Introduction

White clover (Trifolium repens L.) is a perennial forage legume that originated from two diploid species and considered an allotetraploid (2n = 4x = 32)1. It is a common component of cultivated pastures in temperate regions globally and represents the most agronomically important species within the Trifolium genus, which comprises nearly 300 species2. Native to the Mediterranean region, specifically southeastern Europe and southern Asia, it is rich in diverse nutrients and minerals, conferring high nutritional, ecological, genetic, breeding, and medicinal value 3–7.

White clover exhibits vigorous growth when mixed with forages of the family Gramineae7, and exhibits excellent resistance to cold, drought, trampling, and weed infestation, traits of importance for cultivar improvement and breeding 7,8,9,10,11,12.

White clover serves as a model system for the study of hydrogen cyanide (HCN) defense evolution13. Cyanogenic polymorphism in some plant species used in agriculture is of interest to breeders because the yield and persistence of the plant are associated with moderate levels of cyanogenesis14. Furthermore, it is also of interest due to potential toxicity in ruminants, thus representing an important descriptor suggested by the International Board for Plant Genetic Resources (IBPGR)15 for white clover. The production of cyanide (HCN) is due to the presence or absence in the plant of two compounds: the enzyme linamarase and the glycoside linamarin. The enzyme linamarase is regulated by the Li locus, resulting in two phenotypes: a dominant one that produces the enzyme and a recessive one. Secondly, the presence or absence of the glycoside linamarin depends on the Ac locus in its dominant form. The presence of both genes in a state of dominance (Ac-Li) results in a cyanogenic phenotype. There are three classes of acyanogenic phenotypes for each plant: a) absence of cyanoglucosides (ac-Li); b) absence of linamarase (Ac-li); and c) absence of both (ac-li). The presence or absence of both compounds affects ruminants, as their rumen contains the enzyme β-glucosidase, which is capable of also hydrolyzing the glycoside and producing cyanide in the rumen, even in the absence of the linamarase enzyme necessary to produce the reaction in the plant itself. Therefore, even if the enzyme is not present in the plant, the reaction can occur in the animal if it has ingested plants with cyanogenic glycoside content.

Both the enzyme and the glycoside are physically separated within the cell: linamarase is located in the walls of the epidermal cells of the leaves, while the glycosides are contained in vacuoles16,17. For the chemical reaction to occur and HCN to be released, the enzyme and substrate must come into contact. This occurs in the plant itself in nature when it suffers some kind of physical damage, which can be caused by biotic factors (e.g., herbivores, parasites) or abiotic factors (e.g., frost, mechanical injury)18. The detection of a cyanogenic phenotype Ac/li is significant because it can result in the release of HCN in ruminants, as they possess the enzyme β-glucosidase.

Morphological and agronomic characters have been widely used for decades to detect genetic variation. These are routine methods with results that are strongly influenced by environmental factors. Molecular markers analysis offers an efficient alternative to this approach19, providing more precise and comprehensive information. The polyploidy and genetic diversity of white clover have hindered rapid progress in genetic and genomic studies12. Compared to related species, the structural and genetic information of the withe clover is limited, especially at the genomic level7. To improve breeding, molecular genetic studies have been conducted. Different molecular markers have been tested, especially microsatellites (SSRs), to understand the genetic variation and polymorphisms existing in clovers2,19,20,21,22,23,24,25,26.

The CIAM-AGACAL’s germplasm bank has one of the most cultivated forage legumes in the world27. The goal of the present study was to evaluate the diversity among cultivars and natural clover populations at morphological and molecular level. A previous study on 13 out of 57 populations of red clover (T. pratense Lam.) conserved at CIAM and 4 commercial cultivars were evaluated by using 12 polymorphic SSR loci and agromorfological traits, which revealed 241 microsatellite alleles with an average of 20.08 alleles per locus25. Two main groups were detected by the Structure software, one of them including local populations and the second clustering cultivars and related populations. Intra-specific variability was found among cultivars and natural populations. A moderate genetic differentiation of Spanish red clover cultivars was observed (Fst = 0.08) between the two main clusters. Finally, genetic and phenotypic variation was related, with the most differentiated cluster having lower growth in the year of sowing and growth habit in early spring before flowering and higher tolerance to pests and diseases.

The first linkage map was constructed with 78 SSR and 57 AFLP markers using an F2 mapping population cross between parental inbred lines. Subsequently, a linkage map suggesting homeologous linkage groups was reported, consisting of 493 SSR loci on 16 homeologous linkage groups spanning 1144 cM12.

The objective of this research is to provide relevant information of the phytogenetic resources of white clover conserved in the Germplasm Bank of the Agricultural Research Centre of Mabegondo (CIAM, Xunta de Galicia) by evaluating cyanogenesis (the ability to produce hydrogen cyanide after tissue damage), agromorphological traits, and molecular markers in 15 local populations and cultivars.

Material and methods

Plant material

The study included a total of 300 samples, which were assessed using 14 SSRs. These samples represented 12 local populations, an experimental selection conducted by CIAM not previously reported (Var.Exp.Tri4, 9079), and two commercially available cultivars (‘Huia’ and ‘California’) (Table 1). Collection of the samples took place in northwestern Spain, specifically in the regions of Asturias, León, and Galicia, between the years 2003 and 2015. The samples were collected at various altitudes ranging from 50 to 1500 m above sea level and are currently conserved in the germplasm bank of the CIAM (Xunta de Galicia). Plants were sown in August 2020 in the greenhouse of the CIAM (A Coruña), in norhwestern Spain (43° 14′ 31.88’’N, 8° 16′ 0.359’’W). The number of samples by local population/ecotype or cultivar amounted to 30 in total (20 for genetic studies and 30 for agromorphological traits).

Table 1 Local populations and cultivars of white clover (Trifolium repens L.) evaluated with 14 SSRs and 13 phenotypic traits.
Full size table

SSRs

The extraction of DNA was conducted using a 0.5–0.75 g sample of young leaves with the "E.Z.N.A.® Plant DNA Kit" from OMEGA Bio-Tek Inc. (Norcross, GA, USA) and the "DNeasy® Plant Mini Kit" from Qiagen (Hilden, Germany). The concentration of genomic DNA was determined using a NanodropTM ND-1000 Spectrophotometer from Thermo Scientific (Wilmington, DE, USA) and diluted to a concentration of 20 ng/μL.

Fourteen SSRs (Simple Sequence Repeats) were chosen for the analyses, based on previous studies25,26,28 (Table S1). The amplification of these fourteen SSRs was performed using three multiplexed PCR reactions, with each reaction using primers labelled with FAM, NED, PET, or VIC fluorophores from PE Applied Biosystems (Warrington, UK). The PCR conditions included an initial denaturation at 94 °C for 5 min, followed by 35 cycles of denaturation at 95 °C for 30 s, annealing at a specific temperature depending on the multiplex set for 90 s, extension at 72 °C for 1 min, and a final extension at 60 °C for 30 min.

The amplification products were diluted and mixed with water. Subsequently, 2 µL of the diluted amplification product was combined with 0.12 µL of 600LIZ size standard (Applied Biosystems, Foster City, CA, USA) and 9.88 µL of formamide. The sizes of the alleles were determined using Peak ScannerTM software from Applied Biosystems.

A Bayesian analysis was conducted using Structure software29,30. The admixture model with unlinked loci and correlated allele frequencies, as defined in Pereira-Lorenzo et al. and Porras-Hurtado et al.31,32, was used to estimate the ancestry membership proportions of the population. A minimum of 20 iterations (30 in this study) were recommended to estimate the proportions accurately. The analysis involved computing unknown reconstructed panmictic populations (RPPs) of genotypes from K = 1 to 15, using the options popinfo = 0 and popflag = 0. These options assumed that the sampled genotypes had unidentified origins and assigned them probabilistically to RPPs based on a qI (probability of membership) of 80%, while a lower probability indicated an admixed genotype. The second-order change of the likelihood function, divided by the standard deviation of the likelihood (ΔK), was also estimated to obtain the best K value supported by the data33. Structure Harvester34 was used for this analysis. Using GenoDive 2.0b.23 software35, pairwise differentiation (Fst-values) was estimated between the reconstructed panmictic populations obtained by Structure.

For each of the 300 unique ecotypes, the frequency of each allele was designated as a variable, with a value of 1 indicating the presence and 0 indicating the absence of the allele. To analyze the unique genotypes of landraces and cultivars, DARwin software version 6.0.010 was utilized to construct a dissimilarity tree, employing the “Dissimilarity for Allelic Data” option. This was followed by the application of the Unweighted Neighbor-Joining method to generate a comprehensive tree encompassing all genotypes. Graphical representations of the PCs included the genetic structure and type of material evaluated with SSRs.

Agromorphological traits including cyanogenesis

In greenhouse conditions, seed trays containing 12 natural populations, two commercial varieties, and an experimental variety obtained at CIAM from Galician ecotypes capable of producing leaves with between 3 and 7 leaflets were sown36. The passport data are shown in Table 1. When the plants were grown, leaf samples were taken for genomic and cyanogenic analysis. Subsequently, individual plants were transplanted to the CIAM field in April 2021 following an experimental design of three complete random blocks using 30 individuals per population for data collection, following the field methodology used by Oliveira et al.37 for ryegrasses and using the descriptors of IBPGR15 for white clover. The edaphoclimatic characteristics for the CIAM field were: soil fertility: 9.845% Aluminum saturation (Alsat) in soil and assimilable elements: 68.65 mg/L Phosphorus (P) and 433.5 mg/L Potassium (K); Mean temperature (T): 12.1 º C; Mean relative humidity (RH) at 1.5 m: 59%; Precipitation (P): 82.1 mm; pH: 5.35.

The variables studied were as follows (in alphabetical order): AFC: Width of the central leaflet (mm), AIN: Abundance of inflorescences (visual scale from 1 = low, 5 = high), ALT: Maximum plant height (cm), ANC: Maximum plant width (cm), CRI: Winter growth (visual scale from 1 = low, 5 = high), CRP: Spring growth (visual scale from 1 = low, 5 = high), ENF: Disease sensitivity (visual scale from 1 = low, 5 = high), FLO: Flowering date (number of days counted from planting date), GRE: Stolon thickness (mm), GRP: Petiole thickness (mm), LFC: Length of the central leaflet (mm), LOP: Length of the petiole (cm), and CP: Cyanogenic potential. Additionally, we recorded the altitude (AL) of the population’s origin in meters above sea level (m.a.s.l.).

For the study of cyanogenesis, plants were selected from each population and cultivar in the greenhouse, and 3 to 4 leaves were analyzed. Three determinations were made. The detection of HCN was carried out following the sodium picrate coloration method proposed by Corkill38 and later modified by Oliveira et al.37. In each test tube, 750 µl of distilled water and 200 µl of toluene were added to the fresh tissue to destroy the membranes and thus bring the enzyme and substrate into contact. The expression of cyanogenesis in white clover shows quantitative variation due to the homozygous or heterozygous condition of the genes39; therefore, the intensity of the coloration of the picrate paper was recorded using a quantitative scale with values ranging from 0 (yellow: no color change) to 3 (dark brown: very intense coloration). The tubes were incubated for two hours at 37 °C, and the result for the detection of Ac/Li cyanotypes was recorded. After 24 h, another recording was made to detect Ac/li cyanotypes, as the glycoside hydrolyzes spontaneously in the absence of the enzyme in a slower chemical reaction. Finally, to distinguish between ac/li and ac/Li cyanotypes, 100 leaves from plants previously determined to be Ac/Li were autoclaved to destroy the enzyme but not the glycoside, and the resulting extract was used as a reagent in plants with negative results. Apple seeds, which are highly cyanogenic, were used as a control for a very positive reaction (value 3) to contrast the maximum value. The cyanogenic potential (CP) of each plant was determined as the sum of the values at 2 h, 24 h, and at 24 h adding the autoclaved extract (maximum value of 9, Table S2).

For statistical analyses of the agromorphological traits, a fixed-effects ANOVA was performed for each variable accordingto the following model Xmjk = μ + Cm + Rj + (CR)mj + Єmjk; where Xmjk is theobservation of the population/cultivar (C) i (i = 1 to 15) in the block or repetition (R) j (j = 1, 2, 3) for morphological traits and the cyanogenic potential, respectively, and the sample k (k = 1 to 28); μ is the mean of all the observations; Cm, Rj, (CR)mj and Єmjk are the effects of the cultivar m, the block or repetition j, the interaction between cultivar and block or repetition, and the error associated to the sample k in the observation mjk, respectively. A Student–Newman–Keuls (SNK) multiple range test was used to.

assess significant differences between means at the 5% level.

Principal components (PCs) were estimated on the correlation matrix of the standardized variables and the first three PCs were used to capture the most variation in the data by using SPSS V.22. Principal components (PCs) were used for an agglomerative hierarchical cluster analysis aimed at identifying groups of similar agronomic types. The squared Euclidean distance was employed as the metric for measuring distances, while Ward’s clustering algorithm, which focuses on minimizing the increase in the sum of squares within clusters, was utilized to amalgamate accessions into clusters. This approach effectively groups accessions in a manner that minimizes within-cluster variance. A specific partition was selected from the resulting classification tree.

Additionally, a fixed-effects ANOVA was performed for each variable according to the following model to test the phenotypic differences between ecotypes/cultivars grouped in RRPs (including admixed genotypes as a group): Xim = μ + RPPi + Єim; where Xim is the observation of the RPPi (i = 1 to 3 for K = 2) and the ecotype/cultivar m (m = 1 to 15); μ is the mean of all the observations; RPPi, and Єim are the effects of the RPPi and the error associated to the ecotype/cultivar m in the observation im, respectively. Graphical representations of the PCs included the genetic structure based on SSRs.

The correlation (Pearson’s r) was estimated for the agromorphological traits and RPPs obtained using the Structure software utilizing SPSS V.22.

Results

SSRs

All local populations and cultivars showed genetic variation, and a total of 300 genotypes were differentiated (Table S3). Among the 300 unique genotypes analysed, 2 exhibited two bands, 74 three bands, and 224 four bands, indicating putative diploids, triploids and tetraploids, respectively. A total of 225 alleles were detected (Table 2), with a minimum of 5 alleles for RCS0035 and a maximum of 34 for GTRS851. The average per locus was 16.07.

Table 2 14 SSRs evaluated in 12 local populations, an experimental selection conducted by CIAM (Var.Exp.Tri4, 9079), and two commercial cultivars (‘Huia’ and ‘California’) of white clover, classified by the two reconstructed populations (RPPs) obtained by Structure software.
Full size table

Genetic structure

Two primary groups were identified (Table S3, Figs. 1, S1): 121 genotypes in RPP1, 140 genotypes in RPP2, and 39 admixed genotypes. RPP1 comprised exclusively of seven local populations: 1812, 1815, 1816, 1817, 1819, 1820, and 1821. All of them were found in Asturias and León, but none in Galicia. In contrast, RPP2 included five local populations (1813, 1814, 1825, 1851, 9079 [Var.Exp.Tri4]) and two commercial cultivars (‘Huia’, code 9009, and ‘California’, code 9010). Population 1818 was predominantly admixed.

Fig. 1
Fig. 1
Full size image

Bayesian analysis was conducted using 14 microsatellites (SSRs) with Structure software on 300 unique genotypes of white clover, including northwestern local populations (RPP1 in red) and commercial white clover cultivars (RPP2 in green).

Populations 1812, 1816, 1819, 1825, 1851, and 9079 grouped all genotypes within the same RPP. Population 1813 had 3 out of 20 genotypes as admixed. Population 1814 had 1 and 2 genotypes out of 20 in RPP1 and admixed, respectively (Table S3). Population 1815 had 2 and 5 genotypes out of 20 in RPP2 and admixed, respectively. Population 1817 had 2 and 2 genotypes out of 20 in RPP2 and admixed, respectively. Population 1818 had 13 genotypes admixed, and 3 and 4 genotypes out of 20 in RPP1 and RPP2, respectively; therefore, it was considered admixed. Population 1820 had 9 genotypes out of 20 as admixed. Population 1821 had 1 and 3 genotypes out of 20 in RPP2 and admixed, respectively. Cultivar ‘Huia’ had 1 genotype out of 20 in RPP1 and another 1 classified as admixed. Cultivar ‘California’ had 1 out of 20 genotypes as admixed. The genetic differentiation between the two RPPs was statistically significant, as indicated by an Fst value of 0.055 (P < 0.001). RPP1 showed higher number of alleles (189) than RPP2 (164), rare alleles (114 and 96, respectively), and exclusive alleles (37 and 22, respectively) (Table 2). The neighbour-joining (NJ) tree, based on the Jaccard Coefficient matrix of 300 clover genotypes, also clustered them in alignment with the structure for K = 2 (Fig. 2).

Fig. 2
Fig. 2
Full size image

Neighbour-joining tree constructed using 14 SSRs across 300 unique white clover genotypes. Commercial cultivars and northwestern local populations associated with RPP2 (K = 2) identified with Structure software are highlighted in green.

Agromorphological traits including cyanogenesis

Average values (Table 3) showed that FLO presented the highest value observed in 1812 (135.5) and the lowest in 9010 (53.4), with an average value of 109.2; ALT varied from 8.5 (1819) to 31.4 (9010), with an average value of 15.3; ANC showed the lowest value being 39.6 (1821) and the highest 79.8 (9010), with an average value of 61.6; AIN varied from 1.9 (1820) to 4.8 (1816), with an average value of 3.6; LFC ranged from 1.4 (1821) to 4.0 (9010), averaging 2.5; AFC ranged from 1.1 (1821) to 3.3 (9010), with an average of 1.9; GRP varied from 1.0 (1819 and 1821) to 2.9 (9010), with an average of 1.6; LOP ranged from 4.5 (1821) to 25.4 (1814), averaging 12.0; GRE spanned from 2.2 (1819 and 1821) to 4.4 (9010), with an average of 3.0; ENF ranged from 2.6 (9010) to 5.0 (1812, 1814, 1815, 1816, and 1819), with an average of 4.4; CRP varied from 2.3 (1821) to 4.5 (9010), averaging 3.4; CRI ranged from 2.4 (1815) to 3.7 (1820), with an average of 3.0; finally, CP varied from 3.0 (1820) to 8.4 (1818), with an average of 5.9. The ANOVA showed significant differences among populations/cultivars, but not the block, while the interaction was significant for most of the traits (Table 3).

Table 3 Local populations and cultivars of white clover (Trifolium repens L.) and commercial cultivars evaluated with 13 phenotypic traits classified by the reconstructed population (RPP) obtained by Structure software.
Full size table

Highest correlations (over 0.7) were obtained between LFC and AFC, GRP, LOP and GRE (0.835, 0.720, 0.717, and 0.719, p < 0.01); AFC and LOP (0.730, p < 0.01); and GRP and GRE (0.764, p < 0.01) (Table S4).

PCA of agromorphological traits revealed that the first three principal components (PCs) accounted for 83% of the total variance, with PC1, PC2, and PC3 explaining 57.7%, 16.8%, and 8.2% of the variance, respectively. The main positive influences on PC1 were LFC, GRE, GRP, AFC, LOP, and lower ALT and CRP, while FLO had a negative effect. For PC2, the main positive influences were AIN, ENF, CP, and ANC, with CRI having a negative effect. PC3 was influenced positively by ENF and negatively by AIN.

RPPs presented significant agromorphological differences for all traits except FLO and CRI. RPP2 showed higher values for most traits except ENF, but there was considerable variability within each RPP (Table 3). The two commercial cultivars (9009, ‘Huia’, and 9010, ‘California’) included in this study showed the lowest incidence of diseases, followed by populations 9079 and 1851 (RPP2). The commercial cultivar ‘California’ (9010) had the most positive value for PC1 due to its significantly shorter flowering date from planting date, and the highest values for LFC, GRE, GRP, AFC, ALT, and CRP. Notably, 1814 (RPP2) had the second-best growth after ‘California’, but with the highest values for ENF. 9079 (Var.Exp.Tri4), also grouped in RPP2, showed high FLO, growth parameter similar to ‘Huia’ but with lower CP.

Populations 1820 and 1821, both grouped in RPP1, exhibited the most negative values for PC2 and positive values for PC3 (Fig. 3), primarily for PC2 due to the lowest values for AIN, but also due to low values of CP, with 1820 and 1821 having the significantly lowest values in this study (CP = 2.97 and 3.28, respectively) (Table 3). Ecotype 1821, collected at the highest elevation in Asturias (1500 m.a.s.l.), presented the lowest growth form for most parameters. Ecotype 1820, also from Asturias (800 m.a.s.l.), differentiated from 1821 due to higher ALT and ANC than 1821. Landrace 1817 from Asturias had the most positive value of PC2, related to one of the highest values of CP and the lowest CRI of all samples evaluated. No significant correlation was found between CP and ENF.

Fig. 3
Fig. 3
Full size image

Principal components (PCs) over 13 agromorphological traits of 12 local populations, an experimental selection conducted by CIAM (Var.Exp.Tri4, 9079), and two commercial cultivars (9009, ‘Huia’, and 9010, ‘California’) of white clover classified by the reconstructed populations (RPPs) obtained by the Structure software (molecular data).

Hierarchical clustering analysis (Fig. 4), which was performed on the first three PCs (83% of the total variance), at the three-cluster cut-off level grouped the commercial cultivars with some local ecotypes in Cluster 1, differentiated the local ecotypes 1820 and 1821 in Cluster 2, and other local ecotypes in Cluster 3. Clusters showed significant differences for each trait, even for FLO and CRI, due to the significantly lowest values of FLO for Cluster 2 and of CRI for Cluster 3, which were not significant for RPPs (Table 3).

Fig. 4
Fig. 4
Full size image

Dendrogram based on the results of a hierarchical cluster analysis using the Ward method over the first three principal components (accounting for 83% of the variance) estimated from 13 agromorphological traits of 12 local populations, an experimental selection conducted by CIAM (Var.Exp.Tri4, 9079), and two commercial cultivars (9009, ‘Huia’, and 9010, ‘California’) of white clover.

Discussion

The average number of alleles per locus (16.07) was double that found in a previous study on white clover (8.66), which also used 14 SSR markers2 but evaluated a more reduced number of samples.

In general, the agromorphological characteristics of the local ecotypes (RPP1) of white clover presented lower growth than the commercial cultivars, as was demonstrated for red clover25. They were also more sensitive to diseases. For example, our results agree with the previous description of cv. ‘California’ as having large leaflets2. Local ecotypes also showed significantly lower GRE, which is related to lower winter survival, as thicker stolons tend to have higher carbohydrate reserves40, and better regrowth and persistence41 However, variability was found among local ecotypes, with some having values similar to the cultivar ‘Huia’ (2.92 mm), but not close to the maximum value of ‘California’ (4.39 mm). The stolon thickness of Grasslands Trophy was considered intermediate, with a value of 2.4 ± 0.16 mm, similar to some of the local ecotypes evaluated in this study.

However, local ecotypes of white clover showed longer FLO than the commercial cultivars, and the opposite was observed with local red clover versus commercial cultivars25. Local ecotypes also showed significantly lower values of CP compared to commercial cultivars, with ‘Huia’ considered to have medium–high cyanogenesis potential42, similar to the experimental local selection 9079, which showed high FLO, similar plant growth to ‘Huia’, but higher ENF and lower CP. The lack of correlation between CP and ENF found in our study was also reported previously1, specifically for pepper spot caused by Stemphylium sarciniforme (Cavara) Wiltshire.

Our results were in agreement with a previous study on agromorphological traits that included some of the local ecotypes and the two commercial cultivars37, but without genetic analysis. That study differentiated three clusters: one including the commercial cultivars with some local ecotypes (Cluster 1), and two others (Clusters 2 and 3) with only local ecotypes. Notably, ecotype 1821 was also differentiated into an independent cluster in the previous results that included this ecotype37. Interestingly, the clusters showed significance for all traits, even for those traits that were not significant for the RPPs (FLO and CRI). Our study also agreed with the higher plant and leaf development observed in the group that included the commercial cultivars.

‘1815’ had similar ANC (maximum plant width) than ‘California’, 78.2 versus 79.8 cm, respectively, but a significantly half lower ALT (maximum plant height) and it could not be classified as T. repens var. giganteum Lagr.-Foss. (the large kind) according to the classification by Fick and Luckow43. Some of them presented similar ANC than ‘Huia’, which was classified as T. repens f. hollandicum Erith ex Jfiv. & Sob (the intermediate kind), with values around 63 cm or slightly higher, such as ‘1820’, ‘1818’, ‘1817’ and ‘1816’, but with lower ALT. ‘1821’, ‘9079’ and ‘1851’ presented the significantly lowest ANC, with ranges 39.6–41.7, which could classify them as T. repens L. f. repens (the small kind).

Furthermore, phenotypic variation inside each RPP was high, which is of interest for breeding programs looking for specific traits in not related germplasm with commercial cultivars. Moreover, as stated by Caradus et al.1, “the contribution of white clover to sustainability and environmental goals is a growing focus of breeding programs”.

Breeding programs should aim to deliver high performance across diverse environmental conditions44,45. Recent studies have demonstrated that white clover is a key component of both permanent and sown pastures and represents an important target in breeding programs46. It contributes to increased dry matter yield, improved forage quality, enhanced persistence, greater drought tolerance, and germplasm conservation12, and serves as a valuable legume component in the establishment of cultivated meadow and pasture agrophytocenoses under temperate climates47.

Comparison with red clover. This study was conducted in parallel with the study of red clover (T. pratense L.) from northwestern Spain25 using the same number of SSR markers (14) and similar phenotypic traits, with the addition of cyanogenic potential evaluated in this study for white clover. The number of local ecotypes evaluated was 13 for both red clover and white clover.

For the same number of unique genotypes analysed (300) in both species, the level of ploidy differed: 63% of the red clover genotypes showed 2 alleles compared to 0.67% in white clover, 20.33% vs. 24.67% with 3 alleles, and 16.67% vs. 74.67% with 4 alleles, respectively. The higher number of diploid genotypes in red clover could explain the greater diversity in the number of alleles. Both white and red clover require cross-pollination48,49, but reproductive success was related to ploidy level, with 0.52 seeds per pollinated floret in diploid and 0.16 in tetraploid cultivars of red clover49. This could also be related with the percentage of populations and commercial cultivars grouping genotypes in the same RPP, lower in the red clover (20%), with higher percentage of diploids and reproductive success, indicating an increased probability of intercrossing in the fields, than in white clover (30%).

Red clover showed a higher number of alleles (273) than white clover (225), and exclusive alleles (80 vs. 64). The percentage of rare alleles (P < 0.05) was similar, 72.5% vs. 68.4%, respectively. The average number of alleles per locus was also higher in red clover, with 20.08 vs. 16.07 alleles per locus.

Only two SSR markers amplified in both species (12 others that amplified in red clover were not useful in white clover): RCS0035, which maintained the tendency of showing a higher number in red clover (14 vs. 5), and RCS0685, which did not follow this trend (5 vs. 7).

When considering only local ecotypes not related to commercial cultivars obtained by Structure software (RPP1), white clover showed a slightly higher number of alleles (189) compared to red clover (172). The percentage of rare alleles (P < 0.05) was higher in white clover (84%) than in red clover (63%), but the number of exclusive alleles was lower (23 vs. 37).

The Fst value between the RPP including only local populations and the RPP with commercial cultivars was 0.086 (p < 0.001) when K = 2 for red clover in Spain25, which is higher than the value for white clover in the present study (0.055, P < 0.001). This difference is related to the higher total number of alleles, exclusive alleles, and rare alleles present in red clover. The Fst value in our study was also lower than that found in Ukraine, where the Fst value was 0.07 between the two main populations of red clover detected by Structure software.

Conclussions

White clover exhibitedlower genetic variation than red clover in northwestern Spain, likely related to the higher number of tetraploids compared to diploids. Nevertheless, it still contained distinct genetic groups of ecotypes conserved at CIAM, which can be employed in breeding programs. The genotypic variation was related to phenotypic variation, wherein the commercial cultivars displayed greater plant and leaf development along with lower sensitivity to diseases. Among the genetic groups composed solely of local ecotypes, two clusters could be phenotypically distinguished: one cluster showed the lowest parameters for plant and leaf development but also had the lowest cyanogenic potential (CP). CP was not related to lower sensitivity to diseases (ENF), a result that was also observed in previous studies.

Analysis of agromorphological and genetic variability revealed significant differences among Trifolium repens genotypes, confirming substantial potential for selection based on agronomic traits and marker-assisted selection (MAS). MAS accelerates conventional breeding by enabling selection at any developmental stage, irrespective of environmental influences, and is particularly advantageous for traits that are difficult to measure. This approach will facilitate the development of genotypes with desired characteristics, such as disease resistance or/and increased yield. Moreover, they will be genetically distinguishable and can be propagated and registered to diversify plant material with local and more adapted selections.

Supplementary Information.

The online version contains supplementary material.