The fourteen selected ISSR primers produced 389 bands with an average of 27.78 bands per primer, of which 313 (80.5%) were polymorphic with an average of 22.36 bands per primer. The different ISSR primers amplified the number of bands from 20 (ISSR-14) to 40 (ISSR-11) having a range of 400 to 1200 bp fragments size. The number of polymorphic fragments detected by each primer ranged from 16 to 27. The most polymorphism was shown by ISSR-5, which showed 92.8% polymorphism. The mean of polymorphism information content (PIC) value for primers was 0.328 which ranged from 0.205 (ISSR-5) to 0.433 (ISSR-15) (Table 1). Figure 1 shows the banding pattern of genotypes generated by the primer ISSR-7.

The average values of observed number of alleles (Na), effective number of alleles (Ne), Nei’s gene diversity (He) and Shannon’s information index (I) for all primers were 2, 1.56, 0.33 and 0.56, respectively.

The similarity matrix of the 27 genotypes from ISSR data using Jaccard’s coefficient showed that the genetic relationships of genotypes were different varying between 0.057 and 0.524 (data not shown). The lowest similarity was revealed between “18-Gr” and “27-II” genotypes (0.057), whereas the highest similarity was between “3-Gs” and “6-Gs” genotypes (0.524). The latest two common bermudagrass accessions, “3-Gs” and “6-Gs”, are tetraploid and are collected from the same origin (Table 2) with some similar morphological characteristics such as length and width of leaf and diameter of rhizome. The very close genetic relationship of these accessions suggests that they are a clonal propagule of a single plant.

The data obtained from ISSR analysis of 27 bermudagrass genotypes was subjected to cluster analysis. In order to recognize the best clustering and similarity coefficient methods and the cophenetic correlation coefficient, a measure of the correlation between cophenetic matrix constructed from tree and similarity matrix, was calculated for each method combination. From different methods, the highest value (r = 0.828) was observed for the un-weighted paired group method with arithmetic average (UPGMA) clustering method based on Jaccard’s similarity coefficient, suggesting that the cluster analysis represents the similarity matrix.

The UPGMA dendrogram grouped the 27 accessions into six clusters considered as A, B, C, D, E and F with a similarity coefficient of 0.23 (Figure 2). Seven tetraploid accessions were grouped together as cluster A of which three (“1-Lr”, “3-Gs”, “6-Gs”) and four (“11-Ns”, “13-Ns”, “15-Bs”, “58-Chr”) of them were from Charmahal Bakhtiari and Isfahan provinces, respectively. These accessions have some similar morphological characteristics such as leaf and rhizome color. Genetic similarity coefficient values for accessions in cluster A ranged from 0.198 to 0.524.

The major cluster B included the largest number of accessions (11 tetraploides) that had some similar morphological characteristics, such as leaf width, length of flower, stem and leaf, and stolon color. The grouping of these accessions from the same province (Isfahan) may reflect their having a common geographic origin. Genetic similarity coefficient values in this group were from 0.128 to 0.458.

Cluster C embraced two diploid accessions (“17-GN1” and “88-Khl”) and one triploid accession (“44-II”). These accessions have also similar morphological characteristics such as width of leaf and stolon color.

Four triploid hybrids (Cynodon dactylon × Cynodon transvaalensis), Tifgreen, Tifdwarf, Tifway and Midlawn were separated from all accessions and grouped together as cluster D.

Two tetraploid accessions, “19-Gcr” and “18-Gr”, collected from Gilan (Annual rainfall 1220 mm) were separated from all accessions in single clusters as labeled E and F, respectively in Figure 2.

Principal coordinate analysis (PCoA) based on genetic similarity metrics was used to visualize the genetic relationships among accessions. The first three eigenvectors accounted for 25.61% of the total molecular variation. Since the original data are not highly correlated in PCoA, the first few PCs do not explain much of the original variation. Assessment of genetic relationships on the basis of the first three PCs could lead to misleading interpretations and therefore analysis of genetic relationships among accessions should be based on cluster analysis as well as the optimal number of PCs that explain the maximum amount of the original data.

Twenty seven bermudagrass genotypes were included in this study. Among them, 23 genotypes were common bermudagrass types (Cynodon dactylon) and four were triploid hybrids of Cynodon dactylon × Cynodon transvaalensis (Tifdwarf, Tifgreen, Tifway and Midlawn). Stolons of twenty-three accessions of Cynodon dactylon (L.) Pers. had been collected from Isfahan, Charmahal Bakhtiari, Gilan and Mazandaran, provinces of Iran (Table 2). Each accession was grown in 20 cm diameter pots under green-house conditions. The green-house was maintained at 25 ± 1 °C. Bermudagrass DNA samples were isolated from fresh leaf tissue according to Dellaporta et al. [32].

Fourteen random ISSR primers (Operan technologies, Inc, Japan) were used to amplify DNA fragments of the selected materials. These primers were selected based on preliminary screening, producing a higher level of polymorphism and reproducible fragment patterns (Table 1). ISSR reactions were conducted in a volume of 25 μL consisting of 2.5 μL of 10× PCR buffer, 1.5 mM MgCl₂, 1 U of Taq DNA polymerase (Roch Co. Germany), 200 μM of dNTPs, 1.25 μM of each primer, and template DNA of approximately 50 ng.

Amplifications were carried out in a Thermal Cycler (T-1 Theroblock, Germany) with the following PCR program: 3 min of initial denaturing at 94 °C, 40 cycles of 94 °C for 45 s, 1 min for annealing at the primer-specific melting temperature, and 72 °C for 4 min; followed by a final extension of 10 min at 72 °C. The amplified fragments were separated on a 1.5% agarose gels in 1× TBE buffer running at 65 V constant for 1.5–2 h and then stained by ethidium bromide (1 μg·mL⁻¹) and photographed under UV light in a gel documentation system (Villber Lourmat, France)

All clearly detectable ISSR product bands were scored as either presence (1), absence (0) or ambiguous (9) of each band for all accessions and the matrix of ISSRs data was assembled. All amplifications were repeated at least twice and only reproducible and well-defined bands were considered for analysis.

Polymorphism information content (PIC) was calculated using the formula PIC = 1 − ∑pi², where pi is the frequency of the ith allele [33]. Jaccard’s similarity coefficient values for each pairwise comparison between accessions were calculated and a similarity coefficient matrix was constructed. This matrix was subjected to an un-weighted paired group method with arithmetic average (UPGMA) [34] to generate a dendrogram using software NTSYS-pc Version 2.02 [35]. The MXCOMP subroutine was used to calculate a cophenetic correlation coefficient between the similarity matrix and respective dendrogram derived matrix to measure suitability-of-fit [36]. Winboot software [37] was used for a bootstrap analysis with 1000 replicates to obtain the confidence of branches of the cluster tree. The frequency of occurrence of each marker in each genotype was computed, to render a matrix of 27 accessions by ISSR markers. These matrices were subjected to principal coordinate analysis (PCoA) to show the relationship between genotypes.

Observed number of alleles (Na), effective number of alleles (Ne), Nei’s gene diversity (He) [38] and Shannon’s information index (I) [39] were estimated for total genotypes using POPGENE software version 1.32 [40].