Southeastern Naturalist P.A. Smallwood, M.D. Caspary, and J.E. Russell 2018 Vol. 17, No. 3 456 2018 SOUTHEASTERN NATURALIST 17(3):456–469 Investigation of Population Structure in the Rare Amsonia ludoviciana Vail (Louisiana Bluestar; Apocynaceae) Patrick A. Smallwood1,*, Melissa D. Caspary2, and James E. Russell2 Abstract - Amsonia ludoviciana (Louisiana Bluestar), in the Apocynaceae family, is a rare herbaceous perennial plant species found in the ecotones of granite rock outcrops of Georgia as well as the pine flatwoods in Louisiana, Mississippi, and Texas. This research focused on using fluorescently tagged primers to amplify polymorphic microsatellite loci as a way to characterize the population genetic structure of the rare Louisiana Bluestar in Georgia. We sampled 6 populations across the Georgia range of Louisiana Bluestar. We employed 8 microsatellite loci to detect variation within and among Louisiana Bluestar populations and to detect patterns in population structure and dispersal. Values of the genetic-diversity estimator Gst suggested that the majority of the sampled populations demonstrated moderate genetic similarity. Further, Geneland software indicated the presence of 3 clusters. From our population genetic analyses, it can also be inferred that there is substantial gene flow between populations of Louisiana Bluestar in the state of Georgia. Introduction Amsonia ludoviciana Vail (Louisiana Bluestar) is a rare herbaceous perennial in the Apocynaceae family known to occur in the shallow soils and open habitat in the ecotones of mafic Lithonia granitic gneiss outcrops of the Piedmont of Georgia (Weakley 2012) and Pinus (pine) flatwoods in Louisiana, Mississippi, and Texas (Doffitt et al. 2014, Lemke 1987, Weakley 2012). In Georgia, rock outcrops are characterized by high solar-irradiance, extreme temperatures, xeric substrates, and an island-like geographic distribution that likely contributes to the radiation of many of the endemic species typical of the habitat (Quarterman et al. 1993, Shure 1999, Wyatt and Allison 2000). Louisiana Bluestar occurs in the ecotone between rock outcrop and forested habitat, growing in the thin soils and diffuse light of the rock margin where water seeps at the rock edge (Allison 2013; Edwards et al. 2013; M.D. Caspary, pers. observ.). Historic occurrences of Louisiana Bluestar are reported in South Carolina, but it is presently only found in Louisiana, Mississippi, Texas, and Georgia (Doffitt et al. 2014, Estill and Cruzan 1999, Weakley 2012). In the state of Georgia, Louisiana Bluestar is known to occur across 33 sites in Dekalb, Rockdale, and Walton counties, excluding historic occurrences documented in Gwinnett County (Fig. 1; Allison 2013, GADNR 2013). All populations within Georgia are located within a 25-km2 cluster. 1Department of Plant Biology, University of Georgia, Miller Plant Sciences Building, Athens, GA 30601. 2School of Science and Technology, Georgia Gwinnett College, 1000 University Center Lane, Lawrenceville, GA 30043. *Corresponding author - Patrick.Smallwood25@uga.edu. Manuscript Editor: Scott Markwith Southeastern Naturalist 457 P.A. Smallwood, M.D. Caspary, and J.E. Russell 2018 Vol. 17, No. 3 The ecotone-associated flora for many outcrops has shifted over time. Fire suppression compounded by exotic plant invasion and anthropogenic impacts are creating a dense understory layer of high competition and low light that discourage establishment and persistence of native outcrop flora (Caspary 2011). Research has shown that narrow endemics, such as the plant species that have specialized on mafic Lithonia gneiss, are particularly vulnerable to extinction and are presently being threatened by adjacent development, conversion to other land-use cover, invasive species, dumping, fire-building, quarrying, recreational impacts, road frontage, vehicular traffic, and grazing by farm animals (Allison 2013, Dury 1974). Most known populations of Louisiana Bluestar are small, from a few plants to tens of individuals (Allison 2013), which may compromise long-term maintenance of genetic variation. Additionally, wind and water are the dominant mechanisms for fruit and seed dispersal on rock outcrops, so dispersal events are typically highly localized around these habitats (Wyatt and Fowler 1977). The Louisiana Bluestar has a pair of long, pubescent, dehiscent follicles (Woodson 1928). Once ripened, the follicle releases its seeds, which are then predominantly gravity dispersed. This fruit type likely leads to a low level of genetic connectivity between populations of Louisiana Bluestar, which could result in inbreeding depression. The island-like geographic distribution and the small size of granite outcrops, may also influence the level of genetic variability within endemic species (MacArthur and Wilson 1967). The pollinator of the species is not known. The blue petals and a 5–9-mm tubular salverform corolla shape (Woodson 1928) indicate a pollination syndrome targeting bees and butterflies. Allison (2013) documented Eurytides marcellus Kramer Figure 1. Map of the range of Amsonia ludoviciana (Louisiana Bluestar) within the state of Georgia. (A) Map of Georgia with the dotted line representing state boarders. (B) Zoomed in map (GADNR 2013). Each square area represents ¼ of a USGS 7.5-minute quadrangle map. Light gray squares indicate 20 years since the last reported sighting of the species, while the dark gray squares have reports of the species presenc e in the last 5 years. Southeastern Naturalist P.A. Smallwood, M.D. Caspary, and J.E. Russell 2018 Vol. 17, No. 3 458 (Zebra Swallowtail) and other butterflies nectaring on the Louisiana Bluestars in Georgia, however it is not known if these plants require a specialist pollinator or rely more heavily on a suite of generalists to serve as vectors for pollen transfer. To our knowledge, the rate of selfing within the species has also never been investigated in this genus, but evidence in related species suggest that selfing rates are expected to be low. Post-zygotic self-incompatibility has been noted within some species in Apocyanaceae, particularly those with evolutionary adaptive traits towards pollen massing (Endress et al. 2007, Wyatt et al. 1998). A previous study investigating microsatellite diversity within a single Apocyanaceae species found observed frequencies of heterozygosity for polymorphic loci ranged from 0.50 to 0.80, except for a single locus which had an observed value of 0.20 (Topinka et al. 2004). This observation may be suggestive of a low rate of selfing when compared to average expected values of plants with a similar life history (gravity dispersed, short lived, eudicot) (Hamrick and Godt 1996). Information about existing threats and population genetic variability of rare plants like Louisiana Bluestar are needed to understand potential long-term survivability of species native to such landscapes and to direct conservation efforts. To this end, the goal of this study was to gain a deeper understanding of genetic diversity and genetic structure within the Georgian populations of the rare Louisiana Bluestar. Methods Plant material collection We selected 6 locations in Rockdale and Walton counties (POP1, POP2, POP3, POP4, POP5, and POP6) as representative study sites across the present-day species range and because these sites had moderate to large, accessible populations of Louisiana Bluestar (10–50 plants). Historically, Louisiana Bluestar was documented in Rockdale, Dekalb, Walton, and Gwinnett counties in Georgia. However, the majority of plants in the present-day distribution are concentrated in Rockdale County, 4 km to the southeast of the historic distribution. All Gwinnett County sites are suspected to be extirpated and the Dekalb County sites are small and were inaccessible at the time of sampling. The distance between sites varied from 2.04 km to 14.60 km. Due to the small size of populations (Table 1), we sampled 12 individuals from each location using 2 medium-sized leaves from each individual. We collected samples during the summer months of 2014. We stored at 4 °C all leaves obtained for DNA extraction until extraction procedures could be carried out. If the extraction procedure was not performed within a month of collection, we stored the sample at -20 °C until the leaves could be processed. DNA extraction We employed the GeneJET Plant Genomic DNA Purification Mini Kit (Thermo Scientific, Waltham, MA) for extraction of genomic DNA (gDNA) from collected leaf material. We followed the provided procedure with the following modifications: liquid nitrogen was not used in the initial grinding of plant material. We placed the leaf sample in a 1.5-ml centrifuge tube, added the provided lysis buffer A Southeastern Naturalist 459 P.A. Smallwood, M.D. Caspary, and J.E. Russell 2018 Vol. 17, No. 3 and lysis buffer B, and ground the sample using the prepared sterilized glass pestle until the solution had taken on a green color. We centrifuged all samples at 17,000 x g during steps 5 and 9 of the DNA-extraction kit protocol. PCR amplification of target microsatellites We selected a total of 8 microsatellite loci (Ake2, Ake3, Ake5, Ake6, Ake7, Ake8, Ake10, and Ake11; see Table 2) due to demonstrated polymorphisms in other Amsonia species (Topinka et al. 2004). We employed PrimeSTAR® HS DNA polymerase (Takara Bio, Mountain View, CA) to perform all PCR reactions. Each reaction contained the following: 1X PrimeSTAR Buffer (5-mM Mg2+), 0.2 mM of each dNTP, 0.2-μM concentration for the respective forward and reverse primers, Table 1. Population sizes of each sampling site (small = less than 15 plants, medium = 15–20 plants, and large = more than 20 plants observed.), observed heterozygosity (Ho), expected heterozygosity (He) (P-value of Ho = He using a 1-sample proportions test (POP1 = 0.121, POP2 = 0.802, POP3 = 0.783, POP4 = 0.091, POP5 = 0.028, POP6 = 0.045), the calculated Fis values with 95% confidence intervals, the total number of private alleles found at each sampling site, the computed allelic richness (Ar) and the 95% confidence interval of Ar of each site, and the effective number of alleles (Ae) along with its standard error. * = 95% confidence interval does not include 0. ** = P-value of a 1-sample proportion test of He = Ho is < 0.05. Population Fis Private Ar Ae Site size Ho He (95% CI) alleles (95% CI) (± SE) POP1 Large 0.479 0.489 0.006 3 3.15 2.217 (-0.203, 0.132) (2.75, 3.50) (± 0.283) POP2 Medium 0.458 0.454 -0.005 4 3.26 2.107 (-0.205, 0.124) (2.75, 3.63) (± 0.254) POP3 Medium 0.464 0.458 0.017 0 3.07 2.096 (-0.181, 0.118) (2.63, 3.38) (± 0.289) POP4 Small 0.552 0.544 0.061 9 4.11 2.453 (-0.278, 0.196) (3.25, 4.75) (± 0.296) POP5 Small 0.300 0.355** 0.129 6 2.66 1.657 (-0.056, 0.234) (2.00, 3.13) (± 0.151) POP6 Large 0.522 0.484** -0.095 8 3.16 2.306 (-0.247, -0.014)* (2.63, 3.63) (± 0.430) Table 2. The number of alleles (Na), effective number of alleles (Nae), the average size of produced amplicons for each study locus, and annealing temperature used. Locus Na Nae Amplicon size (min–max in bp) Annealing temp. (°C) Ake2 9 2.165 (± 0.322) 200–225 56.2 Ake3 6 2.165 (± 0.215) 120–222 56.2 Ake5 7 1.255 (± 0.081) 138–274 60.9 Ake6 10 1.879 (± 0.215) 204–242 60.9 Ake7 10 2.493 (± 0.275) 193–223 60.9 Ake8 8 2.325 (± 0.143) 95–111 56.2 Ake10 10 2.802 (± 0.561) 109–169 60.9 Ake11 7 1.659 (± 0.309) 162–210 56.2 Southeastern Naturalist P.A. Smallwood, M.D. Caspary, and J.E. Russell 2018 Vol. 17, No. 3 460 and 5 μl of template DNA produced from the DNA extraction procedure. We brought the reactions to a final volume of 25 μl through the addition of moleculargrade ddH2O. PCR reactions were performed using the thermocycle protocol of Topinka et al. (2004); however, we changed the annealing temperature to optimize each individual microsatellite reaction (Table 2). Following amplification, we electrophoresed all products on a 2% agarose gel at 90 V for 1.5 h to verify the presence of the desired product. Tagging of PCR products with fluorophores We used the microsatellite amplicons as template DNA for a second PCR procedure, hereafter referred to as the PCR-tagging reaction (PCRtag), in order to generate fluorescently tagged DNA fragments. We diluted the PCR products from the first reaction to a 1:100 concentration using molecular grade ddH2O and used these diluted PCR products as the template DNA in the PCRtag reaction. We carried out PCRtag reactions using the same concentration of reagents as in the original PCR, with the exception of the forward primer. The PCRtag required 2 forward primers: (1) the original forward primer with a tagging sequence attached to the 5' end (5'-CAGGACCAGGCTACCGTG-original primer sequence for target loci-3'; Blacket et al. 2012), and (2) the tagging sequence with 1 of 4 fluorophores—either FAM, VIC, NED, or PET—attached at the 5' end. PCRtag cycle 1 created the attached tagging strand (forward primer 1) that was replicated in cycle 2. The replicated strand was subsequently used as the template strand for the incorporation of the fluorophore (forward “tagging” primer 2) in PCRtag cycle 3. The final concentration of the forward primer was 0.06 μM, whereas the tagging primer was 0.08 μM, allowing a 0.4 tagging primer: 0.3 forward primer: 1 reverse primer-volume ratio to be maintained, as suggested by Blacket et al. (2012). We verified successful amplification through use of a 2% agarose gel using electrophoresis as described above. All fluorescently tagged oligonucleotides were synthesized by Eurofins Genomics (Huntsville, AL). The 2-step PCR reaction methodology employed here was developed by others to lower costs associated with the development process of primers (Culley et al. 2013). The 2-step PCR methodology used with Louisiana Bluestar, however, incorporated the use of the normal, nonfluorophore-containing primers to initially detect non-specific binding of primers to other regions of the genome. We employed this modification to the 2-step PCR amplification used by Culley et al. (2013) due to the observation that when all 3 primers involved in the tagging process were used together, multi-banding resulted from non-specific primer annealing. The use of a normal forward primer, lacking the tagging sequence, allowed the amplification of only the locus of interest. Once we verified the success of the first PCR, the observed non-specific binding could be overcome by the massive number of copies. DNA fragment analysis of amplicons After verification of a successful PCR reaction, both Ake7 and Ake8 amplifications from a respective individual were multiplexed in a 1:1 ratio. The fluorescently tagged Ake2, Ake3, and Ake4 were also multiplexed together; the Ake5, Ake6, Southeastern Naturalist 461 P.A. Smallwood, M.D. Caspary, and J.E. Russell 2018 Vol. 17, No. 3 Ake10, and Ake11 amplicons were multiplexed into a separate sample. We diluted the multiplexed samples by a 1:10 volume ratio using molecular-grade ddH2O. After sample preparation, we added to the Ake 7:Ake 8 multiplex sample a GGF 500 ROX labeled size standard produced by Georgia Genomics Facility (Athens, GA). The remaining multiplex samples had a Gene Scan LIZ 600 standard added, followed by analysis through use of an ABI 3730XL Genetic Analyzer (Thermo- Fisher Scientific, Waltham, MA). To determine the genotype of each individual, we analyzed the produced .FSA files using the open-source program STRand (Toonen and Hughes 2001). Analysis of differentiation and gene flow Once we had determined individual genotypes, we calculated allelic richness (Ar), observed and expected heterozygosity (Ho and He, respectively), inbreeding coefficient (Fis), and genetic differentiation (Gst) in the R package diveRsity (Keenan et al. 2013). We used 9999 bootstraps to determine the 95% confidence interval for Fis and Gst and calculated the effective number of alleles (Ae) for each population, along with standard errors in GenAlEx 6.5 (Peakall and Smouse 2006, 2012). We also employed this program to calculate genetic distances between all pairwise individuals, which we subsequently used in a principal coordinates analysis (PCoA). We conducted a mantel test (GenAlEx) using 9999 permutations to analyze isolation by distance. We examined how populations were clustered using a 100,000-step Markov Chain Monte-Carlo after a burn-in period of 50,000 steps in the R package Geneland (Guillot et al 2012). The following relationship was used to determine the number of individuals migrating between each pairwise set of population (Nm): Gst = 1 / (4Nm + 1) We determined the 95% confidence intervals for Nm estimates by using the above relationship on the upper and lower bounds of the 95% confidence interval for the Gst for the respective pair of populations. We determined the population genetic structure of A. ludoviciana in the program STRUCTURE version 2.3.4 using a burn in of 10,000 steps and 100,000 MCMC steps (Pritchard et al. 2000). We found the optimal number of clusters to be K = 4 through the use of STRUCTURE HARVESTER version 0.6.94 (Earl et al. 2012), which follows the methods of Evanno et al. (2005). The ancestry model used in the analysis was the admixed model with the assumption of independent allele frequency. We employed STRUCTURE PLOT version 2.0 to visualize the resulting Q matrices (Ramasamy et al. 2014). Results Microsatellite diversity The average number of alleles (Na) was 8.36 and the average number of effective alleles (Nae) was 2.139 (± 0.120) across the 8 loci used (Table 2). Expected Southeastern Naturalist P.A. Smallwood, M.D. Caspary, and J.E. Russell 2018 Vol. 17, No. 3 462 heterozygosity (He) varied from 0.355 to 0.544, with population 5 producing the lowest He and population 4 the highest (Table 1). Observed values of Ar fell between 2.66 (95% CI = 2.00–3.13) and 4.11 (95% CI = 3.25–4.75). Fis values varied from -0.095 (95% CI = -0.247 to -0.014) to 0.129 (95% CI = -0.056–0.234), with POP6 the only 95% confidence interval to not include 0. The number of private alleles at each site was 3 for POP1, 4 for POP2, 0 for POP3, 9 for POP4, 6 for POP5, and 8 for POP6. Ae values varied from 1.657 (SE = ±0.151) to 2.453 (SE = ±0.296). All population-level values of Ae (including the range from the standard error) existed within the 95% confidence interval for Ar. Population differentiation and gene flow Values of Gst varied from 0.0450 (95% CI = 0.0212–0.0806) and 0.1817 (95% CI = 0.1262–0.2448) with the global value 0.1918 (95% CI = 0.1615–0.2262), and Nm varied from 1.13 (95% CI = 0.77–1.73) to 5.31 (95% CI = 2.85–11.54) (Table 3). The mantel test revealed a pattern of isolation by distance ( P < 0.0001). We observed no strong clustering in the PCoA (Fig. 2). However, some patterns among populations emerged within the plot. Individuals from POP2, POP3, and POP5 were freely interspersed together in the negative region of principal coordinate analysis (PCA) 1. Individuals from POP1 were in this area as well, though some individuals were also placed on the positive side of PCA 1 intermingled with POP4 and POP6. Points from POP4 were heavily skewed towards the positive side of PCA2. POP4 appears to be the most isolated population, with few individuals from other populations interspersed within the area which POP4 occupies. Points Table 3. Calculated values of Gst (below the diagonal) and the calculated values for Nm, based upon the values of the pairwise Gst (above the diagonal). Numbers within the parentheses are the lower and upper bounds of the 95% confidence interval, respectively. POP1 POP2 POP3 POP4 POP5 POP6 POP1 3.09 5.31 2.78 2.02 4.39 (1.85, 5.72) (2.85, 11.54) (1.63, 5.46) (1.27, 3.38) (2.37, 9.22) POP2 0.0748 3.18 1.69 4.42 1.87 (0.0419, (1.92, 6.19) (1.19, 2.48) (2.29, 11.65) (1.24, 2.83) 0.1193) POP3 0.045 0.0729 2.27 1.56 1.96 (0.0212, (0.0388, (1.48, 3.70) (0.98, 2.59) (1.24, 3.25) 0.0806) 0.1152) POP4 0.0826 0.1289 0.0994 1.13 1.87 (0.0438, (0.0917, (0.0633, (0.77, 1.73) (1.25, 2.93) 0.1330) 0.1734) 0.1449) POP5 0.1102 0.0535 0.1379 0.1817 1.49 (0.0688, (0.0210, (0.0879, (0.1262, (1.00, 2.27) 0.1640) 0.0986) 0.2034) 0.2448) POP6 0.0539 0.1179 0.1132 0.1180 0.1436 (0.0264, (0.0813, (0.0715, (0.0787, (0.0994, 0.0954) 0.1675) 0.1679) 0.1666) 0.1995) Southeastern Naturalist 463 P.A. Smallwood, M.D. Caspary, and J.E. Russell 2018 Vol. 17, No. 3 from POP6 are intermixed with others from POP1, POP3, and a solitary point near the POP4 region. Individuals within POP3 and POP4 were predominantly assigned to the first cluster in the STRUCTURE analysis, with the exception of 2 individuals within POP4 (Fig. 3). These 2 individuals were the only ones that were assigned to the 4th cluster within the data set. STRUCTURE also showed that POP2 and POP5 mostly Figure 3. The plotted results from the STRUCTURE analysis. The bars along the bottom of the figure display which vertical bars are representative of individuals from a given sampling site. The shading of the bars represents which genetic cluster to which STRUCTURE assigned an individual. Individuals can be assigned to multiple genetic clusters, with part of the genotype suggesting one cluster and another part of the genotype suggesting the individual belongs to a different cluster. In total there were 4 clusters which were inferred from the Evanno method. However, 1 cluster only had 2 individuals. Figure 2. Principal coordinate analysis based upon the pairwise genetic distances between all individuals. Principle coordinate 1 described 17.32% of the variation in the data, while principle coordinate 2 described 12.89%. Southeastern Naturalist P.A. Smallwood, M.D. Caspary, and J.E. Russell 2018 Vol. 17, No. 3 464 contained individuals from the same cluster as well. The 2 remaining populations (POP1 and POP6) were not dominated by a single cluster and instead showed a mixed cluster composition. The analysis using Geneland produced 3 separate spatial clusters (Fig. 4). The 3 western populations (POP3, POP6, and POP1) were placed into a single cluster. The eastern POP4 was placed into a singular cluster while the 2 central populations (POP2 and POP5) were clustered together. Figure 4. The generated map of cluster membership using the R package Geneland. Three clusters of populations were inferred form this analysis. The map in panel A shows the inferred boarders of the 3 clusters. Within this map, cluster 1 is the darkest grey, cluster 2 is the medium grey, and cluster 3 is the lightest grey. The remaining maps show the probability that an individual found in a given location is a member of the respective cluster, with panel B representing cluster 1, panel C representing cluster 2, and panel D representing cluster 3. The lighter the color the higher the probability of belonging to that cluster. Lines within these maps show the probability topology. Axis units have been omitted to protect the locations of the sampled populations. Southeastern Naturalist 465 P.A. Smallwood, M.D. Caspary, and J.E. Russell 2018 Vol. 17, No. 3 Discussion When compared to other species that are endemic to the granite outcrops in the Southeastern US, the Louisiana Bluestar displays higher levels of gene flow between populations and higher genetic diversity. Most outcrop species capable of self-pollination have high levels of differentiation, with Fst and Gst values varying from 0.217 to 0.693 and He values of 0.020–0.069 (Koelling et al. 2011, Wyatt 1997, Wyatt et al. 1992). The data presented here for the Louisiana Bluestar, however, seem more in line with the self-incompatible species that occur on outcrop sites, which have Fst and Gst values from 0.077 to 0.235 and He values between 0.142 and 0.681 (Gevaert et al. 2013, Godt and Hamrick 1993). While the mating system for the Louisiana Bluestar is not currently known, this is evidence that there may be a self-incompatibility system present. Louisiana Bluestar produces a follicle, which dehisces to release seed. Seeds dispersed through this method would typically not move far from the maternal plant, thus having little impact on the movement of genes between populations. Populations of the Louisiana Bluestar appear to be much less differentiated when compared to other perennial plants that utilize a gravity seed-dispersal strategy. Such plants have been reported to have an average Gst of 0.248, which is higher than the upper bound of the 95% confidence interval for the Louisiana Bluestar (0.2262; Hamrick and Godt 1996), and which may mean that pollinators are supporting genetic connectivity between populations. Louisiana Bluestar also appears to be more genetically diverse with an estimated He of 0.464, compared to an average estimated He of 0.174 among ecological cohorts (Hamrick and Godt 1996). This finding, along with the high Nm values, is suggestive of notable levels of gene flow between populations. The majority of the loci investigated in this study showed much lower values of Nae in comparison to the number of observed alleles (Na), indicating that there is a high number of rare alleles. This result could be suggestive of a higher level of diversity throughout the species’ range in Georgia because we sampled only 6 exterior populations for this study (Slatkin 1985). Further investigations into the range of alleles at these loci within the populations of Louisiana Bluestar throughout Georgia may be a worthy avenue of future research. There is also the possibility of intragenic hybridization. We observed no co-occurring Amsonia species at these sites, suggesting that this hybridization could only occur through long-distance pollen flow. Eight of the Ake microsatellite loci described by Topinka et al. (2004) appear to be polymorphic within Louisiana Bluestar. Topinka’s 4 remaining Ake loci (Ake1, Ake4, Ake9, and Ake12) either failed to demonstrate polymorphism or failed to amplify. To our knowledge, this is the first study to identify polymorphic microsatellite loci for the Louisiana Bluestar. Interestingly, the Louisiana Bluestar showed a higher number of alleles at the microsatellite loci used within this study than any of the other 6 Amsonia species in which these microsatellites have been examined (Topinka et al. 2004). For these other species, most loci contained 2 to 4 alleles, in comparison to the 6 to 10 alleles observed for Louisiana Bluestar. Southeastern Naturalist P.A. Smallwood, M.D. Caspary, and J.E. Russell 2018 Vol. 17, No. 3 466 While the Evanno method inferred that there were 4 clusters, Geneland only detected 3. It should be noted however that only 2 individuals were actually assigned to the fourth cluster in the STRUCTURE analysis. These individuals could be migrants from one of the populations not surveyed within this study. When these 2 samples are ignored, then both methods are in agreement with only 3 genetic clusters existing. It also appears as though these 3 clusters are split geographically, with a western, eastern, and central cluster. The STRUCTURE plot also suggested that the migration events are coming from the eastern and central clusters to the western ones, and not in the opposite direction. POP1 appears to be the population most involved in interpopulation gene movement, with the highest mean value of Nm. Further, the PCoA analysis also shows that individuals from POP1 are intermingled with individuals from the other populations in ordination space, suggesting there are individuals within this population which are genetically similar to other individuals across all populations. This situation could be the result of migratory pollen coming into this population, which would also explain why it is composed of such a mixture of STRUCTURE-based genetic clusters. Another population that appears to function as a genetic migratory destination is POP6; this may just be a result of the high level of genetic intermixing between geographically proximate POP1 and POP6 (Gst = 0.0539, Nm = 4.39). Our analyses seem to be in disagreement over the interaction between POP3 and POP4. Although the STRUCTURE plot shows that all of the individuals in these 2 populations should be from the same genetic cluster, these are 2 of the more geographically distant populations. Further, both the PCoA and Geneland analyses seem to agree that these 2 populations are genetically dissimilar. The PCoA seems to be in agreement with Geneland and suggests that individuals in POP3 are more similar to POP5. It is also strange that POP3 is more similar to POP4 in the STRUCTURE analysis than it is to the geographically adjacent POP5. Why would such a strong barrier exist between these 2 populations, but allow the more distant populations to be similar to each other? While interspersed populations of Louisiana Bluestar may act as genetic stepping stones between POP3 and POP4, why would they not also allow genetic homogenization between all 3 populations? Overall, there appears to be a high level of diversity and connectivity within the Georgian populations of Louisiana Bluestar. However, throughout the species’ range, there is a marked increase in urbanization and land-use change (Caspary 2011). The question of how the genetic makeup of the Louisiana Bluestar will change with time remains. One of the populations resides a short distance from a neighborhood that was under development and undergoing expansion during the time of collection. Amsonia species are perennial plants; thus, the data presented here are likely more representative of the historic trends than a snapshot of current events. The relative contributions of seed and pollen movement cannot be determined from the data presented here. Chloroplast-DNA data would be useful for future study to determine the relative extent of seed dispersal (Ennos 1994). Southeastern Naturalist 467 P.A. Smallwood, M.D. Caspary, and J.E. Russell 2018 Vol. 17, No. 3 Acknowledgments The authors acknowledge Dr. Fengjie Sun for his expertise and assistance and the Georgia Gwinnett College Seed Fund for supporting this research. We also acknowledge Chris Doffitt, with the Louisiana Heritage Program, and Theo Witsell, with the Arkansas Natural Heritage Commission, for generously sharing plant material for this research. The Georgia Department of Natural Resources, Non-game Conservation Section also provided collection permits and location information for the work to continue. Literature Cited Allison, J.R. 2013. Status of rare plant species on outcrops of Lithonia Gneiss and on granite outcrops in Heard County. Georgia Department of Natural Resources. Social Circle, GA Blacket, M.J., C. Robin, R.T. Good, S.F. Lee, A.D. Millers. 2012. Universal primers for fluorescent labeling of PCR fragments: An efficient and cost-effective approach to genotyping by fluorescence. Molecular Ecology Resources 12:456–463 . Caspary, M. 2011. Effects of prescribed fire, invasive species, and geographic distribution patterns on granite rock outcrop plant communities in the Georgia Piedmont. Ph.D. Dissertation. University of Georgia, Athens, GA. Culley, T.M., T.I. Stamper, R.L. Stokes, J.R. Brzyski, N.A. Hardiman, M.R. Klooster, and B.J. Merritt. 2013. An efficient technique for primer development and application that integrates fluorescent labeling and multiplex PCR. Application in Plant Sciences 1(10): 1300027. DOI:10.3732/apps.1300027. Doffitt, C., C. Allen, P. Lewis, and D. Lewis. 2014. Amsonia ludoviciana (Apocynaceae) new to the flora of Texas (USA.). Journal of the Botanical Research of Texas 8(2):663–664. Dury, W.H. 1974. Rare species. Biological Conservation 6:162–169. Earl, D.A., and B.M. vonHoldt. 2012. STRUCTURE HARVESTER: A website and program for visualizing STRUCTURE output and implementing the Evanno method. Conservation Genetics Resources 4(2):359–361. Edwards, L., K. Kirkman, and J. Ambrose. 2013. The Natural Communities of Georgia. The University of Georgia Press, Athens, GA. 800 pp. Endress, M.E, S. Liede-Schumann, and U. Meve. 2007. Advances in Apocynaceae: The enlightenment, an introduction. Annals of the Missouri Botanical Garden 94(2):259–67. Ennos, R.A. 1994. Estimating the relative rates of pollen and seed-flow migration among plant populations. Heredity 72:250–259. Estil, J.C., and M.B. Cruzan. 1999. Phytogeography of rare plant species endemic to the southeastern United States. Castanea 66:3–23. Evanno, G., S. Regnaut, and J. Goudet. 2005. Detecting the number of clusters of individuals using the software STRUCTURE: A simulation study. Molecular Ecology 14(7):2611–2620. Georgia Department of Natural Resources (GADNR). 2013. Nongame conservation section biotics database. Wildlife Resources Division, Social Circle. Available online at http:// www.georgiawildlife.com. Accessed: 6 April 2016. Gevaert, S.D., J.R. Mandel, J.M. Burke, and L.A. Donovan. 2013. High genetic diversity and low population structure in Porter’s Sunflower (Helianthus porteri). Journal of Heredity 104(3):407–415. Godt, M.J.W., and J.L. Hamrick. 1993. Genetic diversity and population structure in Tradescantia hirsuticaulis (Commelinaceae). American Journal of Botany 80(8):959–966. Southeastern Naturalist P.A. Smallwood, M.D. Caspary, and J.E. Russell 2018 Vol. 17, No. 3 468 Guillot, G., S. Renaud, R. Ledevin, J. Michaux, and J. Claude. 2012. A unifying model for the analysis of phenotypic, genetic, and geographic data. Systematic Biology 61(6):897–911. Hamrick, J.L., and M.J.W. Godt. 1996. Effects of life-history traits on genetic diversity in plant species. Philosophical Transactions: Biological Sciences 351(1345):1291–1298. Keenan, K., P. McGinnity, T.F. Cross, W.W. Crozier, and P.A. Prodöhl. 2013. DiveRsity: An R package for the estimation and exploration of population-genetics parameters and their associated errors. Methods in Ecology and Evolution 4(8):782–788. Koelling, V.A., J.L. Hamrick, and R. Mauricio. 2011. Genetic diversity and structure in two species of Leavenworthia with self-incompatible and self-compatible populations. Heredity 160:310–318. Lemke, D.E. 1987. Recent collections and a redescription of Amsonia ludovicinana Vail (Apocyanaceae). Sida 12:343–346. MacArthur, R.H., and E.O. Wilson. 1967. The Theory of Island Geography. Princeton University Press, Princeton, NJ. 224 pp. Peakall, R., and P.E. Smouse. 2006. GENEALEX 6: Genetic analysis in Excel. Population genetic software for teaching and research. Molecular Ecology Notes 6:288–295. Peakall, R. and P.E. Smouse. 2012. GENEALEX 6.5: Genetic analysis in Excel. Population genetic software for teaching and research: An update. Bioinformatics 28:2537–2539. Pritchard, J.K., M. Stephens, and P. Donnelly. 2000. Inference of population structure using multilocus genotype data. Genetics 155:945–959. Quarterman, E., M.P. Burbanck, and D.J. Shure. 1993. Rock outcrop communities: Limestone, sandstone, and granite. Pp. 35–86, In W.H. Martin, S.G. Boyce, and A.C. Echternacht (Eds.). Biodiversity of the Southeastern United States. Wiley, New York, NY. 528 pp. Ramasamy, R.K., S. Ramasamy, B.B. Bindroo, and V.G. Naik. 2014. STRUCTURE PLOT: A program for drawing elegant STRUCTURE bar plots in user-friendly interface. SpringerPlus 3(431). DOI:10.1186/2193-1801-3-431. Shure, D.S. 1999. Granite outcrops of the southeastern United States. Pp. 99–118, In R.C. Anderson, J.S. Fralish, and J.M. Baskin (Eds.). Savannas, Barrens, and Rock Outcrop Plant Communities of North America. Cambridge University Press, Cambridge, UK. 484 pp. Slatkin, M. 1985. Rare alleles as indicators of gene flow . Evolution 39(1):53–65. Toonen, R.J., and S. Hughes. 2001. Increased throughput for fragment analysis on ABI Prism 377 Automated Sequencer using a membrane comb and STRandsSoftware. Biotechniques 31:1320–1324. Topinka, J.R., A.J. Donovan, and B. May. 2004. Characterization of microsatellite loci in Kearney’s Bluestar (Amsonia kearneyana) and cross-amplification in other Amsonia species. Molecular Ecology Notes 4:710–712. Weakley, A. 2012. Flora of the Southern and Mid-Atlantic States. University of North Carolina Herbarium, Chapel Hill, NC. 1225 pp. Woodson, R.E. 1928. Studies in the Apocynaceae. III. A monograph of the genus Amsonia. Annals of the Missouri Botanical Garden 15(4):379–434. Wyatt, R. 1997. Reproductive ecology of granite outcrop plants from the southeastern United States. Journal of the Royal Society of Western Australia 80:123–129. Wyatt, R., and J.R. Allison. 2000. Flora and vegetation of granite outcrops in the southeastern United States. Ecological Studies 146:409–433. Wyatt, R., and N. Fowler. 1977. The vascular flora and vegetation of the North Carolina granite outcrops. Bulletin of the Torrey Botanical Club 104:245–253. Southeastern Naturalist 469 P.A. Smallwood, M.D. Caspary, and J.E. Russell 2018 Vol. 17, No. 3 Wyatt, R., E.A. Evans, and J.D. Sorenson. 1992. The evolution of self-pollination in granite outcrop species of Arenaria (Caryophyllaceae) VI. Electrophoretically detectable genetic variation. Systematic Botany 17:201–209. Wyatt, R., A.L. Edwards, S.R. Lipow, and C.T. Ivey 1998. The rare Asclepias texana and its widespread sister species, A. perennis, are self-incompatible and interfertile. Systematic Botany 23:151–156.