QTL meta-analysis in perennial ryegrass
As a pasture crop species, a predominant focus on vegetative yield-related characters has been observed during trait-dissection studies of perennial ryegrass, leading to identification of a large number of leaf/pseudostem and plant mass-related QTLs (Table
1). In contrast, as a grain crop, panicle/flower and seed-related traits have received more attention in rice QTL identification activities
. In the present study, lower QTL numbers were identified on LGs 5 and 6. In comparison, a meta-study for hexaploid wheat grain yield-related QTLs identified relatively smaller numbers of QTLs were identified on the homoeologous 5, 6 and 7 groups of chromosomes
, which exhibit extensive macrosynteny with perennial ryegrass LGs 5, 6 and 7
. Hence, despite divergent trait-specific biases between perennial ryegrass and hexaploid wheat, a similar chromosomal distribution pattern of QTLs was exhibited (Table
1). In a previous study of grain yield under drought stress conditions, conservation of QTL locations between different Poaceae species was observed
. As perennial ryegrass and wheat are relatively closely related within the cool-season grasses, the similarity of QTL distribution patterns between these two species suggests that conserved regions corresponding to wheat homoeologous chromosomes 5, 6 and 7 show lower importance than others for agronomic traits, including both vegetative and seed yield characters. Due to a large genome size, the perennial ryegrass genome has not yet been completely sequenced and assembled. Full assembly of genome sequence information from chromosomes that are rich in important QTLs may be more valuable, and should perhaps be prioritised, in comparison to that from other chromosomes.
The bibliographic survey identified putatively conservation of QTL locations under different environmental and across different genetic backgrounds. Plant height QTLs on LG1 were reported in three distinct studies, and heading date QTLs on LGs 4 and 7 were identified with various parental combinations at multiple geographic locations, although further analysis is required to determine whether the common QTLs are controlled by identical genetic factors
[13, 15, 18, 29, 32, 42]. The two pathogen resistance MQTLs are also putatively conserved under multiple environmental conditions and genetic backgrounds. Conversely, evidence was also obtained for a relatively large number of QTLs that are either genotype- or environment-specific. QTL analysis studies with two-way pseudo-testcross populations have demonstrated the presence of QTLs only on single parental genetic maps for traits measured under identical environmental conditions
[2, 17, 32]. Several studies also subjected single populations to QTL analysis under various environmental conditions, and reported environment-specific QTLs
[33, 42, 45, 46]. The p150/112 mapping population was developed for the activities of the International Lolium Genome Initiative (ILGI) and was subjected to QTL analysis for traits such as leaf length, leaf width, and variation for heading date in both Japan and the UK, identifying unique QTLs at the two geographic locations
[13, 15]. Leaf length and width QTLs were identified on LGs 5 and 3, respectively under Japanese conditions, while QTLs for both traits were found on LG7 in the UK-based trial. Only a single heading date QTL on LG4 was detected in Japan, while two QTLs on LGs 4 and 7 were found in the UK, probably associated with vernalisation genes (Vrn-1 and Hd3a orthologues, respectively). These results suggest that although stable QTLs may be detected under different environmental and genetic backgrounds, QTL identification largely depends on both genetic and environmental factors in perennial ryegrass.
The frequency distribution of Vp demonstrated in this study (Figure
1) was also similar to that obtained from a previous study in rice, in which the mean Vp value was calculated to be c. 13%, based on a sample of 231 QTLs
. In both studies, although the distribution range was skewed towards lower Vp values, a considerably small number of QTLs were classed in the 0-5% category. The probable presence of loci of minor effect, which could be excluded from identification due to the requirement for threshold LOD values for QTL detection, was also described for the rice study, and such minor undetected QTLs are also likely to be present in perennial ryegrass. Although F2 and BC1 genetic mapping populations have been generally employed for rice, construction of perennial ryegrass linkage maps has been commonly based on use of one-way and two-way pseudo-testcross strategies, due to the effect of an outbreeding reproductive habit. These crossing formats may not achieve such precise estimation of QTL effects as the F2 and BC1 designs, due to complexity of genetic background
. Despite this difference, the distribution patterns of Vp values were largely similar between the two species.
Due to the relatively small sizes of discovery populations (typically in the range from 100–200 genotypes) that have been used for trait-dissection in perennial ryegrass, the magnitudes of QTL effects have probably been over-estimated. Several studies have identified failures to deliver anticipated genetic gains through marker-assisted QTL selection, due apparently to both over-estimation and imprecise estimates of location
[63, 64]. The basis of these problems has been extensively discussed, and has in most cases been attributed to the influence of experimental population size, the so-called Beavis effect
[65–67]. QTL identification in progressively larger population sets, up to 500–1000 individuals, has been theoretically and empirically demonstrated to enhance the accuracy of QTL effect measurement. Alternatively, more accurate estimates of locus-specific effect are likely to derive from implementation of genome-wide association studies (GWAS). For example, a GWAS for 14 agronomic traits in rice identified six characters associated with colours, grain quality and grain width that exhibited a small number of significant loci with large effects, while the remaining traits were influenced by multiple loci with relatively small effects
. Equivalent studies in perennial ryegrass might be anticipated to generate similar results.
The BioMercator software assisted the melding of linkage maps resulting from distinct studies. This process, however, was not fully accomplished in the present study, except for the p150/112, AU6, NA6, WSCF2, MFA, MFB and SB2 x TC1 maps, due to insufficiency of common genetic markers. In previous studies, non-functional DNA-based markers, such as genomic DNA-derived SSR, AFLP and restriction site-associated DNA (RAD) systems, were predominantly used
[18, 32, 37]. Such assays are not ideally suited to comparative mapping studies, as multiple locus amplification is often observed for genomic DNA-derived SSR markers, and both AFLPs and RADs are more genotype-specific than functional markers
[11, 31, 37]. Enrichment of functional markers is hence essential for a further meta-analysis. A recent study assigned over 700 gene-derived markers to perennial ryegrass LGs with public release of the corresponding information
. The outcomes may permit efficient functional marker enrichment in specific chromosomal regions of interest.
Prediction of candidate gene status
Two putative MQTLs were identified for pathogen resistance (Table
2). Both mqResis-2 and mqResis-6 were identified as consensus loci containing both grey leaf spot and crown rust resistance QTLs, implying non-specific activities for several pathogens, rather than race-specific resistance QTLs. Through the process of genetic map alignment and MQTL analysis, additional functionally associated markers that are putatively linked to the QTLs were identified. Information from functional markers may support development of novel flanking DNA-based markers for a given target locus based on a comparative genetics approach, enabling candidate gene-based selection and association genetics studies
[15, 70]. Although further characterisation is required, both MQTLs and flanking functional markers may be useful for deployment in perennial ryegrass breeding.
Comparative analysis demonstrated close proximity between genetic markers related to the DGL1, Ph1 and OsPIPK1 ortholoci and the corresponding perennial ryegrass QTLs. This observation suggests that the DGL1 and Ph1 ortholoci are related to, and may provide candidate genes for, the herbage yield-related QTLs on LG3. In a previous study, the CDO795-linked heading date QTLs were suggested to be equivalent to a rice heading date QTL, dth3.3 (Gramene QTL Acc. ID AQFE011)
[15, 71]. The physical location of the OsPIPK1 gene was located in the candidate interval (5.7 Mb) of dth3.3. These results suggest that the perennial ryegrass OsPIPK1 ortholocus may be related to the heading date QTLs on LG4. For both yield and flowering time traits, plausible evidence for related candidate genes has been obtained.
In contrast, markers linked to the EUI1 and D3 ortholoci were located over 10 cM distant from the maximum LOD values for the target QTLs. In a wide range of plant species, genes causing variation in quantitative traits have been identified to be located within genetic distances of less than 3 cM from the LOD maximum location
. It seems, therefore, unlikely that LpEUI1 and LpD3 genes are plausible candidates for QTL function. For issues arise for candidate genes associated with disease resistance.
The Pi37 and Pbi genes encode NBS-LRR proteins
[51, 59]. Molecular studies have shown a rapid evolutionary rate and limited cross-species synteny of NBS-LRR genes
[51, 59, 73, 74]. The comparative approach may not be so effective for such species-specific genes, due to unresolved paralogous relationships between species, and hence accounting for the failure of putative ortholoci to map in regions predicted on the basis of conserved synteny.