Background

The genus Paspalum L. is an important member of the Paniceae tribe (Poaceae) and includes between 330 and 400 species, most of which are native to the tropical and subtropical regions of the Americas [1–3]. Paspalum species can be found in diverse habitats, such as subtropical rainforests, savannas, marshes and dunes, but they are more frequently found in the natural grasslands of eastern Bolivia, Paraguay, central and southern Brazil, northern Argentina and Uruguay [3]. The main center of origin and diversity of the genus is considered to be located in the South American tropics and subtropics [1], primarily in central Brazil, where numerous species appear to be associated with savannas and rocky terrain. Brazil harbors the greatest number of Paspalum species [4, 5], with approximately 220 species that exist in nearly all herbaceous plant communities within different ecosystems [6].

Because of this degree of complexity, the division of the Paspalum genus into subgenera, sections or informal groups has been proposed by many authors and has been extensively discussed [1–3, 7–16]. Currently, four subgenera are recognized within Paspalum: Paspalum subg. Anachyris Chase, P. subg. Ceresia (Pers.) Rchb., P. subg. Harpostachys (Trin.) S. Denham and P. subg. Paspalum[3], the last of which consists of approximately 265 species divided into 25 informal groups [1, 14]. However, such rankings have been based on morphological similarities, and the evolutionary and genetic relationships between these groups are not always clear. Several species, mainly within Paspalum subg. Paspalum, are of economic importance for foraging, turf and ornamental purposes [17] in different parts of the world. Dallisgrass (Paspalum dilatatum Poir., Dilatata group) and bahiagrass (P. notatum Flüggé, Notata group) are especially important and are widely used for forage, mainly in the southern United States of America (USA) [18, 19]. P. atratum Swallen (Plicatula group) has been the object of growing interest for use as forage in areas that are subjected to periodic flooding in Florida (USA), northeastern Argentina, Brazil, Thailand, the Philippines and Australia [18]. In addition, Paspalum scrobiculatum L. (‘Kodo millet’, Plicatula group) is cultivated in India as a cereal crop [3, 20], and Paspalum vaginatum Sw. (Disticha group) [21] and P. notatum[19] are widely grown as turf grass.

Because of the widespread interest in several species of this genus, many accessions have been conserved in germplasm banks and distributed throughout various countries for cultivar development and cytogenetic studies. However, the use of some accessions is restricted due to difficulty with accurate taxonomical identification. In Brazil, despite constant evaluations of germplasm banks and taxonomic reviews [22], many species of Paspalum, especially those included in the Plicatula Group, remain unidentified. This deficit presents a problem because the correct identification of germplasms and the quantification of their variability are necessary for the development of conservation and breeding programs.

The complexity of the taxonomical classification of Paspalum germplasms results from the complex evolutionary history of the group. Polyploidy occurs at a high frequency within the genus [23, 24] and has played a crucial role in the evolution of Paspalum. Most species have × = 10 as the basic chromosome number, with ploidy levels that range from diploid to hexadecaploid [25]. Diploid species are not rare within the genus [3, 25–32], but nearly 80% of the investigated species are polyploids, among which 50% are tetraploid, and most of these tetraploids are apomictic [24, 33]. Many Paspalum species comprise sexual diploid and apomictic polyploid cytotypes, and several have been shown to have arisen through natural hybridization [34]. However, interspecific hybridization and allopolyploidy are not always morphologically evident in Paspalum species; therefore, additional methods of taxonomic classification are required [35].

Many taxonomical and species characterization problems arise as a consequence of the great morphological variation present in agamic complexes. P. notatum is a good example of this situation, because it forms an agamic complex and presents wide morphological variation [7, 36, 37]. This species is a perennial rhizomatous turf and forage grass and is recognized as a major constituent of the native grasslands of the New World, being found from Central Eastern Mexico to Argentina and throughout the West Indies [1]. This grass is economically important and is widely used for forage production, mainly in the southern USA [19].

Various morphological and cytological forms are recognized within P. notatum[7, 37]. This species includes several genotypes, which differ in both their ploidy levels and their reproductive systems. The diploid cytotype (2n = 2× = 20) is sexual and self-incompatible [38, 39], whereas the tetraploid cytotype (2n = 4× = 40) is a self-compatible pseudogamous aposporous apomict [40]. Apomixis in tetraploid P. notatum can be either obligate or facultative [24]. In botanical terms, the tetraploid cytotypes are usually considered to be the typical form of P. notatum; as such, they form the variety notatum. On the other hand, the diploid cytotypes are classified as belonging to the saurae variety based on their distinct morphological characteristics [36]. In addition to these two widely recognized varieties, Döll [41] proposed the variety latiflorum, which was accepted by some taxonomists [36, 42, 43] until the mid-1980s. However, P. notatum var. latiflorum is not currently recognized.

Molecular makers are of great value in plant studies and have been used for multiple purposes, including estimating the genetic diversity, determining accession relationships, elucidating evolutionary relationships, aiding in the taxonomic classification of many plants [44–49] and aiding in the identification of botanical varieties [50, 51]. In addition, utilizing molecular markers that regularly identify many genetic polymorphisms at low taxonomic levels makes it possible to address the relationship between morphological and genotypic variation.

The molecular markers that are more informative for the discrimination of closely related genotypes include microsatellite sequence markers [52–54]. Microsatellites are tandem repeat sequences of 1 to 6 nucleotides that are widely distributed in the genome [55]. However, these markers are usually species specific, and in the Paspalum genus, few microsatellite markers are available for P. vaginatum[56], P. dilatatum and the related species [57], P. notatum[58] and P. atratum[59]. However, primers designed for source species have been successfully employed to amplify nuclear SSRs in closely related taxa when the DNA regions that flank the microsatellite loci are sufficiently conserved [57, 59–67].

The main objectives of this study are (1) to evaluate the informative potential of SSRs for genetic discrimination in different species of Paspalum using markers developed for P. atratum and P. notatum; (2) to conduct an in-depth study of the extent, distribution and structure of the genetic variation of P. notatum in a South American collection of this species; and (3) to evaluate the genetic patterns and relationships among different morphological types found in this species and correlate them with their geographic distribution.

Methods

Plant material and DNA extraction

A total of 214 accessions of Paspalum (177 accessions sampled from 35 species and 37 unclassified accessions) were included in this study. These accessions have been maintained in vivo at Embrapa Southeast Livestock, located in São Carlos, São Paulo (SP), Brazil and at the Federal University of Rio Grande do Sul in Porto Alegre, Rio Grande do Sul (RS), Brazil (Additional file 1). Fifty-seven P. notatum accessions from germplasm collections were evaluated in a more detailed analysis in this study (Additional file 2). The genomic formula, chromosome number, type of polyploidy and mode of reproduction are presented for all Paspalum species investigated in this study (Additional file 3).

Genomic DNA from each sample was isolated from lyophilized young leaf tissue using the cetyltrimethyl ammonium bromide (CTAB) method proposed by Doyle and Doyle [68] with minor modifications. DNA concentrations were estimated by comparison with known concentrations of λ DNA on 0.8% agarose gels.

Development of new SSR markers for Paspalumand SSR analysis

A microsatellite-enriched genomic library was obtained for P. notatum using the method described by Billotte et al. [69]. The steps for obtaining the clones, sequencing and analysis, the determination of the criteria for selecting the microsatellites, the primer design and the amplification were conducted according to previously described methodology [58]. For some loci, a touchdown protocol was used, as previously described [70]. Only the strongest bands were considered because of the possibility that lighter bands could be stutter bands that result from the slippage of Taq polymerase during the PCR amplifications [63, 71].

Cross-species amplification was tested for 214 accessions that correspond to 35 Paspalum species, with a set of 23 SSR primer pairs developed for P. atratum and P. notatum[58, 59] and new SSRs developed in this work. An SSR was considered to be transferable when a band of the expected size was amplified via PCR and an SSR pattern was observed. To avoid false negatives, primers that resulted in null alleles in some of the samples were tested at least twice. The microsatellite loci that were amplified successfully and that showed readable electrophoretic patterns and an absence of nonspecific products were used for the genetic evaluation of the germplasm collection.

A more detailed and accurate analysis of the P. notatum accessions was conducted using 15 microsatellite markers developed by Cidade et al. [58, 59] and 15 new SSRs developed in this work.

The total Paspalum germplasm and the P. notatum accessions were genotyped and analyzed in the same manner. No assumptions were made about the genetic nature of the examined alleles because of the high ploidy levels of the majority of the Paspalum accessions. Hence, each SSR allele was treated as dominant in this study due to the high ploidy levels of the samples [72]; these alleles will hereafter be referred to as bands. Each allele was scored as 1 (present) or 0 (absent) and was arranged in a matrix. Non-amplified loci were scored as missing data. Summary statistics, including the total number of bands, the number of bands per locus, and the polymorphism information content (PIC) [72] were determined based on the data matrix that documented the microsatellite genotyping of each locus. To verify the information obtained for these SSR loci for future genetic studies of other Paspalum species, the number of bands and the PIC values were calculated for each marker within each species.

Clustering and population structure analyses

For the analysis of the Paspalum germplasm, genetic similarity (GS) indices were calculated for all the possible pairwise comparisons using Dice’s similarity coefficient [73]. For calculation of the P. notatum GS indices, Jaccard´s similarity index was used [74]. All of the calculations were performed with the software package NTSYS-pc 2.1 [75]. For both analyses, clustering was performed using the Unweighted Pair Group Method with Arithmetic Mean (UPGMA) [76], and the significance of the cophenetic correlation was tested with the Mantel correspondence test [77]. Additionally, a PCO was conducted.

An unrooted dendrogram was constructed for both analyses using the weighted neighbor-joining method (NJ) with DARwin 5.0.157 software [78], and trees were drawn with Figtree v. 1.3.1 software [79]. The bootstrap method was employed to evaluate the reliability of the tree topology. Bootstrap calculations were performed using BOOD 3.0 [80] and DARwin 5.0.157 [78] software based on 1000 replications. The cophenetic coefficient between the matrix of genetic similarity and the dendrogram was computed using NTSYS-pc 2.1 software [75].

The program STRUCTURE, version 2.3.2 [81–83], was used to detect the population structure and to assign individuals to subpopulations. This program employs model-based clustering in which a Bayesian approach identifies groups based on compliance with Hardy–Weinberg equilibrium and linkage equilibrium. In this study, membership in each genotype was tested for the range of genetic groups, from K = 1 to K = 12, with the admixture model and without prior information regarding their origin, using 300,000 replicates for burn-in and 500,000 replicates for Markov Chain Monte Carlo processes for each run. Analyses were conducted with correlated allele frequencies (AFC) for the putative populations. The final subgroups were determined based on the ad hoc measure ΔK [84], which was calculated using an application on the STRUCTURE Harvester website [85].

The binary matrix derived from the SSR data was examined via an analysis of molecular variance (AMOVA) using Arlequin 3.11 software [86]. The partitioning of the genetic variation within Paspalum was calculated, and the significance of each variation was estimated non-parametrically using 1000 permutations. An AMOVA was also performed to evaluate the differentiation between the P. notatum groups obtained from STRUCTURE. Furthermore, the genetic distances among these groups were calculated as the Slatikin distance and pairwise ϕST.

Morphological traits of Paspalum notatum

Data regarding nine morphological traits from [87] and unpublished data from Batista (2008) were used for the morphological analysis of P. notatum. The following vegetative traits were included in this analysis: leaf sheath length (LSL), leaf sheath width (LSW), leaf blade length (LBL), leaf blade width (LBW) and leaf blade hairiness (LBH). In addition, the following four reproductive traits were examined: inflorescence stem length (ISL), average length of inflorescence branches (ALIB), spikelet length (SL) and spikelet width (SW). The plant material was botanically identified according to Parodi [36], Barreto [7] and Canto-Dorow [37].

A data matrix was created that included 46 accessions and nine morphological traits, which comprised a mixed matrix that was composed of one qualitative and eight quantitative traits. The data were standardized by dividing each real value by the standard deviation of the trait. A PCO was performed, and descriptive statistics were calculated using GENES software [88]. Pairwise matrices of Euclidean distance were calculated using NTSYS-pc 2.1 software [75]. Based on the results obtained via principal coordinate analysis (PCA) and PCO, a graph was constructed using STATISTICA software, version 7.0 [89].

To estimate the correlation between the morphological distances and molecular similarities, a Jaccard’s similarity matrix was constructed based on molecular genotyping of the morphologically evaluated accessions. Correlations were calculated using the Mantel test [77] in NTSYS-pc 2.1 software [75].

Geographic origin of Paspalum notatum

The geographic coordinates of the P. notatum accessions were used to construct a matrix of linear distances (in kilometers) between pairs of genotypes using Geographic Distance Matrix Generator software [90]. The correlation between this matrix and the genetic similarity matrix (generated using microsatellites) was tested with NTSYS-pc 2.1 software using the Mantel test. The geographic distribution of the accessions was visualized with the DIVA-GIS program, version 5.2.0.2 [91].

Results

Development of new SSR markers for Paspalumand SSR analysis

Seventeen new SSR polymorphic loci were developed for the investigated Paspalum species (Table 1), among which 15 were isolated from P. notatum and two from P. atratum. A total of 121 bands were scored; the number of bands per locus ranged from 4 to 15, with an average of 7.12 bands per locus (Table 2).

Table 1 New SSR loci developed for Paspalum
Full size table
Table 2 Characteristics of the microsatellite loci from Paspalum atratum and P. notatum that were used in diversity analyses and cross-species amplification tests
Full size table

Of the 23 microsatellite primer pairs tested for transferability to other species, 12 (52%) successfully amplified their respective loci in most species (Additional file 4). Exceptions were found for the PA01-C8 locus, which showed weak amplification in Paspalum ovale Nees ex Steud., and the PN03-A6 locus, which showed weak amplification in Paspalum pumilum Ness, Paspalum subciliatum Chase, and P. ovale. All 12 of the SSRs that transferred to other species revealed a high degree of SSR polymorphism within the surveyed species, which allowed an analysis of the organization of genetic diversity within the Paspalum germplasm to be conducted. The mean PIC was 0.65, and an average of 15.58 bands per locus was found (Table 2). The PA02-B1, PA01-B7, PN03-C7, PN03-A6, and PN02-B5B loci were the most informative, with more than 20 bands being amplified for each of these loci (Additional file 4).

Among the P. notatum accessions, the 30 microsatellite primer pairs evaluated amplified a total of 208 bands, with an average of 6.93 fragments per locus and a mean PIC of 0.64 being obtained (Table 2). The combined electrophoretic profile of all of the loci revealed 44 distinct genotypes, which suggests that there were duplicates or clones in the analyzed material.

Clustering analyses in the Paspalumgenus

All of the analyses performed supported the formation of three major groups in the evaluated Paspalum germplasm (Figures 1, 2 and 3), which corresponded to the Paspalum botanical groups Notata and Plicatula and a group formed by all the other species of Paspalum. Because of the differences between the matrices obtained when analyzing different microsatellite loci, we will address only the major groups and the most robustly supported groups derived from distinct methods of analysis.

Figure 1
figure 1

Principal coordinate analysis based on Dice’s similarity matrix for the 214 accessions from the Paspalum species. Principal coordinate analysis (goodness-of-fit 0.86 (P < 0.001)) based on Dice’s similarity matrix for the 214 accessions from the Paspalum species from the analysis of 187 SSR bands. The three ordination factors together explain 40.21% of the variation in the data matrix. “A” is a bi-dimensional scatter plot, and “B” is a tri-dimensional scatter plot.

Full size image
Figure 2
figure 2

Detailed unrooted neighbor-joining tree based on Dice’s similarity coefficient for the 214 Paspalum accessions. A detailed unrooted neighbor-joining tree based on Dice’s similarity coefficient for the 214 Paspalum accessions on the basis of 187 SSR bands.

Full size image
Figure 3
figure 3

STRUCTURE analysis for the 214 Paspalum accessions. A bar plot obtained from a model-based ancestry analysis of the 214 Paspalum accessions implemented within the STRUCTURE software.

Full size image

The PCO showed a goodness-of-fit value of 0.86 (P < 0.001) and indicated the existence of three main groups (Figure 1), which correspond to the Plicatula group (except for some dispersed species) and the Notata group (P. notatum, P. pumilum, P. subciliatum, Paspalum cromyorhizon Trin. ex Döll and Paspalum ionanthum Chase, with the last two being grouped more distantly from the other species in this group), and to a third group that is composed of other species from different botanical groups of Paspalum (Additional file 1). The three-dimensional PCO scatter plot (Figure 1B) showed the same trend of a tripartite division; however, the species of the Plicatula group were more dispersed. The three principal vectors, PC1 (26.77), PC2 (9.52) and PC3 (3.92), accounted for 40.21% of the variation. The NJ trees based on Dice´s similarity coefficient support the results of the PCO analysis, identifying three main groups – the Notata Group, the Plicatula Group and a third group formed by species from other botanical groups (Figure 2).

The Bayesian analysis of the population structure using a model-based approach provided support for a genetic division of the Paspalum germplasm into three distinct groups (Figure 3), with K = 3 (Additional file 5). Out of the twelve runs for K = 3, the run with the highest likelihood value was selected to assign posterior membership coefficients (Q) to each accession, which supported the results obtained from the other methods of analysis, described previously (Figure 3).

The inter-species pair-wise genetic dissimilarity among the 214 accessions that belong to different Paspalum species varied from 0.00 among some P. notatum accessions to 0.80 among Paspalum conjugatum P.J.Bergius and Paspalum rhodopedum L.B.Sm. & Wassh., with an overall average of 0.43 being obtained.

Population structure in Paspalum notatum

The Bayesian analysis performed with STRUCTURE software revealed two main values of K (Additional file 6). Figure 4 shows sub-figures for two K values up to the number of groups detected in UPGMA because both levels of subdivisions were instructive. K = 2 separates individuals from S. Tome (Argentina) from the accessions from other localities, and the K = 7 groups show correspondence to the botanical varieties of P. notatum. The accessions from groups A, B, C and D belong to the variety notatum, while the accessions from group E and some of the accessions from groups F and G belong to the variety latiflorum. The saurae variety accessions belong to the same genetic group (G). The number of accessions that correspond to each botanical variety and to each of the seven STRUCTURE groups with more than a 50% membership probability is shown in Additional file 7. The M54 accession from the Guaíba region, RS, Brazil, showed a mixed group membership, which corresponded to both groups D (0.49) and G (0.487); however, this accession exhibits morphological traits that are more similar to group G. Group F corresponds to the accessions from S. Tome (Argentina), which separated when K = 2.

Figure 4
figure 4

STRUCTURE analysis of P . notatum . Structure analysis with K = 2 and K = 7 populations for the collection of P. notatum showing the sub-divisions within each variety, which are indicated.

Full size image

Similarity and distance-based methods (PCA, NJ, UPGMA) (Figures 5, 6, 7) revealed accession relationships that were consistent with the STRUCTURE-based membership assignment for most of the accessions (using K = 7, Figure 4), as indicated by the colors in the UPGMA dendrogram (Figure 6). On the other hand, the groups formed in PCA (Figure 7) indicate a correlation between the botanical varieties observed in P. notatum. The coefficients of similarity for all of the P. notatum accessions ranged from 0.17 to 1.0, with an average value of 0.42 being obtained. The cophenetic correlation coefficient in the UPGMA dendrogram was 0.92, which suggests that there were low levels of distortion between the matrix and the phenograms.

Figure 5
figure 5

Principal component analysis based on Jaccard’s similarity for P . notatum . PCA based on Jaccard’s similarity, which resulted from 57 accessions of Paspalum notatum. The branch colors correspond to the colors of the STRUCTURE clusters from Figure 4, K=7.

Full size image
Figure 6
figure 6

Dendrogram based on Jaccard´s similarity coefficient for 57 Paspalum notatum accessions. An UPGMA dendrogram based on Jaccard´s similarity coefficient for 57 Paspalum notatum accessions. The branch colors correspond to the colors in the STRUCTURE clusters from Figure 4, K=7.

Full size image
Figure 7
figure 7

Principal component analysis based on nine morphological traits for P . notatum . A principal component analysis was based on nine morphological traits in 46 Paspalum notatum accessions. The branch colors correspond to the colors in the STRUCTURE clusters from Figure 4, K=7.

Full size image

The AMOVA showed highly significant (P < 0.001) differentiation among groups A, B, C, D, E, F and G, which indicates that approximately 47% of the observed variation can be explained by the subdivision of the accessions into these seven groups (Table 3).

Table 3 AMOVA for the groups of Paspalum notatum accessions detected via STRUCTURE analysis of 30 SSR markers
Full size table

In the NJ (not shown) and UPGMA trees (Figure 6), greater genetic divergence was observed among the saurae variety accessions in comparison to other varieties, and saurae behaved as an outgroup in the dendrogram. This variety is distinguished from the others mainly by its ploidy level (diploid) and reproductive system (sexual self-incompatible), while P. notatum var. notatum and P. notatum var. latiflorum are either obligate or facultative tetraploid apomictics.

Exclusive bands were found between the different varieties (Table 4); however, these bands were not present in all of the accessions of a given variety but were instead differentially distributed between the corresponding gene groups.

Table 4 Exclusive bands revealed by SSR markers in Paspalum notatum varieties
Full size table

Genetic diversity and relationships among the seven groups of Paspalum notatumgroups

Descriptive statistics for the P. notatum groups detected via STRUCTURE analysis (Table 5) revealed that groups E and G exhibited the highest genetic diversity. The overall ϕST among the seven groups was 0.47 (95% confidence interval, p < 0.001), with the ϕST value for each cluster ranging from 0.46 to 0.52. Pairwise comparisons on the basis of the ϕST values can be interpreted as standardized population distances between any two groups. The pairwise ϕST values obtained in this work ranged from 0.181 between groups E and G to 0.803 between groups B and F (Table 6). The genetic distance data agreed with the ϕST estimates. The smallest genetic distance was observed between groups E and G (0.222), whereas groups F and B presented the greatest genetic distance (4.075).

Table 5 Descriptive statistics for the entire group of Paspalum notatum accessions and for the groups detected via STRUCTURE analysis of the 30 SSR markers
Full size table
Table 6 Genetic distances between the Paspalum notatum groups based on STRUCTURE analysis
Full size table

Morphological and molecular results from P. notatum

The PCA results indicated three main groups (Figure 7), which correspond to the three botanical varieties accepted by Parodi [36]. The features that contributed most to the morphological divergence between the groups were ALIB and LBL in principal component 1 (PC1) and SL and SW in principal component 2 (PC2), which together accounted for more than 55% of the detected variation. The group shown in dark green in the upper graph was botanically identified as var. latiflorum due to the size of its spikelets and the length of its inflorescence branches. However, the LBL and LBW were smaller than in other accessions that belong to var. latiflorum, which likely accounts for this difference. The results of the descriptive analysis (Additional files 8 and 9) and correspondence between informal Paspalum botanical groups and the STRUCTURE analysis groups (Additional file 10) are presented.

The correlation between the morphological traits and the SSR markers indicated a weak correlation between the morphological and molecular data (r = 0.42; p < 0.001).

Geographic distance and genetic dissimilarity in Paspalum notatum

The correlation between the geographic distance and the genetic similarity matrix was estimated for all the accessions and revealed no significant association between genetic and geographic variation based on r = 0.0017 for the Mantel test (Additional file 11).

Discussion

Microsatellite markers

The P. notatum and P. atratum marker sets revealed polymorphisms in many of the other Paspalum species analyzed, which suggests that they are useful in genetic analyses of the Paspalum genus.

Some SSRs were highly polymorphic among the evaluated loci, and according to Huang et al. [92], the correlation between the number of bands revealed per locus and the PIC value is equivalent to Nei’s gene diversity [93]. Testolin et al. [94] reported that highly polymorphic markers are promising for use in DNA fingerprinting to identify plant varieties or cultivars. We recommend that a subset of these loci be used in future fingerprinting studies, with a preference for loci that are easy to score and that exhibit high heterozygosity, as indicated by high PIC values.

The genetic structure of the three genetic groups was confirmed using an AMOVA, where 46.83% of the variation was explained by the division into the three groups (Table 7). This finding suggests that the different species could contain unique alleles or genotypes and that these data can discriminate between taxa, especially the more distant taxa. Similar results were obtained by Jungmann et al. [47], who found that 44% of the variation in apomictic Brachiaria humidicola (Rendle) Schweick accessions could be explained by the subdivision of the germplasm into five groups.

Table 7 AMOVA for the Paspalum accessions based on 187 SSR bands
Full size table

The proportion of detected polymorphisms decreased as the genetic distance increased. This finding corroborates other results previously obtained in both plant [61, 66] and animal studies [95]. We observed a reduced number of bands at some loci in species that were more distantly genetically related to the species for which the primers were designed (P. notatum and P. atratum). We suggest that this reduction might have been caused by the presence of null alleles or mutations that could have arisen in the primer-binding sites for the SSRs. Homoplasy has been frequently detected among cross-species-amplified SSR markers [96], and it tends to complicate the interpretation of SSR variations. Non-amplifying or null alleles are not commonly found in plant species, but some examples of these phenomena have been observed [97]. The occurrence of null alleles could cause heterozygotes to be confused with homozygotes, which are usually inferred from significant heterozygote deficits. Undetected null alleles in populations can have dramatic effects on the interpretation of genotype frequency distributions and could lead to mistaken interpretations about the level of inbreeding in a population [98]. Therefore, care should be taken when using the studied loci to analyze species that are distant from the Notata and Plicatula groups, as they could lead to incorrect estimates. However, all of the examined loci were generally useful as an initial approach for the evaluation of Paspalum germplasm, especially for species that were more genetically distant from P. atratum and P. notatum.

We observed the presence of exclusive bands between different varieties of P. notatum, which suggests that the use of these microsatellite markers can benefit taxonomic classification and breeding programs. These bands were not present in all the representatives of the same variety but were shared between accessions from the same gene pool. More detailed morphological and molecular characterizations of these groups should be performed to determine whether these characteristics are related to specific cultivars or biotypes, which we could not detect in our work.

Cluster analysis in the Paspalumgenus

Our data support the usefulness of SSRs as a tool for the identification of genetic diversity and global similarity among members of a species. Therefore, we adopted a phenetic approach, which evaluates the genetic similarities between individuals.

All of the analyses identified three main groups (Figures 1, 2 and 3). This trend toward the formation of three main groups could have resulted from the low representative numbers of some species and groups as well as the nature of the markers used. These markers tended to be more conserved in the more genetically distant species of taxa for which the primer sets were originally developed (P. notatum and P. atratum, from the Notata and the Plicatula groups, respectively).

In this study, the species of the Notata group that form the subgenus Paspalum are represented by five taxa (P. pumilum, P. notatum, P. subciliatum, P. cromyorhizon and P. ionanthum). The former two species were clearly distinguishable at the molecular level, whereas P. pumilum, P. notatum, P. subciliatum were grouped together. These results are in accordance with phylogenetic studies based on internal transcribed sequences (ITS) and chloroplast DNA (cpDNA), which support a close genetic relationship between P. pumilum, P. notatum and P. subciliatum[99–101] (these species form a monophyletic clade). Moreover, whether P. cromyorhizon and P. ionanthum should be included in the subgenus Paspalum remains uncertain, but evidence provided by cpDNA sequences [99] supports a close relationship between these two species. P. pumilum, which is a sexual autogamous diploid species, showed less genetic divergence from P. notatum when compared with other species from the group. P. subciliatum exhibited a relatively intermediate degree of genetic divergence from the other species in the Notata group, whereas P. cromyorhizon and P. ionanthum were classified as being somewhat more distant from the other group members (Figures 1, 2 and 3).

In this study, the Plicatula group showed high genetic variability within its species, but there was no clear distinction between different species (Figures 1, 2 and 3). The lack of a clear delineation between the different taxa of the Plicatula group is most likely caused by the genetic closeness of these species as well as the interspecific hybridization that has occurred throughout their evolution, which favors allele sharing. The complexity of the Plicatula was described in a taxonomic review of this group [22]. Many species of the Plicatula group are morphologically polymorphic, including P. atratum, Paspalum plicatulum Michx., Paspalum compressifolium Swallen, and Paspalum limbatum Henrard. Oliveira [22] suggested that the current circumscription adopted for the separation of some species of the Plicatula group could include different taxa and that the pattern of evolution in the group is reticulated, with no well-defined morphological boundaries occurring between taxa. Although the species of this group have not been clearly delineated, the species of the Plicatula group were consistently grouped together in our analyses, with few exceptions. Polyploidy and hybridization are strategies that have played a role in the evolution of the species of the Plicatula group, as in most species of Paspalum, and should be accounted for when defining a species in this group. The species comprising the Plicatula group were considered to be those with populations that were formed by different co-specific cytotypes with at least a basic polyploid series of 2n = 20 or 40 [22].

Few conclusions could be made regarding the third main group, because it was composed of species from the other Paspalum botanical groups that were not included in the first two groups because of the small number of representative species. Previous molecular phylogenetic studies [99] have suggested a greater genetic distance between the third main group compared to the Plicatula and Notata groups, which explains the lower frequency of microsatellite polymorphisms observed for species from the third group. However, there was a clear distinction between the species of this group. Thus, the markers employed in this study have proven to be effective in distinguishing the different Paspalum species and can assist in the identification of distinct species when combined with other molecular techniques.

The accessions from subgenus Anachyris were grouped and presented a close genetic similarity to the species from the Paniculata group (Paspalum juergensii Hack. and Paspalum paniculatum L.) (Figures 2 and 3). Our results corroborate the monophyly of the subgenus Anachyris[99] and the existence of close genetic relationships between Anachyris and Paspalum falcatum Nees ex Steud. (Falcata), Paspalum paucifolium Swallen (Eriantha), Paspalum humboldtianum Flüggé (Subgenus Ceresia), Paspalum quarinii Morrone and Zuloaga (Quadrifaria), P. juergensii and P. paniculatum (Paniculata).

The Virgata and Dilatata groups represent species complexes that have been extensively studied and have been shown to be entirely composed of allopolyploid species based on cytogenetic evidence [27, 102–112]. Paspalum conspersum Schrad., P. dilatatum and Paspalum pauciciliatum (Parodi) Herter exhibit genomic formulae of IIJ2J2, IIJJX and IIJJ, respectively (Additional file 3). The donor species for genome I is Paspalum intermedium Munro ex Morong (Quadrifaria group), and the donor species for genome J is P. juergensii (Paniculata group). Although the Dilatata group has the genomic formulae IIJJ (Additional file 3), the two species from this group included in this study were not genetically close to the Quadrifaria and Paniculata groups, which include the donor species for genomes I and J, respectively.

The classification of Paspalum into two informal groups is controversial. P. conspersum (Virgata) and Paspalum exaltatum J. Presl (Quadrifaria) clustered together in the UPGMA dendrogram (Figure 6) and were shown to be genetically related to Paspalum regnellii Mez (Virgata). The genetic proximity of these species corroborates previous phylogenetic studies [99, 113]. The Virgata group should be considered monophyletic, and a number of additional species, such as P. exaltatum, should be included in it [113]. Moreover, their data supported the idea that P. regnellii should be excluded from the Virgata group and included in the Macrophylla group.

Paspalum maritimum Trin. (Caespitosa) is considered to be allopolyploid, with a predominant bivalent chromosome pairing [114], and this species clustered close to P. conjugatum and Paspalum corcovadense Raddi. P. conjugatum and P. corcovadense were grouped on the basis of the dendrogram and were shown to be genetically distant from the other species studied, corroborating previous phylogenetic studies [99].

Diversity and population structure in Paspalum notatum

Population structure analysis using STRUCTURE software corroborates the most supported groups found in the UPGMA analysis. Group F is the most consistent group throughout all the analyses, and it corresponds to all the accessions from S. Tome (Argentina) (Figures 4, 5 and 6). The K = 7 analysis confirmed the differences among the genotypes that are dispersed throughout southern South America (Additional file 11). The methodology used to analyze the molecular data in this work has been successfully employed for understanding how genetic diversity is organized in other economically important crops [47, 115–117].

Populations of predominantly apomictic grasses are predicted to exhibit low levels of within-population genetic variation due to founder effects and a lack of recombination in offspring, whereas high levels of differentiation among populations are expected because of the limited gene flow via pollen and the divergence of populations over time caused by selection, drift and the accumulation of new mutations [118]. Our work focused on evaluating the relationships between accessions, which could represent distinct populations. Consequently, we expected to detect a high level of differentiation between individuals due to reproductive and geographic isolation. However, a high level of differentiation between individuals was not observed for all the accessions, such as the accessions from gene pools C and D, which are widely distributed but are genetically close. Furthermore, some accessions showed a large amount of genetic variation, especially those derived from Rio Grande do Sul (Brazil) and Argentina. The studied species is widely distributed in the Americas, and Argentina is considered to be its center of origin [119]. The most widely distributed accessions of P. notatum most likely originated from a few apomictic individuals who were better adapted to different environmental conditions, as reported for Pennisetum setaceum, an apomictic grass that has an invasive capacity and can colonize large areas with only a single clone [120].

Moreover, P. notatum comprises sexual diploid and facultative apomictic tetraploid individuals, and diversity could be especially high in the apomictic species complexes that are derived from multiple, independent hybridization events. Gene flow in this species occurs between tetraploid and diploid cytotypes when they are in sympatry [44]; thus, diploid commercial cultivars can introduce variability into an apomictic complex. Furthermore, facultative apomictic tetraploid species can retain residual sexuality such that their occasional hybridization and genetic recombination introduces variability into natural populations. The better-adapted genotypes are then fixed via apomixis, which maintains the maternal genotype and heterozygosity through successive generations [121].

Morphological and molecular results in P. notatum

The great morphological variability of P. notatum in Brazil has been previously described [7, 37], and the morphological and genetic data reported here confirm these findings. Our genetic analyses show a weak correlation with phenotypic data, possibly as a consequence of the small number of morphological characters that we were able to use in this work, combined with the phenotypic plasticity of the morphological characteristics, which were mainly vegetative. Our results show that there is a tendency toward clustering the P. notatum accessions following the botanical variety classifications; however, our data could not confirm the informal intraspecific categories proposed by other authors because the morphological features employed to make these distinctions are continuous and often overlapping or are otherwise incapable of reflecting the total observed variation.

Using spikelet size as a diagnostic characteristic, Döll [41] described the variety latiflorum, which features spikelets that are orbicular-elliptical in shape and larger than the spikelets of the notatum variety. Paspalum notatum var. latiflorum Döll was previously accepted and cited by some authors [36, 42, 43], but this classification is not currently regarded as valid. Parodi [36] studied the variety latiflorum and proposed a new variety, P. notatum var. saurae, which is accepted and used to this day. To differentiate these varieties, Parodi [36] employed spikelet size and the length and number of inflorescence branches, which are diagnostic characteristics that exhibit little or no phenotypic plasticity. The morphological and molecular evidence obtained in this study supports the division of the species into three varieties because there is a strong tendency for accessions of the same variety to remain grouped into similar gene groups. Out of the seven distinct gene pools identified using Bayesian STRUCTURE software analysis, four represent the botanical variety notatum, and three pools correspond to the latiflorum variety.

Sexual diploid strains grow only in a limited area in Argentina, which is considered to be the center of origin for P. notatum[44, 45]. Some tetraploids that share the diploid gene pool might have maintained the genetic patterns of the Argentine diploid populations or might have arisen via independent self-polyploidization events. The diploid accessions (P. notatum var. saurae) were grouped into a single gene pool, which suggests that they share a common origin. These accessions could have derived from the cultivar Pensacola, which originated in Argentina [44, 119], which is in agreement with the existence of gene flow between tetraploid and diploid cytotypes of P. notatum when they are sympatric [44]. However, a group of tetraploid accessions from western Rio Grande do Sul (Brazil) and Uruguay that were botanically identified as P. notatum var. latiflorum share the same gene pool with P. notatum var. saurae.

According to Quarín [24], autotetraploid species of Paspalum might have appeared in a two-step process. Crossbreeding sexual diploids sometimes develop aposporous embryo sacs. Occasionally, an unreduced oosphere is fertilized by the reduced spermatic nuclei of a diploid, forming a triploid (2n, 2× + n = 3×). The triploids then produce offspring that, when fertilized by diploid pollen, give rise to tetraploid (2n, 3× + n = 4×) genotypes.

Geographic distance and genetic dissimilarity in Paspalum notatum

Our data suggest that there is no correlation between genetic dissimilarity and geographic distance among the P. notatum accessions. This arrangement might occur due to an apomixis and polyploidy-induced buffer that restrains the genetic status of the species and promotes the fixation of genetic fitness over a wide range of environments [45]. Moreover, the incongruence between the genetic and geographic data might also be a consequence of anthropogenic dispersal, because P. notatum is a turf and forage grass that is widely sown commercially and can be accidentally dispersed by cattle or other grazing animals. Although no correlation between genetic and geographical data was found in our analysis, a tendency for accessions from the same region to share the same gene pool was evident (Additional file 11).

Conclusions

In summary, our molecular genetic approach using cross-species amplification was proven to be useful for distinguishing the different taxa of Paspalum that were examined, with the exception of those that belong to the complex Plicatula group. The use of a molecular genetics approach that employs microsatellite markers in an initial assessment of germplasms was shown to be useful in species identification and in evaluating the possibility of successful species hybridization.

The methodological approach applied in this study allowed us to understand the genetic organization of different P. notatum genotypes. Comparing the genetic organization with the morphological characteristics provided evidence that supports the existence of three botanical varieties (P. notatum var. notatum, P. notatum var. saurae and P. notatum var. latiflorum[36]), which are represented by different gene groups, with few exceptions. However, the wide genetic variation within each variety deserves more detailed study. We suggest using the methodology that we have proposed in this work to aid in species classification, especially for plants that exhibit wide morphological variation and are difficult to identify. In addition, the microsatellites developed and used throughout this work can be employed for taxonomic classification, breeding programs and DNA fingerprinting for P. notatum cultivar identification.