II. Objectives, breeding methodologies and progress in grass breeding 12
III. Markers and their suitability for genetic analysis in grasses 13
IV. Associating markers to genes for ‘sustainability’ in the grasses 13
V. The genetic control of flowering time 16
VI. Target traits: resilience against climate change and resistance to abiotic stresses 17
VII. Target traits: carbohydrates and sustainable ruminant animal production 19
VIII. Target traits: grass biomass, climate change and energy sustainability 21
IX. Conclusions 21
Here, we review the current genetic approaches for grass improvement and their potential for the enhanced breeding of new varieties appropriate for a sustainable agriculture in a changing global climate. These generally out-breeding, perennial, self-incompatible species present unique challenges and opportunities for genetic analysis. We emphasise their distinctiveness from model species and from the in-breeding, annual cereals. We describe the modern genetic approaches appropriate for their analysis, including association mapping. Sustainability traits discussed here include stress resistance (drought, cold and pathogeneses) and favourable agronomic characters (nutrient use efficiency, carbohydrate content, fatty acid content, winter survival, flowering time and biomass yield). Global warming will predictably affect temperature-sensitive traits such as vernalisation, and these traits are under investigation. Grass biomass utilisation for carbon-neutral energy generation may contribute to reduced atmospheric carbon emissions. Because the wider potential outcomes of climate change are unpredictable, breeders must be reactive to events and have a range of well-characterised germplasm available for new applications.
Crop domestication has been viewed as the single most important event in human cultural development over the last 10 000 years, with grasses in the broad sense being most prominent (Buckler et al., 2001). Until recently, trait selection through both domestication and breeding has concentrated on similar objectives, with the result that grass crops (for example maize, wheat and rice) derived from a common ancestor may share variants of the same genes (Ahn et al., 1993). Despite large variations in genome size and incidents of genome rearrangements, an improved understanding of conserved syntenic relationships between grass species has given rise to the ability to transfer knowledge at the genomic level across species. However, targets for crop development are now changing, in part a consequence of the success in the western world of crop production, but also the result of the negative impacts that industrial and agricultural practices have had on the environment. New opportunities arise for exploitation of novel traits in grass design to redress these anthropogenic effects. Such traits may be derived from the native forage grasses that comprise vast areas of our agricultural land and are closely related to the cereals.
Agricultural land in the UK covers over 18.4 Mha, which represents 76% of the total land area. Most of this (65%) is grassland, and the largest part of this (31%) is the rough grazing that predominates in the cool, wet uplands of the North-west. The short growing season and acid soils support vegetation of low forage value like Molinia, Deschampsia, Nardus and Festuca. Permanent grasslands over 5 yr old account for c. 29% of total land use and consist of mixtures of sown species and volunteer grasses such as Poa, Agrostis and Holcus. High-quality grasslands under 5 yr old occupy only c. 7% of the total land and have been sown primarily with Lolium (ryegrass) species (considered the ideal grasses for European agriculture), with or without Trifolium repens (white clover) (http://www.iger.bbsrc.ac.uk/SAGES2/sages2.html). These permit the highest levels of grass and livestock production, and are also concentrated in the maritime climates of the western part of the UK, where rainfall is generally well distributed, soils have good water-holding capacity, and growth is not limited by environmental extremes. Although Lolium species are the most important grasses used for re-seeding, other species are sown including Dactylis glomerata, Phleum pratense, Festuca pratensis and Festuca arundinacea, but these account for only 6% of total grass seed usage in the UK. However, they are used to a much greater extent in countries with more extreme weather patterns in Northern Europe, the USA and Japan, because they are more tolerant to abiotic stresses than Lolium. Some of the climate change scenarios predicted for the UK could significantly affect Lolium-based grassland systems which lack resilience to abiotic stresses. Subject to their capabilities for plant hybridisation (Humphreys MW et al., 1998), or alternatively a change in public acceptance that enables use of transgenic technologies, they could provide allelic variants for improved resistance in Lolium.
Although rarely out of today's news, the concept of sustainability for agricultural systems in the UK is not new. It has long been recognised that a balance must be sought between management practises that encourage agricultural production and those that safeguard its continuity. The importance of growing forage grass in combination with legumes to maintain fertility was recognized as early as the 17th century (Blith, 1652). The economic value of grasslands per se was not considered before Stapledon & Davies (1948) and Davies (1960) at the Welsh Plant Breeding Station (now IGER). Then the need for the UK to be self-sufficient in food production during and immediately following the Second World War led to increasing use of nitrogen and herbicides to maximise forage yield. Increasing dry matter yield (DMY) was the major objective of grass breeding programmes, and mean productivity of forage grass cultivars increased by 0.5% per year, with the best varieties in recent years yielding 20–30% more DMY than varieties bred 20 to 30 years ago (Sodin, 1991). However, intensive systems resulted in overproduction of milk and beef and led to excessive costs, pollution of soil, water and atmosphere, loss of biodiversity, loss of genetic diversity within the grassland crop species itself, and an overall reduction in landscape quality (Nö sberger & Staszewski, 2002), thus compromising the environmental balance. Although yield remains important, forage quality is becoming an ever greater selection criterion (Humphreys MO & Theodorou, 2001) in recognition of the need for production of healthy and safe food for the consumer. Food shortages in the UK are unlikely in the foreseeable future, and changing governmental priorities that put increased emphasis on the delivery of grassland systems, providing affordable safeguards for the environment in addition to the European Union's Common Agricultural Policy (CAP) reforms and withdrawal of farm subsidies, have led to changes in grassland research priorities.
Forage and turf grasses for the future must utilise resources (nutrients and water) more efficiently and must also confer measurable benefits in terms of environmental quality. A holistic approach in grass research and development must take account of impacts on biodiversity, including those within the soil and rumen microbe populations. Improved efficiency in energy and protein transfer between grasses and the feeding animal not only benefits animal production, but also the environment, through reduced N pollution and improved air, soil and water quality. Grassland management must be devised to enhance amenity value and maintain an attractive landscape as well as supporting efficient agricultural output. Farming systems need to be viewed as desirable ecosystems rather than food factories, and studied in a way where the sum of the biologically active chemicals can be monitored and influenced. There is a need to understand complex interactions of crop ecosystems in terms of fluxes of energy, resources and genetic information. At a fundamental level, ecosystems obey the laws of thermodynamics, so we need to understand the bioenergetics of light interception, conversion into biomass, partitioning of that biomass into useful products, and subsequent degradation.
In addition to requirements for improved grassland sustainability, plant breeders and researchers need to take account of the impacts of climate change. A recent international symposium on the stabilisation of greenhouse gases at the Meteorological Office of the Hadley Centre (2005) highlighted these impacts, including some that may actually be beneficial to agricultural production in mid- and high latitudes. However, overwhelming evidence is accumulating that indicates the overall dangers to world ecosystems, including some new disturbing evidence that the oceans are increasing in acidity, reducing their capacity to remove CO2 from the atmosphere and thereby affecting the entire marine food chain. EU figures for European grasslands show that the 2003 droughts had an extremely high impact on 17%, high impact on 46% and moderate impact on 37% of yields of all its pasture areas. In the UK, the whole of Wales and central and southern England constituted areas of high loss of forage yield.
The UK climate is predicted to include warmer, wetter winters and hotter, drier summers, with increasing frequency of extreme weather patterns (Harrison et al., 2001). The outcomes include water surplus during periods of extreme rainfall and water deficit during the summer months. Water impacts on many aspects of environmental systems, encompassing drought, efficiency of water usage, water pollution, saline water, flooding, interactions with high and low temperature, soil, etc. Because grasslands comprise such large areas of UK land, the differing abilities of species or ecotypes to withhold or release water, and the effects they have on soil structure and stability, can and will play a significant part in the UK's abilities to withstand climate extremes. There is also a need for further understanding of the grass plant's response to climate change (including pests and diseases) and to adopt appropriate crop/management systems. Information is required on factors that make crops more tolerant to changing conditions (temperature, drought, pests, disease, etc.). The impacts of changes in winter temperatures on flowering time in grass species or even their capability for flowering, given that a prolonged low temperature is a prerequisite for successful vernalisation, are still unknown. With warmer winters, grass genotypes may need to be designed for greater dependency on photoperiod response rather than low temperatures to stimulate the onset of flowering (Harrison et al., 1997). For grass breeders, geneticists and physiologists, this represents a major challenge but also an opportunity to design grasses capable of withstanding these climate perturbations. The nature of these challenges will depend on the impact and speed of climate change which could be dramatic (leading to serious food shortages) or gradual (allowing time for geneticists and plant breeders to respond).
Informed use of biotechnology requires basic knowledge of relevant traits and their genetic, biochemical and physiological characteristics. The grasses within the Lolium–Festuca species complex contain an exceptionally wide range of variation, including adaptations to contrasting temperatures, day length, rainfall and soils, thus providing a rich pool of desirable traits. Interspecific and intergeneric recombination rates in Lolium–Festuca are the highest amongst monocot crops, allowing introgression to be exploited as a highly efficient research and breeding tool, and assisting in exchanges of genomic information across all monocot species (King et al., 1999; Humphreys MW et al., 2003). The obligate outbreeding habit of Lolium and Festuca species has resulted in highly heterogeneous genetic backgrounds which undoubtedly can, and frequently will, affect the efficacy of gene expression following transfer from one species to another during introgression breeding programmes. The recent development of reliable diagnostic polymerase chain reaction (PCR)-marker-based tools allows for more precise targeting of genes in donor species and their transfer to Lolium through introgression over generations for both simple (Moore et al., 2005) and complex (Humphreys J et al., 2005) traits. Festuca-derived alleles that could reduce both forage production and quality can simultaneously be excluded (Humphreys MW et al., 2003). These developments offer the prospect that, despite their outbreeding and heterogeneous nature, future breeding and design of grass cultivars with improved sustainability will be precise, predictable and effective.
As a response to the changing climate, both political and actual, some of the developments in research that are aimed at safeguarding the future of the UK's grassland agricultural industry are reviewed below. We describe progress in grass breeding, a review of the development of marker technologies and their suitability for targeting genes that can contribute to improved sustainability and then be assembled in a more predictive manner, and selected target traits associated with improved resource use efficiency and resilience against stresses of climate change.
II. Objectives, breeding methodologies and progress in grass breeding
The breeding of forage grasses began more recently than in most other agricultural species, and seeds of Lolium varieties that were superior to traditional commercial strains or to wild populations were not available to UK farmers until 1939. In Germany, farmers continued sowing natural populations of L. perenne as late as the 1980s (Spatz et al., 1987). Systematic forage grass breeding in the UK began at the Welsh Plant Breeding Station in 1919 with the production of the first L. perenne variety S-23 in 1931. Since then, the Station has released a large number of grass varieties bred to have improved persistence, yield and forage quality.
Most gains in grass breeding have been achieved by combining desirable genes through sexual recombination and selection. Grasses are predominantly outbreeders; therefore, population improvement based on some form of recurrent selection of individual spaced plants or their half-/full-sib families is the main means for their improvement. In the last decade, the breeding methods used in grass breeding have been extensively reviewed (Vogel & Pedersen, 1993; Casler, 2000; Wilkins & Humphreys MO, 2003). In addition to recurrent selection, tetraploidy is another method used widely in Lolium breeding. Chromosome doubling of parents by colchicine treatment has been used to create tetraploid Lolium and Festulolium varieties, and to facilitate introgression from Festuca species into Lolium (Humphreys MW, 1989; Humphreys MW & Pašakinskienė, 1996). Stable hybrid Lolium varieties are produced by crossing chromosome-doubled L. multiflorum and L. perenne (Jones & Humphreys MO, 1993). Such hybrids combine the good characteristics of L. multiflorum (e.g. rapid establishment and early growth characteristics) and L. perenne (persistency, stress tolerance and leafiness). Combining the forage quality of Lolium with the better stress tolerance and persistency of related Festuca species has been a goal for plant breeders for a number of years. Cultivar development has thus far harnessed entire genomes of Lolium and Festuca species. Hybrid fertility has been achieved through amphiploidy by chromosome doubling, thereby encouraging homologous chromosome pairing and disomic inheritance (Breese et al., 1981; Zwierzykowski et al., 1993). However, these hybrids tend to be unstable (Canter et al., 1999), leading to increased sterility, and can incorporate gene combinations that reduce forage quality. Stable tetraploid hybrids between F. pratensis and L. perenne (Lewis, 1983) are grown successfully in Sweden, and hybrids between F. pratensis and L. multiflorum (Zwierzykowski et al., 1993) are cultivated in central Eastern Europe and USA.
Until the 1980s, the main aim of grass breeding was to improve persistency and DMY. Breeding quality traits into Lolium varieties was made possible with advancements in techniques to measure traits such as in vitro dry matter digestibility (Jones & Hayward, 1975). Developments in near-infrared spectroscopy (NIRS) enabled analysis of herbage samples for a number of quality traits simultaneously (Brown et al., 1990). A recent report (Wilkins & Lovatt, 2004) suggested that water soluble carbohydrate (WSC) concentration, which can improve the efficiency of energy and protein utilisation in ruminant animals, has been increased by 37 g kg−1 whilst also increasing DMY by 13% in L. perenne varieties bred between 1991 and 2000. The study also reported increases in WSC (35 g kg−1) and DMY (10%) in tetraploid hybrids developed between the years 1988 and 1994. In addition to improvement in DMY, persistency and quality aspects, traits for efficient use of inputs (e.g. nitrogen) and a long growing season are also being bred into modern grass varieties (Wilkins & Lovatt, 1989; Wilkins et al., 2000).
Technological developments in the last 20 to 30 years have also played a vital role. For example, the plot harvester (http://www.haldrup.dk) has enabled direct selection for yield and persistency, and computing technologies have enabled the application of NIRS to herbage analysis. A prototype of a field scanner known as IMSPECTOR recently developed at Plant Research International at Wageningen in the Netherlands (http://www.plant.wur.nl) has the potential to measure herbage quality traits together with a range of other agronomic traits under field conditions. Genetic maps and quantitative trait locus (QTL) analysis including linkage disequilibrium (LD) can locate genes and linked markers associated with important agronomic traits to facilitate QTL introgression and selection in crop breeding programmes. Lolium genetic maps are now well aligned with maps of other species in the Poaceae (Jones et al., 2001) and are being used in comparative analysis to identify gene orthologues in other crops such as rice and Arabidopsis (Armstead et al., 2004; Cogan et al., 2005; Jensen et al., 2005). Better markers, together with high-throughput genotyping technologies and improved marker assisted selection (MAS) (Knapp, 1998; Morgante & Salamini, 2003; Peleman et al., 2003; Yadav et al., 2003) and backcrossing methods (Frisch & Melchinger, 2001), will increase the scale and precision of marker implementation in grass breeding. MAS will particularly improve the efficiency of conventional plant breeding in situations where the traits to be selected are difficult or expensive to evaluate, where the traits are expressed late in the growth cycle, when several genes controlling the same character need to be pyramided, when numbers of traits are needed to be improved simultaneously, or when the traits to be selected are controlled by recessive alleles. Most traits that we manipulate in our modern grass breeding programmes fall into one or more of these categories, and advances in genomics will lead to improved efficiency of forage and amenity grass breeding.
III. Markers and their suitability for genetic analysis in grasses
A wide range of marker types is available for genetic mapping and most have been employed with temperate forage grasses (Xu et al., 1995; Hayward et al., 1998; Bert et al., 1999; Jones et al., 2002a). Each type has different advantages and disadvantages. The main characteristics that need to be considered are ease of use (amount of DNA required and length of procedure), transferability across mapping families and species in order to permit alignment with other maps, and whether markers are dominant or codominant. The earliest markers were isozymes: these require no DNA, are robust, codominant and transferable across species, but are severely restricted in numbers. Restriction fragment length polymorphisms (RFLPs) are good anchor probes for map alignment and are codominant, but require large amounts of DNA and are time-consuming to use. PCR-based markers such as random amplification of polymorphic DNA (RAPDs) and amplified fragment length polymorphisms (AFLPs) are quick and easy to use and only require small amounts of DNA. However, they are dominant markers and have poor transferability across mapping families. Simple sequence repeats (SSRs; also called microsatellites) and tagged sites (e.g. sequence tagged sites (STSs), expressed sequence tags (ESTs) and sequence characterised amplified regions (SCARs)) which may be associated with a known function are high-throughput PCR-based markers requiring small amounts of DNA and are usually codominant. Single nucleotide polymorphisms (SNPs) are nowadays the markers of choice because they are allele-specific.
IV. Associating markers to genes for ‘sustainability’ in the grasses
1. Mapping genes for complex traits: QTL mapping
Genetic markers for traits of interest offer the opportunity to increase the precision of selection and therefore the rate of improvement in breeding programmes. One approach to identify such markers is QTL mapping. The first requirement for QTL analysis is a genetic linkage map. Linkage mapping of the major human food crops is well advanced. Most of the important species are self-fertile, have been highly inbred and consequently have a relatively narrow genetic base. It is easy to produce well-defined F2 or recombinant inbred line (RIL) mapping families. In contrast, most temperate pasture grasses are outbreeders. This has the advantage that it is generally easy to find a large range of available variation for traits of interest. The disadvantage is that the production of suitable mapping families is more difficult. For this reason, linkage mapping and QTL analysis in forage grasses has lagged behind work in the related cereals. However, considerable progress has been made in recent years, particularly within Lolium and Festuca. Advances in marker technologies and mapping software mean that there is now a choice of mapping family structure suitable for work with outbreeding grasses. However, these vary in the ease with which they can be produced, the marker types which are useful with them and the purposes for which they are most suited.
Two-way F1 pseudo testcross families (paircrosses: type CP in some mapping software) are the easiest to produce. Two unrelated highly heterozygous parents are crossed to produce an F1 family. Self-incompatibility systems are unlikely to be a problem, so seed set is good and the family will contain a large number of individuals. Furthermore, segregation distortion is uncommon. However, construction of the linkage map is complex because up to four alleles may be segregating in the family at any given locus. Co-dominant markers are therefore more informative than dominant markers, but as increasing numbers of high-throughput PCR-based codominant markers become available this becomes less of a problem. QTL analysis is less straightforward than other mapping family structures and does not provide information on dominance, but is now possible with recently developed software. Linkage maps for Lolium and Festuca species have been published based on this type of family (Alm et al., 2003; Inoue et al., 2004a; Warnke et al., 2004).
If a homozygous or near homozygous tester is available then a one-way F1 pseudo testcross family can be produced. Doubled haploid testers are available for L. perenne and this is the mapping family type employed by the International Lolium Genome Initiative (ILGI) Lolium reference map (Jones et al., 2002b). Dominant markers are informative for linkage mapping with this family structure and this allowed early use of PCR-based markers (Bert et al., 1999). Although the variation in traits may be more limited than with an F2 family, QTL analysis is straightforward. However, dominance is not determined and family members are not suitable for further crossing to test effects. An F2 mapping family structure has advantages for QTL analysis and has been employed with Lolium and Festuca species (Chen et al., 1998; Armstead et al., 2002). Such families are not easy to produce because self-incompatibility problems can limit the success of selfing the F1 generation to produce the F2. One approach used to overcome this has been to partially inbreed contrasting lines in order to select parents with some resilience to selfing for the F1 cross (Armstead et al., 2002). This can lead to higher levels of segregation distortion in the mapping family, but this may be no more of a problem than with one-way F1 pseudo testcross families. Dominant markers are not informative on F2 families, so work was previously limited by the requirement for time-consuming RFLP analysis. However, the higher levels of recombination observed in F2 families (Armstead et al., 2002) and the ability to determine dominance increase the efficiency of and knowledge gained from QTL detection. Family members are also suitable for direct crossing to test effects.
With recent advances in technology, genotyping is quick, but phenotyping can still be very time-consuming. It is necessary to determine quickly and easily the measured parameters that represent the trait of interest so that large numbers of plants can be assessed in a short period of time. Samples for later analysis must be rapidly taken and frozen. It may also be necessary to consider how the trait of interest is affected by the environment. If the expression of a trait is highly dependent on environment, then phenotyping must be carried out in different environments in order to gain a full picture of the genetic control of the trait.
Most current QTL analysis software packages offer QTL detection procedures on different levels of complexity. Single locus analysis does not require any linkage map information and is performed by one-way analysis of variance or the Kruskall–Wallis rank sum test. It associates individual markers with QTL effects. Simple interval mapping constructs a ‘QTL likelihood map’ and requires a reliable linkage map, but this need not necessarily have a high density of markers in all regions. Composite interval mapping can refine the position of the QTL by using information from cofactors, but does require a denser linkage map. Most groups working with Festuca and Lolium currently have maps suitable for composite interval mapping (e.g. Van Loo et al., 2003; Armstead et al., 2004; Yamada et al., 2004).
The earliest attempts to identify QTL in forage grasses were published over 10 years ago (Hayward et al., 1994). Although these were based on a partial linkage map, the presence of isozyme loci close to some regions of interest may allow approximate alignment of QTL against the more complete maps currently available. However, relatively few QTL studies were subsequently published, although numerous groups have reported preliminary data at meetings (Ghesquiére et al., 2001; Humphreys M & Turner, 2003; van Loo et al., 2003; Rognli et al., 2003). Inoue et al. (2004b) found QTL for morphological traits relating to lodging in L. multiflorum. QTL for morphological traits have also been reported in L. perenne (Yamada et al., 2004), along with some information on physiological traits such as winter hardiness and WSC content. The next few years should see the considerable advances being made in this field appearing in the literature.
Once the positions of relevant QTL are known, markers can be identified for selection purposes. These markers may be random or informative markers such as RFLPs and SSRs underlying the QTL. Crosses can be designed to test the effects of the QTL on phenotypes. Such marker-selection crosses have been carried out in L. perenne with some success (Humphreys M & Turner, 2003; Humphreys M et al., 2003; van Loo et al., 2003). The resulting differences in phenotype may be small if selections are based on a single QTL and if segregation occurs at other regions of the genome controlling the trait. Selections based on QTL indices have showed greater effects. When random markers are used to characterise QTL regions, it is advisable to identify fairly large regions of chromosome to be sure of transferring the trait of interest. In future, it may be desirable to identify candidate genes under QTL to use as markers. This would allow selections based on smaller regions of the chromosome and would minimise the possibly of undesirable transfer of linked traits. In this context, the synteny of species within the Poaceae will prove useful. The ILGI reference map is aligned with the Triticeae consensus map and hence with rice. Workers have already started to make use of the resources available in the rice genome (Armstead et al., 2004, 2005). In addition, gene-derived functional markers (Anderson & Lübberstedt, 2003; Faville et al., 2004; Saha et al., 2004) are becoming available in Lolium.
The limitation of QTL analysis for characterising traits of interest in outbreeding species like Lolium and Festuca comes from the narrow genetic base of mapping families. Any one mapping family can only describe a fraction of the variation available in the species. Different regions of the genome may be identified as controlling a given trait in different mapping families of the same species, even if, in a given cross, one QTL appears to explain a large part of the variation. This is clearly demonstrated by considering heading date QTL in Lolium, where QTL have been identified on linkage group 4 (Yamada et al., 2004), linkage groups 6 and 7 (Inoue et al., 2004b), linkage groups 2, 4 and 7 (Armstead et al., 2004), and linkage groups 2, 4, 6 and 7 (Jensen et al., 2005). Other approaches, such as introgression mapping and the recent development of association mapping, may be better at providing means to access the full range of variation available.
2. Mapping genes for complex traits: introgression mapping
A concern for many crops, although not the grasses described herein, is that much of the genetic variation for improving abiotic stress tolerance has been lost during domestication and selection in modern breeding, leaving only pleiotropic effects of a relatively small number of genes for adaptation. Fortunately, the outbreeding nature of the grasses and the availability of highly heterogeneous ecotypes provide us with a vast array of genetic variation, including adaptations to most climatic and edaphic conditions encountered throughout Europe and across many other regions of the world. Gene introgression from species of related grasses, which are capable of interpollination and of retaining fertility as hybrids, means that this range of genetic variation can be dissected and manipulated. Lolium–Festuca hybrids are highly amenable to introgression mapping to identify major gene complexes involved in trait expression (Humphreys MW et al., 2003), and some major advances in breeding for stress resistance have been achieved in the forage grasses. Chromosome 3 of these species has received considerable attention. King et al. (2002) demonstrated the use of the high recombination frequency in these species to produce a recombination series for F. pratensis chromosome 3 and combined physical and linkage mapping to locate sites of genetic markers along the entire chromosome length.
3. Mapping genes for complex traits: association analysis
The development of high-throughput technology for molecular marker analysis over the last 10 years has made feasible association analysis for the dissection of complex traits of agronomic importance. This approach avoids the time and effort needed to produce mapping populations and uses populations without prior knowledge of pedigree. It exploits the accumulation of mutation and recombination events over all the generations since a particular association appeared in the population. This affects the level of resolution achievable; it has been estimated that in highly diverse maize populations QTL can be mapped with 5000-fold higher resolution than F2 populations (Thornsberry et al., 2001). Furthermore, a range of alleles can be assessed, representing a wider range of genetic variation. Association mapping has been most extensively used in human genetics to identify genes responsible for major diseases, principally because of the lack of large mapping families. There have been notable successes in identifying and cloning major genes responsible for diastrophic dysplasia (Hästbacka et al., 1994) and cystic fibrosis (Kerem et al., 1989), amongst others. The marker density available from the several million human SNPs makes it technically possible to undertake association mapping of QTL using whole-genome scans (Goldstein et al., 2003), although there are still issues concerning the power of such analyses (Carlson et al., 2003; Goldstein et al., 2003).
Association mapping relies on LD, which is the nonrandom distribution of alleles at different loci. If two loci A and B are bi-allelic, with alleles A and a at the first locus and B and b at the second, then LD between the two loci can be described as DAB = pAB–pApB, where pAB is the observed haplotype frequency and pA and pB are the respective allelic frequencies. The LD is the difference between the observed and expected haplotype frequencies. (For discussion of statistical significance as relates to LD measurements, see Hill & Robertson, 1968; Weir, 1996; Hedrick, 2000.) LD between two allelic variants occurs if they are physically so close that recombination between them is rare, even over many generations, resulting in a common history of mutation and recombination (Flint-Garcia et al., 2003). The recombination rate is therefore important in determining the extent of LD between two loci. The higher the recombination rate (i.e. the larger the genetic distance between two loci), the faster the LD decays. The breeding system of a species also influences the rate of decrease in LD. Because the effective recombination rate is lower in inbreeding than in outbreeding species, one would in general expect LD to extend further in inbreeding species (Hedrick, 2000). In plants, LD has been most studied in maize and in Arabidopsis thaliana. In the former species, LD decayed within 1500 bp in diverse inbred lines or germplasm (Remington et al., 2001; Tenaillon et al., 2001), but extended further in some loci examined in elite inbred lines (Ching et al., 2002; Jung et al., 2004). This has been attributed to differences in the diversity of the germplasm studied, to differences in selection pressure between the different loci studied, and to variation in recombination frequencies in different parts of the genome (Flint-Garcia et al., 2003; Gaut & Long, 2003). Two studies show that LD in A. thaliana decays within approximately 250 kb (Hagenblad & Nordborg, 2002; Nordborg et al., 2002). The more extensive LD observed in this species compared with diverse maize germplasm is consistent with its inbreeding habit. Other evolutionary forces, (such as genetic drift, selection, mutation and gene flow or population admixture) also influence LD, either directly or indirectly. Genetic drift becomes important in small populations, leading to the loss or fixation of alleles, and this increases LD (Weir, 1996; Hedrick, 2000). While mutations are sources of new LD, selection and population admixture can both act to maintain LD, and this can occur whether or not the loci are linked.
There are two main strategies for association or LD mapping: whole-genome scans and candidate gene approaches. Whole-genome scans rely on a large number of markers evenly spaced throughout the genome. Estimates of how many markers are needed in whole-genome scans vary from 130 000 to more than 700 000 to be sure of finding associations (Goldstein et al., 2003; Carlson et al., 2003). In genomes where LD extends for large distances, fewer markers are needed to detect associations, but the downside is that the resolution becomes poorer, making the approach unsuitable for fine-mapping. A number of association analyses have been carried out in grass species based on molecular marker data covering the whole genome. Virk et al. (1996) found associations between six morphological traits and RAPD markers in rice, and in wild barley Pakniyat et al. (1997) found associations between AFLP markers and salt tolerance, whereas Kraakman et al. (2004) used modern spring barley cultivars for LD mapping of AFLP markers associated with yield and yield stability. In L. perenne, two studies by Skøt et al. (2002, 2005a) demonstrated association of AFLP markers with cold tolerance and flowering time, respectively, and three of the markers showing significant LD with flowering time could be mapped close to a major flowering time QTL on chromosome 7. Skøt et al. (2005b) also analysed the extent of LD that exists within a locus encoding an alkaline invertase gene in L. perenne. The pattern of LD over this 5500 bp locus shows a decline to an r2 value of less than 0.2 within approximately 1000 bp (Fig. 1). If this is a general pattern in L. perenne, this species would appear to have similar short-distance LD to that found in diverse maize germplasm or diverse inbred lines (Flint-Garcia et al., 2003). More work is needed to ascertain how general this pattern is.
Currently, work is underway to evaluate candidate gene approaches in LD mapping in L. perenne and other outbreeding species. In this type of association mapping, candidate genes are selected on the basis of their likely involvement in the trait of interest. The selection can be based on biochemical and physiological knowledge of trait function, perhaps in combination with QTL mapping and/or expression data from functional genomic analyses. Such an analysis has the potential to identify functional allelic variants responsible for some of the genetically based variation in the phenotype (see Pritchard et al., 2000; Flint-Garcia et al., 2003). This approach is still in its developmental stages in plant species, but has already proved to be useful in analysing two different maize populations (Thornsberry et al., 2001; Wilson et al., 2004), and has the potential to be highly useful in forage grasses.
V. The genetic control of flowering time
Plant phenology, especially flowering time and extent, has probably more impact than all other traits on grassland production (both forage and seed), forage quality and persistency. It is also one of the most sensitive traits to changes in climate or stress. The main cues for heading are day length and temperature. In order to achieve a fully saturated flowering response, most temperate perennial grasses require these cues in the form of a primary induction of short days and cold temperatures (vernalisation) followed by a secondary induction of longer days and higher temperatures. There is, however, a great deal of variation in these requirements particularly relating to the latitude of origin of a genotype (Heide, 1994; Aamlid et al., 2000). The control of flowering time is both important for seed production, where it is essential in outbreeders to coordinate flowering for effective pollination, and in forage production, where delayed flowering is desirable to extend the period for vegetative growth. In this case, vegetative meristems produce increased leaf material and the delayed flowering time reduces the number of flowering heads. For reviews of this subject, see Izawa et al. (2003), Putterill et al. (2004) and Laurie et al. (2004).
An understanding of the genetic control of flower induction in dicots comes from work on the model species Arabidopsis. This was initiated by Koornneef et al. (1991), who identified 11 loci involved in promoting late flowering under long day conditions of an early-flowering ecotype. Orthologues of some of the genes described by this and later studies have been identified in rice and other grass genomes, and analogous modes of action have been determined experimentally, particularly for some of the genes involved in the photoperiodic induction of flowering (reviewed by Hayama & Coupland, 2004). QTL associated with the genetic map positions of some of these orthologous genes have been identified both in rice (Yano et al., 2000) and in forage grasses (Armstead et al., 2004, 2005).
Vernalization-dependent flowering has also been studied extensively in Arabidopsis (for reviews, see Henderson et al., 2003; Amasino, 2005), though the physiological parallels with monocot species are less clear than with photoperiodic induction and orthologues of key regulatory genes in the Arabidopsis vernalisation pathway (e.g. FLOWERING LOCUS C (FLC), FRIGIDA (FRI) and VERNALIZATION 2 (VRN2)) have yet to be identified in grass species (Laurie et al., 2004). Two vernalisation-related (VRN) genes (unrelated to the Arabidopsis VRN genes) have been positionally cloned in cereal species (Yan et al., 2004), and alleles of these are associated with the major vernalisation effects in wheat and barley and have been shown to cosegregate with major QTL for vernalisation in cereals and forage grasses (Laurie, 1997; Börner et al., 1998; Jensen et al., 2005). Thus, while comparative studies on vernalisation between dicots and monocots are, at the moment incomplete, direct comparisons can be made between moncot species which are informative to both forage grass and cereal geneticists.
In addition to the genes involved in vernalisation and photoperiodic induction, there are many other genes which contribute to the process of successful flowering (Goto et al., 2001; Bommert et al., 2005). As an example, Jensen et al. (2001) identified a TERMINAL FLOWER 1-like gene from L. perenne (LpTFL1) and demonstrated disruption of wild-type flowering phenotype, complementation of a tfl-1 mutant and tissue specific expression in Arabidopsis. In this fashion, as our ability to transfer knowledge from model species to crop species and from dicot to monocot increases, many other genes will be validated and become targets for the manipulation of grass flowering phenotypes.
VI. Target traits: resilience against climate change and resistance to abiotic stresses
Recent developments in understanding the genes that underpin control of abiotic stress resistance in forage grasses have been reviewed extensively elsewhere (Alm et al., 2005; Humphreys MO & Humphreys MW, 2005; Yamada et al., 2005) and therefore are not described in detail here. The emphasis here is on progress in the design of a grass genotype with increased resilience following the onset of summer or winter stresses. Plants respond to environmental change as individuals through phenotypic plasticity and in populations through selection and associated evolutionary processes. It is not always easy to determine the genetics underlying adaptive processes because environmental factors may be complex or not clearly defined. Several genes may be involved in a response to a given factor or the same gene(s) may be involved in different adaptive responses. Traditional methods of evaluating phenotypic plasticity in terms of genotype–environment interactions (e.g. Finlay & Wilkinson, 1963) equate ‘environment’ to the mean performance of all genotypes in response to a particular site, management, or year. Yield of individual varieties is plotted against a measure of ‘environment’ and yield potential defined as the highest yield in the best ‘environment’. Yield stability of a particular genotype should be measured across environments and may be equated to stress resistance (Thomas, 1997).
Drought is a complex phenomenon, but an overriding requirement is that the crop should survive and re-grow rapidly when autumn rains set in. However, of equal importance from an agronomic perspective is the demand that yields should not be reduced greatly during mild drought, a response that occurs in many Mediterranean grasses such as F. glaucescens (Humphreys MW et al., 1997), which become ‘quiescent’ at the onset of drought. Drought resistance results from a combination of traits that are not all independent from one another and a balance must be sought depending on the severity of the stress. Plants need to adjust transpiration to absorption. Transpiration is increased by leaf area, whereas water absorption is increased by root depth. Leaf area cannot be compromised because a minimum productivity is required during drought. Hence, maximum root depth should be combined with optimum leaf expansion and good control of water loss per unit leaf area (via the cuticle and stomata). As targets for grass breeding, four main groups of traits were identified in order of importance: (a) floral phenology, which determines indirectly the amount of growth devoted to roots and the density of vegetative tillers; (b) root depth and water status; (c) leaf production and extension; and (d) regulation of transpiration (Humphreys MW et al., 2004; http://www.iger.bbsrc.ac.uk/SAGES2/sages2.html).
To survive the winter, a plant must evolve mechanisms whereby sensitive tissues can avoid freezing or undergo cold hardening compatible with the normal variations of the local climate, coordinate the induction of the tolerance at the appropriate time, maintain adequate tolerance during times of risk, and properly time the loss of tolerance and resumption of growth when the risk of freezing has passed (Guy, 1990). Development of winter hardiness requires exposure of plants to low nonfreezing temperatures, typically 0–10°C, and shortened photoperiod. Many physiological and biochemical changes occur during cold acclimation (CA), including slowed or arrested growth, reduced tissue water content, altered cell pH, protoplasm viscosity and photosynthetic pigments, reduced ATP levels (Levitt, 1980), transient increases in ABA (Chen et al., 1983), changes in membrane lipids (Uemura & Steponkus, 1994), accumulation of compatible solutes including proline, betaine, polyols and soluble sugars, and accumulation of antioxidants (Tao et al., 1998). Considerable resources are necessary to sustain and protect plant metabolism under low temperature stress, and for recovery subsequent to the onset of more benign growth conditions. Temperate grasses store fructans, a soluble polymer capable of rapid polymerization and depolymerization. The partitioning of solutes is important because survival from freezing depends on survival of apices, particularly the lateral buds rather than mature leaf tissue (Eagles et al., 1993). The rate and extent of de-hardening is also critical, and in the UK plants are frequently compromised owing to unpredictable temperature fluctuations (Gay & Eagles, 1991). Eagles (1989) suggested that the nature of an adaptive CA process would vary with the stability and predictability of winter conditions in a particular environment. In stable and predictably cold continental climates where the onset of freezing temperatures is rapid, a photoperiod-triggered and rapid acclimation process is desirable, whilst in the more variable and less severe conditions of a maritime climate such as the UK, a temperature-dependent response might enable plants to exploit a mild autumn or spring by continuing to grow. With the event of climate change and warmer winters, this conclusion might change, and a strategy for winter survival based more on response to photoperiod rather than low temperature might gain importance as a target trait.
CA and freezing tolerance are the result of a complex interaction between low temperature, light and photosystem II (PSII) excitation pressure. At low temperatures, plants have two principal difficulties. The first is maintenance of cell membranes in a fluid state. This can be compromised further by ice formation (Thomashow, 1999). The second problem relates to thermodependency of photosynthetic electron transport and carbon fixation, which are slowed at low temperature (Guy, 1990). The PSII reaction centre is the key site for regulation of light energy and also the main site of photoinhibitory damage. The redox state of PSII reflects fluctuations in the photosynthetic energy balance and so acts as a sensor of any environmental stresses that disturb that balance. Changes to the redox state of PSII, triggered by a low temperature shift, have been proposed to be one of several temperature-sensing mechanisms involved in CA (Rapacz, 2002). Recent evidence has indicated a clear relationship in Lolium–Festuca hybrids between energy dissipation before winter through a lower maximum quantum yield (FV/FM) of PSII and improved winter survival (Rapacz et al., 2004). Furthermore, CA-derived energy dissipation through nonphotochemical quenching (NPQ) appears common to a number of Festuca species and its expression relates to their freezing tolerance (M. Rapacz, Agricultural University of Cracow, Poland, pers. comm). On the other hand, NPQ appears unaltered in Lolium species grown under the same CA conditions and this may in part explain its lower freezing tolerance.
The first QTL analysis for drought-resistance and winter-hardiness traits in Lolium or Festuca has recently been produced (Alm et al., 2005). This involved use of a F. pratensis molecular map (Alm et al., 2003), and the conclusions drawn largely support those derived from alternative introgression breeding approaches. For example, QTL for survival and recovery following severe drought were found along the entire length of Festuca chromosome 3, a source of Festuca genes that have led to an enhanced expression for this trait once transferred to Lolium (Humphreys MW & Pašakinskienė, 1996; Humphreys J et al., 2005). Chromosome 3 of Lolium and Festuca species shares considerable synteny with rice chromosome 1 (Jones et al., 2002b), with orthologous markers distributed at the same location along the majority of each chromosome. Rice chromosome 1 (http://www.gramene.org) includes QTL for root development traits and for osmotic adjustment, traits known to contribute to the drought resistance of Festuca species compared with Lolium (Thomas, 1997). Lolium has several QTL associated with forage quality located on chromosome 3 (Cogan et al., 2005) which could imply some deleterious affects on forage digestibility following gene transfer from Festuca species. However, the intergeneric drought-resistant introgression lines developed at IGER have shown no compromise (or, in one case, only a little) to forage quality, and indeed in that case there was an increased contribution to forage yield resulting from increase in leaf width (M. W. Humphreys, unpublished results). Yamada et al. (2004) described coincident QTL for plant height, tiller size and leaf length on Lolium chromosome 3. The rice SD1 semidwarfing gene, which launched the ‘green revolution’, encodes a gibberellin biosynthetic enzyme (GA20ox), and was assigned to the long arm of rice chromosome 1 (Sasaki et al., 2002). A CAPS marker developed for the L. perenne ortholocus of the GA20ox gene was mapped to chromosome 3 close to the plant height QTL, in a region of conserved synteny with rice chromosome 1 (Yamada et al., 2005). The finding provides evidence for the utility of the candidate-gene-based marker approach in determining the major genes that underlie QTL expression and as such provide targets in strategies for genotype design for plant improvement.
Research to locate regulatory genes is underway using the F. pratensis monosomic chromosome substitution lines available at IGER (King et al., 2002), and evidence is increasing that regulatory genes associated closely with the vernalisation locus Vrn-1 which in L. perenne and F. pratensis is on chromosome 4 (Jensen et al., 2005) are implicated. Comparative mapping with heterologous wheat anchor probes has indicated that the important freezing-tolerance QTL Frf4–1 on chromosome 4 of F. pratensis was orthologous to the frost-tolerance loci Fr1 and Fr2 in wheat (Rognli et al., 2003). These loci are associated closely with Vrn-1 and indicate the importance of this region of the genome in freezing tolerance in the Poaceae. Other mechanisms aimed at avoidance of damage to PSII from freezing temperatures have been demonstrated in Festuca. A representational difference analysis identified amplified up-regulated cDNA fragments from cold-induced F. pratensis seedlings (Canter et al., 2000). In particular, a homologue of the chloroplast encoded gene psbA which codes for the D1 protein of photosystem II (PSII) was recovered. The D1/D2 protein dimer at the core of PSII appears to be crucial in maintaining the integrity of the complex (Mattoo et al., 1989).
Successful breeding depends on broad understanding of the genetic architecture of relevant traits. Genes with major effects and genes contributing to the expression of quantitative traits have both a role in controlling abiotic stress tolerance. Useful information on the genetic basis of abiotic stress tolerance has also been obtained by moving genes between plants of the same or closely related species. Chromosome 3 has provided a rich source of allelic variants for stress resistance and genes for drought resistance have been transferred to Lolium genotypes from F. arundinacea (Humphreys MW & Thomas, 1993; Humphreys MW & Pašakinskienė, 1996), from F. glaucescens (Humphreys J et al., 2005) and for freezing tolerance (Grønnerod et al., 2004).
VII. Target traits: carbohydrates and sustainable ruminant animal production
In temperate regions, the primary agricultural use of grasses is as animal fodder. Following the epidemics of bovine spongiform encephalopathy and foot-and-mouth disease in the UK, there has been an increased public and political interest in feeding forage in preference to manufactured concentrate or food wastes. However, feeding forage has some associated nutritional and environmental problems which need to be addressed in the design of new varieties.
A fundamental component of all animal feed is carbohydrate, either as free sugar or as polysaccharide. Nutritionally, carbohydrate provides metabolic energy and a source of carbon skeletons for general biosynthesis. In the ruminants, nitrogen conversion is a microbial process taking place in the foregut. An abundant supply of readily available energy as fermentable sugar, present simultaneously (‘in synchrony’) with nitrogen, is vital for microbial growth and efficient ruminant nutrition (Miller et al., 1999). Dietary synchrony maximises nitrogen conversion to milk and meat, with a concomitant reduction of nitrogenous excretion and environmental pollution. In concentrate-fed animals, dietary synchrony is rarely problematic, because feed composition is controlled. Conversely, forage composition, particularly with respect to carbohydrate, is intrinsically variable, on hourly, daily and seasonal timescales (Humphreys MO, 1989; Cairns, 2003). In the field, the ratio of energy as carbohydrate to unit nitrogen varies continuously. A general outcome of feeding forage is asynchrony of energy and nitrogen resulting in energy limitation and ultimately inefficient of microbial nitrogen conversion. It follows that in designing forage grasses for sustainable ruminant production, increasing the energy content of forage is of primary concern and increasing the concentration, availability and uniformity of fermentable sugar are the principal aims. Substantial quantities of sugar exist in two major pools in plant material. Firstly, in the fibrous structural material of the cell walls: this is quantitatively a high and fairly constant proportion of forage (approximately 0.1 kg kg−1 fresh mass: Clissold et al., 2004) and is composed predominantly of cellulose and hemicellulose. (In the grasses, the latter is a heteropolymer of two sugars: arabinose and xylose accounting for c. 45% of wall carbohydrate.) From a nutritional point of view, however, the structural polymers are relatively indigestible and unavailable for microbial metabolism. In the second pool are the reserve carbohydrates which are synthesised and accumulated during the photoperiod. These may be subsequently used by vegetative tissues in the short term for energy provision when current photosynthesis is low or absent (e.g. at night or in roots) or in the longer term for the growth of perenniating organs (e.g. tillers and seeds). The reserve carbohydrates of temperate grass forage are sucrose, fructan (polyfructosyl sucrose) and, to a lesser extent, starch (polymeric glucose). In the above-ground, grazed vegetative tissues, their accumulation is directly related to the balance between current photosynthesis and overall sink demand. In consequence, their content varies as a complex function of a number of environmental factors (irradiance, temperature, water availability, etc.). Mean total concentrations may reach 0.020–0.030 kg kg−1 fresh mass in the field (Humphreys MO, 1989). Unlike the structural carbohydrates, the reserve carbohydrates are readily available for fermentation.
The structural polysaccharides present in grass forage, at 3 to 5 times the maximal concentration of the combined reserve carbohydrates, represent a large potential energy source. Unfortunately, much of the fibre passes intact through the digestive tract of ruminants and is excreted to the land together, ironically, with any unconverted nitrogen. The simple expedient of making this material more available for fermentation could substantially improve nitrogen nutrition from forage. The mechanistic basis for this low digestibility is not fully understood, although it is likely that the accessibility of cell wall polysaccharides to degradative enzymes is a key determinant (Bunzel et al., 2003). Hence, an understanding of cell wall structure and, ultimately, cell wall biosynthesis will underpin plant-based efforts to improve digestibility. Little is known of the genetics and biochemistry of cell wall synthesis in plants in general (Bolwell, 2000). Less is known about the monocots, and much of this relates to model tissues (such as maize coleoptiles); there is a paucity of information for the mature grazed tissues of forage grasses. What little is known for grass leaves shows that arabinoxylan is the predominant noncellulose structural polysaccharide, and xylan is known to be an antifeedant: its concentration correlates with lower forage digestibility (Bolwell, 2000). In the rumen, the interaction of arabinoxylan with cellulose may sterically inhibit the access of degradative enzymes to polymeric substrates. An additional structural feature of the grass cell wall is lignification by phenolic cross-linking. Ferulate bridges (c. 2% by mass of the wall) covalently bind adjacent arabinoxylan chains and close the wall structure, thereby restricting the entry of degradative enzymes (Bolwell, 2000; Bunzel et al., 2003).
It may therefore be desirable to select for plants with reduced arabinoxylan and lignin in the cell wall (without, obviously, compromising vital plant structure). To this end, lignin has been examined in a Festulolium backcross family as reported by Ralph et al. (2004), but little information is provided therein. An approach to genetically modify grass with a fungal ferulic acid esterase is under development at IGER (M. M. Bunafina et al., IGER, pers. comm.). The esterase is expressed and contained within the vacuole. On consumption, the enzyme is released, and able to break the ferulate ester bridges and open the wall structure to digestion. Preliminary studies indicate an improvement in forage digestibility in these transformants.
The use of functional markers, i.e. the genetic mapping of structural genes for the cell wall components, is currently not possible because the complexity of cell wall biosynthesis has led to the identification of very few enzymes and genes for hemicellulose biosynthesis (Bolwell, 2000). The current lack of understanding of the biochemistry and genetics of plant cell wall synthesis offers a number of opportunities for future research. In the absence of a full understanding of wall biochemistry, the empirical approach afforded by QTL mapping is likely to be more rewarding. Provided that measurable variation exists, it is possible to genetically map and select traits without any mechanistic understanding of their origin. A number of surrogate traits have been used to this end: neutral detergent fibre (NDF) has most commonly been used as an indicator of cell wall content in forages for QTL analysis. In the temperate forage grasses, QTL have been identified on chromosomes 1, 2, 6, 7 (in common with QTL for WSC) and also on chromosome 3 (Humphreys M & Turner, 2003; Van Loo et al., 2003; Cogan et al., 2005). In some cases, NDF QTL colocate with WSC QTL illustrating the likely close relationship between different carbohydrate fractions in plant tissues. More detailed work has been carried out with forage maize and numerous QTL for NDF, ADF and lignin have been identified (Mechin et al., 2001; Cardinal et al., 2003; Krakowsky et al., 2003; Ralph et al., 2004). The phenolic acids p-coumaric acid and ferulic acid were also examined by Fontaine et al. (2003). It may be possible to identify markers from theses studies of maize which could be useful in the analysis of the digestibility of temperate forages.
The various direct measures of digestibility (DMD, dry matter digestibility; OMD, organic matter digestibility; CWD, cell wall digestibility) which integrate the effects of the different nutritional components in forage have also been examined in Lolium by a number of groups. Most groups have used only one analysis technique and found rather few QTL. However, taken together, it is clear that QTL for digestibility have been identified on chromosomes 1, 2, 5, 6 and 7; the same linkage groups as for WSC and NDF (Van Loo et al., 2003; Vandewalle et al., 2003; Cogan et al., 2005). It is not yet clear exactly how QTL for the component traits relate to QTL for digestibility, but it would seem likely that marker selections for any of a range of traits might be used to improve forage quality. Again, more detailed work has been carried out in forage maize, and QTL for various measures of digestibility have been identified (Lübberstedt et al., 1998; Mechin et al., 2001).
The reserve carbohydrates are readily available to the rumen microflora, but are present in lower average concentration by factor of 3 to 5 compared with wall carbohydrate. Improvements in forage quality via carbohydrate reserves can be achieved by increasing the concentration of WSC and/or starch in the grazed component of the plants. Under some physiological conditions, temperate grass leaves can transiently accumulate 90 g kg−1 fresh mass as WSC (Cairns, 2003), so it is clear that the tissue has the natural capacity to sustain higher soluble sugar concentrations.
Genetic variation in WSC content in forage grass has been recognised for some time (Humphreys MO, 1989) and conventional plant breeding has resulted in a number of commercial high-sugar varieties (e.g. AberDart). Significant improvements in ruminant productivity have been reported when fed with high-sugar grass (Miller et al., 1999). Crosses between high- and low- sugar genotypes have produced F2 mapping families which segregate for WSC content. QTL for WSC have been identified by a number of groups working with different species and mapping families, but some reports contain insufficient information to compare QTL positions (Hu et al., 2002; Bhowmik et al., 2005). The first association of WSC with a molecular marker was shown by Hayward et al. (1994), working with Lolium. They demonstrated a close linkage with the isozyme locus phosphogluco-isomerase (PGI/2), which has now been mapped to chromosome 1 of the Triticeae-aligned Lolium map (Armstead et al., 2002). The importance of this region of the Lolium genome has been confirmed by Humphreys M & Turner (2003) and Humphreys M et al. (2003). QTL have also been identified on chromosomes 2, 5, 6 and 7 (Humphreys M et al., 2003; Vandewalle et al., 2003) for Lolium. These QTL often colocate with clusters of QTL for some of the various sugars that comprise the WSC fraction (Humphreys et al., 2003). In most cases, the main sugar explaining the WSC QTL is fructan (L. B. Turner, unpublished data). In addition, a region of chromosome 2 has been shown to be important for the regulation of WSC accumulation during drought (Guthridge et al., 2003). It may be that the QTL represent control regions remote from the structural genes.
On the mechanistic level, the biochemistry of WSC accumulation is relatively well advanced (compared with wall synthesis). Conventional biochemistry and studies of gene expression have led to the isolation of a number of candidate genes (e.g. Gallagher et al., 2004) for use as probes for mapping and then comparison with QTL for high WSC. The first attempts to identify the candidate genes for WSC QTL by this route have not been particularly successful. Several putative structural genes of fructan synthesis have been mapped in L. perenne (Wei et al. 2000; Lidgett et al., 2002; Johnson et al., 2003). However, some confusion arises from the considerable sequence homology between fructosyltransferases and invertases (sucrose breakdown), and it is probable that most of the mapped genes are in fact invertases (Gallagher et al., 2004). Currently there is little evidence for fructosyltransferase involvement in fructan QTL (L. B. Turner, unpublished data). It is possible that further mapping of fructosyltransferase genes may show closer associations. Alternatively, the QTL could result from variation in fructan breakdown by fructan hydrolases, or from the activity of regulatory genes. Alkaline invertase has been shown to map close to sugar QTL on chromosome 6 (Humphreys M et al., 2005), but as its role in cell metabolism has not been clarified, the relevance of this remains unclear.
QTL for starch content in Lolium have been identified on chromosome 3 (Turner et al., 2003); these explain up to 70% variation in leaf and tiller base tissues in both spring and autumn. As these starch QTL do not coincide with identified QTL for WSC, independent selection of starch and WSC may be possible. This holds out the possibility of selection in favour of each trait to tailor forage for different purposes. For example, starch persists during silage manufacture, whilst WSC does not. High-starch grass could permit the development of silage in which a higher proportion of fermentable energy persists for animal consumption.
Approaches using genetic modification have been less successful than conventional genetics in attempts to increase fermentable sugar content in plant tissues. Bacterial genes coding for levan synthesis (soluble polymeric fructose) have been introduced into a range of plants, including L. multiflorum (Ye et al., 2001: reviewed by Cairns, 2003). The levan polymer is not degraded in plant tissue, but can be metabolised by the rumen microflora. Transformation was used in an attempt to shift carbon partitioning in favour of increased WSC accumulation. However, transgenic levan accumulation rates were generally low, and the presence of levan in plant tissue was associated with a number of adverse developmental effects (Cairns & Perret, 2005). Levan transformation does not currently offer an efficient means of improving forage quality.
In the context of forage grass use as an animal feed, NUE can be considered at both the plant and the animal scale. Little information on QTL mapping is available for either. Van Loo et al. (2003) examined plant NUE traits using Lolium and found QTL on chromosomes 1, 2, 4 and 5. The NUE of animal systems depends on C : N ratios (dietary synchrony) in feed and their interactions during the utilisation of nutrients in the rumen. Vandewalle et al. (2003) report a crude protein QTL on chromosome 4 for Lolium. Crude protein QTL have also been identified on chromosomes 2 and 3 (M. O. Humphreys, IGER, unpublished data). At present, it is not possible to assess the feasibility of concurrently improving the NUE at both plant and animal scales.
VIII. Target traits: grass biomass, climate change and energy sustainability
Within the limit of the available solar energy (c. 10 GWh ha−1 yr−1) plants have the potential to contribute to energy provision, to replace fossil sources of energy and to concomitantly reduce carbon emissions. Of the grasses, the East Asian giant perennial, C4 genus, Miscanthus is under consideration as a high-yield, low-input (N and P) and water-use-efficient biomass crop for electricity generation in temperate regions (Clifton-Brown et al., 2004 and references therein). The canes are grown in dense stands in short rotation, mechanically coppiced, chipped, dried, burned, and the heat released used to drive turbines and to provide local heating. In the UK, Bical Ltd. (http://www.bical.net) report a gross average yield of 0.083 GWh ha−1 yr−1 equivalent to 0.8% of incident solar energy. As part of a mixed approach with other technologies, Miscanthus could make a useful contribution, especially if grown on unproductive set-aside land which currently amounts to c. 10% in the UK and Europe. This view is confirmed by Clifton-Brown et al. (2004), who estimate that 10% of agricultural land under Miscanthus could provide 9% of the electricity demand for the European Union (based on consumption in 2000).
Biomass production from Miscanthus currently uses a single clone of one triploid genotype: M. × giganteus Greef and Deuter., which is seed-sterile and cold sensitive, both characters disadvantageous for commercial growing. Attempts to improve these traits and to generally broaden the genetic base are in progress: other species (M. sinensis, M. sacchariflorus and hybrids) have been trialled in Europe (Clifton-Brown et al., 2004) to assess their potential utility. Genetic mapping and QTL analyses using Miscanthus sinensis (the contributor of one of the two subgenomes of M. × giganteus) have also been reported recently: 20 potential QTL for yield and its components (e.g. stem yield, leaf yield) (Atienza et al., 2003a) and 11 QTL for other agronomic traits (height and stem diameter) (Atienza et al., 2003b) were identified, which explain much of the variation. Factors determining combustion quality were also examined and QTL identified for chlorine and potassium content (Atienza et al., 2003c) and calcium, phosphorus and sulphur content (Atienza et al., 2003d).
The principles of sustainability have been understood by plant breeders for many years. As a result, traditional breeding has already been successful in producing germplasm and also commercial varieties of grasses with traits for enhanced agricultural sustainability. An example is high-sugar perennial ryegrass for improved ruminant nitrogen conversion and reduced nitrogen emissions. That grass breeding can provide sustainable solutions to defined agronomic and agro-environmental problems is clearly demonstrated. However, future climatic outcomes are difficult to predict. Faced with a range of unpredictable outcomes, the implications for plant improvement are clear: we need to be in a position to tailor plants to fit whatever new environments may result. The widespread adaptation of ecotypes of Lolium and Festuca species to the highly contrasting growing conditions that are found in Europe and elsewhere, harnessed to the increasing access through genomics to the genes that underpin these adaptive capabilities, provides an almost limitless resource for plant geneticists and physiologist to operate in tandem to design grass genotypes to combat the various challenges grasslands will face over the coming years.