Genetic structure of coexisting wild and managed agave populations: implications for the evolution of plants under domestication

Agave inaequidens is used to produce mescal, and a management gradient from gathered wild, silvicultural and cultivated plantations has been documented. We expected different levels of genetic diversity and structure associated with management. Through 10 nuclear microsatellite loci we compared population genetics parameters and found that A. inaequidens exhibits high levels of genetic diversity (He=0.707) and moderate genetic structure (FST=0.112) with no differences among wild and managed populations. Bayesian analysis indicated that genetic clusters best fit with the corresponding habitats where populations grow. Natural mechanisms of gene flow and movement of agave propagules among populations by people explain these patterns.


Introduction
Domestication is a gradual and continuous process through which plants undergo phenotypic and genetic changes, mainly resulting from artificial selection favouring organisms with features desirable, necessary or interesting to humans (Darwin 1859;Casas et al. 2007;Pickersgill 2007). Since Darwin's time (Darwin 1859(Darwin , 1868, domestication has been a useful model for studying a variety of questions in evolutionary biology (Gepts 2004;McKey et al. 2012). Understanding the relationship between domesticated and wild individuals provides valuable opportunities for studying evolutionary processes in nature, especially in areas where the ancestor and descending organisms coexist and are available for comparison (Olsen and Wendel 2013). The most common pattern of domestication involves a reduction of genetic variation in domesticated populations compared with wild ones, mainly due to artificial selection (Sauer 1972;Doebley 1989;Olsen and Wendel 2013). However, some studies in plants have documented that such a trend may vary according to the life history of the plant species analysed, as well as the context of human cultures involved in use and selection of those species (Casas et al. 2006(Casas et al. , 2007Vargas-Ponce et al. 2009;Parra et al. 2010;Aguirre-Dugua et al. 2013).
Mesoamerica is a main centre of diversity of useful plant species (Caballero et al. 1998). It is also the centre of domestication of nearly 200 economically important plant species worldwide, such as maize, beans, chilli peppers, cocoa, cotton, prickly pears (Vavilov 1951;Harlan 1975) and among them several species of agaves (Gentry 1982;Colunga-GarcíaMarín et al. 2007). In addition, numerous native plant species (.800) are at various stages of domestication in traditional farming systems, such as home gardens and other agroforestry systems, most of them still unstudied (Colunga-GarcíaMarín and Zizumbo-Villarreal 1993;Casas et al. 1997Casas et al. , 2007Caballero et al. 1998;Blancas et al. 2010). In these systems, it is possible to find coexisting wild, semi-domesticated or incipiently domesticated, and fully domesticated plants belonging to a single species (Casas et al. 2007). In Mesoamerica, people practice in situ management of wild plants in forests, forest patches and agroecosystems, which includes several types of interaction: letting stand, encouraging growth and special care and protection of more favourable plants. These interactions in some cases involve artificial selection, resulting in documentable domestication processes (Colunga-GarcíaMarín and Zizumbo-Villarreal 1993;Casas et al. 1996Casas et al. , 2007. Agaves belong to the family Asparagaceae, which comprises 9 genera and 300 species (Eguiarte et al. 2000;Rocha et al. 2006). The genus Agave is relatively recent, about 10 million years (my) old, with .160 species, 75 % of them occurring in Mexico, where it has shown a high adaptive radiation in diverse ecosystems (Eguiarte et al. 2000). Among the most meaningful adaptations of the genus, several authors (Rocha et al. 2006;Eguiarte and Souza 2007) have highlighted the morphological and physiological defences against predators, and substantial diversity of reproductive mechanisms and pollinators. Agaves have high chromosome number and large genome size, nearly 64.3 % of species being polyploid (Goldblatt 1980;Palomino et al. 2007), showing different ploidy levels. For instance, for Agave cupreata Trel. et Berg., Palomino et al. (2012) reported diploids, tetraploids, pentaploids and hexaploids. For A. tequilana Weber, Bennett et al. (2000), Robert et al. (2008) and Simpson et al. (2011) recorded that genome size varies from 2940 to 77 458 million base pairs. In a review considering 22 species of Agave, Eguiarte et al. (2013) estimated that, on average, genetic diversity of this genus is high and genetic differentiation is moderate compared with other outcrossing monocots (Hamrick and Godt 1996).
About 102 taxa of agaves have been reported as being used in Mexico (Colunga-GarcíaMarín et al. 2007;Torres et al. 2015a), most notably as human food. Before maize was adopted as the main Mesoamerican crop, agaves were a principal source of carbohydrates for peoples of western Mexico and an area that is now the southwest USA, who consumed the stems, leaf bases and flower stalk cooked in stone ovens (Callen 1965;Smith 1986;Hodgson 2001;Zizumbo-Villarreal et al. 2013).
At present, flower buds, flower stalk and leaf bases are edible, whereas the sap, called aguamiel, is collected for drinking fresh or for preparing the fermented drink called pulque. In addition, various distilled beverages (generically called mescal, including the appellations of origin Tequila, Bacanora and Raicilla) are prepared with fermented cooked stems. Other important uses are as a source of fibre, and as fodder. The activity that has gained the highest economic importance in recent decades is the production of mescal, spirit drinks produced from the central corm of the plant. For the production of this beverage, the use of 53 taxa has been recorded in Mexico (Colunga-GarcíaMarín et al. 2007;Torres et al. 2015a), most of them extracted from forests. The selection and harvesting of agave corms are carried out just before sexual reproduction takes place, which poses a high risk for numerous populations of several species (Delgado-Lemus et al. 2014).
Only four Mexican Agave taxa and two of the USA referred to as crops have been studied in terms of the consequences of divergence in genetic diversity caused by domestication and management practices. Among the species studied are A. fourcroydes Lem., which was domesticated by the Pre-Columbian Maya of the Yucatá n Peninsula for fibre, and their putative ancestor A. angustifolia Haw. (Gentry 1982). Using isozyme markers, Colunga-GarcíaMarín et al. (1999) found that A. fourcroydes exhibits lower levels of genetic diversity than wild A. angustifolia populations, probably due to the predominant asexual reproduction of a clone selected for commercial monoculture plantations since the beginning of the 20th century, in addition to the disappearance of the traditional management. In southern areas of the state of Jalisco, the complex of taxa related to A. angustifolia has been used to produce mescal and tequila. Currently, it is possible to find wild varieties, traditionally managed varieties (of the species A. angustifolia and A. rhodacantha Trel.) and one variety that is predominately found in commercial monoculture plantations (A. tequilana var. azul). Using inter simple sequence repeat markers, Vargas-Ponce et al. (2009) found that traditional landraces of A. angustifolia have genetic diversity levels similar to those recorded in wild populations. In contrast, genetic diversity in A. tequilana was markedly low. These trends can be explained as a result of the differences between traditional management and commercial monocultures, and the predominant asexual method used to propagate agave crops. In southeast Arizona, A. parryi Engelm. and A. p. var. huachucensis (Baker) Little were cultivated since prehistory, as sources of food and fibre. A number of relict populations from ancient cultivated areas remain in modern landscapes. Parker et al. (2010Parker et al. ( , 2014 evaluated the genetic divergence between wild and anthropogenic populations (relicts of ancient managed populations) through isozymes and microsatellites. In this case, the genetic diversity in the anthropogenic populations was lower than that of wild populations, a trend that can be explained due to dominance of asexual reproduction in anthropogenic populations.
In the cases mentioned, the general trend observed is that cultivated populations exhibit lower levels of diversity than wild populations. The most frequent explanation for this pattern is that genetic diversity in wild populations has been modelled for millions of years, whereas managed populations most commonly involve a fraction of such variation in time periods much shorter. However, in the case of the complex A. angustifolia in southern Jalisco, it is clear that traditional management allows the occurrence of high genetic diversity levels in the crop lands, similar or even higher than in the wild, through the constant let standing or encouraging of plants from wild populations already growing in the crop land, or introducing them to these areas. A similar pattern was also reported for several columnar cacti species (Casas et al. 2007;Parra et al. 2008Parra et al. , 2010. Agave inaequidens Koch is distributed mainly in pine and pine-oak forests of the Trans-Mexican Volcanic Belt. Historically, this species was consumed as food and its sap extracted for drinking fresh or fermented (Gentry 1982). At present, it is used mainly for producing mescal and fibre. In Michoacá n, this species is found in a gradient of management intensity with populations occurring in wild habitats as part of natural forests, but also under silvicultural or in situ management, through which people leave some individuals standing when the forest is cleared, and deliberately propagate agaves in the cleared areas in order to increase their population density (Torres et al. 2015b). In addition, some people cultivate this agave away from its natural habitat (ex situ cultivation), by transplanting saplings taken from their wild populations and others produced in seedbeds or nurseries, but this practice is relatively new, no more than 30 years old (Figueredo et al. 2014;Torres et al. 2015b).
Based on previous ethnobotanical and morphological studies, we (Figueredo et al. 2014) documented the morphological variation and differential use of variants in populations of A. inaequidens. We also documented that cultivated populations are composed of individuals from different wild populations, propagated by both seeds and vegetative propagules (Figueredo et al. 2014;Torres et al. 2015b). This background leads us to hypothesize that the high morphological variation observed in cultivated and managed populations would be accompanied by high genetic diversity, similar to that found in wild populations of this species. Furthermore, given the dynamics of formation of crops of this species, and the relatively short time this activity has been carried out, we expected to record high levels of genetic diversity in both managed in situ and cultivated populations. We also expected to document high levels of gene flow among managed and unmanaged populations and low genetic structure among populations.
Our study aimed to (i) assess levels of genetic diversity in populations of A. inaequidens within a gradient from low to intensive management; (ii) evaluate the levels of genetic structure and divergence between wild, silvicultural managed and cultivated populations of A. inaequidens and (iii) estimate gene flow among populations under different management practices, in order to identify possible sources of the gene pools.

Study species
Agave inaequidens is recognized with the common name of 'maguey alto', which makes reference to its large size, or 'maguey bruto', which makes reference to the caustic property of its tissue, because of the presence of saponins and other secondary metabolites that can cause dermatitis (Gentry 1982). This species is endemic to Mexico, naturally growing throughout the Trans-Mexican Volcanic Belt (Gentry 1982). Its distinctive characteristic is teeth dimorphism (successively one shorter and one larger) along the leaves' edges. It is monocarpic, with yellow AoB PLANTS www.aobplants.oxfordjournals.org flowers, protandrous and xenogamous, with pollination conducted by bats and also diurnal animals (Leó n-Jacinto 2013). Fruits are capsules producing shiny black seeds dispersed by the wind (Gentry 1982;Leó n-Jacinto 2013). This species has .30 different documented uses (Torres et al. 2015b). In the past, this agave was more important as food, since its inflorescences and stems were cooked in underground ovens and consumed. Its sap was extracted for drinking fresh or fermented (Gentry 1982). Nowadays, those uses are rather rare, and its main use is mescal production and, in some areas, the extraction of its fibres for manufacturing cords (Valenzuela-Zapata et al. 2011).

Study area
We sampled a total of 16 populations of A. inaequidens: 6 wild, 7 cultivated and the only 3 silvicultural in situ managed populations accessible in the state of Michoacá n ( Fig. 1). Wild populations were growing in pine-oak, oak and subtropical forests. Seven cultivated populations are found growing in orchards together with other species of agaves and fruit trees, where part of the forest was removed for cultivation. Two silvicultural managed populations found in the region grew in secondary forests (natural forest cover has decreased) and one in pastureland (Table 1). We collected tissue of young healthy leaves of 19 -30 individuals per population. Tissue samples were kept in silica gel until extraction of total DNA.

DNA extraction, amplification, marker screening and data quality
Total DNA was extracted following the cetyltrimethylammonium bromide protocol for plants (Lefort and Douglas 1999). We used nuclear microsatellites as genetic markers. Eleven different oligonucleotides or primers were assayed, which were originally developed for the studies of A. parryi, A. parryi var. huachucensis and A. palmeri (Parker et al. 2010(Parker et al. , 2014Lindsay et al. 2012).
Polymerase chain reaction (PCR) was conducted using the QIAGEN (www.qiagen.com) Multiplex kit at a final volume of 5 mL, including 2.5 mL of Master Mix (with HotStartaq DNA polymerase, Multiplex PCR buffer, 3 mM MgCl 2 and deoxynucleotides), 1 mM of each primer, 1.5 mL sterilized water and 0.5-1.0 mL of 50-100 ng mL 21 DNA. For all microsatellite loci, amplifications were carried out with a EscoSwiftMaxPro thermal-cycler, using the following protocol: initial heat activation for 15 min at 95 8C, and then 35 cycles of denaturing at 95 8C for 1 min, annealing at 60 8C for 1 min 30 s for all microsatellite loci and extension at 72 8C for 1 min. We included a final step of extension at 72 8C for 7 min. PCR products were mixed with formamide and the size standard Gene Scan LIZ-500 (Applied Biosystems) and denatured for 5 min at 95 8C. PCR products were analysed in a 3130xl Genetic Analyser (Applied Biosystems) sequencer. The resulting electropherograms were analysed using the program Peak Scanner (Applied Biosystems).
Possible genotyping errors due to the presence of null alleles, large alleles or stuttering were identified using the MicroChecker 2.2.3 software (Van Oosterhout et al. 2004) with 1000 bootstrap simulations and a confidence interval (CI) of 95 %. Deviations from Hardy-Weinberg Equilibrium (HWE) were examined for all loci in each population using the exact test with Arlequin version 3.11 (Excoffier et al. 2005). Deviations from linkage equilibrium (LE) were estimated using Genepop on the Web program with Fisher's method for each pair of loci (Raymond and Rousset 1995).

Genetic diversity
Number of alleles per locus (Na), effective number of alleles (Ne) per locus, expected heterozygosity (He), unbiased expected heterozygosity (uHe) and observed heterozygosity (Ho) were estimated using GenAlEx (Peakall and Smouse 2006). Kruskal-Wallis one-way analysis of variance range tests were performed to identify significant differences in levels of diversity according to the management type for each of the estimated parameters.

Genetic structure
Following Weir (1996), the fixation index (F ST ) was calculated with the program FreeNA using the excluding null alleles method assuming null alleles, with 10 000 bootstrap repetitions (Chapuis and Estoup 2007 (1967) were estimated for each pair of populations using the including null alleles correction described in Chapuis and Estoup (2007). From a matrix of Nei's genetic distances (D), we constructed a dendrogram through the unweighted pair group method with arithmetic mean (UPGMA) method with 1000 bootstrap replicates of the original matrix with TFPGA 1.3 (Miller 1997) and the MEGA program (Tamura et al. 2013). Isolation by distance and Mantel tests were examined with isolations by distance webb service (Jensen et al. 2005) using Nei's genetic distances.

Genetic differentiation and genetic flow
The program STRUCTURE version 2.3.4 (Pritchard et al. 2000;Falush et al. 2003) was used to perform the Bayesian clustering, in which individuals are probabilistically assigned to one of K predefined groups to identify the optimal number of genetic groups (Evanno et al. 2005). The optimum group number (K) was determined varying K from 1 to 17, with 10 runs for each K value, in order to determine the maximum value of the posterior probability [ln P(K)]. Every run consisted of 5.0 × 10 4 burn-in and 10 6 periods of MCMC repetitions after the burn-in. We used the admixture model with correlated allelic frequencies without prior information. The number of subpopulations was additionally estimated based on the approach of Evanno et al. (2005) using the software STRUCTURE Harvester (Earl 2011). In order to align the cluster membership coefficients of the 10 structure runs and to graphically display the results, we used the programs CLUMPP version 1.1.2 (Jakobsson and Rosenberg 2007) and Distruct version 1.1 (Rosenberg 2004). Analyses of molecular variance (AMOVAs) were used to test for genetic differences firstly among the populations under the three types of management and then according to the genetic grouping resulting from the Bayesian clustering. For these tests, we used the stepwise mutation models with the program Arlequin version 3.11 (Excoffier et al. 2005). Migration rate (M ¼ m/m, where m is the migration rate per generation and m is the mutation rate) paired in both directions, and Theta (Q ¼ 4N e m, where N e is the effective size of population) was estimated through MIGRATE-N Felsenstein 1999, 2001), based on maximum likelihood using the Brownian method and a constant mutation rate (m). From the values of M and Q, we estimated the gene flow or number of migrants per population (Nm). The effective population size (Ne) per population was estimated through an average mutation rate for microsatellites (5 × 10 24 , Schlötterer 2000; Selkoe and Toonen 2006).

Genetic diversity
In total, we successfully amplified 10 microsatellite loci. No evidence for large-allele dropout was detected in any locus using the software Micro-Checker. Null alleles seemed to be present in 14 populations for loci 6 and 7 (APARLC20, APARLC21, Lindsay et al. 2012), whereas populations CSAH1 and M1 had no evidence of null alleles. Significant deviations from HWE were recorded associated with heterozygote excess in the wild populations SPIE, SCUA, SICU and SPB; the cultivated populations CROC1, CROC2, SAHC1 and LH and in the silvicultural managed populations M1, M2 and SSAH1. Heterozygote deficiency was recorded in two wild populations (SPR and SSAH2) and three cultivated populations (CSAH2, CROC3 and TC). A global test of LE indicated that genotypes at one pair of loci (APAR2-12 and APARLC28, Lindsay et al. 2012) are not independent (P ≤ 0.05), but identical results were found in the rest of the analysis when either of these loci was removed.
The Na ranged from 4.0 to 8.6, with the highest value found in the wild population SPIE and the lowest value recorded in the wild population SSAH2 and the cultivated population CSAH2 (Table 2). The Ne ranged from 2.37 to 5.04, with the highest value found in the wild population SPB. The values of Ho ranged from 0.2 to 0.87, with the highest value recorded for the cultivated population CSAH1. He ranged from 0.449 to 0.754, with the highest value observed in the cultivated population CLH. For uHe, estimates ranged from 0.461 to 0.774; in this case, the managed population M1 had the highest values. An interesting point to emphasize is that the cultivated population CSAH2 exhibited the lowest values in all the diversity parameters estimated (Table 2) The UPGMA dendrogram based on genetic distances among populations (Fig. 2) did not cluster populations according to their management type, but several clusters were partially concordant with the geographic location of several populations. Populations SSAH2 and CSAH2, located in the western part of the sampling area, were separated in a group, different from the remaining populations. These populations were also separated in the dendrogram from the SSAH1 and CSAH1. The next group included the four populations of region of Sahuayo in the western part of the study area, separated from the remaining populations. Another group included population CTC and the two silvicultural managed populations M1 and M2. The remaining populations were grouped neither according to their geographic distances nor according to their management category. The regression coefficient of the linear regression analysis of F ST /(1 2 F ST ) on ln of geographic distance (Fig. 3)   The highest value of DK revealed that the most probable number of genetic groups was K ¼ 5 (Fig. 4A). Figure 4B shows the proportion of ancestry of each population and individual plants based on these groups. Three wild populations and one cultivated population (SPIE, SCUA, SICU and CROC1, respectively) had a higher proportion of the orange genetic group. The green genetic group comprised two wild and two cultivated populations (SPR, SPB, and CROC3, CROC2, respectively). The brown genetic group (SSAH1 and CSAH1) and the pink genetic group (SSAH2 and CSAH2) are markedly different, because their members had a very low proportion of ancestry from other genetic groups. The remaining populations were included in the grey genetic group (CLH, CTC, M2 and M1) (Fig. 5). Overall, the grouping pattern under this analysis was very similar to that found through the UPGMA, in which the groups did not correspond to the categories of management but rather are associated with geographic location and other ecological aspects like life-history characteristics.
The AMOVA carried out to test whether populations had hierarchical genetic structure according to the management type categories was not significant (R ST ¼ 0; P ≥ 0.05). When AMOVA was conducted using the STRUCTURE-defined populations, results indicated that most of the variation occurs among groups (59.32 %; P ≤ 0.001, Table 3), followed by the variation within populations (36 %, Table 3). Total differentiation was high and significant among groups (F CT ¼ 0.5926; P ≤ 0.001) and within populations (F ST ¼ 0.6322; P ≤ 0.001), whereas differentiation among populations (F SC ¼ 0.0972; P ≤ 0.001) was moderate but significant.     Byers et al. 2014), which have also been studied by using microsatellite loci. The dynamic movement of propagules among populations of A. inaequidens involved in silvicultural management and cultivation could be one of the causes determining high levels of genetic diversity in these populations. Overall high levels of genetic diversity and the moderate levels of genetic structure among populations of A. inaequidens are consistent with several life-history traits of this species: predominant sexual reproduction, the outcrossing breeding system (Leó n-Jacinto 2013) and involvement of long-distance pollinators (bats and birds) (Leó n-Jacinto 2013).
Although morphological studies have reported significant phenotypic differences between wild and cultivated populations of A. inaequidens (Figueredo et al. 2014), this pattern was not consistent with the results of our genetic structure analysis using microsatellite loci. No genetic divergence was detected among populations according to management practice, which indicates that management and artificial selection documented to occur on A. inaequidens by other studies (Figueredo et al. 2014;Torres et al. 2015b) have not affected genetic diversity within the species. Artificial selection could be a recent process in this species or not strong enough to have caused genetic differentiation between wild populations and populations exposed to management practices, because human-induced and naturally occurring gene flow may rapidly counteract the effects of artificial selection. Also, it could be that there may be phenotypic plasticity in the genus since phenotypic differences documented previously could be largely influenced by local environments, and therefore, artificial selection would be acting on features that have a low heritability. Or, finally, the neutral markers used are not efficient to identify artificial selection on the phenotypic features evaluated.

Genetic diversity in A. inaequidens
Despite its relatively recent origin (approximately 10 my old), the genus Agave has a high adaptive radiation and an active process of speciation, with a diversification rate (0.32 species per my), only comparable with the rosetophyllous silversword (0.56 species per my) (Baldwin and Sanderson 1998;Eguiarte and Souza 2007). Often, the highest levels of genetic diversity have been found in the centre of origin, and levels decrease as long as the distance increases until the extremes of the area of distribution (Eckert et al. 2008;Figueredo and Nassar 2011;Parra et al. 2015). The main area of diversification of the genus Agave is central Mexico, where A. inaequidens occurs. This fact could help to explain the high levels of genetic diversity of this species.
Polyploidy is a source of genetic variation in plants. Nearly 70 % of Agave species are polyploid. We do not discard the possibility that A. inaequidens populations have different ploidy levels, similar to what has been recorded in A. cupreata (Palomino et al. 2012) and A. parryi (Parker et al. 2014). However, this supposition is yet to be confirmed through cytogenetic and flow cytometric studies. Generally, polyploids show the giant phenotype, which includes large size plants. It is possible that in A. inaequidens, polyploidy has an influence on one of the target features of artificial selection in favour of more productive agaves, originally for more edible matter, more aguamiel and pulque production, and more recently for more efficient mescal production. However, the microsatellites tested showed patterns typical of diploid plants.
Other aspects that can contribute to the high levels of genetic diversity and substantial gene flow are life-history traits (Hamrick et al. 1992;Hamrick and Godt 1996). Agave inaequidens is self-incompatible, with a xenogamous pollination system, and bat pollinated. All these traits can promote gene flow and high levels of genetic variability within populations (Horner et al. 1998;Rocha et al. 2006;Eguiarte and Souza 2007;León-Jacinto 2013;Parra et al. 2015).
Although, in general, we found high levels of genetic diversity in all studied populations, we identified two populations with comparably lower levels of genetic variation: SSAH2 and CSAH2. Possibly, the differences are associated with anthropogenic effects. The wild population SSAH2 has been under an intensive extraction regime for mescal production for .40 years, suffering significant population decline. In addition to agave extraction, the population has also been affected by livestock raising, because animals may consume saplings and trample seedlings, thus affecting recruitment of new plants (Torres et al. 2015a), as also documented for A. potatorum Population CSAH2 is composed of individual agaves that form a live fence, and plants of the fence are used for extracting sap. The process of sap extraction enhances asexual reproduction, forming axillary suckers, and apparently, the new plants forming the fence are clones maintained for several generations (Torres et al. 2015b). In other species of agave (A. tequilana, A. fourcroydes and A. parryi), it has been similarly reported that the low levels of genetic diversity are due to asexual propagation (Colunga-GarcíaMarín et al. 1999;Vargas-Ponce et al. 2009;Parker et al. 2010Parker et al. , 2014.

Structure and gene flow among studied localities
We recorded a significant, although slight, pattern of isolation by distance for the set of localities sampled; however, we noticed that this pattern seems mainly driven by distances from the most distant localities. In addition, Bayesian grouping appears to be explained by geographic distance and the type of ecosystem where agaves grow. These results are supported by the AMOVA, indicating that 60 % of the variation occurs among these five genetic groups. Populations from Sahuayo grow in subtropical scrub forest at the western area of Michoacá n. These populations are the most genetically distinctive, and there is substantial gene flow among them. These populations could be adapted to a drier and warmer environment. Morphological differences between these populations and populations occurring in oak and pine-oak forests were reported by Figueredo et al. (2014).
High levels of historical gene flow were detected among most populations examined, independently of their management type. As referred to above, bat pollination makes possible a long-distance exchange of genes among populations, since Leptonycteris yerbabuenae can potentially fly nearly 100 km in one single night (Horner et al. 1998). But it is also relevant to indicate that the agave handlers are moving seeds and vegetative propagules from forest to forest, from forests to plantations and among plantations themselves, thus promoting gene flow among populations. The populations with fewer incoming migrants per generation were located in the Sahuayo region, which can be explained by the relatively long distance from other populations.

Implications for management and conservation
An important aspect associated with management of agaves is that cultivation is conducted mainly in agroforestry systems where sexual and asexual propagules and even plantlets transplanted from different wild and managed sources are maintained together. This management system determines that these managed populations represent important reservoirs of genetic variation. Some handlers sow seeds and then transplant saplings to the forest (Torres et al. 2015a), which increases the probability of establishment of new agave plants in forests with this management regime. In addition, cultivation of agave together with other crops and other wild plants allows high plant heterogeneity, which can significantly contribute to control pests, an aspect that has been documented for other Agave species ) and crops (Root 1973). Populations of the western area that grow in drier ecosystems are more vulnerable, particularly because of low population density, continual plant extraction and absence of plans for population recovering. The forms of management developed in the eastern areas should be considered for adoption in the western areas. Exchange of management experiences is possible and necessary. This study identified some effects of the management regimes carried out in the eastern zone of the study area. It is still necessary to evaluate the population genetic patterns of A. inaequidens in other regions of Mexico. Particularly interesting would be those areas where the species is used mainly for the production of fibre, which apparently is an activity that has decimated the agave populations, and where no reforestation or cultivation planning has occurred (Valenzuela-Zapata et al. 2011).

Conclusions
Agave inaequidens is a species with high levels of genetic diversity along the entire management gradient. Population genetic structure in relation to management practices was moderate, because management is relatively recent, and artificial selection can be rapidly counteracted by high gene flow.
Although we recorded high levels of genetic diversity in general in this species, some populations appear to be at risk, because of low population sizes and levels of genetic diversity, in addition to the continuous removal of individuals for the production of mescal, without a strategy for reforestation or farming.

Contributions by the Authors
C.J.F., main author, involved in the study design, conducting of field and laboratory work, literature review and general data collection, systematization and analysis, wrote the first draft and concluded the final version of this paper. A.C., main coordinator-supervisor of the research project, contributed with original data and the designing of all the researches providing the information for the current analysis; participated in systematization and analysis of data and reviewed several drafts of the manuscript and concluded the final version this paper. P.C.-G., J.M.N. and A.G.-R. contributed to designing and following progress of the research and fieldwork and data analyses. All authors contributed to writing, reading and approved the final manuscript.
AoB PLANTS www.aobplants.oxfordjournals.org for her, before being published; Ignacio Torres for recommendations of populations for sampling and field assistance and Alejandro Leon Jacinto for field assistance; Carolina Larios and Dolores Lugo Aquino for laboratory assistance, and Francisco Alberto Rivera for his valuable help with data analysis.

Supporting Information
The following additional information is available in the online version of this article -Appendix S1. Migration rates (M ¼ m/m) paired in both directions and parameter Theta (Q ¼ 4N e m) for 16 populations of A. inaequidens. + ¼ receiving population. Top row of the table shows donor populations, and first column to the left shows populations that receive migrants.
Appendix S2. Effective population size (Ne) and the number of migrants (Nm) for 16 populations of Agave inaequidens. + ¼ receiving population. In the table, the donor population is shown in the top of the table and the population that receives is shown in the left of the table.