doligez & Joly 1997; Chung et al. 2000; Isagi et al. 2000). It is, therefore, appealing to apply these methods to study the impact of human practices on the genetic structure of V. paradoxa. Our main objectives were to characterize the level and organization of genetic diversity within V. paradoxa populations, and to assess the impact of human practices on the spatial and temporal genetic structure. We attempted to answer the following questions:

1 What is the impact of farming practices on the level of genetic diversity? 2 Does human activity lead to genetic differentiation among cohorts? 3 How has human activity influenced the spatial pattern of genetic diversity? To this end, we analysed the pattern of variation in microsatellite markers in three cohorts, comparing a population located in forest, which should be representative of natural conditions, with a population subject to the field/fallow cycle in the parkland system.

Materials and methods Model species The main parkland tree species in soudanian zones of subSaharan Africa, Vitellaria paradoxa reaches 20 m in height and 1 m in diameter at chest height. Recognized as a longliving savannah tree species (> 200 years), its natural range extends from the eastern part of Senegal to the high plateau of Uganda forming an almost unbroken belt 5000 km long and 500 km wide. Its population density varies greatly according to land use, locality and ecological conditions. It mostly reproduces sexually and is mainly insect pollinated, flowering and fruiting from December to May with some geographical variations. It fruits more in fields and young fallow than in forest because of reduced competition, ploughing and crop fertilization, and shows intense regeneration in fallow. Harvesting of fruit occurs mainly between July and September The species is barochore but its seeds can also be dispersed by animals such as birds, monkeys and rodents, and also by humans.

Site and experimental design The study was carried out in Sikasso, southern Mali, on territory belonging to Mperesso village. This village is located in the north of the region (12°16′ N, 5°19′ W) where the climate is south soudanian. During the decade 1982– 92 mean annual rainfall was 680 mm with a minimum of 435 mm and a maximum of 950 mm. Two parcels of land, 2 and 7 ha, separated by 1.2 km, were selected on the basis of two criteria: land use (fallow and forest stands) and a relatively high density of adult trees. Each stand is part of two different cycles. In the first, there is a succession of cultivation (field), regeneration and growth (fallow), and a reduction in tree density by thinning in preparation for re-use as a field. This cycle is shown in Fig. 1(a) by the distribution of tree circumference at chest height (CBH) which exhibits two modes: the first (15 cm) reflecting the distribution of regeneration since the field was left as fallow, and the second (95–105 cm) that of © 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 1231–1240

W E A K I M P A C T O F P A R K L A N D I N S H E A T R E E D I V E R S I T Y 1233

DNA extraction and microsatellite-enriched library development

Fig. 1 Distribution of the tree circumference at chest height in the (a) fallow and (b) forest stands.

adult trees already present during the field stage, the lack of intermediate sizes reflects the low regeneration potential during field stages. In forest, the cycle corresponds to that for the classical dry forest with weak but constant regeneration and slow growth, as well as selective tree cutting for various purposes (mainly for local craft). This cycle results in fewer individuals in the first size classes compared with fallow and the absence of individuals in size classes > 125 cm (Fig. 1b). In each stand, individuals with a CBH > 10 cm were marked, mapped and their CBH measured. Populations from the two stands were divided into three cohorts according to flowering and fruiting ability, which correlated well with tree size (Kelly, personal communication). Therefore, we recognized: (i) adults (CBH > 30 cm), (ii) juveniles (CBH 10–30 cm); and (iii) natural regeneration (from height > 1 cm to CBH < 10 cm; which include seedlings and saplings). The samples were exhaustive for adults and juveniles and the sampled regeneration (not mapped) was collected under marked adult trees. Samples size, mean CBH, density and the area of each parcel are given in Table 1.

Stand

Cohort

Sample size (N)

Fallow

AD (CBH > 30 cm) JUV (10 cm ≥ CBH ≥ 30 cm) NR (CBH < 10 cm) AD (CBH > 30 cm) JUV (10 cm ≥ CBH ≥ 30 cm) NR (CBH > 30 cm)

68 63 90 78 52 90

Forest

Two or three leaves were harvested from each tree and dried in silica gel. DNA was extracted from dried leaves following the Matab method developed for Theobroma cacao (Bousquet et al. 1990). Dried leaf samples (150 mg) were flash-frozen in a tube filled with liquid nitrogen and mixed to a homogenous slurry with 5 mL DNA extraction buffer (100 mm Tris–HCl, pH 8.0, 20 mm EDTA, 1.4 m NaCl, 1% PEG 6000, 2% mixed alkyl trimethyl amonium bromide, 0.5% sodium sulphite). The tube was then incubated at 74 °C for 30 min. Samples were washed with wet chloroform (CIAA, 24:1) to remove cellular debris and protein. Following centrifugation at 9000 g for 15 min, the liquid phase was transferred into 15 mL tubes, and 5 mL of ice-cold isopropanol was added to precipitate the DNA. The resulting DNA pellets were resuspended in 400 µL sterile water and then stored at 4 °C until required. Microsatellite-enriched libraries were achieved following the method developed for SSR markers in tropical crops; the two principles of the technique are hybridization with biotin-labelled oligorobe followed by the capture of selected sequences using streptavidin-coated magnetic beads. The procedure is detailed in (Billotte et al. 1999). After shearing the DNA with an endonuclease RsaI, fragments were ligated to adaptators Rsa21 5′-C TC T TGC T TACGCGTGGACTA-3′ and Rsa25 5′-TAGTCCACGCGTAAGCAAGAGCACA-3′. Fragments of DNA containing (GA) microsatellite were selected and PCR-amplification was performed using primers Rsa21 and purified PCR products were cloned into pGEM-T Easy vector. Cloned fragments were then used to transform competent DH5µ Escherichia coli strain. Plasmids from positive clones containing a microsatellite sequence were extracted and sequencing was performed using the universal T7 and M13 reverse sequencing primers and the dideoxydye terminator method. Primers complementary to flanking regions of the repeats were designed for 96 clones containing microsatellite sequence using oligo v4 software (National Bioscience, Inc). Primers pairs were generated to produce

Mean CBH (cm)

CV of Area CBH (%) (m2)*

Density (N/ha)

78.57 17.79 — 51.66 19.12 —

43.46 28.04 — 42.21 28.50 —

32 30 3305† 11 7 218†

*Parameter is given for the whole stand and not by cohort. †Density for natural regeneration is estimated including seedlings. AD: adults; JUV: juveniles, NR: natural regeneration. © 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 1231–1240

21 000

73 500

Table 1 Mean circumference (CBH), approximate area and estimates of density per stand and cohort

1234 B . A . K E L L Y , O . H A R D Y and J . - M . B O U V E T amplified DNA fragments between 95 and 350 bp in length. Thirteen polymorphic nuclear loci with sizes ranging from 109 to 226 bp were obtained and sampled populations were screened using these loci (B5, E4, E11, E6a, E6b, B3, H4, D10, G7, D6, F1, F5 and E5). This number of microsatellite loci allows a good assessment of population genetic structure.

Data analysis Six loci (E6b, E4, D6, D10, G7 and F1) showed suspicious results regarding FIS values in a primary analysis. E6b, E4, D6 , D10 and G7 gave systematically positive and high values of FIS which could be due to null alleles effect, whereas F1 gave systematically negative and very low values which may result from gel misreading. Congruently, when checking individuals that were undetermined at one or two loci, we found that E6b, G7 and F1 had the highest rate of missing data (42, 24 and 19%, respectively, compared with a mean of 5% for the other loci), suggesting the presence of null alleles at fairly high frequency (resulting in frequent homozygotes for null alleles). Therefore, these three loci were eliminated. E4, D6 and D10 presented a normal rate of missing data, suggesting a frequency of null alleles < 10–20%, and numerous alleles. Hence, it was decided to remove them when calculating multilocus FIS, but to use them to calculate the diversity parameters, the FST and spatial genetic structure parameter in order to keep a good precision in the estimates and sufficient testing power. Whereas a low frequency of null alleles has a significant impact on FIS, it has a weak effect on FST or Fij estimates (see below) (Hardy 2000), so that removing E4, D6 and D10 was not necessary.

Genetic diversity analysis Standard genetic diversity parameters were determined for each cohort (adults, juveniles and regeneration) in forest and fallow: number of alleles per locus (A0), allelic richness (Rs), which measures the number of alleles independent of sample size (Petit et al. 1998), and the expected heterozygosity or gene diversity (HE) using an unbiased estimator (see Nei 1987). Rs is based on the rarefaction index and consists of estimating the expected number of alleles in a subsample of 2n genes, given that 2 N genes have been sampled (N ≥ n). n was fixed as the smallest number of individuals typed for a locus in a sample and Rs was calculated as:   Rs = ∑ 1 − i=1   nu

 2N − Ni    2n     ,  2 N    2n     

where Ni is the number of alleles of type i among the 2 N genes. Comparisons of cohorts within and among stands according to these parameters (A0, R s, H E), were made using the software fstat v2.9.3 (Goudet 2001). The same program was used to test Hardy–Weinberg equilibrium, heterozygote deficit and population differentiation. All these tests are randomization based, i.e. datasets fitting the null hypotheses to be tested are generated by randomising the appropriate units (alleles or genotypes) and statistics are calculated on these randomised datasets to assess their frequency distributions. To test for heterozygote deficit relative to Hardy–Weinberg expectations, alleles were permuted among individuals within samples and FIS was used as a statistic for comparison. To test for population differentiation, genotypes were permuted among samples. To check if there is a signature for recent bottleneck events, we used the software bottleneck (Cornuet & Luikart 1996) which compares gene diversity observed ( HE ) with that expected from the number of alleles per locus (A0) when population size remains constant and for a given mutation model. After a bottleneck, one expects A0 to decrease more than HE, so that the observed HE should be higher than expected HE on basis of A0. As the microsatellite mutation model is thought to be intermediate between a stepwise mutation model and an infinite allele mutation model, we tested bottleneck events assuming each of these models. We used the Wilcoxon sign-rank test as suggested by Cornuet & Luikart (1996).

Genetic structure analysis The genetic structure of populations was assessed by Fstatistics computed using a nested anova (Michalakis & Excoffier 1996). The level of significance of F-statistics was obtained after 10 000 permutations of individuals (FST), or genes (FIS) using spagedi v1.0 (Hardy & Vekemans 2002).

Spatial genetic structure analysis The spatial distribution of alleles in adults and juveniles of each stand (fallow and forest) was assessed using pairwise kinship coefficient Fij (Ritland 1996) between individuals. ∑

Fij =

∑l  

a

∑ci ∑cj(xlcia xlcja/pla )  − 1  ∑ci ∑cj 1  , ∑l(ml − 1)

where xlcia is an indicator variable (xlcia = 1 if the allele on the chromosome c at locus l for individual i is a, otherwise xlcia = 0), pla is the frequency of allele a at locus l in the reference sample (fallow or forest), ml is the number of different © 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 1231–1240

W E A K I M P A C T O F P A R K L A N D I N S H E A T R E E D I V E R S I T Y 1235 alleles found in the sample at locus l, and Σci stands for the sum over homologous chromosomes of individual i. The spatial structure was assessed: (i) by averaging Fij values over pairs of individuals separated by a given distance class in a way equivalent to a spatial autocorrelation analysis; (ii) by regressing Fij values on ln(dij), where dij is the Euclidian distance between i and j. To compare cohorts within each stand, we distinguished the following types of pairwise comparisons: adult–adult, juvenile–juvenile and adult–juvenile. The averaged Fij estimators per distance class and the regression slopes were tested against the null hypothesis of random distribution of genotypes by randomising 1000 times the spatial positions using spagedi v1.0 (Hardy & Vekemans 2002).

Results

Table 2 Allele frequencies per locus and cohort in forest and fallow. The different alleles are expressed in base pairs Forest Locus

Alleles

AD

JUV

NR

AD

JUV

NR

B5

157 161 112 114 116 118 119 120 121 123 125 127 224 228 230 120 122 124 126 151 155 214 216 218 222 218 220 222 224 226 228 114 116 117 118 119 120 124 201 205 209 240 242

0.270 0.730 0.016 0.008 0.065 0.016 0.032 0.169 0.645 0.008 0.008 0.032 0.676 0.324

0.217 0.783

0.283 0.717

0.286 0.714

0.270 0.730

0.176 0.824

0.012 0.071

0.007 0.033 0.020 0.100 0.180 0.600 0.040

0.020

0.019 0.083

0.077

0.065 0.102 0.694 0.028

0.070 0.049 0.775 0.028

E4

E11

Genetic diversity of the markers Loci E4, D10 and D6 were the most polymorphic, displaying 5–10 alleles (Table 2). For each locus, one or two alleles had a higher frequency. In most cases, these dominant alleles were the same for cohorts and stands. The distribution of rare alleles (frequency < 0.01) was different for the two stands. In the forest the highest number of rare alleles (5) was found in adults and no rare allele was found in juveniles, whereas in fallow the reverse was observed (i.e. three rare alleles in juveniles and no rare allele in adults).

E6a

B3 H4

D10

Genetic diversity and inbreeding: comparison between stands Table 3 shows the averaged value over loci of four parameters of genetic diversity within each stand. Considering the adults, the expected heterozygosity (HE) was higher in forest, whereas the inbreeding coefficient (FIS) was higher in fallow. The difference (Table 4) was significant for HE but not for FIS. Over the pooled sample of cohorts, the test was not significant for Rs, HE or FIS (Table 4).

D6

Genetic diversity and inbreeding: comparison among cohorts

E5

The total number of alleles (all loci) was similar for the three cohorts within each stand. In forest, the highest mean number of alleles (A0) and mean allelic richness (Rs) were observed in the regeneration and the highest mean gene diversity (HE) was observed in juveniles. In fallow, the three parameters were higher for juveniles (Table 3). Randomization-based tests (Table 4) showed that, in forest, Rs and HE were not significantly different between cohorts. In fallow, HE was significantly different between adults and juveniles and marginally significant differences were found in the comparisons adult cohort vs. juvenile cohort © 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 1231–1240

Fallow

F5

N

0.008 0.142 0.708 0.142

0.202 0.179 0.488 0.036 0.012 0.625 0.375

1.000 0.346 0.615 0.038

0.156 0.700 0.144 0.011 0.989 0.219 0.703 0.078

0.032 0.016 0.177 0.371 0.226 0.177

0.097 0.153 0.167 0.444 0.083 0.056

0.026 0.092 0.158 0.592 0.092 0.039 0.539 0.008 0.453 1.000

0.015 0.197 0.167 0.545 0.061 0.015 0.679

(78)

0.321 0.974 0.026 (52)

0.020 0.655 0.331 0.014 0.013 0.133 0.713 0.140 0.013 0.987 0.306 0.625 0.069 0.049 0.069 0.389 0.382 0.069 0.042 0.007 0.027 0.062 0.226 0.610 0.048 0.021 0.582 0.021 0.397 0.987 0.013 (90)

0.039 0.167 0.745 0.020 0.010

0.053 0.224 0.211 0.395 0.039 0.079

0.009 0.639 0.333 0.028 0.027 0.282 0.509 0.182 0.025 0.975 0.209 0.745 0.036 0.009 0.028 0.148 0.361 0.315 0.093 0.056

0.015 0.015 0.147 0.721 0.074 0.029 0.676 0.029 0.294 1.000

0.111 0.102 0.157 0.565 0.056 0.009 0.649 0.018 0.333 1.000

0.068 0.096 0.158 0.596 0.082

(68)

(63)

(90)

0.680 0.300 0.020 0.039 0.275 0.618 0.069 0.037 0.963 0.147 0.809 0.044

0.701 0.278 0.021 0.007 0.201 0.634 0.157 0.021 0.979 0.188 0.715 0.097 0.021 0.336 0.192 0.288 0.082 0.082

0.658 0.041 0.301 1.000

N: sample size for adults (AD), Juveniles ( JUV) and natural regeneration (NR).

according to Rs and juvenile cohort vs. regeneration cohort according to HE (P = 0.05). Values of FIS varied according to cohorts in the two stands and positive or negative values per locus (not shown) were met without any real trend. In forest, the lowest multilocus FIS was observed for the regeneration (FIS = −0.002)

1236 B . A . K E L L Y , O . H A R D Y and J . - M . B O U V E T Table 3 Levels of gene diversity and individual inbreeding within cohorts and stands. The multilocus averages of the following statistics are reported: number of alleles detected (A0), allelic richness (Rs), expected heterozygosity (HE), and inbreeding coefficient (FIS). A randomization procedure was used to test if FIS deviated significantly from zero Forest

N A0 Rs HE FIS

Fallow

AD

JUV

NR

Total

AD

JUV

NR

Total

78 3.8 3.48 0.42 − 0.018 ns

52 3.5 3.39 0.44 − 0.075 ns

90 4 3.63 0.42 − 0.002 ns

220 4.2 3.51 0.43 − 0.026 ns

68 3.6 3.52 0.38 0.024 ns

63 4.2 3.93 0.44 0.073 ns

90 3.4 3.29 0.39 0.036 ns

221 4.0 3.45 0.41 0.045†

N: sample size for adults (AD), Juveniles ( JUV), natural regeneration (NR) and the pooled sample (TOTAL). Levels of significance: †P < 0.10; ns, not significant.

Table 4 Differences in levels of gene diversity and inbreeding among cohorts and among stands as measured by Rs, HE and FIS. Levels of significance were obtained after 5000 permutations using fstat (Goudet 2001)

Comparisons

Mean Diff. (Rs)

Mean Diff. (HE)

Mean Diff. (FIS)

Differences among cohorts within stand Forest AD vs. NR − 0.15 ns 0.00 ns 0.007 ns AD vs. JUV 0.09 ns − 0.02 ns 0.073 ns JUV vs. NR − 0.24 ns 0.02 ns − 0.080 ns Fallow AD vs. NR 0.23 ns − 0.01 ns − 0.014 ns AD vs. JUV − 0.41† − 0.06** 0.050 ns JUV vs. NR 0.64 ns 0.05† 0.038 ns Differences among stands within cohort or for the pooled cohorts (Total) AD FO vs. FA − 0.04 ns 0.04* − 0.035 ns JUV FO vs. FA − 0.54 ns 0.00 ns − 0.151 ns NR FO vs. FA 0.34 ns 0.03 ns − 0.042 ns TOTAL FO vs. FA 0.06 ns 0.02 ns − 0.070 ns AD, adults; JUV, juveniles; NR, natural regeneration; FO, forest stand; FA, fallow stand. Levels of significance: †P = 0.05; *P < 0.05; **P < 0.01; ns, not significant.

but the FIS of the three cohorts were not significant. In fallow, the lowest FIS was observed for the adults (FIS = 0.024) but the three cohorts did not exhibit FIS values significantly different from zero. When the whole population was considered, fallow displayed the highest multilocus FIS (FIS = 0.045), which was significant at the 10% level (Table 3). No significant difference in FIS was found between cohorts within the two stands and between stands within cohorts (Table 4).

demographically stable population (Wilcoxon sign-rank tests were all nonsignificant at the 5% level). Hence, there is no sign of the occurrence of recent bottleneck events.

Temporal genetic structure — F-statistics among cohorts within each stand Average within-cohort inbreeding level (FIS) was higher in the fallow (FIS = 0.045) than in the forest (FIS = –0.025) (Table 5). Although FST values were estimated with high precision (SE = 0.003 and 0.004 for forest and fallow stands, respectively), differentiation among cohorts was weak and not significant (Table 5). This result was confirmed in forest when computing pairwise FST between cohorts; only the comparison of adults and juveniles was marginally significant: FST (P-value) = 0.0114 (0.0521), 0.0042 (0.2177) and 0.0079 (0.0717) for adult–juvenile, adult–regeneration and juvenile –regeneration, respectively. In fallow, FST among cohorts was not significant, but pairwise FST values, estimated with 10 loci, suggest that juveniles and regeneration were differentiated: FST (P-value) = 0.0056 (0.2099), 0.0027 (0.4075) and 0.0081 (0.0413) for adult–juvenile, adult–regeneration and juvenile–regeneration, respectively.

Genetic structure — F-statistics between stands within each cohort All FIS and FST values were positive and gave low values. Only FST for adults (FST = 0.026) was significant (Table 6). Thus, cohorts of the two stands did not display any heterozygote deficit and were weakly genetically differentiated.

Bottleneck effects

Spatial genetic structure

Whatever the mutation model assumed for the microsatellites (infinite allele mutation model or stepwise mutation model), gene diversity found in each cohort and each stand was consistent with the number of alleles per locus for a

The spatial genetic structure within and between cohorts within the forest and fallow stands are depicted in Fig. 2. Globally, the genetic structure was more pronounced in the fallow stand, where it appeared within as well as between © 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 1231–1240

W E A K I M P A C T O F P A R K L A N D I N S H E A T R E E D I V E R S I T Y 1237 Table 5 Temporal genetic structure as measured by F-statistics among cohorts within each stand. Significance levels are obtained after 10 000 permutations. Standard errors (SE) of the multilocus estimates were obtained by jack-knifing over loci Forest (3 cohorts: AD, JUV and NR)

Fallow (3 cohorts: AD, JUV and NR)

Locus

FIS

FST

FIS

FST

B5 E4‡ E11 E6a B3 H4 D10‡ D6‡ F5 E5 All SE

− 0.0121 ns 0.1221** − 0.0631 ns − 0.0593 ns − 0.0042 ns − 0.0069 ns 0.1934*** 0.1561** − 0.0153 ns 0.4974* − 0.025 ns 0.014

− 0.0027 ns 0.0114 ns − 0.0056 ns − 0.0076** − 0.0019 ns 0.0000 ns 0.0356** 0.0007 ns 0.0077 ns − 0.0045 ns − 0.001 ns 0.003

0.0107 ns 0.2640*** − 0.0259 ns 0.0146 ns − 0.0214 ns 0.0843 ns 0.1869*** 0.1720** 0.1445*

0.0107 ns 0.0049 ns − 0.0043 ns 0.0092 ns − 0.0055 ns − 0.0005 ns 0.0197* 0.0047 ns − 0.0081 ns

0.044† 0.032

0.001 ns 0.004

Levels of significance: †P < 0.10, *P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant. ‡Loci suspected to display null alleles that were not used for the calculation of the multilocus FIS.

cohorts, than in the forest stand, where adults showed no spatial structuring. This is confirmed by the regression slopes of Fij on ln(distance) which were equal to: (i) in forest, −0.00005 (P = 0.97) within adults, −0.00697 (P = 0.012) within juveniles and −0.00653 (P = 0.003) for adult–juvenile pairs; (ii) in fallow, −0.01141 (P = 0.013) within adults, −0.02147 (P < 0.001) within juveniles and −0.02034 (P < 0.001) for adult– juvenile pairs. Curiously, in the fallow stand, Fij values decreased sharply at distances > 120 m, whatever the pairwise comparisons. A closer look at the data showed that these values resulted from a subgroup of six adults and eight juveniles located at one extremity of the stand. When these 14 individuals were removed from the sample, the genetic structure appeared very similar to that in the forest stand (Fig. 2, lower), although it remained somewhat more marked: regression slopes of −0.00645 (P = 0.055) within adults, −0.01177 (P = 0.003) within juveniles and −0.00721 (P = 0.018) for adult–juvenile pairs. It is noteworthy that, in both stands, mean Fij at short distance (< 15 m) was higher for adult–juvenile pairs than within adults or juveniles. This makes sense under limited seed dispersal because a proportion of such adult–juvenile pairs may correspond to mother– offspring pairs.

Discussion Genetic diversity in Vitellaria paradoxa To our knowledge, this is the first study using microsatellites to investigate the genetic diversity and spatial © 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 1231–1240

Fig. 2 Mean pairwise kinship coefficients between individuals according to distance in the forest stand (upper), the fallow stand (middle), and in the fallow stand when a subsample of individuals were removed (lower, see text). Adult–adult (square), juvenile– juvenile (circle) and adult–juvenile (diamond) pairs are distinguished. Filled symbols show values departing significantly (α = 5%) from a random spatial distribution of genotypes.

genetic structure of a savannah woody tree species. In the two stands, genetic diversity parameters of Vitellaria paradoxa were low compared with microstaellite marker results for other tropical tree species with the same pollination and seed-dispersal systems. For instance, the mean allele number, A0, varied from 3.4 to 4.2 for cohorts in the two stands, while Aldrich et al. (1998) observed for Symphonia globulifera, a tropical forest tree species with a 4–5 cm diameter, single-seeded fruit mainly dispersed by bats in a forest stand, a mean allele number of 11.66, 13 and 16 for adults, saplings and seedlings, respectively. For the same cohorts, S. globulifera had a mean expected heterozygosity (HE) of 0.819, 0.823 and 0.775, which were much higher than those observed for V. paradoxa in the two stands. In tropical forest trees (Hall et al. 1994), as well as in temperate evergreen woody plants (Soo Oh et al. 1996), genetic diversity parameters did not vary greatly according to cohorts, and this was also observed for V. paradoxa. With regard to the impact of human practices, it is interesting to note that genetic diversity parameters in the two stands were not significantly different (Table 4). Hence, human practices do not cause a significant bottleneck, and/or gene flow between forest stands and field/fallow

1238 B . A . K E L L Y , O . H A R D Y and J . - M . B O U V E T Table 6 F-statistics for comparison between stands within each cohort. Significance levels are obtained after 5000 permutations Cohort

FIS

FST

AD JUV NR

0.001 ns 0.014 ns 0.016 ns

0.026** 0.0005 ns 0.010 ns

Level of significance: **P < 0.01; ns, not significant. AD: adults; JUV: juveniles; NR: regeneration.

stands is sufficient to compensate any human-mediated bottleneck effect.

Inbreeding level of Vitellaria paradoxa The multilocus FIS values for each cohort were close to 0 and not significant at the 5% level in forest and fallow, suggesting that selfing is weak or absent in this tree species. In fallow, however, for the total population, FIS was positive and significant at the 10% level. Many authors have attributed significantly positive FIS values to population substructure and inbreeding (Ueno et al. 2002). Positive FIS could be explained by more selfing in fallow than in forest, but we do not have reliable information on the mating system of this tree species in the two stands to confirm this. FIS values close to 0 were similar to those observed for other tree species with the same pollination and seed-dispersal systems such as S. globulifera (FIS = 0.02; Aldrich et al. 1998) and Cecropia obtusifolia (FIS = −0.081; Alvarez-Buylla et al. 1996), a pioneer tropical tree species with bird- and mammaldistributed seeds and some gravity-dispersed seed. Although the multilocus FIS values over the seven loci were not significantly different from zero, they decreased weakly with age in the two stands (Table 3). This trend is similar to that found in S. globulifera (FIS = 0.20, 0.10 and 0.02 for seedlings, saplings and adults, respectively) and C. obtusifolia (FIS = 0.065, −0.005 and −0.081 for the same cohorts). For these species, however, FIS values decreased more strongly with age than was the case for V. paradoxa. Hamrick et al. (1993) attributed a decrease in FIS with age as being primarily the result of a loss of fine-scale genetic structure with age. Such a pronounced effect does not occur in V. paradoxa, unless it happens at an earlier stage, from seed to seedling. In the case of V. paradoxa, FIS values close to zero for juveniles and the regeneration might be a consequence of a recent increase in gene flow from fields surrounding the two stands, which show a higher flowering and fruiting yield than the forest. Indeed, 100– 150 years ago, conditions were very different as there were few cultivated areas (less pollen and fewer seeds produced), more abundant vegetation (high density per hectare), and fewer dissemination agents, such as humans, resulting in less gene flow and weaker seed dispersal than today.

Genetic differentiation between stands The multilocus FST values between stands were significantly positive, whatever the cohort (Table 6), suggesting that the two stands were differentiated, but this differentiation was very weak (low FST and two of three not significantly different from zero). This suggests that these populations have the same genetic basis and have shared the same evolutionary processes, being in the same geographical space. Moreover, one cannot exclude that the forest was also used as a field a long time ago. This is the case for most of the stands considered as ‘forest’ in soudanian and soudano –sahelian zones, and so this term should be used with caution. In addition, gene flow via pollen and/or seeds may be sufficient to compensate for differentiation by drift. Note that low FST values are usually noted for tree species, most of them exhibiting > 90% of the total genetic variation within a population (Ledig 1986; Cheliak et al. 1988).

Temporal genetic structure Differentiation among cohorts within each stand was weak. Significantly positive F ST values among cohorts would have been expected if: (i) young cohorts had many genes coming from a differentiated population; and/or (ii) young cohorts showed a marked reduction in gene diversity relative to the parental population. The absence of a clear differentiation among cohorts is thus consistent with the low differentiation observed among stands and the absence of a reduction in gene diversity in young cohorts.

Spatial genetic structure In studies of spatial genetic structure, many authors have reported either a weak or even a lack of spatial structure for various tropical or temperate tree species, e.g. Quercus spp. (Berg & Hamrick 1994; Streiff et al. 1998), Psychotria officinalis (Loiselle et al. 1995), Rhus spp. (Chung et al. 1999). Their results are explained by limited seed dispersal and extensive gene flow. Other authors (Dewey & Heywood 1988; Loiselle et al. 1995; Chung et al. 1999, 2000) reported that for woody insect-pollinated species with seeds widely and independently dispersed by birds, only weak spatial genetic structure will result. For C. obtusifolia, Epperson & Alvarez-Buylla (1997) found a very weak spatial structure in juveniles and adults but strong genetic spatial autocorrelations among seedlings, which they explain as a consequence of thinning processes. A lack of spatial structure was found for other tree species by Sokal & Oden (1978), Waser (1987), Dewey & Heywood (1988), Doligez & Joly (1997) and Chung et al. (2000). These authors explained their results by extensive gene flow, wide seed dispersal, self-incompatibility and dispersal agents. © 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 1231–1240

W E A K I M P A C T O F P A R K L A N D I N S H E A T R E E D I V E R S I T Y 1239 In the forest stand of V. paradoxa, no significant spatial structure was found for adults, but juveniles showed a spatial structure and were also more closely related to nearby adults, demonstrating the impact of limited seed dispersal (Fig. 2). In the fallow stand, a spatial structure was found for both adults and juveniles, and juveniles were also more closely related to nearby adults (Fig. 2). The extent of structuring was more pronounced in the fallow stand than in the forest stand, although this difference was attenuated when a differentiated subsample of individuals from the fallow was removed. Different factors could be responsible for a stronger spatial genetic structure in the fallow stand: the high degree of fruiting (no crown competition between trees, good soil conditions) favouring the important natural regeneration of siblings in the neighbourhood of the mother trees as seeds of V. paradoxa are mostly dispersed by gravity, and the restricted seed dissemination by secondary dispersal agents such as farmers who abandoned the fallow and devoted their time to the exploitation of fields. By contrast, in forest, the low degree of fruiting of trees (strong competition) reduced the number of siblings in the neighbourhood of the mother trees. The higher species richness of the forest ecosystem and the habit of the dispersal agents may also lead to a reduction in spatial genetic structure. For example V. paradoxa fruit eaters (birds, bats and even humans working in forest gathering firewood) pick fruits from different mother trees and disperse them in the forest. Important regeneration of V. paradoxa found under other tree species in the forest supports this hypothesis. As seed dissemination of V. paradoxa occurs during the rainy season, another possible factor enhancing seed dispersal is surface water flow mixing seeds several metres apart. We observed this phenomenon during fieldwork, particularly in the forest stand where a smooth slope favours surface water flow.

The impact of humans on the pattern of genetic variation in Vitellaria paradoxa The field/fallow cycle was heavily influenced by human practices, leading to low species diversity, high tree selection, low density and high fruit production. Nevertheless, the impact of these practices on the pattern of genetic variation at microsatellite loci in V. paradoxa appears to be limited. Genetic diversity indicators were not reduced in the fallow stand, inbreeding level was weakly affected, the fallow stand was little differentiated from the forest stand and cohorts were also little differentiated. There was, however, somewhat more spatial genetic structuring in the fallow stand, which might reflect a difference in seed dispersal patterns due to human practices. Frequent gene flow via seed and/or pollen is likely to be a determining factor buffering the possible impact of human practices on genetic differentiation. It must, however, be kept in mind that © 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 1231–1240

microsatellites are representative only of neutral genetic variation. Hence, they can be powerful for investigating the impact of humans on neutral diversity, mating system, drift and gene flow, but not on phenotypic characters that are subject to selection. Human practices surely changed substantially the nature of the selection processes acting on V. paradoxa trees in the field/fallow cycle. More research is needed to explore this impact using larger samples and modelling on longer cycles. Other markers associated with adaptive phenotypic difference may also be used in future (SNPs).

Acknowledgements We are grateful to the two anonymous reviewers for their comments to improve the manuscript. We are very grateful to Céline Cardi for having carried out a huge laboratory work for the development of microsatellites for the first time regarding the tree species (Vitellaria paradoxa). Many thanks to EU for financial support, to IER (Mali) and Cirad-forêt (Montpellier, France) for all logistic support which allowed us to undertake field and laboratory studies in a convivial atmosphere. Our thanks are also addressed to all those who contributed at any time, in any way to this work and particularly to the farmers of Mperesso who allowed us to carry out this study on their land and who participated in field activities.

References Aldrich PR, Hamrick JL, Chavarriaga P, Kochert G (1998) Microsatellite analysis of demographic genetic structure in fragmented populations of the tropical tree Symphonia globulifera. Molecular Ecology, 7, 933–944. Alvarez-Buylla EA, Chaos A, Pinero D, Garay AA (1996) Demographic genetic of a pioneer tropical tree species: patch dynamics, seed dispersal, and seed banks. Evolution, 50, 1155–1166. Austerlitz F, Mariette S, Machon N, Gouyon P-H, Godelle B (2000) Effects of colonization processes on genetic diversity: differences between annual plants and tree species. Genetics, 154, 1309 – 1321. Bagnoud N, Schmithüsen F, Sorg J-P (1995) Les parcs à Karité et Néré au Sud-MALI. Analyse du Bilan économique des arbres associés aux cultures. Bois et Forêts Des Tropique, 244, 9–23. Berg EE, Hamrick JL (1994) Spatial and genetic structure of two sand hills oaks: Quercus laevis and Quercus margaretta (Fagaceace). American Journal of Botany, 81, 7–14. Billotte N, Lagoda PJL, Risterucci A-M, Baurens F-C (1999) Microsatellite enriched libraries: applied methodology for the development of SSR markers in tropical crops. Fruits, 54, 277–288. Boffa JM (1998) Agroforestry Parklands Systems. FAO, Rome. Bousquet J, Simon L, Lalonde M (1990) DNA amplification from vegetative and sexual tissues of trees using polymerase chain reaction. Canadian Journal of Forestry Research, 20, 254–257. Cheliak WM, Wang J, Pitel JA (1988) Population structure and genetic diversity in tamarack, Larix laricina (Du Roi) K. Koch. Canadian Journal of Forestry Research, 18, 1318–1324. Chung MG, Chung JM, Chung MY, Epperson BK (2000) Spatial genetic distribution of allozyme polymorphisms following clonal and sexual reproduction in populations of Rhus javanica (Anacardiaceae). Heredity, 84, 178–185. Chung MG, Chung JM, Epperson BK (1999) Spatial genetic structure of allozyme polymorphisms within populations of Rhus trichocarpa (Anacardiaceae). Silvae Genetica, 48, 223–227.

1240 B . A . K E L L Y , O . H A R D Y and J . - M . B O U V E T Chung MG, Epperson BK (2000a) Clonal and spatial genetic structure in Eurya emarginata (Theaceae). Heredity, 84, 170 –177. Chung MG, Epperson BK (2000b) Spatial genetic structure of allozyme polymorphisms in a population of Eurya japonica (Theaceae). Silvae Genetica, 49, 1 – 4. Cornuet JM, Luikart G (1996) Description and power analysis of two tests for detecting recent population bottlenecks from allele frequency data. Genetics, 144, 2001 – 2014. Dewey SE, Heywood JS (1988) Spatial genetic structure in a population of Psychotria nervosa I. Distribution of genotypes. Evolution, 42, 834–838. Doligez A, Joly H (1997) Genetic diversity and spatial structure within a natural stand of a tropical forest tree species, Carapa procera (Meliaceae), in French Guiana. Heredity, 79, 72 –82. Ennos RA (1994) Estimating the relative rates of pollen and seed migration among plant populations. Heredity, 72, 250 – 259. Epperson BK (1993) Recent advances in correlation analysis of spatial patterns of genetic variation. Evolutionary Biology, 27, 95–155. Epperson BK, Alvarez-Buylla ER (1997) Limited seed dispersal and genetic structure in life stages of Cecropia obtusifolia. Evolution, 51, 275–282. Geburek T, Tripp-Knowles P (1994) Genetic architecture in bur oak, Quercus macrocarpa (Fagaceae), inferred by means of spatial autocorrelation analysis. Plant Systematics and Evolution, 189, 63–74. Goudet J (2001) FSTAT, a Program to Estimate and Test Gene Diversities and Fixation Indices, Version 2.9.3. http://www.unil.ch/izea/ softwares/fstat.html. Hall P, Chase MR, Bawa KS (1994) Low genetic variation but high population differentiation in a common tropical forest tree species. Conservation Biology, 8, 471– 482. Hamrick JL, Murawski DA, Nason JD (1993) Influence of seed dispersal mechanisms on the genetic structure of tropical tree populations. Vegetation, 107 – 108, 281 –297. Hardy OJ (2000) Structure spatiale de la variation génétique et relations entre cytotypes dans des populations d’un complexe polyploïde: Centaurea jacea L. s.l. PhD thesis, Université Libre de Bruxelles, Brussels. Hardy OJ, Vekemans X (2002) SPAGEDI 1 .0 Freeware Program for Spatial Pattern Analysis of Genetic Diversity. Université Libre de Bruxelles. Hossaert-Mckey M, Valero M, Magda D, Jarry M, Cuguen J, Vernet P (1996) The evolving history of a population of Lathyrus sylvestris. Evidence from temporal and spatial genetic structure. Evolution, 50, 1808–1821. Isagi Y, Kanazashi T, Suzuki W, Tanaka H, Abe T (2000) Microsatellite analysis of the regeneration process of Magnolia obovata Thunb. Heredity, 84, 143 –151. Kitamura K, Shimada K, Nakashyma K, Kawano S (1997a) Demographic genetics of the Japanese beech, Fagus crenata, at Ogowa Forest Preserve, Ibaraki, Central Honshu, Japan. I. Spatial genetic structuring in local populations. Plant Species Biology, 12, 107–135. Kitamura K, Shimada K, Nakashyma K, Kawano S (1997b) Demographic genetics of the Japanese beech, Fagus crenata, at Ogowa Forest Preserve, Ibaraki, Central Honshu, Japan. II. Genetic structuring among size-classes in local populations. Plant Species Biology, 12, 137–155. Knowles P, Perry DJ, Foster HA (1992) Spatial genetic structure in two tamarack (Larix laricina) populations with differing establishment histories. Evolution, 46, 572–576. Ledig FT (1986) Heterozygosity, heterosis, and fitness in outbreeding plants. In: Conservation Biology (ed. Soulé ME), pp. 77–104. Sinauer Associates, Sunderland, MA.

Leonardi S, Raddi S, Borghetti M (1996) Spatial autocorrelation in a Norway spruce (Picea abies) population. Canadian Journal of Forestry Research, 26, 63–71. Levin DA, Kerster HW (1974) Gene flow in seed plants. Evolutionary Biology, 7, 139–220. Loiselle BA, Sork VL, Nason J, Graham C (1995) Spatial genetic structure of a tropical understoty shrub, Psychotria officinalis (Rubiaceae). American Journal of Botany, 82, 1420–1425. Michalakis Y, Excoffier L (1996) A genetic estimation of population subdivision using distances between alleles with special reference for microsatellite loci. Genetics, 142, 1061–1064. Nei M (1987) Molecular Evolutionary Genetics. Columbia University Press, New York. Perry DJ, Knowles P (1991) Spatial genetic structure within three sugar maple (Acer saccharum Marsh) stands. Heredity, 66, 137–142. Petit RJ (2001) Chloroplast DNA variation in European white oaks. Phylogeography and patterns of diversity based on data from over 2600 populations. Forest Ecology and Management, 5692, 1–22. Petit RJ, El Mousadik A, Pons O (1998) Identifying populations for conservation on the basis of genetic markers. Conservation Biology, 12, 844–855. Ritland K (1996) Estimators for pairwise relatedness and individual inbreeding coefficients. Genetical Research, 67, 175 – 185. Schaal BA (1980) Measure of gene flow in Lupinus texensis. Nature, 284, 450–451. Schnabel AR, Laushiman RH, Hamrick JL (1991) Comparative genetic structure of two co-occuring tree species, Maclura pomifera (Moraceae) and Gleditsia triacanthos (Leguminoseae). Heredity, 67, 357–364. Sokal RR, Oden NL (1978) Spatial autocorrelation in biology; 2. Some biological implication and applications of evolutionary and ecological interest. Biology Journal of Linnaean Society, 10, 229 –249. Soo Oh G, Kang SS, Chung MG (1996) Temporal genetic structure in Camellia japonica (Theaceae). Genes and Genetic System, 71, 9 – 13. Streiff R, Labbe T, Bacilieri R, Steinkellner H, Glössl H, Kremer A (1998) Within-population genetic structure in Quercus robur L. & Quercus petraea (Matt.) Liebl. assessed with isozymes and microsatellites. Molecular Ecology, 7, 317–328. Ueno S, Tomaru N, Yoshimaru H, Manabe T, Yamamoto S (2002) Size class differences in genetic structure and individual distribution of Camellia japonica L. in a Japanese old-growth evergreen forest. Heredity, 89, 120–126. Waser NM (1987) Spatial genetic heterogeneity in a population of the mountain perennial plant Delphinium nelsonii. Heredity, 58, 249–256. Young AC, Merriam HG (1994) Effect of forest fragmentation on the spatial genetic structure of Acer saccharum Marsh. (sugar maple) populations. Heredity, 72, 201–208.

This study is part of BA Kelly’s PhD thesis on the impact of agroforestry practices on genetic structure in Vitellaria paradoxa. This study was part of the INCO project founded by European Union for the ‘Improved management of agroforestry parkland system in Sub-Saharan Africa’. Institut d’Economie Rurale (IER) in Mali where the study was undertaken was one of the contractors of the INCO project in addition to Cirad-forêt where laboratory work was done and where JM Bouvet is the responsible of genetic resources programme. OJ Hardy is a postdoctoral researcher from the Belgian National Fund for Scientific Research (FNRS).

© 2004 Blackwell Publishing Ltd, Molecular Ecology, 13, 1231–1240