There remains a great deal of genetic variation for seed yield even in elite breeding populations of the Lolium forage grasses (Elgersma, 1990). The major focus for selection has been on the improvement of the vegetative sward (Wilkins, 1991). Furthermore, in contrast to self-fertile annual grain crops, outbreeding perennial species carry high mutational load manifested by, among other things, low seed set caused by gamete and seed developmental abnormalities (Charlesworth, 1989). However, dependable and predictable seed yield is a vital component of the production of commercially viable forage grass varieties, as farm-scale propagation is achieved almost exclusively via seed. Consequently, this report of the position of QTLs influencing seed yield in Lolium populations has potential significance for the future manipulation of this important trait.
Initial analyses of these ryegrass mapping families suggested that, while the two traits were measured in different years, the genomic regions associated most closely with seed set and heading date were closely linked. Consequently, one of the main aims of the present study was to carry out a detailed analysis to determine if this linkage could be further clarified. Additionally, previous analysis of heading date in the F2/WSC populations highlighted the close syntenic relationship between the region of ryegrass LG7 which contained the major heading date QTL and the region of rice LG6 which contains the Hd3 and Hd1 heading date QTLs and their controlling genes, Hd3a and Hd1 (Yano et al., 2000; Monna et al., 2002; Armstead et al., 2004, 2005). Recently, Ji et al. (2005) and Qiu et al. (2005) identified that this same region of rice 6 also contained the S5n (S5n) gene, originally identified by Ikehashi & Araki (1988), which confers wide compatibility in rice and is particularly useful in overcoming the fertility barrier in indica/japonica hybrids. While, in general, there are few problems in crossing between ryegrasses, it is well established that the species possesses a high mutational load leading to a high magnitude of inbreeding depression for many traits (Corkill, 1956), probably because of the high likelihood of the build-up of somatic deleterious mutations resulting from the extended vegetative growth phases associated with perennial crop management. These mutations are maintained by an effective outcrossing system. Therefore, it is expected that seed setting is always considerably short of its full potential. In rice, the combination of an S5i and an S5j allele from indica and japonica subspecies results in partial fertility and seed setting of c. 50%, as opposed to the expected seed setting of (or nearly) 100% (Qiu et al., 2005). Although the overall seed setting figures for our perennial ryegrass populations were lower than this (as a consequence of the outcrossing nature of the species), the QTLs on LG7 accounted for a similar difference in magnitude for partial fertility. Consequently, another major aim of the present investigation was to determine if the orthologous region to in ryegrass could be identified; if its position was consistent with the known syntenic relationships between rice and ryegrass; and if this locus might contribute to the genetic control of seed yield in the forage grasses.
The association of seed set, heading date and flag leaf dimension QTLs on chromosome 7
In an outcrossing species with an effective self-incompatibility system, it might be expected that seed setting could be affected simply by pollen limitation through the spatial and temporal separation of pollen and receptive compatible stigma. In terms of spatial separation, this is unlikely to affect a particular heading date phenotype over another. In the worst case, pollen limitation might randomly affect some plants more than others in a particular experimental design and reduce the size of QTLs, or even make them less likely to be identified, but it would not affect the positioning of the QTLs. Temporal effects are also likely to be minimal as, although specific heading dates are scored for each genotype, based on the ear emergence date of the first three inflorescences, individual perennial ryegrass plants will flower over a much longer period, with pollen being produced over a period of 2–3 wk ensuring panmixis. Furthermore, in the ILGI population, the difference in heading date between the early and the late Hd3 marker genotypes is only 3 d; in the F2 family, heading date range is much greater, but all of the F2 plants capable of producing viable pollen were in fact self-fertile (Thorogood et al., 2005), ensuring close proximity of stigma with compatible self pollen. As an empirical test of the independence of seed set and flowering time per se, correlations were determined between seed set and flowering time for Hd3 marker genotype classes ‘aa’, ‘ab’ and ‘bb’ for the F2/WSC family and ‘aa’ and ‘ab’ for the ILGI family. All correlations, although negative, were nonsignificant. The correlations for the three F2/WSC genotype classes were −0.344, −0.212 and −0.196, with 22, 64 and 93 degrees of freedom, respectively, and those for the ILGI family were −0.070 and −0.233, with 54 and 43 degrees of freedom, respectively.
As grass fertility is affected by numerous environmental factors, such as nitrogen availability (Griffith et al., 1997), water availability (Martiniello, 1998) and disease incidence (Barker et al., 2003), the timing of flowering in relation to these external factors is of possible significance. Therefore, a gene which has a major influence on the timing of flowering could have a considerable indirect influence on seed yield. This is more likely in the F2/WSC family with a wide heading date range, whereas the smaller heading date range in the ILGI family implies a smaller window for variable environmental influence on seed set associated with flowering time.
It is interesting to note that the QTLs for seed set are very similar in magnitude in both families, accounting for 17 and 21% of the variance in the F2/WSC and ILGI families, respectively (Table 3), in contrast to the heading date QTLs that account for 77.9 and 20.2% of the total variation for the trait. Heading date in the F2/WSC and ILGI families was not measured in the same year as seed set but shows consistent patterns of heading date phenotype segregation from year to year when measured at the same site (M. Humphreys, pers. comm.). Furthermore, heading date has been shown to have reasonably high heritability (0.70 and 0.50) in two parent-offspring regression studies (Rogers, 1989; Wedderburn et al., 1992), and Cooper (1954) showed that although genetic variation can be revealed if temperature and photoperiod conditions are changed, heading behaviour is consistent when growing populations on the same site under similar environmental conditions, as is the case with the mapping populations in this study. It seems unlikely, therefore, that seed set would be equally affected by environmental factors in both populations when the potential for environmental variation, as determined by the flowering date QTLs, is so much larger in one of the populations than the other. Therefore, we propose an independent genetic mechanism influencing seed yield in both families. Figure 1 illustrates the LOD profiles for heading date, seed set and seed weight/inflorescence for the ILGI family on LG7 when analysed using MQ (with automatic cofactor selection) and WQC. MQ suggests that heading date and seed set colocalize, and seed weight/spikelet is slightly offset. By contrast, WQC suggests that seeds/inflorescence and seed weight/inflorescence colocalize and that heading date is slightly offset. From these results, and bearing in mind the different years in which the traits were measured, it is not possible to draw a firm conclusion as to whether the locus that is influencing heading date (for which Hd3a is a candidate gene) is the same locus that is directly influencing seed set. If two loci are involved then S5n would be a candidate gene for the latter trait. It is also quite possible that there might be a direct or indirect interaction between a pathway integrator gene involved in flowering induction (Hd3a) and a gene involved in fertility which would contribute to the problems involved in resolving the effects of two closely linked loci in this region.
High seed set is, in both families, associated with early flowering. If it is found that a separate locus (such as an S5n homologue) is responsible for seed-set determination, then the linkage between heading date and seed set could theoretically be broken, allowing plant breeders to select high seed-setting plants of both late and early flowering types. However, if both traits are determined directly by the Hd3a locus, they may be restricted (at least at this locus) to selecting for early flowering genotypes.
It is of interest to note that the S5n allele restores fertility over the semisterile condition of indica/japonica rice hybrids even in heterozygous form (Yanagihara et al., 1995), which is in accordance with our observations of dominance of high seed-setting-associated marker alleles over low ones in our F2/WSC ryegrass mapping family.
Differential resource allocation was approached specifically through a concurrent QTL analysis of flag leaf morphology, and the coincidence of QTLs for flag leaf dimensions in both families with the LG7 QTL for heading date and seed setting might partly explain the sizeable association of flag leaf width with seed yield in the closely related grass species Festuca pratensis, observed by Fang et al. (2004). It is likely that flowering phenology will have a significant influence on the development of vegetative organs. Wider flag leaf widths in the F2/WSC family associated with the ‘Aurora’ parent-derived alleles of the LG7 markers are also associated with the early ‘Aurora’ parent-derived flowering alleles, concomitant with early flowering being associated partly with faster growth rates. However, the relationship between flag leaf length and the heading date QTL on LG7 in the ILGI family is such that longer leaves are associated with later flowering. Therefore, there does not appear to be a direct functional link between heading date and flag leaf size, at least in this family. However, it is worth noting that the QTLs for leaf length and leaf width on LG7 of both families can be associated with the positions of Hd3a and/or S2539/Rz144, markers that are associated with the flowering control genes Hd3a (FT) and Hd1 (CONSTANS), respectively. Thus, there may be a possible linked induction of leaf growth and flowering onset.
There also does not appear to be a simple physiological relationship (i.e. in supply of photosynthate from the flag leaves) between the flag leaf size and seed set, even though the QTLs for these traits on LG7 are closely associated. Although higher seed set is associated with larger leaves in the F2/WSC family, the opposite is the case in the ILGI family. Thus, the association is most likely the result of genetic linkage, rather than physiological coupling.
Quantitative trait loci analyses of inflorescence number and number of spikelets per inflorescence did not reveal any significant QTLs for either family (unpublished results). With no discernible genetic variation for these traits, we assume that, at least in these mapping families, they cannot influence seed set through diversion or dilution of resources that would otherwise be used to maximize seed set.
Genetics and comparative genomics of seed set and heading date QTLs on chromosome 4
The pollen viability data are strongly indicative of a major recessive gene underlying the QTL on LG4 in the F2/WSC family (Table 5). Increased pollen viability derives from the ‘Perma’ (‘bb’) parent. Obligate selfing over four generations was practised to derive the two parent inbred lines of the F2/WSC population, and widespread pollen sterility was observed in the ‘Aurora’ lines (M. Humphreys, pers. comm.). The normal outcrossing nature of perennial ryegrass results in a high mutational load and considerable inbreeding depression upon selfing, thought to be a result of exposure of deleterious recessive genes. Although we cannot be certain, we postulate that the original ‘Aurora’ parent was heterozygous (normal phenotype) for the viability factor and, further, that the F1 plant, selfed to produce the F2/WSC mapping population, was also heterozygous (normal phenotype). The fact that a seed-set QTL is also revealed in this location despite viable pollen being readily available in an open pollinated situation indicates that female gametophyte and/or embryo development is also affected. An anther dehiscence QTL is also present on LG4 and this effect is also associated with the pollen viability/seed-set region (Table 3). Furthermore, poor anther dehiscence appears to be a recessive trait associated with poor pollen viability deriving from the ‘Aurora’ parent (Table 5). However, MQ mapping positioned the major anther dehiscence effect away from the pollen viability/seed-set region and most closely associates it with marker R2702B and a second, less significant QTL for pollen viability (Table 3). Thus, LG4 of the F2/WSC family may contain two distinct loci which can have significant effects on male and female gametophyte development.
In contrast to LG7, on LG4 the loci affecting seed set and heading date in the F2/WSC family were genetically separated, with 8 cm between the two QTL maxima. While the exact position of the heading date QTL on LG4 of the F2/WSC family is always likely to be problematic, given the high percentage variance accounted for by the QTL on LG7 in that family, the fact that a QTL was identified in approximately the same position in the ILGI family does support this location. No seed-set QTL was detected on LG4 in the ILGI family, so the position of the seed-set QTL in the F2/WSC family is not corroborated by the two crosses. However, comparative genomic/QTL analysis with rice does indicate some interesting parallels. It has been established that there is a syntenic relationship between ryegrass LG4 and rice LG3 (Jones et al., 2002; Sim et al., 2005), and 20 comparative markers mapped to ryegrass LG4 in the present study could be aligned with the rice 3 pseudomolecule on the basis of physical position after BLASTN alignments (Fig. 2), although the relationship is by no means absolute, as nine comparative markers could not be assigned to the rice 3 pseudomolecule (F2/WSC: Rz395, PSR922, RZ537, CDO1380; ILGI: BCD1421, PSR305, C764, CDO241, PSR922, PSR144). However, using the relationship described in Fig. 2 as a guide, there are a number of seed yield-associated and heading date QTLs that have been mapped to rice LG3 whose position can be inferred on ryegrass LG4. These results suggest that the ryegrass LG4 seed-set QTL may be equivalent to the rice LG3 gn3 (grains/panicle), spkfrt (spikelet fertility) and gy (grain yield/plant, main effect and epistatic interaction) QTLs; additionally, a recent fine-mapping study (Jing et al., 2007) identified that the rice pollen fertility locus, S33(t), was also associated with this region. A similar comparative analysis also indicates that the ryegrass LG4 heading date QTL may be equivalent to the dth3.3 and dthd days to heading QTLs (see Fig. 2 for references). It has been noted previously that ryegrass LG4 shows a degree of synteny with both Triticeae LG4 and LG5 (Jones et al., 2002; Alm et al., 2003; Jensen et al., 2005) and there are a number of QTLs for seed yield-related traits and heading date that map to these chromosomes in both wheat and barley (see Graingenes; Borner et al., 2002; Pillen et al., 2003; Quarrie et al., 2005). Unfortunately, the lack of common markers between the ryegrass mapping populations and many Triticeae experimental populations means that it is not possible to infer the comparative positions of QTLs with any certainty. However, Bins 5.02–5.03 of maize which include the interval CDO795-CDO542 (see Maize Bins QTL 2005/Cornell Wilson 1999 comparison at http://www.gramene.org; Wilson et al., 1999) are associated with a number of grain yield and seed-weight QTLs, and hence a possible comparative relationship may exist (Veldboom & Lee, 1994; Melchinger et al., 1998, Gramene QTL Acc. ID AQFS1064, AQFS976, AQFS1248 and AQFS977).
Figure 2. Diagrammatic representation of the relative genetic positions of markers mapped onto Lolium perenne chromosome 4 with their physical position on the rice 3 pseudomolecule (bold type) in relation to the inferred positions of selected rice LG3 quantitative trait loci (QTLs). Markers in normal type to the left of the rice 3 axis have been associated with the QTLs on the right of the axis (italics) in rice. LOD profiles are for heading date (solid line) and seeds per spikelet (dashed line). ×, genetic position of L. perenne markers not aligned with rice LG3 pseudomolecule. Hd6, heading date 6 (Yamamoto et al., 2000); gy, grain yield per plant (Li et al., 2001); spkfrt, spikelet fertility (Gramene QTL Acc. ID AQCU108) dthd, days to heading (Gramene QTL Acc. ID AQFW130); dth3.3, days to heading (Thomson et al., 2003); gn3, grains per panicle (Xing et al., 2002); gw3.1, grain weight (Thomson et al., 2003); yld3.1, yield per plant (Thomson et al., 2003); f3, hybrid fertility (Wang et al., 1998); gpp3.1, grains per panicle (Septiningsih et al., 2003); spp3.1, spikelets per panicle (Septiningsih et al., 2003); Hd8, heading date 8 (Takeuchi et al., 2003); Hd9, heading date 9 (Lin et al., 2002).
Download figure to PowerPoint
Heading date QTLs on chromosomes 4 and 7
The major QTL on LG7 of the F2/WSC population and its association with the positions of the gene Hd3a and Hd1 has been discussed previously (Armstead et al., 2004, 2005). However, it is worth noting that other studies have also detected a QTL in a similar position (Inoue et al., 2004; Jensen et al., 2005). Interestingly, the study that Yamada et al. (2004) undertook on the ILGI population in Japan identified a QTL on LG4, but they reported no QTL on LG7. This underlines the importance of environmental conditions on the genetic control of flowering time, even when putatively identified major genes are concerned.
Quantitative trait loci on LG4 have also been identified in a number of studies on ryegrass populations (Inoue et al., 2004; Yamada et al., 2004; Jensen et al., 2005) and the closely related grass species meadow fescue (Festuca pratensis) (Ergon et al., 2006). Jensen et al. (2005) demonstrated that the ryegrass equivalent of the Triticeae VRN1 vernalization gene mapped to LG4 and could be associated with a QTL for vernalization responsiveness; Shinozuka et al. (2005) showed that the LpCk2α-1 locus also mapped to LG4 and was associated with days to heading. Previously, the lack of common markers between studies has made it difficult to draw many cross-family and/or cross-environment conclusions. However, both VRN1 and LpCk2α-1 have been mapped on the F2/WSC population and the results indicate the QTL peak on LG4 of the F2/WSC population is not directly associated with these two candidate genes. In addition, comparative mapping with rice indicates that it is probably not equivalent to any of the identified rice heading date QTLs Hd6/OsCk2α (Yamamoto et al., 2000; Takahashi et al., 2001), Hd8 (Takeuchi et al., 2003) or Hd9 (Lin et al., 2002). For the ILGI population, the heading date QTL detected on LG4 in the present study colocalized with the LG4 QTL in the F2/WSC family. However, in the study of Yamada et al., 2004) on the ILGI population with phenotype assessment in Japan, the QTL peak was associated more directly with the putative position of VRN1, indicating a second environmental influence on QTL position in this family. Future work within the Lolium/Festuca complex will allow us to clarify further the genetic basis of heading date determination on LG4.
This study has identified two genomic regions with major influence on reduced seed-setting ability. The first, on LG4, identified as a recessive mutation in a finished variety of perennial ryegrass and which results in almost complete sterility, is an example of the high mutational load that outcrossing perennial ryegrass carries. The second, on LG7, although clearly at least partially recessive in nature, is more difficult to identify phenotypically and requires accurate seed-set determination and, in future, an anatomical investigation of the seed developmental process. Evidence from the only two populations that have so far been studied for seed setting in perennial ryegrass suggests that there is selectable variation at this locus that could significantly increase seed-setting ability in commercial cultivars. Molecular markers for this gene, including those developed in this study, will be useful for identifying favourable allelic variants in these and other populations, with the aim of incorporating them into different breeding populations. In rice, it is likely that the S5n locus is determined by one of five candidate genes (Qiu et al., 2005) and, although we cannot preclude the involvement of the Hd3 gene, the same gene may well be responsible for the seed-setting QTL on ryegrass LG7. Further research to identify allelic variation within the orthologous ryegrass genes, either by direct sequencing or by utilizing an eco-tilling approach (Mejlhede et al., 2006), will prove useful towards both validating the gene/trait relationship and developing molecular markers for selection within ryegrass populations.
The discovery of loci with a major effect on seed setting also has a practical application: much breeding effort has been focused on improving the vegetative characteristics of grasses used for forage (Wilkins & Humphreys, 2003) and ornamental and sports turf. Yet seed yield is also an important trait, as seed growers are reluctant to grow poor-yielding cultivars even if the contracting seed company is willing to pay a premium price for the seed produced. Selection for seed production traits that are dependent on preferable allocation of resources from vegetative to reproductive organs will inevitably be detrimental to agronomic traits determined by good vegetative yield and quality, and so will be negatively correlated with agronomic performance (Wilkins, 1991). However, seed set (i.e. the proportion of florets that produce a seed, sensu strictu caryopsis) and seed retention are two reproductive traits that are independent of vegetative growth performance traits (Wilkins, 1991) which breeders of outcrossing forage crops should focus on.