The role of anthropogenic dispersal in shaping the distribution and genetic composition of a widespread North American tree species

Abstract Dispersal and colonization are among the most important ecological processes for species persistence as they allow species to track changing environmental conditions. During the last glacial maximum (LGM), many cold‐intolerant Northern Hemisphere plants retreated to southern glacial refugia. During subsequent warming periods, these species expanded their ranges northward. Interestingly, some tree species with limited seed dispersal migrated considerable distances after the LGM ~19,000 years before present (YBP). It has been hypothesized that indigenous peoples may have dispersed valued species, in some cases beyond the southern limits of the Laurentide Ice Sheet. To investigate this question, we employed a molecular genetics approach on a widespread North American understory tree species whose fruit was valued by indigenous peoples. Twenty putative anthropogenic (near pre‐Columbian habitations) and 62 wild populations of Asimina triloba (pawpaw), which produces the largest edible fruit of any North American tree, were genetically assayed with nine microsatellite loci. Putative anthropogenic populations were characterized by reduced genetic diversity and greater excess heterozygosity relative to wild populations. Anthropogenic populations in regions that were glaciated during the LGM had profiles consistent with founder effects and reduced gene flow, and shared rare alleles with wild populations hundreds of kilometers away (mean = 723 km). Some of the most compelling evidence for human‐mediated dispersal is that putative anthropogenic and wild populations sharing rare alleles were separated by significantly greater distances (mean = 695 km) than wild populations sharing rare alleles (mean = 607 km; p = .014). Collectively, the genetic data suggest that long‐distance dispersal played an important role in the distribution of pawpaw and is consistent with the hypothesized role of indigenous peoples.


| INTRODUC TI ON
Dispersal and colonization are critical ecological processes that affect the ability of organisms to undergo range shifts as they track a changing environment and thus to persist over evolutionary time scales. During ice age glacier advances, most cold-intolerant Northern Hemisphere plant species retreated to southern refugia and during subsequent warming periods there was extensive northward migration and often range expansion (Davis & Shaw, 2001;Hewitt, 2000;Lyford et al., 2003). Understanding historical patterns of plant dispersal and mechanisms responsible for movement during the warming climate that ensued after the last glacial maximum (LGM; Clark et al., 2009) allows inferences and predictions regarding their capacity for range modification in response to future climatic shifts.
Sessile plants disperse their seeds by abiotic means (i.e., barochory, hydrochory, or anemochory) and/or biotic vectors (epiand endozoochory) with some species having more than one dispersal mechanism. Interestingly, some taxa with limited seed dispersal, such as large-seeded tree species without an obvious dispersal vector, appear to have migrated considerable distances after glacial retreat in the late Pleistocene (~19,000 years before present [YBP]), a phenomenon known as "Reid's Paradox" (Clark et al., 1998;MacDougall, 2003). In 1899, Clement Reid was the first of many to note the surprisingly rapid rate of northward migration from glacial refugia by large-seeded trees, such as oaks, into formerly glaciated regions of Great Britain (Clark et al., 1998).
Some have hypothesized that humans may have mediated dispersal of valued tree species beyond the limits of their southern glacial refugia (Abrams & Nowacki, 2008;Delcourt & Delcourt, 2004;MacDougall, 2003).
There is considerable evidence that pre-Columbian peoples had a marked impact on the North American landscape prior to European arrival (Abrams & Nowacki, 2008;Denevan, 1992;Pyne, 1983;Stewart, 2002). However, little attention has been paid to the possible role of native peoples in the dispersal of useful plant species (but see Abrams & Nowacki, 2008;Cottrell et al., 2002;Delcourt & Delcourt, 2004;MacDougall, 2003;Petit et al., 2002) and some population genetic studies have intentionally avoided species valued or impacted by humans (Petit et al., 2003). One of the few exceptions involving a North American tree species used genetic markers and packrat middens (pollen and macrofossils) to investigate an isolated (>200 km) northern population of Pinus edulis (Colorado pinyon pine: Bentancourt et al., 1991). The authors concluded that this isolated population originated by long-distance dispersal by Native Americans who valued pine nuts as a food source. Additional phytogeographic evidence of human-mediated dispersal in North America comes from tobacco (Nicotiana sp.) and bottle gourd (Lagenaria siceraria; Asch, 1995). Both species are thought to have been absent from the area presently known as Illinois (U.S.A.) until their introduction by external contacts and subsequent cultivation as useful specialty-plant species. Wykoff (1991) asserted that black walnut (Juglans nigra), some hickories (Carya spp.), oaks (Quercus spp.), and medicinal plant species were likely introduced into what is now recognized as New York state (U.S.A.) prior to European contact.
At least 15,000 YBP an ice-free corridor connected Alaska and the rest of the continental United States, permitting human migration into this region. Earlier arrival of humans in North America ~16,000 YBP could also have been possible along the North Pacific coast (Erlandson et al., 2015). It is generally agreed that indigenous peoples inhabiting North America manipulated their environment to manage resources soon after their arrival ~15,000 (Delcourt & Delcourt, 2004;Morse & Morse, 1983) to 17,000 YBP (Adovasio et al., 1978;Richland et al., 2007); however, debate remains about the uniformity of their influence over space and time (Muñoz et al., 2014).
While Denevan (2011) estimated two million indigenous people lived in North America in 1492, when Europeans commenced colonization, there is much controversy over the extent of anthropogenic impact on the present distribution of plant species. Some investigators maintain that impacts by indigenous peoples were ubiquitous, with few ecosystems unaffected (Abrams & Nowacki, 2008;Denevan, 1992;Kay, 2002;Krech, 1999). However, after the introduction of Old World diseases by Spanish explorers and the resulting decimation of indigenous populations (Lovell, 1992), North American ecosystems recovered somewhat, which may have obscured indigenous imprints on the landscape (Denevan, 1992).
Others have argued that landscape alteration near densely settled areas showed greater impacts, while more remote, sparsely inhabited areas exhibited little anthropogenic alteration (Muñoz et al., 2014;Parker, 2002;Vale, 1998Vale, , 2002. Thus, plants used by indigenous people may show signs of manipulation in proximity to settlements but may exist in an un-manipulated state farther away (e.g., Parker et al., 2014). Muñoz et al. (2014) contended that agricultural and silvicultural impacts were greatest near settlements, as well as along riparian corridors and along trade routes. By approximately 2,000 YBP, indigenous societies throughout the river valleys of Eastern North America were interacting widely, erecting monuments, and producing sophisticated material cultures (Wright, 2017). agricultural fields associated with settlements were established in the floodplains of larger rivers in the Midwest and the south (Fritz, 1990;Scarry & Scarry, 2005). Some cultures relied heavily on nutritious nuts (i.e., balanocultures) of mast-producing species such as chestnut, hickory, and oaks and are thought to have planted orchards and manipulated the distribution of some tree species (Abrams & Nowacki, 2008). Acorns served as a particularly important dietary staple for many indigenous peoples (Bainbridge, 1985;Logan, 2005), and widespread acorn use almost certainly predates widespread corn use (Bainbridge, 1985). There were many other crops that were cultivated as well (e.g., Delcourt & Delcourt, 2004;Mueller, 2017;Schroeder, 1999). Inadvertent or intentional anthropogenic seed dispersal, particularly in heavily impacted portions of the landscape, may have contributed to postglacial maximum range expansion (e.g., Abrams & Nowacki, 2008;Keener & Kuhns, 1998;White, 1906;Wykoff, 1991) and help explain Reid's Paradox. It is important that this question be investigated in more depth in order to develop a better understanding of plant dispersal and factors that have shaped the distribution of plant taxa.
Asimina triloba (Annonaceae), whose common name is pawpaw or Indian banana, has the largest (5-16 cm long, 3-7 cm wide) edible fruits of any native North American tree. The flavorful fruits (described as a cross between banana, mango, and pineapple) were prized as a food source by native peoples (e.g., Brooks & Johannes, 1990;Keener & Kuhns, 1998;Peterson, 1991), and the fibrous bark was used to make rope and cloth (Peattie, 1950). It may also have been utilized medicinally (Krochmal & Krochmal, 1973;Peterson, 1991). In recent years, there has been considerable research on promising anti-cancer properties of A. triloba seeds and bark (e.g., Zhao et al., 1993).
Seeds or carbonized remains of A. triloba have been found along the Mississippi and Ohio Rivers and at numerous archaeological sites in Arkansas, Illinois, Indiana, Kansas, Kentucky, Mississippi, and Missouri (Blasing, 1986;Cutler & Blake, 1976;Gilmore, 1931;Jones, 1936;King, 1982;Waselkov, 1977;Wedel, 1943;Yarnell & Watson, 1966) and may have been dispersed by various cultures (Brooks & Johannes, 1990;Keener & Kuhns, 1998;Peterson, 1991;Wykoff, 1991). The first written record of A. triloba comes from Hernando de Soto's expedition in 1541 across Southeastern North America, which reported widespread planting of the tree by indigenous tribes of the southeast and mentions its flavor and fragrance (Pickering, 1879). Some authors have asserted that A. triloba was grown by pre-Columbian people (Cai et al., 2019;Hormaza, 2004;Peattie, 1950;Peterson, 1991), although there is no known evidence for the cultivation of A. triloba orchards by Native Americans beyond such written reports. There is however molecular evidence for pre-Columbian dispersal of Annona cherimola between Central and South America (Larranaga et al., 2017), a species that shares some attributes of A. triloba (i.e., a woody perennial that produces large, edible, and nutritious fruits).
It has been hypothesized that the remarkably wide distribution of pawpaw throughout much of eastern North America relative to its seven North American congeners, which are restricted to the southeastern United States, is due in part to its spread by indigenous peoples who may have actively traded propagules, encouraged its growth (e.g., Abrams & Nowacki, 2008;Keener & Kuhns, 1998;MacDougall, 2003), and/or inadvertently dispersed its seeds (White, 1906). White (1906) for example observed Native Americans using pawpaw seeds as game pieces, contending that this game is centuries old and that Sac and Fox Indians may have unintentionally dispersed the seeds. That the fruit cannot easily be stored or transported suggests the advantages of planting pawpaw seeds along well-traveled routes and/or near settlements.
Its occurrence north of the southern Pleistocene glacial boundary suggests that its post-LGM dispersal could have been humanmediated. While pawpaw was clearly used by indigenous peoples in North America, there has been debate regarding the relative importance of natural versus anthropogenic dispersal (Keener & Kuhns, 1998;Murphy, 2001) with more authors favoring an anthropogenic role (e.g., Abrams & Nowacki, 2008;Keener & Kuhns, 1998;MacDougall, 2003).
The overarching goal of our research is to employ a molecular genetics approach to test the hypothesis that indigenous pre-Columbian people contributed to the dispersal and range expansion of a valued North American understory tree. To address this question, we used genetic signatures contained within con-

| Study species
Asimina triloba (L.) Dunal (Annonaceae) is the most widespread member of the genus, occurring in 26 states in the eastern United States, ranging from Texas and Iowa to the eastern seaboard and north into southern Ontario, Canada (Little, 1977). This understory tree, with trunks ≤20 cm diameter at breast height (DBH; Wyatt, personal observation), generally occurs at elevations <350 m above sea level, primarily in mesic, alluvial forests along rivers and creeks (Freeman & Hulbert, 1985;Murphy, 2001;Pomper et al., 2010). Vegetative spreading occurs via root-suckers (Keener & Kuhns, 1998), resulting in dense patches and genets that persist for a long time (Hosaka et al., 2005). While there are no reports of longevity of individual ramets or genets, we documented 32 annual growth rings in one stem from Georgia with a DBH of 11.4 cm.
Flowers, which are strongly protogynous (Lagrange & Tramer, 1985), emerge in the spring after the last frost, and flowering on a given stem continues for approximately a month (Wyatt, personal observation). Flowering is also highly synchronous (Lagrange & Tramer, 1985). Asimina triloba pollinators are weak flying insects such as flies (including Drosophila spp.), beetles, and small Lepidoptera (Goodrich et al., 2006;Robertson, 1928;Willson & Schemske, 1980). Fruit set tends to be low (Lagrange & Tramer, 1985;Willson & Schemske, 1980), with only 0.4% of flowers and 15% of stems producing fruit in a central Illinois population (Willson & Schemske, 1980). Willson and Schemske (1980) found that fruit set was subject to pollen-limitation as well as resource limitation related to the size of the fruits. Asimina triloba produces the largest edible fruits (≤1 kg; Darrow, 1975) of any native North American tree. Lagrange and Tramer (1985) also report fruit set relative to the number of flowers ranging from 0.5% to 3.4% in Tennessee and 0% in Toledo, Ohio.
The fruit is consumed by various animals with seeds reported in the scat of raccoons (Murphy, 2001;Willson, 1993;Willson & Schemske, 1980;Yeager & Elder, 1945), opossums, red fox (Murphy, 2001;Willson, 1993), and coyotes (Cypher & Cypher, 1999; Wyatt, personal observation) suggesting that these animals may mediate seed dispersal. There are reports of black bear in Virginia and North Carolina eating the fruit but seeds have not been documented in scat (Willson, 1993). Turkeys also consume the fruits (Murphy, 2001) although the seeds are digested and destroyed in the process (Wyatt, personal observation). Deer are unlikely vectors as there is no evidence that they consume the fruit other than in captivity (Murphy, 2001;Wyatt, personal observation). It is believed that historically the fruit was primarily consumed by megafauna that are now extinct or extirpated from their former range (Janzen & Martin, 1982;Poor, 1984) and that these megafauna may have mediated long-distance dispersal of pawpaw. Mammoths America (Fields et al., 2012;Janzen & Martin, 1982;Poor, 1984) as recently as 12,700 YBP (Perrotti, 2018). Evidence suggests that horses and mastodons, for example, underwent annual migrations in Southeastern North America in excess of 150 km (Hoppe & Koch, 2007). Bison also migrated extensively before their numbers were decimated (Kauffman et al., 2021).

| Sampling and genotyping
Leaf tissue samples were collected from 82 mapped populations of A. triloba distributed throughout the species' range (Tables 1 and 2;   Table S1; Figure 1). Populations were separated from one another by a mean of 681.3 km (range = 0.2 to 1,845.8 km). Twenty populations were classified as of putative anthropogenic origin based on their location <1 km from documented pre-Columbian habitations (i.e., villages or mound sites). We selected this objective criterion based on Muñoz et al. (2014)

who indicate that impacts on plants by indigenous North
Americans were localized around settlements as well as along travel corridors. Proximity to pre-Columbian settlements increases the likelihood that populations are anthropogenic, but anthropogenic origins become less probable with increasing distance from former settlements. Six of the 20 putative anthropogenic populations (LM1, LM2, LM3, MC1, MC2, and MC3) are located near archeological sites where seeds or carbonized remains of A. triloba have been found (Keener & Kuhns, 1998;Yarnell & Watson, 1966). Sixty-two populations were designated as "wild" as there are no known indigenous sites or trade routes located within 1 km of these populations. We compared the elevation of putative anthropogenic and wild populations and tested for significance with a two-sample t test assuming equal variances.
Leaf samples were collected from a mean of 28.1 individuals (range = 5-50) from each anthropogenic population and a mean of 29.6 individuals (range = 10-62) from each wild population (Table 2).
Samples were taken from mature (≥3 m in height) stems separated by ≥10 m when possible to avoid sampling multiple ramets belonging to the same genet. Samples were snap-frozen in liquid nitrogen and transported to the University of Georgia for genetic analyses.
Because A. triloba propagates vegetatively, we tested whether duplicate multilocus genotypes (MLGs) within populations represented clones. To estimate the likelihood of two individuals within a population having identical MLGs by chance, we calculated the probability of identity (PI) for each population which reflects the number of loci, allele frequencies, and sample size. The multilocus probability of identity (PI m ) was calculated as PI m = Π s (PI s ), where and p i and p j are the frequencies of the ith and jth alleles at a locus, respectively (Gonzales et al., 2008;Paetkau & Strobeck, 1994). The (Parker et al., 1998). Duplicate MLGs that represented clones were removed, leaving one individual per genet for subsequent analyses.

PI was adjusted for sample size (N) by PI
The number of MLGs and the number of MLGs adjusted for population sample size (sMLG; i.e., rarefaction) were calculated using Fstat (Goudet, 1995(Goudet, , 2001. For the rarefaction analysis, all populations were treated as consisting of 10 individuals to reflect the number of individuals in the smallest population after SEL A (N = 5), LM2 A (N = 6), and STO W (N = 9) were omitted. Genetic diversity statistics were estimated separately for putative anthropogenic populations and wild populations using GenAlEx v. 6.51b2 (Peakall & Smouse, 2006 Wright, 1922Wright, , 1951. Deviations from Hardy-Weinberg expectations were tested for significance using chi-square (Li & Horvitz, 1953).
A Bonferroni correction for multiple comparisons was applied using FSTAT (Goudet, 1995(Goudet, , 2001. Correlations between MLGs and F IS values, as well as between latitude and F IS values, were tested for significance using linear regression analyses. Genetic structure among populations was estimated using Nei's (1973) G ST . Pairwise G ST values were calculated for all possible pairs of populations in GenAlEx.v. 6.5.1b2 (Peakall & Smouse, 2006. A two-sample t test assuming equal variances was used to test for significant differences between mean pairwise G ST values and mean pairwise geographic distances among anthropogenic populations versus wild populations.  Bennett (1986) For the nine putative anthropogenic populations located at, or within, the boundary of the Wisconsin glaciation we used a twosample t test assuming equal variances to assess significance be- Isolation by distance (IBD; Wright, 1943) was estimated using linear regression analysis of G ST / (1 − G ST ) against log geographic distance (Rousset, 1997) for all 82 populations as well as separately for anthropogenic and wild populations. Significance was tested using a Mantel test in GenAlEx.v. 6.5.1b2 (Peakall & Smouse, 2006. structure v. 2.3.4 (Pritchard et al., 2000), a Bayesian clustering approach, was used to estimate levels of genetic admixture among all 82 populations and the number of genetically distinct clusters (K).
Levels of admixture and K were also estimated separately for putative anthropogenic populations as a part of the present study and wild populations as reported by Wyatt et al. (2021). Ten independent simulations at each K-value from 1 to 20 were run, using a burn-in of 500,000 repetitions and a run length of 1,000,000 Markov chain Monte Carlo (MCMC) iterations. The admixture model was chosen to infer alpha (α). We also employed a model based on correlated allele frequencies with no a priori assumptions regarding sampling locations. The Evanno et al. (2005) method was used to determine the optimal number of genetic clusters (K) in Structure Harvester (Earl & vonHoldt, 2012).
The occurrence of rare alleles in <10% of all wild and putative anthropogenic populations (i.e., alleles present in 2-8 populations) was used to more closely investigate relationships between putative anthropogenic and wild populations. A two-sample t test assuming equal variances was used to assess whether the number of rare alleles in anthropogenic versus wild populations differed significantly.
We also used linear regression analyses to test whether there was a significant relationship between latitude and the number of rare alleles in anthropogenic and wild populations.
We were particularly interested in examining rare alleles shared by putative anthropogenic and wild populations. We estimated pair-

| Fruit set
To better evaluate whether populations of putative anthropogenic origin and populations at the edge of the species range exist beyond or at the edge of their ecological limit, we collected fruit set data.
For a species to persist in a particular landscape, individuals must be able to survive and reproduce.   Table S1).

| Genetic diversity
Null allele frequencies were nonsignificant for all loci and populations (p = .05). Mean probability of identity (PI) for anthropogenic and wild populations was 1.46 × 10 -4 and 1.15 × 10 -3 , respectively (  Figure S1). When adjusted for sample size, the mean number of sMLGs was also significantly lower for anthropogenic populations (4.6 and 7.0, respectively; p = .029; A total of 160 alleles occurred across all 82 populations (Table 2) Table 2).
F IS values were highly significant for 50% of anthropogenic populations (p < .005) and significant for 19% wild populations (p < .05; Table 2). For nearly all populations with an inbreeding coefficient that differed significantly from HWE, there was excess heterozygosity. There was a significant positive correlation between MLGs and     (Table 3;   Table S4; Figure 2; Figure S2). The next most likely number of clusters was K = 4 with mean Ln P(D) of −63,063.9 and ΔK of 8.5 (Table 3;

Figures S2 and S3). Caution is required in interpretation of Structure
results for all 82 populations as well as the 62 wild populations as the presence of significant IBD violates one of the underlying assumptions of Structure (Pritchard et al., 2000). Simulations for the 20 putative anthropogenic populations also found K = 2 to be optimal, but the clusters are not geographically distinct: populations with a higher genetic assignment to one cluster are often geographically embedded in the other cluster (Figure 3).

| Rare alleles
A total of 54 rare alleles occurred in 75% (15) of anthropogenic and 81% (50) of wild populations. Only 24 rare alleles were shared by both anthropogenic and wild populations, and no rare alleles occurred exclusively in anthropogenic populations ( Figure S4). There were also fewer rare alleles in anthropogenic populations (mean = 1.9 vs. 2.9; p = .071). The number of rare alleles per population decreased significantly with increasing latitude for both anthropogenic (r = −0.679; p = .001) and wild populations (r = −0.302; p = .017; Figure 4).
As predicted, the pairwise G ST values between wild populations that share rare alleles (W RA ) were significantly smaller than for all wild populations (W; 0.166 and 0.228, respectively; p = 1.37 × 10 -23 ) and pairwise geographic distances for W RA were significantly smaller (607.1 and 660.3 km, respectively; p = .014; Table 4). Wild and anthropogenic populations that shared rare alleles (W -A) RA had a significantly higher pairwise G ST values than W RA (0.185 vs. 0.166; p = .012) but were also separated by significantly larger pairwise geographic distances (695.3 vs. 607.1 km; p = .014; Table 4).
Anthropogenic wild populations with which they shared rare alleles were separated by a larger mean pairwise distance (723.1 vs. 687.7 km; p = 0.345) and had a significantly higher mean G ST (0.216 vs. 0.176; p = 0.018) (Table 4) than anthropogenic populations in areas not formerly glaciated and wild populations that share rare alleles.
Regarding the inverse relationship between genetic diversity and excess heterozygosity, wild populations of Fagus in Europe reveal a similar pattern (Comps et al., 2001). Excess heterozygosity appears to be correlated with asexual reproduction (Ballou et al., 2003;Birky, 1996;Bryzski & Culley, 2011;Meloni et al., 2013) which agrees with our finding of significantly higher clonality and excess heterozygosity in anthropogenic populations of A. triloba (Table 2).
Excess heterozygosity may result from dominance of well-adapted, highly heterozygous genotypes or the introduction of a few seeds from multiple genetically divergent source populations. If humans intentionally transported the fruits or seeds from wild populations, it is likely that propagules were selected on the basis of favorable traits (i.e., fruit production, fruit size, flavor, and/or fiber quality; Peattie, 1950;Peterson, 1991) possibly associated with higher heterozygosity.
The significant decline in F IS values with increasing latitude in anthropogenic populations is also consistent with studies showing that populations at the edge of a species' range, where they are at their ecological limit, tend to have more asexual reproduction (Billingham et al., 2003;Dorken & Eckert, 2001;Eckert, 2002;Garcia et al., 2000;Silvertown, 2008). That this inverse relationship was highly significant in anthropogenic populations but not in wild populations supports the hypothesis that the more northern populations resulted from anthropogenic dispersal into habitat at A. triloba's ecological limit. The low fruit set documented in these northern populations is consistent with low rates of gene flow and possible marginal ecological conditions for A. triloba. White (1906) also reported that flower and fruit production in A. triloba decreased with increasing latitude north of the Ohio River mouth (Cairo, Illinois) and that its vegetative growth at the northern edge of its contemporary range is beyond its fruiting limit. Murphy (2001) further mentions the reduced size of A. triloba in northern populations, although he attributes this to possible overgrazing by a deer population that has been unnaturally large since 1912 (Bartlett, 1958). ~12,000 YBP (Janzen & Martin, 1982;Perrotti, 2018;Poor, 1984).   (Table 4).
Anthropogenic populations in formerly glaciated landscapes and the wild populations with which they share rare alleles, (A G -W) RA are separated by the largest mean geographic distance investigated (Table 4).

Subsequent to extinction of North American megafauna and the
LGM, the primary nonanthropogenic candidates for dispersal are waterways, raccoons, opossums, turkeys, red foxes, and coyotes.
However, hydrochory cannot explain the northward migration of A.
triloba. Many of the anthropogenic populations occur west of the Appalachians along the Mississippi/Ohio river drainages where the directionality of water flow is from the northeast to the southwest.
While this makes hydrochory an unlikely mechanism of northward dispersal, the distribution of anthropogenic populations is consistent with known pre-Columbian trade routes (e.g., Forsberg, 2003;Muñoz et al., 2014). Major rivers east of the Appalachians flow from west to east or northwest to southeast, also precluding the possibility of northward dispersal by hydrochory. While raccoons consume pawpaw, they are unlikely to mediate long-distance dispersal due to their generally small home ranges of 1-3 km (maximum = 6.4 km; Stuewer, 1943;Butterfield, 1944;Feldhamer et al., 1982). Opossums are also unlikely long-distance seed dispersers with average home ranges of 64.4 ha and 141.6 ha for females and males, respectively, and average nightly movements of 1,465 m (females) and 1,835 m (males; Ryser, 1995 In summary, this study represents the first investigation into human-mediated dispersal of a useful understory tree species in Eastern North America using genetic markers. Perhaps the most salient finding was that nearly all genetic diversity statistics are higher, although not significantly so, in wild populations than putative anthropogenic populations, except H O which is significantly lower in wild populations. Lower genetic diversity in anthropogenic populations of A. triloba is likely associated with founder events and dispersal of seeds or ramets from wild populations to areas isolated and/ or less ecologically suitable. The high levels of genetic structure observed in northern anthropogenic populations suggest independent dispersal and colonization events from southern source populations and low levels of subsequent gene flow between these populations, which is consistent with the expectation of human-mediated dispersal into northern habitats. As we learn more about movement and trade practices of indigenous people in Eastern North America, it has become increasingly evident that they were exchanging goods over distances of 1,000 km as early as the Archaic Period (~8,000-3,000 YBP; Sanger et al., 2018). The preponderance of anthropogenic populations along major river valleys is consistent with known trade corridors and continental-scale exchange along large mid-continental rivers including the Mississippi and Ohio (e.g., Fowler, 1975;Smith, 2010).
Given the active and widespread trade practices of indigenous peoples of Eastern North America as well as the duration of such practices, human-mediated dispersal is a viable explanation for the patterns of genetic variation we detected. murpheyi (Parker et al., 2007), which are believed to have been introduced into Arizona by pre-Columbian peoples because of their value as a reliable source of food, fiber, and medicine. Both species are associated with indigenous settlements that were abandoned centuries ago due to a drying climate and have persisted through clonal spreading with no evidence of sexual reproduction (Parker et al., 2007). It is thought that Arizona is outside the normal range of these species, precluding sexual reproduction. While the low pawpaw fruit set we documented represents a snapshot of one reproductive season, it is consistent with the complete absence of A.

Populations of
triloba seeds in refuse middens from Archaic sites around Lake Erie and northward (Keener & Kuhns, 1998). Such absence is consistent with the arrival of A. triloba at these locations more recently than 3,000 YBP and/or with the persistence of these populations primarily through asexual reproduction. This work elucidates the possible origins of populations of a widespread Eastern North American tree species at the northern limit of its range. Furthermore, closer examination of populations of A. triloba that are designated as wild for genetic profiles more similar to anthropogenic populations may reveal hitherto undetected locations of pre-Columbian settlements and trade routes, thus furthering our understanding of indigenous cultures (e.g., Parker et al., 2014).

ACK N OWLED G M ENTS
This research was supported by an award to DWT by the University of Georgia Office of the Vice President for Research.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest. Writing-review & editing (lead).

DATA AVA I L A B I L I T Y S TAT E M E N T
Genetic data are accessible at Dryad (https://doi.org/10.5061/ dryad.5x69p 8d3g).