Genetic similarity and unweighted pair group method using arithmetic average clustering.

SSR marker allele phenotypes (Supplementary Table 2) were scored for presence/absence and assembled into a binary data matrix. Genetic similarity and clustering methods were performed using the Numerical Taxonomy System, NTSYSpc Version 2.21m (Rohlf, 2011). Pairwise comparisons of the proportion of shared alleles between individual genotypes (plants) were determined by the Jaccard (1908) coefficient, and ordered into a similarity matrix using the SIMQUAL module. Cluster analysis was generated from the similarity matrix by the unweighted pair group method using arithmetic averages (UPGMA) algorithm option in the SAHN module to create a dendrogram. The COPH (cophenetic values as a matrix) and MXCOMP (Mantel test statistic) modules were used to test the goodness of fit between the UPGMA dendrogram and the original similarity matrix (cophenetic correlation) using 999 permutations for the matrix correlations. The data set was resampled 1000 times using the RESAMPLE module with the bootstrap option. Bootstrap values for the dendrogram were calculated using the CONSENS module using the majority rule method with a minimum support value set to 0.500.

Model-based clustering.

Model-based clustering analysis using a Bayesian algorithm was applied to infer the genetic structure and to define the number of clusters (kentucky bluegrass classification types/populations) in the data set using the computer program STRUCTURE 2.3.3 (Falush et al., 2003; Pritchard et al., 2000). In this analysis, genotyped individuals are allocated to a predetermined number of clusters/populations (K), where (K) is chosen in advance and can be varied across different runs. Plants can have membership in several clusters with the membership coefficient of individuals equaling 1.0 across clusters. This method uses a Markov chain Monte Carlo algorithm to estimate the allele frequencies in each of the (K) clusters/populations and, for each individual, the proportion of its genome derived from each cluster/population (q_k). We assumed that all loci were independent and in linkage equilibrium. An admixture ancestry model was used and allele frequencies were correlated with a burnin length of 20,000 iterations followed by 50,000 run iterations at each (K). For other settings, program defaults were used, and no prior population information was assumed to define the clusters (K). (K) values were set from two to 25 with 20 replicate runs at each value of (K). The most parsimonious number of clusters/populations was identified using the maximal value of the average estimated log probability Pr(X|K) output from 20 independent runs at each (K) value (Pritchard et al., 2000). The wild ecotype of P. annua was excluded from this analysis.

Kentucky bluegrass classification type revision.

Based on a combination of known pedigrees, the results of the UPGMA cluster analysis, and the results of the model-based clustering analysis, we assigned all 247 individuals of the current study into 16 revised kentucky bluegrass classification types and one outgroup (P. annua). The 247 kentucky bluegrass cultivars, experimental selections, and collections are listed in Table 1 according to this newly revised/updated classification scheme. This newly revised classification scheme is, therefore, a representation of genetic relatedness based on the current SSR marker data and known pedigrees and differs from the previous PTM kentucky bluegrass classification system (Supplementary Table 3) that was based on pedigree, turf performance data, and morphological characters.

Analysis of molecular variation.

The revised/updated classification types, exactly as described in Table 1, were treated as populations in an analysis of molecular variation (AMOVA), performed in GenAlEx 6 (Peakall and Smouse, 2006), to examine the distribution of variation among and within populations (classification types) and to assess the interpopulation pairwise genetic distance (Φ_ST). Statistical significance was tested by random permutation with the number of permutations set to 999. Kentucky bluegrass entries classified as “other” type (Table 1) and the wild ecotype of P. annua were excluded from this analysis.

Simple sequence repeat markers.

The 25 SSR markers used in the current study produced 401 allele phenotypes in the 247 entries (Supplementary Table 2). The number of allele phenotypes for individual SSR markers ranged from seven to 25 with an average of 16.04 allele phenotypes per SSR marker. Additional details about individual SSR markers can be found in Honig et al. (2010).

Unweighted pair group method using arithmetic average clustering.

The results of the UPGMA clustering analysis are presented in Fig. 1, Panel a. The Mantel test for the goodness of fit between the UPGMA dendrogram and the original similarity matrix (cophenetic correlation) was r = 0.91, where an r value above 0.90 indicates a very good fit (Rohlf, 2011, user documentation). Bootstrap values for the delineation of the major kentucky bluegrass classification types are shown, whereas additional bootstrap values are excluded for clarity of the figure (Fig. 1, Panel a). Cultivars and accessions in the UPGMA diagram are color-coded according to a combination of known pedigrees and the newly revised SSR classification type assignment.

View Full Size

View Full Size

View Full Size

View Full Size

Unweighted pair group method using arithmetic average (UPGMA) and model-based clustering analysis of 247 kentucky bluegrass cultivars, experimental selections, collections, and one outgroup entry using 25 microsatellite [simple sequence repeat (SSR)] markers. (Panel a) Diagram is a UPGMA dendrogram based on the average Jaccard similarity of 401 SSR alleles between individuals, color-coded according to known pedigrees. Black color indicates unknown pedigree. Bootstrap values (based on 1000 replicate runs) for major groupings are shown in parenthesis. (Panel b) Estimated population (kentucky bluegrass classification type) structure output from STRUCTURE 2.3.3 for kentucky bluegrass cultivars, experimental selections, and collections using 25 microsatellite (SSR) markers. Each individual is represented by a horizontal colored line, which can be partitioned into (K) segments that represent the individual’s estimated membership fractions in (K) clusters. Color coding is based on the output from STRUCTURE 2.3.3. Groupings of cultivars from this model-based clustering analysis most closely resembled pedigree relationships and the groupings delineated in the UPGMA dendrogram at (K) = 14.

Citation: HortScience horts 47, 9; 10.21273/HORTSCI.47.9.1356

• Download Figure
• Download Figure
• Download Figure
• Download Figure
• Download figure as PowerPoint slide

The UPGMA clustering analysis grouped the kentucky bluegrass entries into several distinct classification types. Bootstrap support exists for the delineation of the traditional kentucky bluegrass types Midnight (0.971), BVMG (1.000), Shamrock (1.000), Julia (0.901), Compact-America (0.799), and Mid-Atlantic types (0.805) (Fig. 1, Panel a). The cultivars Bodacious, Boomerang, and Cheetah as well as the experimental selections SRX27921, PpH7907, and PpH7929 were part of a supported group in (Fig. 1A, Panel a). The pedigrees of these entries can all trace, in part, back to the cultivar Cynthia (Supplementary Table 4). Bootstrap support (0.978; Fig. 1A, Panel a) and unique morphological characteristics (e.g., lower seed yield compared with BVMG type) provide justification for a new Cynthia classification type. The cultivars Allure, Fairfax, and Serene as well as the experimental selections B545 and B543 were part of a bootstrap-supported (1.000) group at the bottom of Fig. 1A, Panel a, and the top of Fig. 1B, Panel a. The pedigrees of these entries can all trace, in part, back to the cultivar Sydsport (Supplementary Table 4) providing justification for a new Sydsport classification type. The majority of European as well as a few Asian plant collections formed three supported clusters in the UPGMA analysis (Fig. 1B–1C, Panel a), providing justification for newly defined classification types referred to as the Eurasian types. Additional new types with bootstrap support included Limousine (0.997) and P-105 (0.992) (Fig. 1B, Panel a and supporting breeding history information in Supplementary Table 4). Cultivars that had previously been referred to as Compact-type cultivars were split into two supported clusters: one cluster containing the cultivars Alpine, Blue Sapphire, Blackstone (Rose-Fricker et al., 2002), and Hampton (Fig. 1B, Panel a) and the other comprised of a number of cultivars and experimental selections closely related to the Compact-America-type cultivars (Fig. 1D, Panel a). A final potential supported new classification type was the Jefferson/Washington type (1.000) (Fig. 1B, Panel a); however, this current classification type would only be comprised of two cultivars in the current data set. Fourteen entries that have low bootstrap support, and consequently did not strongly cluster with defined classification groups, were classified as “other” type (Table 1).

The three clusters of Eurasian collections and cultivars were basal, being most closely related to the outgroup, P. annua (Fig. 1, Panel a). The remainder of the groups and cultivars, which were predominantly developed from U.S. breeding programs, appeared to be derived from the more basal Eurasian cultivars. These derived classification types appeared to form two distinct larger groupings: one group comprised of the Midnight, BVMG, Cynthia, Sydsport, Shamrock, Limousine, and P-105 classification types (as well as the potential Jefferson/Washington classification type) (Fig. 1A–B, Panel a) and the other large group comprised of the Mid-Atlantic, Compact, and Compact-America classification types (Fig. 1C–D, Panel a).

The majority of cultivars and accessions were uniquely identified with the current panel of 25 kentucky bluegrass SSR markers. Exceptions included three clusters of cultivars within the Midnight classification type (Fig. 1A, Panel a). The first of these clusters consisted of the cultivars NuGlade (Brede, 2001a), Courtyard (Brede, 2011), Awesome (Brede, 2011), and Award (Brede, 2001b); the second consisted of the cultivars Tsunami (Brede, 2004a), Alexa (Brede, 2006a), Everest (Brede, 2006b), Freedom III (Brede, 2006c), and Beyond (Brede, 2004b); and the third consisted of the cultivars Excursion (Brede, 2004c) and Barrister (Brede, 2006d) (Fig. 1A, Panel a).

Model-based clustering.

The results of the model-based clustering (STRUCTURE) analysis are presented in Fig. 1, Panel b. Color coding is separated based on the groupings produced from the STRUCTURE output. The maximal value for the first plateau of the average estimated log probability Pr(X|K) [used to identify the most parsimonious number of clusters/populations (K)] from 20 independent runs at each (K) occurred at (K) = 14 (Supplementary Fig. 1). This (K) value very closely matched known pedigree relationships (Supplementary Table 4) and the major kentucky bluegrass classification groupings delineated in the UPGMA dendrogram (Fig. 1, Panel a). Other values of (K), particularly those at (K) = 11, 12, 13, 15, 16, and 17, were also considered; however, the biological interpretation (congruence with pedigree information and the UPGMA analysis) of the current data were most appropriate at (K) = 14 (Pritchard et al., 2000). At (K) = 14, clusters identified by STRUCTURE were very closely matched to pedigree (Supplementary Table 4) and clusters in the UPGMA dendrogram (Fig. 1, Panel a) for the Midnight, BVMG, Cynthia, Sydsport, Shamrock, Limousine, P-105, Julia, Compact-America, and Mid-Atlantic classification types. The STRUCTURE analysis split the Eurasian classification type into three distinct clusters (which were also supported by the UPGMA clustering analysis) (Fig. 1B–C). A number of cultivars and accessions that had been traditionally classified as Compact cultivars were found to have highly diverse membership in multiple (K) clusters. This included the cultivars Alpine, Blue Sapphire, Blackstone, and Hampton (Fig. 1B, Panel b) as well as the cultivars Blacksburg II, Rita, and Ascot and the experimental selections PST-H5-35 and NA-K992 (Fig. 1D, Panel b). The STRUCTURE analysis at (K) = 14 did not uniquely identify the Jefferson/Washington classification type, as identified in the UPGMA clustering analysis (bootstrap value 1.000; Fig. 1B). Entries defined as “other” type (Table 1) exhibited a combination of both high population level admixture in the STRUCTURE analysis and low bootstrap support in the UPGMA analysis (Fig. 1).

In some instances, the STRUCTURE analysis (Fig. 1, Panel b) provides additional detail not captured in the UPGMA analysis (Fig. 1, Panel a), specifically in relation to interbreeding or admixture between classification types. This was evident in some of the recognizable “hybrid” cultivars or selections exhibiting population-level admixture, where the admixture pattern matched known pedigree records (Supplementary Table 4). For example, ‘Chicago II’ (Brede, 2004d) (Fig. 1A, Panel b) was a cross between ‘Midnight’ (Meyer et al., 1984) × ‘Limousine’ (Supplementary Table 4); the STRUCTURE analysis showed admixture between these two classification types for this cultivar (Fig. 1A, Panel b). Similar results, where STRUCTURE admixture accurately reflected known pedigrees (Supplementary Table 4), were observed for Bd0384, A05-313, SRX26351, ‘Freedom II’ (Brede, 2004e), Alpine (Fig. 1B, Panel b); A98-183 (Fig. 1C, Panel b); A00-1400, ‘Royale’, ‘Diva’, ‘Sonoma’ (Ford et al., 2004), A03-132, ‘Durham’, ‘Shiraz’, A03-66, Bd992103, A99-2427, PST-604, A00-1254, and A03-37 (Fig. 1D, Panel b).

In some instances, lower admixture vs. higher admixture in individual entries in the STRUCTURE analysis (Fig. 1, Panel b) helped to visually clarify relationships between major classification types and sister entries or sister clades in the UPGMA analysis (Fig. 1, Panel a). For example, the Midnight-type cultivars exhibited zero to relatively low levels of admixture with other populations from Midnight through Chicago II (Fig. 1A, Panel b). This grouping of entries had bootstrap support of 1.000 in the UPGMA analysis (Fig. 1A, Panel a). The cultivar Skye, defined here as a Midnight-type cultivar, could alternatively be considered sister to the other Midnight-type entries because of lower UPGMA similarity and lower bootstrap support when this cultivar was included in the Midnight classification type (Fig. 1A, Panel a). Skye exhibited a relatively high level of admixture in the STRUCTURE analysis (Fig. 1A, Panel b), which was consistent with the relationship of Skye to the remainder of the Midnight-type entries in the UPGMA analysis (Fig. 1A, Panel a). Other examples of STRUCTURE admixture accurately reflecting sister grouping relationships in the UPGMA analysis included the core Shamrock type entries vs. A01-701, Bd0384, A05-313, and A97-857 (Fig. 1B) and a core set of Compact-type entries that includes ‘Wildwood’, ‘Hallmark’, PST-B4246, PST-YorkHarbor4, SRX2114, and PST-B5125 vs. Blacksburg II, ‘Baronette’, ‘Rita’, PST-H535, NA-K992, and Ascot (Fig. 1D).

Kentucky bluegrass classification type revision.

The combination of pedigree information, UPGMA cluster analysis, and model-based clustering analysis all supported the proposed revised kentucky bluegrass classification system outlined in Table 1. With just a few exceptions, kentucky bluegrass pedigree (Supplementary Table 4), UPGMA clustering analysis (Fig. 1, Panel a), and model-based clustering analysis at (K) = 14 (Fig. 1, Panel b) produced similar grouping results for the Midnight, BVMG, Cynthia, Sydsport, Shamrock, Limousine, P-105, Julia, Compact-America, Compact, and Mid-Atlantic classification types. Three clusters composed of predominantly Eurasian cultivars and collections were supported in the model-based clustering and UPGMA clustering. The potential Jefferson/Washington classification type had strong bootstrap support in the UPGMA analysis (1.000) (Fig. 1B) but had no support in the model-based clustering analysis. Taken together with pedigree, the combination of the results of these analyses indicated that the revised 14 supported classification types (not including the potential Jefferson/Washington type and the catchall Other type) described in Table 1 are a reasonable representation of the genetic relationships among kentucky bluegrass classification types, cultivars, experimental selections, and collections.

Analysis of molecular variance.

The AMOVA was conducted treating the newly revised classification types (as listed in Table 1) as populations. The “other” type entries (Table 1) were not included in the AMOVA, because this classification group is not a genetically related population. The AMOVA of the remaining 15 revised classification types (Table 1) was, in part, meant to serve as a validation step for the newly revised classification scheme. AMOVA results showed that the majority of the SSR marker variation (52%) observed among the kentucky bluegrass individuals from the 15 populations was accounted for by within population variance, although a significant portion (48%) was attributable to differences between populations. Within-population variance was spread relatively uniformly among the various classification types (data not shown); however, the Midnight type had the lowest within population variance (data not shown).

Pairwise Φ_ST values derived from the AMOVA highlighted a large number of differences between classification types when individual pairs of classification types were compared (Table 2). Only one pairwise Φ_ST comparison, between the Jefferson/Washington type and the P-105 type, was not significant (P > 0.05), whereas all other pairwise Φ_ST comparisons were significant at P < 0.05 (Table 2). This indicates that the vast majority of classification types were differentiated from other classification types, the Φ_ST distance values are not random, and the AMOVA supported the revised classification scheme presented in Table 1.

Table 2.

Pairwise Φ_ST values calculated by analysis of molecular variation (AMOVA) illustrating differences between populations (kentucky bluegrass classification types) of kentucky bluegrass (Φ_ST values are given below the diagonal and the associated P values are given above the diagonal).^z

View Table

Validation and revision of kentucky bluegrass classification types.

The UPGMA depiction (Fig. 1, Panel a) of the genetic relatedness of kentucky bluegrass cultivars, experimental selections, and collections, based on SSR marker data, represents a cogent argument for the grouping of kentucky bluegrass entries into several distinct populations/classification types. With only a few exceptions, model-based clustering (Fig. 1, Panel b) corroborated the classification type grouping scenario depicted in the UPGMA analysis. Both of these analyses were in agreement with known pedigree information (Supplementary Table 4) and, therefore, supported a revision of the previous PTM kentucky bluegrass classification system, resulting in a newly proposed classification scheme outlined in Table 1. The results of the AMOVA (Table 2) validated the revised classification system, indicating that the supported proposed types were composed of distinct populations or genetically related classification groups.

A significant improvement over the PTM kentucky bluegrass classification system was the dramatic reduction in the number of entries in the current study from the “other” type (Supplementary Table 3 vs. Table 1, respectively). In the current study, the combination of low bootstrap support in the UPGMA analysis and high population admixture in the STRUCTURE analysis was used to define entries belonging to the “other” type (Table 1). Part of the original definition of the “other” type was that this group possessed traits that are intermediate between two or more of the defined classification types (Murphy et al., 1995); thus, high population-level admixture, in combination with poor UPGMA clustering resolution, can be used as a means to define “other” type cultivars/accessions. This significantly reduced the number of entries that were considered “other” type, because the vast majority of entries clustered in defined classification groups. It is interesting to note that the STRUCTURE analysis showed some weak associations between the “other” type entries and defined classification groups. For example, common or shared STRUCTURE admixture components likely explained why H02-603, A04-1427, ‘Misty’, and 99AN53 grouped with the BVMG-type entries (Fig. 1A); HV140, A04-1268, A98-3320, and A98-3367 grouped with the Cynthia-type entries (Fig. 1A); ‘NorthStar’ grouped with the Limousine-type entries (Fig. 1B); and A94MH94 grouped with the Julia-type entries (Fig. 1B). However, all of the “other” type entries that have these associations with defined classification types are best described as trends as a result of low bootstrap support in the UPGMA analysis (Fig. 1, Panel a).

The UPGMA cluster analysis (Fig. 1, Panel a) and model-based clustering (Fig. 1, Panel b) provide support for the traditional classification types of Midnight, BVMG, Shamrock, Julia, Compact-America, and Mid-Atlantic. New classification types with UPGMA and model-based clustering support include Cynthia, Sydsport, Limousine, and P-105 (Fig. 1). These analyses were in agreement with pedigree information (Supplementary Table 4).

The new Eurasian classification types were split into three supported clusters in the STRUCTURE analysis (Fig. 1B–C, Panel b), which were also supported in the UPGMA analysis (Fig. 1B–C, Panel a). The split Eurasian groups indicate that there was significant genetic diversity in the Eurasian collections and that continuing to collect Eurasian plant material could lead to novel germplasm sources and new kentucky bluegrass classification types. Although the genetic evidence clearly separated these new Eurasian types, it is difficult to assign specific attributes to these groups as a result of heretofore limited information about the performance and morphological characteristics of these collections. That being stated, one interesting observation was the presence of a higher percentage of entries that were older direct ecotypic selections from the United States or could be traced back to older direct ecotypic selections from the United States (Supplementary Table 4) in the third Eurasian cluster (Fig. 1C) vs. the first and second Eurasian clusters (Fig. 1B–C). For example, the cultivar Kenblue (Fig. 1C) was previously classified as a Common (or Mid-Western) ecotype. The original description of the Common type (Bara et al., 1993) stated that cultivars in this group were commonly selected from naturalized ecotypes surviving within old pastures. Additional entries in the third Eurasian cluster that fit this description were H94-305, A98-183, ‘Eagleton’ (Hurley et al., 1997), and ‘Wellington’ (Brentano, 2004) (Fig. 1C; Supplementary Table 4). A plausible explanation for these entries grouping together, within the third Eurasian type, is that these cultivars, although clearly ecotypic selections made in the United States, were all descendants from remnant populations of pasture grasses originally seeded by early European settlers. The third Eurasian cluster could, therefore, represent the genetic cluster most closely related to some of the earliest Eurasian germplasm introduced into the United States. SSR marker genotyping of additional Common-type cultivars and other direct ecotypic selections from the United States will be needed to support this hypothesis.

The Compact-type was split among multiple clades in the UPGMA analysis (Figs. 1B and 1D, Panel a) and multiple (K) groupings in the model based clustering analysis (Figs. 1B and 1D, Panel b). This was interesting in that the current SSR marker data indicated that the Compact type, as previously defined in the PTM classification scheme, may not represent a distinct genetic group. The original description of Compact-type cultivars included cultivars that exhibited a “low compact growth habit, forming a highly attractive turf” (Bara et al., 1993). Although pedigree information was considered when assigning cultivars to this grouping, a number of genetically divergent cultivars may have been included in the Compact group based simply on a common low, compact morphological growth habit. The current data set indicated that the cultivars Alpine, Blue Sapphire, Blackstone, and Hampton (Fig. 1B) have a different genetic background than the remainder of the Compact-type entries (Fig. 1D). These four cultivars exhibited relatively high population-level admixture with the third Eurasian classification type (Fig. 1B, Panel b). These results may follow pedigree information for two of the cultivars, in which Alpine was a cross between Warren’s A-25 × a collection from Iceland, and Blue Sapphire was a cross between a derivative of Wildwood × ‘Baron’ (Hurley and Ghilsen, 1980) (Baron is a European cultivar) (Supplementary Table 4). In the future, admixture with the Eurasian 3 type could be a basis for creating a new classification type for these cultivars and any new hybrids created from this group.

The cultivars Jefferson (Bonos et al., 2003) and Washington form a distinct group in the UPGMA analysis (bootstrap support of 1.000 for this two cultivar clade, but much weaker bootstrap support when grouped with the P-105 type cultivars) (Fig. 1B, Panel a); however, the STRUCTURE analysis did not uniquely identify this group, and the AMOVA pairwise Φ_ST comparisons indicate that although Jefferson/Washington were significantly different from most other groups, they were not significantly different from the P-105 classification type (Table 2). These discrepancies indicate that Jefferson/Washington should be referred to as a potential or incipient classification type. This terminology may be justified based on the fact that the morphological and performance characteristics of Jefferson and Washington are very different from P-105 type cultivars and also generally very different from all other kentucky bluegrass classification types (authors’ personal observations). An additional line of evidence that provides some support for a potential Jefferson/Washington classification type included a higher level of population-level admixture in Jefferson and Washington vs. the P-105 type entries (Fig. 1B, Panel b). Finally, recent breeding efforts, using Jefferson and Washington as parents, have produced new experimental selections (not included in the current study) that might later be grouped in this classification type (authors’ personal observations). SSR marker genotyping of these additional selections will be required to determine if additional support can be found for this potential classification type.

Genetic relationships among classification types.

Overall, the genetic relationships among kentucky bluegrass classification types (populations) were more difficult to define than the delineations of the individual classification types/populations. This was evident in the comparatively lower bootstrap values of groupings larger than the main kentucky bluegrass classification types in the UPGMA diagram (Fig. 1, Panel a). Additional SSR marker genotyping could improve resolution; however, lower bootstrap support for relationships among classification types could also be the result of reticulate relationships resulting from repeated hybridization between classification types in kentucky bluegrass breeding programs. That being stated, there were still a number of genetic relationships among kentucky bluegrass classification types that were supported by a combination of UPGMA clustering analysis (Fig. 1, Panel a), model-based clustering analysis (STRUCTURE) (Fig. 1, Panel b), AMOVA results (Table 2), and pedigree relationships (Supplementary Table 4).

The main grouping of the three Eurasian types (Fig. 1B–C, Panel a), being most closely related to the outgroup, P. annua, was basal to the remainder of the kentucky bluegrass cultivars, experimental selections, and collections in the current study. The UPGMA placement of the Eurasian collections as basal to the remainder of the entries was supported by the AMOVA results, where the pairwise Φ_ST genetic distance values were always lowest between the Eurasian 2/Eurasian 3 groups and all other classification types (Table 2). These results suggest that Eurasia was the likely center of origin for the germplasm in the current study. This grouping scenario fits with observations that kentucky bluegrass is native to the old world, being distributed naturally throughout the temperate and cooler regions of Europe and Asia (Bashaw and Funk, 1987) and with the proposed center of origin of the genus Poa, which based on morphological, cytological, and species diversity is considered to be Eurasia (Huff, 2003).

The Julia classification type was closely aligned with the main grouping of Eurasian entries (Fig. 1B, Panel a). There was bootstrap support (0.848) for the grouping of Julia-type cultivars and accessions with the first Eurasian grouping (Fig. 1B, Panel a). This grouping scenario follows pedigree information, because ‘Julia’ originated as a collection from northern Germany (Alderson and Sharp, 1994).

A large grouping with bootstrap support includes the Midnight, BVMG, Cynthia, Sydsport, and Shamrock classification types (Fig. 1A–B, Panel a). This large lineage was composed of a mixture of cultivars developed in U.S. breeding programs, cultivars of European origin, and a limited number of individual Eurasian plant collections (now classified as “other” type) weakly aligned with the BVMG and Cynthia classification types. The individual Eurasian collections aligned with the BVMG and Cynthia classification types followed pedigree information, because Cynthia originated in England (Alderson and Sharp, 1994) and Baron (BVMG type) originated in Holland (Hurley and Ghilsen, 1980). The cultivar Sydsport, which originated in Sweden (Alderson and Sharp, 1994), was one of the earlier improved kentucky bluegrass cultivars used in U.S. breeding programs, likely accounting for the placement of the new Sydsport classification type in this section of the UPGMA diagram, which includes both U.S. and European germplasm (Fig. 1A–B, Panel a). The cultivar Shamrock (Baily et al., 1995) (type name for the Shamrock type) was a single plant progeny from A80-336, pollinated in a polycross that included the cultivar Sydsport, which likely explains the proximity of the Shamrock and Sydsport classification types in the UPGMA clustering analysis (Fig. 1B, Panel a).

The cultivar Midnight (type name for the Midnight type) originated from the progeny of F64-603, a selection made from an old lawn in Washington, DC, crossed with the cultivar Glade (Alderson and Sharp, 1994; Jacklin et al., 1977). There were no clear pedigree relationships that explained the placement of the Midnight classification type in this large lineage, because both F64-603 and Glade were direct ecotypic selections (Alderson and Sharp, 1994; Meyer et al., 1984); however, it is very interesting that the Midnight type was placed within this lineage, as opposed to the lineage that includes the Compact and Compact-America types (Fig. 1A vs. Fig. 1D). Earlier PTM kentucky bluegrass classification systems implied that the Midnight type was related to the Compact and Compact-America classification types based on the inclusion of the name “Compact” associated with the Midnight type name (see Supplementary Table 1). The current SSR marker data dispute this based on the location of the Midnight type in the UPGMA analysis (Fig. 1A, Panel a) and comparatively large pairwise Φ_ST genetic distance values calculated by AMOVA (Table 2) between the Midnight type and all other classification types in the current study. Taken together, these findings suggest that the “Compact-Midnight” type was one of the most distinct classification groups relative to all other revised classification groups, and that from this point forward, this group should be named the Midnight type, because the “Compact” designation for this specific type seems to refer only to growth habit and not genetic relationships among groups.

Another large lineage depicted in the UPGMA diagram was comprised of the Mid-Atlantic, Compact, and Compact-America classification types (Fig. 1C–D, Panel a). There was bootstrap support for the association between the Compact and Compact-America classification types but lower bootstrap support when including the Mid-Atlantic type with the latter classification types (Fig. 1D, Panel a). Although bootstrap support in the UPGMA analysis was lower for all three classification types together, it is interesting to note that this large lineage was almost exclusively comprised of cultivars and accessions from U.S. breeding programs with the possible exceptions of ‘Baronie’ and Baronette (unknown pedigrees). This situation was different from the remaining large lineages, which were a mixture of U.S. and Eurasian, or only Eurasian, germplasm.

An association between the Compact and Compact-America types has existed in the kentucky bluegrass classification system since 1995 (Murphy et al., 1995). An early description of the Compact-America type was written as “within the compact type, a number of cultivars exhibit growth and performance characteristics similar to the cultivar ‘America’” (Funk et al., 1982) (Murphy et al., 1995). At the time, this description was intended to imply that the Compact-America type was a newly defined subtype within the Compact type. As numerous new cultivars were developed within the Compact-America subtype, this new group came to be considered a separate classification grouping. The UPGMA cluster analysis indicated a genetic relationship between the Compact and Compact-America classification types (0.909 bootstrap support for the combined groups) (Fig. 1D, Panel a). This was different from the previously described situation between the Compact and Midnight classification types in that the assumed relationship between the Compact and Compact-America classification types in previous PTM kentucky bluegrass classification systems appears to have a genetic basis. The delineation between the Compact and Compact-America classification types was also interesting. There was a core set of Compact-type entries in Wildwood through PST-B5125 (1.000 bootstrap support) (Fig. 1D, Panel a). Blacksburg II through ‘Moonshadow’ (Bonos et al., 2005) exhibited various levels of admixture between the two classification types (Fig. 1D, Panel b), whereas the remainder of the Compact-America types, from ‘Bedazzled’ (Bonos et al., 2008) through ‘Glenmont’, exhibited very low to zero levels of admixture between the two classification types (Fig. 1C–D, Panel b).

The remaining large lineage in the UPGMA diagram included the Limousine, Jefferson/Washington, and P-105 classification types as well as the Compact-type cultivars Alpine, Blue Sapphire, Blackstone, and Hampton (Fig. 1B, Panel a). Relative to some of the other larger lineages, this grouping was not supported in the UPGMA analysis (approaching minimum bootstrap support of 0.500) (Fig. 1B, Panel a). Although better described as a trend than a strong relationship, it is possible that these entries were grouping in this location as a result of an association with the Eurasian and Julia classification types. The cultivar Limousine was originally a European cultivar (collection from Germany) (Alderson and Sharp, 1994), whereas the entries NorthStar through ‘P-105’ (Hurley et al., 2000) exhibit various levels of admixture with the third Eurasian classification type (Fig. 1B, Panel b).

Kentucky bluegrass classification types not included in the current classification revision.

There were four previous classification types that were not represented in the current classification revision based on SSR markers (Supplementary Table 3 vs. Table 1, respectively): CELA (‘Challenger’, ‘Eclipse’, ‘Liberty’, ‘Adelphi’) type, Common type, High Density (Aggressive) type, and texas bluegrass × kentucky bluegrass hybrids. As discussed previously, a number of the Common-type cultivars/accessions in the current study clustered with the third Eurasian type. SSR marker genotyping of additional Common-type cultivars will be needed to determine if all of these cultivars should remain in the third Eurasian type. High Density (Aggressive) type cultivars were underrepresented in the current study. Bonos et al. (2000) reported that this type may only be related by the common growth characteristic of high shoot density, indicating that this type may not be a genetically related group. This observation was supported by the fact that Limousine and ‘Julius’ (both previously High Density) are now classified as Limousine-type cultivars (Table 1; Fig. 1), whereas NorthStar (previously High Density) is now classified as an “other” type cultivar (Table 1; Fig. 1). SSR marker genotyping of additional cultivars that were previously considered to be High Density (Aggressive) type cultivars will be needed to determine whether this group should be retained or removed in the revised classification system. The CELA type, represented by the type cultivars Challenger (Meyer et al., 1987), Eclipse (Funk et al., 1981), Liberty (Brilman et al., 1989), and Adelphi (Funk et al., 1973), was also underrepresented in the current study. The situation with this group was different from the High Density (Aggressive) type in that the CELA type may have been a genetically related classification type. The cultivars Adelphi and Eclipse shared a common pedigree (Alderson and Sharp, 1994) as did Challenger and Liberty (Alderson and Sharp, 1994). It will be problematic to determine the genetic relationships of the CELA classification type using SSR markers because many of the cultivars in this group currently have very limited or no commercial production and/or have infrequently been used in the development of new cultivars, meaning that this group has effectively been discontinued. The final classification type, texas bluegrass × kentucky bluegrass hybrids, was not included in the current study because hybrid cultivars in this classification type may be populations rather than apomicts.