Poa pratensis L. is a forage and turf grass species well adapted to a wide range of mesic to moist habitats. Due to its genome complexity little is known regarding evolution, genome composition and intraspecific phylogenetic relationships of this species. In the present study we investigated the morphological and genetic diversity of 33 P. pratensis accessions from 23 different countries using both nuclear and chloroplast molecular markers as well as flow cytometry of somatic tissues. This with the aim of shedding light on the genetic diversity and phylogenetic relationships of the collection that includes both cultivated and wild materials. Morphological characterization showed that the most relevant traits able to distinguish cultivated from wild forms were spring growth habit and leaf colour. The genome size analysis revealed high variability both within and between accessions in both wild and cultivated materials. The sequence analysis of the trnL-F chloroplast region revealed a low polymorphism level that could be the result of the complex mode of reproduction of this species. In addition, a strong reduction of chloroplast SSR variability was detected in cultivated materials, where only two alleles were conserved out of the four present in wild accessions. Contrarily, at nuclear level, high variability exist in the collection where the analysis of 11 SSR loci allowed the detection of a total of 91 different alleles. A Bayesian analysis performed on nuclear SSR data revealed that studied materials belong to two main clusters. While wild materials are equally represented in both clusters, the domesticated forms are mostly belonging to cluster P2 which is characterized by lower genetic diversity compared to the cluster P1. In the Neighbour Joining tree no clear distinction was found between accessions with the exception of those from China and Mongolia that were clearly separated from all the others.
Citation: Raggi L, Bitocchi E, Russi L, Marconi G, Sharbel TF, Veronesi F, et al. (2015) Understanding Genetic Diversity and Population Structure of a Poa pratensis Worldwide Collection through Morphological, Nuclear and Chloroplast Diversity Analysis. PLoS ONE 10(4): e0124709. https://doi.org/10.1371/journal.pone.0124709
Academic Editor: Manoj Prasad, National Institute of Plant Genome Research, INDIA
Received: November 18, 2014; Accepted: March 8, 2015; Published: April 20, 2015
Copyright: © 2015 Raggi et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
Data Availability: All relevant data are within the paper and its Supporting Information files.
Funding: The research was funded by the Italian Ministry of Education and Scientific Research (MIUR) “Genetic aspects of seed production: genomic and functional characterization of candidate genes for apomictic reproduction” and by the Dipartimento di Scienze Agrarie Alimentari e Ambientali of Università degli Studi di Perugia. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors confirm that Emidio Albertini is an Academic Editor and that this does not alter the authors' adherence to PLOS ONE Editorial policies and criteria.
Poa pratensis L., also known as Kentucky bluegrass, is a hardy, persistent, and attractive forage and turf grass species that is well adapted to a wide range of mesic to moist habitats. It belongs to the Poaceae (subfamily Pooideae, tribe Poacea), a family with more than 500 described species  and can hybridize with P. secunda, P. arctica, P. alpina, P. nervosa, P. reflexa and P. palustris  to form allopolyploids, many of which are facultative apomicts . Hybridization is hypothetically facilitated by apomixis, as asexual reproduction provides an escape from the typical reproductive problems associated with unbalanced or mixed genomes. Bashaw and Funk  suggested that the great adaptiveness of this species is likely associated with its variable ploidy, ranging from 2n = 22  to 2n = 147 , and its versatile mode of reproduction . In fact the perpetuation of a single genotype through apomixis is an advantage in determining the P. pratensis circumpolar distribution  while sexual reproduction represent an important source of new genetic variation [3,9]. Studying a core collection of this species, Wieners and collaborators reported that the majority of the populations contained facultative apomicts (with a combination of reduced, zygotic and unreduced, parthenogenic embryo production), in addition to obligate sexual or obligate apomictic accessions .
An allopolyploid origin of P. pratensis is supported by its great polyploid and aneuploid variation, a consequence of which is that little is known regarding its evolution, genome composition and intraspecific phylogenetic relationships. Based on morphological, cytological and species diversity, it is generally accepted that P. pratensis originated mainly from Europe and Asia [11,12], followed by introduction to North America by European settlers in the Seventeenth Century . According to other authors, some populations found in remote mountain meadows of western USA or the Appalachian mountains might be considered native [14,15] together with some ecotypes identified in the Rocky Mountains . As karyology is by definition difficult in Poa, flow cytometry has been employed to investigate its genome size [10,17,18] and of other turfgrass species . Since variation in DNA content may not be accounted for using different molecular markers, but may nonetheless have a significant role in determining the ultimate phenotype, this rapid and inexpensive method is useful in polyploids . In addition genome size per se has adaptive significance influencing the phenotype by the expression of its genic content and by the physical effects of its mass and/or volume , and may additionally affect cross compatibility between genotypes. Finally, in angiosperms, DNA amount can correlate with a several relevant characters such as minimum generation time and ecological behaviour , and therefore deserves some attention in phylogenetic studies.
Kentucky bluegrass cultivars and accessions have previously been characterized using random amplified polymorphic DNA (RAPD) [21–23], inter-simple sequence repeat (ISSR)  and, only very recently, using simple sequence repeats markers (SSR) [25,26]. Although microsatellites are codominant markers, the difficulty in identifying the number of allele copies in polyploids limits their use. Here we attempt to overcome this hurdle by investigating the morphological and genetic diversity of 33 P. pratensis accessions using flow cytometry, sequencing of the trnL-F chloroplast region, in addition to nuclear and chloroplast microsatellite markers.
Materials and Methods
Plant materials, morphological characterization and total DNA extraction
The phenotypic and genomic variability of 33 cultivated and wild accessions of P. pratensis from 23 different countries and 4 continents (Europe, Asia, America and Africa) were investigated (Fig 1; Table 1).
Geographical origin of the 33 accessions of P. pratensis used in this study, along with information of biological status. The figure was modified from Wikipedia and was for representative purposes only.
Seeds were sown in jiffy pots and, and 30 plants were grown for each accession. Total genomic DNA was extracted from 0.5 to 1 g of fresh leaf tissue per plant using the Genelute Plant genomic DNA miniprep kit (Sigma Aldrich) according to the manufacturer’s handbook. DNA concentration was assessed through optical density reading (DU650 spectrophotometer, Beckman) and confirmed by agarose gel electrophoresis.
For morphological characterization, 8 randomly chosen plants per accession were transplanted to the field in a randomized design and evaluated for: growth habit at flowering (scoring plants from 1 = erect to 9 = prostrate), rust resistance/tolerance (1 = r esistant/tolerant—9 = highly susceptible), leaf colour (1 = light green—9 dark green), stem length (mm), width and length of the flag leaf (mm), leaf texture (mm), spring plant regrowth (1 = slow—9 = very fast regrowth) and spring growth habit (1 = erect—9 = prostrate) after cutting. Some of the traits were evaluated according to the UPOV Guidelines . Data were statistically analyzed using univariate one way ANOVA and multivariate discriminant analyses. A stepwise discriminant procedure was carried out in order to look at the most useful traits to use in a phenotypic characterization and for distinguishing populations on the basis of discriminant functions.
Flow cytometry analysis of somatic tissue
Karyotyping was difficult and inaccurate due to the high numbers of chromosomes and variable ploidy levels, hence we estimated genome size through flow cytometric measurements of relative total DNA content in somatic tissues. Fifteen plants from each accession were grown in a greenhouse at the Leibniz Institute of Plant Genetics and Crop Plant Research (IPK, Gatersleben, Germany), and an average of seven plants were randomly selected for flow cytometry. Nuclei were collected from young leaves of single plants by manual chopping them with a razor blade in the following staining buffer: 100 mM Tris-HCl, 5.3 mM MgCl2, 86 mM NaCl, 0.03 mM sodium citrate, 7.3 mM Triton X-100, 3 μM 4‘-6-diamidino-2-phenylindole (DAPI), pH 7.0. After filtration through nylon tissue (30-μm mesh) to remove large debris the tubes were stored in the dark on ice for 1 to 4 h before measurement . Fluorescence intensity of DAPI–stained nuclei was measured on a Ploidy Analyser PA (Partec, Münster, Germany).
For each sample the standardized 2C DNA content was obtained by using the following equation: position of sample main peak times position of the internal control main peak (a triploid Boechera holboellii genotype) divided by mean position of the peak of the internal control. The repeated measure of the internal control over multiple days allowed to standardize the main peak position of each P. pratensis samples. The final DNA quantification in pictograms was then estimated through a simple linear regression based on four accessions (PI 286210, PI 298096, PI 355956 and PI 601107) characterized by Wieners and collaborators  and common in both studies (R2 = 0.91, P<0.05). Since individual plants within the same accession might show different somatic tissue DNA content, particularly when multiple reproductive pathways occur, standard errors would be misleading and were therefore not calculated . Differences between wild, cultivated and accessions with uncertain improvement status were analysed by one way ANOVA.
trnL-F region amplification and sequencing
A fragment of the chloroplast trnL-F region was PCR amplified from DNA extracted from 70 randomly chosen individuals representing all 33 collected accessions using primers “c” (CGAAATCGGTAGACGCTACG) and “f” (ATTTGAACTGGTGACACGAG) of Taberlet and collaborators . Ten μl of the PCR products were electrophoresed on a 1.5% (w/v) agarose gel and visualized by ethidium bromide staining. Successful amplifications were purified using Qiaquik PCR purification Kit (Qiagen) following manufacture instructions. Sequencing reactions were performed for both DNA strands using the ABI PRISM BigDye Terminator Cycle Sequencing Ready Reaction Kit (Applied Biosystems) on an ABI 3700 genetic analyser according to the manufacturer’s instruction. Amplification primers used for PCR were also used for sequencing. All sequence alignments were carried out using MEGA version 4 software .
trnL-F region CAPS marker development and analysis
Since only one Single Nucleotide Polymorphism (SNP) was detected in the chloroplast sequenced region (see below) we developed a Cleaved Amplified Polymorphic Sequences (CAPS) marker to test the allelic condition at the locus. PCR primers were developed using PRIMER 3  from multiple alignments of sequences obtained in this study and those retrieved from public databases, and the restriction endonuclease was selected according to the CAPS differential cleavage sites. The allelic condition for the developed CAPS marker was tested on eight individuals randomly chosen from each accession (hereafter population) for a total of 264 individuals. PCR amplifications were carried out using the developed cpCAPfor and cpCAPrew (Table 2) in a 20 μl volume containing 25 ng genomic DNA, 1× PCR buffer (Invitrogen), 200 μM dNTPs, 0.5 μg of Bovine Serum Albumin (BSA), 10 pMol of the cpCAPsf and cpCAPsr primers, and 1 U of Taq DNA polymerase (Invitrogen).
All reactions were performed using the following cycling parameters: 94°C for 5 min, followed by 38 cycles of 94°C for 30 s, 51°C for 30 s, and 72°C for 45 s. Aliquots of 11 μl PCR product were digested with 2 units of BsmI in 1× Buffer R for 2 h at 37°C in a total volume of 25 μl. After digestion, 15 μl of the reaction was electrophoresed on a 2% (w/v) agarose gel and visualized by ethidium bromide staining.
Nuclear and chloroplast SSR analyses
A total of 11 nuclear and two chloroplast single sequence repeat (SSR) markers were used to genotype the 264 P. pratensis samples tested for the CAPS marker allelic condition (Table 3).
Concerning microsatellites derived from P. alpina and P. arachnifera, PCR amplification was performed in a total volume of 20 μl containing 30 ng genomic DNA: 1× PCR buffer (New England Biolabs), 200 μM dNTPs, 3,5 pM of the Forward and Reverse primers and 0,75 U of Taq DNA polymerase (New England Biolabs). Since primer were not originally developed for P. pratensis a touchdown PCR method was used to optimize yield of amplified DNA and amplification specificity . The reactions were performed using the same cycling parameters: 94°C for 3 min, followed by 10 cycles of 92°C for 30 s, 58°C for 40 s (decreasing of 0.7°C for each cycle) and 72°C for 60 s and 25 cycles of 92°C for 30 s, 50°C for 30 s and 72°C for 60 s. All others amplifications were performed in a total volume of 12 μl containing 15 ng genomic DNA: 1× PCR buffer (Sigma Aldrich), 1.5 mM MgCl2, 200 μM dNTPs, 4 pM of the Forward and Reverse primer and 0,75 U of Taq DNA polymerase (Sigma Aldrich). The following PCR cycling parameters were used: 94°C for 3 min, 35 cycles of 94°C for 30 s, 56°C for 30 s, 72°C for 30 s, and an extension final step of 72°C for 20 min. Amplification products were resolved and analyzed with an ABI 3100 Genetic Analyzer (Applied Biosystems), using Dye Set G5 (ABI) labelled primers. Size calculation and data record of generated amplicons were performed using GeneMapper 3.0 software (Applied Biosystems).
Population structure analysis
Considering variable ploidy in P. pratensis and the fact that it was not possible to infer relative dosage of PCR products, nuclear SSR alleles were scored as dominant markers (i.e. presence/absence of amplicons). We carried out the analysis by using the software STRUCTURE ver. 2.3.4 [33,34] to investigate the genetic population structure of the 264 P. pratensis individuals from 33 populations. This software has been successfully used to evaluate genetic structure of individuals on different apomictic species, such as for example those belonging to Crataegus, Brachiaria, Panicum, Taraxacum, and Paspalum genera [35–39]. The approach uses a Markov Chain Monte Carlo (MCMC) algorithm to group individuals into populations (clusters) on the basis of multilocus genotype data [33,34]. In this study the procedure that accounts appropriately for the genotypic ambiguity inherent in dominant markers  was followed. The number of clusters (K) was estimated by carrying out ten independent runs for each K (from 1 to 10), using 30,000 MCMC repetitions and 30,000 burn-in periods, assuming correlated allele frequencies (Run 1). No a priori population information was used. The percentage of membership (q) of each individual in each of the inferred K clusters was computed by one additional STRUCTURE run for 100,000 MCMC repetitions and 100,000 burn-in periods (Run 2).
Two different approaches were used to set the number of clusters (K), the ad hoc statistic ΔK  and the change in H’ (average symmetric similarity coefficient) from CLUMPP software . The latter indicates the goodness of fit to assign each individual to a given cluster(s), consistently across STRUCTURE independent runs. For each K, the results of Run 2 were compared with those obtained by merging the 10 initial STRUCTURE simulations (Run 1) using the CLUMPP software. CLUMPP was used with the Greedy Algorithm method.
Data on chloroplast DNA variation for both cpSSR (atpB-rbcL) and CAPS marker (trnL-F) were used to classify individuals into different groups on the basis of their relative alleles.
Genetic diversity and phylogeographic analysis
Genetic diversity analysis was carried out using i) the nuclear SSR raw data of the 33 populations and ii) the clusters identified by the software STRUCTURE. For the former analysis, the gene diversity Hj, analogous to the unbiased expected heterozygosity He , was computed from allele frequencies estimated following the approach of Lynch & Milligan . The estimates were obtained by AFLPSURV , a software using a Bayesian method based on a non-uniform prior distribution of the allele frequencies , and able to produce unbiased estimates in dominant markers . For the latter analysis, the FK parameter for each of the K clusters  was estimated using the software STRUCTURE. The parameter FK represents the estimated drift from the inferred common ancestor of all populations, thus similar to FST but specific for each cluster and expected to be proportional to the divergence from a common ancestral population. A low FK value indicates little drift away from the ancestral state. The Tukey-Kramer HSD (Honestly significant difference) procedure was used to test for differences in gene diversity estimates among the 33 populations, while the Wilcoxon signed-ranks non-parametric test for two groups, arranged for paired observations, (i.e. one pair of estimates for each locus) , and Bonferroni correction for multiple comparisons was used to test for significant differences among clusters identified by STRUCTURE analyses.
Finally, to further shed light on the relationships among populations, a consensus NJ tree based on FST estimates between the 33 populations was obtained; 500 matrices of FST were computed by bootstrapping over loci by using AFLP-SURV  on the basis of allele frequencies estimated by the Zhivotovsky  approach. The CONSENSE procedure implemented in the PHYLIP software package  was used to infer bootstrap values.
The association between the clusters, as obtained by a genetic diversity analysis at nuclear SSR, and the chloroplast data, for both cpSSR (atpB-rbcL) and CAPS marker (trnL-F), was tested by analysis of contingency tables with the likelihood ratio chi-squared (χ2) test. To evaluate whether the nuclear DNA content was significantly associated with membership coefficients of identified K clusters and alleles of chloroplast markers a bivariate and one way analysis were performed. Finally, a correlation between morphological traits and DNA content was carried out on mean population values. All the test were performed using the JMP 8.0 software (SAS Institute Inc. 2008, Cary, NC, USA).
The phenotypic characterization of 32 P. pratensis populations (no plants of XD accession survived transplanting) showed highly significant differences for each of the examined traits (P<0.001; Table 4).
When populations of known origin were separated into cultivated and wild groups, traits such as spring growth habit, leaf colour, leaf texture and spring regrowth, showed significant differences (P<0.05), with cultivated accessions more prostrate, more tolerant to rust, having a darker green colour and slower spring regrowth as compared to wild accessions (Table 4). Stem length, a trait normally used to characterize a variety for registration procedures, was higher in wild compared to cultivated populations. Leaf texture and width of the flag leaf were higher in cultivated accessions, while no statistical significances were found for inflorescence pigmentation and flag leaf length (Table 4).
The stepwise discriminant procedure included all recorded traits and increased in each step the averaged scored canonical correlation (P<0.001). Out of the ten possible discriminant functions found, the first eight were highly significant, with the first three able to explain over 66% of the total variability (29, 23 and 14%, respectively). The first discriminant function was highly and positively correlated with spring growth habit (0.61), the second negatively correlated with leaf colour (-0.70) and the third positively correlated with pigmentation of the inflorescence (0.64) and spring regrowth (0.60). The centroids of the class means, plotted according to the first three functions, showed that cultivated and wild accessions were clearly separated by the first discriminant function (cultivated accession being more prostrate than wild); no clear cut was observed according to the second discriminant function, although the leaf colour of some cultivars (B, N and V) were generally darker than many wild accessions (E, XC, XI). The third function was able to separate the cultivated from the uncertain origin basically for the presence of inflorescence pigmentation (Fig 2). Accession V, the cultivar Princeton from the USA, was unique for its prostrate growth habit and dark green colour of the leaves. Similar characteristics were shown by accession I and XM, both of uncertain improvement status and from East European countries (Poland and Czeck Republic, respectively). Also landrace M from Hungary clustered apart and was characterized by a light green colour, early spring regrowth and intense inflorescence pigmentation.
Genome size analysis
The total variation in terms of DNA content among 176 P. pratensis genotypes (JG and V were not included) ranged between ~4.08 pg (accession B) to ~18.66 pg (XO) (Fig 3), while the mean DNA content of each accession ranged from ~4.91 (accession M) to ~12.69 pg (XO) (Table 1). The overall data reveal high variability both within and between populations (Fig 3).
Box plot representation of Nuclear DNA content (pg/2C) of successfully characterized P. pratensis accessions. The “whiskers” are drawn from the top of the box up to the largest data point less than 1.5 times the box height from the box (the "upper inner fence"), and similarly below the box. Values outside the inner fences are shown as circles, values further than 3 times the box height from the box are shown as stars.
Uniform DNA content within accessions were found in M, R, S, JD, JE, XA, XD, XG, XL, XM and XP, while all others showed two distinct groups. DNA content of plants that showed contrasting values compared to the majority of plants of the same accession is reported in brackets in Table 1 last column.
Considering only the somatic DNA content of the majority of plants present in each accession, no significant differences were detected between cultivated, wild and populations of uncertain improvement status. The distribution of DNA content of the accessions clearly shows a bi-modal distribution, with two major peaks, around 6–7 and 9–10 pg values (S1 Fig). According to the relationship between DNA content and chromosome number, developed for Kentucky bluegrass by Eaton and collaborators , the most common chromosome numbers in the collection are between 52 and 60 followed by those between 60 and 68. DNA content was correlated with some morphological traits: in cultivated materials it was negatively correlated with the spring regrowth (-0.66) and with the length and width of the flag leaf (-0.67 and -0.71, respectively), while in wild populations it was negatively correlated with the intensity of the green colour (-0.56) and positively correlated with the rust tolerance (0.54) (all P<0.05).
Using two chloroplast SSR primer sets, 5 discernible DNA fragments were obtained: the atpB-rbcL locus was polymorphic with four different alleles (A, B, C and D, respectively) (Fig 4, chloroplast DNA variation part) while rpoC2-rps2 was monomorphic and therefore not included in successive analysis.
Molecular characterization of the 264 genotypes used: chloroplast data, each individual is represented by a vertical line coloured according to the allele type of the relative locus; nuclear data, each individual is represented as a vertical line, partitioned into K coloured segments, which represent the individually estimated membership fractions in the two clusters identified.
The portion of the sequenced trnL-F region yielded sequences ranging from 561 to 566 bp. The alignment of the produced sequences with those available in the NCBI database (AY061951 and AY504641) revealed the presence of only 2 polymorphisms: i) a 5 bp indel (AATAT) at position 328 bp found only in Mongolian (XD) and Indian (XO) populations, and ii) a transversion from adenine to cytosine at position 359 bp (S2 Fig). In the 264 genotypes tested with the CAPS marker developed in this study, adenine type incidence was 51.1% (135 genotypes) while cytosine type incidence was 48.9% (Fig 4, chloroplast DNA variation part). According to SNP allelic state, the 33 studied populations were grouped in: i) adenine containing (13 populations, 39.4%); ii) cytosine containing (10 populations, 30.3%) and iii) adenine and Cytosine containing group (10 populations, 30.3%).
Scorable amplicons were produced for 9 of the 11 nuclear SSR, as loci Poa310 and Poa406 produced unclear electrophoretic profiles (Table 3). Locus SSR AE11 was monomorphic, while the remaining loci were polymorphic with a maximum of 18 alleles (for Poa290 and Poa415) and an average of 10.1 (Table 3). The number of alleles per genotype varied from a minimum of 1 (as for several genotypes at Poa001, Poa002, Poa290, Poa410, Poa414 and AE11 loci) to 10 different alleles. A total of 91 different alleles were detected.
Population structure, genetic diversity and phylogeographic analysis
The ln likelihood estimates increased progressively at increasing values of K (S3A Fig). On the basis of the method of Evanno and collaborators  (S3B Fig), the genotypes were split into K = 2 groups (hereafter referred to as cluster P1 and P2). A lower peak of ΔK was identified at K = 5 (S3B Fig). The highest H’ value from CLUMPP was that for K = 2 (H’ = 0.76), confirming the indication obtained from the Evanno’s approach. As shown in S3C Fig, the H’ estimates for other K values were comprised between 0.59 at K = 3 and 0.45 at K = 7, thus not supporting any lower hierarchical structure. The percentages of memberships (q values) at K = 2 were then recalculated by a new run based on 100,000 repetitions and burn-in periods. The q values so obtained were consistent with those of CLUMPP (H’ = 0.82), hence we used those from the STRUCTURE’s run. A threshold value of q = 0.70 was used to assign genotypes to one of the inferred K clusters (Fig 4).
At K = 2 the cluster P1 included 93 genotypes, 63 (67,7%), 17 (18.3%), and 13 (14%) were wild, cultivated and of uncertain improvement status, respectively. Seventy-eight (83.9%) and 15 (16.1%) P1 genotypes showed the ‘A’ allele and the ‘C’ allele of the CAP locus, respectively. Cluster P2 included more genotypes (158), and of these 61 were wild (38.6%), 71 cultivated (44%) and 26 had uncertain biological status (16.5%). Most of them (69%) carried the ‘C’ allele, with the remaining (31%) carried the ‘A’ allele of the CAP locus. Thirteen genotypes were admixed (0.30<q<0.70), among them 12 were wild and one with a doubtful biological status (Fig 4). Wild materials were assigned to both P1 and P2 clusters (mean qP1 = 0.51 and mean qP2 = 0.49), while cultivated genotypes were mostly assigned to cluster P2 (mean qP2 = 0.79; Fig 5).
Molecular characterization of the P. pratensis genotypes grouped for biological status: wild (W), cultivated (C) and Unknown (Un). Chloroplast DNA variation: each group of individuals is represented as a vertical line coloured according to the percentage of different alleles detected at the relative locus. Nuclear DNA variation: each group of individuals is represented as a vertical line, partitioned into K coloured segments, according to the percentage of estimated membership in the two clusters identified.
At the chloroplast level for the locus CAP (trnL-F) we found contrasting results for wild and cultivated materials, with the wild mostly represented by genotypes carrying the ‘A’ allele (71%) and the cultivated mostly represented by genotypes carrying the ‘C’ allele (82%). Finally, the DNA content was not significantly correlated with qP1 values and the two chloroplast loci. Cluster P1 was characterized by a significantly higher genetic diversity as compared to cluster P2 (P<0.05, Wilcoxon signed-ranks non-parametric test and Bonferroni correction; Table 5).
Moreover, cluster P1 showed a lower FK estimate (0.09) as compared to cluster P2 (FK = 0.18). A significant FST value was found between P1 and P2 clusters (FST = 0.09, P< 0.05). All populations showed similar level of gene diversity (mean Hj = 0.22), with few exceptions (Hj of O population was significantly higher than that of Q, JF, XP and XQ populations; S1 Table).
The NJ tree (Fig 6) revealed the relationships between the 33 populations analysed. No clear distinction was found between them, indeed bootstrap values resulted lower than 50% for almost all the nodes of the tree. However, it is worth noting that accessions from China and Mongolia clustered apart from the other populations (100% of bootstrap value). China (XC) and Mongolia (XD) populations, along with those included in their more closely related clusters (XO, JB, A, XG, R, P, Q, XQ populations) are mostly wild materials, with only populations A (Turkey) and XO (India) being of uncertain biological status. Interestingly, they were assigned to cluster P1 (the only exception is population XQ), they carried the allele ‘A’ at the locus CAP and showed a high variability at the locus cpSSR with the presence of all the four alleles (Fig 6).
Neighbor-joining tree based on nuclear SSR data compared with the percentage of membership for the 33 accessions computed for the two clusters identified and the allelic state of the two chloroplast loci. Only bootstrap values higher than 50% are showed.
The main aim of the present study was to characterize, through both morphological evaluation and genetic diversity analyses, a large sample of worldwide distributed wild and cultivated Poa pratensis encompassing 23 different countries of Europe, Asia, America and Africa. The gained knowledge about their morphology, DNA content, level and structure of genetic diversity and genetic relationships are important both for interpretations of the evolutionary history of P. pratensis, while also providing useful data for breeders.
The analysis of the morphological data showed that the most relevant traits able to distinguish cultivated from wild forms of P. pratensis were spring growth habit and leaf colour (Table 4). This was confirmed both considering these traits independently, as well as considering them together, where both traits resulted correlated with the first two discriminant functions able to explain 52% of total variation. These traits are some of the most important traits that underwent selection during domestication, first, and breeding later . In particular, rust tolerance, prostrate growth habit and dark green colour are among the most important traits selected by breeders for any new turf variety . In fact, the dark green colour and rust tolerance are commonly associated with a healthy turf, while the prostrate growth habit with a better ground cover. Moreover, shorter basal internodes (short stature), higher tiller density, narrow leaf blades, high ground cover, all characteristics related to growth habit, were pursued since domestication [50,51]. Thus, our phenotypic analyses provide evidence that breeding selection for growth habit and leaf colour contributed to the cultivated phenotype. In contrast, traits such as inflorescence pigmentation and flag leaf length were apparently not influenced by anthropic selection pressures, as individuals with extreme values were found almost equally distributed between wild and cultivated forms (Table 4).
In recent years, flow cytometry has become the preferred technique for estimating nuclear DNA content because of its ease of utilization and accuracy [52,53]. The DNA content values found in this study were similar to those reported by other authors. Studying 38 accessions of a USDA core collection of Kentucky bluegrass, Wieners and collaborators  observed values ranging from 4.85 to 15.70 pg. In a different study on twenty-two commercial cultivars of this species, values ranging from 5.39 to 17.69 pg were reported . The continue distribution of DNA content (S1 Fig) of the studied collection suggests the presence of several aneuploids in both wild and cultivated P. pratensis accessions and confirm the high genome variability of this species [10,18]. In fact, as suggested in previous studies DNA content can be efficiently used as a means to determining ploidy level in Kentucky bluegrass .
Genome size estimation in a polyploid species can be very important since molecular markers alone can fail in detecting the diversity caused by gene dosage effect, which may, in turn, affect the phenotypic diversity between accessions. Accordingly, we identified correlations between genome size and morphological traits related to both plant morphology, growth habit and rust tolerance in both wild and cultivated P. pratensis accessions. This supports previous studies which showed that variation of nuclear DNA content is associated with morphological characters in different plant species, for example negative correlations between nuclear DNA content and mean leaf length and width have been reported in Dasypyrum villosum (species of annual grass in the Poaceae family) . The overall flow cytometry data produced in this study revealed the existence of both intra-and inter-population variation in P. pratensis DNA content. As suggested by Mowforth and Grime  intra-population variation in DNA content can arise through opposing forces: i) natural selection acting to reduce DNA amount, ii) the disposition of some genomes to accumulate “selfish” DNA, and iii) the existence of ecological advantages or disadvantages associated with large and small DNA amounts. The same authors postulated that in in P. annua, individuals with large genomes have an advantage under winter and early spring conditions, sufficient to guarantee their survival compared to faster-growing individuals characterized by small genomes. The negative correlation between genome size and spring-regrowth detected in this study may suggests a similar role for the DNA content in the distribution of P. pratensis, with respect of climate and growing season length. On the other hand, the presence of individuals with contrasting genome size in cultivated accessions (Fig 3), subjected to intense breeding selection, as documented in this study may also suggest the natural disposition of Poa genome to accumulate DNA. This tendency is even more clearly evident in wild populations where as many as 10 out of 16 (62.5%) were found to be characterized by at least two ploidy levels. In fact, as early suggested by Bashaw and Funk, the great adaptiveness of this species could be closely related to its variable ploidy and by the presence of asexual reproduction that provides an escape from the typical reproductive problems associated with unbalanced genomes .
Chloroplast genes have been used in numerous phylogenetic studies and have proven to be useful in distinguishing different species belonging to the same genus. Patterson and collaborators used chloroplast markers to dissected the relationships among different species of the complex Poa genus . Data produced in this study on both wild and cultivated P. pratensis accessions, showed low polymorphism level in the trnL-F region. This condition was suggested by Stoneberg Holt and collaborators  studying three P. pratensis varieties. However, our data (S1 File), originated from a world wild collection, allow to speculate a situation of widespread homogeneity in this species.
The low diversity here observed could be both related to the low nucleotide substitution rate in plastid DNA and to the mode of reproduction of the species. In fact, the existence of two different asexual reproduction mechanisms (i.e. apomixis and rhizomes production) could strongly reduce plastid diversity within accessions that can derive from one or very few maternal individuals clonally propagated. It is noteworthy that wild and cultivated accessions were well separated according to the allelic state of the SNP position identified in the chloroplast sequence suggesting a strong phylogenetic value for polymorphisms in this region.
Differently from the chloroplast analysis, the high level of polymorphism observed at the nuclear level confirms the suitability of SSR markers in genetic studies of Kentucky bluegrass. In fact, the number of alleles detected per SSR locus (from 1 to 18) was similar to that reported by Honig and collaborators (7–25) . The high level of genome variability here reported, in terms of somatic DNA content and SSR allelic diversity, is not surprising considering the high variable ploidy level repeatedly documented for the species [10,18], the coexistence of apomictic and sexual reproduction in wild populations [7,9] and the pollen recognition system, which confers a strong ability to hybridize and retain alien genomes , all factors contributing to the extreme complexity of the genome of this species .
Only a modest relationship between genetic divergence and geographical origins was reported in P. pratensis by several authors [22,23]. A general lack of separation between cultivated and wild materials at molecular level was reported by Johnson and collaborators , who evaluated a USDA P. pratensis collection using RAPD markers and found a wide overlap of 17 commercial cultivars with clusters formed by wild accessions. Our results are consistent with these findings; indeed both the NJ and population structure analyses indicated that a clear distinction between wild and domesticated was not found (bootstrap values in NJ are almost all lower than 50% and, for STRUCTURE, wild and domesticated were attributed to both the two clusters, even if domesticated were mostly attributed to P2). The main reason of a similar genetic structure could rely on breeding activities used to develop new cultivars of Kentucky bluegrass starting from ecotypes. Finally, the widespread distribution of some ecotypes can generate confusion on the improvement status and on the geographical origin of some accessions that can be wrongly considered native of a given place.
However, the analyses of both chloroplast and nuclear data highlighted some differences between wild and cultivated materials. In particular, considering the chloroplast cpSSR a strong reduction of variability was detected in cultivated materials, where only two alleles were conserved out of the four detected in wild accessions. Concerning the chloroplast locus trnL-F most of the cultivated accessions showed the ‘C’ allele (82%), while, by contrast a predominance of the ‘A’ allele (71%) was found for the wild accessions. A higher variability was present at nuclear level for wild forms, where the two different clusters were equally represented (average qP1 = 0.51 and qP2 = 0.49), while the domesticated forms were mostly belonging to cluster P2 (average qP2 = 0.79), which showed also a significantly lower genetic diversity compared to the cluster P1. Thus, in spite of specific characteristics of the species that, as mentioned above, could make difficult a distinction between wild and domesticated forms, effects of domestication and breeding, were detected at molecular level by our analysis, with a certain loss of genetic variability in domesticated forms as expected due to founder effects and selection at the target loci.
The analyses at molecular level showed also interesting results related to the origin of the species. In this regard, we first focused our attention on the NJ tree. It showed the relationships between the different populations, with the main result being a clear and well supported subdivision, due to the 100% bootstrap value between the two wild populations from Asia (Mongolia XD and China, XC) and all the others. These two populations were included in cluster P1, the one with a lower FK estimate compared to cluster P2 . FK parameter is similar to FST but it is specific for each population and is expected to be proportional to the divergence from a common ancestral population, thus suggesting a more ancient origin of P1 germplasm. Moreover, Mongolian and Chinese populations carried the ‘A’ allele at the chloroplast locus trnL-F, the most represented allele in wild forms and also among the individuals assigned to cluster P1. All these outcomes seem to indicate China and Mongolia as the most probable geographical areas of origin of the species.
Population XO (India) and JB (USA, NJ) were those most closely related to China and Mongolia in the NJ tree, even if their position was not statistically supported by bootstrap. Although both populations belong to cluster P1 and carry the ‘A’ allele, JB is a cultivated form and XO is of uncertain biological status, hence it is difficult to clearly understand their role and origin.
Within the limits of resolution of the present study, the molecular evidences here presented suggest Mongolia and China as the most probable center of origin of P. pratensis. However, further studies, including a wider sample of P. pratensis accessions from Asia would be needed to obtain a more detailed picture on the topic.
S1 Fig. Nuclear DNA content distribution.
Histograms representation of Nuclear DNA content (pg/2C) distribution of successfully characterized P. pratensis accessions.
S2 Fig. trnL-F chloroplast region sequences alignment.
Portion of the trnL-F sequences alignment showing both 5bp indel and SNP position. In the alignment, P. pratensis individuals names are built by population name and individuals number (i.e. B29 is individual 29 from population B). For reference sequences Gene Bank accession number is reported.
S3 Fig. Inference of the most probable K-value.
a) Average ln likelihood value over 10 runs ± SD for increasing K-values, from 1 to 10. b) ΔK values over 10 runs for increasing K-values, from 2 to 9. c) Average symmetric similarity coefficient (H’) for K from 2 to 10.
The work carried out here is an integral part of L. Raggi Ph.D. investigations. Authors wish to thank COST Action FA0903. Authors wish also to thank Prof. Mario Falcinelli, Dr. Luca Pallottini and Dr. Paolo Pesaresi.
Conceived and designed the experiments: EA TFS FV. Performed the experiments: L. Raggi EA GM L. Russi. Analyzed the data: L. Raggi EB L. Russi. Contributed reagents/materials/analysis tools: EA TFS FV. Wrote the paper: L. Raggi EB L. Russi EA.
- 1. Soreng R. Chloroplast-DNA phylogenetics and biogeography in a reticulating group: study in Poa (Poaceae). Am J Bot 1990; 77: 1383–1400. Available: http://www.jstor.org/stable/10.2307/2444749. Accessed 9 May 2013.
- 2. Welsh S, Atwood N, Goodrich S, Higgins L. A Utah Flora (Great Basin Naturalist Memoirs Number 9). Provo, UT: Brigham Young University. 1987.
- 3. Clausen J. Introgression facilitated by apomixis in polyploid Poas. Euphytica 1961; 10: 87–94. Available: http://link.springer.com/article/10.1007/BF00037208. Accessed 9 May 2013.
- 4. Bashaw E, Funk C. Apomictic grasses. In: Fehr WR, editor. Principles of Cultivar Development. Vol. 2 Crop Species. New York, NY: Macmillan. 1987; pp. 40–82.
- 5. Muntzing A. Further studies on apomixis and sexuality in Poa. Hereditas 1940; 26: 115–188. Available: http://onlinelibrary.wiley.com/doi/10.1111/j.1601-5223.1940.tb03230.x/abstract. Accessed 9 May 2013.
- 6. Löve A. Preparatory studies for breeding Icelandic Poa irrigata. Hereditas 1952; 38: 11–32. Available: http://onlinelibrary.wiley.com/doi/10.1111/j.1601-5223.1952.tb02912.x/abstract. Accessed 14 May 2013.
- 7. Åkerberg E. Apomictic and sexual seed formation in Poa pratensis. Hereditas 1939; 25: 359–370. Available: http://onlinelibrary.wiley.com/doi/10.1111/j.1601-5223.1939.tb02702.x/abstract. Accessed 9 May 2013.
- 8. Kellog E. Variation and species limits in agamosphermous grasses. Syst Bot 1990; 15: 112–123.
- 9. Huff D, Bara J. Determining genetic origins of aberrant progeny from facultative apomictic Kentucky bluegrass using a combination of flow cytometry and silver-stained RAPD markers. Theor Appl Genet 1993; 87: 201–208. Available: http://link.springer.com/10.1007/BF00223765. pmid:24190213
- 10. Wieners RR, Fei S, Johnson RC. Characterization of a USDA Kentucky Bluegrass (Poa pratensis L.) Core Collection for Reproductive Mode and DNA Content by Flow Cytometry. Genet Resour Crop Evol 2006; 53: 1531–1541. Available: http://www.springerlink.com/index/10.1007/s10722-005-7766-0. Accessed 25 September 2012.
- 11. Hartley W. Studies on the origin, evolution and distribution of the Gramineae. IV. The genus Poa L. Aust J Bot 1961; 9: 152–161.
- 12. Huff D. Kentucky bluegrass. In: Casler M, Duncan R, editors. Turfgrass biology, genetics, and breeding. John Wiley & Sons Hoboken, NJ. 2003; pp. 27–38.
- 13. Carrier L, Bort L. The history of Kentucky bluegrass and white clover in the United States. Agron J 1916; 8: 256–266.
- 14. Cronquist A, Holmgren N, Holmgren P, Reveal J. Intermountain Flora: Vascular Plants of the Intermountain West, U.S.A., Volume 6. New York, NY,: Columbia University Press. 1977.
- 15. Duell R. The bluegrasses. In: Heath M, Barnes R, Metcalfe D, editors. Forages—The science of grassland agriculture. Ames, IA: Iowa State University. 1985; pp. 188–197.
- 16. Soreng R. Poa L. in New Mexico, with a key to middle and Southern Rocky Mountain species (Poaceae). West North Am Nat 1985; 45: 395–422. Available: https://ojs.lib.byu.edu/ojs/index.php/wnan/article/viewArticle/2034. Accessed 9 May 2013.
- 17. Murovec J, Kastelec D, Vilhar B. High variability of nuclear DNA content in cultivars and natural populations of Poa pratensis L. in relation to morphological characters. Acta Biol Cracoviensia 2009; 51: 45–52. Available: http://www.ib.uj.edu.pl/abc/pdf/51_2/06_murovec.pdf. Accessed 14 May 2013.
- 18. Eaton T, Curley J, Williamson R, Jung G. Determination of the level of variation in polyploidy among Kentucky bluegrass cultivars by means of flow cytometry. Crop Sci 2004; 44: 2168–2174. Available: https://www.crops.org/publications/cs/abstracts/44/6/2168. Accessed 14 May 2013.
- 19. Arumuganathan K, Tallury S, Fraser M. Nuclear DNA content of thirteen turfgrass species by flow cytometry. Crop Sci 1999; 39: 1518–1521. Available: https://www.soils.org/publications/cs/abstracts/39/5/1518. Accessed 28 March 2013.
- 20. Bennett M, Leitch I. Nuclear DNA amounts in angiosperms. Ann Bot: 1995; 113–176. Available: http://rstb.royalsocietypublishing.org/content/274/933/227.short. Accessed 10 September 2014.
- 21. Curley J, Jung G. RAPD-Based Genetic Relationships in Kentucky Bluegrass. Crop Sci 2004; 44: 1299–1306. Available: https://www.agronomy.org/publications/cs/abstracts/44/4/1299?access=0&view=article. Accessed 22 May 2013.
- 22. Fard JR, Zamani Z, Moghaddam MRF, Kafi M. Evaluation of genetic diversity among some genotypes of Kentucky bluegrass by RAPD molecular markers. Hortic Environ Biotechnol 2012; 53: 298–303. Available: http://link.springer.com/10.1007/s13580-012-0120-5. Accessed 22 May 2013.
- 23. Johnson RC, Johnston WJ, Golob CT, Nelson MC, Soreng RJ. Characterization of the USDA Poa pratensis collection using RAPD markers and agronomic descriptors. Genet Resour Crop Evol 2002; 49: 349–361.
- 24. Goldman JJ. The use of ISSR markers to identify Texas bluegrass interspecific hybrids. Plant Breed 2008; 127: 644–646. Available: http://doi.wiley.com/10.1111/j.1439-0523.2008.01526.x. Accessed 9 May 2013.
- 25. Honig JA, Bonos SA, Meyer WA. Isolation and Characterization of 88 Polymorphic Microsatellite Markers in Kentucky Bluegrass (Poa pratensis L.). Hortscience 2010; 45: 1759–1763.
- 26. Honig J, Averello V, Bonos SA, Meyer WA. Classification of Kentucky Bluegrass (Poa pratensis L.) Cultivars and Accessions Based on Microsatellite (Simple Sequence Repeat) Markers. Hortscience 2012; 47: 1356–1366. Available: http://hortsci.ashspublications.org/content/47/9/1356.short. Accessed 25 October 2012.
- 27. UPOV (Union Internationale Pour La Protection Des Obtentions Vegetales). Guidelines for the conduct of tests for distinctness, homogeneity and stability: Kentucky bluegrass: (apomictic varieties). UPOV web site (http://www.upov.int/about/en/list_publications.html).: UPOV web site (http://www.upov.int/about/en/list_publications.html). Available: UPOV web site (http://www.upov.int/about/en/list_publications.html). 1990.
- 28. Matzk F, Meister A, Schubert I. An efficient screen for reproductive pathways using mature seeds of monocots and dicots. Plant J 2000; 21: 97–108. Available: http://www.ncbi.nlm.nih.gov/pubmed/10652155. pmid:10652155
- 29. Taberlet P, Gielly L, Pautou G, Bouvet J, Biologie L De, et al. Universal primers for amplification of three non-coding regions of chloroplast DNA. Plant Mol Biol 1991; 17: 1105–1109. pmid:1932684
- 30. Tamura K, Dudley J, Nei M, Kumar S. MEGA4: molecular evolutionary genetics analysis (MEGA) software version 4.0. Mol Biol Evol 2007; 24: 1596–1599. Available: http://www.ncbi.nlm.nih.gov/pubmed/17488738. Accessed 14 May 2013. pmid:17488738
- 31. Rozen S, Skaletsky H. Primer3 on the WWW for general users and for biologist programmers. In: Krawetz S, Misener S, editors. Bioinformatics Methods and Protocols: Methods in Molecular Biology. Totowa, NJ: Humana Press. 2000; pp. 365–386.
- 32. Don RH, Cox PT, Wainwright BJ, Baker K, Mattick JS. “Touchdown” PCR to circumvent spurious priming during gene amplification. Nucleic Acids Res 1991; 19: 4008. Available: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3318507&tool=pmcentrez&rendertype=abstract. pmid:1861999
- 33. Pritchard JK, Stephens M, Donnelly P. Inference of population structure using multilocus genotype data. Genetics 2000; 155: 945–959. Available: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1461096&tool=pmcentrez&rendertype=abstract. pmid:10835412
- 34. Falush D, Stephens M, Pritchard JK. Inference of population structure using multilocus genotype data: linked loci and correlated allele frequencies. Genetics 2003; 164: 1567–1587. Available: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1462648&tool=pmcentrez&rendertype=abstract. pmid:12930761
- 35. Lo EYY, Stefanović S, Dickinson TA. Population genetic structure of diploid sexual and polyploid apomictic hawthorns (Crataegus; Rosaceae) in the Pacific Northwest. Mol Ecol 2009; 18: 1145–1160. Available: http://www.ncbi.nlm.nih.gov/pubmed/19243504. Accessed 3 November 2014. pmid:19243504
- 36. Jungmann L, Vigna BBZ, Boldrini KR, Sousa a CB, do Valle CB, et al. Genetic diversity and population structure analysis of the tropical pasture grass Brachiaria humidicola based on microsatellites, cytogenetics, morphological traits, and geographical origin. Genome 2010; 53: 698–709. Available: http://www.ncbi.nlm.nih.gov/pubmed/20924419. Accessed 3 November 2014. pmid:20924419
- 37. Vigna BBZ, Jungmann L, Francisco PM, Zucchi MI, do Valle CB, et al. Genetic Diversity and Population Structure of the Brachiaria brizantha Germplasm. Trop Plant Biol 2011; 4: 157–169. Available: http://link.springer.com/10.1007/s12042-011-9078-1. Accessed 6 November 2014.
- 38. Majeský L, Vašut RJ, Kitner M, Trávníček B. The pattern of genetic variability in apomictic clones of Taraxacum officinale indicates the alternation of asexual and sexual histories of apomicts. PLoS One 2012; 7: e41868. Available: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3411577&tool=pmcentrez&rendertype=abstract. Accessed 27 October 2014. pmid:22870257
- 39. Cidade FW, Vigna BB, de Souza FH, Valls JFM, Dall’Agnol M, et al. Genetic variation in polyploid forage grass: assessing the molecular genetic variability in the Paspalum genus. BMC Genet 2013; 14: 50. Available: http://www.biomedcentral.com/1471-2156/14/50. Accessed 3 November 2014. pmid:23759066
- 40. Falush D, Stephens M, Pritchard JK. Inference of population structure using multilocus genotype data: dominant markers and null alleles. Mol Ecol Notes 2007; 7: 574–578. Available: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1974779&tool=pmcentrez&rendertype=abstract. Accessed 28 February 2013. pmid:18784791
- 41. Evanno G, Regnaut S, Goudet J. Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study. Mol Ecol 2005; 14: 2611–2620. Available: http://www.ncbi.nlm.nih.gov/pubmed/15969739. Accessed 28 February 2013. pmid:15969739
- 42. Jakobsson M, Rosenberg NA. CLUMPP: A cluster matching and permutation program for dealing with label switching and multimodality in analysis of population structure. Bioinformatics 2007; 23: 1801–1806. pmid:17485429
- 43. Nei M. Estimation of average heterozygosity and genetic distance from a small number of individuals. Genetics 1978; 89: 583–590. Available: http://www.genetics.org/content/89/3/583.short. Accessed 14 May 2013. pmid:17248844
- 44. Lynch M, Milligan B. Analysis of population genetic structure with RAPD markers. Mol Ecol 1994; 3: 91–99. Available: http://onlinelibrary.wiley.com/doi/10.1111/j.1365-294X.1994.tb00109.x/abstract. Accessed 14 May 2013. pmid:8019690
- 45. Vekemans X. AFLP-SURV, version 1.0. Distributed by the author. Laboratoire de Génétique et Ecologie Végétale, Université Libre de Bruxelles, Belgium. 2002.
- 46. Zhivotovsky LA. Estimating population structure in diploids with multilocus dominant DNA markers. Mol Ecol 1999; 8: 907–913. Available: http://www.ncbi.nlm.nih.gov/pubmed/10434412. pmid:10434412
- 47. Krauss SL. Accurate gene diversity estimates from amplified fragment length polymorphism (AFLP) markers. Mol Ecol 2000; 9: 1241–1245. Available: http://www.ncbi.nlm.nih.gov/pubmed/10972764. pmid:10972764
- 48. Sokal R, Rohlf F. Biometry: the Principles and Practice of Statistics in Biological Research. 4th ed. New York: Freeman WH and Co. 2012.
- 49. Felsenstein J. Phylogeny Inference Package (PHYLIP). Version 3.5. Distributed by the author. Department of Genetics. University of Washington, Seattle, USA. 1993.
- 50. Casler MD. Perennial grasses for turf, sport and amenity uses: evolution of form, function and fitness for human benefit. J Agric Sci 2006; 144: 189. Available: http://journals.cambridge.org/abstract_S0021859606006137. Accessed 5 November 2014.
- 51. Harlan J. Crops and Man, 2nd Edition. Madison, WI.: American Society of Agronomy. 1992.
- 52. Rayburn AL, Auger JA, Benzinger EA, Hepburn AG. Detection of Intraspecific DNA Content Variation in Zea mays L. by Flow Cytometry. J Exp Bot 1989; 40: 1179–1183. Available: http://jxb.oxfordjournals.org/content/40/11/1179.short. Accessed 14 May 2013.
- 53. Heslop-Harrison J. Flow cytometry and genome analysis. Probe (Lond) 1995; 5: 14–17.
- 54. Caceres ME, De Pace C, Scarascia Mugnozza GT, Kotsonis P, Ceccarelli M, et al. Genome size variations within Dasypyrum villosum: correlations with chromosomal traits, environmental factors and plant phenotypic characteristics and behaviour in reproduction. Theor Appl Genet 1998; 96: 559–567. Available: http://link.springer.com/article/10.1007/s001220050774. Accessed 17 September 2014.
- 55. Mowforth M, Grime J. Intra-population variation in nuclear DNA amount, cell size and growth rate in Poa Annua L. Funct Ecol 1989; 3: 289–295.
- 56. Patterson JT, Larson SR, Johnson PG. Genome relationships in polyploid Poa pratensis and other Poa species inferred from phylogenetic analysis of nuclear and chloroplast DNA sequences. Genome 2005; 48: 76–87. pmid:15729399
- 57. Stoneberg Holt S, Horová L, Bures P, Janecek J, Cernoch V. The trnL-F plastid DNA characters of three Poa pratensis (Kentucky bluegrass) varieties. Plant Soil Environ 2005; 51: 94–99. Available: http://agris.fao.org/agris-search/search/display.do?f=2005/CZ/CZ0506.xml;CZ2005000531. Accessed 14 May 2013.
- 58. Wedin W, Huff D. Cool-Season Forage Grasses. In: Moser L, editor. Cool-Season Forage Grasses. Madison, WI: American Society of Agronomy Monograph Series. 1996.
- 59. Porceddu A, Albertini E, Barcaccia G, Falistocco E, Falcinelli M. Linkage mapping in apomictic and sexual Kentucky bluegrass (Poa pratensis L.) genotypes using a two way pseudo-testcross strategy based on AFLP and SAMPL markers. Theor Appl Genet 2002; 104: 273–280. Available: http://www.ncbi.nlm.nih.gov/pubmed/12582697. pmid:12582697