Genome-wide markers for seed yield and disease resistance in perennial ryegrass

Kristina Ja?kūn, Vilma Keme?yt, Andrius Aleliūnas, Grazǐna Statkevicǐūt

Institute of Agriculture, Lithuanian Research Centre for Agriculture and Forestry (LAMMC), Instituto av. 1, Akademija LT-58344, Lithuania

Keywords:Crown rust Puccinia coronata f. sp. lolli Flag-leaf length Genome-wide association study (GWAS)Seed yield Lolium perenne L.

ABSTRACT The market success of perennial ryegrass(Lolium perenne L.)cultivars depends on sufficient seed production,as they are propagated by seed. However, breeding for high quality forage production reduces seed yield, and breaking the negative correlation would help to overcome the problem. The foliar disease crown rust is another factor affecting reproductive capacity and thereby seed yield. We evaluated seed yield-related traits and resistance to crown rust in a collection of commercial cultivars and ecotypes of perennial ryegrass and identified genome-wide markers associated with the traits. The study revealed high variation between the ecotype and cultivar groups as well as between years. A genome-wide association study identified 17 DNA single-nucleotide polymorphisms(SNPs)of which eight were associated with crown rust and nine with flag-leaf length. The SNP markers were located within or near predicted genes functioning in defense against pathogens. The identified genes are strong candidates for a further in-depth functional study to continue unravel determination of leaf architecture and crown rust resistance in perennial ryegrass.

1. Introduction

Perennial ryegrass(Lolium perenne L.)is the predominant forage grass in temperate pasture agriculture for milk and meat production. Superior digestibility and high grazing tolerance, complemented by rapid establishment and adequate seed production[1,2], make perennial ryegrass predominant for agriculture and recreation [3]. It is a self-incompatible diploid species known for large morphological variation and high genetic diversity in populations[4,5].Although studies of abiotic stress tolerance and disease resistance [6–9] have also revealed high genetic diversity, little of this variation has been captured in breeding material and used for improving cultivars [10]. Knowledge about genetic diversity in available germplasm can support breeders’ decisions on the selection of cross combinations from large sets of parental genotypes for widening the genetic basis of breeding material and thus improving cultivars.

Because commercial perennial ryegrass cultivars are propagated by seed, sufficient seed production is necessary for competitiveness and commercial success of new cultivars introduced into today’s market. Although agricultural practices and environment strongly influence seed yield,genetic components such as number of reproductive tillers,inflorescence size,and spikelet number also determine seed yield[2].However,the use of forage crops for grazing or high-quality feed production has directed breeding towards higher leafiness and nutritive value as well as high dry-matter yield [11], with seed yield not a breeding target of first priority[12]. Improvement of dry matter digestibility also demands a reduced proportion of generative tillers, leading to unacceptable seed yields [13]. The way to overcoming this problem could be the development of genotypes that produce fewer but highly fertile tillers [12].

Fungal diseases also may lead to severe losses of seed yield, as they affect plant growth, development, and fitness. Crown rust(caused by Puccinia coronata f. sp. lolli) is a foliar disease that disrupts photosynthesis and accumulation of water-soluble carbohydrates,impairing yield-related traits such as tillering,root growth,and regrowth after cutting[14],as well as persistency[15].In some perennial ryegrass cultivars,it can reduce yields by up to 30%[16];moreover, it damages forage quality, potentially leading to health problems in grazing cattle [17] and liver damage in dairy calves[18].The effect of crown rust on reproductive capacity of the plants is also well documented; infected plants produce fewer tillers and smaller, lighter seeds, reducing seed yield [14,19,20]. Because the disease also reduces the vigour of perennial ryegrass and lowers its competitiveness, the composition of plant mixtures gradually changes in long-term swards [15,19].

Increased tolerance and resistance to crown rust is fairly highly heritable (0.46) and can be controlled by breeding [21]. Such breeding has conventionally involved phenotypic recurrent selection at the single-plant level. As perennial ryegrass germplasm is highly variable for this trait and disease resistance is correlated between single plants and swards, it can be effectively exploited in breeding programs, resulting in superior cultivars [16].

The fastest and most effective way to develop improved cultivars is the combination of phenotypic selection and molecular breeding tools. Genetic markers are known for their use in identification of quantitative trait loci (QTL) associated with agronomic traits in biparental populations.Examples include studies of crown rust resistance,heading date,seed yield,frost tolerance,leaf length and width, and other morphological traits [22–25]. However,biparental populations do not represent all breeding material, so that the transfer to other genetic backgrounds of DNA markers linked to identified QTL is often problematic [26]. Another approach, known as association analysis, overcomes the restrictions of biparental QTL mapping by employing unrelated genotypes for identifying population-wide marker-phenotype associations [5,8,27,28]. The strength of this technique is utilization of historical recombination patterns that happen to occur within a set of genotypes and detection of correlations between genotypes and phenotypes within these genotypes[29].The greatest advantage of this approach is that genotypes forming association populations and used for identification of marker-trait associations can be directly incorporated in breeding programs.

The objective of this study was to identify genetic loci involved in disease resistance and formation of seed yield in perennial ryegrass. The experimental approach was to perform genotyping-bysequencing and association analysis in a collection of commercial cultivars and ecotypes and identify genome-widely distributed markers associated with resistance to crown rust and flag-leaf architecture.

2. Materials and methods

2.1. Plant material

An association mapping panel of perennial ryegrass comprising 216 accessions including 95 cultivars and 121 natural ecotypes was used. Most of the cultivars were of European origin, except for one each from Japan and New Zealand and seven from the USA. The group of cultivars comprised a subgroup of 48 foragetype cultivars and one of 47 turf-type cultivars.The ecotype group was also divided into two subgroups based on their origin. Sixtyseven ecotypes from Ukraine and two from Slovakia comprised the subgroup of continental ecotypes(EcoCont),whereas 37 genotypes from Lithuania, two from Poland, three from Latvia, and 10 from the Kaliningrad region of the Russian Federation formed the subgroup of maritime ecotypes (EcoMar). A detailed description of the association mapping panel is presented in Ja?kūnet al.[6].

2.2. Design of the field experiment

The panel was established in the field in 2013 as described by Statkevicǐūtet al. [30] and maintained until 2020 by vegetative propagation every three years. Four replicates of each accession were planted at 50 × 50 cm distances using a randomized complete block design in Akademija (Kedainiai district, Lithuania).The soil of the experimental site was Endocalcari-Epihypogleyic Cambisols. Lithuania is located in the nemoral zone, which is defined by a cool temperate climate with a 190–195-day growing season[31].Daily meteorological data(mean,minimum,and maximum temperatures, precipitation, and snow cover) over the period of 2013–2014 and 2019–2020, was recorded at the meteorological station in Akademija. The years of the study were characterized by contrasting meteorological conditions, as shown in Fig. S1.

2.3. Phenotyping of traits

Flag-leaf length (FLL), flag-leaf width (FLW), inflorescence length (IL), number of spikelets (SN), seed weight per plant (PS),and seed weight per inflorescence (IS) were recorded in 2013 and 2014, and infection by crown rust in 2019 and 2020. Flagleaf architecture traits, in particular FLL and FLW were measured at full emergence of five inflorescences and the stems were labeled.The inflorescences of the labeled stems were collected at full ripening stage, air dried and the seed yield-related traits IL, SN and IS were assessed. PS was determined by harvesting an individual plant at its full ripening stage, drying it to constant weight, and threshing the seeds. Crown rust and leaf spot disease were evaluated in the middle of September and scored as 1=no rust damage,5 = 25%, and 9 = more than 75%.

The statistical analysis was conducted in the R statistical environment (version 4.0.2; [32]). Analysis of variance and post-hoc Tukey HSD tests were performed with the agricolae R package[33]. As some of the traits did not show normal distributions, the phenotypic data were transformed using transformation methods implemented in bestNormalize R package [34]. Ordered quantile normalization was applied for crown rust data,IS data were transformed using Standardized log_b (x + a) transformation, PS by Standardized Yeo-Johnson transformation, and FLL by sqrt (x + a)transformation. The relationships between the traits were evaluated as Pearson correlation coefficients for each trait pair.

2.4. Genotyping-by-sequencing library preparation, sequencing and single nucleotide polymorphism discovery

High-quality DNA was extracted and genotyping-bysequencing libraries were prepared and sequenced as described in Ja?kūnet al. [6]. Briefly, genomic DNA was digested with PstI and the restriction products were ligated to unique barcoded adapters and sequenced on an NextSeq 2000 (Illumina, San Diego, C A,USA) sequencing system.After initial quality checking,reads were demultiplexed and mapped to the perennial ryegrass draft genome[35]. Variants were called with Genome Analysis Toolkit [36] and then filtered by minimum read depth of 5, minimum GQ score of 30, and minor-allele frequency (MAF) below 0.05. Variant sites with more than 50% of missing data were excluded.

2.5. Population structure and genome-wide association study

Characterization of population structure in the perennial ryegrass association panel and genome-wide association study(GWAS) analysis were performed as described in Ja?kūnet al.[6]. Population structure was characterized by principal component analysis (PCA) using the ‘‘prcomp” function in the R stats package.Genotypic missing data were replaced using imputes with the mean for each marker implemented in R hmisc package [37]and then subjected to PCA. Marker-trait associations were identified using transformed phenotypic data values. GWAS was performed for each trait separately using BLINK (Bayesianinformation and Linkage-disequilibrium Iteratively Nested Keyway) [38] and Mixed Linear Model (MLM) [39] implemented in GAPIT (https://github.com/jiabowang/GAPIT3). For each trait, the optimal number of covariates to be included in GWAS models was determined by model selection using the Bayesian information criterion(BIC).The P values for marker-trait associations were corrected by the FDR method [40]. Quantile-quantile (QQ) plots(Fig.4)were used to assess the number and magnitude of observed associations between SNPs and the traits.

3. Results

3.1. Phenotypic variation for crown rust and seed yield-related traits

All phenotypic traits but disease resistance varied substantially within the panel(Figs.1 and 2).Very low variation for both disease infection as well as PS was observed between groups within years,though the traits differed significantly among the years of the study (Figs. 1C, D, and 2D).

High variation for FLL and FLW, IL, SN, PS and IS was found between the groups of ecotypes and cultivars and among experimental years (Figs. 1 and 2). In the whole panel, the values for FLL ranged from 3.0 to 21.5 cm with means of 9.9 cm for 2013 and 13.4 cm for 2014. A significant difference (P ＜0.01) was observed between the groups of ecotypes and cultivars within the year of investigation as well as between the years (Fig. 1).FLW of cultivars varied (P ＜0.01) over years and was significantly narrower (mean value of 0.41 cm in 2013 and 0.45 cm in 2014)than FLW of ecotypes (mean value of 0.47 cm in 2013 and 0.49 cm in 2014, P ＜0.01). All seed yield-related traits: IL, SN, PS and IS, showed significantly lower (P ＜0.01) values in 2013 than in 2014 with the ecotypes performing better than the cultivars(Fig. 2). In the whole panel, IL varied from 6.60 cm to 28.54 cm with means of 14.60 in 2013 and 20.89 in 2014.The highest differences were observed for PS,when the mean for ecotype group was 2.14 g and for cultivar group was 1.47 in 2013, while in 2014 the ecotypes produced 26.21 g of PS and cultivars produced 16.99 g.

As expected, there were strong to moderate correlations between FLL and FLW in both experimental years, r = 0.80 in 2013 and r=0.55 in 2014,P ＜0.01(Fig.3).Strong positive and significant correlations were found between FLL and IL (r = 0.75 in 2013 and r = 0.68 in 2014), whereas the correlation with SN was moderate (r = 0.53 in 2013 and r = 0.50 in 2014). FLL, was moderately correlated with PS (r = 0.53) and IS (r = 0.59) in 2014, but in 2013 the correlation was weak or lacking(Fig.3).The relationship among IL and FLW was significant and moderate, though in the season of 2013 it was higher than in 2014, r = 0.66 and r = 0.45 respectively. All correlations are shown in Fig. S2.

3.2. Genome-wide association analyses

Quality control and subsequent filtering of phenotypic and genotypic data resulted in 188 diploid perennial ryegrass accessions suitable for analysis. Removal of variant sites with MAF ＜0.05 left 23,989 GSB markers for further analysis.SNP alleles were used to infer the population structure, which was visualized by PCA and showed some population stratification as described in Ja?kūnet al. [6] where the first, second, and third principal components accounted for respectively 2.9%,1.8%,and 1.3%of observed genetic variance.

After correction for multiple testing, 17 significant associations(FDR-adjusted P ＜0.05) between SNPs and analyzed plant phenotypic traits were identified(Table 1);magnitude of observed associations between SNPs and the traits is presented in QQ plots(Fig. 4). Nine significant associations were detected for FLL(Fig. 4A) and eight for crown rust (Fig. 4B). The majority of significant associations were located in predicted genes.

Fig. 3. Correlations among studied traits. Pearson correlation coefficients (P ＜0.01) between traits are given in light and dark gray circles, where light gray denotes an estimated correlation for 2013 and dark gray for 2014;n.s.indicates not significant.FLL,flag-leaf length;FLW,flag-leaf width;IL,inflorescence length;SN,spikelet number;IS, seed weight per inflorescence; PS, seed weight per plant.

The most significant (FDR adjusted P = 2.482348E-06) variant site associated with FLL was identified within a predicted gene for BTB/POZ and MATH domain-containing protein 2-like. Three other significant hits were also located within predicted genes,for calmodulin-binding transcription activator 3 (FDR adjusted P = 6.694763E-06), PSK SIMULATOR 2-like (FDR adjusted P = 1.515762E-05), and eukaryotic translation initiation factor 3 subunit A-like (FDR adjusted P = 0.000196842519) (Table 1).

Table 1 Most significant marker–trait associations for crown rust resistance and flag-leaf traits in perennial ryegrass.

The most significant variants associated with crown rust were also located within predicted genes, for SUPPRESSOR OF ABI3-5-like, alpha/beta-Hydrolases superfamily protein, CHD3-type chromatin-remodeling factor PICKLE, QWRF motif-containing protein 7-like, ankyrin repeat domain-containing protein 2A-like, and transport inhibitor response 1-like protein.

Fig.1. Variation of flag-leaf architecture traits and disease infection between the forage cultivar(n=48),turf cultivar(n=47),maritime origin ecotype(EcoMar,n=52)and continental origin ecotype(EcoCont,n=67)subgroups in the perennial ryegrass panel.Flag-leaf length(cm)and width(cm)are shown in(A)and(B)over the period 2013–2014. (C) and (D) represent plant infection with crown rust (scores) and leaf spot (scores) over the period of 2019–2020. Lowercase letters above violin plots denote differences between subgroups in the years 2013 and 2019, and uppercase letters, in 2014 and 2020 (Tukey HSD, P ＜0.05). FLL, flag-leaf length; FLW, flag-leaf width; IL,inflorescence length; SN, spikelet number; IS, seed weight per inflorescence; PS, seed weight per plant.

Fig.2. Variation in seed yield-related traits between the forage cultivar(n=48),turf cultivar(n=47),maritime origin ecotype(EcoMar,n=52)and continental origin ecotype(EcoCont,n=67)subgroups in the perennial ryegrass panel over the period of 2013–2014.Inflorescence length(cm)and number of spikelets are shown in(A)and(B).(C)and(D)show seed weight per inflorescence(g)and seed weight per plant(g).Lowercase letters above violin plots indicate differences between subgroups in 2013 and uppercase letters those in 2014(Tukey HSD,P ＜0.05).FLL,flag-leaf length;FLW,flag-leaf width;IL,inflorescence length;SN,spikelet number;IS,seed weight per inflorescence;PS,seed weight per plant.

4. Discussion

Commercial success of the grass varieties is partly determined by seed yield. The seed yield itself is a complex trait, consisting of several components, all of which are determined by the interplay of plant genetic background and environmental factors, such as abiotic stresses, disease pressure and agro-technology. Number of reproductive tillers and inflorescences as well as their length are among the main morphological characters defining seed yield [2].With increasing temperature and daylength after a winter period,a secondary induction of tillers begins, resulting in development of a new inflorescence. The contrasting growing conditions in the years of the study affected plant growth, resulting in differences in the measured traits. Cool and late spring of 2013 delayed the start of the vegetative season and led to fewer generative tillers,followed by shorter IL, than in 2014 (Fig. 2). The difference in IL did not strongly affect SN, so that differences in this trait among the years of the study were modest. This also can be said about IS, though fewer and shorter spikes combined with unfavorable spring conditions led to an extremely low PS in 2013. The genetic background of the accessions also played an important role along with environmental impact. The turf-type cultivars were the least productive with respect to PS or IS as well as to other seed yieldrelated traits in both 2013 and 2014, whereas both EcoMar and EcoCont groups showed higher mean values. The results suggest considering ecotypes as important donors in breeding programs for improved seed yield.The higher seed yield of the ecotypes than of the cultivars may be the consequence of breeding for higher biomass yield,compromising seed yield or SI alleles.Another factor in seed production is flag-leaf size[5]which reduces the rate of seed abortion in meadow fescue by reallocating assimilates via the stems to the inflorescence during anthesis [41]. The phenomenonis well described in cereals [42–44] and in turfgrasses it may be even more important,given that the vegetative tiller,forming biomass, competes with the generative organs, although the seed head may also function in seed filling.

Fig. 4. Quantile–quantile-plots for flag-leaf length (A) and crown rust (B) traits after correction for population structure.

In this study, 17 genome-wide markers were associated with FLL and crown rust and were located within or near predicted genes. The majority of the predicted genes for FLL are associated with plant growth and proliferation. In plants, a large expansion of the MATH-BTB family occurred in the grasses.BTB proteins have diverse functions in plants and may act as negative regulators of nitrate uptake [45], and the effect of their genes on plant growth was described in Bauer et al. [46]. Calmodulin-binding transcriptional activator 3 (CAMTA3) functions in plant growth through auxin- and brassinosteroid-mediated signaling pathways, with mutant plants showing dwarfism [47]. It has been speculated[46] that CAMTA3 serves as a positive regulator of plant growth.A gene for PSK SIMULATOR 2,belonging to a plant-specific protein family and tagged by a significant marker, is also involved in cell proliferation.Its mutants were reported[48]to show reduced proliferation and premature leaf growth arrest, possibly under metabolic control. Another gene for eukaryotic translation initiation factor 3 (eIF3) influences growth and development [49]. The eIF3 complex initiates translation of mRNAs involved in cell proliferation, including cell cycling, differentiation, and apoptosis [50].

Disease affects plant persistency as well as reproductive capacity,resulting in fewer tillers and smaller seeds[17].The crown rust fungus infects the grass throughout the year,but under Lithuanian conditions the highest pressure is observed in August [51]. However, meteorological conditions in the study years were not favorable for either crown rust or the leaf spot pathogen and the infection occurred in the middle of September. The damage was not severe and did not distinguish between cultivar or ecotype groups (Fig. 1). Even though cultivars are bred for some degree of crown rust resistance, they did not show greater resistance than the ecotypes, revealing some resistance to crown rust among the latter.

Breeders rely mostly on phenotypic selection based on longterm field experiments with natural infection by a wide variety of pathogen races,but variation in growing conditions from season to season may interfere with selection.Modern genomic tools may eliminate environmental effects and allow continuous and efficient selection [8,23]. In this study, eight genes were associated with crown rust in perennial ryegrass. The most significant marker was located within the gene for suppressor of abi3-5. The gene encodes a splicing factor that affects alternative splicing of ABI3[52] which is important for plant immunity, as it is required for constitutive defense responses [53]. Another candidate gene for pathogen resistance encoded ankyrin repeat domain-containing protein 2A, an essential molecular chaperone for peroxisomal ascorbate peroxidase 3 [54]. The protein may act in disease resistance via regulation of antioxidant metabolism [55]. A gene for transport inhibitor response 1-like (TIR1-like) protein was also associated with crown rust in perennial ryegrass. TIR1 is one of the receptors for the plant hormone auxin [56]. It has been suggested [57] that pathogens negatively regulate TIR1. This regulation might lead to auxin signaling repression and subsequent restriction of pathogen growth. Transport inhibitor response 1(TIR1) functions in defense against pathogens [58].

Agriculturally important traits, such as disease resistance and plant architecture, are complex traits, governed by many genetic loci. Genome-wide association studies provide a powerful tool to identify the genes underlying complex traits and therefore is widely used in plant research [59]. Eight SNP markers associated with FLL were detected within or near predicted genes and one was located within an uncharacterized gene sequence. Eight SNP markers associated with rust resistance were identified,all of them located within predicted genes. The identified genes are strong candidates for a further in-depth functional study to continue elucidating the determination of leaf architecture and crown rust resistance in perennial ryegrass.Data availability

Phenotypic data and SNP profiles are available at Figshare repository, https://doi.org/10.6084/m9.figshare.12433715.v3.

CRediT authorship contribution statement

Conceptualization, Funding acquisition, Project administration, Supervision, Writing – original draft, Writing– review & editing.Investigation.Andrius Aleliūnas:Formal analysis, Writing – original draft.Grazǐna Statkevicǐūt:Formal analysis, Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This study was carried out in the framework of the long-term research program ‘‘Genetic determination of the traits of agricultural and forest plants, development of modern cultivars” and‘‘Genome-wide functional analysis of perennial ryegrass for improved growth under water limiting conditions (GrowGene)”project funded by the Research Council of Lithuania (S-MIP-17-24). The authors are grateful for Prof. Bruno Studer for critical review and suggestions that helped to improve the manuscript.

Appendix A. Supplementary data

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2021.07.005.