Agrostis stolonifera is largely cultivated for cool-season turf that is used extensively on golf course greens. Improvement of lines has been pursued for many years through selective breeding and more recently with molecular breeding (Chakraborty et al., 2006; Engelke and Arnold, 1993; Engelke et al., 1995; Krans et al., 2002). Creeping bentgrass has also been genetically engineered for experimental testing of various traits including resistance to herbicides, pests and abiotic stressors (Chai et al., 2002, 2003; Fu et al., 2005, 2007; Hartman et al., 1994; Lee et al., 1996).
Numerous research projects have shown that A. stolonifera and certain other Agrostis and Polypogon species are potentially cross-compatible. Thus, to varying degrees they can interpollinate to produce viable, fertile progeny (Belanger et al., 2003a, 2003b; Hart et al., 2009; Wipff and Fricker, 2001; Zhao et al., 2007). Hybridization within the bentgrass species complex is important for several reasons. Cytological and molecular evidence suggests that ancestral hybridization and reticulate evolution have resulted in the allopolyploid Agrostis species present today (Jones, 1956a, 1956b, 1956c; Vergara and Bughrara, 2003; Wipff and Fricker, 2001). From a contemporary perspective, the capacity for extant Agrostis species to form hybrids may be commercially advantageous in terms of broadening the germplasm resources that can be utilized in crosses for bentgrass cultivar development (Belanger et al., 2003b, 2004; Rotter et al., 2009). However, flow of proprietary transgenes into non–genetically engineered bentgrass crop fields could result in regulatory and economic issues for developers and growers. Furthermore, ecological concerns regarding fertile transgenic Agrostis hybrids exist because transgene flow, establishment of genetically engineered plants outside of cultivation, and persistence of such plants in the environment have recently been documented for creeping bentgrass (Reichman et al., 2006; Watrud et al., 2004; Zapiola et al., 2008).
Overall, little is known about the genetic diversification among and within A. stolonifera populations. Even less is known about the gene flow rates between creeping bentgrass populations and those of other species in the Agrostis hybridizing complex. Molecular markers are useful tools for studying these relationships. In particular, improved population-level molecular markers are needed for several Agrostis species for which phenotypic plasticity and subtle distinctions between morphological characters (many of which are small floral structures) can make hybrid identification difficult and uncertain, especially for intraspecific crosses (Vergara and Bughrara, 2004). Molecular markers developed specifically for A. stolonifera could be used to measure the genetic connectivity between crop and wild (unmanaged, native, and naturalized populations in nonagronomic habitats) creeping bentgrass populations. Polymorphic neutral markers are also necessary to utilize the well-developed techniques of population genetics that are designed to determine the selective effects on novel alleles within populations as well as track the history of allele movement (Schoen et al., 2008). Such markers could be used to clarify the phylogenetic origin of A. stolonifera, and they may help with product identification and breeding of bentgrass cultivars.
A fundamental challenge to DNA fragment–based population structure analyses of allopolyploid species is scoring individuals as homozygous or heterozygous for loci that are present in more than one genome of the species. This is particularly true for markers that often produce several bands per individual such as restriction fragment length polymorphisms (RFLPs), randomly amplified polymorphic DNA (RAPDs), and amplified fragment length polymorphisms (AFLPs) even in diploid organisms. Although several of these marker systems have previously been developed for other types of research on Agrostis species, each has limited suitability for the current needs (Caceres et al., 2000; Casler et al., 2003; Chakraborty et al., 2005; Reichman et al., 2006; Scheef et al., 2003; Vergara and Bughrara, 2003, 2004; Warnke et al., 1997, 1998; Zapiola et al., 2010). A summary of earlier marker types and their applications are available online in Supplement 1. Therefore, our primary objective in this research was to identify a set of genome-specific, codominant microsatellite (simple sequence repeat [SSR]) markers that are diallelic, producing one or two alleles per A. stolonifera individual per locus. While planning our development strategy, we considered two key points that have been inferred about the genomic organization and evolutionary history of A. stolonifera and its close relatives.
First, early cytological work on interspecific Agrostis hybrids identified the shared genomes among the crossed species as A. canina, A1A1; A. capillaris, A1A1A2A2; A. stolonifera, A2A2A3A3; and A. gigantea, A1A1A2A2A3A3 (Jones, 1956a, 1956b, 1956c). There was significant homology noted between the A1 and A2 genomes, while A1 was only slightly homologous with A3, which was thought to be derived from an unknown diploid ancestor.
Second, AFLP results have been used to describe the phylogenetic relationships for accessions of several Agrostis species from around the world (Vergara and Bughrara, 2003). Based on their observation that A. transcaspica Litv. was closely related to the Turkey–Eurasian lineage of A. gigantea, Vergara and Bughrara hypothesized that A. transcaspica was the diploid ancestral source of the A3 subgenome found in A. stolonifera and A. gigantea.
To test that hypothesis, we focused our efforts on identifying A3 genome-specific diallelic SSR loci from A. stolonifera and A. transcaspica SSR-enriched libraries. With such markers, fragment size data from A. stolonifera populations could then be analyzed as if this species was diploid instead of allotetraploid, and the markers would be expected to have the following cross-species reactivity:
Diallelic SSRs identified from the A3 subgenome of A. stolonifera will have orthologs in A. transcaspica and A. gigantea, but not in A. capillaris or A. canina.
Diallelic SSRs identified from the A3 subgenome of A. transcaspica will have orthologs in A. stolonifera and A. gigantea, but not in A. capillaris or A. canina.
Diallelic SSRs identified from the A2 subgenome of A. stolonifera will have orthologs in A. gigantea, A. capillaris, and possibly A. canina, but not in A. transcaspica.
MATERIALS AND METHODS
The cultivars and accessions of each Agrostis species utilized within this project are presented in Table 1 Unmanaged populations of A. stolonifera were identified in diverse nonagronomic habitats in Oregon based on morphological characteristics described in Hitchcock (1950) The Rainbow, Baskett Slough, and Ona Beach collection sites were selected to represent presumably isolated populations. The Rainbow riparian site (44°44′45″ N, 121°13′40″ W) was at an elevation of 418 m with Era sandy loam (coarse-loamy, mixed, superactive, mesic Vitritorrandic Haploxeroll) and fluvents near Madras, OR. The Baskett Slough wetland site (44°58′60″ N, 123°14′56″ W) was located 161 km west of Rainbow, over the Cascade Mountain Range in the Willamette Valley at an elevation of 70 m in Waldo silty clay loam (fine, smectitic, mesic Fluvoquentic Vertic Endoaquoll). The Ona Beach site (44°31′21″ N, 124°4′17″ W) was 83 km west of Baskett Slough, over the Coastal Mountain Range at 1.5 m above mean sea level where the soil was predominantly Waldport fine sand (mixed, isomesic Typic Udipsamment) (USDA-NRCS, 2010). The Ona Beach habitat is similar to that of the tidal flats near Coos Bay, OR, from where the phenotypically variable A. stolonifera cultivar Seaside originated (Duble, 1996; Hyslop, 1930; Madison, 1971; Warnke et al., 1997).
|Species||Accession or cultivar||Source|
|A. stolonifera||cv. Penn A-4||Turf Seed, Inc.|
|A. stolonifera||cv. Crenshaw||Kansas State University|
|A. stolonifera||CBG-1 mixed crop population||Seed Research of Oregon, Inc.|
|A. stolonifera||Ona Beach; wild-collected||Oregon Coast|
|A. stolonifera||Baskett Slough; wild-collected||Willamette Valley, Oregon|
|A. stolonifera||Rainbow; wild-collected||Deschutes River bank, central Oregon|
|A. gigantea||cv. Reton||University of Connecticut Farm|
|A. gigantea||Major||Major Michael Donnelly Walking Trail, South Windsor, CT|
|A. capillaris||cv. Exeter||University of Connecticut Farm|
|A. capillaris||cv. Game||Oregon State University Seed Lab|
|A. canina||cv. Vesper||University of Connecticut Farm|
|A. canina||PI 189141 cv. Novobent||USDA GRIN|
|A. transcaspica||PI 283174||USDA GRIN|
Growth of Seeds from Commercial, Academic, and Government Sources
Seeds from six Agrostis species were individually planted in water-moistened potting media (Seedling Mix #1, OBC Northwest, Canby, OR) in tagged trays and grown in greenhouses set to a 16-h light (388 μmol m−2 sec−1)/8-h dark photoperiod. Relative humidity in the greenhouse was kept at approximately 70%, and temperature was set to 20°C during the light period and 10°C during the dark period. Lids were left on the trays for about 8 d, until the seeds germinated. Leaf tissue was harvested from individual plants after 4 wk of growth and was stored at −80°C until DNA extractions were performed.
Growth of Propagules from A. stolonifera Populations
Stolon segments were collected from creeping bentgrass plants in the field and placed in sealed plastic bags. Because A. stolonifera spreads aggressively by vegetative means, samples were collected no less than 5 m from one another to maximize the likelihood of collecting distinct individuals. Sealed plastic bags with stolon samples were stored at 4°C until planting. Individual stolon segments were planted in 10.2-cm (4-inch) plastic pots in water-moistened potting media and placed in growth chambers and grown as above. After approximately 4 wk, leaf tissue was harvested from the plants for DNA extraction.
Crosses to Produce Intraspecific A. stolonifera Hybrids
Agrostis stolonifera is predominantly an out-crossing species with little to no self-fertilization (Belanger et al., 2003a; Wipff and Fricker, 2001). A single plant from the Ona Beach A. stolonifera population was used as a maternal parent in crosses with creeping bentgrass plants resistant to glufosinate [2-amino-4-(hydroxymethylphosphinyl)butanoic acid], allowing herbicide resistance to be used as a selectable marker for intraspecific hybridization. Glufosinate-resistant creeping bentgrass plants containing the bar gene were used as paternal plants (Fu et al., 2005, 2007). Maternal and paternal plants were placed in close proximity to one another in enclosed hybridization chambers in a greenhouse, and one to several panicles from each plant were taped together in a pollination bag (no. 421, Lawson Bags, Northfield, IL). The pollination bags were left on the plants until panicles were harvested approximately 8 wk later. Plants were watered throughout the period before seed harvest.
Seeds and chaff from the maternal parent (Ona Beach non–glufosinate-resistant) were removed from the panicles by rubbing the panicles by hand through a 1.0-mm mesh sieve. Seeds and chaff passing through the 1.0-mm mesh sieve were mixed in plastic bags with 100 mL of autoclaved sand and about 20 mL of water and were chilled at 4°C for 7 to 10 d. The chilled moist seed and sand mixture was spread onto the surface of a plastic tray filled with moistened potting media and grown as described above.
Germinated seedling counts were made about 3 wk after trays were planted, after which time the trays were sprayed with Liberty herbicide (glufosinate-ammonium; Bayer CropScience, Research Triangle Park, NC) at 2× field application rate (f.a.r. = 1005 mL [34 fluid ounces] Liberty per acre) using a track sprayer (RC-500-100-EP, Mandel, Guelph, ON, Canada) to identify which seedlings were glufosinate resistant (bar +). Two weeks after being sprayed, surviving plants were transplanted from the seedling trays into potting media (Bedding Mix #2, OBC Northwest). These plants were confirmed as hybrids by testing for the presence of the PAT protein indicating the presence of the bar gene using Trait-Chek LL test strips (Strategic Diagnostics, Newark, DE).
Genomic DNA from individual plants was extracted using DNeasy Mini Plant Kits (Qiagen, Valencia, CA) following the manufacturer's protocol. DNA concentrations were measured with a NanoDrop ND-1000 (Thermo, Waltham, MA). Genomic DNA was stored at −20°C until used in analyses described in following sections.
Molecular ID Confirmation
To check the species-level identifications of wild-collected A. stolonifera samples and other accessions used in this study, matK chloroplast DNA regions from three randomly chosen individuals from each population were directly sequenced by methods described in Reichman et al. (2006) Sequences were compared to existing matK accessions in GenBank by BLAST searches (Altschul et al., 1990), and a novel sequence for A. transcaspica was deposited in GenBank under accession number GU338987. Sequences that were identical to existing GenBank accessions for the same taxa were not uploaded to the database. Insertions and deletions within the aligned sequences from this study were coded as present or absent (Simmons and Ochoterena, 2000). Sequences were then compared through parsimony analysis using PAUP* v4.03b10 (Swofford, 2003) with methods and settings as in Reichman et al. (2006)
Creation of SSR Libraries
Three genomic libraries, enriched for SSRs, were created from A. stolonifera ‘Crenshaw’, wild A. stolonifera (Ona Beach), and A. transcaspica (PI 283174) using the FIASCO protocol, adaptors, and primers (Zane et al., 2002). Genomic DNA from each source that was digested with MseI and simultaneously ligated to the MseI adaptor was pre-amplified with MseI-N primers. Amplicons were purified with Qiagen Gel Extraction Kits per the manufacturer's instructions and were separately hybridized with the following biotinylated probes: (AAC)8, (AAG)8, (AC)13, (AG)12, (AGG)7, (AT)13, (ATG)8, and (GAAT)8 (Operon, Huntsville, AL). All steps of the hybridization, streptavidin bead capture, magnetic separation, and washes conformed to the Zane et al. (2002) protocol.
Eluted DNA was precipitated, resuspended, and re-amplified using the MseI-N primer as in Zane et al. (2002) The SSR-enriched amplicons were purified as described above, then TA cloned (Invitrogen, Carlsbad, CA). Cells from transformed colonies were lysed by boiling in nuclease-free water and screened by standard 25-μL polymerase chain reaction (PCR) amplifications using M13F and M13R vector primers to detect inserts. Polymerase chain reaction products from positive clones were purified as above and were labeled by thermal cycling using BigDye v3.1 chemistry (Applied Biosystems, Carlsbad, CA) with the M13F and M13R primers, and sequence data were collected on an ABI 3100 Genetic Analyzer. Sequence contigs for each clone were assembled using Seqman (DNAStar, Madison, WI). Consensus sequences for each clone were annotated with SeqBuilder (DNAStar) and all direct repeats were mapped in GeneQuest (DNAStar). Simple sequence repeat loci specific primers were designed with PrimerSelect (DNAStar).
Polymerase chain reaction conditions for each of the primer pairs were then optimized using Mastercycler gradient thermal cyclers (Eppendorf, Hauppauge, NY). All optimized thermal profiles consisted of an initial 95°C for 3 min, 35 cycles of 95°C for 30 s, touchdown of −1°C per cycle for the first 10 cycles to base annealing temperatures for 30 s and 72°C for 1 min, with a final extension of 72°C for 7 min. Reactions were performed in a 25-μL total volume containing 40 ng DNA, 125 μM of each dNTP, 0.4 μM primers, 1× buffer, and 0.04 U μL−1 Taq polymerase (Roche, Indianapolis, IN). Polymerase chain reaction products were purified, cloned, and sequenced as before to confirm that the correct target loci were being amplified and to identify loci that showed polymorphism based on variations in the SSR repeat number. Sequences for target loci and putative orthologs were aligned with Clustal W in MegAlign (DNAStar). Selected primer pairs were then modified for automated fragment data collection. The forward primers were given either 6FAM or 5HEX 5′ fluorescent tags. The reverse primers were modified with a 5′-GTTTCTT “pigtail” to reduce stutter peaks during data collection (Brownstein et al., 1996).
Modified primer pairs were then used to amplify 40 ng of genomic DNA from each individual within populations of interest using the optimized PCR conditions that had been identified for each pair above. The PCR products were diluted to 1/10 and 1/20 with nuclease-free water. Diluted PCR products were combined with appropriate fluorescently tagged molecular weight markers ROX 400HD, ROX 500 (Applied Biosystems), or MapMarker 1000 (BioVentures, Murfreesboro, TN) and HiDi formamide (Applied Biosystems), denatured per the manufacturer's instructions, and then separated by capillary electrophoresis. Size data were collected on an ABI 3100 sequencer using the Foundation Data Collection application (Applied Biosystems). Fragment data was then imported into GeneMapper 4.0 (Applied Biosystems) for genotyping analysis. Separate size standard files had to be created and verified for each run that utilized MapMarker 1000.
Allele Statistics and Analysis of Molecular Variance
POWERMARKER v3.2.5 (Liu and Muse, 2005) was used to generate summary statistics including allele size range, number of alleles, observed heterozygosity, and expected heterozygosity. Pairwise linkage disequilibrium and deviations from Hardy–Weinberg equilibrium (HWE) were evaluated within populations at the 0.05 nominal level for multiple tests using sequential Bonferroni corrections within FSTAT v22.214.171.124 (Goudet, 2001). Significance of pairwise linkage disequilibrium was based on consensus of four analyses of the data with 4320 permutations per run. Significance of deviations from HWE was based on consensus of four analyses of the data with 1080 randomizations per run. FSTAT was also used to calculate allelic richness with rarefaction (El Mousadik and Petit, 1996; Hurlbert, 1971) and to make statistical comparisons of allelic richness between groups. P-values were obtained after 10,000 permutations of the data. The Arlequin v3.5 (Excoffier et al., 2005) analysis of molecular variance (AMOVA) module was used to assess the statistical significance of the percentage of variation of alleles at SSR loci among groups of populations, among populations, and within populations. AMOVA significance tests were based on 16,002 permutations of the data.
Genetic Clustering and Ordination Analyses
To test the robustness of our marker data to detect creeping bentgrass population structures, cluster analyses were run on 87 A. stolonifera individuals from the six sources listed in Table 1 Bayesian clustering analyses were executed using STRUCTURE v2.3.1 (Hubisz et al., 2009; Pritchard et al., 2000). Runs had the following parameters in common: length of burn-in period = 100,000, number of Markov chain Monte Carlo reps after burn-in = 100,000, admixture model default settings, allele frequencies assumed to be correlated, computing the probability of the data for estimating K (inferred number of clusters), 20 iterations per K level for K = 2 through 8. A plateau of the Ln P(D) term (estimated ln probability of the data) across all iterations within each K was used to infer the maximum number of clusters represented in the data. The CLUMPP v1.1.2 (Jakobsson and Rosenberg, 2007) Greedy algorithm was used to permute the STRUCTURE ancestry coefficients for runs at a particular K level to those of the run with the highest LnP(D) at that K CLUMPP was then used to evaluate the G′ (similarity coefficients) for all pairs of runs within the same K CLUMPP executions of 10,000 configurations of runs and clusters were repeated four times per K Outputs from STRUCTURE and CLUMPP were used to create Population Q and Individual Q matrices respectively that were run within DISTRUCT v1.1 (Rosenberg, 2004) to generate cluster graphics.
To provide a contrasting measure of the ability of the nine SSRs to detect divergence between the A. stolonifera populations that we sampled, a principal coordinates analysis (PCoA) was performed using modules of NTSYSpc v2.21c (Rohlf, 2009). Data were transformed into a binary matrix for the presence or absence of each allele at all loci for all individuals. The binary data were then converted into allele frequencies with Simgend using the Rogers' modified distance coefficient (Rogers-W) (Rohlf, 2009; Wright, 1978), which assumes no knowledge of the evolutionary forces acting on the samples being compared (Reif et al., 2005a) and has been shown to be suitable for divergence studies using SSR markers (Balestre et al., 2008). A double center matrix was created by transformation with Dcenter Eigen vectors and values were computed with Eigen and graphics were generated with Mod3D.
STRUCTURE was also used to estimate the ancestry coefficients for the intraspecific A. stolonifera hybrids that were created through the crosses described above. Analyses under the USEPOPINFO model were conducted on 37 individuals consisting of the single Ona Beach maternal parent, nine Crenshaw parental parents, 14 F1 hybrids, and an outgroup of 13 CBG-1 individuals. PopFlags defined the Ona Beach, Crenshaw, and CBG-1 as the “learning samples,” while the hybrids were left as unknowns. Twenty runs with the same burn-in and Markov chain Monte Carlo reps as before were executed at K = 3. The individual ancestry coefficients were also permuted with CLUMPP and graphics were generated with DISTRUCT.
Molecular Confirmation of Agrostis Sample IDs
Results from species identifications based on matK sequences of our samples were consistent with morphological taxonomic identifications of A. stolonifera made in the field and with the stated species of the creeping bentgrass cultivar seed accessions. Furthermore, the relative positions of our A. stolonifera, A. canina, A. gigantea, and A. capillaris samples within a Most Parsimonious Tree (Supplement 2) were in agreement with the recent phylogenetic analyses of Agrostis species (Rotter et al., 2010). These results indicate that the samples used in this study were correctly identified to the species level.
Characterization of Diallelic Polymorphic SSR Loci in A. stolonifera
We enriched nuclear microsatellite libraries directly from accessions of A. stolonifera and from the diploid A. transcaspica for clones containing AC, AG, AAC, AAG, ATG, AGG, and GAAT motif repeats. Six hundred ninety genomic clones were sequenced. Primers designed from 88 clones that contained microsatellites were screened for amplification specificity in creeping bentgrass plants from the six sources described above. Forty-two primer pairs correctly amplified orthologous loci that were cloned, sequenced, and evaluated for size heterogeneity between individuals within the A. stolonifera populations. Fragment analysis data were collected on these same loci and individuals using 55 fluorescently tagged primer pair combinations. From these, nine diallelic polymorphic loci were identified for A. stolonifera (Table 2 ).
|Locus/GenBank||Source||Insert repeat motif||Primer sequences (5′-3′) with “pigtails” underlined||TD|
|AgrosSSR1||A. stolonifera Ona Beach||(AC)10||F: FAM-GGCAACCTCAGTCACACCCTCTCC||63|
|AgrosSSR2||A. stolonifera Ona Beach||(G)11, (CAA)65(CAC)2(CAA)14||F: FAM-CAACGCCTGGTGGATGAAG||58|
|AgrosSSR3||A. stolonifera Ona Beach||(CT)20,(CA)8||F: FAM-GTGCCGAGTACAATTGTGGAAGGAGAG||63|
|AgrosSSR4||A. stolonifera Ona Beach||(AAC)12||F: FAM-GTCGCCAAACACTAAAACTCAGAAAAG||60|
|AgrosSSR5||A. stolonifera cv. Crenshaw||(A)9, (CA)8||F: HEX-GCACAGGCGGCCGTCTTTATTT||63|
|AgrosSSR6||A. stolonifera cv. Crenshaw||(CAA)17||F: HEX-ACCCCAATTGAATAGGAGTAGG||60|
|AgrosSSR7||A. stolonifera cv. Crenshaw||(AAC)5A(AAC)15||F: FAM-AGTGTTTCCCTGAGGCCCCTGAGTTTC||58|
|AgrosSSR8||A. transcaspica PI 283174||(GAA)24||F: FAM-ACCGTGAATGCAGTAATGA||52|
|AgrosSSR9||A. transcaspica PI 283174||(ATG)6(AGG)9||F: HEX-AGAGATTTGGAGACGCCGCTGGAGATG||58|
Five of the 92 A. stolonifera plants produced more than two alleles at various, but not all loci. Fragment size data from these plants were excluded from further analyses. In total, 160 SSR alleles were characterized in A. stolonifera (Table 3 ). The number of alleles per locus ranged from 45 (AgrosSSR2) to 7 (AgrosSSR5, AgrosSSR9). For AgrosSSR2 and 7, which had the highest allele numbers and widest allele size ranges, additional sequencing demonstrated that the peaks called as alleles were in fact orthologs (GenBank PopSets 284808624 and 284808637). AgrosSSR2 had the greatest allelic richness per locus for all populations at 10.87, while AgrosSSR7 had the least at 3.77. The mean availability of loci across all samples was 0.9438.
|Penn A-4||Crenshaw||CBG-1||Baskett Slough||Ona Beach||Rainbow|
|(n = 15)||(n = 15)||(n = 13)||(n = 14)||(n = 17)||(n = 13)|
|Locus||a||Allele range (bp)||r||HO||HE||A(p)||HO||HE||A(p)||HO||HE||A(p)||HO||HE||A(p)||HO||HE||A(p)||HO||HE||A(p)|
|Crop summary: Total A(p)||135(26)||Wild summary: Total A(p)||155(58)|
|Rs crop||3.9||Rs wild||4.5|
Based on analyses of the data using FSTAT with a Bonferonni adjusted P-value for the 5% nominal level of P = 0.000231, AgrosSSR 2, 3, 4, 6, and 7 only showed significant linkage disequilibrium within the Baskett Slough population. Linkage disequilibrium was not detected between the loci in the remaining five sampled populations. The only population for which several loci (AgrosSSRs 1–4 and 9) appeared to deviate from HWE (Bonferonni adjusted P-value for the 5% nominal level of P = 0.00093) was Ona Beach.
When examining the number of private alleles present within the crop and wild groups, 19% of the alleles found in the crop A. stolonifera were unique, whereas 37% of alleles found in the wild collected creeping bentgrass were unique. All three crop subpopulations (Penn A-4, Crenshaw, and CBG-1) had 18 to 20% unique alleles. The percentage of unique alleles found in each of the wild subpopulations (Baskett Slough, Ona Beach, and Rainbow) was higher than any of the crop subpopulations. Ona Beach had the highest number of unique alleles with 44%, while Rainbow and Baskett Slough populations had 36 and 25%, respectively. Because the number of private alleles observed can be highly dependent on sample size, we also evaluated allelic richness with rarefaction (El Mousadik and Petit, 1996; Hurlbert, 1971). Although the unweighted allelic richness averaged over all loci for the crop group (3.9) was lower than that of the wild group (4.5), that difference was not statistically significant.
By contrast, AMOVA results as a weighted average over all loci (Table 4 ) indicated that there was a small (3.8%) but significant (P ≤ 0.05) variation between the crop and wild groups. Substantially more of the percentage variation of the data was present among the six populations (24.3%) and within the populations (71.9%), both of which had P-values ≤ 0.001.
|Source of variation||Percentage variation||F-statistic|
|Between groups (crop and wild)||3.80||FCT: 0.03796|
|Among populations within groups||24.29||FSC: 0.25246|
|Within populations||71.92||FST: 0.28084|
Agrostis stolonifera Population Clusters
At the broadest view of the data (K = 2) STRUCTURE resolved two clusters separating crop and wild samples. CLUMPP permutations of the STRUCTURE membership coefficients revealed that there were two dominant cluster modes for runs at K = 2, both of which generally distinguished crop and wild genotypes. Eleven runs, including the highest probability run presented in the upper K = 2 panel in Fig. 1A , had pair-wise G′ similarity coefficients ≥0.9768. An additional three runs with G′ ≥ 0.9965 had the mode presented in the lower K = 2 panel of Fig. 1A The main distinction between these two modes was membership assignment for certain individuals from the Baskett Slough population.
Simple sequence repeat data were examined to a maximum possible population state of K = 8, but the highest inferred cluster number for which there was a relative plateau in LnP(D) values across STRUCTURE runs was K = 6. As before, CLUMPP permutations also identified two main cluster modes of membership coefficients at K = 6. Five runs at this level, including the highest probability run presented in the upper K = 6 panel in Fig. 1A, had pair-wise G′ ≥ 0.9921. An additional six runs were nearly identical (G′ ≥ 0.9882) to the membership coefficients shown in the lower K = 6 panel. Both modes assigned the majority of individuals to clusters that corresponded to their sources. Again, the most notable discrepancy between the two modes was with regard to individuals from Baskett Slough.
A contrasting visualization of the degree of relationship between A. stolonifera individuals is represented by the PCoA of the SSR data shown in Fig. 1B For all 87 individuals from the six populations, the first three dimensions of the ordination explain 10.9, 10.5, and 8.6% of the variation within the data. The six clusters derived by PCoA are largely in agreement with those identified by STRUCTURE at K = 6, particularly for the second mode at that level. The PCoA placed 11 Baskett Slough individuals within the Rainbow group. The remaining Baskett Slough individuals clustered separately.
Identification of Intraspecific Hybrids
Fragment analysis data was collected for each of our nine SSR loci from 14 PAT-positive progeny plants that were grown from seeds formed on the PAT-negative Ona Beach parent plant. This single wild parent plant acted as a recipient of pollen from transgenic Crenshaw plants used in the controlled crosses to produce known intraspecific hybrids. Because the Ona Beach population was well resolved from Crenshaw in the STRUCTURE runs at K = 6 and in the PCoA above, the expectation for the intraspecific hybrids was that alleles from Ona Beach and Crenshaw parents would be observed in each individual. From the 30 alleles scored for the hybrids across all loci, 11 were present in the Ona Beach parent and 21 were distributed between nine Crenshaw individuals. Four of the alleles were shared by the Ona parent and certain Crenshaw plants and an additional three were unaccounted for (not observed in any parents). One Crenshaw plant and the Ona Beach parent did not amplify with AgrosSSR4 primers, although this locus did amplify correctly in 13 of 15 Crenshaw plants and 13 of 17 Ona Beach plants. Two additional Crenshaw plants did not amplify with AgrosSSR8 primers. Overall, there was a mean of six loci per hybrid for which one allele per locus matched the Ona parent and the other allele matched at least one of the Crenshaw plants. The highest probability STRUCTURE run for SSR data from the Ona Parent, the Crenshaw parents, the hybrids, and CBG-1 outgroup individuals at K = 3 is shown in Fig. 1C CLUMPP permutations of STRUCTURE outputs demonstrated that for all 20 runs at K = 3, pair-wise G′ ≥ 0.8388. This single clustering mode consistently showed hybrid genotypes as admixed between parent genotypes.
Cross-Species Amplification of Orthologs
Seven of nine markers (all except AgrosSSR4 and 6) amplified and were polymorphic in the other Agrostis species included in this study (Table 5 ). The amplification of orthologs in these taxa was confirmed by sequencing, as was previously done for A. stolonifera (GenBank PopSets 284808620, 284808624, 284808629, 284808633, 284808637, 284808644, and 284808647). All seven markers were diallelic in A. gigantea For this species 58 alleles were amplified and AgrosSSR8 was the most polymorphic with 13 alleles. Six of the seven SSRs (AgrosSSR2, 3, 5, 7–9) were also diallelic in A. transcaspica, but did not amplify in A. canina or A. capillaris Twenty-one SSR alleles were identified in A. transcaspica, and as with A. gigantea, AgrosSSR8 was the most diversified with seven alleles. From these six loci, AgrosSSR2, 3, 5, and 7 were developed from A. stolonifera, while AgrosSSR8 and 9 originated from A. transcaspica clones. AgrosSSR1, which was derived from A. stolonifera, amplified in A. canina, A. capillaris, and A. gigantea but not in A. transcaspica
|A. canina (n = 29)||A. capillaris (n = 30)||A. gigantea (n = 30)||A. transcaspica (n = 16)|
|Locus||GenBank accession||Allele range||a||GenBank Accession||Allele range||a||GenBank Accession||Allele range||a||GenBank Accession||Allele range||a|
The primary objective of this study was to identify and develop diallelic, genome-specific microsatellite loci for the allotetraploid creeping bentgrass (A. stolonifera), which is an important agricultural turf crop that hybridizes with many different related species. We have described the development of nine diallelic SSR markers and have tested their utility for resolving genetic structures of A. stolonifera populations, for intraspecific hybrid identification, and for cross-species reactivity. The species of samples used in this study were confirmed by morphological and chloroplast matK sequence-based identifications.
Creeping Bentgrass Differentiation Detected by SSRs
Crop species, particularly those that have undergone a long history of domestication, may have less genetic variation than wild or unmanaged populations (Hoisington et al., 1999; Lee, 1998; Reif et al., 2005b). The degree of polymorphism present in the microsatellite markers developed in this study allowed us to assess the relative genetic variation present within a representative panel of both crop and wild creeping bentgrass populations. While it might be expected that the greatest genotypic uniformity would occur within the A. stolonifera cultivars due to selective breeding, that was not apparent for our samples.
The data collected with nine SSR markers suggest that the domesticated individuals used in our representative panel are still closely related to the naturalized creeping bentgrass populations that we sampled. Nevertheless, the results of our AMOVA, STRUCTURE analyses, and PCoA indicate that our markers were sufficiently polymorphic to detect divergence between crop and wild groups. They have also shown to be useful for quantifying diversification among and within A. stolonifera populations. Because SSRs tend to be rapidly evolving loci, they are well suited for detecting both relatively recent divergence and admixture between populations.
Our wild creeping bentgrass populations were purposefully chosen from sites that were geographically distant from each other. Both STRUCTURE and PCoA resolved clusters that largely matched the sampling locations, with the exception of the anomalous groupings for individuals from Baskett Slough. Our SSR results suggest that the Baskett Slough population may consist of recent migrants from two distinct sources, one of which is genetically similar to individuals sampled from the Rainbow site. A founder effect (Mayr, 1942) could explain the apparent linkage disequilibrium between our markers observed within the Baskett Slough population.
Although additional low levels of admixture between our wild samples were indicated by the STRUCTURE results, this should not necessarily be interpreted as evidence of current gene flow between these widely distributed populations. It is possible that the shared alleles are common in the naturalized creeping bentgrass populations of Oregon in general. However, in circumstances in which gene flow is occurring between sympatric populations, the signature of admixture that is detectable with the markers developed here is likely to be more pronounced.
Marker Detection of Agrostis Hybrids
In the Agrostis system, morphological distinctions are often inadequate for determining species identification (Madison, 1971; Vergara and Bughrara, 2004), much less verifying the presence of hybrid individuals. We have shown that the SSR markers developed here are sufficiently polymorphic to detect the genotypes of F1 intraspecific creeping bentgrass hybrids. Additional testing is needed to determine their ability to identify backcrossed hybrids as well. A higher number of Agrostis SSRs may be needed to distinguish individuals derived from more advanced generations of backcrossing as has been shown for RAPD and RFLP markers applied to other species (Boecklen and Howard, 1997), but this remains to be seen.
Cross-Species SSR Orthologs and the Possible Diploid Source of the A3 Subgenome
The contrasting patterns of amplification of orthologs (confirmed by sequencing) in species with different but overlapping genomic complements match all three of our predictions in the Introduction Four diallelic loci that were identified from A. stolonifera (A2A2A3A3) had orthologs in A. transcaspica (putative A3A3) and A. gigantea (A1A1A2A2A3A3), but not in A. capillaris (A1A1A2A2) or A. canina (A1A1 or possibly A2A2). Two diallelic loci that were identified from A. transcaspica had orthologs in A. stolonifera and A. gigantea, but not in A. canina or A. capillaris. One diallelic locus identified from A. stolonifera had orthologs in A. canina, A. capillaris, and A. gigantea, but not in A. transcaspica The results suggest that AgrosSSR2, 3, 5, and 7–9 are A3 genome specific, and that AgrosSSR1 was isolated from the A2 subgenome of A. stolonifera These findings are consistent with the Rotter et al. (2010) conclusion that A. canina contains the A2 genome. Additional A2–specific SSR loci may be isolated in the future from A. canina Mapping of our SSR loci to the previously reported linkage groups within the A. stolonifera genome (Chakraborty et al., 2005) is needed to confirm the genomic distribution of these loci. Still, the cross-reactivity of our SSR loci indicates potential utility of several markers beyond A. stolonifera, particularly for analyses of A. gigantea population structure and A. gigantea × A. stolonifera hybrid zones. Our results also provide evidence supporting the hypothesis that A. transcaspica is the ancestral source of the A3 subgenome found in A. stolonifera and A. gigantea It is possible that A. transcaspica may have been the diploid paternal parent of A. stolonifera
We have described the characterization, applications, and cross-species utility of nine nuclear SSR markers developed for A. stolonifera Additional diallelic, codominant markers are needed for population analyses of other allopolyploid plant species. Genome-specific, diallelic markers allow population genetic data of strict allopolyploids to be analyzed as if these species were diploids. Markers isolated from the different subgenomes of a strict allopolyploid may also provide contrasting insights regarding the genetic diversity of populations and the evolutionary history of the species. The development of such markers will be more challenging for segmental allopolyploid species such as Agrostis castellana Boiss. and Reut., in which pairs of subgenomes may contain at least partially homoeologous chromosomes or there may be more than two copies of each subgenome. High-throughput, whole genome sequencing will make this process much more efficient. The SSRs characterized here are expected to complement molecular markers that have been previously developed for A. stolonifera.
Supplemental Information Available
Supplemental information associated with this manuscript is located at http://www.crops.org/publications/cs