ISSR and AFLP analysis of the temporal and spatial population structure of the post-fire annual, Nicotiana attenuata, in SW Utah

Background The native annual tobacco, Nicotiana attenuata, is found primarily in large ephemeral populations (typically for less than 3 growing seasons) after fires in sagebrush and pinyon-juniper ecosystems and in small persistent populations (for many growing seasons) in isolated washes typically along roadsides throughout the Great Basin Desert of the SW USA. This distribution pattern is due to its unusual germination behavior. Ephemeral populations are produced by the germination of dormant seeds from long-lived seed banks which are stimulated to germinate by a combination of unidentified positive cues found in wood smoke and the removal of inhibitors leached from the unburned litter of the dominant vegetation. Persistent populations may result where these inhibitors do not exist, as in washes or along disturbed roadsides. To determine if this germination behavior has influenced population structure, we conducted an AFLP (244 individuals), ISSR (175 individuals) and ISSR+ AFLP (175 individuals) analysis on plants originating from seed collected from populations growing in 11 wash and burns over 11 years from the SW USA. Results Genetic variance as measured by both ISSR and AFLP markers was low among sites and comparatively higher within populations. Cluster analysis of the Utah samples with samples collected from Arizona, California, and Oregon as out-groups also did not reveal patterns. AMOVA analysis of the combined AFLP and ISSR data sets yielded significantly low genetic differentiation among sites (Φct), moderate among populations within sites (Φsc) and higher genetic differentiation within populations (Φst). Conclusions We conclude that the seed dormancy of this post-fire annual and its resulting age structure in conjunction with natural selection processes are responsible for significantly low among sites and comparatively high within-population genetic variation observed in this species.

seasons) after fire in sagebrush and pinyon-juniper ecosystems, in small persistent (for >3 growing seasons) populations in isolated washes, and as a roadside weed after new construction in a previously undisturbed area [2,[4][5][6][7][8][9]. Positive and negative control by environmental signals over germination from long-lived seed banks (estimated to be minimally 150 y [10] can account for its occurrence in these habitats. Specifically, dormant N. attenuata seeds are stimulated to germinate by unidentified factors in wood smoke [9] but are inhibited by factors, including ABA and 4 terpenes (bornane-2,5-dione, 1,8-cineole, βthujaplicin and camphor [11] which leach from the litter of the dominant vegetation. Genotypes of N. attenuata produce seeds that vary in their genetically-determined primary dormancy [9]. Regardless of their degree of primary dormancy, seeds that are shed in unburned habitats with significant accumulations of litter develop strong secondary dormancy in response to the negative germination cues. If the seeds are shed into habitats without significant litter accumulations (e.g. in washes or roadside habitats), seeds without dormancy germinate. When fires pyrolyze the litter layer, removing the germination inhibitors and saturating the soils with smoke-derived germination stimulants, the seed bank responds with a dramatic, synchronized germination response the following growing season during favorable moisture and thermal regimes.
This well-characterized germination behavior likely affects the genetic structure of this potential annual. Genetic structure of a population results from mutations, gene flow (as mediated by pollen and seed dispersal), drift, and selection, all acting in the context of an organism's life history traits [12]. Genetic differentiation may be more prevalent between primarily dormant and nondormant populations, namely between plants found ephemerally (in burns) and those occurring more persistently (in washes). Within the ephemeral populations, the number of plants in the population will vary in relation to the size of the burn and the distribution of the seed bank. Because pollinators must locate these ephemeral populations in a landscape that may be largely composed of other plant associations, out-crossing may not be prevalent. Flowers of N. attenuata are self-compatible and outcrossing does not significantly affect seed production, seed mass or viability [13] indicating that this species relies on selfing as its primary form of reproduction. Selfing may keep genetic variation low, especially within populations. Persistent populations are more likely to experience outcrossing, owing to their predictability. These considerations in combination with the annual life cycle of plants in washes in contrast to the 7 -150 year life cycle of plants growing in burns may increase the genetic differences among populations found in burns and washes.
Here we examine the genetic structure of N. attenuata plants from wash and burn populations in the SW Utah ( Fig. 1; Table 1) to determine if the particular germination behavior of this species has left signatures in the plant's population structure. We use an AFLP (amplified fragment length polymorphism) analysis, based on the selective polymerase chain reaction (PCR) amplification of restriction fragments from a total digest of genomic DNA [14] and an ISSR (inter-simple sequence repeats) analysis in which bands are generated by single primer PCR that amplifies products between two simple sequence repeats [15]. Both procedures produce reproducible markers useful for the quantification of genetic polymorphism within species [16].
Specifically, we compare plants growing from seeds collected from 11 large populations after fires, from small populations in 10 washes, from plants in transects across 5 large burns, and from plants growing in specific areas over 10 years during which a small wash population erupted into a large burn population as a result of a fire and returned to become a small wash population. By analyzing the genetic diversity across these N. attenuata populations, we aimed to answer the following questions: 1) Are plants growing in burn and wash populations genetically distinct? 2) Are plants growing in the same washes genetically similar through different years? 3) What is the genetic makeup of plants found growing across large burns and geographically adjacent populations? While genetic diversity among the various Nicotiana species has been studied with RAPD [17] and AFLP [18] markers, and with peroxidase isozymes [19], this is the first effort to study the spatial and temporal population structure of a native Nicotiana species.

SET-I
Set I (Table 1) consisting of 244 individuals, which was used only for AFLP analysis, produced a total of 207 loci (data not shown). This data was used for separate dendrogram and principle co-ordinate (PCO) analyses. The Jaccard similarity index [20] based on unweighted pair group method average (UPGMA) dendrogram revealed a lack of distinct spatial or temporal structure and had brush-or star-structures with nodal bootstrap values of less than 60% (data not shown). The samples collected from the greatest spatial distances, namely California, Oregon and Arizona, did not form separate clusters from any of the Utah populations. A cluster analysis of 10 wash and 5 burn populations all grown from seed collected in 1999 (Table 1) revealed no clustering based either on type of population (wash or burn) or geographic location (data not shown).. Some structure was identified when each time series at particular locations (Motoqua, DI Ranch, Shivwits Reservation) were analyzed separately ( Fig.  2A,2B,2C), but the nodes did not correspond to a particular growing season. No genetic differentiation was observed from a cluster analysis of plants collected from Motoqua before (1990) during (1995)(1996) and after (1999) a fire-associated population explosion (Fig. 2B). A similar lack of structure was found in the time series analysis from the DI Ranch ( Fig. 2A) and Shivwits roadside washes (Fig. 2C). Similarly, the PCO did not yield any apparent population structure or site-or population-specific grouping associations.

SET-II
Set II (Table 1) is a subset of Set I ( Table 1) and consists of 175 individuals analyzed by AFLP and ISSR (either combined or separate) and this dataset was used for dendrogram and PCO analyses. Combined (AFLP + ISSR) Location of Nicotiana attenuata populations from which seeds were collected between 1988-1999 in Southwestern Utah Figure 1 Location of Nicotiana attenuata populations from which seeds were collected between 1988-1999 in Southwestern Utah. See Table 1 for number of plants grown for DNA extraction from each location for the AFLP and ISSR analysis. Locations labeled with circles represent single-site or -time collections, while squares signify multiple-site or -time collections UPGMA dendrograms based on the Jaccard similarity index calculated from an AFLP analysis of N. attenuata populations col-lected at three different sites  Table 1. While substantial genetic variation was found, this variation was not organized in time.
analysis revealed a total of 286 loci of which 268 were polymorphic (93.70%). Here, the AFLP analysis showed higher percent polymorphic loci than did the ISSR analysis (96.1% and 87.5%, respectively; Table 2). Interestingly in the AFLP analysis, the primer and the restriction enzyme combinations that produced the lowest number of loci also delivered the highest rate of polymorphism ( Table 2). It produced an average of 68.7 loci per primer combination with a high percentage of unique bands (65 in the 0-10% frequency class; Fig. 3) and a high frequency of commonly shared loci (42 in the 91-100% frequency class; Fig. 3). The ISSR analysis, on the other hand, produced 16 loci per primer with a predominance of commonly shared loci (29 in the 91-100% frequency class as compared to 14 in the 0-10% frequency class; Fig. 3). Dendrograms and PCO produced from this data set had the same overall characteristics as those produced from showed same structure nature Set I.

Heterozygosity
A Bayesian approach [21] was used for heterozygosity calculations. The total heterozygosity as measured from the combined AFLP and ISSR data set of Utah collections (SET-II,  Table 3).

AMOVA
AMOVA analysis was performed separately for AFLP, ISSR and combined analysis of plants collected from Utah (168  individuals from SET-II, Tables 1, 4). The combined data set was also used to partition variation between wash and burn populations and to examine the effects of the collection year. In separate analyses, ISSR revealed higher variance than did the AFLP in the among-sites, amongpopulation, and within-site categories; whereas, variation in the within-population category from the AFLP analysis was higher than that from the ISSR analysis (Table 4). All values except the among-site category in the AFLP analysis (p < 0.05) revealed highly significant differences at p < 0.001. AFLP and ISSR data was combined for an AMOVA analysis of all analyzed Utah populations. From this analysis, all three Φ categories were highly significant (p <   Table 4) among sites (Φct), among populations within sites (Φsc) and within populations (Φst) values were 0.046, 0.116, and 0.156, respectively. Table 4 reveals low genetic differentiation among sites and a relatively high genetic differentiation within populations. Pair-wise genetic distances (pair-wise Φst) were calculated from the AMOVA. Of the 300 comparisons from the 25 populations, 220 showed highly significant differences and 29 were significant at the p = 0.05 level (Table 5) [see Additional file 1]. Very low among-site variation (0.18 %) was obtained when samples were compared as being derived from either burn or wash populations (Table 4). To determine the effect of collection year, all individuals were grouped according to their collection year; an AMOVA analysis revealed low (3.77 %) variance within years at p < 0.5 significance level ( Table 4).
The AMOVA analysis had sufficient statistical power to detect small differences among populations, which accounted for 0.23 to 16.48 % of the variation, but this was dwarfed by the much larger genetic variation within populations, which ranged from 81.20 to 99.77 % (Table  4). This dramatic high degree of within-population variance was found in all populations. Again, populations from Goldstrike Canyon had the lowest among-population variance (0.23%) and highest within population variance (99.77%; Table 4). The Goldstrike populations were located in a narrow canyon produced by a stream and in this region seeds are likely transported among the populations during spring floods.
Mantel test were conducted to analyze isolation by distance using pair-wise Φst values obtained by AMOVA (ver 1.55). The Φst values from the AFLP, ISSR and the combined data sets were separately correlated with geographical distance and all revealed non significant correlations (AFLP, r = 0.099, p = 0.81; ISSR, r = 0.122, p = 0.89 and AFLP + ISSR, r = 0.019, p = 0.63)

Discussion
The analysis revealed high levels of heterozygosity, with total heterozygosity from all populations (0.2771 ± 0.0018) being higher than that from comparable analyses  [24]. The ISSR analysis (0.2452 ± 0.0056) yielded estimates of heterozygosity that were comparable with the AFLP analysis (0.2432 ± 0.0024) ( Table 3) despite the basic difference in the logic of the two procedures. ISSRs are designed to span a repeat region of the genome whereas AFLP is designed to randomly sample the full genome [16] and most plant genomes are thought to evolve faster in the repeat regions [25]. However, despite the differences in absolute estimates of genetic variation, both procedures produced the same conclusion.
The principle conclusion of this study is that large amount of genetic variation measured by AMOVA, within populations at a particular area significantly dwarfed that observed among sites, among populations growing in burns or washes, or collected during subsequent years growing at a given site. The conclusion that the genetic variation between neighbors is greater than that found between temporally-or spatially-separated populations, is dramatically reflected in the plants sampled along transects through the Pahcoon Springs burn. Only a small fraction (12.6%) of the total genetic variance is found among the 8 sub-samples from the extreme corners of this burn that covered more than 5000 hectares [26], while the majority (87.94 %) was found among plants growing within 10 m 2 of each other. Pahcoon, was the site from which highest number of populations were analyzed. (9 populations), whereas, from other sites smaller numbers of populations were analyzed.

NS = Not significant
A number of factors, including N. attenuata's unusual seed germination mechanisms and the irregular nature of fires, natural selection, gene flow mediated by pollination or the relocation of seeds via mammal-vectored transport could account for the lack of population structure and each deserve further discussion.
Dormancy is a major adaptive response of native plants that allows them to cope with environmental variation and provide a means of habitat selection [27][28][29]. Moreover, dormancy is likely to influence the genetic structure of populations [30,31]. Seed banks serve as repositories of genetic diversity for most species. Many seeds use cues as general as temperature, photoperiod, moisture, or their own age to trigger germination and initiate vegetative growth [32,33]. To cope with the lack of reliability of these proximate signals and the unpredictability of the post-germination environment, some species may have evolved "bet-hedging" strategies, whereby only a certain fraction of the dormant seed bank germinates under favorable conditions. This has been experimentally shown by various researchers. In Plantago lanceolata [30], Calluna vulgaris [34], Clarkia springvillensis [35] and Lesquerella fendleri [36] it has been demonstrated that the seed banks have less genetic differentiation than do the adults of a given population. This strategy provides a statistical solution to the problem of cueing germination with unreliable signals [37,38]. Other species, however, use specific signals to time their germination with particular niches. Those species that specialize in the immediate post-fire habitat are a particular case in point [39]. Studies on another fire-dependant plant, Grevillea macleayana [40], which also has a long-lived seed bank, showed Fst (0.218) that were comparable to those measured in this study for N. attenuata, but had variable heterozygosity (H obs = 0.248 -0.523). Another major difference from the current study was that G. macleayana showed significant isolation by distance.
Seed dormancy increases the effective generation time of this annual plant and by doing so, prevents genetic decay and inhibits the formation of spatial structure between geographically distinct populations [12]. Additionally, a long-lived seed bank results in the overlap of generations [41], which has similar effects and additionally reduces the ability of genetic drift to drive unique alleles to fixation. Operating under the assumption that the synchronized germination response observed after fires represents a synchronized germination of cohorts from the seed bank, we examined populations that occurred over a [6][7][8][9][10][11] year interval at the same location ( Fig. 2A,2B,2C) to determine if temporally-defined genetic structure occurred in the populations, but none were found. This suggests that seed banks have a more complicated germination response whereby only fractions of a cohort may germi-nate at any particular interval and hence may represent a combination of "bet-hedging" [33] and the chemicallycued germination of the seed bank.
N. attenuata has all the characteristics of species pollinated by moths at night (white fragrant flowers scenting and becoming receptive at night) but day-active humming birds (Selaphorus sp.) and bumblebees (Bombus sp.) are also known to visit the flowers [42]. Despite these traits that are thought to facilitate out-crossing, 16 years of field work with the Utah populations have revealed that the vast majority of seeds produced are the result of self-pollination. No evidence exists for inbreeding depression in plants self-pollinated for more than 20 generations (I. T. Baldwin, unpublished results). However, the plant species likely enjoys sporadic bursts of cross pollination during the rare outbreaks of hawk moths (Hyles lineata and Manduca species(observed once in 16 years of observation at the study sites [43]. The amount and distance of gene flow that occurs during these rare events is not known. In the wind pollinated species such as Zea mays maximum distance of pollen dispersal was observed to be 18 m achieving outcrossing rate of not more than 1%; insect pollination does not substantially increase this rate [44]. Hence in comparison to seed dispersal these events are likely to have a minor effect on the homogeneity of populations [12]. Seeds of N. attenuata are small (160 µg) and could be dispersed by wind, water transport and animals, but none of these mechanisms are well documented. The seeds are eaten by various ground squirrels [45] but are not known to survive a transit through the digestive track. The greater heterogeneity within populations and low genetic differentiation among populations found along the stream in the Goldstrike canyon (Table 5) [see Additional file 1] suggests water transport may not be important. While seeds tend to be dispersed from the plant upon maturation of the seed capsules, the N. attenuata calyx is sticky and glandular and could be dispersed by adhering to animals. However, the plants ability to produce the defense secondary metabolite nicotine in substantial quantities in its calyx [9] may be a much more important determinant of its long distance transport. Native Americans are known to have smoked leaves and seed capsules for recreational and medicinal purposes, buried their dead with leather pouches containing N. attenuata seeds, burned the sagebrush to promote its growth and are likely to have transported seeds throughout its range in North America [46]. Hence, movement of N. attenuata genotypes across the landscape by humans who were smoking this plant may have contributed to the lack of correlation between geographical and pair wise Φst values, as reflected by Mantel test for isolation by distance.
In summary, we conclude that the unusual nature of the N. attenuata populations from Utah revealed by AFLP and ISSR analysis is a likely result of combination of random dispersal by humans and its seed dormancy.

Conclusions
We conclude that the genetic structure of N. attenuata populations in Utah showed: 1) high similarity across collection sites; 2) small difference between populations growing in burns or washes; 3) small differences between growing seasons; and 4) large difference between individuals growing within populations.

Seed sources
Seeds were from individual-and multiple-plant samples collected from 1988-1999 from the southwestern USA (Table 1; Fig. 1). A majority of the seed collections (244) originated from a 1500 km 2 region of the SW corner of Utah (T38S R10W-T43S R19WUSA). Collections from Arizona (Flagstaff), Oregon (Eugene) and California (Sequoia Natl. Park) served as out-groups. In Utah, seeds were collected from plants growing at 6 locations for a number of years and were used for a time series analysis ( Table 1). One of these areas (Motoqua), the region surrounding a small wash population that had been sampled in 1990, was struck by lightening at the end of the growing season in 1994 (August) and 1163 hectares were burned. During the 1995-6 growing seasons, large populations of more than 100,000 plants were found, but by 1999 only a small population remained in the original wash. At this site, seeds were collected during the population explosion as well during the contraction of the population at this site (Table 1). A fire during the 1998 growing season at Pahcoon Springs created a large population (covering more than 5,000 hectares) in the 1999 growing season which was sampled in 8 locations: seeds from 10 individual plants growing along each of 4 line-transects with an inter-plant distance of 10 m and 10 plants growing within a 10 m 2 area at 4 locations were sampled to provide a small-scale spatial analysis of genetic variation for this population.

Plant material
Seeds (10 seeds per plant collected) were exposed for 1 h to 100 µL liquid smoke (House of Herbs INC., Passaic, NJ, USA): water (1:300, v/v) in 1-mL shell vials and 5 seedlings were planted in soil and grown to the rosette-stage in a glasshouse. Leaves from one plant randomly selected plant from each collection were harvested for DNA extraction.

DNA extraction
Leaves were flash-frozen in liquid nitrogen, ground to powder and suspended in 750 µL of 100 mM Tris/50 mM EDTA (pH 8.0), containing 250 µg/mL RNase A. Eight µL liquid laundry detergent (Ariel, Procter & Gamble, Schwalbach, Germany) were added. After 60 min incubation at 60°C and subsequent addition of 80 µL of 5 M NaCl, the suspension was centrifuged for 5 min at 16,000 × g. The supernatant was removed and extracted with phenol/chloroform. The DNA was precipitated with 600 µL isopropanol, pelleted by centrifugation at 16,000 × g for 5 min, washed with 200 µL 70% ethanol and dissolved in 50 µL of water. The purity and concentration of the extracted DNA were assessed by electrophoresis on a 1% agarose gel and optical density spectrometric measurements. Both AFLP and ISSR procedures were performed on the same DNA samples.
The separate PCR amplification products generated by each of the three primer combinations were loaded together with a ROX 500 GeneScan size standard onto the ABI Prism 310 automated genetic analysis system as described in the manufacturer's instructions. The samples were run with the following GeneScan settings: "GS STR POP 4 (1 mL) D" module; 150000 V run Voltage; 5 second sample injection; 60°C gel temperature, and 9 m Watt's laser power. The distinct emission spectra of the three fluorescently labeled Eco RI-primer types made it possible to distinguish the DNA fragments resulting from each of the different primer combinations separately while the samples were being separated in the same electrophoresis capillary.
Collection of raw data and size alignment of the AFLP fragments was performed with ABI Prism GeneScan Analysis Software (Applied Biosystems) with the internal standard. Aligned data were subsequently imported into Genographer [49] for band calling. Each AFLP locus with an intensity ≥ 150 fluorescence units was scored with the 'thumbnail' option of genographer and converted into a 1/0 binary data matrix, which was used for further analysis.

ISSR procedure
The PCR reaction (25 µL) contained 20 ng genomic DNA, buffer (10 mM Tris-HCl pH 8.3, 50 mM KCl, 1.5 mM MgCl 2 ) 0.8 U Taq DNA polymerase (Eppendorf) 0.1 mM dNTPs, 0.3 µM primer. After 5 min initial denaturation at 94°C, 45 cycles of 1 min denaturation, 45 s annealing at 50°C, and 2 min for extension at 72°C, were followed by 5 min final extension in the PCR cycling program. A total of 55 primers were screened and 5 primers (Table 2) were selected because they reproducibly produced distinct banding patterns. The amplified products were separated on 2.0% agarose gel (28 samples plus 2-1 Kb ladder standards on each gel) in 0.5 × TAE buffer and bands were detected by ethidium bromide staining. The PCR reaction and separation of PCR products was carried out in duplicate for each DNA sample and only reproducible bands were scored manually as present (1) and absent (0).

Data analysis
Pair-wise genetic similarity was calculated with the Jaccard coefficient [20]. The resulting matrix was processed for dendrogram construction using the UPGMA (unweighted pair group method average) clustering method and PCO (principle co-ordinate analysis) options of software MVSP (Multi-Variate Statistical Package ver 3.13: [50]) program. The entire AFLP (244 individuals. SET-I) and ISSR+AFLP (175 individuals, SET II) data sets were analyzed individ-ually and the 175 individuals (Table 1) those were used in both procedures were combined for clustering analysis. Subsequently, the SET I data was analyzed for each time series separately (Fig. 2) Genetic diversity was estimated for SET II (without Oregon, Arizona and California individuals) (168 individuals) as heterozygosity using the Bayesian approach of Holsinger et. al [21]. For this analysis, the analysis program, Hickory (ver 1.0) [23], was used with the full model. Several runs were carried out with default sampling parameters (burn-in = 50,000, sample = 250,000 and thin 50) to ensure consistency of results. Since, dominant markers (AFLP and ISSR) are used in conjunction with a largely sefling species, we used an approach that does not assume Hardy Weinberg equilibrium (Holsinger [21]).
The SET II was used to calculate molecular variance from the combined and separate AFLP and ISSR data sets (168 individuals) as partitioned into individual and population components with an AMOVA (ver1.55: [51]). We also calculated variation between different locations, burns and washes, by collection year and population, separately. Φ values generated by the AMOVA program were used to estimate pair-wise genetic diversity, which is an analogue of the F-statistic. The Mantel permutation test was used to correlate pair-wise Φst values obtained by separate analyses of the AFLP, ISSR and combined data sets with geographic distance.

Authors' contributions
RB carried out the entire ISSR analysis, the analysis of the AFLP data and contributed to writing the manuscript. DS grew the plants, extracted the DNA and conducted the AFLP analysis. CAP collected seeds, grew the plants and extracted the DNA. ITB was responsible for coordinating the study, collecting seeds for the analysis, and wrote the manuscript.