Common bermudagrass, Cynodon dactylon var. dactylon is an economically and ecologically important warm-season perennial grass that has been widely cultivated for turf, forage, soil protection, and remediation in the world1. The grass is the only Cynodon taxon that enjoys globally remarkable distribution between 45° S. Lat. and 53° N. Lat2. and harbors enormous genetic diversity3,4. In modern plant breeding programs, C. dactylon is often crossed with African bermudagrass (C. transvaalensis Burtt-Davy) in generating vegetatively propagated, fine-textured turfgrass cultivars5 or with C. nlemfuensis var. nlemfuensis Vanderyst in developing highly digestible and high yielding forage cultivars6. Collectively, forage and turf bermudagrass crops have been grown on ~20–25 million hectares in the USA, demonstrating its substantial economic value7. Although it is difficult to estimate the acreage of C. dactylon grown in many other countries, its use for turf and forage is extensive.

The base chromosome number of Cynodon species is x = 98. A series of euploidies including diploid (2x = 18), triploid (3x = 27), tetraploid (4x = 36), pentaploid (5x = 45), and hexaploid (6x = 54) have been reported in the genus with samples collected worldwide over the past century8,9,10,11,12,13,14,15,16. Nevertheless, tetraploid is the predominant ploidy in C. dactylon var. dactylon and in the Cynodon genus in general2,12,13,14,15. Harlan and de Wet2 proposed “all tetraploids are autotetraploids” in C. dactylon var. dactylon and that these probably were derived from the diploid var. aridus (2x = 18) on the basis of morphological, cytogenetic, and rhizomatous similarity between the two taxa.

DNA markers are powerful tools that have been employed in C. dactylon to investigate genomic structure and constitution, and to construct genetic maps. In one study, an F1 progeny population derived from C. dactylon ‘T89’ (4x = 36) × C. transvaalensis ‘T574’ (2x = 18) was analyzed to construct two genetic maps, one for each parent, with single-dose restriction fragments17. They reported that the C. dactylon T89 exhibited “polysomic inheritance of an autotetraploid” based on the predominance of coupling linkages over repulsion. Using the same progeny population, Harris-Shultz et al.18 observed disomic inheritance for multiple markers on seven of 34 linkage groups (LGs), and therefore reported that T89 may be a segmental allotetraploid or an allotetraploid, rather than an autotetraploid. This population was also used in a recent investigation by Khanal et al.19 who contributed additional SSR markers to the previous maps. Their study indicated that the C. dactylon parent was in a process of diploidization after a whole genome duplication (WGD) event (i.e., an autotetraploid behaving as a segmental polyploid), as some markers showed polysomic inheritance19. In a different line of experiments, C. dactylon SSR markers showed disomic inheritance in two first-generation selfed (S1) progeny populations20. Between the two S1 populations, the one derived from selfing of ‘A12359’ had less segregation distortion than that derived from ‘Zebra’. Consequently, the A12359 population was used for subsequent research including the development of an SSR marker-based genetic map21. All SSR markers showed disomic inheritance in the mapping population21. Obviously, more research is needed to definitively reveal the genome constitution and subgenome differentiation of C. dactylon.

Previous genetic mapping research in C. dactylon has generated valuable information, yet a relatively small number of markers, specifically 291 loci in T8919 and 252 loci in A1235921, have been mapped. Third-generation marker systems, such as genotyping-by-sequencing (GBS) can generate a large volume of single nucleotide variants (SNVs), which can be used to establish saturated genetic maps. Accordingly, the major objective of the present study was to construct a high-density genetic map using GBS generated SNVs. C. dactylon is a member of the genus Cynodon, subtribe Eleusininae, tribe Cynodonteae, subfamily Chloridoideae, and family Poaceae22,23, and the evolution history of Cynodon is elusive24. As all SNVs in this study were developed from C. dactylon DNA, the tagged sequences would provide landmarks for comparative genomic analysis. Therefore, the second objective was to explore the evolution pathway of C. dactylon as compared with fully sequenced genomes of selected species in the grass family. The findings would add to the knowledge pool in the evolution of the grass family.


GBS and SNV calls

One hundred and thirty S1 progeny and the parent A12359 were sequenced with GBS. A total of 884,977,554 reads were generated, of which 846,616,373 reads (96%) matched the ApeKI enzyme cut site remnant (CWGC) and the 96 barcodes. Sequenced reads were trimmed to 64 bp in length, and a total of 232,166 SNVs were obtained with loose SNV calling criteria (minor allele frequency ≥ 0.01). After removing genotype data with < 6 reads or a missing rate > 10%, a total of 39,013 SNVs were retained for further filtering. A chi-square test was then conducted on each retained SNV to examine the expected segregation ratio of 1:2:1, and the 7443 SNVs meeting this criterion at P > 0.01 were retained for genetic mapping.

Genetic map

The filtered 7443 SNV markers and 266 SSRs from Guo et al.21 were used for LG construction. Markers with segregation distortion (P < 0.01) were excluded during the framework LG construction. Initially, 18 LGs were made using the maximum likelihood mapping algorithm, of which 16 LGs showed extensive genomic synteny with eight of the 10 O. thomaeum chromosomes (Chr. 1–7, 9). The remaining two LGs showed clear synteny with O. thomaeum chromosome 8 but with a small number of markers in each LG, resulting in large inter-maker spacing. Therefore, to retrieve more markers for these two small LGs, a new dataset was created as follows: markers of these two small LGs were combined with segregation distorted markers (P < 0.01) showing significant alignment with O. thomaeum chromosome 8 (i.e., comparative genomics approach as described in Methods). Two LGs, 15 and 16, were obtained from this new dataset. After removing redundant markers with a marker interval < 0.4 cM, each LG was recalculated using the regression mapping algorithm. The maximum likelihood algorithm was efficient in ordering markers but often generated a severely expanded genetic map, whereas the regression mapping algorithm was computationally intensive but generated a reasonable map length. Finally, a total of 3544 markers (3322 SNVs and 222 SSRs) were resolved into 18 LGs. The genetic map spanned 1467.3 cM with an average inter-marker spacing of 0.41 cM (Table 1, Fig. 1a). Based on comparative mapping results, extensive genomic collinearity was observed between C. dactylon and O. thomaeum (Fig. 1b). On average, each of the first nine O. thomaeum chromosomes matched two LGs of C. dactylon. Therefore, we were able to determine the two homeologous LGs in C. dactylon based on their homology to the O. thomaeum chromosomes (Table 1). We further numbered C. dactylon LGs and designated their orientation based on homology with O. thomaeum chromosomes. The number of markers per LG ranged from 91 on LG 16 to 291 on LG 1. The length of each LG ranged from 55.26 cM for LG 16 to 129.95 cM for LG 4. Two genetic gaps (a distance between two adjacent markers larger than 15 cM) were detected on the 18 LGs. They were located on LG 4 with a genetic distance of 20.67 cM and LG 8 with a gap of 34.12 cM.

Table 1 Statistics summary for the Cynodon dactylon linkage groups (LGs) of single nucleotide variants (SNVs) and simple sequence repeats (SSRs) as related to chromosomes of Oropetium thomaeum.
Fig. 1: Genetic map of common bermudagrass (Cynodon dactylon) and its comparison with O. thomaeum.
figure 1

a A high-density genetic map of common bermudagrass ‘A12359’. Genetic distance is shown on the left in centimorgans (cM) and linkage group (LG) numbers are shown at the bottom. b Genomic synteny plot between C. dactylon and Oropetium thomaeum. Horizontal axis shows genetic position of markers on 18 C. dactylon LGs; vertical axis shows the map position of markers aligned to the 10 O. thomaeum chromosomes. Each dot represents a single marker.

Analysis of segregation distorted markers

All 39,013 SNV markers were initially used to analyze the distribution of segregation distortion. Due to computational constraint, redundant markers with genetic distances < 1.5 cM were removed after the first round calculation with the maximum likelihood mapping algorithm, resulting in 2904 markers retained. Using these 2904 markers, of which 1861 (64.08%) were distorted markers, we recalculated the 18 LGs under the regression algorithm (Fig. 2). The number of distorted markers on each LG are given in Table 1. The segregation distorted markers were observed on all 18 LGs but not randomly distributed. It appears that most of the segregation distorted markers formed clusters on each of the LGs (Fig. 2). Approximately 80% and 82% of the total mapped markers on LG 5 and LG 6 were segregation distorted, respectively. The distorted markers on LG 5 expanded the linkage group by ~71 cM compared with that from the non-distorted map (Fig. 2), which was primarily due to their aggregation on one telomeric region (Fig. 2). Similarly, the distorted markers on LG 6 elongated the LG by ~90 cM compared with its counterpart from the non-distorted map, and these markers mostly appeared to intermix with the non-distorted markers. Comparing the LGs with and without distorted markers, we observed that the distorted markers did not dramatically change the marker orders on the LGs.

Fig. 2: Distribution of segregation distortion markers on Cynodon dactylon linkage groups and its alignment to Oropetium thomaeum chromosome 10.
figure 2

Segregation distortion markers on linkage groups (LGs) 1–18. 3-1, 4-1 5-1, and 6-1 are the LGs mapped with the SNV markers aligned to Oropetium thomaeum chromosome 10. The numbers on the top row indicates the LGs and left ruler shows the genetic distance in centimorgans (cM). Magenta dots aligned with red lines indicates the same markers mapped on both the LGs.

Comparative mapping

Among the three Chloridoideae species C. dactylon, O. thomaeum, and Zoysia japonica, it is evident that a high level of chromosome-level collinearity exists between C. dactylon and O. thomaeum (Fig. 1b) and that a similar level of collinearity exists between Z. japonica and C. dactylon (Supplementary Fig. 3A). The results are not unexpected since O. thomaeum and C. dactylon are classified into the same tribe Cynodonteae23. O. thomaeum (2x = 20) is an emerging model for desiccation tolerance and genome size evolution in grasses and has an extensive degree of chromosome-level collinearity with Sorghum bicolor (2x = 20)25,26. Among the 3322 SNVs mapped on this C. dactylon genetic map, 702 SNV tagged sequences (21.13%) shared high synteny with O. thomaeum genome (Table 2). A clear 2:1 chromosome correspondence ratio was observed between the 18 LGs of C. dactylon and the chromosomes 1–9 of O. thomaeum. However, no discernable chromosome-level correspondence between C. dactylon LGs and O. thomaeum chromosome 10 was found (Fig. 1b). The comparative analysis also showed frequent chromosomal rearrangements in C. dactylon relative to O. thomaeum. Obvious gaps were observed at centromeric regions for each homeologous LG pair. Interestingly, inverted insertions of ancestral chromosomal segments were observed in several C. dactylon LGs (e.g., LGs 3, 15, 16, 17, 18), and C. dactylon LG 6 exhibited syntenic relationship with the long arm of O. thomaeum chromosome 3, whereas LG 8 showed synteny with telomeric regions of O. thomaeum chromosome 4 in an inverted fashion. C. dactylon LGs 15 and 16 were obtained after incorporating segregation distorted markers (P < 0.01) and showed less pronounced genomic synteny with O. thomaeum chromosome 8.

Table 2 Comparative genomics analysis between Cynodon dactylon and six other species in the Poaceae family.

In order to identify the possible reason for the absence of genomic synteny between C. dactylon LGs and O. thomaeum chromosome 10, we further investigated all raw SNV tagged sequences that aligned to O. thomaeum chromosome 10 and constructed genetic maps using these SNVs. A total of 2048 SNV-tagged sequences were aligned to O. thomaeum chromosome 10, of which 320 SNVs were successfully grouped into four new LGs (Supplementary Data 2). Among these 320 mapped SNVs, 32 were mapped on the segregation distorted map, including 7 on LG 3, 6 on LG 4, 6 on LG 5, and 13 on LG 6 (Fig. 2). Thus, these four LGs were named as LG 3-1, LG 4-1, LG 5-1, and LG 6-1, accordingly. When these four LGs were compared with O. thomaeum chromosome 10, they showed two homeologous groups, with LG 3-1 and LG 4-1 corresponding to one arm of O. thomaeum chromosome 10, and LG 5-1 and LG 6-1 corresponding to the other arm (Fig. 3). As expected, a high degree of chromosome-level collinearity was observed between Z. japonica and C. dactylon (Supplementary Fig. 3A). The basic chromosome number of C. dactylon is 9, while it is 10 in Z. japonica. Each C. dactylon LG pair corresponded to one pair of Z. japonica chromosomes except for Z. japonica chromosomes 15 and 16 (Supplementary Fig. 3A). Such mutual correspondence implied their divergent speciation from two common ancestors as Z. japonica is an allotetraploid species27. Similarly to those observed between C. dactylon and O. thomaeum, numerous chromosomal rearrangements differentiate these two species. Among 3322 SNVs in the C. dactylon genetic map, 703 (21.16%) were unambiguously aligned to Z. japonica genome sequence (Table 2).

Fig. 3: Comparison between Oropetium thomaeum chromosome 10 and four segmental Cynodon dactylon linkage groups (LGs).
figure 3

A total of 320 SNVs that aligned to O. thomaeum chromosome 10 were investigated and formed four LGs of Cynodon dactylon.

Similarly, syntenic blocks and collinearity were also observed between C. dactylon and the selected grass species in other subfamilies (Fig. 4, Supplementary Fig. 3A–C). Panicoideae is the sister subfamily of Chloridoideae. Three Panicoideae grasses, S. bicolor (2x = 20), Setaria italica (2x = 18), and Miscanthus sinensis (2x = 38) were employed in the comparative analysis (Supplementary Fig. 3B–D). In brief, each S. italica chromosome corresponded to two C. dactylon LGs (Supplementary Fig. 3B), and each S. bicolor chromosome corresponded to two C. dactylon LGs except for chromosome 8 (Supplementary Fig. 3C). M. sinensis has one more haploid chromosome than C. dactylon, and the synteny analysis revealed mutual correspondence between these two species as that observed between C. dactylon and Z. japonica (Supplementary Fig. 3A, D).

Fig. 4: Comparison between Cynodon dactylon and Oryza sativa genomes.
figure 4

a Comparison between C. dactylon linkage groups (LGs) and O. sativa chromosomes. b Circos plot showing O. sativa chromosome 9 inserted into the centromeric region of chromosome 6 to form C. dactylon LGs 3 and 4; O. sativa chromosome 10 inserted into the centromeric region of chromosome 2 to form C. dactylon LGs 1 and 2. The gray filled segments represent C. dactylon LGs and color filled segments represent O. sativa chromosomes.

Oryza sativa (2x = 24) is a member of Ehrhartoideae (syn. Oryzoideae), representing the phylogenetically most distant species with C. dactylon as compared with the other five species in this study. Nine of the 12 O. sativa haploid chromosomes showed 1:2 correspondence with C. dactylon 18 LGs except for O. sativa chromosomes 9, 10, and 12 (Fig. 4a). The O. sativa chromosome 9 showed local synteny with the central part of C. dactylon LGs 3 and 4. In similar fashion, the O. sativa chromosome 10 was related to the central part of C. dactylon LGs 1 and 2. The O. sativa chromosome 12, however, did not exhibit any significant connection to C. dactylon LGs. In summary, between C. dactylon and the six other grass species, highest syntenic levels were observed between C. dactylon and the other two Chloridoideae grasses O. thomaeum (21.13%) and Z. japonica (21.16%; Table 2), followed by three Panicoideae grasses S. italica (13.96%), S. bicolor (12.34%), M. sinensis (12.01%), and tailed by O. sativa (8.31%).


To the best of our knowledge, we developed the most-dense genetic map for C. dactylon using SNVs generated through GBS. Compared with the tetraploid C. dactylon maps of Khanal et al.19 and Guo et al.21 this map has increased marker density by more than tenfold. The average inter-marker spacing in this map was 0.41 cM, while those values from Khanal et al.19 and Guo et al.21 were 12.5 cM and 4.3 cM, respectively. The tag sequences of mapped markers in this study have been used to facilitate a whole-genome assembly of C. dactylon by resolving uncertainties in assembling large scaffolds due to high heterozygosity in the genotype A12359 (Y.Q.W. and J.F., personal communication). The tag sequences of this high-density genetic map also provided valuable community resources for marker-assisted selection (MAS) and genetic diversity studies.

Segregation distortion is a common phenomenon in plants and animals. In this study, we had to include severely segregation distorted markers (P < 0.01) to obtain LGs 15 and 16. Similarly in Miscanthus, it was necessary to include segregation distorted markers to obtain the ‘missing’ LG 1528. Cynodon dactylon is largely an outcrossing species29, whereas some genotypes such as A12359 can produce selfed progeny20. Although this self-compatible parent produced hundreds of S1 progeny, many gametes with low fitness or carrying potential lethal alleles in homozygous loci might have perished, and thus were not represented in this S1 population, leading to varying degrees of segregation distortion. Since this population was derived from selfing a single parent, the segregation distortion observed in this study would be more extensive than populations derived from crossing two parents.

The interest concerning the genome of Cynodon species started in the early 1930s when C. dactylon was recognized as an important turf and forage species30. Investigations indicated that tetraploid C. dactylon var. dactylon is the only taxon of the genus, having worldwide geographic distribution11. The cosmopolitan form has contributed the most germplasm/genes to developing modern forage and turf cultivars. One fundamental question regarding the worldwide tetraploid C. dactylon genome is whether it is an autopolyploid with polysomic inheritance vs. an allopolyploid with disomic inheritance. Chromosome pairing behavior in meiosis may have provided insights into chromosome affinity. Forbes and Burton8 reported that dactylon cv. Coastal had 0.15 I’s, 16.00 II’s, and 0.96 IV’s and that PI 226011 had 1.36 I’s, 15.69 II’s, 0.18 III’s, and 0.68 IV’s. Hanna and Burton31 observed similar 0.1–1.72 I’s, 16.13–17.45 II’s, 0–0.15 III’s and 0.26–0.39 IV’s in three tetraploid dactylon cultivars (Coastal, Midland, and Suwanee). The results may suggest higher affinity/homology between two chromosome pairs, resulting in the tetravalent formation in meiosis. Harlan and de Wet2 performed a large investigation of meiotic chromosome behavior in C. dactylon var. dactylon germplasm, including all three races (tropical, temperate, and seleucidus) collected from multiple countries. Their study indicated that among 50 hybrids, 11 had regular meiosis with 18 II’s; 13 showed slightly irregular meiosis with 0–2 I’s, 17–18 II’s, and 0–1 IV’s; 19 exhibited irregular meiosis with 2–6 I’s, 14–18 II’s and 0–2 VI’s; and seven demonstrated very irregular meiotic behavior with 0–8 I’s, 10–18 II’s, and 0–2 VI’s. The regular meiotic events were observed in hybrids between parental plants collected from the most similar origins, while the most irregular ones were found in progeny of parents from more distant locations with a few exceptions2. It appears that tetraploid C. dactylon has a complex genome with various homologies between two subgenomes, which may have evolved over time under different conditions through natural selection.

Guo et al.20 studied segregation patterns of SSR markers in two S1 progeny populations and reported that the two parental C. dactylon plants were allotetraploidy. Using one of the two populations, they mapped 249 SSRs into 18 LGs. Among the mapped markers, 154 non-distorted loci had the 1:1 ratio of coupling vs. repulsion, suggesting the allopolyploid origin of the parent A1235921. Roodt and Spies32 reported that one tetraploid C. dactylon was allotetraploidy while another was segmental allotetraploidy. On the basis of observing two strong and two weak signals of 45S rDNA and 5S rDNA FISH probes on metaphase chromosomes in mitosis and 18 bivalents of chromosome pairing in meiosis of one tetraploid C. dactylon, Gong et al.33 indicated that the tetraploid C. dactylon had two different subgenomes. The findings in this study lend strong evidence for disomic inheritance and allotetraploid in C. dactylon. Since C. dactylon is adapted to a large range of environments in the world, it is likely that in the evolutionary process, accumulation of genomic changes under local natural selection pressures has resulted in the complexity of genome structure and constitution in the species.

Genomic synteny and collinearity is a common feature in the grass family (Poaceae), which is unarguably the most important plant family with about 12,000 species that are classified into 12 subfamilies23,34. In the present study, we compared C. dactylon with O. thomaeum and Z. japonica of the subfamily Chloridoideae; S. bicolor, S. italica, and M. sinensis of Panicoideae; and O. sativa of Ehrhartoideae (Table 2). In addition, both Chloridoideae and Panicoideae belong to the ‘PACMAD’ clade (Panicoideae, Arundinoideae, Chloridoideae, Micrairoideae, Aristidoideae and Danthonioideae), whereas Ehrhartoideae belongs to the ‘BEP’ clade [Bambusoideae, Ehrhartoideae (formerly Oryzoideae) and Pooideae]. Therefore, it is not unexpected that C. dactylon showed high genomic syntenic relationships with O. thomaeum and Z. japonica due to their close phylogenetic relationship within the Chloridoideae subfamily (Table 2). The levels of genomic synteny between C. dactylon and the three Panicoideae species were similar, and the results agreed with the relatively close phylogentic relationship between the subfamilies Chloridoideae and Panicoideae. The lowest level of genomic synteny between C. dactylon and O. sativa further agreed with the most distant phylogenetic relationship between subfamilies Chloridoideae and Ehrhartoideae.

Chloridoideae subfamily is yet to be explored in genomics and evolutionary studies23. Most comparative genomic analyses between Chloridoideae grasses have used S. bicolor, a well sequenced and studied Panicoideae grass, as reference. Sorghum is in the Panicoideae subfamily, which may have diverged from the common ancestor of Chloridoideae ~31 million years ago. Despite this divergence, VanBuren et al.26 demonstrated the ten chromosomes in O. thomaeum are largely collinear to the corresponding ten chromosomes in S. bicolor, though large-scale inversions and translocations were identified. Wang et al.35 reported WGD in Z. japonica relative to S. bicolor. In this study, we also observed that each S. bicolor chromosome except for chromosome 8 showed collinear synteny with two C. dactylon LGs (Supplementary Fig. 3C). Within the Panicoideae subfamily, multiple studies have confirmed the WGD in Miscanthus relative to S. bicolor36,37,38.

It has been proposed that the common ancestor of grasses most likely had a basic chromosome number of seven39,40. The basic chromosome number increased to 14 via one WGD, and then reduced to 12 via two nested chromosome fusions (NCFs)41. O. sativa genome has a basic chromosome of 12 and has been believed to likely resemble the 12 paleo-ancestor chromosomes (ρ) and provides an important reference for comparative genomics studies35. In this study, the comparative genomic analysis between C. dactylon and O. sativa revealed paleo-ancestor chromosome 9 (ρ9) inserted into the centromeric region of paleo-ancestor chromosome 6 (ρ6) to form C. dactylon LGs 3 and 4 (ρ6-ρ9-ρ6; Fig. 3b); and paleo-ancestor chromosome 10 (ρ10) inserted into the centromeric region of paleo-ancestor chromosome 2 (ρ2) to form C. dactylon LGs 1 and 2 (ρ2-ρ10-ρ2). Such paleo-chromosome merges during speciation have been demonstrated in many other grass species. In Z. japonica (x = 10), ρ6-ρ9-ρ6 and ρ2-ρ10-ρ2 have been proved to be involved in the 12-to-10 process35; in Eleusine coracana, ρ6-ρ9-ρ6, ρ2-ρ10-ρ2, and ρ5-ρ12-ρ5 have been reported to be engaged in the 12-to-9 process42; in S. bicolor, ρ7-ρ9-ρ7 and ρ3-ρ10-ρ3 were demonstrated in the 12-to-10 process43; and in S. italica, ρ7-ρ9-ρ7, ρ3-ρ10-ρ3, and ρ5-ρ12-ρ5 were demonstrated in the 12-to-9 process44. In the Chloridoideae subfamily, all three species E. coracana, C. dactylon, and Z. japonica have undergone ρ6-ρ9-ρ6 and ρ2-ρ10-ρ2 paleo-chromosome merges. With this high-density genetic map, we discovered the tetraploid C. dactylon had the same two merges (ρ6-ρ9-ρ6 and ρ2-ρ10-ρ2), resulting in chromosome number reduction from 12 to 10.

An interesting finding from this study was that no genomic synteny was found between C. dactylon LGs and O. thomaeum chromosome 10, and between C. dactylon LGs and O. sativa chromosome 12 in the initial analysis (Figs. 2, 4). Both O. thomaeum and O. sativa have high-quality genome sequences25,44. The comparison between O. thomaeum and O. sativa genome sequences revealed that O. thomaeum chromosome 10 is largely syntenic with O. sativa chromosome 12 (Supplementary Fig. 3E). This phenonmenon echoes our findings that no genomic correspondence was found between C. dactylon LGs with O. thomaeum chromosome 10 and O. sativa chromosome 12 (Figs. 1, 4). Evolutionary studies suggested that the ancestors of all extant diploid angiosperm species underwent whole-genome duplication, which is consistent with the progenitors of extant diploid plants experiencing polyploidization followed by diploidization45,46,47,48. Diploidization includes loss of entire chromosomes or large segments of chromosomes, losses of one copy of duplicated genes, and various chromosome rearrangements. However, the loss of a whole chromosome corresponding to the paleo-ancestor chromosome 12 and O. thomaeum chromosome 10 is unlikely to have occurrd during C. dactylon genome evolution. The four new LGs (LG 3–1, LG 4-1, LG 5-1, and LG 6-1 in Fig. 2) mapped with the SNVs aligned to O. thomaeum chromosome 10 (Fig. 3) indicated that O. thomaeum chromosome 10 was divided into two parts, and segmentally disseminated into two C. dactylon homoleogous chromosome pairs. LGs 3 & 4 are one pair and LGs 5 & 6 another. This result indicated that the chromosome reduction from 10-to-9 may have happened before the formation of the allotetraploid C. dactylon.

The ρ6-ρ9-ρ6 and ρ2-ρ10-ρ2 paleo-chromosome rearrangements are common in the Chloridoideae subfamily. The composite structures of bermudagrass LGs 3, 4 (one half of paleo-ancestor chromosome 12 inserted into 6, ρ6-ρ12-ρ6) and LGs 5, 6 (the other half of paleo-ancestor chromosome 12 into 1, ρ1-ρ12-ρ1) are unique to C. dactylon. The new findings in the current study represented a different evolution event in C. dactylon from the finger millet ρ5-ρ12-ρ5 chromosome rearrangement within the same subfamily41. To summarize the results in a phylogenetic context, the evolutionary model to demonstrate the chromosome number reduction for C. dactylon was modified from previous publications35,49 of sequenced and mapped grasses in the Panicoideae, Chloridoideae, and Ehrhardoideae subfamilies (Fig. 5). Evidently, the paleo-chromosome 12 has played an important role in evolution within Chloridoideae. The high-density genetic map developed in this study provides a clear view of the allotetraploid genome structure of C. dactylon. The base number of nine chromosomes of the species was demonstrated to be derived from an ancestor of 10 chromosomes through the two chromosomal rearrangement events (ρ1-ρ12-ρ1 and ρ6-ρ12-ρ6). The study confirmed that two NCFs (ρ6-ρ9-ρ6 and ρ2-ρ10-ρ2) happened to all the species tested in the Chloridoideae subfamily. As whole genome sequences are available in the Cynodon genus, more in the Chloridoideae subfamily and Poaceae family, further investigations might unveil mechanisms in chromosome evolution. The high-resolution genetic map will be a powerful tool for establishing associations between phenotypes and genotypes, and performing MASs in breeding new cultivars of turf and forage bermudagrass.

Fig. 5: Evolutionary model for bermudagrass (Cynodon dactylon).
figure 5

Bermudagrass (highlighted in magenta color) has unique chromosome rearrangements as compared with sequenced and fine-mapped major grasses in the Panicoideae, Chloridoideae, and Ehrhartoideae subfamilies. Chromosomes in rearrangement events are linked with lines. NCF nested chromosome fusion, IV inversion, CFI chromosome fission, CEJ chromosome end–end joining, Mya millions of years ago, WGD whole-genome duplication. Modified from previous publications35,49.


Cynodon dactylon mapping population

A first-generation S1 population derived from the C. dactylon genotype A12359 (2n = 4x = 36) was used in this study13,20. Detailed information of the population development was described by Guo et al.21. In brief, 130 randomly selected individuals from the S1 population were employed for GBS and genetic map construction.

DNA extraction, library construction, and genotyping by sequencing

Young leaf samples of the progeny and parent were sent to the Genomic Diversity Facility at Cornell University, Ithaca, NY, where DNA extraction, library preparation, and sequencing were performed. Briefly, genomic DNA was extracted with Quick-DNA™ Plant/Seed 96 Kit (Zymo Research, Irvine, CA), and GBS libraries were prepared using the ApeKI enzyme system50. In order to improve read depth and reduce missing data, DNA was replicated twice for each progeny and 14 times for the parent, which was distributed in all libraries. A total of three GBS libraries were sequenced on a HiSeq 2500 (Illumina, San Diego, CA) with 100 bp single-end reads.

Single nucleotide variant calls

Raw sequence data were investigated for base quality using FastQC v0.11.7 ( All three GBS libraries were sequenced with high-quality reads, and only the two ends of reads in each library showed slightly lower quality score (Supplementary Fig. 1). For SNV calling, the Universal Network Enabled Analysis Kit (UNEAK) pipeline in TASSEL 3.051 was implemented due to its ability to distinguish heterozygous genotypes from paralogous loci in organisms lacking reference genome sequences such as C. dactylon in this study. Sequenced reads that matched the ApeKI enzyme cut site remnant (CWGC) and 96 barcodes were trimmed to 64 bp in length, and SNVs were called using the UNEAK pipeline51. Initially, SNVs were obtained with loose SNV calling criteria (minor allele frequency ≥ 0.01, missing data per site <90%). In order to reduce heterozygotes undercalling (i.e., incorrectly call a heterozygous genotype as homozygote due to low coverage), a minimum of six reads per SNV was imposed to call heterozygotes with a theoretical accuracy of 96.88% by assuming a binomial distribution of reads at two alleles for each locus36,52. Genotype data with each ≤5 reads were converted to missing data. Filtered SNVs and previously genotyped SSR markers21 were assembled into a single dataset for genetic map construction.

Linkage map construction

To identify potential seed contaminants (i.e., non selfed progeny) in this population, we performed a principal coordinates analysis (PCoA) across 131 individuals (130 progeny and their parent) using SNV and SSR markers. PCoA was conducted in R (version 3.5.0; R Core Team, 2014) using dist and cmdscale functions. PCoA confirmed that these 130 progeny were truly selfed from A12359 (Supplementary Fig. 2). A linkage map was constructed with JoinMap 553. A chi-square test was conducted on each SNV to test for the Mendelian segregation ratio of 1:2:1, and segregation distorted markers (P < 0.01) were excluded from the initial linkage group construction because distorted markers affect accuracy of linkage mapping by introducing errors in map distance estimation and marker ordering54,55. All markers were initially formatted to the <hk × hk> segregation type (i.e., heterozygous in the parent) according to JoinMap 5.0 user manual in Excel, and missing data were scored as “–”. Markers were first assigned to LGs based on a minimum log-likelihood of the odds (LOD) value of 10.0, and then the LOD threshold was decreased progressively to integrate ungrouped markers. A maximum likelihood mapping algorithm was used to order markers within each LG with the following parameters: linkages with a recombination frequency smaller than 0.4 and a LOD value larger than 1; goodness-of-fit jump threshold for removal of loci = 5; and number of added loci after which to perform a ripple = 1. Kosambi’s mapping function56 was used to calculate inter-marker distance in centimorgans. To reduce redundant markers, one of each marker pair having an interval <0.4 cM was removed. The linkage map was constructed again using the same parameters except the maximum likelihood mapping algorithm was changed to a regression mapping algorithm. The LGs data were loaded to MapChart 2.3257 to make the linkage group chart of the genetic map (Supplementary Data 3).

Analysis of segregation distorted markers

Segregation distortion is a very common phenomenon in plant species. To analyze segregation distortion, an additional genetic map was constructed without removing segregation distorted markers using the same method as described above except that more redundant markers were removed. A marker was excluded if the distance between it and the adjacent markers on an LG (made under the maximum likelihood mapping algorithm) was <1.5 cM. This genetic map, containing only the segregation distorted markers, was made with MapChart 2.3257 and was compared with the previous one lacking severe segregation distorted markers.

Comparative mapping between C. dactylon and other grasses

Cynodon dactylon belongs to the subfamily Chloridoideae, which is composed of more than 1600 species, one of the largest grass subfamilies and is yet underexplored in terms of genomics and evolutionary studies23,35. To fill this gap and to provide further insight into the evolutionary history of grasses, we analyzed the genomic synteny and collinearity between the SNV-associated genomic sequences of C. dactylon and those of several other sequenced grass species. The 64 bp sequences of SNVs arranged on this C. dactylon genetic map were aligned against the reference genome sequences of Oropetium thomaeum v2.026, Zoysia japonica r1.158, Sorghum bicolor v2.043, Setaria italic v2.244, Oryza sativa v7.059, and Miscanthus sinensis v7.1 (DOE-JGI, (sequences of mapped SNVs used in comparison analyses given in Supplementary Data 1). C. dactylon, O. thomaeum, and Z. japonica belong to the subfamily Chloridoideae; S. bicolor, S. italic, and M. sinensis belong to the subfamily Panicoideae; O. sativa is a member of Ehrhartoideae (syn. Oryzoideae). Reference genome sequences of these grass species were downloaded from Phytozome v12.1 ( except for O. thomaeum v2.0, which was downloaded from CoGe (, and Z. japonica r1.1, which was downloaded from Zoysia Genome Database ( Sequences of SNV markers were originally stored in the Tags on Physical Map file from the UNEAK pipeline, and were then formatted to the standard FASTA file using python program TagDigger60. Alignment was performed with Bowtie261 with stringent criteria: -D 20 -R 3 -N 1 -L 18 -i S,1,0.50–local. Genomic synteny plots between C. dactylon and the aforementioned six grass species were made in R (version 3.5.0; R Core Team, 2014) with customized scripts. Circos plots were generated using the circlize package62.

Statistics and reproducibility

All statistical tests were performed using R (version 3.5.0) and base packages. Results are reproducible using the bioinformatic scripts and raw sequencing data that deposited on GitHub and NCBI, respectively.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.