Dissecting the regulation of fructan metabolism in perennial ryegrass (Lolium perenne) with quantitative trait locus mapping


Author for correspondence: L. B. Turner Tel: +44 (0) 1970 823144 Fax: +44 (0) 1970 823243 Email: lesley.turner@bbsrc.ac.uk


  • • Quantitative trait locus (QTL) mapping, which can be a useful tool for dissecting complex traits, has been used here to study the regulation of fructan metabolism in temperate forage grasses.
  • • An F2 mapping family, derived from a high water-soluble carbohydrate (WSC) × low WSC cross, was used to map fructans and the other components of WSC (sucrose, glucose and fructose) in leaves and tiller bases of perennial ryegrass (Lolium perenne) in spring and autumn. To characterize regions of the genome that control basic carbohydrate metabolism, a strategy to minimize the impact of genotype (G) × environment (E), and E-effects on the characterization of G-effects, was adopted.
  • • Most traits were highly variable within the family. There was also considerable year-to-year environmental variation. However, significant genetic effects were detected, and several traits had high broad-sense heritability. QTL were identified on chromosomes 1, 2, 5 and 6. Leaf and tiller base QTL did not coincide. Individual QTL explained between 8 and 59% of the total phenotypic variation in the traits.
  • • Fructan turnover, metabolism and their genetic control, and the effect of environment, are discussed in the context of the results.

amplified fragment length polymorphism


degree of polymerisation

G × E

genotype by environment


high-performance liquid chromatography


log-odds score


quantitative trait locus (loci)


restriction fragment length polymorphism


simple sequence repeat


sequence-tagged site


water-soluble carbohydrate.


Temperate grasslands support most of the world's milk and meat production. Presently c. 75% of feed requirements are obtained from grass and forage, although this varies from 60% of the feed for some dairy cows up to 90% for sheep (Wilkins & Humphreys, 2003). Feed costs have been estimated to represent 35% of the total cost of ruminant production in the UK, so the nutritional value of fodder has a major impact on the efficiency and profitability of UK livestock production. One important aspect of nutritional value for ruminants is the availability of an easily fermentable energy source in the rumen which can be supplied by forage water-soluble carbohydrate (WSC) (Miller et al., 2001). More than 80% of agricultural forage seed usage in the UK is ryegrass (Burgon et al., 1997), and perennial ryegrass (Lolium perenne L.) is the most important species. It accounted for 78% of the grass seed produced in the UK in 2003 (NIAB Seed Statistics; NIAB, Huntingdon Road, Cambridge). Traditional plant-breeding approaches have been successful in producing high-sugar ryegrasses that have been shown to improve protein utilization by ruminants, and boost milk and meat production whilst reducing nitrogen losses in waste products (Miller et al., 2001). Greater understanding of the underlying regulation of WSC content will benefit future breeding programmes.

Starch, a polymer of glucose, is the major form in which carbon is stored in the reserve tissues of plants. However c. 12–15% of higher plants store an alternative polysaccharide, fructan (a polymer of fructose), which is synthesized directly from sucrose and is soluble. Fructan-accumulating plants are found in several plant groups, including the Compositae, Liliaceae and Gramineae (Pollock & Cairns, 1991). The fructans accumulated by the temperate forage grasses and cereals are distinct from those of the other groups (Pollock & Cairns, 1991), and the regulation of fructan metabolism in grasses is still little understood. Studies of reserve deposition following artificial manipulation of sink strength have produced information on the physiology, biochemistry and enzymology of fructan synthesis in these particular circumstances (Pavis et al., 2001; Cairns et al., 2002; Cairns & Gallagher, 2004). Nonetheless, much more remains to be explained, and there appears to be little or no published information on the endogenous regulation of fructan content.

Quantitative trait locus (QTL) mapping can be a useful tool for dissecting complex traits. It requires variation for the trait of interest, appropriate mapping families with suitable marker maps and reliable measures of phenotype. Outbreeding plants, like the temperate forage grass perennial ryegrass (L. perenne), are ideally suited for such work because they retain a broad genetic base. Indeed, perennial ryegrass has been shown to contain considerable variation for WSC content (Humphreys, 1989; Turner et al., 2001, 2002). Genetic mapping in forage grasses has been relatively under-developed compared with that of the major cereal species. However, the recent publication of L. perenne genetic maps, which include a range of markers that have been mapped in other species, has led to the development of linkage maps suitable for both QTL mapping and comparative genomics (Armstead et al., 2002, 2004; Jones et al., 2002).

The efficiency of QTL mapping is sometimes limited by inaccurate phenotypic measures. The WSC content of perennial ryegrass can be rather unpredictable in ‘natural’ environments. This may well be a reflection of the many factors that can affect WSC metabolism. Irradiance, photoperiod, diurnal regulation, temperature, water availability, nutrient supply, timing of flowering, pests, diseases, and interactions between such abiotic and biotic factors, may all alter the WSC content of vegetative tissues of temperate forage grasses and cereals (Hayes et al., 1993; Radojevic et al., 1994; Smith et al., 1998; Thomas & James, 1999; Wang et al., 2000; Thorsteinsson et al., 2002). Consequently, any study to dissect the underlying regulation of fructan metabolism must take account of its interaction with the environment. With unlimited resources the ideal would be to run replicated experiments at different times and in different locations and/or environments. Phenotyping then becomes a seriously rate-limiting step. In practice it can be possible to choose an experimental design that minimizes undesirable effects in the context of current experimental aims (Borevitz & Chory, 2004), for example combining replication with natural variation in environmental conditions.

The objective of the work reported here was to characterize regions of the genome that have basic control over carbohydrate (and particularly fructan) metabolism in order to give further insight into its regulation in ryegrass. Therefore, the aim was to maximize the impact of genetic effects, and minimize the impact of genotype by environment (G × E) and environmental effects, in the detection of QTL. The strategy chosen was to replicate testing of genotypes over years (i.e. to collect one sample per genotype each year for several years), and look for reproducible effects. In this way, confounding effects from short-term environmental variation would be minimized. Two different tissues (leaves and tiller bases) and two sampling times (spring and autumn) were included to give some information on tissues with different roles and different carbon status.

Materials and Methods

Plant material

The mapping family (WSC F2) used was that previously described by Armstead et al. (2002). This is an F2 population of 188 plants produced by self-pollinating a single F1 hybrid plant, which was obtained by crossing individual genotypes from partially inbred-lines of the varieties Perma and Aurora. The individual members of the family were maintained in 15-cm diameter pots of Humax John Innes No3 with wetting agent (Richardsons Moss Litter Co. Ltd, Carlisle, UK) in a frost-free, unlit glasshouse throughout the year. The plants were renewed each year from a small subset of tillers. Samples for trait data were taken only from plants in the F2 mapping family. Parents of the original F1 plant, used to produce the mapping family, were inbred for several generations and suffered from inbreeding depression, as expected in an outbreeding species such as Lolium. Consequently, unlike the F2s, the parent plants were very small and weak, and hence unsuitable for phenotypic analysis. Data on carbohydrates in the noninbred parent source varieties have previously been published (Turner et al., 2002), which provide a better indication of the parent trait values.

Tiller base and leaf material (Turner et al., 2001) for carbohydrate analysis were sampled from October 1997 to March 2001. Autumn samples were taken during the last week of October or the first week of November, and spring samples were taken during the last week of February or the first week of March. One replicate (i.e. one sample per genotype) was taken each year for 3 yr. To minimize diurnal changes in WSC content during the sampling period, dull, cool days were chosen for sampling. Samples were taken between 13:30 and 16:30 hours, usually within 2.5 h. All material was immediately frozen in liquid nitrogen, stored at −80°C, freeze dried and then finely chopped before extraction.

Leaf material for DNA extraction was also frozen in liquid nitrogen and, if necessary, stored at −80°C until extraction.

Carbohydrate extraction and analysis

Carbohydrates were extracted following the procedure described by Turner et al. (2002). Sugars were separated and quantified by isocratic high-performance liquid chromatography (HPLC) on a 300 × 7.8-mm column of Aminex HPX87-C (Bio-Rad, Hemel Hempstead, UK) at 85°C with degassed water as a mobile phase, at 0.6 ml min−1. The column was protected by an in-line 0.5-µm filter and a Carbo-C guard column (Bio-Rad). Samples from 1997, and some from 1998, were analysed using a Shodex RI-71 refractive index (RI) monitor, 1350 series pump, AS-100 auto-sampler and high-resolution liquid chromatography instrument interface (all Bio-Rad). The remainder were separated using the same chromatographic system (column and RI monitor) using an ASI-100 automated sample injector, P580 pump and Chromeleon instrument interface (all Dionex, Cambridge, UK). The column separated fructan molecules with a degree of polymerization (DP) of 3 from other short-chain fructan molecules and from long-chain polymers. In this communication, DP3 molecules have been described as ‘oligofructan DP3’, other short-chain molecules as ‘oligofructan DP > 3’ and long-chain high-molecular-weight fructan molecules as ‘polymeric fructan’. Total WSC was measured directly on the same extracts by using the anthrone colorimetric assay (Yemm & Willis, 1954).

DNA extraction and molecular marker analysis

DNA was extracted from the plant material as described by Hayward et al. (1998) and by using the Qiagen DNEasy Plant Mini Kit (Qiagen, Crawley, UK). The marker analysis data for restriction fragment length polymorphisms (RFLPs), amplified fragment length polymorphisms (AFLPs), some simple sequence repeats (SSRs), sequence-tagged site (STS) gene probes and isozymes were previously described by Armstead et al. (2002, 2004). The marker data set has been extended to include 132 new SSRs produced by ViaLactia Biosciences (Auckland, New Zealand) and used under licence. Most of these have the prefix ‘rv’, the remainder have syntax of the form ‘code number base repeat code number’. These were amplified individually using fluorescent primers (thermal cycling was performed starting with 10 min at 95°C, followed by 10 cycles of 60 s at 94°C, 30 s at annealing temperature, 60 s at 72°C with the annealing temperature decreasing by 1°C per cycle, followed by 25 cycles of 30 s at 94°C, 30 s at the final annealing temperature, 30 s at 72°C and finally 10 min at 72°C; the annealing temperatures were dependent on the primer pair used) and multiplexed on an ABI 3100 sequencer (Applied Biosystems, Warrington, UK).

Linkage map

The new SSR markers were integrated into the recently published version of the map for this F2 family (Armstead et al., 2004) using joinmap® 3.0 (Van Ooijen & Voorrips, 2001). Markers were grouped with a minimum log-odds score (LOD) of 3 and the map order was determined using the following parameter settings: pairwise marker data with a recombination frequency of < 0.4 and LOD of > 1, and a jump in goodness-of-fit of < 5. One marker, rv1208, which just exceeded the goodness-of-fit criterion, was allowed on chromosome 6 as it was the only marker on the chromosome that did not meet the criteria and omitting it made no significant difference to the map order or map distances. Forty-two of the new ViaLactia SSR, and 21 markers from the previous map (15 AFLPs, four RFLPs and two other markers), were excluded from the map.

Statistical analysis

All statistical analyses were performed with the aid of genstat® for Windows®, Version 8.1 (Payne et al., 2005). The Anderson-Darling test was used to assess whether the trait data for the mapping family showed normal distributions. Correlation and analysis of variance (anova) analyses were carried out using the menu-driven procedures standard within the programme. The product moment correlation coefficient was calculated for pairwise combinations. Trait data were subjected to one-way anova, treating year as a random effect and genotype as a fixed effect. Broad-sense heritability was calculated from the anova mean square values, according to the formula:

h2 = [(genotype ms − residual ms)/n]/{[(genotype ms − residual ms)/n] + [(residual ms)/n]},

where ms is the mean square value and n is number of blocks or replications.

QTL analysis

QTL analysis was carried out using MapQTL, version 4.0 (Van Ooijen et al., 2002) with the population structure set as F2. Additive [(µA – µB)/2] (µA is the estimated mean of the distribution of the quantitative trait associated with the ‘a’-genotype; µB for the ‘b’-genotype) and dominance [µH−(µA +µB)/2] (µH for the ‘ab’-genotype) effects were fitted by the software. Each trait was analysed by interval mapping data from individual years and from the mean of the years. A QTL was declared when the LOD score was > 3.0 for the mean, the QTL was detected in at least two individual years with a LOD score > 2.0, and the QTL position was supported by Kruskal–Wallis nonparametric single-locus analysis. The comparison of Kruskal–Wallis output with interval mapping included consideration of the pattern of the test statistic (K*) when markers were arranged in map order. Declared QTL were further resolved by composite mapping [multiple-QTL mapping (MQM) in MapQTL 4.0], using cofactors chosen with the help of the automatic cofactor selection option within the software with a threshold of 0.02.


WSC content

Correlations between years for total WSC are reported in Table 1. Data for polymeric fructan (not shown) were similar to those for anthrone, whereas data for sucrose, glucose and fructose (not shown) were more like the HPLC total. All the correlation coefficients were numerically low; however, except in the case of tiller bases in the autumn, most were significant. Higher correlations were observed for directly measured total WSC (anthrone) than for derived WSC (HPLC), indicating that significant accumulation of errors sometimes occurred when individual peaks were summed. The lowest correlation coefficients were those calculated for the autumn tiller base data. A fourth year was analysed to examine the lack of correlation between these data sets. All correlations were again very low or zero, confirming that tiller base carbohydrate content has a low repeatability from year to year.

Table 1.  Correlations between replicates from different years for total water-soluble carbohydrate (WSC) of perennial ryegrass (Lolium perenne) leaves and tillers measured directly by anthrone or calculated as the sum of the major components from HPLC analysis
  LeavesaTiller basesa
Year 1Year 2Year 3Year 1Year 2Year 3Year 4
  • a

    Spring or autumn, as listed in the first column of the table.

  • n= 188 for all traits. The threshold significance value for P < 0.05 is 0.144; for P < 0.01 is 0.188; and for P < 0.001 is 0.239.

Total WSC (anthrone)
 SpringYear 11.000   1.000   
Year 20.2041.000  0.235 1.000  
Year 30.2770.3151.000 0.288 0.3671.000 
 AutumnYear 11.000   1.000   
Year 20.2121.000  0.072 1.000  
Year 30.2020.4951.000 0.190 0.1291.000 
Total WSC (HPLC)
 SpringYear 11.000   1.000   
Year 20.2111.000  0.110 1.000  
Year 30.1110.0991.000 0.210 0.3531.000 
 AutumnYear 11.000   1.000   
Year 20.1331.000  0.015 1.000  
Year 30.1450.2181.000−0.100−0.0271.000 
Year 4    0.051 0.1090.0111.000

The genetic and environmental (year) variances were partitioned by anova. Only 3 years’ data for tiller bases in the autumn were used in order to balance the analysis. The partitioning of G × E effects was not possible, but these must be smaller than the residual mean squares, which also include sampling and analysis errors and random plant-to-plant variation, so their order of magnitude can be inferred. Mean square values for example traits are shown in Table 2. Year effects for total WSC were large, genetic effects significant and G × E effects must be relatively small. Consequently, the broad-sense heritability of these traits was mostly moderate to high (Table 3). In general, the highest heritabilities were calculated for total WSC and polymeric fructan. Values for tiller bases in the autumn were low.

Table 2.  Partitioning of genetic and environmental variation by analysis of variance (anova) data
  1. Data for the total water-soluble carbohydrate (WSC) of perennial ryegrass (Lolium perenne) leaves and tillers, measured directly by anthrone or calculated as the sum of the major components from HPLC analysis, are expressed as mean squares for 188 genotypes and 3 yr.

Total WSC (anthrone)SpringLeaves 29861344551926
Tiller bases  9076061752648
AutumnLeaves 35004956371995
Tiller bases 19124295545497
Total WSC (HPLC)SpringLeaves 62634066394064
Tiller bases 13156570583653
Tiller bases 25528456725063
Table 3.  Broad-sense heritability values for water-soluble carbohydrate (WSC) component traits in perennial ryegrass (Lolium perenne) leaf and tiller base tissues from mapping family plants sampled in the spring and autumn over a time-period of 3 yr
 LeavesTiller bases
  1. DP, degree of polymerisation; DP3, degree of polymerisation of 3; DP > 3, degee of polymerisation > 3; hmwt, high molecular weight.

Oligofructan DP30.1600.29600.154
Oligofructan DP > 30.5570.5850.0590.478
Polymeric fructan hmwt0.4450.6980.4620.304
Total WSC (HPLC)0.3880.4610.4820.107
Total WSC (anthrone)0.5680.6460.5710.425

The distributions of the components of WSC (Table 4) showed a considerable range of WSC content within the mapping family for both leaves and tiller bases in both spring and autumn. Leaves showed a 20% variation in WSC content in spring and 24% in autumn based on anthrone data. The variation for tiller bases was greater, with values of 31% and 34%, respectively. This variation reflected the large range observed in the content of all the WSC components measured. Nonnormal distributions were observed in a number of cases, particularly for the more minor components (DP3 oligofructan, DP > 3 oligofructan, glucose and fructose), which were often undetectable in a number of samples. In most cases the data could be normalized or near-normalized (significance level reduced to P < 0.05), either by omitting significant outliers, transforming to natural logarithms or both. The only exceptions were leaf oligofructan DP3 in spring and autumn, and leaf polymeric fructan in the spring. Autumn tiller base oligofructan DP > 3 data were not analysable as there were too many ‘not detected’ entries. However the data were not highly skewed, and in all cases transforming the data had no effects on the positions of the QTL and little effect on their magnitude. Consequently, QTL analysis on the original data is shown for all traits.

Table 4.  Distributions of components of water-soluble carbohydrate (WSC) content in perennial ryegrass (Lolium perenne) leaves and tillers
  TraitMinimum valueMaximum valueNormality testSkewKurtosis
  1. DP, degree of polymerisation; DP3, degree of polymerisation of 3; DP > 3, degee of polymerisation > 3; NA, not analysable; ND, not detected.

  2. Significant deviations from a normal distribution were tested by the Anderson–Darling procedure. The minimum and maximum values (mg g−1 dry matter) given for the 188 genotypes in the mapping family disregard outliers identified by the normality test.

LeavesSpringOligofructan DP3ND 39.4P < 0.01 0.61 0.11
Oligofructan DP > 3ND 51.8P < 0.01 0.81 0.43
Polymeric fructan  6.9139.7P < 0.01 0.60−0.36
Sucrose 17.4 95.6NS 0.33−0.10
GlucoseND 42.0P < 0.01 0.63 0.21
FructoseND 38.3NS 0.17 0.12
Total WSC (HPLC) 77.7306.5NS−0.40 0.50
Total WSC (anthrone) 87.1285.6P < 0.01 0.70 1.30
LeavesAutumnOligofructan DP3ND 44.0P < 0.01 0.31−0.59
Oligofructan DP > 3ND 68.7P < 0.01 1.01 0.48
Polymeric fructan 29.9258.7P < 0.01 0.43−0.20
Sucrose 15.4 77.8P < 0.01 1.13 2.90
Glucose 10.8 36.2P < 0.01 1.75 8.14
Fructose 14.5 49.6P < 0.01 1.73 9.86
Total WSC (HPLC) 94.7355.8NS 0.30−0.40
Total WSC (anthrone) 90.4328.9P < 0.01−0.50 1.20
Tiller BasesSpringOligofructan DP3ND 80.5P < 0.01 2.42 7.16
Oligofructan DP > 3ND101.8P < 0.01 1.17 1.27
Polymeric fructan 15.6229.7NS 0.27 0.91
Sucrose  8.2 53.4NS 0.34 0.40
Glucose  7.0 31.2P < 0.05 0.56 0.67
Fructose 27.2 64.9P < 0.01 2.00 9.05
Total WSC (HPLC)144.6418.9P < 0.05 0.70 3.10
Total WSC (anthrone) 87.7403.2NS 0.50 1.40
Tiller BasesAutumnOligofructan DP3ND 25.9P < 0.01 1.81 7.00
Oligofructan DP > 3ND 49.4NA 3.4310.31
Polymeric fructan125.7417.3NS 0.30 0.60
Sucrose 11.9 30.8NS 0.07−0.24
Glucose  3.2 20.7NS 0.11 0.05
Fructose  5.2 30.0NS 0.21 0.22
Total WSC (HPLC)215.4465.0NS 0.20 0.50
Total WSC (anthrone)197.3541.0NS 0.40 0.80

Table 5 gives mean WSC contents for the mapping family. Tiller bases in the autumn had the highest WSC content, with at least 74% being polymeric fructan. The anthrone total WSC was higher than the HPLC total in this tissue, in contrast to the other data sets where it was lower. (The slightly higher HPLC totals in most instances probably arose from the accumulation of minor errors in the individual totals.) This suggests that autumn tiller bases contained in the order of 12% anthrone-reactive material, which is currently unidentified. The tiller base WSC content was reduced by 44% in the spring; 63% was polymeric fructan and the contents of the other fractions had increased. Oligofructans increased by five-fold (DP3) to 11-fold (DP > 3). The fructose content increased threefold and was more than twice that of glucose. Leaves always contained less total WSC and less polymeric fructan than tiller bases. Leaf oligofructan, mono- and disaccharide contents were higher than tiller base in the autumn, but these differences had largely disappeared by the spring.

Table 5.  Water-soluble carbohydrate (WSC) content (mg g−1 dry matter) of perennial ryegrass (Lolium perenne) leaf and tiller base tissues from mapping family plants in the spring and autumn
 LeavesTiller bases
  1. Data are expressed as mean values and standard errors (n = 188) of mean plant values.

  2. DP3, degree of polymerisation of 3; DP > 3, degree of polymerisation > 3; hmwt, high molecular weight.

Oligofructan DP3 17.55 ± 0.57 16.39 ± 0.77 17.20 ± 0.78  3.34 ± 0.30
Oligofructan DP > 3 20.24 ± 1.10 13.55 ± 1.13 25.50 ± 1.65  2.22 ± 0.56
Polymeric fructan hmwt 53.09 ± 2.30161.89 ± 3.44132.60 ± 3.08275.49 ± 2.92
Sucrose 53.93 ± 1.15 32.77 ± 0.67 25.79 ± 0.58 20.54 ± 0.28
Glucose 19.29 ± 0.50 20.19 ± 0.39 18.69 ± 0.41 10.66 ± 0.22
Fructose 21.97 ± 0.41 27.06 ± 0.48 39.99 ± 0.49 13.71 ± 0.26
Total fructan 83.93 ± 2.21190.98 ± 2.77175.30 ± 2.86278.58 ± 3.10
Total WSC (HPLC)179.30 ± 3.57272.01 ± 2.82261.36 ± 3.53324.42 ± 2.95
Total WSC (anthrone)169.40 ± 2.88226.45 ± 3.17208.86 ± 3.32369.91 ± 4.61

There were significant correlations between some of the different components of total WSC content (Table 6). In the spring, DP3 oligofructan was correlated with mono- and disaccharide content, but not with the other fructan fractions in leaves. DP > 3 oligofructan was negatively correlated with polymeric fructan, but had no significant relationship with any other component. Polymeric fructan was correlated with all the measured components except DP3 oligofructan. The small sugars (glucose, fructose and sucrose) were all correlated with each other. Some of these relationships were maintained in the autumn, but DP3 oligofructan was no longer correlated with glucose and fructose, DP > 3 oligofructan was correlated with fructose, and polymeric fructan was negatively correlated with all the other fractions measured. Tiller base data (not shown) showed similar relationships, although the correlation coefficients were numerically slightly lower.

Table 6.  Correlations between the components of water-soluble carbohydrate (WSC) content of perennial ryegrass (Lolium perenne) leaf tissues in (a) spring and (b) autumn
 DP3DP > 3hmwtSucroseGlucoseFructose
  1. Oligofructan DP3 (DP3), oligofructan DP > 3 (DP > 3), polymeric fructan (hmwt). n = 188. Threshold significance value for P < 0.05 is 0.144; for P < 0.01 is 0.188; for P < 0.001 is 0.239.

 DP3 1.000     
 DP > 3 0.045 1.000    
 hmwt 0.136−0.531 1.000   
 Sucrose 0.313−0.070 0.5151.000  
 Glucose 0.352−0.051 0.4380.2881.000 
 Fructose 0.252 0.004 0.4140.2100.7991.000
 DP3 1.000     
 DP > 3 0.187 1.000    
 hmwt−0.188−0.780 1.000   
 Sucrose 0.245 0.103−0.1571.000  
 Glucose 0.108 0.193−0.1950.2541.000 
 Fructose 0.102 0.311−0.3250.2230.8731.000

Linkage map

The genetic map (Table 7) used for QTL analysis in this study identified seven chromosomes and covered a total genetic distance of 628 cM with an average marker density of one marker every 2.8 cM. The largest distances between consecutive markers were 21 cM on chromosomes 1 and 5. Significant (P < 0.05) segregation distortions for individual markers were observed on all chromosomes, but were particularly prevalent on chromosomes 5, 6 and 7. Two chromosomes contained previously mapped markers with functional attributes pertaining to sucrose metabolism. The alkaline invertase loci (EMBL: Ac. AJ003114), Alkinv 1/4 and Alkinv 2/3, were on chromosome 6, and Inv1:2 (EMBL: Acs. AJ532551 and AJ532552) was on chromosome 7.

Table 7.  Genetic linkage map of the Lolium perenne F2 mapping family
Distance (cM)MarkerDistance (cM)MarkerDistance (cM)MarkerDistance (cM)MarkerDistance (cM)MarkerDistance (cM)MarkerDistance (cM)Marker
  • Chromosome designations conform to the Triticeae numbering.

  • *

    Markers with a distorted segregation ratio (P < 0.05).

 0.0rv0913 0.0M4-136 0.0C30  0.0PSR580 0.0GSY60.2* 0.0CDO542* 0.0INV1:2*
 1.8PGI10.2CDO38* 6.6LPSSRK1A11  2.1rv0992 3.1rv0250* 9.4CDO395* 5.6R2869
 3.4rv139121.2rv007515.5LPSSRK1A03*  4.5E42M3311 4.0rv0814*12.2rv1423* 8.8B6102*
 7.9CDO58023.8FpAFPH21.6rv1133  7.1rv0068 4.4CDO1380.2*25.2E39M4908*11.0CDO545*
13.3rv065925.1E39M580522.9CDO328 11.7PSR115 4.4CDO127*30.1E41M571216.7OSW*
15.4E42M330825.6FpAFPD25.3E39M4907 19.6rv0454 4.5F29-1*34.9rv120819.1rv0908*
15.7rv030128.3BCD85527.4rv1144* 27.7rv0966 6.6rv0757*35.5RZ8725.1rv1284
16.0rv003329.3R334928.1E42M3306 30.1rv0684 8.1rv0184*38.8M65-1*26.3C764*
16.002ga130.1Fp12.129.5CDO455 31.8Rz395 9.1RZ206*41.7Alkinv.1/4*26.5rv0009_2
16.6rv025331.1Fp12.2*31.6PSR394 36.0BCD808 9.8BCD1087*43.4B610426.504-ga1_2
16.6rv025231.7Rz395.132.8rv0154 37.4RYE12*11.8PSR574*44.4rv0641*28.9rv1171*
17.2E38M470933.6E40M591033.3rv0009_1 41.9PSR92212.6rv0495*45.2rv006730.0rv1411*
17.5rv0810_1*34.0CDO5933.304ga1_1 42.5C746*17.0rv0260*46.1rv030731.708ga1*
18.1E36M550334.7CDO395.2*34.6F29-1b* 45.3E41M570818.7rv0082*47.2CDO71831.7rv0141*
18.9rv043434.9LPSSH01A0736.113ca1 46.5E41M570223.4B6103*47.4rv017132.0rv0810_2
19.3E36M550835.6CDO36536.7rv0629* 48.2E38M470625.8rv0340*47.8Alkinv.2/3*32.8RZ144*
21.8E41M570936.0rv011637.4WG889 49.0CDO79529.3rv1112*50.4rv044934.5E40M5909*
23.9rv0105*36.2rv010938.7BCD828 49.4E38M470531.8E40M590750.6rv041435.6rv0479*
26.4OSE36.3CDO385.239.2E40M5901* 50.2rv038052.8R271051.4C3736.2E38M4710*
29.583ca136.8rv006240.5rv1042 51.0ADH56.3E40M590553.8rv0196*40.1rv0311*
29.7rv032737.6C451*42.3rv1063 51.8RZ53758.9E36M551256.9CDO686*40.6rv0440*
30.7B6101*38.022taga142.8GOT/3* 53.1Fp4165.2RZ404*59.7RYE1441.3E41M5710*
34.6BCD738*38.917ca143.8C250 53.6rv0074  60.7RYE641.6CDO385.1*
45.1rv024440.4rv003744.7E42M3302 55.2CDO1380.1  63.6RYE541.8GSY60.1*
52.9R295342.4E41M571347.2PRO* 57.9R2702B  72.5CDO51642.521ga1*
73.7CDO20244.3E38M470450.5rv1131 59.6CD0938  83.1rv0739*43.0PSR690*
86.7E36M551147.7rv097958.4E36M5509* 62.9rv1190    44.7RZ952*
  48.6CDO141760.9CDO345* 63.7PSR163    46.1C390*
  48.8rv012263.5E38M4712* 65.8E39M4905    48.8rv0262*
  49.4rv128264.2rv0203 66.6E42M3303    51.9rv0728
  49.9E39M490666.814ga1* 67.8E42M3313    52.7rv0264*
  59.0rv115468.3C949 85.3rv0190    54.1LtCOa*
  60.4rv095972.1E36M5513 86.4CDO20    55.9E41M5701*
  61.5E40M5908*73.0rv1046* 87.6B6105    57.4LtCOb*
  62.8M15-18579.025ca1* 90.9rv0061    58.9E39M5802*
  65.2E36M5510*90.5UNI-001 93.2E39M4904    66.5rv0692*
  69.8PSR540  111.8rv0382_2    72.4rv0248*
  76.3E41M5705        82.0ACP*
  84.1rv0188        94.6CAT


QTL were detected for polymeric fructan in tiller bases and leaves in both spring and autumn sampling (Table 8, Fig. 1). In most cases these QTL were also observed as total WSC QTL, reflecting the high proportion of total WSC measured as polymeric fructan and the correlation between the two traits (data not shown). Tiller base and leaf QTL were located on different chromosomes. The polymeric fructan QTL for tiller bases on chromosome 1 was observed in spring and autumn. In contrast, although leaf polymeric fructan QTL were on chromosome 6 in spring and autumn, they were not at the same location. The QTL for polymeric fructan in autumn-sampled leaves coincided with a QTL for DP > 3 oligofructan, with the positive alleles in the opposite linkage phase.

Table 8.  Multiple quantitative trait locus mapping (MQM) of quantitative trait loci that are detected by interval mapping of mean trait data and that are supported by interval mapping of individual years and single-locus analysis
TissueTraitLinkage group2-LOD interval (cM)Distance (cM at peak)   EffectsPercentage variation explained
WSC fractionFlanking markersLODAdditiveDominance
  1. Positive values of additive effects indicate that the allele from Aurora, the high-sugar parent, confers the positive effect.

Leaf SpringPolymeric fructan636–4038.8RZ87Alkinv.1/4 7.2−17.88−9.6016.3
Total WSC (HPLC)636–4038.8RZ87Alkinv.1/4 5.8−24.74−18.4115.0
Leaf AutumnOligofructan DP > 3615–1917.2rv1423E39M490824.4−17.29−16.8559.0
Polymeric fructan611–1812.2CDO395E39M490822.5+42.75+40.5943.2
Glucose229–3130.1R3349Fp12.2 8.6+15.67−14.8216.1
Glucose643–4645.2B6104rv0307 6.1−2.65−1.9814.7
Fructose643–4645.2B6104rv0307 5.1−2.90−2.7712.4
Total WSC (anthrone)611–1712.2CDO395E39M490815.9+34.66+30.1138.7
Tiller Base SpringPolymeric fructan114–1615.7rv0659rv0033 9.9+28.81+1.0121.0
Glucose5 0–4 3.1GSY60.2rv0814 6.5−3.71+0.3414.9
Total WSC (anthrone)114–1615.7rv0659rv0033 7.0+27.30+3.8716.2
Tiller Base AutumnPolymeric fructan1 0–2 1.8rv0913rv1391 3.5−4.53−25.00 8.2
Polymeric fructan115–1615.7E42M330802ga1 3.7+31.22−34.96 8.9
Figure 1.

Locations of quantitative trait loci (QTL) for water-soluble carbohydrate (WSC) and its components on the ryegrass (Lolium perenne) chromosomes. Bars represent the 2-LOD interval from multiple-QTL mapping (MQM) for leaf tissue in the spring (white background) and autumn (light grey background) and tiller bases in the spring (mid-grey background) and autumn (dark grey background). C, total WSC; PF, polymeric fructan; OF, oligomeric fructan DP > 3; S, sucrose; G, glucose; F, fructose.

The traits showed not only additive, but also dominance, effects at the QTL positions detected. The latter made a considerable contribution to the percentage of the variation explained. High fructan content was conferred by the allele from Aurora, the high-sugar parent, in most instances. However, the high-sugar allele was derived from Perma at the fructan QTL for leaves in the spring on chromosome 6. The high-sugar alleles for the two QTL on chromosome 1 for tiller base polymeric fructan content in the autumn were from opposite linkage phases, with the QTL at 2 cM deriving the high-sugar allele from the Perma parent. Additionally, there were strong dominance effects at this QTL, with the high-sugar allele being recessive.

In the autumn, there were coincident QTL for glucose and fructose in leaves on chromosome 6. The interval mapping profiles for the two sugars were extremely similar, suggesting that they may be controlled by the same gene. An invertase would be one possible candidate for such a pleiotropic effect. The relationship of the alkaline invertase loci to these sugar QTL is shown in Fig. 2.

Figure 2.

The relationship between the glucose (solid line) and fructose (dashed line) quantitative trait loci on chromosome 6 from interval mapping for ryegrass (Lolium perenne) leaf tissue in the autumn and an alkaline invertase marker. The solid bar represents the 2-LOD interval from multiple-QTL mapping (MQM).

Overall, the QTL explained between 8 and 59% of the total phenotypic variation in their respective traits. The smallest effects were for the polymeric fructan QTL in repulsion on chromosome 1. The largest effects were for the fructan (and total WSC) QTL near the top of chromosome 6. There were only two instances in which more than one QTL was detected for a trait, namely leaf glucose in the autumn and tiller base polymeric fructan in the autumn. Therefore, a large proportion of the total phenotypic variation was unexplained by the QTL.


Fructan tissue content, turnover and metabolism

It is often stated that fructan is found predominantly as a reserve carbohydrate in leaf sheaths and leaf bases of temperate forage grasses, and that little occurs in expanded leaf blades (Lidgett et al., 2002; Gallagher et al., 2004). However, this is based mainly on evidence from controlled environment studies on intact plants when overall plant carbohydrate status can be low (Pavis et al., 2001). It is clear from the current measurements that, whilst tiller bases (leaf sheaths and the base of the expanding leaf) had the higher reserves, significant amounts of polymeric fructan often accumulated in leaf blades. This may have been the result of conditions in the glasshouse, where higher irradiances and variable temperatures were experienced. Cool temperatures before the sampling periods could reduce sink activity from growth. Leaf blades are known to accumulate high concentrations of polymeric fructan when photosynthate supply is maintained and sink activity is low, for example after excision when all sinks have been removed (Cairns et al., 2002; Cairns & Gallagher, 2004). Additionally, material was sampled during the second half of the photoperiod when WSC levels are elevated (Cairns, 2003). In most of the tissues sampled, polymeric fructan was the major component of total WSC, and this relationship was also demonstrated by the QTL analysis; total WSC QTL always coincided with fructan QTL. In the autumn, the youngest fully expanded leaf contained, on average across the mapping family, c. 160 mg g−1 dry matter as fructan, which was 59% of the value for tiller bases. Total WSC content for the same tissue was 226 mg g−1 dry matter. This is comparable with published values for herbage cut from the field to assess nutritional status for animal feeds (Radojevic et al., 1994; Wilkins, 2002; Jafari et al., 2003). These reports are for mixed tissue which, when harvested in a predominantly nonreproductive state, is mostly leaf blade, although it may contain a small percentage of leaf sheath material.

Nonstructural plant carbohydrates are involved in many aspects of metabolism and consequently are in a constant state of flux. Fixed point analyses, as presented in this article, do not provide direct evidence of rates of synthesis and breakdown. Nevertheless, the data are the product of these activities and some inferences can be drawn. In the autumn, the mean polymeric fructan concentration across the mapping family was high. It constituted the major part of the total WSC pool and was negatively correlated with all other sugar components, indicating active fructan synthesis. If the low concentrations of small sugars in tiller bases resulted from rapid polymer synthesis, this might establish a sink for the carbon influx necessary to build up carbohydrate reserves in high-WSC-accumulating plants. Many hardy plants lay down reserves in autumn to enable them to maintain basal metabolism in periods of adverse weather during winter (Sagisaka, 1995). A polymeric fructan QTL coincided with an oligomeric fructan QTL in the opposite linkage phase on chromosome 6 for leaf tissue. This suggests that some impairment to polymer extension controlled by this region of the genome may limit polymeric fructan, and therefore total WSC, accumulation in low-WSC plants. High leaf fructan was not associated with low tiller base reserves. In fact, there were positive correlations between leaf and tiller base WSC in both spring and autumn (data not shown). A high WSC genotype provides high-sugar forage for animal feed and maintains high plant reserves, which may be important for plant growth, persistence and stress tolerance traits. This would negate the suggestion that fructan accumulates in leaves as a result of low sink activity in tiller bases, although growth rates may well have decreased.

In the autumn there was more fructose than glucose in both tiller bases and leaves, which indicates a low rate of fructan breakdown, concurrent with the high rate of synthesis. This fructan turnover was similar in both leaves and tiller bases (i.e. in nonstorage and storage tissues), and thus did not appear to be related to any diurnal mobilization of polymers in leaves. In spring, the high fructose : glucose ratio indicated that significant breakdown of fructan reserves was occurring in tiller bases. Polymeric fructan was positively correlated with most other components; high-fructan plants produced high concentrations of sugars for metabolism as growth rates increased. The data provide no evidence on whether fructan synthesis was also occurring in tiller bases at this time.

There was evidence that the high WSC trait was maintained across periods of time, although the QTL positions for different sampling times mostly did not coincide, indicating genetic control from different regions of the genome. High carbohydrate content in the spring was related to the quantity of reserves laid down in the autumn, but additional factors were involved as the correlation coefficients were numerically low. That for the tiller base polymeric fructan content was 0.209 (P < 0.01). QTL at the same position on chromosome 1 did explain some of the variation in both spring and autumn. However, in the autumn there was an additional, adjacent QTL in the opposite linkage phase. Variation in rates of fructan breakdown over the winter may have affected the fructan content in the spring, but this was not detected as a separate QTL. The correlation for leaves was lower (r = 0.149; P < 0.05) but the QTL did not coincide in this instance.

Effect of G × E interactions

The broad-sense heritability of forage WSC content has often been estimated to be relatively high (Humphreys, 1995; Jafari et al., 2003) but the replication in many studies has been within one environment. Although the ranking of varieties for WSC content in the field remains fairly constant, absolute amounts may vary considerably within and between years (Wilkins & Lovatt, 2003; Wilkins et al., 2003). The current study confirms the importance of genetic, environmental and, to a lesser extent, G × E interactions on carbohydrate content.

The correlations between years were numerically low, although usually statistically significant. However, this was mainly an effect of environment reducing the size of genotypic differences, rather than G × E interaction effects on genotype ranking, and significant genetic effects could be partitioned for most traits. These genetic effects were confirmed by the QTL analysis, and QTL were identified for most traits that showed a heritability > 0.45. The exceptions were oligofructan DP > 3 in leaves in the spring and tiller bases in the autumn. In the case of tiller bases in the autumn, this may well be a consequence of the large number of genotypes with an oligofructan DP > 3 content below the threshold of detection. Additionally, in some instances the QTL probably explained as much of the variation as might reasonably have been expected from the heritability values. The large QTL for oligofructan DP > 3 found for leaf tissue in the autumn, explaining 59% of the variation, is directly comparable with the calculated heritability of 0.585 for the trait. The two glucose QTL identified for this same tissue together explained just over 30% of the variation, and the calculated heritability of this trait was 0.398. However, as the mapping family used in the study comprised 188 genotypes, smaller QTL may have been missed (Hyne et al., 1995). Also, when considering the QTL that were identified, it should be remembered that population size may have limited the precision of QTL positioning and QTL effects may have been overestimated. The low heritability for tiller base carbohydrates in the autumn was not expected, but indicates the importance of recent and current environmental conditions on the extent of reserve deposition before winter. The QTL detected for this tissue were the smallest in the study, although they did again explain a reasonable part of genetic variation in the mapping family, as indicated by the calculated heritability. In fact, although few of the QTL explained a significant proportion of the total phenotypic variation, many did explain a good part of the genetic variation expected from the heritability values.

Fourteen QTL were detected with the current protocols designed to characterize primarily genetic effects. In only two cases was more than one QTL detected for a trait in two or more years. This is small compared with the total number of the QTL that were observed in the data with a LOD threshold of 2.0. Fifty-four additional QTL were identified in only 1 yr, 11 of these were large enough to be seen also in the mean data and many instances of multiple QTL per trait occurred. Summing these QTL for a given trait in one particular year explained a higher proportion of the total phenotypic variation. For example, six QTL for leaf polymeric fructan content in the spring in year two explained a total of 34.5% of the variation, compared with 16.3% for the main genetic QTL. Four QTL for leaf fructose content in the spring in year three explained a total of 44.5% of the variation, although no QTL were identified that were consistent over years. Similarly, for leaf fructose content in the spring in year three, four QTL explained 28.0% of the variation in total. A significant proportion of the total phenotypic variation unexplained by the genetic QTL is environmentally determined. Many of the QTL from single years were for oligofructans and small sugars, and in some cases coincided with the QTL clusters found in this study, suggesting that environment can have a particular effect on the composition of the minor components of WSC. Kamoshita et al. (2002) also report that many putative QTL may be identified in only one environment when traits which are strongly affected by environment are examined. In contrast, the proportion of repeatable QTL is higher for agronomic traits, such as heading date and plant height which are more highly conserved, as Lu et al. (1997) found that 14 out of 22 QTL were detected in two or more environments.

Genetic control and candidate genes

The clusters of WSC content QTL reported here are located in regions of the genome that have previously been identified as important in preliminary analyses of single-year data from the WSC F2 mapping family (Humphreys et al., 2003). It is probable that these represent the regions of major control in this material. When an LOD threshold of 4.0 was applied to individual years to look for substantial transitory effects, just five further QTL were identified. Four of these fell within the clusters reported here. Only one (for leaf total WSC measured by anthrone in the spring) occurred in a new area on chromosome 3.

To the best of our knowledge, no other fructan or WSC QTL have been reported for Lolium. The map used in this study was developed from previous maps produced for this family (Armstead et al., 2002, 2004). The total genetic coverage of the current map was equivalent to that described for other recent versions of the map for this population (Armstead et al., 2004; Skøt et al., 2005). The fact that the addition of 90 new markers did not cause overall map expansion (total genetic map distances for cited studies were 628 and 631 cM, respectively) indicates both reliable genotype scoring and good coverage of the recombinational distance within the L. perenne genome. This allows alignment with the International Lolium Genome Initiative reference family (Jones et al., 2002) and consequently allows comparisons of QTL positions across a range of Gramineae, as a result of synteny. However, there appears to be little published information on the positions of fructan or WSC QTL in grasses and related cereals. Hayes et al. (1993) mapped crown fructan content in barley to (Triticeae) chromosome 5, but the lack of common markers makes it difficult to compare the exact position with the only (glucose) QTL found on chromosome 5 in the current study. Leaf sheath and culm nonstructural carbohydrate contents were mapped to chromosomes 5 and 11 in rice, a nonfructan-containing plant, in different years (Ngata et al., 2002), but the syntenic relationship between rice chromosome 11 and L. perenne is, again, not clear. The QTL region on rice chromosome 5 can be aligned with L. perenne chromosome 1, but this alignment is with the other end of the chromosome from the Lolium fructan QTL.

Some care is needed when mapping QTL that are affected by developmental stage. If flowering time varies in a mapping population then it is possible to pick up anomalous QTL related to flowering/heading loci in other phenotype data (Yin et al., 2004). The family used in this study has been shown to segregate markedly for heading date. However, in the current study, no carbohydrate QTL were found on chromosome 7 where the major heading date QTL, accounting for up to 70% of the variation, was located (Armstead et al., 2004). The QTL reported here can therefore be considered to be independent of time-to-flowering differences.

The search for candidate genes for fructan metabolism can begin once QTL positions have been established. In ryegrass, fructans are thought to be synthesized by a range of fructosyltransferases (Pavis et al., 2001). A number of attempts have been made to isolate and map these. However there is considerable sequence homology between fructosyltransferases and invertases, and critical examination of the literature and information in databases has suggested that most of the mapped genes are, in fact, invertases (Gallagher et al., 2004). LpFT1 was designated as a fructosyltransferase on publication (Lidgett et al., 2002) and mapped to chromosome 7. However, it is most likely to be an acid invertase, possibly identical to the Inv 1:2 clone (Gallagher et al., 2004) on chromosome 7. This shows no change in expression in conditions that induce fructan accumulation, and does not correspond to fructan QTL in the WSC F2 mapping family. LpFT2 was initially classified as a fructosyltransferase in the EMBL database (accession AY082350), but later redesignated as an invertase and mapped to chromosome 6 (Johnson et al., 2003). This gene does underlie large fructan QTL in leaves in the autumn. Some cell wall invertases have been shown to have considerable sequence homology with fructan hydrolases (Van den Ende et al., 2004), and the fructan breakdown rate could have a causal relationship with a fructan QTL. However, as LpFT2 is a soluble acid invertase that is predominantly expressed in leaf sheaths (Johnson et al., 2003; Gallagher et al., 2004), it is unlikely to have a role here. The only published mapped gene with functionally verified activity in fructan metabolism is the barley 6-SFT (Wei et al., 2000). However, this maps to chromosome 7, in a position corresponding to LpFT1/Inv 1:2 (Lidgett et al., 2002) distant from any currently detected fructan QTL, and is considered by some to be absent from Lolium (Pavis et al., 2001). There is therefore currently little evidence for fructosyltransferase involvement in these fructan QTL. Further isolation and mapping of fructosyltransferase genes from Lolium or related grasses may show closer associations. Alternatively, the QTL could result from variation in fructan hydrolases or from the activity of regulatory genes.

The one candidate gene directly mapped under a QTL in this study was alkaline invertase. Alkaline invertase is localized in the cytosol, and although its role in cell metabolism remains unclear (Roitsch & González, 2004), its action is to break down sucrose to fructose and glucose. The two alkaline invertase loci on chromosome 6 were identified by a cDNA clone (Gallagher & Pollock, 1998) used as an RFLP probe. Neither locus fell within the interval defined by MQM mapping, but Alkinv 2/3 was close to the peak of the interval-mapped QTL. This locus was the only one amplified by polymerase chain reaction (PCR) primers designed from sequence information (L. Skøt & J. Gallagher, pers. comm.) and is therefore likely to represent the position of the gene. Alkinv 1/4 may be a related locus with sufficient sequence homology for the cDNA probe to hybridize. No corresponding QTL for sucrose was found, but sucrose is also metabolized through other pathways. The allele for high monosaccharide content came from Perma, indicating high alkaline invertase activity in plants with lower WSC content, but higher growth rates (Turner et al., 2002).

Potential applications for fructan QTL and markers

The QTL which explained a significant proportion of the variation in a trait were for polymeric fructan and total WSC in leaves in the autumn, and there may be potential for marker selection here. It is advantageous that these QTL were for leaf tissue as, except during flowering periods, this is the fraction that constitutes the major part of animal fodder. It is unfortunate that few of the new SSRs on the map fall within this region of the genome as they offer opportunities for rapid high-throughput screening. In general, markers could prove particularly valuable in cases like that found on chromosome 1 for tiller base fructan content in the autumn, where two QTL were in repulsion and the traits showed a strong dominance effect, with high fructan being recessive. However, these QTL may help to identify genes controlling carbohydrate traits, thus providing essential information about the basic biology of forage grasses that will be useful in improvement programmes.

Finally, it should be noted that as an outbreeder L. perenne contains considerable variation for a wide range of traits. This article presents data based on only one cross. It is therefore necessary to test these QTL effects in marker-designed crosses, and also to consider what other variation may be available at these loci.


We thank Kevin Smith from Agriculture Victoria and Markku Farrell from IGER for their assistance with some aspects of the experimental work, Ruth Sanderson (IGER) for assistance with statistical analysis, Joe Gallagher (IGER) and Oene Dolstra (Plant Research International, Wageningen) for advice during the preparation of the manuscript, and the BBSRC RASP Initiative, EC FAIR CT98-4063 and DEFRA for funding.