Genetic studies of various Prosopis species (Leguminosae, Section Algarobia) co‐occurring in oases of the Atacama Desert (northern Chile)

Abstract In the Atacama Desert from northern Chile (19–24°S), Prosopis (Leguminosae) individuals are restricted to oases that are unevenly distributed and isolated from each other by large stretches of barren landscape constituting an interesting study model as the degree of connectivity between natural populations depends on their dispersal capacity and the barriers imposed by the landscape. Our goal was to assess the genetic diversity and the degree of differentiation among groups of Prosopis individuals of different species from Section Algarobia and putative hybrids (hereafter populations) co‐occurring in these isolated oases from the Atacama Desert and determine whether genetic patterns are associated with dispersal barriers. Thirteen populations were sampled from oases located on three hydrographic basins (Pampa del Tamarugal, Rio Loa, and Salar de Atacama; northern, central, and southern basins, respectively). Individuals genotyped by eight SSRs show high levels of genetic diversity (H O = 0.61, A r = 3.5) and low but significant genetic differentiation among populations (F ST = 0.128, F ST‐ENA = 0.129, D JOST = 0.238). The AMOVA indicates that most of the variation occurs within individuals (79%) and from the variance among individuals (21%); almost, the same variation can be found between basins and between populations within basins. Differentiation and structure results were not associated with the basins, retrieving up to four genetic clusters and certain admixture in the central populations. Pairwise differentiation comparisons among populations showed inconsistencies considering their distribution throughout the basins. Genetic and geographic distances were significantly correlated at global and within the basins considered (p < .02), but low correlation indices were obtained (r < .37). These results are discussed in relation to the fragmented landscape, considering both natural and non‐natural (humans) dispersal agents that may be moving Prosopis in the Atacama Desert.


| INTRODUC TI ON
The Atacama Desert in northern Chile extends for more than 1,000 km along the western coast of South America and is one of the most arid deserts on Earth. The extreme aridity is primarily caused by the cold water of the Humboldt Current, running parallel to the Chilean and southern Peruvian coasts and preventing precipitation in the coastal areas. Moreover, aridity is intensified by the Andes cord that produces a rain-shadow effect, blocking moisture from the Amazon basin (Houston & Hartley, 2003). However, during summer some of this moisture passes over the Andes, intensifying discharges of water flows through deeply incised gorges and shallow riparian and underground water systems (Jayne et al., 2016;Vuille et al., 2012).
In the Atacama Desert, within 19-24°S, three hydrological basins can be recognized: Pampa del Tamarugal, Loa River, and Salar de Atacama (Mortimer, 1980;Nester et al., 2007). In Pampa del Tamarugal or northern basin, the vegetation corresponds to the "Desierto del Tamarugal" (Gajardo, 1994) and is distributed in areas where the water table is relatively shallow or receives riparian flows originated in the Andes by summer precipitations (Barros, 2010). This basin has evidence of several human settlements since the Late Pleistocene, and it is connected by a small ravine with the lower course of the Loa River, nearby its mouth in the Pacific Coast (Nester et al., 2007;Pfeiffer et al., 2018). The Loa River or the central basin is the only exoreic stream traversing the desert from the Andes to the Pacific Ocean. It is placed between the Pampa del Tamarugal to the north and the Salar de Atacama to the southeast. This basin was used as a natural and cultural corridor since pre-Columbian times (Marquet et al., 1998;Núñez, 1971), passing beyond different vegetation belts, although riparian plants predominate along the watercourse (Gajardo, 1994;Villagrán & Castro, 2004). Finally, the Salar de Atacama is an endorheic basin within the Prepuna foothills (Arriagada et al., 2006;Marazuela et al., 2019). Different quebradas flow into this system giving rise to several oases with riparian vegetation (Gajardo, 1994;Villagrán & Castro, 2004). This area was also key in the history of pre-European societies (Núñez & Santoro, 1988).
The genus Prosopis (Leguminosae) includes 44 species distributed in Southwest Asia, Africa, and predominantly America (Burkart, 1976). Prosopis species are important plants in arid and semiarid regions, and many species from one particular Section, Algarobia (algarrobo), are considered multipurpose trees. From an economic perspective, humans use it as food as the raw pods can be eaten directly or processed within a variety of culinary preparations (Capparelli, 2007). Their pods and leaves provide forage for animals, and their flowers are melliferous. The wood is useful for diverse manufactures, having higher caloric value than other sympatric species (Burkart, 1976;Cony, 1996;Roig, 1993;Villagrán & Castro, 2004). From an ecological view, they can grow on sandy soils and contribute to stabilize dunes, combat desertification, and reforest degraded areas (Bessega et al., 2019;Burkart, 1976;Cony, 1996;Roig, 1993).

six species from Section
Strombocarpa and Algarobia can be found (P. strombulifera, P. burkartii, P. tamarugo, P. chilensis, P. flexuosa, and P. alba), but Muñoz (1981) also cited P. nigra for the north of Chile. While the Pampa del Tamarugal forest is dominated by three species from Section Strombocarpa (P. burkartii, P. strombulifera, and the endemic P. tamarugo), individuals belonging to different species from Section Algarobia (P. alba, P. flexuosa, and P. chilensis) can be found up to 3,000 masl and are distributed from the Pampa del Tamarugal to the Salar de Atacama basin (Figure 1; Barros, 2010;Carevic et al., 2012). Natural algarrobo populations are restricted to oases unevenly distributed and isolated from each other by large stretches of barren landscape. This area is categorized as an absolute desert since plant life is practically absent in much of its extension (Gajardo, 1994;Gayo, Latorre, Jordan, et al., 2012;Latorre et al., 2002Latorre et al., , 2005. The effects produced by landscape isolation may be comparable with those occurring on forest populations due to urbanization, overexploitation, and land conversion into crop plantations, in terms of habitat fragmentation, mating system changes, and gene flow restriction (Cascante et al., 2002;Jump & Penuelas, 2006). When populations become genetically isolated, they are at risk of losing the genetic diversity that is critical to their long-term survival (Sork & Smouse, 2006). As a consequence of the isolation, an immediate loss of alleles due to the reduction in the population is expected, with the consequence of inbreeding, population divergence increase, and genetic diversity reduction within population patches (Lowe et al., 2005). However, the longevity of the trees and effective seed and pollen dispersal (when possible) can enhance their resistance to the negative effect of the forest fragmentation (Jump & Penuelas, 2006).
In trees, gene flow is mediated by mobile structures with reproductive function as seeds, pollen, and vegetative propagules, while the adult stage is sessile. Limitation in the dispersal distance of seeds and/or pollen over time will produce a particular genetic structure consisting of decreasing relatedness of individuals with increasing geographical distance (Loiselle et al., 1995). The natural Prosopis populations in Atacama Desert constitute an interesting study model as the degree of connectivity between populations depends on their dispersal capacity and the barriers imposed by the landscape.
Prosopis populations are expected to be structured because pollen and seed dispersals are limited (Bessega, Ferreyra, Julio, et al., 2000;Bessega et al., 2011Bessega et al., , 2017. The endozoic seed dispersal determines that seeds from the same mother plant are eaten by small-and medium-sized herbivores (Reynolds, 1954) and transported away jointly. They are then deposited in dung, and full-or half-sib seeds tend to germinate together in a narrow area. Besides, the pollination is entomophilous (Genisse et al., 1990) favoring crosses among near neighbor plants.
The main objective of this study was to assess the genetic diversity and the degree of differentiation among groups of Prosopis individuals of different species from Section Algarobia and putative hybrids (hereafter populations) co-occurring in the isolated oases from the Atacama Desert and determine whether genetic patterns are related to dispersal barriers. For this, we characterized by microsatellites the variability, genetic differentiation, and genetic structure among 13 populations considering three distant basins.
Through these analyses, we addressed the following questions: (a) Are there differences in the genetic diversity parameters among populations? (b) How is the genetic variation distributed considering the different hierarchical levels (basins, populations and individuals)?
(c) What is the level of differentiation and gene flow among populations? (d) Is there a genetic pattern compatible with isolation by distance (IBD)? (e) Is there genetic structure evidence? As the Atacama Desert provides habitat fragmentation that could affect the natural gene flow, given by the large stretches of hyperarid landscape, we discuss the genetic results considering the dispersal barriers that separate the populations.
Given that Prosopis populations are restricted to isolated oases, the current hypothesis is that the gene flow restrictions among different populations would determine significant genetic differentiation among populations. Additionally, the dispersal limitation described for Prosopis species (Bessega, Ferreyra, Julio, et al., 2000;Bessega et al., 2011Bessega et al., , 2017 and the barriers imposed by the landscape should be reflected in population genetic structure.

| Study sites and sampling
The sampled range in our study extends from 19 to 24°S (Figure 1 century that we did not sample as natural populations were our study goal (Barros, 2010).
We sampled thirteen natural populations of Prosopis species from Section Algarobia (Table 1; Figure 1). We collected fresh young leaves and pods for species identification. Each sampled tree was deposited at the INTA, Hurlingham BAB herbarium, Buenos Aires, Argentina. Species identifications were done considering leaves, spines, fruits, and tree form following Burkart (1976), Burkart and Simpson (1977), and Palacios and Brizuela (2005) proposals. For each sample, we had considered habit form, leaf division, leaflets (length, width, distance between leaflets), spines (types and length), and fruits (shape, length, and epicarp color). Likewise, the identification was corroborated with the type specimen of the taxa available at https://plants.jstor.org/. As the identification of hybrids was based on morphological traits, not confirmed by molecular data, the term "putative hybrid" is used throughout the manuscript. Following Vilardi et al. (1988) method, which recommends sampling trees separated more than 30-50 m from each other to avoid collecting genetically related material, we did not gather all the trees present in each oasis, but the maximum possible. Six to seventeen adult trees (diameter at breast height [DBH] > 80 cm) were sampled in each location because populations were small, giving a total of 126 adult individuals. Geographical coordinates were recorded for each sampled tree using GPS Garmin Etrex 20, datum WGS84.

| Microsatellite analysis
Total DNA was isolated from leaf material using DNA easy Plant Mini kit (Qiagen Inc.). We used a total of eight microsatellites; four developed by Mottura et al. (2005) for P. chilensis and P. flexuosa: Mo08, Mo09, Mo05, and Mo13, and four developed by Bessega et al. (2013) for P. alba and P. chilensis: GL9, GL12, GL8, and GL21. The SSRs were amplified using forward primers labeled with a fluorescent dye (6-FAM and HEX, Invitrogen). The PCR amplifications were carried out in a 50 µl reaction volume containing 10-30 ng DNA, 0.6 µM each primer, 0.2 mM dNTPs, 0.3 U Taq DNA polymerase (Invitrogen), and 1.5 mM MgCl 2 .
A T100 Thermal Cycler (Life Science Research, Bio-Rad) was used for amplifications with a cycling profile of initial denaturation at 94°C for 5 min followed by 35 cycles at 94° for 45 s denaturation, primerspecific annealing temperature (56°-59°) for 45 s and at 72°C for 45 s extension, and a final extension step at 72° for 10 min. PCR products were electrophoresed using Macrogen service (www.dna.macro gen. com) and sized using GENEMARKER ver 1.91 (SoftGenetics, 2019).

| Data analysis
2.3.1 | Genetic diversity and inbreeding coefficient F IS Linkage disequilibrium was tested by the index of association (I A ) and a slightly modified statistic which is independent of the number of loci (r d ) (Agapow & Burt, 2001). was considered significant when the zero is not contained in its 95% confidence interval (based on 2,000 bootstrap resampling).
A, %TA, and H O were estimated using the package diveRsity (Keenan et al., 2013) of R software; A r was estimated using ASDE (Szpiech et al., 2008) that allows the correction for samples bias; and H E was estimated using SPAGeDi (Hardy & Vekemans, 2002) that performs the Nei (1978) gene correction for small number of individuals.

| Genetic differentiation
From the microsatellite's marker data, three differentiation coefficients were estimated. F ST (Wright, 1951) was estimated with the package hierfstat (Goudet, 2005) of the software R, and its significance was determined by the G test (test.g function). In order to avoid bias induced by the presence of null alleles, F ST excluding null alleles (F ST-ENA ) was also estimated using the software FreeNA (Chapuis & Estoup, 2007). D JOST (Jost, 2008) and its significance were calculated using 1,000 permutations as implemented in GenAlEx 6.5 (Peakall & Smouse, 2012). F ST , F ST-ENA , and D JOST were also estimated within each sampled basins of the Atacama Desert (northern, central, and southern) and also as pairwise population genetic differentiation indices. Heat map of F ST pairwise comparisons between populations was conducted using the levelplot function of the lattice package (Sarkar, 2008) of the software R. Gene flow (Nm) was calculated assuming drift-migration equilibrium using the formula: (Slatkin & Barton, 1989 (Pritchard et al., 2000) and by the ΔK statistic (Evanno et al., 2005), based on the secondorder rate of change in the log-likelihood of data between successive K values.

| Correlation between genetic and geographic distance
In order to test isolation by distance (IBD), two pairwise population distance matrices (genetic distances and geographic distances) were constructed and compared through the Mantel tests with 9,999 permutations using the software GenAlEx 6.5 (Peakall & Smouse, 2012).
A pairwise, individual-by-individual (N × N) genetic distance matrix based on multilocus genotypes was calculated (GD). For a singlelocus analysis, with i-th, j-th, k-th, and i-th different alleles, a set of Peakall et al., 1995;Smouse & Peakall, 1999). Geographic distances were calculated from the coordinates of sampled localities.

| Genetic diversity and inbreeding coefficient F IS
Individuals sampled were determined as P. alba, P. chilensis, P. flexuosa, and P. nigra and just one putative hybrid was identified in ZAPI, QUIN, QUIS, YAYE, TULO, CAMA, and TILO (Table 1).
The analysis of linkage disequilibrium in the whole sample yielded significant results (p = .001), which are attributable to four populations (TILI, TARA, CANC, and VVJO). In the other nine populations, the association index was nonsignificant (Table 2); however, after applying Bonferroni's correction for multiple tests, linkage disequilibrium was significant in only one population suggesting that the eight loci markers were genetically independent.
The eight SSR loci analyzed in the populations were highly vari- In five populations, F IS was significant (according to their CI 95% ;   Table 4).

| Genetic differentiation
The genetic differentiation between populations from the same or different geographical basins showed the same tendency by F ST , F ST-ENA , and D JOST (Tables S1-S3), pairwise genetic differentiation expressed by F ST was small to moderate (F ST < 0.17), but a trend toward higher differentiation between populations from different basins was observed (Tables S1-S3; Figure 2 (Nei, 1978); F IS , inbreeding coefficient, F IS was considered significant (*) when zero is not included in 95% confidence intervals (CI 95% ). SD in parenthesis.

| Genetic structure among populations
The STRUCTURE analysis revealed an optimal number of subpopulations at K = 2 by the ΔK criteria and K = 4 by the mean loglikelihood of data for each value of K (Pritchard et al., 2000) in both no-admixture and admixture models performed ( Figure S1a,b, respectively). Consequently, we analyzed the structure distribution at K = 2 and 4 for both models (Figures 4 and S2, respectively).
According to Figure

| Correlation between genetic and geographic distance
We found significant correlation between genetic and geographical distances both when all sites were included (r = .37, p = .000) and when the analysis was performed considering the northern (r = .12, p = .020), central (r = .29, p = .001), and southern (r = .15, p = .000) basin populations separately ( Figure 5). This result is consistent with the model of isolation by distance although the correlations obtained can be considered very low, especially within basins. Saidman & Vilardi, 1987). As a consequence of geographic isolation, it is expected that genetic diversity within populations declines, associated with genetic drift or inbreeding (Bessega et al., 2018;Grivet et al., 2008). However, it may be that the geographic isolation consequences have not yet produced detectable effects on genetic diversity due to the long generation time of Prosopis species.

| D ISCUSS I ON
Genetic diversity can be spatially structured at different levels, such as landscape, population, or between nearby individuals, due to different ecological process of habitat characteristics such as population density and community structure operating in natural populations (Zeng et al., 2012). at the uppermost hierarchical level as populations from northern, central, and southern basins were not grouped forming clusters. In agreement, low levels of variation were explained by AMOVA between basins (7.40%). Isolation between the studied basins can be attributed to ancient biogeographical processes linked to aridity fluctuations and geomorphological events that may have driven biological differentiation (Baranzelli et al., 2014;Ossa et al., 2013;Viruel et al., 2012). Significant differentiation patterns between the northern and southern groups of P. chilensis population in the Coquimbo Region (Chile) were attributed to differentiation dated more than 1 million years ago that has been blurred by more recent gene flow (Moncada et al., 2019). A regional examination of the Atacama Pacific Paleosurface (Evenstar et al., 2017)  When variation distribution is analyzed at lower hierarchical levels, most of the variation (79.3%) was found at the lowest level (individuals), which is expected in outcrossing species as Prosopis (Bessega et al., 2019;Chequer Charan et al., 2020;Roser et al., 2017). However, the AMOVA indicated that the genetic F I G U R E 5 Pairwise genetic distances and geographical distances obtained from SSR considering all (a), the northern (b), the central (c), and the southern (d) populations studied in the Atacama Desert variation between populations within basins was low and significant (7.47%) and almost the same as the percentage obtained considering the between-basin component (7.4%). But the migrants per generation estimations within basins were up to four times higher than the Nm estimate obtained for the total basins (Nm = 2.6, 6.9, and 2.5 within northern, central, and southern basins, respectively).
Although Mantel's tests were also significant, indicating that nearby populations tend to be genetically more similar than expected by chance and suggesting gene flow restriction both at short and long distances, the low correlation estimates suggest a high degree of gene flow occurring within basins. According to these results, the geographic isolation that occurs between the populations within each of the basins seems to be able to be crossed by the natural dispersers of Prosopis easier than the large distances that separate the different basins considered here. These results are compatible with endozoic seed dispersal associated with short-distance spread described for Prosopis (Burkart, 1976;Keys, 1993;Mares et al., 1977;Reynolds, 1954). Although native vectors for Prosopis have not been exhaustively studied in the area, these might be foxes, small rodents, and birds (Bessega et al., 2006;Burkart, 1976;Campos & Ojeda, 1997;Carevic et al., 2019;Maldonado et al., 2014 (Shimada & Shimada, 1985). and finally results in inbreeding depression (Frankham et al., 2002).
Here, no general trend to heterozygote excess or deficiency was de-  (Correa & García, 2014;Molina, 2017;Núñez, 1976;Núñez & Briones, 2017). Consequently, there are various species living together in the oases and an unnatural gene flow pattern is produced.
The admixture pattern detected in the central basin is also compatible with its position within the Loa River course, which functioned as a sociocultural corridor throughout millennia. Quillagua oasis has been defined as a node of interregional interchange, whereas caravans and humans often passed in the transit between the coast and the highlands (Agüero et al., 2006;Briones, 2006;Berenguer & Pimentel, 2017;Cases & Montt, 2013;Gallardo, 2017;Martínez, 1998;Pimentel et al., 2017;Sanhueza, 1992). During historical times, the miner industry heavily disturbed the landscape especially within the northern basin. This industry reduced the use of algarrobo to coal and wood, but also to forage for European herbivores, especially mules which transit thoroughly from the coast to beyond the Andes (Carmona, 2018). These animals might have favored the encroachment of some species within oases (Brown & Archer, 1989); however, we cannot rule out that the patterns found were produced also in pre-Columbian times, where a long history of interchange has been acknowledged from archaeological evidences (Uribe et al., 2020).
Historically, Atacama populations benefit from both algarrobo and chañar (Geoffroea decorticans) trees (Martínez, 1998). A recent study by Contreras et al. (2018) on the genetic structure of eight Geoffroea decorticans populations from the Atacama Desert has concluded that at least two different origins could explain the genomic differences in chañar populations in northern Chile. Similarly, if we consider the ΔK criteria, our structure results let us propose that two origins are also possible for algarrobo in Atacama (K = 2). A first approximation could be to link the two genetic groups and the altitude at which the different populations studied are found (Table 1).
However, the simple observation of the altitudes and the genetic groups rules out this possibility. A plausible proposition to consider may be that northern basin may have received influence of populations from Bolivia and/or Peru, whereas southern and central samples could have been influenced by Argentinean populations despite the geographical Andes barrier. Although future studies may confirm this proposal, this interpretation is supported by archeological, ethnographical, and ethnohistorical data that show that in the Loa River and Salar de Atacama basin, people were interdigitated with communities from South Lipez (Bolivia Highlands), although communities from the Salar the Atacama basin were also connected and interdigitated with northwestern Argentina, especially from Toconao to the south (Hidalgo, 1995;Martínez, 1998).
In summary, we have assessed the genetic diversity of Prosopis populations (Section Algarobia) located in the oases of the Atacama Desert considering the genetic variability and its distribution, genetic differentiation, gene flow among populations, IBD patterns, and genetic structure. We have discussed the genetic patterns in reference to ecological, historical, and sociocultural characteristics.
Based on the analyses of genetic parameters, we propose that the genetic pattern may be the consequence of natural gene flow together with anthropic transport of a wide variety of algarrobo.

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