Unveiling the genetic determinants of germination efficiency in common bermudagrass: A genome‐wide association study

Abstract

Common bermudagrass [Cynodon dactylon (L.) Pers.] is an economically and ecologically important warm‐season perennial species widely used for turf, forage, and soil conservation and remediation. Seeding offers economic and practical advantages over vegetative propagation for establishing common bermudagrass. However, the adoption of seeded cultivars is limited by slow germination speed and low germination rates. The genetic basis behind these traits in common bermudagrass remains elusive. Accordingly, the objective of this study was to evaluate the genetic and phenotypic variation and identify genetic loci associated with seed germination‐related traits in a diverse common bermudagrass panel. A diverse panel of 216 genotypes was formed for a genome‐wide association study (GWAS). Seeds for each genotype of the panel were collected in both 2022 and 2023, and germination tests for each year were conducted separately in a randomized complete block design with three replications (100 seeds per replicate) in petri plates inside a growth chamber. The germination process was phenotyped by counting germinated seeds every 3 days from the beginning to determine the germination rate and estimate total germination percentage over a 21‐day period. The panel underwent genotype‐by‐sequencing, and 21,810 high‐quality single‐nucleotide polymorphisms (SNPs) were retained for GWAS analysis. GWAS indicated that 52 unique SNPs were associated with the seed germination traits, of which six were consistent over the 2 years. Twenty candidate genes linked to the consistent SNPs were identified to be involved in seed germination. These findings add valuable information on genetic mechanisms for seed germination and its rapidness, and provide a foundation for developing breeder‐friendly markers to improve seed germination in the species.

Core Ideas

• Early germination rate: seeds germinated between days 3 and 6 exhibited the highest reliability estimate (82%).

• Genome‐wide association study discovered 52 unique significant single‐nucleotide polymorphism (SNP) markers associated with germination traits in common bermudagrass.

• Six SNP markers were consistently associated with germination traits across 2 years.

• Consistent marker‐trait associations were linked to candidate genes involved in hormonal signaling, dormancy regulation, and calcium signaling.

Plain Language Summary

Common bermudagrass is a highly versatile plant used for turf and forage; however, some seeded varieties germinate slowly and unevenly, limiting their broader use. Understanding the genetic basis of seed germination would provide useful information in developing improved cultivars. The research aimed to evaluate a common bermudagrass collection for rapid and reliable germination to discover associated molecular markers. Fifty‐two genomic regions were identified that are linked to seed germination traits, six of which were consistent over 2 years. Many of these identified molecular markers were linked to candidate genes having roles in seed dormancy regulation and hormonal signaling. These findings provide valuable genetic information and tools for breeding bermudagrass varieties with faster and more reliable seedling establishment.

Abbreviations

BLINK

Bayesian‐information and Linkage‐disequilibrium Iteratively Nested Keyway

BLUE

best linear unbiased estimator

DOG1

delay of germination 1

GAPIT

genomic association and prediction integrated tool

Ger12_15D

seed germinated between day 12 to day 15

Ger15_18D

seed germinated between day 15 to day 18

Ger18_21D

seed germinated between day 18 to day 21

Ger3_6D

seed germinated between day 3 to day 6

Ger3D

seed germinated by day 3

Ger6_9D

seed germinated between day 6 to day 9

Ger9_12D

seed germinated between day 9 to day 12

GRI

germination rate index

GWAS

genome‐wide association study

LD

linkage disequilibrium

MAS

marker‐assisted selection

MTA

marker‐trait association

PCA

principal component analysis

PVE

phenotypic variation explained

QTL

quantitative trait locus

SNP

single‐nucleotide polymorphism

TGP

total germination percentage

1. INTRODUCTION

Common bermudagrass [Cynodon dactylon (L.) Pers. var. dactylon] is a warm‐season perennial grass widely adapted in tropical, subtropical, and warm temperate regions worldwide. It is the best known and economically important due to its multiple use, such as for turf, forage, and soil conservation and remediation (Taliaferro, 1995). Common bermudagrass has several advantageous traits used as turfgrass, such as exceptional drought resistance relative to all other turfgrass species, fast recovery from damage, thriving in hot climates, and high turfgrass quality of modern cultivars, making it a widely preferred choice for sports fields, golf courses, home lawns, and other applications. Common bermudagrass is also the most widely used perennial grass for grazing and hay production in the southern United States due to its positive yield response to fertilization, improved digestibility in new cultivar development, and resilience nature (Hill et al., 2001).

Common bermudagrass is an allogamous and largely self‐incompatible species that can be propagated by seed and vegetatively by sod, plugs, and sprigs (Tan et al., 2014; Wu & Taliaferro, 2009). Compared to transplanting by sod, plugs, or sprigs, seeding is a more cost‐effective and less‐labor‐intensive solution to establish bermudagrass (Baltensperger, 2014; Patton et al., 2004). However, due to self‐incompatibility nature, seed‐propagated populations are genetically heterogenous. In addition, seeds are easier to store and transport than sod, plugs, and sprigs. Since the 1910s, bermudagrass seed production industry has gradually developed in Arizona, California, and more recently in Oklahoma, with an annual gross yield between 7 and 9 million kg (Baltensperger, 2014). Since the 1980s, bermudagrass breeding programs have made significant improvement in turfgrass quality with the development and release of seed‐propagated cultivars Princess‐77 at New Mexico State University and Yukon and Riviera at Oklahoma State University (Baltensperger, 1989; Baltensperger et al., 1993; Taliaferro et al., 2004). However, the potential drawback of seed‐producing bermudagrass is the slow seed germination of some cultivars. If common bermudagrass seed germinates slower than weed, their seedlings could fail to compete with weed seedlings. Weed seedlings normally grow more aggressively, resulting in poor bermudagrass stands, reduced growth, and decreased turfgrass quality (McCalla et al., 2004). In addition, quick germination of bermudagrass seed may facilitate the early establishment, which is crucial in the transitional zone to avoid cold damage during the first winter (Patton et al., 2008; Schiavon et al., 2016).

Seed germination is a physiological process that is influenced by genetic factors. Although there is limited information on heritability of seed germination in Cynodon species, related seed yield traits like seed set percentage, inflorescence prolificacy, etc. have revealed high heritability (up to 0.78) (Tan et al., 2022). Germination proceeds after overcoming dormancy and can be defined as a series of events that starts with the quiescent dry seed absorbing water and ends with the embryonic axis becoming longer (Bewley & Black, 1994). The radicle's penetration of the surrounding seed structures is typically a visible indicator of completion of germination. Seed germination and its earliness are important agronomic traits that impact vegetative growth, quality, and overall yield of a crop. Early germination and high germination rate are crucial traits in the seeded bermudagrass improvement pipeline. To assess the speed of germination and overall germination rate in seeded bermudagrass, several studies have explored different cultivars subjected to different environmental factors such as temperature, light, season, salinity, etc. (Batista et al., 2015; Deaton & Williams, 2013; Giolo et al., 2020; Jellicorse et al., 2009). The ideal daytime/nighttime temperatures for bermudagrass seed germination have been reported to be 35/20°C (day/night), even though it can germinate at a range of temperatures from 15°C to 35°C (Ahring & Todd, 1978; Evers & Parsons, 2009). Under field conditions, the best timing for seeding bermudagrass is late spring or early summer (April and May) in a transition‐zone environment (Patton et al., 2008; Richardson et al., 2004). Substantial variations for germination exist within Cynodon species (Giolo et al., 2014). However, the genetic basis underlying the germination in common bermudagrass remains largely unexplored.

Recent advancements in sequencing technologies and molecular tools have offered genetic dissection of important traits in common bermudagrass, including studies on morphological traits (Singh, 2023), cadmium tolerance (Xie et al., 2015), salt tolerance (Shao et al., 2023), cold tolerance (X. Huang et al., 2022), and heat tolerance (Amini et al., 2023). These studies highlighted the significance of association mapping in examining the marker‐trait link in the species. However, a comprehensive association mapping focusing on seed germination traits has not been conducted in common bermudagrass. Genome‐wide association studies (GWASs) are a powerful approach that enable genetic dissection of complex traits, detecting variation in a diverse population based on the linkage disequilibrium (LD) (X. Huang et al., 2010; Zhao et al., 2011). Recently, reference genome sequences of common bermudagrass have been published (Wang et al., 2024; B. Zhang et al., 2022), providing a solid base of genome resources for conducting GWAS analysis of seed germination traits in common bermudagrass. Therefore, the objectives of this study were to (1) evaluate genetic and phenotypic variation present in a diversity panel of bermudagrass and (2) explore marker‐trait association (MTA) for traits related to seed germination in the species.

Core Ideas

• Early germination rate: seeds germinated between days 3 and 6 exhibited the highest reliability estimate (82%).

• Genome‐wide association study discovered 52 unique significant single‐nucleotide polymorphism (SNP) markers associated with germination traits in common bermudagrass.

• Six SNP markers were consistently associated with germination traits across 2 years.

• Consistent marker‐trait associations were linked to candidate genes involved in hormonal signaling, dormancy regulation, and calcium signaling.

2. MATERIALS AND METHODS

2.1. Plant materials

The study utilized a diverse panel of 216 common bermudagrass genotypes, representing nine breeding populations (P1–P9) developed over the last 15 years at Oklahoma State University (OSU) (Table S1). Hexaploid common bermudagrasses were excluded from the populations due to poor seed set. The selected tetraploid populations comprised diverse historical synthetics derived from intercrosses of multiple parental accessions originally collected from China and European countries. Individual genotypes were selected from the populations based on their morphological traits for turfgrass use following greenhouse evaluation and subsequently established in a field nursery. Detailed information of selected 216 genotypes, including their IDs, populations, origin, and breeding history is provided in Table S1. Fifty mature seed heads were collected randomly from each of the selected common bermudagrass genotypes in both 2022 and 2023 in a field nursery established in 2020. The seed heads were immediately kept in labeled paper bags and kept at room temperatures allowing moisture loss for a week. The seed heads from each paper bag were rubbed manually in pans lined with rubber ridged matting to effect threshing of seed. After threshing, clean healthy seeds were obtained by screening followed by air separation with a Model B South Dakota seed blower (Seedburo Equipment Co.).

2.2. Germination test and data collection

A laboratory experiment was conducted to assess the speed of germination and overall germination percentage of the seed samples. The experiment was carried out from January to March of 2023 and 2024 respectively in a randomized complete block design with three replications. Each replication included 100 seeds for each genotype kept on filter paper treated with 0.2% KNO₃ in petri dishes (100 mm diameter × 15 mm height) which promotes uniform germination and transferred to 4°C for 2 weeks using a standard protocol (Association of Official Seed Analysts, 2007). Subsequently, the seed samples in petri dishes were moved into a growth chamber (Percival Scientific, Inc.) with an alternate temperature regime, 16‐h at 20°C in dark and 8‐h at 35°C with light. On the first day in the growth chamber, 0.2% KNO₃ solution was applied again in each petri dish to promote germination followed by maintenance of moisture with distilled water. The Petri dishes were kept in slanting position with the support of wooden planks to avoid overflooding of water. The seeds were arranged in the upper half of the petri dishes by maintaining uniform spaces. The seeds were monitored daily ensuring filter paper was wet for 21 days. Germinated seeds were counted every 3 days for 21 days. Seeds were considered germinated when the radicle became visible to the naked eye. During counting, petri dishes were removed from the growth chamber temporarily. After each count, germinated seeds were removed from petri dishes.

The seed germination traits were assessed for each petri dish and classified into specific time intervals, days to 50% germination (D ₅₀), germination rate index (GRI) and total germination percentage (TGP). The seed germination rate at different time intervals were recorded as number of seed germinated from 0 to 3 days was designated as Ger3D, followed by Ger3_6D (3–6 days), Ger6_9D (6–9 days), Ger9_12D (9–12 days), Ger12_15D (12–15 days), Ger15_18D (15–18 days), and Ger18_21D (18–21 days). These interval‐based germination rates were applied to monitor the germination process closely over a 21‐day test period. For downstream interpretation, these traits were grouped into three developmental phases: early germination (Ger3D, Ger3_6D), mid‐phase germination (Ger6_9D, Ger9_12D and Ger12_15D) and late‐phase germination (Ger15_18 and Ger18_21D).

GRI was calculated using an equation according to Maguire (1962).

$$\mathrm{GRI}=\frac{n1}{d1}+\frac{n2}{d2}+\cdots +\frac{ni}{di}$$

where $ni$ is the number of germinated seeds at time $i$, and $di$ is number of days since the test was started to the $i$th observation. High GRI value indicates high germination speed.

TGP was calculated by using the following equation:

$$\mathrm{TGP}=\frac{nf}{n}\times 100\%$$

where $nf$ is the total number of germinated seeds at the end of the test and $n$ is the total number of seeds in the test.

D ₅₀ was calculated based on the equation provided by Braun et al. (2023): y = Bottom + (Top − Bottom)/[1 +(D ₅₀/𝑥)^Hillslope] using customized scripts in R (R Core Team, 2018), where y is the cumulative germination percentage, Bottom and Top are the lower and upper asymptotes, respectively, x is time in days, and Hillslope is a parameter describing the steepness of the curve. D ₅₀ values were then derived directly as the fitted model parameter representing the time (in days) at which germination reached 50% between the asymptotes.

2.3. Statistical analysis of germination data

Analysis of variance (ANOVA) was conducted on seed germination rate at different time intervals, D ₅₀, TGP, and GRI using the MIXED Procedure within SAS 9.4 (SAS Institute). Variance components were estimated by using Type III moment estimation method (Q. Jiang, Webb, et al., 2014). Reliability estimates (i ²) was calculated for seed germination rate at different time intervals, TGP, and GRI as follows: $i^2={\mathrm{\sigma }}_G^2/({\mathrm{\sigma }}_G^2+{\mathrm{\sigma }}_{GY}^2/Y+{\mathrm{\sigma }}_{GR}^2/R+{\mathrm{\sigma }}_E^2/\mathrm{RY}),$ where ${\mathrm{\sigma }}_G^2$, ${\mathrm{\sigma }}_{GY}^2$, ${\mathrm{\sigma }}_{GR}^2$, and ${\mathrm{\sigma }}_E^2$ represent genotypic variance, genotypic × year interaction variance, genotypic × replication interaction variance and error variance, respectively, and $Y$ and $R$ represent the number of years, and the number of replications, respectively. Pearson correlation coefficients among all the studied traits were estimated based on PROC CORR within SAS 9.4.

2.4. Genotyping‐by‐sequencing (GBS), single‐nucleotide polymorphism (SNP) alignment, and LD analysis

Approximately 50 mg fresh leaf tissues collected from each of the 216 genotypes maintained in separate pots in a greenhouse at the OSU Agronomy Research Station were kept in an ice box and stored in −80°C until shipping to the University of Wisconsin‐Madison Biotechnology Center for DNA extraction and GBS. DNA was extracted from each plant following a standard cetyltrimethylammonium bromide method (Doyle & Doyle, 1987). Genomic sequencing libraries were prepared by using the GBS protocol (Elshire et al., 2011). Briefly, samples of genomic DNA were cleaved enzymatically using a methylation‐sensitive restriction enzyme, ApeKI (New England Biolabs). Following enzymatic cleavage, the sticky ends of DNA were ligated to barcoded adaptors. After barcode ligation, samples were pooled, DNA was PCR amplified, purified, and quantified before sequencing. The evaluation of completed libraries was conducted using the Agilent Bioanalyzer High Sensitivity Chip (Agilent Technologies, Inc.) for quality and Qubit dsDNA HS Assay Kit (Life Technologies) for quantity. Finally, sequencing was performed on an Illumina NovaSeq 6000 with a maximum read length of 250 bp for single‐end sequencing. Under quality control of sequence data, a Phred score of 20 and a trimming software called Skewer (H. Jiang, Lei, et al., 2014) was used to remove any sequencing adapters and low‐quality bases. Because a high‐quality genome assembly for bermudagrass was not available at the time of population sequencing, SNP discovery was done using the Universal Network‐Enabled Analysis Kit GBS pipeline which was a part of Tassel V5.0 bioinformatics analysis package. The raw genotypic dataset contained missing genotype calls represented by “N.” No genotype imputation was performed. Raw SNPs quality filtering was done removing SNPs with minor allele frequency <0.05 with keeping other tassel setting on default. The remaining SNPs were filtered again to improve the quality of called SNPs with the following steps: 1. the genotypes with the reading depth of 1x and 2x were converted into missing data and 2. SNPs with >20% of missing data were removed. The filtered SNP genotypic data used in this study are provided in Table S2.

The retained high‐quality SNPs after the final refinement process were aligned to the recently published bermudagrass reference genome (B. Zhang et al., 2022) using BLASTN v2.8.1 with a threshold of 5 × 10⁻⁶. Alignment with the lowest expect value (E‐value) was retained and the physical positions of SNPs in relation to the reference genome were extracted and set in the original HAPMAP file for further GWAS analysis. SNP distribution over common bermudagrass chromosomes was visualized using CMplot R package (Yin, 2020), where the bin size was set to 1 Mb to get a detailed resolution of marker distribution.

LD level in the 216 bermudagrasses was evaluated using pairwise comparisons in TASSEL 5.0 (Bradbury et al., 2007). The LD decay plot was extracted from the LD analysis (pairwise r ² values) from the TASSEL 5 across the whole genome in R. This analysis helped to define the window in which to perform a putative candidate gene search.

2.5. Population structure analysis

The population structure of the GWAS panel was determined using the Bayesian clustering approach in STRUCTURE version 2.3.4 (Pritchard et al., 2000) using filtered SNPs (21,810). The parameters to determine population structure were set as follows: the length of 10,000 burnin period, and the Markov Chain Monte Carlo number was 10,000 repetitions. The K‐value was set between 1 and 10 to generate a suitable number of subpopulations. The result of STRUCTURE were then evaluated in a web‐based program, called STRUCTURE HARVESTER (Earl & von Holdt, 2012) and the best K‐value was determined by Evanno test approach (Evanno et al., 2005).

2.6. GWAS analysis and candidate genes identification

The R package lme4 was implemented within the META‐R software (Alvarado et al., 2020) to calculate best linear unbiased estimates (BLUEs) of all the studied traits by fitting the model:

$$Y_{ij}=\mathrm{\mu }+\mathrm{Re}{\mathrm{p}}_i+\mathrm{Ge}{\mathrm{n}}_j+{\mathrm{\epsilon }}_{ij}$$

where Y_ij is the trait, μ is the overall mean, Rep _i is the effect of ith replicate, Gen _j is the effect of jth genotype, and ε_ij is the error effect related to ith replication and jth genotype. The raw phenotypic data used to calculate BLUE values are available in Tables S3 and S4. The calculated BLUE values for number of seed germinated at different time intervals, D ₅₀, GRI, and TGP were provided in Tables S5 and S6 and were subsequently used in GWAS analysis to determine MTAs using the genomic association and prediction integrated tool (GAPIT) package (Lipka et al., 2012) of R software using mixed linear model (MLM) (J. Yu et al., 2006). The MLM is the most used model for association mapping traditionally; however, the single‐locus model MLM is designated to test one marker at a time, which is more likely to identify false negatives (Wen et al., 2018). To overcome this issue, multi‐locus model Bayesian information and linkage disequilibrium‐iteratively nested keyway (BLINK) was added to detect MTAs (M. Huang et al., 2019). BLINK model does not assume a uniform distribution of causal genes across the genome, and it focuses on individual markers rather than groups of markers (bins). Studies reported that the BLINK model is the statistically most powerful and computationally efficient model, identifying more true associations than other models in GAPIT3 (M. Huang et al., 2019; Wang & Zhang, 2021). MLM and BLINK, representing traditional standard and newer statistically and computationally efficient approaches, have been used successfully in previous mapping studies (Lakkakula et al., 2025). A false discovery rate of 0.05 was used to correct for multiple testing to confirm significant MTAs (Benjamini & Hochberg, 1995). According to this threshold, SNPs with −log₁₀ (p) > 5.5 (p < 3.16E‐06) were considered statistically significant. For each of the significant quantitative trait loci (QTLs), designated QTL names were provided following the guidelines described by McIntosh et al. (2013) with some modifications in the format as QTrait.OSU.chrom.Mbps, where "QTrait" represents primary trait associated with the SNP, “OSU” represents OSU, “chrom” represent chromosomal number, and “Mbps” represents the physical position aligned to the bermudagrass reference genome.

The bermudagrass annotated reference genome (B. Zhang et al., 2022) was used for candidate gene search. The significant SNP markers consistently associated with the traits studied over 2 years were selected for this search. The genomic regions spanning ± 17 kbp flanking regions of the consistent significant QTLs were decided based on the LD decay analysis. Based on the fully annotated reference genome, the full protein sequences for each gene were searched against the comprehensive NCBI non‐redundant protein sequence database using BLASTP program from the BLAST version 2.10.1 on the OSU Pete high‐performing computer to functionally annotate bermudagrass genes. Candidate genes were then identified based on their documented roles in biological process related to germination, dormancy, seed development, or hormonal signaling in different crops.

3. RESULTS

3.1. Phenotypic variability and reliability estimates

The summary statistics for seed germination rate at different time intervals, D ₅₀, TGP, and GRI are provided in Table 1. The mean values for Ger3D, Ger3_6D, Ger6_9D, TGP, and GRI were higher in 2024 than those in 2023. The ranges were also higher for all recorded traits in the second year, except for Ger 15_18D, Ger18_21D, and D ₅₀. Among the seed germination counting days, the highest seedling counts were observed for the seed germinated between days 3 and 6 (Ger3_6D), which also exhibited the greatest variability (SD 16.5 in 2023; 16.1 in 2024).

TABLE 1.

Descriptive statistics for 10 germination traits in 216 selected bermudagrass genotypes across 2 years.

Trait	Year	Mean	Max.	Min.	SD	Skewness	Kurtosis
Ger3D	2023 2024	1.0 1.6	12.0 22.0	0.0 0.0	1.7 3.1	3.1 3.2	13.0 12.6
Ger3_6D	2023 2024	41.9 47.2	78.0 80.0	4.0 6.0	16.5 16.1	0.1 −0.3	−0.7 −0.7
Ger6_9D	2023 2024	17.8 20.4	36.0 44.0	4.0 1.0	5.8 8.8	0.2 0.0	0.1 −0.6
Ger9_12D	2023 2024	8.8 7.4	22.0 25.0	1.0 0.0	4.7 5.4	0.3 0.9	−0.4 0.4
Ger12_15D	2023 2024	4.0 2.9	16.0 17.0	0.0 0.0	2.9 3.1	1.0 1.6	1.4 2.7
Ger15_18D	2023 2024	2.3 1.3	12.0 11.0	0.0 0.0	2.0 1.8	1.3 2.1	2.8 5.8
Ger18_21D	2023 2024	1.3 0.8	6.0 5.0	0.0 0.0	1.5 1.1	1.5 1.8	1.7 3.3
D 50	2023 2024	7.9 6.5	24.0 16.0	4.0 4.0	3.3 1.8	2.2 1.6	6.0 4.1
TGP	2023 2024	77.1 81.8	94.0 98.0	20.0 16.0	11.6 10.8	−1.4 −2.1	3.4 7.5
GRI	2023 2024	10.5 11.6	16.0 17.0	2.2 2.2	2.4 2.2	−0.4 −0.6	0.2 1.2

Note: Skewness refers to the asymmetry of the data distribution; Kurtosis describes the peakedness of the data.

Abbreviations: D _50, days to 50% germination; Ger12_15D, seed germinated between day 12 to day 15; Ger15_18D, seed germinated between day 15 to day 18; Ger18_21D, seed germinated between day 18 to day 21; Ger3_6D, seed germinated between day 3 to day 6; Ger3D, seed germinated by day 3; Ger6_9D, seed germinated between day 6 to day 9; Ger9_12D, seed germinated between day 9 to day 12; GRI, germination rate index (seeds/day); SD, standard deviation; TGP, total germination percentage (%).

Estimates of variance components for the ten traits recorded in the 216 genotypes and their reliability estimates (i ²) are presented in Table 2. Genotypic variance estimates $({\mathrm{\sigma }}_G^2)$ were significant (p < 0.05 to p < 0.0001), and year variance estimates $({\mathrm{\sigma }}_Y^2)$ were significant (p < 0.01 to p < 0.0001) for all the traits. For most traits, the variance estimates for genotype × replication $({\mathrm{\sigma }}_{\mathrm{GR}}^2)$ were negative and therefore considered to be zero, whereas Ger3D showed small (0.69) but statistically significant ${\mathrm{\sigma }}_{\mathrm{GR}}^2$(p < 0.001). These negative variance components were considered as zero during reliability estimation. The variance estimates for genotype × year were significant (p < 0.0001) for Ger3D, Ger9_12D, D ₅₀, TGP, and GRI. Reliability estimates (i ²) for the recorded germination traits ranged from 0.23 to 0.82, suggesting some traits were more easily influenced by environments than others. The large genotype × year variance for TGP resulted in a low reliability estimate of 0.23. To improve seed germination, the results highlighted that Ger3_6D is a better selection trait than TGP.

TABLE 2.

Estimates of variance components and reliability estimates for tested traits in selected common bermudagrass genotypes.

Traits	Variance components						Reliability estimates (i 2)
	σ 2 G	σ 2 Y	σ 2 GY	σ 2 R	σ 2 GR	σ 2 E
Ger3D	1.65 ****	0.18 **	3.03 ****	0.34 ****	0.69 ***	3.45	0.42
Ger3_6D	200.84 ****	13.30 ****	3.65	36.48 ****	0	246.74	0.82
Ger6_9D	30.28 ****	3.01 ****	0	17.07 ****	0	130.48	0.58
Ger9_12D	14.77 ****	0.80 ****	3.98 ****	0.52 ****	0	21.52	0.73
Ger12_15D	4.90 ****	0.55 ****	1.17	0.93 ****	0	8.34	0.71
Ger15_18D	2.03 ****	0.42 ****	0	0.08 ****	0	5.19	0.70
Ger18_21D	0.62 ****	0.10 ****	0.09	0.11 ****	0	2.58	0.57
D 50	3.28 ****	0.89 ****	1.33 ****	0.45 ****	0	6.23	0.67
TGP	23.08 **	10.42 ****	80.46 ****	5.28 ****	0	70.02	0.31
GRI	2.66 ****	0.61 ****	2.18 ****	0.59 ****	0	3.23	0.62

Abbreviations: D ₅₀, days to 50% germination; Ger12_15D, seed germinated between day 12 and day 15; Ger15_18D, seed germinated between day 15 and day 18; Ger18_21D, seed germinated between day 18 and day 21; Ger3_6D, seed germinated between day 3 and day 6; Ger3D, seed germinated by day 3; Ger6_9D, seed germinated between day 6 and day 9; Ger9_12D, seed germinated between day 9 and day 12; GRI, germination rate index; TGP, total germination percentage.

^*, ^**, ^***, and ^**** denote significant difference at the p < 0.05, p < 0.01, p < 0.001, and p < 0.0001.

3.2. Correlation analysis

There was genotype × year interaction; therefore, Pearson correlation analysis was performed separately for 2023 and 2024 using phenotypic BLUEs. The correlation coefficients among the seed germination rate at different time intervals, D ₅₀, TGP, and GRI are presented in Table 3 (2023) and Table 4 (2024). Positive correlations were found between early germination traits in both years, including Ger3D with Ger3_6D (r = 0.35‐0.39, p < 0.0001), Ger3D with GRI (0.47‐0.51, p < 0.0001), Ger3_6D with GRI (0.88‐0.95, p < 0.0001), and Ger3_6D with TGP (0.42‐0.72, p < 0.0001). Early germination traits (Ger3D and Ger3_6D) were negatively correlated with germination rates at later stages than 6 days, with correlation coefficients ranging from ‐0.17 to 0.76 (p < 0.0001). D ₅₀ was negatively correlated with Ger3D, Ger3_6D, TGP, and GRI (r = ‐0.32‐0.88, p < 0.0001). After day 9, the germinated seeds showed low to strong positive correlations among themselves in both years (r = 0.38‐0.81, p < 0.0001). Mid‐to‐late germination stages showed mostly negative correlations with TGP and GRI (r = ‐0.10 to ‐0.63, p < 0.0001), which were consistent in both years. A strong correlation was found between TGP and GRI, with correlation coefficient ranging from 0.69 to 0.87 (p < 0.0001) across years.

TABLE 3.

Pearson correlation coefficients among seed germination traits in the year 2023.

Trait	Ger3D	Ger3_6D	Ger6_9D	Ger9_12D	Ger12_15D	Ger15_18D	Ger18_21D	D 50	TGP
Ger3_6D	0.39 ****
Ger6_9D	−0.33 ****	−0.25 ***
Ger9_12D	−0.47 ****	−0.69 ****	0.43 ****
Ger12_15D	−0.39 ****	−0.72 ****	0.14	0.68 ****
Ger15_18D	−0.30 ****	−0.66 ****	0.06	0.62 ****	0.73 ****
Ger18_21D	−0.17 **	−0.53 ****	−0.07	0.38 ****	0.52 ****	0.56 ****
D 50	−0.32 ****	−0.80 ****	−0.05	0.50 ****	0.65 ****	0.63 ****	0.57 ****
TGP	0.17 **	0.72 ****	0.31 ****	−0.11	−0.29 ****	−0.28 ****	−0.33 ****	−0.75 ****
GRI	0.47 ****	0.95 ****	0.01	−0.52 ****	−0.63 ****	−0.58 ****	−0.51 ****	−0.85 ****	0.88 ****

Abbreviations: D _50, days to 50% germination; Ger12_15D, seed germinated between day 12 and day 15; Ger15_18D, seed germinated between day 15 and day 18; Ger18_21D, seed germinated between day 18 and day 21; Ger3_6D, seed germinated between day 3 and day 6; Ger3D, seed germinated by day 3; Ger6_9D, seed germinated between day 6 and day 9; Ger9_12D, seed germinated between day 9 and day 12;

GRI, germination rate index; TGP, total germination percentage.

^**, ^***, and ^****denote significant difference at p < 0.01, p < 0.001, and p < 0.0001.

TABLE 4.

Pearson correlation coefficients among seed germination traits in the year 2024.

Trait	Ger3D	Ger3_6D	Ger6_9D	Ger9_12D	Ger12_15D	Ger15_18D	Ger18_21D	D 50	TGP
Ger3_6D	0.35 ****
Ger6_9D	−0.55 ****	−0.52 ****
Ger9_12D	−0.44 ****	−0.76 ****	0.61 ****
Ger12_15D	−0.34 ****	−0.73 ****	0.43 ****	0.81 ****
Ger15_18D	−0.26 ****	−0.65 ****	0.31 ****	0.71 ****	0.76 ****
Ger18_21D	−0.29 ****	−0.61 ****	0.31 ****	0.65 ****	0.68 ****	0.71 ****
D 50	−0.43 ****	−0.88 ****	0.37 ****	0.74 ****	0.76 ****	0.69 ***	0.67 ****
TGP	−0.01	0.42 ****	0.37 ****	0.13	0.04	0.00	−0.02	−0.43 ****
GRI	0.51 ****	0.88 ****	−0.27 ****	−0.54 ****	−0.55 ****	−0.50 ****	−0.49 ****	−0.88 ****	0.70 ****

Abbreviations: D ₅₀, days to 50% germination; Ger12_15D, seed germinated between day 12 and day 15; Ger15_18D, seed germinated between day 15 and day 18; Ger18_21D, seed germinated between day 18 and day 21; Ger3_6D, seed germinated between day 3 and day 6; Ger3D, seed germinated by day 3; Ger6_9D, seed germinated between day 6 and day 9; Ger9_12D, seed germinated between day 9 and day 12; GRI, germination rate index; TGP, total germination percentage.

^**, ^***, and ^**** denote significant difference at p < 0.01, p < 0.001, and p < 0.0001.

3.3. GBS and SNP calling

The initial sequencing run produced 1,937,123,132 raw reads in total. After applying quality control measures, a total of 968,561,566 reads (50.0%) were retained as high‐quality sequences. The number of good‐quality reads per sample ranged from 2.66 million to 13.27 million, with an average of 4.26 million raw reads per sample. This high‐quality dataset generated a total of 292,506 Mbp of sequence data. After trimming and SNP calling, a total of 446,652 SNPs were generated, and 21,810 high‐quality SNP markers were retained and utilized for further analysis after stringent SNP filtering.

3.4. Population structure analysis

The STRUCTURE result based on the high‐quality SNP markers illustrated that the GWAS panel of 216 bermudagrass genotypes was diverse, with multiple clusters represented by different colors (Figure 1a). The Evanno method showed the highest ΔK at K = 5, indicating that there were five subpopulations in the panel (Figure 1b). According to Kinship matrix (Figure 1c), the genotypes in the panel were clearly stratified into five different clusters. The MLM and BLINK models automatically constructed Q matrix (five PCs) and K matrix in GAPIT to correct for population stratification and kinship. The principal component analysis (PCA) demonstrated that the first three principal components collectively accounted for 23.22% of the total genetic variance (Figure 1d). Notably, PC1 explained 10.1% of the variation alone, the most significant component, suggesting a significant contribution in the population. Upon closer investigation of 3D PCA biplot (Figure 1d), distinct separation and clustering of genotypes were found, indicating there are distinct genetic groups in the population.

FIGURE 1.

Population structure and genetic similarity analysis of 216 common bermudagrass genotypes. (a) Population structure of 216 bermudagrass individuals; (b) Magnitude of ΔK determined by the Evanno method depicting the population stratified into five clusters; (c) Kinship matrix depicting the relationship among bermudagrass individuals; (d) The first three principal components explain about 23.22% of genetic variation in the panel.

3.5. SNP alignment, density plot, and LD analyses

The physical locations in chromosomes for the high‐quality SNP markers were obtained from aligning with the reference genome. Chromosome 2B accumulated the highest number of SNP markers (2555), whereas the least number of SNP makers (292) were aligned to chromosome 1A (Table S7). The retained high‐quality SNP markers showed a broad distribution across the bermudagrass chromosomes, with higher marker densities toward the distal ends of chromosome arms (Figure 2a). Our LD analyses exhibited that LD curve dropped below the selected threshold of r ² = 0.1, and LD decay point across the whole genome was 17,354 bp (Figure 2b).

FIGURE 2.

Marker density and linkage disequilibrium (LD) decay analysis of the common bermudagrass genome. (a) Distribution of 21,810 single‐nucleotide polymorphism (SNP) markers across chromosomes; (b) LD decay plot showing average LD decay distance at 17,354 bp across the whole bermudagrass genome.

3.6. GWAS analysis

Since the significant genotype × year interaction was observed, we conducted GWAS analysis for 2023 and 2024, 2 years separately. The GWAS analysis across the 2 years identified a total of 52 QTLs (Table S8). In 2023, a total of 32 QTLs were identified, of which six QTLs were identified by MLM and 25 QTLs by BLINK. One QTL (QTGP.OSU.2B.9) was identified by both models in 2023. Similarly in 2024, 26 QTLs were identified, of which the MLM identified four QTL, whereas the BLINK identified 19 QTL. In 2024, three QTLs (QTGP.OSU.2A.5, QGer.OSU.4B.4, and QTGP.OSU.2A.5) were identified by both models. Notably, six QTLs (QGer.OSU.4B.4, QTGP.OSU.2A.5, QTGP.OSU.2A.7, QTGP.OSU2B.9, QTGP.OSU.8A.4, and QGerOSU.4A.13) were consistently significant across 2 years for different traits (Table 5). All MTAs, designated QTL, and test statistics are listed in Table S8.

TABLE 5.

Summary of consistent significant single‐nucleotide polymorphism (SNP) markers associated with seed germination traits.

SNP	Designated QTL	Chr	Position (bp)	Year	Traits	Model	‐log10(p value)	PVE (%)
TKT2d53375	QGer.OSU.4B.4	4B	4,380,918	2023	Ger15_18D	MLM	6.68	11.85
				2024	Ger12_15D, Ger15_18D	MLM + BLINK	5.96–9.67	9.51–53.05
TYCafdb47b	QTGP.OSU.2A.5	2A	4,673,066	2023	TGP	MLM	7.22	9.81
				2024	TGP	MLM + BLINK	8.92–15.04	1.43–18.23
CSTc98f92ae	QTGP.OSU.2A.7	2A	6,843,760	2023	TGP	MLM	7.22	9.81
				2024	TGP	MLM	8.92	18.23
GRT89691388	QTGP.OSU.2B.9	2B	8,925,345	2023	TGP	MLM + BLINK	6.87–14.93	11.71–13.75
				2024	TGP	MLM	6.87	13.41
CMA78f76704	QTGP.OSU.8A.4	8A	4,446,764	2023	TGP	MLM	7.22	9.81
				2024	TGP	MLM	8.92	18.23
GRCf8290962	QGRI.OSU.4A.13	4A	12,819,722	2023	Ger3_6D, GRI	BLINK	6.77–7.11	7.57–9.73
				2024	Ger15_18D, Ger18_21D, GRI, D 50	BLINK	6.17–10.42	6.29–58.25

Abbreviations: BLINK, Bayesian information and linkage‐disequilibrium iteratively nested keyway; Chr, chromosome; D ₅₀, days to 50% germination; Ger12_15D, seed germinated between day 12 and day 15; Ger15_18D, seed germinated between day 15 and day 18; Ger3_6D, seed germinated between day 3 and day 6; GRI, germination rate index; MLM, mixed linear model; PVE, phenotypic variance explained by SNP; QTL, quantitative trait loci.

3.6.1. Seed germination MTAs at different time intervals

A total of 40 MTAs were identified for the number of germinated seeds recorded at different time intervals, days 0–21, across 2 years (Table S8). For early germination (Ger3D and Ger3_6D), a total of 27 MTAs (two by MLM and 25 by BLINK) were detected across 2 years. In 2023, the MLM identified two MTAs, while BLINK identified nine MTAs for Ger 3D. Notably, QTL, QGer.OSU.4B.6, associated with Ger3D identified by BLINK in 2023, showed the highest effect explaining 39.33% of the phenotypic variance. In 2024, the BLINK identified eight MTAs for Ger3D, in which QTL, QGer.OSU.7A.21, accounted for highest effect of 33.49% of phenotypic variance. For Ger3_6D, BLINK identified a total of eight MTAs in 2 years. Among these, QGer.OSU.8B.19, identified in 2024, accounted for the highest phenotypic variance explained (PVE) value of 33.85%. QTL, QGer.OSU.4A.13, associated with Ger3_6D, was also detected for GRI, D ₅₀, Ger15_18D and Ger18_21D across years.

For mid‐ to late‐phase germination, 10 QTL (one by MLM and nine by BLINK) were detected in 2023, while six QTLs (two by both MLM and BLINK and an additional four by BLINK only) were detected in 2024. Notably, QTL, QGer.OSU.4A.13, was consistently detected across years associated with Ger3_6D, Ger15_18D, Ger18_21D, GRI, and D ₅₀, explaining up to 58.26% of phenotypic variance. For Ger15_18D, QGer.OSU.4B.4 was identified in both years: firstly in 2023 by the MLM and again in 2024 by both the MLM and the BLINK, explaining up to 53.05% phenotypic variance (Figure 3c,d). Similarly, in 2024, another QTL, QGer.OSU.3B.13, was identified by both MLM and BLINK. For the latest phase (Ger18_21D) of germination, six QTLs were identified in 2023 and one in 2024, respectively, by the BLINK, with phenotypic variance explained up to 58.25%.

FIGURE 3.

Manhattan and quantile‐quantile (Q–Q) plots demonstrating consistent single‐nucleotide polymorphisms (SNPs) over 2023 and 2024 for different studied traits. The Y‐axis shows the −log₁₀(p) values, and the X‐axis shows the chromosomal positions of markers. The solid horizontal middle line is the threshold line of −log₁₀(p) = 5.5. Q–Q plots show the observed and expected p‐values. (a) Manhattan and Q–Q plots demonstrated four significant SNPs associated with total germination percentage (TGP) in 2023 (mixed linear model [MLM]); (b) Manhattan and Q–Q plots demonstrated four significant SNPs associated with TGP in 2024 (MLM); (c) Manhattan and Q–Q plots demonstrated one significant single‐nucleotide polymorphism (SNP) associated with seed germination between day 15 and day 18 in 2023 (MLM); (d) Manhattan and Q–Q plots demonstrated one significant SNP associated with seed germination between day 15 and day 18 in 2024 (MLM).

3.6.2. Days to 50% germination

In 2023, the BLINK model identified a single QTL, QD50.OSU.2A.15, which explained 67.16% of the phenotypic variance (Table S8). Similarly, in 2024, three QTLs were identified by the BLINK model, explaining up to 28.06% of the phenotypic variance. Notably, QTL, QGRI.OSU.4A.13, associated with D ₅₀, was also identified for early‐ and mid‐phases germination rate and GRI across years.

3.6.3. Germination rate index

GWAS analysis identified two QTLs associated with GRI (QGRI.OSU.4A.13 and QGRI.OSU.9B.1) under the BLINK model in the year 2023, explaining up to 18.61% of the phenotypic variance. In 2024, the BLINK model identified two QTLs (QGRI.OSU.2A.8 and QGRI.OSU.4A.13), which explained 12.94% and 23.39% of the phenotypic variance. QGRI.OSU.4A.13 was consistently detected across years for GRI (Figure S1) and was previously detected for multiple germination traits, including D ₅₀.

3.6.4. Total germination percentage

The GWAS analysis of TGP detected a total of seven QTLs, among, which four QTLs were consistently significant across both years (Figure 3a,b). All these four QTLs, QTGP.OSU.2B.9, QTGP.OSU.8A.4, QTGP.OSU.2A.7, and QTGP.OSU.2A.5, were identified by MLM model across 2 years. These QTLs accounted for moderate level of phenotypic variance for TGP, ranging from 9.81% to 18.23%, suggesting their consistent control of TGP across environments. Additional QTL, QTGP.OSU.4B.22, identified in 2023 under BLINK model, accounted for PVE value of 78.52%. In 2024, both the MLM and BLINK identified a distinct additional QTL on chromosome 2B, QTGP.OSU.2B.8 and QTGP.OSU.2B.14, respectively.

3.6.5. Candidate gene identification

Candidate gene identification was performed for six MTAs, which were consistent across years. For each of these MTAs, all genes located within a 17 kb upstream and downstream interval based on LD decay were extracted from the bermudagrass reference genome sequence (B. Zhang et al., 2022). Twenty candidate genes were identified, including genes that were reported to be related to hormonal regulation, dormancy regulation, calcium signaling, cell wall modification, and carbohydrate metabolism (Table 6).

TABLE 6.

Candidate genes associated with seed germination traits within consistent quantitative trait loci (QTL) regions.

Designated QTL	Chr	Physical interval (Mbp)	Query gene	E‐value	Annotation
QTGP.OSU.2A.5	2A	4.656–4.690	Cd2A1G005560.1	0	CBL‐interacting protein kinase 19
			Cd2A1G005580.1	1.73E‐45	Protein DOG1‐like 1
QGer.OSU.4B.4	4B	4.364–4.398	Cd4B1G004940.1	8.52E‐68	Blue copper protein precursor
			Cd4B1G004970.2	0	Cyclic nucleotide‐gated ion channel 2‐like
			Cd4B1G004980.2	0	Alpha‐1,4‐glucan phosphorylase
			Cd4B1G004990.1	1.30E‐129	EcWRKY‐16, partial
QTGP.OSU.2B.9	2B	8.908–8.943	Cd2B1G010520.1	6.80E‐125	Calcyclin‐binding protein
			Cd2B1G010530.3	9.00E‐162	Katanin p60 ATPase‐containing subunit A‐like 2
			Cd2B1G010540.1	0	Gibberellin 2‐beta‐dioxygenase 3
			Cd2B1G010550.1	0	Potassium channel KAT3‐like
			Cd2B1G010560.2	0	Protein BIIDXI‐like
QTGP.OSU.8A.4	8A	4.429–4.464	Cd8A1G005450.1	1.72E‐63	TIM23‐3‐like
			Cd8A1G005460.3	0	Cellulose synthase‐like protein H1 isoform X1
QTGP.OSU.2A.7	2A	6.826–6.861	Cd2A1G008250.2	0	Pectinesterase isoform X1
			Cd2A1G008260.3	0	Probable glutamate carboxypeptidase LAMP1
			Cd2A1G008270.2	0	Tryptophan–tRNA ligase
			Cd2A1G008280.2	0	NADP‐dependent malic enzyme‐like
			Cd2A1G008290.2	0	Protein cereblon homolog isoform X3
			Cd2A1G008300.2	0	NADP‐dependent malic enzyme‐like
			Cd2A1G008310.1	3.02E‐55	Histidine‐containing phosphotransfer protein 3
QGRI.OSU.4A.13	4A	12.803–12.837	Cd4A1G012670.1	0	Glutamyl‐tRNA (Gln) amidotransferase subunit A
			Cd4A1G012680.1	0	Subtilisin‐like protease SBT1.7

Abbreviations: CBL, calcineurin B‐like; Chr, chromosome; DOG1, delay of germination 1; TIM, translocase of the inner membrane.

4. DISCUSSION

Establishment of bermudagrass by seed provides an effective solution for consumers as it is relatively inexpensive, poses a low risk of pathogens, is easy to transport, and requires minimal soil preparation (Patton et al., 2006). Rapid and high germination percentages are key desirable traits for seeded cultivars. In this study, total seed germination percentage among the genotypes varied from 20% to 94% in 2023 and 16% to 98% in 2024, showing much larger ranges than previously reported under similar germination temperatures (Munshaw et al., 2014). Although the overall range was large, only five genotypes in 2023 and four genotypes in 2024 exhibited a germination percentage below 50%. In 2023, 111 genotypes (51.4%) and in 2024, 65 genotypes (30.1%) had a germination percentage lower than 80%, which is the minimum total germination required for Certified class seed in bermudagrass (Oklahoma Crop Improvement Association, 2025). ANOVA indicated that the genotype effect is significant for each of the 10 germination‐related traits. Reliability estimates for the traits showed substantial variation, with TGP showing lowest reliability (31%) and Ger3_6D showing the highest (82%). Notably, D ₅₀ showed high reliability (67%) and exhibited negatively high correlation with GRI and TGP, suggesting that genotypes reaching 50% germination earlier tend to have faster and overall high germination performance. The findings suggest that using germination rate at 6 days and D ₅₀ are more reliable and informative to select fast and high germination seeded common bermudagrass than TGP.

GWAS for complex traits requires a high marker density to ensure adequate genome coverage. Previous GWAS studies in grass species have commonly utilized tens of thousands of SNPs to get good mapping resolution, such as 21,648 SNPs reported by Jaškūnė et al. (2020), 150,325 SNPs by Pantaliao et al. (2016), and 13,230 SNPs by Bushman et al. (2024). To get comparable marker density, we tested different minimum read depth thresholds during SNP filtering. Higher read depth threshold (≥5X) substantially reduced the number of retained SNPs, resulting in very low marker density for GWAS. But a minimum read depth of ≥3X retained enough SNPs, ensuring good genome coverage, and was therefore utilized in this study. Although higher read depth thresholds improve genotype accuracy, especially in polyploid plant species with high degrees of heterozygosity, they substantially reduce marker density. The limitations associated with low read depth are allelic dropout and the miscalling of heterozygotes (Cooke et al., 2016). In our evaluation, average heterozygosity increased from 0.35 at a ≥3X read depth threshold to 0.55 at a ≥5X, indicating improved heterozygote detection but at the expense of number of retained SNPs. In addition to heterozygosity miscalling, the low read depth in this study can also affect correct estimation of LD decay rates, likely resulting in rapid LD decay, which is not true biologically (Bilton et al., 2018). Therefore, the relatively rapid LD decay in this study (∼17 kb) should be interpreted cautiously, as it might reflect genotypic uncertainty due to low read depth threshold. Overall, the ≥3X read depth used in this study represents a compromise between a marker density comparable to previous GWAS studies and minimizing genotyping errors.

Several studies have performed genetic dissection of seed germination and related traits in grass species through QTL mapping and GWAS. The genetic basis of germination was studied in Oryza sativa, where seven QTLs associated with germination percentage and putative genes encoding hydrolytic enzymes and kinases crucial for germination were mapped successfully (Panahabadi et al., 2022). Chromosomes harboring QTL for germination traits between rice and common bermudagrass were compared based on their chromosome correspondence reported by Fang et al. (2020). Chromosomes 4A and 4B carrying two consistent QTL in common bermudagrass correspond to chromosome 3 harboring one QTL in rice. Similarly, common bermudagrass 2A and 2B chromosomes harboring three QTLs correspond to chromosome 6 of rice, which carries one QTL. But chromosome 8A of common bermudagrass, in which one consistent QTL was identified, corresponds to rice chromosome 11, which has no reported QTL related to seed germination. BLAST alignment of protein sequences from candidate genes nearby consistent QTL from this study identified homologous genes in rice associated with seed germination and dormancy, highlighting functional conservation of genes. However, no clear positional orthology was identified between QTL identified in this study and those reported in the study of rice. Similarly, QTL associated with germination were mapped to chromosomes 2, 4, and 3 in Hordeum vulgare (Edney & Mather, 2004). Evidence from these crops supports that germination related traits are under polygenic control while also strongly influenced by the environment (Koller & Kozlowski, 1972). In common bermudagrass, genetic studies have been primarily targeted on agronomic traits like establishment rate (Guo et al., 2017), winter hardiness and drought resistance (S. Yu et al., 2025), morphological traits (Khanal et al., 2019), and cadmium tolerance (Xie et al., 2015). To our knowledge, this is the first genetic dissection study targeting seed germination traits in common bermudagrass. Building upon genetic studies on related species, we implemented a comprehensive GWAS approach that identified 52 QTLs associated with studied seed germination traits by utilizing 216 bermudagrass genotypes across 2 years. Notably, six markers appeared consistently across the years, highlighting their stability and biological relevance. Among these stable markers, five were detected by MLM model and four by BLINK. This only a few consistent QTLs across years highlighted the high environmental sensitivity of seed germination traits. Similar findings have been reported in other crops like Coffea arabica, where differences in moisture level, seed surface area to volume ratio, and environment during seed harvesting were shown to influence overall germination performance across years (Nasiro et al., 2017). For early germination rate and D ₅₀, four QTLs each on chromosomes 2A and 4A were discovered, indicating that these chromosomes harbor key regions in regulating early stage of germination. The stability and strength of QTL in these chromosomes make them candidate regions for use in marker‐assisted selection (MAS) to improve germination rapidity in bermudagrass. Additionally, QTL QGer.OSU.4B.4 was consistently associated with Ger15_18D across both years and both models. Interestingly, the same QTL was also identified for Ger12_15D, suggesting it as a major locus in regulating multiple phases of the germination process in bermudagrass. This locus potentially points to a central regulator of mid and late phase of germination. Similar phase‐specific regulatory patterns have been reported in Arabidopsis thaliana, where particular loci govern early, mid, and late phase of germination (Holdsworth et al., 2008; Nakabayashi et al., 2005). Additionally, seven QTLs were identified linked to TGP, with four QTL—QTGP.OSU.2A.5, QTGP.OSU.2A.7, QTGP.OSU.2B.9, and QTGP.OSU.8A.4—consistently present across both years and models. The strong and consistent associations of these MTAs suggest that they are important in controlling the overall germination success in common bermudagrass, making them valuable genomic regions to utilize during bermudagrass breeding in improving germination efficiency. QGRI.OSU.4A.13 was noteworthy because this QTL was associated with multiple phases germination, D ₅₀, and GRI across years, which suggests that this locus has a role in regulating both the overall success and speed of germination. To utilize these identified QTL, genetic markers can be designed and validated under the field conditions.

The candidate gene search identified several putative genes that regulate seed germination and other seed‐related traits. While examining genetic basis for germination, Panahabadi et al. (2022) in rice and Zuo et al. (2019) in wheat reported the underlying physiological mechanisms for seed germination, such as ABA‐mediated dormancy, gibberellin catabolism, and seed coat development, which align well with the findings in our current study. The delay of germination 1 (DOG1)‐like 1 gene is an important one. In Arabidopsis thaliana, DOG1‐like genes in clades 1–4 maintain seed dormancy function while regulating ABA sensitivity (Ashikawa et al., 2013). Ashikawa et al. (2010) reported that although the amino acid sequence identity is lower than Arabidopsis DOG1, DOG1‐like gene has the same dormancy function due to the conserved domains. Similarly, second candidate gene annotated as gibberellin 2‐beta‐dioxygenase 3 (GA2ox3) is also crucial. GA2ox3 catalyzes the deactivation GA and regulate GA‐ABA levels, which is important in controlling seed dormancy and hypocotyl elongation in barley (Gubler et al., 2008). Park (2015) reported the presence of elevated expression of GA2ox3 in the embryo of non‐dormant barley grains, which reflects their roles in regulating bioactive GA level and thereby germination. Another candidate gene, calcineurin B‐like‐interacting protein kinase‐19 (CIPK19), supports the regulation of Ca²⁺‐mediated signal transduction in bermudagrass germination. Overexpression of TaCIPK19 significantly increased germination rate in Triticum aestivum, particularly in drought conditions (Y. N. Wu et al., 2023). Since CIPK19 is required for pollen tube extension and polar cell growth in Arabidopsis according to Zhou et al. (2015), it is likely that CIPK19 gene plays an analogous role and regulates radicle protrusion in bermudagrass seed. Alpha‐1,4‐glucan phosphorylase and cellulose synthase‐like H1 are two additional candidate genes with roles in cell wall modification and carbohydrate metabolism that may control endosperm weakening and energy metabolism, which are critical in seed germination (Bao et al., 2025; Cuesta‐Seijo et al., 2017).

The consistent QTL and relevant candidate genes identified in this study offer promising sources for MAS to accelerate the development of rapid and uniformly germinating bermudagrass cultivars. Future research like allele‐specific marker development and validation is needed to confirm the presence of these loci in diverse bermudagrass populations. Additionally, this study provided foundation for future gene‐expression studies or gene editing in common bermudagrass to elucidate the molecular mechanisms controlling seed germination.

5. CONCLUSION

Seeded‐type bermudagrass is a viable option for establishing plantings for turf, forage, soil erosion control, and other uses. Using a diverse panel, this study revealed large variations in seed germination traits, which were significantly related to genotypes and influenced by environmental interaction with varying magnitude. Germination rates from day 3 to 9 and days to 50% germination were important in selecting and improving total seed germination percentage due to their medium to high reliability estimates and significant correlation with TGP. TGP itself, however, had a low reliability estimate, suggesting its limited value in improving seed germination in common bermudagrass. It was identified that 52 SNPs are significantly associated with seed germination traits through GWAS in the bermudagrass panel. Candidate genes identified from consistent QTL regions may have roles in controlling seed germination traits. The significant and consistent SNPs have potential to be utilized in developing PCR‐based markers for breeding new seeded‐type bermudagrass cultivars with improved seed germination traits.

AUTHOR CONTRIBUTIONS

Bigul Thapa Magar: Conceptualization; data curation; formal analysis; investigation; methodology; software; writing—original draft; writing—review and editing. Shuhao Yu: Formal analysis; software; supervision; writing—review and editing. Mingying Xiang: Supervision; writing—review and editing. Million Tadege: Supervision; writing—review and editing. Yanqi Wu: Conceptualization; formal analysis; funding acquisition; investigation; methodology; project administration; resources; supervision; writing—review and editing.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Figure S1. Manhattan and quantile‐quantile (QQ) plots demonstrating a consistent SNP across years associated for GRI.

Figure S2. Manhattan and quantile‐quantile (QQ) plots demonstrating significant SNPs in the year 2023 associated for: (a) Ger3D (BLINK) (b) Ger3_6D (BLINK) (c) Ger6_9D (BLINK) (d) Ger9_12D (BLINK) (e) Ger18_21D (BLINK) (f) D₅₀ (BLINK) (g)TGP (BLINK) (h) Ger3D (MLM).

Figure S3. Manhattan and quantile‐quantile (QQ) plots demonstrating significant SNPs in the year 2024 associated for: (a) Ger3D (BLINK) (b) Ger3_6D (BLINK) (c) Ger6_9D (d) Ger12_15D (BLINK) (e) Ger15_18D (BLINK) (f) Ger18_21D (BLINK) (g) D₅₀ (BLINK) (h) TGP (BLINK).

Supplemental Table S1. List of common bermudagrass genotypes used in the GWAS panel.

Supplemental Table S2

Supplemental Table S3. Raw phenotypic data for seed germination traits recorded in 2023 across bermudagrass genotypes.

Supplemental Table S4. Raw phenotypic data for seed germination traits recorded in 2024 across bermudagrass genotypes.

Supplemental Table S5. Phenotypic BLUE values for seed germination traits recorded in 2023 across bermudagrass genotypes.

Supplemental Table S6. Phenotypic BLUE values for seed germination traits recorded in 2024 across bermudagrass genotypes.

Supplemental Table S7. Distribution of high‐quality SNPs across common bermudagrass's 18 chromosomes.

Supplemental Table S8. Significant SNP markers associated with studied germination related traits.

Thapa Magar, B. , Yu, S. , Xiang, M. , Tadege, M. , & Wu, Y. (2026). Unveiling the genetic determinants of germination efficiency in common bermudagrass: A genome‐wide association study. The Plant Genome, 19, e70219. 10.1002/tpg2.70219

DATA AVAILABILITY STATEMENT

All phenotypic and genotypic data supporting the findings during this study are included in this article and its supplementary information files.