## Abstract

Background and Aims

The Tehuacán Valley in Mexico is a principal area of plant domestication in Mesoamerica. There, artificial selection is currently practised on nearly 120 native plant species with coexisting wild, silvicultural and cultivated populations, providing an excellent setting for studying ongoing mechanisms of evolution under domestication. One of these species is the columnar cactus Stenocereus pruinosus, in which we studied how artificial selection is operating through traditional management and whether it has determined morphological and genetic divergence between wild and managed populations.

Methods

Semi-structured interviews were conducted with 83 households of three villages to investigate motives and mechanisms of artificial selection. Management effects were studied by comparing variation patterns of 14 morphological characters and population genetics (four microsatellite loci) of 264 plants from nine wild, silvicultural and cultivated populations.

Key Results

Variation in fruit characters was recognized by most people, and was the principal target of artificial selection directed to favour larger and sweeter fruits with thinner or thicker peel, fewer spines and pulp colours others than red. Artificial selection operates in agroforestry systems favouring abundance (through not felling plants and planting branches) of the preferred phenotypes, and acts more intensely in household gardens. Significant morphological divergence between wild and managed populations was observed in fruit characters and plant vigour. On average, genetic diversity in silvicultural populations (HE = 0·743) was higher than in wild (HE = 0·726) and cultivated (HE = 0·700) populations. Most of the genetic variation (90·58 %) occurred within populations. High gene flow (NmFST > 2) was identified among almost all populations studied, but was slightly limited by mountains among wild populations, and by artificial selection among wild and managed populations.

Conclusions

Traditional management of S. pruinosus involves artificial selection, which, despite the high levels of gene flow, has promoted morphological divergence and moderate genetic structure between wild and managed populations, while conserving genetic diversity.

## INTRODUCTION

Domestication is a continuous evolutionary process guided by humans, mainly through artificial selection (Darwin, 1859). It involves a variety of mechanisms, but from about 2300 crop species cultivated throughout the world (Reid and Miller, 1989) such mechanisms have been documented for relatively few species. Numerous ongoing processes of domestication have yet to be studied, and they could be a valuable source of information to understand the diversity of forms through which humans drive plant evolution. Of particular importance are incipient domestication processes occurring in hundreds of species throughout the world, as they could help to reconstruct and understand the earlier phases of domestication that originated agriculture.

Mesoamerica, the cultural region between southern Mexico and northern Costa Rica, is recognized as one of the main centres of domestication of plants in the world (Vavilov, 1951; Harlan, 1975). There, native peoples have domesticated more than 200 plant species and currently conduct incipient domestication of several hundred species (Casas and Parra, 2007). This area therefore provides an excellent setting for studying ongoing mechanisms of both advanced and incipient processes of domestication. Domestication of several species of Mesoamerican columnar cacti has been studied (Casas et al., 2007) and it makes them appropriate systems for analysing the consequences of artificial selection in a gradient of management intensity. The present study documents the case of Stenocereus pruinosus, one of the most intensely managed species of cacti in the region.

Columnar cacti are dominant components of several vegetation types in arid and semi-arid zones, which cover nearly half of the Mexican territory, and they have been used as main plant resources since the first stages of human occupation of the area during prehistory (Smith, 1967; Casas and Barbera, 2002). The Tehuacán Valley has one of the highest concentrations of columnar cacti species in the world (a total of 20 species) and is one of the more representative regions of the human culture using and managing these plants for food, fodder, medicine, fuel, house construction, gardens and fencing (Casas et al., 1999a).

Fruits of S. pruinosus have the highest quality and economic value compared with those of other columnar cacti in the regions where it is distributed, and they are commonly gathered in wild populations which form part of thorn-scrub and tropical dry forests (Casas et al., 1999a; Luna-Morales and Aguirre, 2001). Luna-Morales and Aguirre (2001) documented the management of this species in home gardens of the La Mixteca Baja region of Oaxaca, and Casas et al. (1999a) and Parra et al. (2008) documented aspects of cultivation of this species in the Tehuacán Valley. These authors found that some populations are under silvicultural management in agroforestry systems, where wild individuals are left standing when land is cleared to establish agricultural fields. This management appears to involve artificial selection favouring survival and propagation of those plants with attributes preferred by people, such as larger and sweeter fruits, thinner or thicker peel with fewer spines, and pulp with different colours (Casas et al., 1999a). Stenocereus pruinosus is also cultivated in traditional gardens through propagation of branches from plants showing desirable phenotypes. Similar silvicultural management and cultivation forms have been documented in other cactus species such as Stenocereus stellatus (Casas et al., 1999b), Escontria chiotilla (Arellano and Casas, 2003), Polaskia chende (Cruz and Casas, 2002), P. chichipe (Carmona and Casas, 2005) and Myrtillocactus schenckii (Blancas et al., 2009). In all these species, silvicultural and cultivated populations in home gardens show significant morphological divergence with respect to wild populations as the managed populations have higher frequencies of the preferred phenotypes.

According to Arias et al. (1997), S. pruinosus may reach 8 m in height, with green branches having a pruinose, whitish apex, and five to eight ribs. Its flowers are 7–10 cm in length, the outermost tepals are brownish green and rigid, whereas the innermost are white and fleshy. Flowers have nocturnal anthesis, are pollinated by bats and have a self-incompatible breeding system (Cortés-Díaz, 1999). Fruits are ellipsoid, edible, with white, yellow, purple, red and orange pulp (Luna-Morales et al., 2001), and produce black seeds. Vegetative propagation is easy and growth relatively fast (Casas et al., 1999a). The species is mainly distributed in the states of Oaxaca, Puebla, Chiapas, Guerrero, Tamaulipas, Veracruz and San Luis Potosí (Parra et al., 2008).

Otero-Arnaiz et al. (2005), Tinoco et al. (2005), Casas et al. (2006, 2007) and Parra et al. (2008) reported that traditional management of several species of columnar cacti maintains high levels of genetic variation in managed populations, similar to those existing in wild populations or even higher in S. stellatus and S. pruinosus. According to these authors, such a pattern is probably due to continual replacement of plants in managed stands, frequently involving the introduction of plants from different towns or regions. It is also probably due to the protection of seedlings and young plants resulting from natural crossing that become established in cultivated areas, all of which favours gene flow among wild and managed populations. The studies mentioned above identified that traditional home gardens and agroforestry systems are important reservoirs of genetic diversity and keystones for programmes of genetic resource conservation.

Low genetic differentiation and high levels of gene flow have been generally found between wild and managed populations of columnar cacti (Casas et al., 2007). In Stenocereus, such gene flow is influenced by the main pollinators, bat species of the genera Leptonycteris and Choeronycteris (Casas et al., 1999b; Arias-Cóyotl et al., 2006), as well as seed-dispersers, including bats, birds and humans (Casas et al., 2007). For S. pruinosus we previously found through isozyme analysis one of the highest values of genetic variation reported for columnar cacti species (Parra et al., 2008), as well as a pattern of genetic differentiation of populations associated with management type. We hypothesized that these patterns were the combined result of the high levels of gene flow which favours variation, and artificial selection favouring population differentiation. To test this hypothesis, the present study examined how management and artificial selection of S. pruinosus is conducted under traditional contexts in the Tehuacán Valley, and whether and to what extent it results in morphological and genetic divergence. The main aims were: (1) to document management types of S. pruinosus in traditional villages of the region, the cultivation techniques and sources of plant material for planting, and the motives and mechanisms through which artificial selection operates; (2) to evaluate the consequences of artificial selection on frequencies of morphological characters and on the degree of phenotypic divergence between wild and managed populations; and (3) to analyse the consequences of management for population genetic diversity, genetic structure and rates of gene flow among populations.

## MATERIALS AND METHODS

### Populations studied

The study was conducted in the villages of San Luis Atolotitlán, Coatepec and Coxcatlán, in the Tehuacán Valley, central Mexico (Fig. 1). Nine populations (three wild, three silvicultural and three cultivated in home gardens) were studied, sampling 30 individual plants per population. Wild populations are found at Santa Lucía and Fiscal within the territory of Coatepec, and at Cueva del maíz within the territory of Coxcatlán. All these populations form part of natural tropical deciduous forest associated with alluvial valleys of ephemeral rivers (Table 1, Fig. 2). In these habitats the columnar cacti Pachycereus weberi, P. hollianus, Escontria chiotilla, Stenocereus pruinosus, and S. stellatus are co-dominant with the trees Prosopis laevigata (Mimosaceae), Cyrtocarpa procera (Anacardiaceae), Ceiba aesculifolia (Bombacaceae), Bursera morelensis (Burseraceae) and Parkinsonia praecox (Caesalpinaceae). Silvicultural populations are located in scattered areas of agroforestry systems used for seasonal agriculture of maize, near the villages of San Luis Atolotitlán, Coatepec and Coxcatlán, which have been used for more than 100 years, with successive use and fallow periods. Silvicultural management involves not felling individuals with useful phenotypes when natural vegetation is cleared for agriculture (Casas et al., 1999a). The cultivated populations comprise stands of S. pruinosus cultivated in home gardens of the villages mentioned, usually associated with other cultivated plant species (Fig. 2).

Fig. 1.

Study area. The Tehuacán–Cuicatlán Valley. Location of the villages and populations of Stenocereus pruinosus studied. WI = Wild Santa Lucía; WII = Wild Fiscal; WIII = Wild Coxcatlán; SI = Silvicultural from San Luis Atolotitlán (S.L.A); SII = Silvicultural Coatepec; SIII = Silvicultural Coxcatlán; CI = Cultivated from San Luis Atolotitlán (S.L.A); CII = Cultivated Coatepec; CIII = Cultivated Coxcatlán.

Fig. 1.

Study area. The Tehuacán–Cuicatlán Valley. Location of the villages and populations of Stenocereus pruinosus studied. WI = Wild Santa Lucía; WII = Wild Fiscal; WIII = Wild Coxcatlán; SI = Silvicultural from San Luis Atolotitlán (S.L.A); SII = Silvicultural Coatepec; SIII = Silvicultural Coxcatlán; CI = Cultivated from San Luis Atolotitlán (S.L.A); CII = Cultivated Coatepec; CIII = Cultivated Coxcatlán.

Fig. 2.

Populations of Stenocereus pruinosus under different management types. (A) Wild population forming part of a tropical deciduous forest, (B) silvicultural managed population in an agroforestry system, (C) cultivated population in a home garden.

Fig. 2.

Populations of Stenocereus pruinosus under different management types. (A) Wild population forming part of a tropical deciduous forest, (B) silvicultural managed population in an agroforestry system, (C) cultivated population in a home garden.

Table 1.

Environmental aspects of the populations of Stenocereus pruinosus studied in the Tehuacán–Cuicatlán Valley

Population Code Location Elevation (m) Vegetation Annual rainfall (mm) Soils
Wild Santa Lucia WI Santa Lucía, Coatepec 1210 Tropical deciduous forest 544·4 Alluvial
Wild Fiscal WII Fiscal, Coatepec 1210 Tropical deciduous forest 544·4 Alluvial
Wild Coxcatlán WIII ‘Cueva del Maíz’, Coxcatlán 1010 Tropical deciduous forest 394·6 Alluvial
Silviculture S.L.A SI San Luis Atolotitlán (S.L.A) 1900 Seasonal maize fields 394·6 Derived from basaltic rocks
Silviculture Coatepec SII Santiago Coatepec 1853 Seasonal maize fields 394·6 Derived from basaltic rocks
Silviculture Coxcatlán SIII Coxcatlán 1000 Seasonal maize fields 544·4 Regosols, calcareous
Cultivated S.L.A CI San Luis Atolotitlán 1903 Home gardens 394·6 Derived from basaltic rocks
Cultivated Coatepec CII Santiago Coatepec 1780 Home gardens 394·6 Derived from basaltic rocks
Cultivated Coxcatlán CIII Venta Salada, Coxcatlán 1145 Home gardens 544·4 Regosols, calcareous
Population Code Location Elevation (m) Vegetation Annual rainfall (mm) Soils
Wild Santa Lucia WI Santa Lucía, Coatepec 1210 Tropical deciduous forest 544·4 Alluvial
Wild Fiscal WII Fiscal, Coatepec 1210 Tropical deciduous forest 544·4 Alluvial
Wild Coxcatlán WIII ‘Cueva del Maíz’, Coxcatlán 1010 Tropical deciduous forest 394·6 Alluvial
Silviculture S.L.A SI San Luis Atolotitlán (S.L.A) 1900 Seasonal maize fields 394·6 Derived from basaltic rocks
Silviculture Coatepec SII Santiago Coatepec 1853 Seasonal maize fields 394·6 Derived from basaltic rocks
Silviculture Coxcatlán SIII Coxcatlán 1000 Seasonal maize fields 544·4 Regosols, calcareous
Cultivated S.L.A CI San Luis Atolotitlán 1903 Home gardens 394·6 Derived from basaltic rocks
Cultivated Coatepec CII Santiago Coatepec 1780 Home gardens 394·6 Derived from basaltic rocks
Cultivated Coxcatlán CIII Venta Salada, Coxcatlán 1145 Home gardens 544·4 Regosols, calcareous

### Ethnobotanical survey

An ethnobotanical survey through semi-structured interviews (Martin, 1997) was conducted with 63 randomly selected households. Interviews were conducted with both men and women of each household from San Luis Atolotitlán, Coatepec and Coxcatlán in order to get information on perception of morphological variation and fruit quality, forms of use, management, and artificial selection. Data on the role of products of this species in the subsistence of local people and on commercialization of S. pruinosus were also documented. Twenty interviews were conducted with people of households from San Luis Atolotitlán and Coatepec who were managing plants of S. pruinosus in agroforestry systems in order to document details on the provenance and management of plants forming part of silvicultural systems.

### Morphometric analyses

Fourteen quantitative morphological characters were analysed in 264 individual plants of the populations studied. In wild populations reproductive plants were sampled randomly along transects following the natural vegetation bordering the course of ephemeral rivers. The sampled area of wild populations averaged 2·2 ha per site. In silvicultural populations, where individual plants are scattered, we sampled all individual plants found until 30 individual plants were documented. The sample area averaged 4 ha per site. In cultivated populations, plants were sampled in a total area of home gardens per village equivalent to the average sampled area of wild populations.

Vegetative parts (branches, spines, areoles) were measured in the field, and average values of 3–5 measurements per individual plant were taken. Biomass of individual plants was estimated using the formula for volume of a truncated cone, v = πh/3 (R2 + Rr + r2), where h is height, R is the diameter of the canopy, calculated as the average of two perpendicular measures, and r is the diameter of the stem at first branch height. Fruit characters (total volume and weight, pulp and seed weight, and seed number) were evaluated in the laboratory; values were averaged per population and per management type. One-way ANOVAs were conducted using JMP 4·0 (SAS Institute, Cary, NC, USA) to test for differences in the particular morphological features studied among populations according to their management type. Tukey–Kramer multiple range tests were also performed to identify those populations and type of management showing significant differences. These tests were directed to visualize trends of variation according to management and artificial selection intensity.

### Population genetics studies

Four primer pairs for polymorphic nuclear microsatellite loci (Pchi50, Pchi60, Pchi54 and Pchi21), designed for Polaskia chichipe (Cactaceae) by Otero-Arnaiz et al. (2004, 2005), were used. Twenty individual plants per population were analysed, chosen among those plants evaluated in morphometric studies. Plant tissue was obtained from stem ribs, frozen in liquid nitrogen and then stored at –70 °C for conservation.

### DNA extraction and genotyping

DNA extraction was performed following Otero-Arnaiz et al. (2004, 2005). For PCR reactions, a QIAGEN multiplex PCR kit was used (www.qiagen.com). All reactions were carried out in a final volume of 5 µL, including 2·5 µL of Master Mix (containing HotStar taq DNA polymerase, Multiplex PCR buffer, 3 mm MgCl2 and dNTPs), 1 µm of each primer, 1·5 µL distilled water and 0·5–1·0 µL of 50–100 ng/μL of template DNA. Amplifications were carried out in a GeneAmp PCR System 2720 Thermal Cycler (Applied Biosystems; www.AppliedBiosystems.com) using the multiplex PCR protocol for amplification of microsatellite loci (QIAGEN Multiplex PCR kit; QIAGEN): 15 min at 95 °C (initial activation step), followed by 30 cycles of denaturing at 94 °C for 30 s, annealing at primer-specific temperatures (56 °C for Pchi50 and Pchi60 and 60 °C for Pchi54 and Pchi21) for 90 s and extension at 72 °C for 60 s. A final extension step at 60 °C for 30 min was included. PCR products were mixed with formamide and Gene Scan LIZ-500 of standard size (Applied Biosystems), and then denatured for 2 min at 95 °C. Analysis of microsatellite loci was conducted using an ABI-PRISM 3100-Avant sequencer (Applied Biosystems) to detect labelled primers and internal size standard. All the fragments obtained were then scored and analysed with GeneMapper 4·0 (Applied Biosystems) software.

### Genetic diversity within populations

For each population we estimated the average number of alleles per locus (A), the average observed heterozygosity (HO) and the average expected heterozygosity (HE; Nei, 1987) using GENEALEX 6·2 (Peakall and Smouse, 2006). Non-parametric Kruskal–Wallis H tests were used to evaluate differences in HO and HE between wild, silvicultural and cultivated populations.

The inbreeding coefficient (FIS) was calculated with GENEPOP version 4·0·SS10 (Raymond and Rousset, 1995) per locus and population to determine deviations from Hardy–Weinberg equilibrium (HWE). The exact test of HWE per population–locus combination and a global test across populations using the multiscore U-test (Raymond and Rousset, 1995) were performed to test for heterozygote deficiency using the Markov chain Monte Carlo (MCMC) method, with 1000 dememorization steps with 100 batches and 5000 iterations (Guo and Thompson, 1992).

In order to explore possible causes of deviations from HWE, we used MICRO-CHECKER (Van Oosterhout et al., 2004) to identify the presence of null alleles at each locus per population. In addition, a survey to identify possible clones (individuals with identical multilocus genotypes) was made with GENEALEX 6·2 (Peakall and Smouse, 2006). For each set of repeated genotypes the value of Psex was obtained, which is an estimate of the probability of obtaining n repeated multilocus genotypes in a sample of size N by sexual reproduction under random mating (Stenberg et al., 2003).

### Genetic differentiation among populations

Population structure and distribution of genetic variance were analysed by analysis of molecular variance (AMOVA) with Arlequin 3·11 (Excoffier et al., 2005), among the predetermined groups of populations (wild, silvicultural and cultivated in home gardens). Population subdivision was calculated by means of F-statistic analogues (Φ-statistics) obtained from the AMOVA. Significance of these statistics was assessed by permuting genotypes over populations (ΦST), among populations within groups (ΦSC) and among groups according to management type (ΦCT) using 503 permutations. Pairwise population genetic differentiation was estimated with both FST and RST according to two different microsatellite mutation models: the infinite alleles model (IAM) and the stepwise mutation model (SMM), respectively. We used both models because there is no consensus on whether the FST- or RST-based approach is better at handling microsatellite data (Otero-Arnaiz et al., 2005). RST may overestimate differentiation if microsatellites mutate by large insertions or deletions (Di Rienzo et al., 1994), whereas FST may underestimate differentiation if mutation is largely stepwise (Slatkin, 1995). Gene flow among populations was estimated from the FST statistics of Wright (1965), following the expression M = Nm = (1 – FST/4FST) with ARLEQUIN (Excoffier et al., 2005).

### Patterns of genetic structure

A pairwise matrix of Nei's unbiased genetic distances (Nei, 1972) among populations was calculated, and a clustering dendrogram was constructed using the UPGMA method with TFPGA 1·3 (Miller, 1997). Confidence levels for the dendrogram were calculated by bootstrapping the original data 1000 times with replacement over all loci. A Mantel test was performed by TFPGA 1·3 to assess the correlation between genetic and geographical distances.

A Bayesian clustering analysis was conducted using STRUCTURE version 2·1 (Pritchard et al., 2000; Falush et al., 2003). In this analysis individuals are assigned probabilistically to one of the predefined K populations (gene pools) to identify the optimal number of genetic groups (Evanno et al., 2005). The optimum number of groups (K) was determined by varying the value of K from 1 to 10 and running the analysis ten times per K value, in order to determine the maximum value of posterior likelihood [lnP(D)]. Each run was performed using 105 burn-in periods and 106 MCMC repetitions after burn-in. We used a model allowing admixture with correlated allele frequencies without any prior information. We determined the most probable K value using the maximum value of ΔK according to Evanno et al. (2005).

In order to determine the geographical location of the main genetic discontinuities among populations, we used the Monmonier's maximum difference algorithm with BARRIER version 2·2 (Manni et al., 2004). This program creates a map of the sampling locations from geographical coordinates. Barriers are then represented on the map by identifying the maximum values within the population-pairwise genetic distance matrix. We used a matrix of Slatkin's linearized FST (Slatkin, 1995) values calculated with ARLEQUIN.

## RESULTS

### Management and artificial selection

People of the Tehuacán Valley recognize and differentially use and manage phenotypic variation of S. pruinosus. All those interviewed know S. pruinosus as ‘pitaya de mayo’ and consume its fruits, and 51 % of households also use its dry branches as fuel for cooking. All recognized morphological variation in fruits, most of them (81 %) distinguishing 3–4 variants, and the rest recognizing 5–6 variants (red, pink, purple, yellow, white, orange). Approximately 30 % of households interviewed recognized sub-variants; for instance, they said that the variant with yellow pulp could be dark or light yellow, waterish or non-waterish yellow, large or small size (‘ant’ or ‘little sparrow’ yellow) and with few or numerous seeds (Fig. 3). Nearly 57 % of interviewees perceive that the largest fruits are produced by plants in home gardens, whereas the rest said that fruit size is similar in both wild and cultivated populations. A few households (approx. 6 %) considered that fruit size and production depend on the amount of rainfall in the preceding year (years with higher rainfall determine production of larger fruits), and age of plants (young individual plants produce larger fruits than older plants).

Fig. 3.

Examples of variants of pulp colour of fruits of Stenocerus pruinosus, the ‘pitaya de mayo’, and sub-variants associated with more specific characters.

Fig. 3.

Examples of variants of pulp colour of fruits of Stenocerus pruinosus, the ‘pitaya de mayo’, and sub-variants associated with more specific characters.

Individuals of almost all (90 %) households said they planted branches of S. pruinosus and to have learned this practice from their relatives. About 40 % of households that cultivate S. spruinosus said they cut branches for planting from specific mother plants according to flavour, colour and size of their fruit, which involves artificial selection. Individuals also use branches that have recently fallen from individual plants, i.e. non-selective planting. Most people planting branches of S. pruinosus use branches from individual plants previously cultivated in home gardens, but individuals of nearly 11 % of households reported bringing branches from wild populations, particularly from the wild populations studied in the sites Río Hondo and Santa Lucía, which are the main wild populations of S. pruinosus close to the village of Coatepec.

People generally considered that branches may produce fruit after 1 year of being planted, but most commonly after 3–5 years. They prefer to plant branches that have already produced fruit because, according to their experience, these branches produce fruit faster than others. Branches planted are 20–100 cm in length. The dry season is the more convenient time to cut branches for planting, in order to have the opportunity to dry the damaged tissue, whereas the best time for planting the branches is the beginning of the rainy season.

Most individuals said they did not place special care on planted branches, but 25 % of households affirmed that they (1) pile ground around the base of the stem, (2) add organic fertilizer, (3) put stones around the tree in order to direct rain water to it, (4) remove weedy plants invading the surrounding area, (5) protect fruits by covering them with clothes in order to prevent their consumption by frugivores, (6) practise occasional manual irrigation and (7) practise pruning of old branches.

People of nearly 60 % of the households interviewed said that in the past S. pruinosus was more actively cultivated than at present; 24 % of households considered that pitaya fruit is more important than in the past because it is commercially grown and provides monetary income, whereas the remaining individuals were not precise in their response. Those who stated that this fruit was more consumed in the past said that this was because no other fruit was available, but with the opening of roads the village received other commercialized fruit that substituted part of the native fruit. Among these respondents, 37 % of households mentioned that consumption of pitaya was more common in the past because it was more abundant; 11 % of households said that pitaya fruits were of better quality in the past when rainfall was higher; 29 % of households said that people had lost interest in cultivating and taking care of these plants and that this had caused a decrease of population densities in cultivated stands; the remaining households said that many pitaya plants were lost when new houses were constructed.

When interviewees were asked whether they let stand individual plants of S. pruinosus in their agricultural fields, 33 % of households answered affirmatively. Nearly half of them said that S. pruinosus plants were planted by them or by their relatives, and that branches arose from home gardens, whereas the other half said that plants were there and they left them standing and planted more branches from them. Nearly 80 % of the households interviewed maintain 1–15 pitaya trees in their parcels, whereas 20 % of households have more than 15 individual plants. According to our observations, sites of agroforestry systems are established in areas different from those characteristic of the wild populations, and it is possible that the populations of agroforestry systems are mostly cultivated populations.

Fruits of pitaya are commercially grown by 35 % of households interviewed, whereas the rest consume all the fruits that are grown. Most people growing pitaya fruit commercially do so within their home towns, but nearly 17 % do it also in the city of Tehuacan and at other localities. Prices of fruit vary according to their size and quality. On average, the price of one fruit is US$0·13. The main producer in Coxcatlán sells one box of large fruits for$13·85 and one box of small fruits for \$7·31.

### Morphological consequences of artificial selection

Table 2 shows that fruits of S. pruinosus were generally larger in populations cultivated in home gardens, intermediate in silvicultural populations and smaller in wild populations. Seed weight did not differ among populations but seed number was significantly higher in cultivated populations than in silvicultural populations; in wild populations the number was intermediate and not significantly different from the other two. Fruit peel was thinner and with fewer areoles in plants cultivated in home gardens as compared with wild and silvicultural fruits. Branch diameter was highest in cultivated populations and lowest in wild populations. Branch number was higher in cultivated populations than in silvicultural and wild populations. Rib depth was higher in cultivated populations and it was similar in wild and silvicultural populations. Plant biomass was similar in all populations, but individuals from wild populations were taller.

Table 2.

Average morphological characters in wild, in situ managed and cultivated populations of Stenocereus pruinosus

Character Wild Silviculture Cultivated F Significance
Fruit volume 61·418 ± 3·656c 83·130 ± 3·537b 129·751 ± 3·988a 64·674 <0·001
Fruit weight 60·748 ± 3·345c 82·432 ± 3·845b 127·366 ± 3·570a 61·694 <0·001
Pulp weight 29·451 ± 1·968c 46·098 ± 2·646b 80·460 ± 2·853a 62·574 <0·001
Seed weight 0·215 ± 0·007a 0·216 ± 0·003a 0·225 ±0·003a 2·286 0·104
Seed number 1353·33 ± 126·29ab 1235·53 ± 52·08b 1587·50 ± 55·12a 9·565 <0·001
Peel thickness 3·881 ± 0·199a 3·371 ± 0·133ab 3·219 ± 0·094b 5·069 0·007
Areoles/cm2 in fruit peel 2·036 ± 0·060a 2·079 ± 0·035a 1·743 ± 0·036 b 23·4588 <0·001
Height 4·751 ± 0·133a 4·176 ± 0·139b 4·385 ± 0·172ab 3·766 0·024
Plant size 14·328 ± 1·620 13·451 ± 1·924a 13·525 ± 1·612a 0·081 0·922
Branch number 33·6222 ± 3·6419 b 33·5952 ± 3·0403b 58·6111 ± 6·4750a 9·553 <0·001
Branch diameter 10·403 ± 0·099 c 11·119 ± 0·150b 11·823 ± 0·113a 35·011 <0·001
Spines/areole 8·029 ± 0·132b 8·933 ± 0·137a 9·185 ± 0·132a 21·061 <0·001
Rib depth 3·778 ± 0·039b 3·851 ± 0·048b 4·118 ± 0·106a 6·405 0·002
Rib width 3·783 ± 0·069a 3·834 ± 0·095a 3·694 ± 0·067a 0·851 0·428
Character Wild Silviculture Cultivated F Significance
Fruit volume 61·418 ± 3·656c 83·130 ± 3·537b 129·751 ± 3·988a 64·674 <0·001
Fruit weight 60·748 ± 3·345c 82·432 ± 3·845b 127·366 ± 3·570a 61·694 <0·001
Pulp weight 29·451 ± 1·968c 46·098 ± 2·646b 80·460 ± 2·853a 62·574 <0·001
Seed weight 0·215 ± 0·007a 0·216 ± 0·003a 0·225 ±0·003a 2·286 0·104
Seed number 1353·33 ± 126·29ab 1235·53 ± 52·08b 1587·50 ± 55·12a 9·565 <0·001
Peel thickness 3·881 ± 0·199a 3·371 ± 0·133ab 3·219 ± 0·094b 5·069 0·007
Areoles/cm2 in fruit peel 2·036 ± 0·060a 2·079 ± 0·035a 1·743 ± 0·036 b 23·4588 <0·001
Height 4·751 ± 0·133a 4·176 ± 0·139b 4·385 ± 0·172ab 3·766 0·024
Plant size 14·328 ± 1·620 13·451 ± 1·924a 13·525 ± 1·612a 0·081 0·922
Branch number 33·6222 ± 3·6419 b 33·5952 ± 3·0403b 58·6111 ± 6·4750a 9·553 <0·001
Branch diameter 10·403 ± 0·099 c 11·119 ± 0·150b 11·823 ± 0·113a 35·011 <0·001
Spines/areole 8·029 ± 0·132b 8·933 ± 0·137a 9·185 ± 0·132a 21·061 <0·001
Rib depth 3·778 ± 0·039b 3·851 ± 0·048b 4·118 ± 0·106a 6·405 0·002
Rib width 3·783 ± 0·069a 3·834 ± 0·095a 3·694 ± 0·067a 0·851 0·428

Values are given ±s.e. Significance tests according to one-way ANOVAs. Different lower-case letters indicate significant differences according to Tukey tests.

### Population genetics

#### Genetic diversity within populations

A total of 24 alleles were recorded for the four loci analysed (Table 3). Average number of alleles per locus (A) within populations was 5·42 (Table 4). All loci were polymorphic in all populations. The average observed heterozygosity (Table 4) in wild populations (HO = 0·670) was higher than in the silvicultural populations (HO = 0·622), which was in turn higher than in populations cultivated in home gardens (HO = 0·555). However, expected heterozygosity was higher in the silvicultural populations (HE = 0·743) than in the wild populations (HE = 0·726), which were in turn higher than in populations cultivated in home gardens (HE = 0·700). The population with the highest expected heterozygosity was the silvicultural population from Coxcatlán (HE = 0·751), followed by the wild population from Coxcatlán (HE = 0·732), and the cultivated populations from the home gardens of San Luis Atolotitlán (HE = 0·647). When the populations were grouped by management type, they showed differences but these were not significant according to the Kruskall–Wallis tests (for HO: H = 4·356, P = 0·11; for HE: H = 3·289, P = 0·19).

Table 3.

Allele frequencies per locus in the populations of Stenocereus pruinosus studied

Locus Allele/N Wild Santa Lucia Wild Fiscal Wild Coxcatlan Silvicultural S.L.A Silvicultural Coatepec Silvicultural Coxcatlan Cultivated S.L.A Cultivated Coatepec Cultivated Coxcatlan
Pchi21 129 0·200 0·300 0·381 0·105 0·262 0·200 0·211 0·263 0·075
131 0·300 0·150 0·143 0·132 0·214 0·400 0·342 0·158 0·325
133 0·200 0·150 0·143 0·026 0·119 0·075 0·000 0·132 0·050
135 0·050 0·100 0·048 0·447 0·024 0·025 0·237 0·105 0·425
137 0·250 0·300 0·286 0·289 0·381 0·300 0·211 0·342 0·125
Pchi54 170 0·000 0·000 0·000 0·067 0·000 0·000 0·000 0·000 0·026
172 0·000 0·000 0·000 0·100 0·000 0·000 0·000 0·000 0·026
174 0·400 0·050 0·184 0·067 0·235 0·188 0·167 0·105 0·763
176 0·333 0·800 0·500 0·433 0·765 0·313 0·222 0·763 0·079
178 0·200 0·125 0·053 0·100 0·000 0·031 0·556 0·079 0·079
180 0·000 0·000 0·263 0·033 0·000 0·188 0·000 0·053 0·026
182 0·000 0·000 0·000 0·000 0·000 0·094 0·056 0·000 0·000
184 0·067 0·025 0·000 0·200 0·000 0·188 0·000 0·000 0·000
Pchi50 223 0·250 0·300 0·286 0·167 0·205 0·225 0·175 0·175 0·079
225 0·275 0·225 0·167 0·139 0·273 0·375 0·200 0·275 0·737
227 0·150 0·125 0·167 0·056 0·114 0·100 0·000 0·075 0·000
229 0·000 0·025 0·048 0·417 0·023 0·025 0·350 0·050 0·079
231 0·325 0·325 0·333 0·222 0·386 0·275 0·275 0·425 0·105
Pchi20 249 0·025 0·300 0·190 0·050 0·000 0·050 0·000 0·025 0·000
251 0·125 0·250 0·333 0·100 0·432 0·325 0·125 0·325 0·050
257 0·000 0·000 0·000 0·025 0·000 0·050 0·000 0·000 0·025
259 0·200 0·175 0·286 0·375 0·227 0·275 0·150 0·350 0·450
261 0·650 0·275 0·190 0·425 0·273 0·275 0·725 0·300 0·475
263 0·000 0·000 0·000 0·025 0·068 0·025 0·000 0·000 0·000
Locus Allele/N Wild Santa Lucia Wild Fiscal Wild Coxcatlan Silvicultural S.L.A Silvicultural Coatepec Silvicultural Coxcatlan Cultivated S.L.A Cultivated Coatepec Cultivated Coxcatlan
Pchi21 129 0·200 0·300 0·381 0·105 0·262 0·200 0·211 0·263 0·075
131 0·300 0·150 0·143 0·132 0·214 0·400 0·342 0·158 0·325
133 0·200 0·150 0·143 0·026 0·119 0·075 0·000 0·132 0·050
135 0·050 0·100 0·048 0·447 0·024 0·025 0·237 0·105 0·425
137 0·250 0·300 0·286 0·289 0·381 0·300 0·211 0·342 0·125
Pchi54 170 0·000 0·000 0·000 0·067 0·000 0·000 0·000 0·000 0·026
172 0·000 0·000 0·000 0·100 0·000 0·000 0·000 0·000 0·026
174 0·400 0·050 0·184 0·067 0·235 0·188 0·167 0·105 0·763
176 0·333 0·800 0·500 0·433 0·765 0·313 0·222 0·763 0·079
178 0·200 0·125 0·053 0·100 0·000 0·031 0·556 0·079 0·079
180 0·000 0·000 0·263 0·033 0·000 0·188 0·000 0·053 0·026
182 0·000 0·000 0·000 0·000 0·000 0·094 0·056 0·000 0·000
184 0·067 0·025 0·000 0·200 0·000 0·188 0·000 0·000 0·000
Pchi50 223 0·250 0·300 0·286 0·167 0·205 0·225 0·175 0·175 0·079
225 0·275 0·225 0·167 0·139 0·273 0·375 0·200 0·275 0·737
227 0·150 0·125 0·167 0·056 0·114 0·100 0·000 0·075 0·000
229 0·000 0·025 0·048 0·417 0·023 0·025 0·350 0·050 0·079
231 0·325 0·325 0·333 0·222 0·386 0·275 0·275 0·425 0·105
Pchi20 249 0·025 0·300 0·190 0·050 0·000 0·050 0·000 0·025 0·000
251 0·125 0·250 0·333 0·100 0·432 0·325 0·125 0·325 0·050
257 0·000 0·000 0·000 0·025 0·000 0·050 0·000 0·000 0·025
259 0·200 0·175 0·286 0·375 0·227 0·275 0·150 0·350 0·450
261 0·650 0·275 0·190 0·425 0·273 0·275 0·725 0·300 0·475
263 0·000 0·000 0·000 0·025 0·068 0·025 0·000 0·000 0·000

S.L.A, San Luis Atolotitlán.

Table 4.

Parameters for estimating genetic variation and clone frequencies in wild, silvicultural and cultivated populations of Stenocereus pruinosus

Population N A HO HE Total number of clones Number of genotypes with clones
Wild Santa Lucia 20·000 4·250 0·692 ± 0·086 0·695 ± 0·056 4·000 2·000
Wild Fiscal 20·000 4·500 0·625 ± 0·145 0·663 ± 0·104 6·000 3·000
Wild Coxcatlán 21·000 4·500 0·701 ± 0·105 0·732 ± 0·024 4·000 2·000
Wild mean 61·000 4·750 0·670 ± 0·102 0·720 ± 0·033 14·000 7·000
Silviculture San Luis Atolotitlán 20·000 5·750 0·679 ± 0·060 0·726 ± 0·020 4·000 2·000
Silviculture Coatepec 22·000 4·000 0·508 ± 0·169 0·638 ± 0·090 5·000 2·000
Silvicultured Coxcatlán 20·000 5·500 0·691 ± 0·060 0·758 ± 0·020 2·000 1·000
Silviculture mean 62·000 6·000 0·622 ± 0·077 0·743 ± 0·019 11·000 5·000
Cultivated San Luis Atolotitlán 20·000 3·750 0·597 ± 0·088 0·646 ± 0·073 8·000 3·000
Cultivated Coatepec 20·000 4·500 0·586 ± 0·122 0·653 ± 0·083 2·000 1·000
Cultivated Coxcatlán 20·000 4·750 0·479 ± 0·187 0·538 ± 0·068 11·000 1·000
Cultivated mean 60·000 5·500 0·555 ± 0·106 0·70 ± 0·031 21·000 5·000
Total mean 58·333 5·417 0·616 ± 0·052 0·724 ± 0·015
Population N A HO HE Total number of clones Number of genotypes with clones
Wild Santa Lucia 20·000 4·250 0·692 ± 0·086 0·695 ± 0·056 4·000 2·000
Wild Fiscal 20·000 4·500 0·625 ± 0·145 0·663 ± 0·104 6·000 3·000
Wild Coxcatlán 21·000 4·500 0·701 ± 0·105 0·732 ± 0·024 4·000 2·000
Wild mean 61·000 4·750 0·670 ± 0·102 0·720 ± 0·033 14·000 7·000
Silviculture San Luis Atolotitlán 20·000 5·750 0·679 ± 0·060 0·726 ± 0·020 4·000 2·000
Silviculture Coatepec 22·000 4·000 0·508 ± 0·169 0·638 ± 0·090 5·000 2·000
Silvicultured Coxcatlán 20·000 5·500 0·691 ± 0·060 0·758 ± 0·020 2·000 1·000
Silviculture mean 62·000 6·000 0·622 ± 0·077 0·743 ± 0·019 11·000 5·000
Cultivated San Luis Atolotitlán 20·000 3·750 0·597 ± 0·088 0·646 ± 0·073 8·000 3·000
Cultivated Coatepec 20·000 4·500 0·586 ± 0·122 0·653 ± 0·083 2·000 1·000
Cultivated Coxcatlán 20·000 4·750 0·479 ± 0·187 0·538 ± 0·068 11·000 1·000
Cultivated mean 60·000 5·500 0·555 ± 0·106 0·70 ± 0·031 21·000 5·000
Total mean 58·333 5·417 0·616 ± 0·052 0·724 ± 0·015

Values are given ±s.e. where appropriate. N = sample size; A = average number of alleles per locus; HO = average observed heterozygosity (by direct counting); HE = average expected heterozygosity (unbiased) by the model of random mating, estimated from the Nei (1987) genetic diversity index.

FIS values were significantly different from zero at loci Pchi54 and Pchi50, indicating deviations from HWE. Most of the deviations were observed in managed populations (silvicultural and cultivated in home gardens), particularly at the Pchi54 locus, whereas only one wild population (from Coxcatlán) showed a significant deviation at this locus.

According to the MICRO-CHECKER analysis, the probability of the presence of null alleles was significant for loci Pchi54 and Pchi50. For Pchi54 the inferred frequency of the null alleles was 11·6 % (involving approx. 21 of the 183 individuals analysed). In contrast, only three individuals probably contained null alleles at locus Pchi50. Also, several groups of individuals with identical multilocus genotypes were found, each including from two to 11 samples (Table 4). These groups were more numerous in populations cultivated in home gardens than in wild and silvicultural populations. The probability that these groups of individuals with identical genotypes arose through sexual reproduction (Psex) was in all cases very low (P < 0·05).

#### Genetic distances

The UPGMA analysis clustered the nine populations into two groups, separating the cultivated population of home gardens from Coxcatlán, which is the most differentiated, from all other populations. The second step recognized a group formed by both the cultivated and the silvicultural populations of San Luis Atolotitlán, separated from another group formed by the remaining populations. Within this latter cluster, the wild population from Fiscal (Coatepec), the silvicultural population from Coatepec and the cultivated population in home gardens of Coatepec grouped together along with the wild population from Coxcatlán. The wild population from Santa Lucia (Coatepec) formed another group with the silvicultural population from Coxcatlán. Bootstrap analysis provided support values above 50 % for most groups (Fig. 4). The Mantel test indicated that there was no significant correlation between genetic and geographical distances (r = 0·165, P = 0·186).

Fig. 4.

UPGMA phenogram of minimum genetic distances (Nei, 1972) among the wild silviculture managed and cultivated populations of Stenocereus pruinosus studied. Proportions of similar replicates after 10 000 permutations are shown for each grouping.

Fig. 4.

UPGMA phenogram of minimum genetic distances (Nei, 1972) among the wild silviculture managed and cultivated populations of Stenocereus pruinosus studied. Proportions of similar replicates after 10 000 permutations are shown for each grouping.

### Genetic differentiation among populations

The AMOVA indicated that there was no variation due to differences between groups (Table 5), while 10·41 % of the genetic variation was due to differences among populations within groups, and almost 91 % was due to within-population variation. Total differentiation among populations (ΦST) was moderate (Wright, 1978) but significant (ΦST = 0·094, P < 0·001).

Table 5.

Molecular analysis of variance (AMOVA) comparing genetic distance between groups of populations (wild, silviculture and cultivated), populations within groups and individual plants within populations

Source of variation d.f. Sum of squares Components of variance Percentage of variation Φ Statistics P
Among groups 10·754 0·000 Va 0·00 ΦCT = –0·009 n.s.
Among populations, within groups 42·142 0·142 Vb 10·41 ΦSC = 0·103 <0·0001
Within populations 357 442·181 1·239 Vc 90·58 ΦST = 0·094 <0·0001
Total 365 495·077 1·367
Source of variation d.f. Sum of squares Components of variance Percentage of variation Φ Statistics P
Among groups 10·754 0·000 Va 0·00 ΦCT = –0·009 n.s.
Among populations, within groups 42·142 0·142 Vb 10·41 ΦSC = 0·103 <0·0001
Within populations 357 442·181 1·239 Vc 90·58 ΦST = 0·094 <0·0001
Total 365 495·077 1·367

Pairwise population differentiation under the IAM and SMM mutation models revealed similar patterns, with values between 0 and 0·279. The strongest differentiation was between the cultivated population in home gardens of Coxcatlán and the wild population from Fiscal (Coatepec) (FST = 0·263, RST = 0·275; P < 0·05), and between the cultivated population from Coxcatlán and the wild population from Coxcatlán (FST = 0·212, RST = 0·279; P < 0·05). The most genetically similar populations were those cultivated in home gardens of Coatepec and the silvicultural population from Coatepec (FST = 0, RST = 0; P > 0·05).

Gene flow was high among all populations, except among the cultivated population in home gardens of Coxcatlán and the other populations. Highest gene flow was detected between the silvicultural population from Coatepec and the cultivated population in home gardens of Coatepec (Nm > 100), the cultivated population in home gardens of Coatepec and the wild population from Fiscal (also from Coatepec) (Nm = 93·109), as well as between the silvicultural population of Coatepec and the wild population from Coxcatlán (Nm = 48·488), and between the silvicultural Coatepec and the wild Fiscal (from Coatepec) populations (Nm = 36·748). The lowest number of migrants per generation was identified between the cultivated population in home gardens of Coxcatlán and the wild population from Fiscal (Nm = 1·402) and the silvicultural population from Coatepec (Nm = 1·788).

### Patterns of genetic structure

The ΔK statistic (Fig. 5) revealed K = 2 to be the optimum value for the number of genetic clusters in the data. However, this analysis also indicated K = 5 as the second most probable number of genetic clusters (Fig. 5). This second result provides information about the substructure within the two groups previously mentioned, and probably better reflects the groups of individuals with identical multilocus genotypes already identified (see above). Figure 6 shows the proportion of ancestry of each population and individual on these five genetic clusters, represented by the green, yellow, blue, red and pink colours, while Fig. 7 depicts the population-level proportions of ancestry as pie charts displayed on a topographic map of the study region. This analysis clearly showed that the population cultivated in home gardens of Coxcatlán is the most divergent in terms of its genetic ancestry. The population cultivated in home gardens and the silvicultural population from San Luis Atolotitlán are similar among themselves and differentiated from the remaining six populations, which show similar proportions of ancestry on each of the five genetic clusters.

Fig. 5.

Estimated number of populations (K) derived from the structure clustering analyses. The magnitude of ΔK was calculated using the method described by Evanno et al. (2005).

Fig. 5.

Estimated number of populations (K) derived from the structure clustering analyses. The magnitude of ΔK was calculated using the method described by Evanno et al. (2005).

Fig. 6.

Genetic clusters obtained with five groups. Each individual plant is represented by one vertical line with K segments coloured proportionally according to their belonging to a genetic cluster. Black lines separate individual plants from different populations, classified according to management type. WI = Wild Santa Lucía; WII = Wild Fiscal; WIII = Wild Coxcatlán; SI = Silvicultural from San Luis Atolotitlán (S.L.A); SII = Silvicultural Coatepec; SIII = Silvicultural Coxcatlán; CI = Cultivated from San Luis Atolotitlán (S.L.A); CII = Cultivated Coatepec; CIII = Cultivated Coxcatlán.

Fig. 6.

Genetic clusters obtained with five groups. Each individual plant is represented by one vertical line with K segments coloured proportionally according to their belonging to a genetic cluster. Black lines separate individual plants from different populations, classified according to management type. WI = Wild Santa Lucía; WII = Wild Fiscal; WIII = Wild Coxcatlán; SI = Silvicultural from San Luis Atolotitlán (S.L.A); SII = Silvicultural Coatepec; SIII = Silvicultural Coxcatlán; CI = Cultivated from San Luis Atolotitlán (S.L.A); CII = Cultivated Coatepec; CIII = Cultivated Coxcatlán.

Fig. 7.

Distribution of populations of Stenocereus pruinosus in the Tehuacán Valley indicating barriers between populations (A) and the frequency distribution of genotypes obtained by Bayesian analysis among populations associated with barriers (B). WI = Wild Santa Lucía; WII = Wild Fiscal; WIII = Wild Coxcatlán; SI = Silvicultural from San Luis Atolotitlán (S.L.A); SII = Silvicultural Coatepec; SIII = Silvicultural Coxcatlán; CI = Cultivated from San Luis Atolotitlán (S.L.A); CII = Cultivated Coatepec; CIII = Cultivated Coxcatlán.

Fig. 7.

Distribution of populations of Stenocereus pruinosus in the Tehuacán Valley indicating barriers between populations (A) and the frequency distribution of genotypes obtained by Bayesian analysis among populations associated with barriers (B). WI = Wild Santa Lucía; WII = Wild Fiscal; WIII = Wild Coxcatlán; SI = Silvicultural from San Luis Atolotitlán (S.L.A); SII = Silvicultural Coatepec; SIII = Silvicultural Coxcatlán; CI = Cultivated from San Luis Atolotitlán (S.L.A); CII = Cultivated Coatepec; CIII = Cultivated Coxcatlán.

The Monmonier's maximum difference algorithm, applied on the matrix of linearized FST values, placed the first genetic barrier between the population cultivated in home gardens of Coxcatlán and the remaining populations (Fig. 7). The second barrier separated the population cultivated in home gardens of San Luis Atolotitlán and the silvicultural population of San Luis Atolotitlán. The third significant barrier was identified between the group of silvicultural and wild populations from Coxcatlán and the wild population from Santa Lucia (Coatepec) on one side and the cultivated and silvicultural populations from Coatepec and the wild population from Fiscal (also from Coatepec) on the other side. This analysis was largely congruent with the results from STRUCTURE and with the UPGMA dendrogram based on genetic distances.

## DISCUSSION

### Management and artificial selection

Our interviews with local people provided evidence that artificial selection of S. pruinosus as traditionally practised is on the decline but that it is being adjusted to meet the demands of new markets. Artificial selection operates on S. pruinosus in both home gardens and agroforestry systems, where branches are planted from wild and silvicultural populations, but mainly from other home gardens. People recognize variation in the quality of fruits produced by individual plants and nearly half of the people interviewed said they selected the variations desirable to cut branches for planting. The cultivation of S. pruinosus in home gardens appears to be losing importance as the majority of people perceive that cultivation and use of this species was more important in the past, and nearly half of those interviewed affirmed that they do not to practise artificial selection; however, our ethnobotanical survey also reveals signs that in fact selection is still occurring, and the conditions exist to potentially enhance it. For instance, almost all people consume the fruit and have cultivated the plant. The new economic context of this species drives management practices and the persistence of traditional artificial selection that maintains the processes of domestication. Therefore, conservation of these genetic resources should consider the promotion of their economic and human cultural importance. The markets probably play dual role, as on the one hand they have determined the introduction of new products and customs motivating substitution of local resources. On the other hand, they have created new opportunities for increasing the economic importance of local resources. Markets mainly enhance large size and other characters associated with the duration and transportation of fruit (peel thickness and thorniness), although other characteristics (pulp colour, flavour and texture) have decreased in value. The outcome of such a double role of the markets will depend on local decisions as well as on the intervention of external institutions that facilitate processes to promote local resources.

Artificial selection is conducted in both in situ and ex situ management forms. Of particular importance is that silvicultural management in agroforestry systems involves artificial selection as it supports the idea that domestication is occurring not only through ex situ cultivation but also under in situ management of wild populations (Casas et al., 2007). In the particular case studied here, people combine the traditional practice of leaving standing preferred phenotypes in areas where vegetation is cleared in addition to cultivation of branches from the local and other populations, mainly from home gardens. This combination of techniques is more significant in S. pruinosus than in any other of the species studied in the region and contributes to explain the morphological and genetic patterns of divergence and gene flow.

### Morphological consequences of artificial selection

Artificial selection is directed at fruit characters, but our study documented that it also significantly influences characters of both vegetative and reproductive plant parts. The comparison of morphological characters among populations within the three management types indicated significant differences mainly among the wild and cultivated populations in home gardens. Morphological divergence is clear in characters such as fruit size, seed number, and thorniness, but our analysis also reveals that plants from wild populations are taller than those from cultivated and silvicultural populations, and that the latter are more robust than the wild plants. Greater height in wild individual plants could be the result of competition for light in dense woods due to the taller vegetation cover existing in those areas, which could enhance increased length of branches, whereas the more robust and branched form of cultivated plants could be an expression of the higher humidity existing in cultivated environments or the pruning of branches practised by people. These morphological patterns are similar to those reported for other columnar cacti species studied by Casas et al. (1999a), Luna-Morales et al. (2001), Cruz and Casas (2002), Arellano and Casas (2003), Carmona and Casas (2005), and Blancas et al. (2009).

Morphological differentiation between wild and managed populations is explainable by artificial selection. Plants producing larger fruits and with other favourable characters are more abundant in managed populations than in wild populations and, therefore, mean values among populations vary. This is clearly illustrated by the population cultivated in home gardens of Coxcatlán, which is the most intensely managed, as was documented by ethnobotanical interviews. Selection of quality characters is crucial to produce fruits with high economic value for commercialization in important markets of the Tehuacán Valley, and in this population pitaya plants produce fruits of the highest quality.

### Population genetic variation

Previous studies on domestication of Stenocereus spp. in the Tehuacán Valley revealed that traditional management is efficient in maintaining and in some cases increasing genetic diversity in managed populations. For instance, analysing isozyme variation of the same populations of S. pruinosus, Parra et al. (2008) found higher values of genetic diversity in cultivated populations. In the present study the highest variation was found in silvicultural populations, but in this and in the previous studies differences among managed and unmanaged populations were not significant and therefore only the trends were identified.

Although in home gardens a high proportion of individuals are clones, the high diversity can be explained because branches are introduced from other home gardens of the same or even different villages, as well as from silvicultural and wild populations. In silvicultural populations, high genetic diversity can be explained because they are originally wild populations and in the agroforestry systems there are conditions for the establishment of natural seedlings with the consequent potential for increasing genetic diversity via gene flow in seeds or pollen. In addition, branches are introduced from home gardens and from wild and other silvicultural populations.

Although levels of genetic diversity are high, a deficiency of heteozygotes was identified at some loci. The positive values of FIS in Pchi54 and Pchi50 indicate potential levels of inbreeding that are inconsistent with the documented outcrossing systems characteristic of all columnar cacti so far studied (Mandujano et al., 2009). This could be due to genotyping errors as microsatellites used in this study are not specific for S. pruinosus and null alleles could be influencing deviations in FIS (Van Oosterhout et al., 2004), but most probably it can also be explained because, as mentioned, clonal propagation is high in cultivated stands (see Table 4).

### Genetic differentiation and population structure

The genetic relationships among populations visualized in the UPGMA dendrogram (Fig. 4) were not completely consistent with those reported through isozyme analysis (Parra et al., 2008), in which clustering was more clearly according to management type. Microsatellites revealed a pattern reflecting both management type and the origin of plants used for planting. Both isozymes and microsatellites were consistent in identifying the population of cultivated plants from home gardens of Coxcatlán as the most differentiated, as well as in recognizing the affinities between the wild populations from Fiscal (Coatepec) and the wild populations from Coxcatlán (Parra et al., 2008). Microsatellites suggested that the wild population from Santa Lucía is more closely related to the silvicultural population from Coxcatlán. In addition, the cultivated and silvicultural populations from San Luis Atolotitlán, and the cultivated and silvicultural populations from Coatepec clustered together probably due to the exchange of branches between populations within each of the two pairs. Overall, the pattern suggests that the genetic similarity among populations can be explained in part by the system of management practices, rather than simply from distances as reported for other columnar cacti species under lower management intensity than S. pruinosus, such as E. chiotilla, P. chichipe, and S. stellatus (Otero-Anaiz et al., 2005; Tinoco et al., 2005; Casas et al., 2006).

Bayesian analysis provided further insight into the genetic structure among and within populations (Fig. 7B). Cultivated and silvicultural populations from San Luis Atolotitlán clearly have important proportions of their ancestry from the same genetic cluster. This is concordant with the possible origin of individuals in the silvicultural population from home gardens of San Luis, as was found in the ethnobotany survey. The silvicultural and cultivated populations from Coatepec are also similar, with almost the same proportions of genetic groups. Moreover, both populations might be related to the wild population from Fiscal (Coatepec), which probably was the source of the majority of individuals. The cultivated population in home gardens of Coxcatlán is distinct from all other populations because of the higher proportion of the genetic group represented in blue, and because it is much more homogeneous. Eleven individuals in this population had an identical multilocus genotype, suggesting that they were clonally derived from the same plant. According to the ethnobotanical survey most of the branches planted came from already standing individuals at this site, and some material came from villages not included in the present analysis, which might also be contributing to the distinctness of the population.

In addition to management practices, another important factor influencing genetic structure of the populations is probably topography, as was revealed by the barriers analysis (Fig. 7). As expected, the strongest subdivision separated the population cultivated in home gardens of Coxcatlán from the remainder, but the geographical location of the barrier coincided with the mountain chain ‘El Laurel’, which reaches 2500 m and probably acts as a barrier to the movement of pollinators and dispersers, even when these are humans. In contrast, the silvicultural and wild populations from Coxcatlán, located near the cultivated population in home gardens of Coxcatlán, form a group with the geographically more distant wild population from Santa Lucía (Coatepec). In this case, the valley of the Calapa River could be a natural corridor for seed and pollen dispersal. Continuous natural gene flow among those distant wild populations has important implications for management, because it would increase the variability of the propagule sources for human management.

### Traditional management and conservation of genetic diversity of S. pruinosus

Traditional management favours variation and promotes conservation of genetic variation. It is significant that, according to our ethnobotanical survey, people differentially use and manage variation. It is part of the prestige of farmers to have in their agroforestry systems and home gardens variants for either direct consumption or commercialization. It is also significant that nearly 11 % of households introduce to their home gardens branches from wild plants as this practice helps to enforce some of the processes that maintain the high levels of genetic diversity and gene flow between wild and cultivated populations that were detected by our population genetics studies. Enrichment of germplasm variation in home gardens has been rarely documented. Among other studies providing information in this respect are those by Scarcelli et al. (2006) and Tostain et al. (2007) with the yam Dioscorea rotundata cultivated by traditional farmers in Benin. The traditional farmers there enrich their cultivated stock of variation with wild materials, increasing genetic variation. This is also the case for Spondias purpurea, documented by Miller and Schaal (2006) among the Maya of Yucatán, Mexico, as well for Stenocereus stellatus documented in central Mexico by Casas et al. (2006).

Overall indices of genetic variation in populations of S. pruinosus indicate that genetic diversity for this species in the Tehuacán Valley is markedly high (HE = 0·724). These indices are congruent with those found through isozyme analysis (Parra et al., 2008), but the microsatellite data revealing higher genetic diversity, as expected with this technique. The only other case of columnar cacti studied through microsatellites is Polaskia chichipe, endemic to the Tehuacán Valley. Otero-Arnaiz et al. (2005) calculated values of HE of 0·507 in in situ managed populations, 0·560 in populations cultivated in home gardens and 0·631 in wild populations. P. chichipe is managed at lower intensity levels than S. pruinosus.

Similar levels of genetic variation have been reported for other domesticated species under traditional management studied through microsatellite markers. For instance, in Manihot esculenta, Elias et al. (2001) reported HE = 0·592 among individuals managed by indigenous people of Makushi, while Pujol et al. (2005) found HE = 0·52 among the Palikur in French Guyana. For Dioscorea rotundata, Tostain et al. (2007) reported HE = 0·57 in Benin. Moreover, Jarvis et al. (2008) after 10 years of studying 27 cultivated species from all continents concluded that traditional management maintains high levels of genetic diversity, particularly in species under clonal propagation. These high levels of genetic diversity demonstrate an unusual pattern, in which managed plants are maintaining important proportions of variability, contradicting the general assumption that managed and cultivated populations present a drastic reduction in genetic variation as a consequence of artificial selection (Hawkes, 1983; Doebley, 1992).

## CONCLUSIONS

The high levels of genetic variation in both agroforestry systems and home gardens documented here seem to be related to the multiple sources of plant material propagated in these areas, the multiple phenotypes attractive to people, and processes that promote gene flow among wild and managed populations through pollination determined by bats, seed dispersal carried out by birds and bats, and episodic recruitment of seedlings, as well as long-distance transportation of branches mediated by humans. It is probable that historical events of seedling establishment have contributed to the genetic diversity that is currently observed in home gardens. Although we did not observe seedlings of S. pruinosus in home gardens and agroforestry systems, there are records of them for Stenocereus stellatus in the Mixteca Baja, where seedlings are tolerated and protected in human made environments (Casas et al., 1997, 1999a). In fact, it could be more common in regions with higher humidity and better conditions for establishment of S. pruinosus seedlings, as observed in localities from southern Oaxaca, where people tolerate and care for seedlings of this species as part of local traditional practices.

Seedling recruitment is an important issue that could have a more important role in the evolution of this species as well as in others propagated vegetatively (Olsen and Schaal, 2007). This still requires much study, but current information confirms that traditional management: (1) determines artificial selection not only through cultivation in home gardens but also through in situ management in agroforestry systems; (2) determines significant phenotypic divergence and a tendency to genetic differentiation between wild and managed populations; and (3) is able to maintain high levels of morphological and genetic diversity, which is crucial for strategies of conservation of genetic resources of this and other native species of Mesoamerica.

## ACKNOWLEDGEMENTS

We thank the Posgrado en Ciencias Biológicas of the National University of Mexico (UNAM) and the National Council of Science and Technology (CONACYT), Mexico for academic and financial support for PhD studies of the first author. The Dirección General de Asuntos del Personal Académico (research project IN219608) and CONACYT (project 103551) and the Royal Botanic Garden, UK provided financial support for field and laboratory work. Edgar Pérez-Negrón gave valuable support in the field, and Heberto Ferreira and Alberto Valencia kindly provided assistance with computing.

## LITERATURE CITED

Arellano
E
Casas
A
Morphological variation and domestication of Escontria chiotilla (Cactaceae) under silvicultural management in the Tehuacán Valley, Central Mexico
Genetic Resources and Crop Evolution
,
2003
, vol.
50
(pg.
439
-
453
)
Arias
S
Gama
S
Guzmán
L
1997

Flora del Valle de Tehuacan. Fascículo 14. Cactaceae A. L. Juss. Instituto de Biología, UNAM.
Arias-Cóyotl
E
Stoner
K
Casas
A
Effectiveness of bats as pollinators of Stenocereus stellatus (Cactaceae) in wild, managed in situ, and cultivated populations in La Mixteca, Baja, central Mexico
American Journal of Botany
,
2006
, vol.
93
(pg.
1675
-
1683
)
Blancas
J
Casas
A
Lira
R
Caballero
J
Traditional management and morphological patterns of Myrtillocactus schenckii in the de Tehuacán Valley, Central México
Economic Botany
,
2009
, vol.
63
(pg.
375
-
387
)
Carmona
A
Casas
A
Management, domestication and phenotypic patterns of Polaskia chichipe (Cactaceae) in the Tehuacán Valley, Central Mexico
Journal of Arid Environments
,
2005
, vol.
60
(pg.
115
-
132
)
Casas
A
Barbera
G
Mesoamerican domestication and diffusion of cacti
Cacti: biology and uses. Los Angeles: University of California Press
,
2002
In
Nobel PS
(pg.
143
-
162
)
Casas
A
Parra
F
LEISA Revista de Agroecología
,
2007

Septiembre 2007
Casas
A
Pickersgill
B
Caballero
J
Valiente-Banuet
A
Ethnobotany and domestication in xoconochtli, Stenocereus stellatus (Cactaceae), in the Tehuacán Valley and la Mixteca Baja, México.
Economic Botany
,
1997
, vol.
51
(pg.
279
-
292
)
Casas
A
Caballero
J
Valiente-Banuet
A
Soriano
JA
Dávila
P
Morphological variation and the process of domestication of Stenocereus stellatus (Cactaceae) in Central Mexico
American Journal of Botany
,
1999
, vol.
86
(pg.
522
-
533
)
Casas
A
Valiente-Banuet
A
Rojas-Martínez
A
Dávila
P
Reproductive biology and the process of domestication of the columnar cactus Stenocereus stellatus in Central Mexico
American Journal of Botany
,
1999
, vol.
86
(pg.
534
-
542
)
Casas
A
Cruse
J
Morales
E
Otero-Arnaiz
A
Valiente-Banuet
A
Maintenance of phenotypic and genotypic diversity of Stenocereus stellatus (Cactaceae) by indigenous peoples in Central Mexico
Biodiversity and Conservation
,
2006
, vol.
15
(pg.
879
-
898
)
Casas
A
Otero-Arnaiz
A
Pérez-Negrón
E
Valiente-Banuet
A
In situ management and domestication of plants in Mesoamerica
Annals of Botany
,
2007
, vol.
100
(pg.
1101
-
115
)
Cortés-Díaz
V
Biología de la polinización de Stenocereus pruinosus en el Valle de Tehuacán-Cuicatlán, México
,
1996

BSc thesis, Escuela Nacional de Estudios Profesionales, Iztacala, México, Universidad Nacional Autónoma de México, México
Cruz
M
Casas
A
Reproductive biology and morphological variation of Polaskia chende (Cactaceae) under domestication in Central Mexico
Journal of Arid Environments
,
2002
, vol.
51
(pg.
561
-
576
)
Darwin
C
The origins of species by means in natural selection or the preservation of favoured races in the struggle for life
,
1859
London
John Murray
Doebley
J
Soltis
PS
Soltis
D
Doyle
JJ
Molecular systematics and crop evolution
Molecular systematics of plants
,
1992
London
Chapman and Hall
(pg.
202
-
222
)
Di Rienzo
A
Peterson
AC
Garza
JC
, et al.  .
Mutational processes at simple-sequence repeat loci in human populations
Proceedings of the National Academy of Sciences of the United States of America
,
1994
, vol.
91
(pg.
3166
-
3170
)
Elias
M
Penet
L
Vindry
P
Mckey
D
Panaud
O
Robert
T
Unmanaged sexual reproduction and the dynamics of genetic diversity of a vegetatively propagated crop plant, cassava (Manihot esculenta Crantz), in a traditional farming system
Molecular Ecology
,
2001
, vol.
10
(pg.
1895
-
1907
)
Evanno
G
Regnaut
S
Goudet
J
Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study
Molecular Ecology
,
2005
, vol.
14
(pg.
2611
-
2620
)
Excoffier
L
Laval
G
Schneider
S
Arlequin (version 3·0): an integrated software package for population genetics data analysis
Evolutionary Bioinformatics
,
2005
, vol.
1
(pg.
47
-
50
)
Falush
D
Stephens
M
Pritchard
JK
Inference of population structure using multilocus genotype data: linked loci and correlated allele frequencies
Genetics
,
2003
, vol.
164
(pg.
1567
-
1587
)
Guo
SW
Thompson
EA
Performing the exact test of Hardy-Weinberg proportions for multiple alleles
Biometrics
,
1992
, vol.
48
(pg.
361
-
372
)
Harlan
J
Crops and man.
,
1975
American Society of Agronomy
Hawkes
JG
The diversity of crop plants
,
1983
London
Harvard University Press
Jarvis
D
Brown
H
Hung-Cuong
P
, et al.  .
A global perspective of the richness and evenness of traditional crop-variety diversity maintained by farming communities
Proceedings of the National Academy of Sciences USA
,
2008
, vol.
105
(pg.
5326
-
5331
)
Luna-Morales
C
Aguirre
JR
Variación morfológica del fruto y domesticación de Stenocereus pruinosus (Otto) Buxb. y S. stellatus (Pfeiff.) Riccob. (Cactaceae) en la Mixteca Baja, México
Revista Fitotecnia mexicana
,
2001
, vol.
24
(pg.
213
-
221
)
Luna-Morales
C
Aguirre
R
Peña
C
Cultivares tradicionales mixtecos de Stenocereus pruinosus y S.stellatus (Cactaceae)
Anales del Instituto de Biología. Serie Botánica
,
2001
, vol.
72
(pg.
131
-
155
)
Mandujano
MC
Carrillo-Angeles
I
Martínez-Peralta
C
Golubov
J
Ramawat
KG
Reproductive biology of Cactaceae
Desert plants. Biology and biotechnology. Dordrecht: Springer
,
2009
(pg.
197
-
230
)
Manni
F
Guérard
E
Heyer
E
Geographic patterns of (genetic, morphologic, linguistic) variation: how barriers can be detected by using Monmonier's Algorithm
Human Biology
,
2004
, vol.
72
(pg.
173
-
190
)
Martin
G
Ethnobotany. Principles and practices.
,
1997
London
UICN Plants & Peoples
Miller
A
Schaal
B
Domestication and the distribution of genetic variation in wild and cultivated populations of the Mesoamerican fruit tree Spondias purpurea L. (Anacardiaceae)
Molecular Ecology
,
2006
, vol.
15
(pg.
1467
-
1480
)
Miller
MP
1997

Tools for population genetics analyses (TFPGA) 1·3. A windows program for the analysis of allozymes and molecular population genetic data. Computer software distributed by author.
Nei
M
Genetic distance between populations
American Naturalist
,
1972
, vol.
106
(pg.
283
-
292
)
Nei
M
Molecular evolutionary genetics
,
1987
New York
Columbia University Press
Olsen
K
Schaal
B
Insights on the evolution of a vegetatively propagated crop species
Molecular Ecology
,
2007
, vol.
16
(pg.
2838
-
2840
)
Otero-Arnaiz
A
Cruse-Sanders
J
Casas
A
Hamrick
JL
Isolation and characterization of microsatellites in the columnar cactus: Polaskia chichipe and cross species amplification within the Tribe Pachycereeae (Cactaceae).
Molecular Ecology Notes
,
2004
, vol.
4
(pg.
265
-
269
)
Otero-Arnaiz
A
Casas
A
Hamrick
JL
Cruse-Sanders
J
Genetic variation and evolution of Polaskia chichipe (Cactaceae) under domestication in the Tehuacán Valley,central Mexico
Molecular Ecology
,
2005
, vol.
14
(pg.
1603
-
1611
)
Parra
F
Pérez-Nasser
N
Lira
R
Pérez-Salicrup
D
Casas
A
Population genetics and process of domestication of Stenocereus pruinosus (Cactaceae) in the Tehuacán Valley, México
Journal of Arid Enviroments
,
2008
, vol.
72
(pg.
1997
-
2010
)
Peakall
R
Smouse
PE
GENALEX 6, genetic analysis in Excel: population genetic software for teaching and research
Molecular Ecology Notes
,
2006
, vol.
6
(pg.
288
-
295
)
Pritchard
JK
Stephens
M
Donnelly
P
Inference of population structure using multilocus genotype data
Genetics
,
2000
, vol.
155
(pg.
945
-
959
)
Pujol
B
David
P
McKey
D
Microevolution in agricultural environments: how a traditional Amerindian farming practice favors heterozygosity in cassava (Manihot esculenta Crantz, Euphorbiaceae)
Ecology Letters
,
2005
, vol.
8
(pg.
138
-
147
)
Raymond
M
Rousset
F
GENEPOP (Version 1·2): population genetics software for exact tests and ecumenicism
Journal of Heredity
,
1995
, vol.
86
(pg.
248
-
249
)
Reid
WV
Miller
KR
Keeping options alive: the scientific bases for the conservation of biodiversity. World Resources Institute
,
1989
Scarcelli
N
Tostain
S
Vigouroux
Y
Agbangla
C
Dainou
O
Pham
JL
Farmers' use of wild relative and sexual reproduction in a vegetatively propagated crop. The case of yam in Benin
Molecular Ecology
,
2006
, vol.
15
(pg.
2421
-
2431
)
Slatkin
M
A measure of population subdivision based in microsatellite allele frequencies
Genetics
,
1995
, vol.
139
(pg.
157
-
162
)
Smith
E
Byers
DS
Plants remains. In:
The prehistory of the Tehuacán Valley. Volume One: environment and aubsistence.
,
1967
Austin:
University of Texas Press
(pg.
230
-
260
)
Stenberg
P
Lundmark
M
Saura
A
MLGsim: a program for detecting clones using a simulation approach
Molecular Ecology Notes
,
2003
, vol.
3
(pg.
329
-
331
)
Tinoco
A
Casas
A
Luna
R
Oyama
K
Population genetics of Escontria chiotilla in wild and silvicultural managed populations in the Tehuacán Valley, Central México
Genetic Resources and Crop Evolution
,
2005
, vol.
52
(pg.
525
-
538
)
Tostain
S
Agbangla
C
Scarcelli
N
, et al.  .
Genetic diversity analysis of yam cultivars (Dioscorea rotundata Poir.) in Benin using simple sequence repeat (SSR) markers
Plant Genetic Resources
,
2007
, vol.
5
(pg.
71
-
81
)
Van Oosterhout
C
Hutchinson
WF
Wills
DP
Shipley
P
MICRO-CHECKER: software for identifying and correcting genotyping errors in microsatellite data. Program Note
Molecular Ecology Notes
,
2004
, vol.
4
(pg.
535
-
538
)
Vavilov
NI
The origin, variation, immunity and breeding of cultivated plants
Chronica Botanica
,
1951
, vol.
13
(pg.
1
-
366
)
Wright
S
The interpretation of population structure by F-statistics with especial regard to systems of mating
Evolution
,
1965
, vol.
19
(pg.
395
-
420
)
Wright
S
Evolution and the genetics of populations, Vol. 4: Variability within and among natural populations
,
1978
Chicago
University of Chicago Press