Genetic structure of two Prosopis species in Chaco areas: A lack of allelic diversity diagnosis and insights into the allelic conservation of the affected species

Abstract The Gran Chaco is the largest continuous region of the South American dry forest, spanning Argentina, Paraguay, Bolivia, and Brazil. Prosopis rubriflora and Prosopis ruscifolia are typical tree species of chaquenian area forests, which have been subjected to continuous fragmentation caused by cattle raising. This study evaluated P. rubriflora and P. ruscifolia in areas with varying levels of disturbance. We investigated the contemporary genetic diversities of both species in areas with distinct anthropogenic disturbances. Even with a lower heterozygote frequency, disturbed areas can provide important storage for alleles, allowing the maintenance of diversity. The genetic diversity of P. rubriflora was surprisingly similar to that of P. ruscifolia (H e = 0.59 and H e = 0.60, respectively) even with very different distribution ranges of both species. However, P. ruscifolia exhibited a higher intrapopulation fixation index than P. rubriflora. P. rubriflora showed evidence of bottlenecking in 64% of the sampled areas, while P. ruscifolia showed such evidence in 36% of the sampled areas. Additionally, P. rubriflora had two distinct populations due to its disjunctive geographic distribution, whereas P. ruscifolia had a single population that exhibited few signs of population structure in some areas, possibly due to the main pollinators presenting a short range of dispersion. Our results suggest that 42 Chaco areas should be conserved to retain the minimum of 500 individuals necessary to maintain genetic diversity for 100–1,000 generations. This study improves our understanding of these two Prosopis species and provides information for the conservation of their genetic diversities.

duced linear leaflets (Burkart, 1976). This species has two flowering peaks, with the first occurring in February and the second occurring in August. Maximum fruiting occurs from October to January, but this species continuously flowers throughout the year at a lower intensity (Stefanello, 2012). Prosopis rubriflora is associated with "arborized stepic savanna," which primarily consists of sparse nanophanerophytes (IBGE, 2012). This species is often found in clusters interspersed with other species that are commonly dominant in conserved areas (Lima, 2012). Prosopis rubriflora is observed in the southern region in Mato Grosso do Sul, Brazil, and in northeastern Paraguay (Burkart, 1976) frequently associated with the chaquenian areas in Brazil (Pott & Pott, 1994) with arboreal physiognomy. This tree has ornamental potential and is recommended for landscape design, as the wood can be used as charcoal and agricultural instruments (Lorenzi, 2002). This species was considered endangered in Paraguay according to the IUCN 1997 list (Walter & Gillett, 1998).
Prosopis ruscifolia is a tree species with a height range of 5-12 m, and its branches are armed with white inflorescences and large ovallanceolate leaflets (Burkart, 1976). This species annually produces flowers and fruits (November to February) and is typically found in forested stepic savanna, which consists of micro-and/or nanophanerophytes of varying densities. A woody, grassy layer is less common (IBGE, 2012). Prosopis ruscifolia commonly borders conserved and disturbed forested areas, and this species has larger distributions in the chaquenian areas of Bolivia, Paraguay, and Argentina (Burkart, 1976), which are associated with the chaquenian areas of Brazil (Pott & Pott, 1994). The wood can be used for making furniture and frames, firewood, activated carbon, and external uses, such as posts.
The pods produced are edible and can be processed into flour or cooked; in addition, they are fodder for livestock (Lorenzi, 2002).
The tree has been indicated for the reforestation of degraded riparian areas and is considered a valuable species for forest restoration in semi-arid chaquenian areas (Blasco, Astrada, & Carenzo, 2006).
Although the Chaco areas in Porto Murtinho are considered to have biological importance (Ministério do Meio Ambiente, 2002), no Brazilian chaquenian areas are currently designated as conservation areas (Pott & Pott, 2003)  Habitat fragmentation reduces the number of individuals and the genetic diversity of the population. Thus, it may reduce the effective population size (N e ) to a level that causes genetic drift in the short term due to the loss of rare alleles (A r ). Habitat fragmentation may cause inbreeding in the long term due to an increased probability of pollination between related and inbred individuals (Kageyama & Gandara, 1998). The allele frequency in large populations tends to be reduced by genetic drift; however, this effect is stronger in smaller populations in which genetic drift causes allele frequencies to change randomly, leading to the loss or fixation of alleles over time due to the limited alleles presented in the parental generation (Ellstrand & Elam, 1993). Furthermore, these new allele frequencies contribute to increased differentiation among populations. Smaller populations trend toward increased individual homozygosity and inbreeding depression, which tends to reduce fitness and affect the fertility and survival of individuals (Charlesworth & Willis, 2009), especially for noninbred species (Ellstrand & Elam, 1993).
In this context, this study aimed to test the following hypotheses: (i) the genetic diversity of P. rubriflora should be lower than that of P. ruscifolia due to the smaller geographic distribution of the former, and (ii) strong genetic structures are present in both species due to the behavior of the main pollinating agent, as described in the literature on Prosopis. Thus, the objectives of this study were to evaluate the genetic diversity and to estimate the genetic structures of the sampled areas containing P. rubriflora and P. ruscifolia to examine whether they have been influenced by distinct anthropogenic disturbances.

| Plant materials
We considered 19 chaquenian areas, also known as stepic savanna, from which samples of P. rubriflora and P. ruscifolia were collected in the Corumbá, Nioaque, and Porto Murtinho counties of the State of Mato Grosso do Sul, Brazil (Figure 1). Stepic savanna is composed of neotropical steppe vegetation cover with low and spiny plants in a grassy savanna (Furtado et al., 1982). The climate is hot and dry, with a reduced annual rainfall of approximately 1,000 mm (Loureiro et al., 1982). Seasonal rains are concentrated from November to February (rainfall ≥100 mm), and a drought occurs from June to September during the dry season (Stefanello, 2012). The soils are saline with limited drainage due to the fragipan horizon, and the land is covered in water for several months. When the soil is drained, it becomes parched (Furtado et al., 1982).
The sampled areas were classified according to the observed levels of anthropogenic disturbance based on field observations and surveys by the farm owners. "Conserved areas" represent no to low disturbance levels, such as cattle breeding activity without any evidence of anthropogenic suppression. The "intermediate disturbance areas" have experienced recent suppression in vegetation (10-15 years) or have a predominance of colonizing species, such as Parkinsonia praecox and Mimosa hexandra Micheli. The "degraded areas" show a predominance of pastures or land in the initial stage of regeneration of woody plants. The coordinates were obtained from a global positioning system (GPS) using WGS-84 datum.
Both species were sampled in the following areas of Porto Murtinho: Fazenda Retiro Conceição (area 1), Fazenda Tereré (areas 1 and 2), and Fazenda Santa Cristina, which did not present a continuous conservation status but rather "conserved" fragments surrounded by higher or lower degrees of anthropogenic disturbance.
Prosopis rubriflora was exclusively located in the following areas: We collected the leaves and leaflets of 241 P. rubriflora individuals and 308 P. ruscifolia individuals, ranging from 16 to 30 trees per area (Tables 2 and 3). The distance between the samples ranged from 10 to 1,470 m; however, we aimed to maintain an average distance of 50 m between two individuals to prevent the collection of related samples. The sampled leaves and leaflets that were collected for genomic DNA extraction were initially stored in silica gel and subsequently deposited at −80°C.
F I G U R E 1 Locations at which samples of Prosopis rubriflora and Prosopis ruscifolia were collected. The sampled areas are represented by triangles: red indicates P. rubriflora, white indicates P. ruscifolia, and the combination of red and white indicates areas where both taxa were collected. On the map, the yellow areas represent the Cerrado domain, the purple areas represent the Pantanal domain, and the dark purple areas represent the priority areas for conservation for the Pantanal domain. The sampled area codes are presented in Table 1. The map was created by the speciesMapper tool, which is available from the speciesLink project (http://splink.cria.org.br/tools)

| DNA extraction, SSR markers, and genotyping procedure
Genomic DNA extraction, fragment amplification, and genotyping were performed according to protocols published by Alves, Zucchi, Azevedo-Tozzi, Sartori, & Souza (2014). Population genotyping was developed with ten SSR markers for the analysis of P. rubriflora (eight specific markers and two markers transferred from P. ruscifolia) and 11 SSR markers for the analysis of P. ruscifolia (nine specific markers and two markers transferred from P. rubriflora) as developed by Alves et al. (2014).

| Intrapopulational analysis
The amplified SSR markers for P. rubriflora and P. ruscifolia from the sampled areas were analyzed for linkage disequilibrium (LD), adherence to Hardy-Weinberg (HW) equilibrium and the frequency of null alleles (NA). LD and HW analyses were performed using Genepop v.1.2 software (Raymond & Rousset, 1995) and Fisher's exact probability test with 10,000 dememorizations and iterations and 100 batches. The frequency of null alleles was calculated using FreeNA software (Chapuis & Estoup, 2007) assuming that n > 0.20 for the presence of NA based on the algorithm of Dempster, Laird, & Rubin (1977).
Wright's F statistic was employed to estimate the population genetic structure by the F IS , the fixation index between populations (θ or F ST ), and the total endogamy (F or F IT ). The number of private alleles (A p ) was estimated with GDA v.1.0 software (Lewis & Zaykin, 2001) using a 95% CI that was determined by bootstrapping with 10,000 replicates.
The apparent outcrossing rate (t a ), the primary parameter for outcrossing into populations, was obtained according to the equation t a = 1−F IS 1+F IS as described by Vencovsky (1994). The effective number TA B L E 1 Sampled areas in which Prosopis rubriflora and Prosopis ruscifolia were collected and their respective preservation statuses (N e ) was estimated using an equation derived from Cockerham (1969) that included both the inbreeding index and the average coefficient of coancestry (Θ) from a generation: N e = 0.5 2n (Tambarussi et al., 2016). Kimura and Crow (1963) defined N e as the size of an idealized population that would have the same amount of inbreeding or random genetic drift as the same population under consideration.
TA B L E 2 Genetic parameters based on ten microsatellite (SSR) loci analyzed in Prosopis rubriflora samples in the 11 areas of arborized stepic savanna  The group coefficient of coancestry for adults within populations was estimated using the coancestry estimated by Loiselle, Sork, Nason, & Graham (1995) implemented in the software SPAGeDi v.

| Bottleneck
Bottleneck analysis was performed using Bottleneck v.1.2.0 software (Piry, Luikart, & Cornuet, 1999) to identify populations under genetic mutation-drift equilibrium (simulated coalescent process) as described by Cornuet and Luikart (1996). We employed the stepwise mutation model (SMM) and the two-phase model (TPM) (12 variations and 95% SMM) with the sign test and the Wilcoxon signed-rank test (two-tailed). The sign test utilizes parametric analysis in which the null hypothesis can be rejected based on excess heterozygosity by considering the difference H o and H e in a significant number of loci in the studied population (Cornuet & Luikart, 1996). The Wilcoxon test compares the observed excess heterozygosity over H e to a null hypothesis and is a more robust and sensitive method (Piry et al., 1999).

| Global structure and migrants
The historic gene flow (N m ) was estimated indirectly according to the Crow and Aoki (1984) , and n is the number of samples.
Nei's genetic distance (Nei, 1978) was estimated to identify similarities or differences between two fragments and complement pairwise F ST . Nei's genetic distance analysis was performed using TFPGA v.1.3 software (Miller, 1997) to generate an unweighted pair-group method of analysis (UPGMA) matrix clustered with 10,000 bootstraps, which provided a basis for the developed dendrogram.
The Bayesian model of the P. rubriflora and P. ruscifolia genetic structures was developed with Structure v.2.3.2 software (Pritchard, Stephens, & Donnelly, 2000) using admixture and allele frequencies that were correlated between populations. The genetic clusters ranged from K = 1 to 15, and each K value was replicated 20 times. The length of the final Markov chain Monte Carlo (MCMC) was 500,000 replicates, with 200,000 replicates for burn-in. The most likely value of ΔK based on Evanno et al. (Evanno, Regnaut, & Goudet, 2005) was estimated using the online tool Clumpak (Kopelman, Mayzel, Jakobsson, Rosenberg, & Mayrose, 2015). Discriminant analysis of principal components (DAPC) analysis was performed using the adegenet package (Jombart, 2008) for R software (R Development Core Team, 2011). Different from Nei's distance, F ST , and Bayesian analysis (Structure), DAPC analysis uses a nonparametric approach. The results obtained from this analysis are presented as multidimensional scatterplots of the principal components.

| Intrapopulational evaluation of Prosopis rubriflora
Based on the genotyping of P. rubriflora individuals, we were able to identify departure from HW equilibrium based on Fisher's exact test at loci Prb2 and Prb4 and possible null alleles in AAL's area (Tables   S1 and S2). However, no evidence of LD for the evaluated loci was observed after Bonferroni correction (1% and 5% p-values = .0002 and .0011, respectively) (Table S3).
We identified 98 distinct alleles in P. rubriflora, with an average of 10 alleles per locus in the 11 sampled areas from the Nioaque and Porto Murtinho counties ( Table 2). The number of alleles ranged from 41 for AAL to 71 for RMS and averaged 58 alleles per area.
The A ri ranged from 3.9 (AAL) to 6.1 (RMS) alleles, and the A e ranged from 23 (AAL) to 36 (RMS and FSV) alleles. The percentage of effective alleles relative to the number of sampled alleles varied from 47% (FTR1) to 65% (FTR2). We detected 12 A p , half of which were detected in the RMS region, followed by FSV (two alleles), FPT1, FTR1, FSM, and AAL (one A p per area); the remaining areas had no A p . Forty-two A r , ranging from 11 (AAL) to 23 (RMS and FTR), were detected in the sampled areas, with an average of 16 alleles for all areas. Therefore, P. rubriflora contained 45% common alleles, 43% A r, and 12% A p ( Table 2).
The average H o was 0.56 and ranged from 0.40 (CI 95% = 0.26-0.51) (AAL) to 0.64 (CI 95% = 0.47-0.80) (FSC). The average H e was 0.59 and ranged from 0.49 (CI 95% = 0.34-0.60) (AAL) to 0.64 (CI 95% = 0.51-0.73) (RMS) ( Table 2). These analyses reveal the current frequency of heterozygotes (H o ) and the expected frequency of heterozygotes (H e ) in a panmictic population according to assumptions of the HW equilibrium model (Frankham, Ballou, & Briscoe, 2008). The H o was higher than the H e in the areas FSC and FTR1; this difference suggests potential pressure in favor of heterozygotes for these areas compared with the remaining sampled areas.
The average inbreeding coefficient (F IS ) was 0.05 and ranged from −0.08 (CI 95% = −0.10 to −0.04) (FSC) to 0.17 (CI 95% = 0.08-0.24) (AAL) ( Table 2). AAL was the only area with an F IS that significantly differed from zero according to the p-value; this difference suggests an excessive quantity of homozygotes in this area compared with the remaining sampled areas (Table S4). The t a ranged from 0.74 (AAL) to 0.96 (FSC) and averaged 0.86; this result suggests that mixed reproductive systems were functioning in most sampled areas, with a strong allogamous tendency for FSC.
The N e parameter suggests that 54% of the genetic contribution arose from individuals in the group of populations (129 of 241 individuals), varying from 45% (AAL) to 59% (FPT1 and FSM) (

| Intrapopulational evaluation of Prosopis ruscifolia
Using the genotyped P. ruscifolia samples, we were able to detect departure from HW equilibrium based on Fisher's exact test for most of the loci, excluding Prb2, Prb4, and Prs3. Possible null alleles were detected in the markers Prs11 (EPM and FRC1) and Prs6 (FTR2) (Tables S5 and S6). After Bonferroni correction (1% pvalue = .00018), evidence of LD was noted according to pairwise analysis of the following loci: Prb4 × (Prs11, Prs6, Prs7, Prs1, Prb2, and Prs2 × (Prs1, Prs11) ( Table S7). The assumed linkage between these loci was not evident after excluding the following areas: FQB, CJR, NSA, and ECD (Table S8). Therefore, the linkage between the loci may be associated with issues in the areas, such as inbreeding, population subdivision, and potential bottlenecks (Slatkin, 2008). In this context, the markers that revealed deviation in the analysis did not need to be discarded.
In the P. ruscifolia dataset, 138 alleles were detected, with an average of 13 alleles per locus for 11 areas in Corumbá (one area) and Porto Murtinho (10 areas). The number of alleles ranged from 4.4 (EPM) to 6.8 (FFL), with an average of 67 alleles per area (  The F IS ranged from 0.01 (CI 95% = −0.05-0.12 and CI 95% = −0.07-0.11, respectively) (FSC and ROE) to 0.28 (CI 95% = 0.17-0.37) (ECD) and averaged 0.12. A total of 63% of the areas presented values that significantly differed from zero, suggesting an excessive number of homozygotes based on the p-value (Table S4). The t a ranged from 0.57 (ECD) to 0.89 (FSC) and averaged 0.77; this result suggests that a mixed reproductive system is present in all areas. The value of N e suggested that 43% of all sampled trees provided genic contributions to the sampled areas, and this percentage ranged from 31% in ECD to 49% in FSC (Table 3).

| Bottleneck analysis
To reduce the number of errors in the analysis, the markers Prs6 and Prs11 for P. ruscifolia were discarded due to their higher proportions of departure from HW equilibrium for most of the evaluated areas (Table S5); according to Luikart and Cornuet (1998)

| Global structure and migrants
The estimated F IS suggested that P. rubriflora has a panmictic genic distribution (0.042) because this value did not significantly differ from zero according to the CI (Table 5)  The AMOVA results demonstrated that most of the genetic variation (92% for P. rubriflora and 94% for P. ruscifolia) is retained within the populations sampled, with smaller portions of 8% and 6% for those two species, respectively ( Table 6).
The Mantel test revealed a positive and significant relationship between geographic and genetic distances based on Nei's distance (r = .15; p-value < .01) for P. ruscifolia and (r = .2; p-value < .01) P. rubriflora, indicating significant isolation-bydistance across all the sampled areas. However, considering only the areas sampled in Porto Murtinho, no such correlation was observed for P. rubriflora (r = .03; p-value = .10) or P. ruscifolia (r = 0.03; p-value = .09).
The pairwise F ST ranged from −0.008 to 0.273 for P. rubriflora and from 0.011 to 0.101 for P. ruscifolia (Table S9) For P. ruscifolia, the level of genetic structure was low for 65% of the area combinations in both sampled counties and intermediary for 35% of the combinations. An intermediary genetic structure was primarily observed for the ECD area (Corumbá); the distance between this area and the other areas in Porto Murtinho ranged from 189 to 244 km. Even with a moderate level of structure, the distinction between the sampled areas of P. ruscifolia was not clear (Table S9).
Nei's genetic distance (Nei, 1978) varied from 0.000 to 0.450 for P. rubriflora ( Figure 2) and was divided into two groups: the first For P. ruscifolia, the variation in Nei's genetic distance (Nei, 1978) between two areas ranged from 0.014 to 0.153. This species presented two major groups, ECD (Corumbá), which is the most dis-  (Figure 7). Similarly, the DAPC also presented two principal components for P. ruscifolia, where ECD was the only area that did not present an overlay of individuals from the other areas ( Figure 8).
TA B L E 6 Analysis of molecular variance (AMOVA) of 11 sampling areas of Prosopis rubriflora and Prosopis ruscifolia

| D ISCUSS I ON
Small and isolated populations may undergo biparental inbreeding, which causes a loss of genetic diversity and inbreeding depression and may reduce the ability of the population to respond to environmental changes (Frankham et al., 2008). Thus, genetic diversity within populations is fundamental for conservation biology; better adaptation potential is expected in populations with high levels of F I G U R E 3 Dendrogram of the 11 sampled areas of Prosopis ruscifolia as determined by Nei's genetic distance.
The sampled area codes are presented in Table 1. Matrix derived from 11 SSR markers as defined by the unweighted pair-group method of analysis (UPGMA) with 10,000 replicates F I G U R E 4 ∆K values for all sampled areas of Prosopis rubriflora and Prosopis ruscifolia. (a) Represents P. rubriflora, and (b) represents P. ruscifolia. The values were determined based on the average of L (K) for 10 resamplings according to the model proposed by Evanno et al. (2005), and the graphics were generated by Clumpak (Kopelman et al., 2015). The K value represents the most likely ∆K according to the highest peak F I G U R E 5 Population structure based on Bayesian analysis of the 11 stepic savanna areas sampled for Prosopis rubriflora. The sampled area codes are presented in Table 1. Each bar represents one sampled individual as estimated by ten SSR markers (n = 241) genetic diversity (Kalinowski, 2004 Table 1. The plots represent the individuals, and the circles represent the groups of area to display more sudden fluctuations. The areas that exhibited significant A ri compared with the other areas, such as FSV and RMS, may receive major influxes of migrants compared with other areas. This influence may be reflected in the A ri , which corrects for differences among samples for all areas (El Mousadik & Petit, 1996). In the same county, the FTR2 area showed high numbers of effective alleles; this possibly reflects the suppression observed in the area and a loss of individuals with A r . Similar results for disturbed areas were also observed for P. reticulata Benth. (Oliveira, 2012) and S. lycocarpum (Moura, 2007).
The lowest allelic diversity was recorded in the disjunct area of Nioaque (AAL); however, this area lacks direct evidence of anthropic disturbance. This population may have been small for a long period and must have experienced both loss and genetic reorganization, which was expected due to genetic drift (Ellstrand & Elam, 1993).
The factors responsible for the low diversity should be related to the geographic distance and the lack of connection between this area and other areas in Porto Murtinho (Figure 1).
For P. ruscifolia, the diversity and richness of alleles were heterogeneous in configuration due to the different degrees of disturbance in the sampling sites or the reproductive system of the species. The area with the greatest diversity of alleles was FFL, which had less apparent disturbance. Although the diversity and A ri in the FQB and ECD areas were similar to the diversity and A ri in the FFL areas, these regions had the highest levels of disturbance of all the sampled areas. Similar results for disturbed areas were recorded by Barreto in D. nigra (Barreto, 2010). The ECD and FQB areas may have also undergone recent suppression, in which most of the P. ruscifolia individuals were preserved; this preservation was unlike that of all other tree species and hindered the detection of any allelic loss. Adult P. ruscifolia are frequently observed in highly disturbed areas, such as pastures, where these trees may not be cut down for practical reasons (to provide shade for livestock), esthetic reasons, or difficulties in cutting.
The genetic diversity values were similar in P. rubriflora and P. ruscifolia; this result was not expected because P. rubriflora had reduced geographic distributions in Brazil and Paraguay compared with P. ruscifolia, which also occurred in Argentina and Bolivia (Burkart, 1976). Our initial hypothesis was that P. ruscifolia would show higher genetic diversity, as noted by Godt (1996) andNybom (2004). Assuming the older and current anthropogenic disturbances were not sufficiently strong to cause similar H e values for both species, the possibility that the combination of biological factors, such as the different flowering periods for these species, were restricted for P. ruscifolia and extended for P. rubriflora could explain the results obtained herein.
The difference between the H e and H o parameters was lower in P. rubriflora than in P. ruscifolia for all areas sampled in Porto Murtinho and resulted in lower deviations from HW proportions and a lower inbreeding coefficient. The inbreeding coefficient significantly differed from zero in P. rubriflora only for the area in Nioaque; the AAL area revealed little disturbance. Because AAL contains a small population and is geographically isolated, stronger evidence of inbreeding was expected, especially in the biparental data and in terms of its structure due to limitations in pollen flow and seed dispersal (Ellstrand & Elam, 1993).
Both AAL and ECD were the only areas outside of Porto Murtinho county where we were able to find P. rubriflora and P. ruscifolia, respectively, in Brazilian chaquenian areas. The disjunctive area of Nioaque, close to Morro do Solteiro, was the only area registered for this county. For Corumbá County, there were a few areas wherein P. ruscifolia was registered in previous years; however, we were unable to find these areas for this study, most likely because they were suppressed months before our visit. Other areas with P. ruscifolia possibly exist south of Corumbá, but as the access for this region is very difficult, we were unable to seek additional areas for this county.
For conservation measures, AMOVA suggests that most genetic variations are retained inside the populations, indicating the importance of protecting the sampled areas to avoid losses in genetic variability for both species. Even though all the sampled areas contain a precious genetic resource, conservation management of the AAL should be more prioritized in the short term due to its small population size and reduced genetic diversity compared with those of the Porto Murtinho areas. This action is necessary to conserve the current genetic diversity and avoid further reductions in responsiveness to environmental changes, which increase the likelihood of local extinction (Frankham et al., 2008) and may be permanent.
In P. ruscifolia, the major differences between H e and H o were reflected in larger deviations from HW proportions and, consequently, higher values of F IS ; these differences suggest endogamy and an intrapopulational structure for 64% of the sampled areas. The isolated population of P. rubriflora in Nioaque had the highest F IS , whereas the distant Corumbá area (ECD) had the highest level of intrapopulational structure for P. ruscifolia. In addition, the geographic isolation of the severely disturbed ECD area may contribute to intrapopulational inbreeding to maximize the genetic structure. Similar levels of inbreeding were detected by Bessega et al. (2000) in P. ruscifolia; these authors reported values of T m that were similar to the average t a obtained in this study for the same species, thereby supporting the results obtained by the apparent crossing rate in this study.
The Pantanal has a recent history of degradation that is signified by N e estimates how many individuals will genetically contribute to the next generation (Nunney & Campbell, 1993), and the estimated values of N e /N were 0.25-1.00 (Ellstrand & Elam, 1993;Nunney & Campbell, 1993). Averages of 43% of trees (in P. ruscifolia), ranging from 31% in ECD to 49% in FSC, and 54% of trees (in P. rubriflora), ranging from 45% (AAL) to 59% (FPT1 and FSM), will genetically contribute to the Chaco areas, and the largest oscillations have generally occurred in disturbed areas or smaller populations. According to Frankham et al. (2008), an estimated N e of 12-1,000 is required to prevent the accumulation of deleterious mutations, whereas an N e of 50 is sufficient to avoid inbreeding depression, and an N e of 500 is sufficient to retain the evolutionary potential over 100-1,000  (Luikart & Cornuet, 1998 As such, the low genetic structure among the Porto Murtinho areas regardless of geographic distance could be due to the connections between the areas, which enable a longer gene flow for both species. Migrants may also contribute to the genetic homogeneity in these areas (Varvio, Chakraborty, & Nei, 1986).
Compared with all the Porto Murtinho areas in all the analyses, the taxon P. rubriflora showed a distinct genetic structure in AAL (Nioaque), supported by all the analyses. This structure is likely the result of geographic distance (ca. 218-267 km) and the lack of connection between these areas; this lack of connection severely limits pollen flow and causes long periods of isolation for AAL. This disjunction appears to have resulted from ancient natural factors rather than from recent anthropogenic fragmentation.
Prosopis grows poorly in acidic and low-phosphorus soils (Pasiecznik et al., 2001), such as soils in the Cerrado (Oliveira, Costa, Santos, & Moreira, 2005;Sano, 1998) between Porto Murtinho and Nioaque; this growth property limits the distribution of this genus in Brazil and in another regions wherein the genus occurs natively or was introduced ( Figure 1). become increasingly isolated and possibly present results similar to those observed in the Nioaque area for P. rubriflora, which has reduced genetic variability and a significant population structure.
We expect the results of this study to help with the allotment of subsidies for decision making and the development of conservation strategies for chaquenian areas. These measures will help to preserve the genetic stock for both Prosopis species and other rare species that occur only in this biome, which is increasingly suppressed by anthropic pressures.

| CON CLUS IONS
The genetic diversity (H e ) of P. rubriflora appears to be similar to that of P. ruscifolia. Although P. ruscifolia has more alleles, most of the alleles are rare and do not increase H e to the same degree. The apparent conservation status of an area can be misleading regarding allelic diversity, and even disturbed areas may have high allelic diversity.
Evidence of a bottleneck was detected for both species, and P. rubriflora was affected in most of the analyzed areas.
High gene flow was observed between populations, and a strong structure was evidenced only in extreme cases, such as populations at a substantial geographic distance and with a lack of connection.
The intrapopulation structure was higher for P. ruscifolia, as expected.
Despite the predominance of bees as pollen dispersal agents, a relatively small structure index was observed between the sampled areas; this small index indicates high gene flow because the connection between areas enables pollen flow. Based on the effective population number in this study, 42 Chaco areas must be preserved to preserve the minimum of 500 individuals needed to maintain genetic diversity and retain the evolutionary potential of both species over 100-1,000 generations. The measures suggested in this study should prevent additional environmental damage that may cause extinction, which would negatively affect the local fauna because P. rubriflora and P. ruscifolia provide important food resources.

ACK N OWLED G M ENTS
This study received financial support from the Fundação de Amparo a Pesquisa do Estado de São Paulo FAPESP (proc. 2008/52197-4) in the form of a graduate fellowship awarded to Fábio de Matos Alves (proc. 2010/51242-6) and from CAPES and Conselho

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N
Alves FM, Zucchi MI, and Souza AP involved in conception and designed the work; Alves FM and Sartori ALB acquired the data; Alves