Leaf Rubisco turnover in a perennial ryegrass (Lolium perenne L.) mapping population: genetic variation, identification of associated QTL, and correlation with plant morphology and yield

Abstract

This study tested the hypotheses that: (i) genetic variation in Rubisco turnover may exist in perennial ryegrass (Lolium perenne L.); (ii) such variation might affect nitrogen use efficiency and plant yield; and (iii) genetic control of Rubisco turnover might be amenable to identification by quantitative trait loci (QTL) mapping. A set of 135 full-sib F₁ perennial ryegrass plants derived from a pair cross between genotypes from the cultivars ‘Grasslands Impact’ and ‘Grasslands Samson’ was studied to test these hypotheses. Leaf Rubisco concentration at different leaf ages was measured and modelled as a log-normal curve described by three mathematical parameters: D (peak Rubisco concentration), G (time of D), and F (curve standard deviation). Herbage dry matter (DM) yield and morphological traits (tiller weight (TW), tiller number (TN), leaf lamina length (LL), and an index of competitive ability (PI)) were also measured. The progeny exhibited continuous variation for all traits. Simple correlation and principal component analyses indicated that plant productivity was associated with peak Rubisco concentration and not Rubisco turnover. Lower DM was associated with higher leaf Rubisco concentration indicating that Rubisco turnover effects on plant productivity may relate to energy cost of Rubisco synthesis rather than photosynthetic capacity. QTL detection by a multiple QTL model identified seven significant QTL for Rubisco turnover and nine QTL for DM and morphological traits. An indication of the genetic interdependence of DM and the measures of Rubisco turnover was the support interval overlap involving QTL for D and QTL for TN on linkage group 5 in a cluster involving QTL for DM and PI. In this region, alleles associated with increased TN, DM, and PI were associated with decreased D, indicating that this region may regulate Rubisco concentration and plant productivity via increased tillering. A second cluster involving QTL for LL, TN, PI and DM was found on linkage group 2. The two clusters represent marker-trait associations that might be useful for marker-assisted plant breeding applications. In silico comparative analysis indicated conservation of the genetic loci controlling Rubisco concentration in perennial ryegrass and rice.

Nitrogen remobilization, perennial ryegrass, QTL analysis, Rubisco turnover.

Introduction

Perennial ryegrass (Lolium perenne L.) is an important pasture feed crop for ruminants in the temperate regions of the world. Attainment of maximal yields in this species depends on high input of inorganic fertilizer nitrogen (N) (Daepp et al., 2001). Dependence on fertilizer N inputs is a major sustainability issue in modern agriculture due to the reliance on fossil fuels for their production, and the associated economic and ecological costs. It is now widely recognized that sustainable crop production will depend on successful development of cultivars that allow high productivity in low soil N environments. Internal recycling of N, particularly N remobilization from senescing plant organs for new tissue growth, is a strategy evolved by plants to adapt to N-limiting conditions (Vazquez de Aldana and Berendse, 1997). Remobilization reduces the demand for exogenous N supply and better knowledge of its genetic regulation may assist the breeding of high-yielding N-efficient cultivars that support profitable and environmentally friendly agricultural systems.

In intact L. perenne plants, most of the N for new leaf growth is supplied by senescing leaves (Santos et al., 2002; Van Drecht et al., 2003). Varietal and inter-specific genetic variation in N remobilization has been reported in forage grasses (Thornton et al., 1993; Wilkins et al., 1997). Such variation provides potential for genetic manipulation to improve plant N economy and N use efficiency. Free amino acids play an important role in N storage in grasses, but some work in L. perenne indicates that soluble proteins are also important reservoirs of N for leaf regrowth (Louahlia et al., 2000). Most of the N remobilized during leaf senescence is contained in the chloroplasts (Feller and Fischer, 1994), with the photosynthetic enzyme Rubisco (EC 4.1.1.39) representing the single greatest component. In C₃ plants, Rubisco accounts for approximately 50% of the total soluble protein and 20–30% of total leaf N (Makino, 2003). The N invested in Rubisco accounts for a substantial proportion of total plant N, and some researchers (Mae et al., 1983; Millard, 1988) have suggested a role in N storage besides its catalytic function in photosynthesis. Rubisco’s hypothesized storage role is an example of N storage by recycling as opposed to reserve formation (Millard, 1988).

Rubisco content increases rapidly during leaf expansion, reaches a maximum around full expansion, and then declines as it is degraded during leaf senescence with the N remobilized for new growth. At any time during leaf development, the leaf Rubisco content/concentration is the result of a balance between its synthesis and degradation (i.e. Rubisco turnover). ¹⁵N tracer (Mae et al., 1983; Makino et al., 1984) and change in transcript abundances (Suzuki et al., 2001) experiments have shown that Rubisco content during leaf expansion is mainly determined by its synthesis, while degradation is the major determinant during senescence. According to Irving and Robinson (2006), the time course of leaf Rubisco content can be described by a log-normal curve which represents the combined effects of simultaneous synthesis and degradation, with Rubisco degradation assumed to occur exponentially according to first-order kinetics. The log-normal curve is described by three mathematical parameters: D (maximum Rubisco content/concentration), F (fitted logistic curve standard deviation, a measure of Rubisco turnover time), and G (leaf age in days at which the maximum Rubisco content/concentration occurs). These parameters provide a way to quantify genotypic differences in Rubisco dynamics.

Despite the potential N storage role of Rubisco, little work has been done to understand how Rubisco turnover affects plant yield. It is known that N nutrition drives plant dry matter production, and in many higher plants photosynthetic capacity correlates positively with leaf N content (Evans, 1989). The effect of Rubisco N on plant yield is likely to be influenced by the effectiveness with which plants are able to balance between N residence time for photosynthesis and N remobilization and re-allocation for new tissue growth. Identification of genetic factors controlling this balance may be of practical importance in breeding N-efficient plants. One way to elucidate this is to study genetic variability in Rubisco turnover in relation to herbage yield which, in L. perenne, is determined by an interaction of many morphological traits (Chapman and Lemaire, 1993; Bahmani et al., 2001). If such variation exists, plants with Rubisco turnover characteristics that support increased herbage yield may then be targeted in marker-aided improvement of N economy and N use efficiency.

DNA markers and linkage maps for L. perenne have been developed (Jones et al., 2002; Faville et al., 2004) and used to locate quantitative trait loci (QTL) for important traits (Yamada et al., 2004; Jensen et al., 2005; Sartie et al., 2011). The association of easily screened DNA markers with traits of interest provides the opportunity for marker-assisted selection breeding. This is expected to improve efficiency in breeding by enabling selection at the seedling stage for traits expressed later in plant life, or reducing the reliance on expensive and time-consuming field selection trials. Examples of application of marker-assisted selection in L. perenne breeding include optimization of genetic diversity of parents in polycross breeding (Kölliker et al., 2005) and, more recently, QTL-marker-assisted selection for nutritive value (Turner et al., 2010). A range of markers located on L. perenne genetic linkage maps have been mapped in some Triticeae cereals, which have led to linkage maps that are also useful for comparative studies (Jones et al., 2002; Sim et al., 2005). Conserved synteny and collinearity among genomes of the grasses make it possible to infer valuable genetic information from well-characterized cereal crops such as rice to L. perenne.

The present study evaluated a L. perenne mapping population for variation in Rubisco turnover parameters as described by Irving and Robinson (2006) and for correlation of Rubisco turnover and herbage yield characteristics. QTL analysis was also used to identify DNA markers linked to Rubisco turnover loci and to morphological traits.

Materials and methods

Plant material, cultivation, and sampling

The mapping population, I × S, consisted of 135 F₁ progeny derived from a cross between a single genotype from the cultivar ‘Grasslands Impact’ (seed parent, I) and a single genotype from the cultivar ‘Grasslands Samson’ (pollen parent, S) (Crush et al., 2007; Sartie et al., 2011). The parents exhibit clear morphological differences with ‘Grasslands Impact’ having a higher number of smaller tillers per plant, longer leaf sheaths, and narrower leaves than ‘Grasslands Samson’ (Sartie et al., 2009). The ‘Grasslands Impact’ parent was infected with a naturally occurring standard (wild-type) endophyte, Neotyphodium lolii. Due to maternal endophyte transmission, the progeny were infected by the same endophyte. Twelve clonal tillers from each of the F₁ and parental lines were transplanted into 1.5 l plastic planter bags on 23 May 2007. The plants were grown under natural light in a glasshouse at the Massey University Plant Growth Unit in a completely randomized block design with three replicates. The mean temperature in the glasshouse was 13.6 °C (range 2.0–24.6 °C). The mean solar radiation was 4.96 MJ m^–2 day^–1 (range 0.3–10.6). The composition of the potting mixture was 55% alluvial, friable silt loam soil and 45% builder’s sand, with a slow-release 8–9-month osmocote 16/3.5/10 fertilizer (3g l^–1 sand/soil mix to supply readily absorbed N, phosphorus and potassium). The plants were watered twice a day by a capillary irrigation system. On 3 July 2007, two tillers on each plant were identified and tagged. The newest expanding leaf on each tagged tiller was identified as the experimental leaf. On the first sampling date, the lengths of all experimental and adjacent leaves on each tagged tiller were measured. The age of the experimental leaf was then estimated as:

where LED is the leaf elongation duration measured as the average time (days) from leaf tip appearance to full elongation in the leaves of 10 non-test plants, L_e is the length (mm) of the experimental leaf, L_m is the length (mm) of the adjacent leaf, and 1.2 is a constant indicating an expected 20% increase in leaf length between successive leaves (Verdenal et al., 2008; Sartie et al., 2009).

On each sampling date (1, 4, 8, 12, 20, 27, 34, and 41 days after the experimental leaf marking), one leaf per genotype per replicate was excised at the ligule, weighed and immediately frozen in liquid N, and stored at –80 °C until further analysis.

Determination of Rubisco

The frozen leaf blade was homogenized in 50mM sodium phosphate (pH 7.5) buffer containing 0.8% (v/v) β-mercaptoethanol, 2mM iodoacetic acid, and 10% (v/v) glycerol (at a ratio of 1:9 g:ml) using a chilled mortar and pestle. To a 200-μl aliquot of this mixture in a 1.5-ml centrifuge tube, 2 μl Triton X-100 was added. The mixture was then vortexed and centrifuged at 10,000 g for 15min at 4 °C. Lithium dodecyl sulphate (8.6 μl) and β-mercaptoethanol (5.6 μl) were added to the supernatant and the samples boiled at 100 °C for 90 s then pulse-centrifuged. The denatured proteins were resolved at room temperature by sodium dodecylsulphate polyacrylamide gel electrophoresis. The resolving and stacking gels contained 12 and 5% acrylamide, respectively. After electrophoresis, the gels were stained with (v/v) 0.25% Coomassie brilliant blue R-250 dye, 20% methanol, and 20% acetic acid at room temperature for 3–5h, then destained with successive changes of 20% methanol and 20% (acetic acid. Sections of the gel corresponding to the large subunit of Rubisco were excised and transferred to a 2-ml centrifuge tube containing 1ml formamide for extraction of the stain from the gel. The tubes were subjected to gentle shaking at room temperature for 12h. The concentration of Rubisco was determined using a spectrophotometer (at 595nm) (Makino et al., 1986). A calibration curve was constructed with partially purified Rubisco from spinach (Sigma-Aldrich). Rubisco concentration of frozen leaves was converted to per unit of leaf dry matter basis using the mean oven (80 °C, 48h) leaf dry matter of four replicate leaves from non-test plants harvested on the same days as those for Rubisco analysis.

Derivation of Rubisco turnover parameters

The Rubisco turnover log-normal curve parameters, D, F, and G (Irving and Robinson, 2006), were derived using least-squares-non-linear regression in Sigma plot (version 11). The ratio of D:F was also calculated to identify genotypes hypothetically expected to have rapid Rubisco turnover, indicating a larger cost of Rubisco synthesis and/or less photosynthetic contribution over the leaf lifespan. Apparent systematic lack of fit to the log-normal curve of Rubisco concentration was noted in some genotypes. To examine this lack of fit, the average leaf Rubisco concentration at the final two time points (days 34 and 41) was calculated to estimate the quantity of Rubisco concentration TL (mg (g leaf dry matter)^–1)) not remobilized in leaf senescence. The highest observed Rubisco reading was calculated as the average of the two Rubisco concentrations on either side of the curve maximum (PK, (mg (g leaf dry matter)^–1).

Evaluation of morphological traits and plant dry matter

Herbage yield and its component morphological traits were evaluated in two replicates per genotype at the end of the experiment, when all Rubisco samples had been collected. Leaf lamina length (LL, mm) was measured as the mean of three youngest fully expanded leaves, one each from three tillers of one genotype in each replicate. Tiller number (TN) was measured as the mean number of tillers counted on three clones of each genotype in each replicate. Plant dry matter (DM, g) was measured as the mean weight of oven-dried (80 °C, 48h) herbage cut at soil level of all plants for which a tiller count had been taken. Mean tiller weight of plants (TW, g) was derived from DM and TN as:

The ability of each plant to increase its own tiller density and so establish greater leaf area at the expense of its neighbours (here designated as productivity index, PI) was derived according to self-thinning theory from TW and TN as follows:

PI = Log(TW) – [–1.5*Log(TN)/A]

where –1.5 is a constant assumed for size/density compensation (Matthew et al., 1995) and A is surface area of the filled planter bag (m²).

Statistical analysis

Data analyses were carried out in SAS (SAS Institute, 2003). Rubisco turnover and morphological data were subjected to one-way ANOVA (Proc ANOVA) to determine trait means and the significance of genotype effects. Broad-sense heritability (H_b) was calculated from the ANOVA data as:

H_b = δ²_g/(δ²_g + δ²ε/n)

where δ²_g is the genetic variance, δ²_ε is the error variance, and n is the number of replications for each genotype (e.g. Turner et al., 2010).

The Anderson Darling statistic (A²-test) was used to assess normality of the averaged data (normal option in Proc UNIVARIATE). Pearson’s correlation coefficients were determined by pairwise comparisons of averaged trait data (Proc CORR) to establish the degree of association among traits. Principal component analysis (Proc PRINCOMP) was performed to obtain fewer components that describe maximum variance in the studied traits and to establish patterns of association among interacting traits.

QTL analysis

An existing consensus genetic linkage map developed for I × S (e.g. Sartie et al., 2011) was used for genetic analysis. This map is based on simple sequence repeat (SSR) and sequence-tagged site markers derived from a proprietary L. perenne expressed sequence tag (EST) resource (Sawbridge et al., 2003; Faville et al., 2004) supplemented by EST-SSRs from tall fescue (Festuca arundinacea Schreb.) (Saha et al., 2004) prefixed ‘nffa’. The I × S map was further enhanced by the addition of 26 SSRs developed from ryegrass GeneThresher ryegrass sequences (marker loci prefixed ‘rv’) (Gill et al., 2006) and seven SSRs from a ryegrass reference set (Bach Jensen et al., 2005). QTL analysis was carried out using MapQTL 4.0 (Van Ooijen et al., 2002). QTL detection followed the process described by Sartie et al. (2011). Each trait was first analysed by simple interval mapping using the phenotypic mean value of each genotype. QTL prediction for interval mapping was declared for a logarithm of odds (LOD) threshold of 2.5 and if the QTL position was supported by Kruskal–Wallis non-parametric single locus analysis. Declared QTL were further resolved by multiple QTL model (MQM) in MapQTL 4.0, using cofactors chosen with the help of the automatic cofactor selection option (Van Ooijen et al., 2002). MQM was performed on a trait-by-trait basis using all identified cofactors for each trait. Permutation analysis (1000 iterations) was used to establish a genome-wide significance (P < 0.05) value defined as a minimum threshold for each trait in MQM (Doerge and Churchill, 1996). For each form of QTL analysis, the maximum LOD value associated with the most closely linked marker and the proportion of the phenotypic variance attributable to the QTL were tabulated. Map positions were defined by the peak ± 2 LOD (Van Ooijen, 1992). QTL nomenclature consisted of the trait acronym followed by the linkage group.

Allelic effects at the detected QTL were estimated using phenotypic trait means for the four QTL genotypes (ac, ad, bc, and bd) automatically calculated in MapQTL 4.0. Maternal effect (difference in effect of the alleles (a and b) inherited from the ‘Grasslands Impact’ parent (I)) and paternal effect (difference in effect of the alleles (c and d) inherited from ‘Grasslands Samson’ parent (S)) and the interaction effect (deviation from additivity, where 0 is complete additivity, assuming heterozygosity in both parents) were estimated as described by Knott et al. (1997) and Sewell et al. (2002):

maternal effect (I) = (ac + ad) – (bc + bd)

paternal effect (S) = (ac + bc) – (ad + bd)

interaction effect (INT) = (ac +bd) – (ad + bc)

For mapped ryegrass ESTs, sequence alignment by BLAST-like alignment tool (threshold values of < E–05; SRA-interacting protein domain >85% over >100bp) was used to estimate identity with homologous rice genome position in the TIGR rice pseudo-molecule assembly hosted at www.gramene.org.

Results

Phenotypic variation and heritability of traits

As illustrated by the wide range of variation, the F₁ perennial ryegrass family segregated for all traits measured in this study (Table 1). For all traits, variation among the progeny exceeded that between the parent plants. ANOVA indicated significant (P < 0.05) phenotypic variation among F₁ progeny for all evaluated traits except time of Rubisco peak, G (Table 1). Broad-sense heritability values of herbage yield parameters were higher than those of Rubisco turnover parameters D, F, and G (Table 1). Normality testing indicated non-normality for all traits except LL, D, and G (Table 1). Non-normality arose from a small number of outliers and, therefore, data transformation was not considered necessary.

Table 1.

Distribution of the traits measured in the 135 Grasslands II mapping population progeny and parent plants. Normality distribution results computed from ANOVA residuals and based on Anderson-Darling (A²) test statistic for normal data distribution. H_b, broad sense heritability; LL, leaf lamina length (mm); TN, tiller number; TW, tiller weight (mg); DM, plant dry matter (g); PI, productivity index; D, maximum Rubisco concentration (mg (g leaf dry matter)^–1); F, Rubisco curve width measure; G, time of D; D/F, D:F ratio; PK, average of the two Rubisco content determinations closest to D (mg (g leaf dry matter)^–1); TL, residual Rubisco content measured of the average of the last two time points (mg (g leaf dry matter)^–1); S, ‘Grasslands Samson’; I, ‘Grasslands Impact’.

Trait	F1 progeny				Parent value		Hb	Normality
	Min.	Max.	Mean ± SD	ANOVA P-value	I	S		Kurtosis	A2	P-value
LL	178.0	433.0	293.0±55.0	<0.01	336.0	321.0	0.62	0.14	0.61	0.11
TN	13.0	106.0	62.0±22.0	<0.01	104.0	52.0	0.57	2.16	1.65	<0.01
TW	8.0	225.0	96.0±32.0	<0.01	80.0	94.0	0.50	13.8	3.39	<0.01
DM	0.2	12.6	4.8±2.8	<0.01	6.7	4.8	0.70	1.70	2.54	<0.01
PI	3.24	5.21	4.5±0.4	<0.01	4.9	4.5	0.66	1.14	3.04	<0.01
D	31.6	64.3	45.6±6.6	<0.01	49.4	42.8	0.39	2.44	0.82	0.04
F	0.6	1.2	0.8±0.1	0.03	0.77	0.84	0.25	7.09	4.36	<0.01
G	9.0	17.9	12.9±1.6	0.96	12.2	13.5	0.24	2.83	0.71	<0.07
D/F	25.8	104.3	56.3±12.4	–	64.0	50.7	–	–	–	–
PK	27.1	77.0	43.2±8.2	–	47.6	45.0	–	–	–	–
TL	11.2	38.7	19.4±4.4	–	13.6	18.8	–	–	–	–

Trait correlations

Table 2 indicates that the correlations among DM and all morphological traits were positive and highly significant (P < 0.01). Rubisco turnover parameters generated numerically low correlation coefficients with DM and all morphological traits. Of the three log-normal curve parameters, D correlated negatively at varying levels with DM and all morphological traits, while F showed a tendency toward a negative correlation (P < 0.1) with DM. The measure of Rubisco concentration at senescence, TL correlated negatively with DM and all morphological traits. DM and its morphological component traits associated more strongly with TL than any of the other Rubisco turnover parameters evaluated in this study. Among Rubisco turnover parameters themselves, D correlated negatively with F and G, and positively with D/F, PK, and TL.

Table 2.

Pearson correlation coefficients showing interrelations of traits measured in the 135 Grasslands II mapping population progeny and parent plants. †, *, **, significance at P < 0.1. P < 0.05, and P < 0.01, respectively. LL, leaf lamina length (mm); TN, tiller number; TW, tiller weight (mg); DM, plant dry matter (g); PI, productivity index; D, maximum Rubisco concentration (mg (g leaf dry matter)^–1); F, Rubisco curve width measure; G, time of D; D/F, D:F ratio; PK, average of the two Rubisco content determinations closest to D (mg (g leaf dry matter)^–1); TL, residual Rubisco content measured of the average of the last two time points (mg (g leaf dry matter)^–1); NS, not significant.

Trait	LL	TN	TW	DM	PI	D	F	D/F	G	PK
TN	0.45**
TW	0.57**	0.32**
DM	0.62**	0.80**	0.66**
PI	0.58**	0.87**	0.67**	0.86**
D	–0.23*	–0.15†	–0.20*	–0.21*	–0.17*
F	NS	NS	NS	–0.14†	NS	–0.24**
D/F	–0.16†	NS	NS	NS	NS	0.81**	–0.74**
G	NS	NS	NS	NS	NS	–0.20*	–0.39**	NS
PK	–0.22**	NS	NS	0.15†	NS	0.78**	NS	0.57**	–0.21*
TL	–0.28**	–0.18*	–0.25**	–0.31**	–0.21*	0.29**	0.27**	NS	0.17*	0.36**

Principal component analysis

Table 3 summarizes the size and sign of coefficients defining the first four of the 11 available principal components (PCs) collectively explaining 83% of data variation. The first principal component (PC1) accounting for 36% of data variation separated genotypes based on plant size. This PC was characterized by a positive contribution from DM and all morphological traits. This was contrasted by lower leaf Rubisco concentration throughout the leaf lifespan (negative coefficients for D, PK, and TL). The negative correlation of D and PK with F was the dominant feature of the second principal component (PC2) explaining 23% of the total data variation. The third principal component (PC3) discriminated genotypes based on the pattern of Rubisco turnover with low values of G (i.e. early peak Rubisco concentration) associated with high levels of Rubisco at all stages of leaf development (coefficients of 0.25, 0.38, and 0.25 for D, PK, and TL, respectively). This PC explained 14% of the total data variation. The fourth principal component (PC4) explained 10% of the data variation and associated high Rubisco concentration at leaf senescence (TL, coefficient 0.74) with a modest tendency for a higher TN (coefficient 0.23). Fig. 1 illustrates the distribution of the genotypes along the first two PCs. Each of the four PCs (results for PC3 and PC4 are not shown) roughly divided the progeny into two equal groups.

Table 3.

Component weights of the first four of the eleven extracted principal components for traits evaluated in 135 genotypes of the Grasslands II perennial ryegrass mapping population and the parent plants. LL, leaf lamina length (mm); TN, tiller number; TW, tiller weight (mg); DM, plant dry matter (g); PI, productivity index; D, maximum Rubisco content (mg (g leaf dry matter)^–1); F, Rubisco curve width measure; G, time of D; D/F, D:F ratio; PK, average of the two Rubisco content determinations closest to D (mg (g leaf dry matter)^–1); TL, residual Rubisco content measured as the average of the last two time points (mg (g leaf dry matter)^–1).

Trait	Principal component
	1	2	3	4
LL	0.38	0.05	0.11	–0.09
TN	0.38	0.18	0.09	0.23
TW	0.36	0.10	0.11	–0.06
DM	0.45	0.18	0.07	0.08
PI	0.44	0.21	0.11	0.17
D	–0.25	0.47	0.25	–0.04
F	0.01	–0.42	0.55	0.15
D/F	–0.16	0.57	–0.16	–0.12
G	0.02	0.04	–0.60	0.55
PK	–0.22	0.40	0.38	0.09
TL	–0.22	–0.01	0.25	0.74
Variation (%)	35.90	23.30	13.90	10.20

Fig. 1.

Principal component scatter plots based on five herbage yield traits and six Rubisco turnover parameters measured in 135 F₁ genotypes and two parent plants of the I × S perennial ryegrass mapping population. The positions of the two parents, ‘Grassland Samson’ (SAM) and ‘Grasslands Impact’ (IMP), are indicated by arrows.

An important feature of the data set was the statistically significant negative correlation of DM and related measures TN, TW, LL, and PI with Rubisco turnover curve parameters D and TL trait associations (Table 2), also deduced from PC1 (Table 3). Since PC1 is a multivariate ranking of the 135 genotypes, it is informative to examine trait means for genotypes with high and low scores for PC1. For 24 genotypes that exhibited large positive scores (≥ +2), mean ± SD of relevant traits were D = 39.5±0.8mg (g leaf dry matter)^–1, TL = 17.2±0.6mg (g leaf dry matter)^–1, and DM = 8.5±0.4g. By contrast, for 27 genotypes with the most negative scores (≤ –2), the corresponding trait means were D = 50.0±1.2mg (g leaf dry matter)^–1, TL = 22.6±1.0mg (g leaf dry matter)^–1, and DM = 1.4±0.2g.

QTL analysis

All linkage groups (LGs) with the exception of LG3 and LG6 displayed significant associations between markers and traits (Fig. 2, Table 4). There was at least one significant QTL detected by MQM for each evaluated trait except Rubisco curve width, F (Table 4). Individually, identified QTL were of low to moderate effect, explaining 7.0–24.1% of the total phenotypic variation (Table 4). Phenotypic variation explained by each trait ranged from 15.6 to 29.3% for Rubisco turnover traits and from 14 to 42.6% for herbage yield traits (Table 4). Allelic effects indicated additive effects at all identified QTL loci, with complete additivity observed for DM and PI QTL on LG2 (Table 4).

Fig. 2.

Locations of QTL for Rubisco turnover and herbage yield traits on the I × S genetic map of perennial ryegrass, based on the results of multiple QTL mapping. Molecular markers are indicated on the left of each linkage group (LG), and QTL are indicated as filled bars on the right. Bar lengths indicate a LOD drop of 2.0 on either side of the maximum likelihood position. Map distances are indicated by the centimorgan (cM) scale at the left. LL, leaf lamina length (mm); TN, tiller number; TW, tiller weight (mg); DM, plant dry matter (g); PI, productivity index; D, maximum Rubisco content (mg (g leaf dry matter)^–1); F, Rubisco curve width measure; G, time of D; D/F, D:F ratio; PK, average of the two Rubisco content determinations closest to D (mg (g leaf dry matter)^–1); TL, residual Rubisco content measured as the average of the last two time points (mg (g leaf dry matter)^–1).

Table 4.

QTL for morphological and Rubisco turnover parameters detected by multiple QTL mapping in 135 F₁ plants of the Grasslands II L. perenne population and the two parent plants. LL, leaf lamina length (mm); TN, tiller number; TW, tiller weight (g); DM, plant dry matter (g); PI, productivity index. D, maximum Rubisco content (mg (g leaf dry matter)^–1; F, Rubisco curve width measure; G, time of D (days); D/F, D:F ratio; PK, average of the two Rubisco content determinations closest to D (mg (g leaf dry matter)^–1; TL, residual Rubisco content measured as the average of the last two time points (mg (g leaf dry matter)^–1; LOD threshold, genome-wide logarithm of the odds score for declaring significant QTL at P ≤ 0.05; I, ‘Grasslands Impact’; S, ‘Grasslands Samson’; INT, interaction; PVE, phenotypic variation explained.

Trait	Linkage group	Nearest marker	Position (cM)	Support interval (± 2 LOD)	LOD threshold	Maximum LOD	Allelic effects			PVE (%)
							I	S	INT
LL	2	pps0113	40.1	38.2 – 42.0	3.91	4.48	285.6	23.7	-116.3	16.2
TN	2	pps0113	37.6	36.1 – 40.0	4.00	4.26	-15.0	24.1	14.7	10.1
	5	rv0340	61.6	57.3 – 69.6		6.56	-6.7	-29.1	20.7	16.4
TW	5	pps0036	57.5	50.0 – 58.1	3.96	4.60	-22.0	-29.9	36.9	14.0
DM	1	pps0290	32.3	29.0 – 35.3	4.03	4.17	-0.8	-0.7	3.6	12.4
	2	pps0113	40.1	37.4 – 40.5		5.07	-2.3	3.8	0.0	8.9
	5	pps1115	72.1	69.4 – 72.4		4.29	-2.9	-8.5	-3.0	21.3
PI	2	pps0113	40.1	37.2 – 41.5	3.90	5.05	-0.2	0.4	0.0	7.0
	5	pps1115	69.6	64.8 – 72.4		5.25	-0.3	-1.1	-0.7	20.3
D	5	rv0340	61.6	59.8 – 63.0	3.97	4.37	0.8	8.6	-3.5	12.5
	7	pps0065	34.2	30.8 – 34.3		4.30	-6.3	-7.0	0.3	12.3
D/F	5	pps0509	3.1	00.0 – 9.70	3.94	3.95	3.1	17.9	8.3	17.3
G	2	pps0514	9.9	00.0 – 21.6	3.90	4.13	1.0	2.2	-0.4	15.3
	7	pps0002	100.5	96.7 – 101.7		4.17	1.2	-1.8	0.7	14.0
PK	4	pps0130	65.0	64.6 – 70.3	3.90	4.69	-37.5	28.8	-32.4	15.6
TL	7	pps0060	34.4	31.1 – 34.8	3.88	4.28	-27.7	13.7	-4.7	24.1

Seven significant QTL for Rubisco turnover curve parameters were identified on LG2, LG4, LG5, and LG7 (Fig. 2, Table 4). Two equal-effect QTL for D were mapped, one each on LG5 and LG7. The support interval of QTL for D overlapped with that of TN QTL on LG5 and TL QTL on LG7 (Fig. 2, Table 4). Two QTL for G, qG-2 (phenotypic variation 15.3%) and qG-7 (phenotypic variation 14%) were identified on LG2 and LG7, respectively. Both QTL were distinctly separated from all QTL for the other parameters evaluated in this study. There was one significant QTL for PK (phenotypic variation 15.6%) and one for D/F (phenotypic variation 17.3%) identified on LG4 and LG5, respectively. Results indicated Rubisco turnover QTL effect was largely driven by alternative alleles from the ‘Grasslands Samson’ parent (Table 4).

A total of nine QTL were identified for herbage yield traits. These included clustering at two chromosomal locations on LG2 and LG5. On LG2, qLL-2, qTN-2, qPI-2, and qDM-2 had either coincidental peaks or support interval overlaps near marker pps0113 (Fig. 2, Table 4). Alternative alleles from the ‘Grasslands Samson’ parent were most influential for all QTL at this locus with the exception of qLL-2, for which alternative alleles from the ‘Grasslands Impact’ parent were more influential (Table 4). The cluster on LG5 involved support interval overlaps of qTN-5, qPI-5, and qDM-5 which were also in close proximity with qTW-5 and qD-5 (Fig. 2, Table 4). The qDM-5 and qPI-5 peaks coincided, and each of their support intervals coincided with that of qTN-5 (Table 4). Estimated allelic effects indicated a significant influence of alleles derived from the ‘Grasslands Samson’ parent for all LG5 cluster QTL. Besides the appearance in clusters, DM had an additional QTL (qDM-1) detected on LG1. This QTL was largely influenced by alternative alleles from the ‘Grasslands Impact’ parent (Table 4).

In silico comparative analysis

Perennial ryegrass LG5 and LG7 on which QTL determining D were detected (Table 4, Fig. 2) have syntenic relationships with rice chromosomes 9 and 8, respectively (Jones et al., 2002; Sim et al., 2005). Ishimaru et al. (2001) mapped QTL for Rubisco content at 25 days and 5 days after heading on chromosomes 9 and 8, respectively. Markers flanking the region of the Ishimaru et al. (2001) Rubisco content on chromosome 9 mapped to a genomic interval of 18–22Mb on the TIGR pseudo-molecule genome assembly (www.gramene.org), with the closest marker C570 at 21.3Mb. The ryegrass QTL identified on LG5 aligned to the region on rice chromosome 9 at approximately 19–22Mb (Table 5). QTL for Rubisco content at 5 days after heading on chromosome 8 was delineated by the region 19–23Mb on the TIGR pseudo-molecule genome assembly, with the closest marker G1073 at 20.7Mb. The ryegrass D QTL on LG7 aligned to the rice genome described by approximately 20–28Mb (Table 5).

Table 5.

Comparative rice genomic information (www.gramene.org) for ryegrass ESTs mapped using embedded SSRs on D QTL-associated regions of LG5 and LG7 in perennial ryegrass mapping population I × S. Ryegrass genetic linkage map information are on the left (EST position reported in centimorgans, cM) and rice genome information, as determined by BLAST-like alignment tool analysis of ryegrass ESTs are on the right. Start, Start position on rice chromosome; Mb, megabase pairs). SSR loci reported at D QTL peak are highlighted bold.

Ryegrass map data			Rice genome
EST-SSR	LG	cM	E value	Chromosome	Start (Mb)
pps0074	5	47.8	8.00E–34	9	14.6
pps0359	5	48.7	5.00E–79	9	14.8
pps0826	5	53.3	5.00E–64	9	16.9
pps0036	5	57.5	No hit
pps0377	5	58.4	5.50E–92	9	17.0
rv1112	5	60.6	2.30E–05	9	19.9
rv0340	5	61.6	No hit
pps0718	5	64.6	No hit
pps1115	5	72.1	8.10E–06	9	22.2
pps0056	5	73.8	2.30E–18	9	22.3
pps0049	7	29.6	4.00E–83	8	26.4
pps0494	7	31.9	2.00E–27	8	27.9
pps0065	7	34.2	0.00E+00	8	27.8
pps0060	7	34.4	1.00E–132	8	20.7
pps0462	7	35.1	0.00E+00	8	21.2
pps0705	7	36.4	3.00E–06	8	13.5
ppt003	7	37.0	1.00E–05	8	3.5

Discussion

Trait variation within the mapping population

Substantial variation was observed within the F₁ mapping population for all evaluated morphological and physiological parameters. This variation is comparable to that reported for a wide range of traits in this species (Crush et al., 2007; Turner et al., 2008; Sartie et al., 2011). All traits exhibited continuous variation (Table 1), indicating polygenic inheritance. The F₁ variation extended beyond the means of the parental genotypes, indicating transgressive segregation or the age difference between the parent plants (older) and the progeny (younger). The low heritability values for the measures of Rubisco turnover indicates a strong environmental influence on Rubisco turnover, even in a controlled glasshouse experiment. For traits with low heritability, progress from phenotypic selection will be slow, because so much of the variation is due to environmental variation or genotype–environment interaction, and will not be passed on to the next generation. Therefore, marker-assisted selection (which can result in greater genetic progress of low heritability traits, Hospital et al., 1997) would be preferential over phenotypic selection for improvement of Rubisco turnover.

Trait associations

The strongest relationships were found among morphological traits and yield (Table 2). The signs of trait correlations were consistent with previous data from a study of the same mapping population (Sartie et al., 2011), except for the positive correlation of TN with LL and TW. In this study, in contrast to that of Sartie et al. (2011), plants were spaced such that interplant competition was minimized, allowing for increased tillering of large plants. Thus, increased yield was a result of both increased yield per tiller and number of tillers (Tables 2 and 4). The link between Rubisco turnover pattern and plant morphology and yield was weak, but one notable finding was that where leaf Rubisco concentration was higher, plants tended to be smaller. This finding was subsequently confirmed in a second experiment with a different plant population (Khaembah, 2009) and also agrees with the findings of Wilkins et al. (2000), who reported a negative correlation of dry matter with leaf N levels (which correlates with leaf Rubisco levels). Possible interpretations of the negative association of DM with leaf Rubisco concentration in this study are: (i) that accumulation of large amounts of Rubisco (more than is needed for photosynthesis) represents an energy cost that negatively impacts on yield; or (ii) there is asynchrony of leaf development and Rubisco turnover. The first interpretation possibly indicates down-regulation of Rubisco activation as reported in previous studies (Ray et al., 2003; Millard et al., 2007; Suzuki et al., 2009). Due to intrinsic limits in light capture above a certain level, increases in Rubisco levels cause a decline in Rubisco activation state. The second interpretation is likely due to lack of synchronization between leaf development and Rubisco turnover, which can potentially lead to retention of high amounts of Rubisco levels in senescent leaves. Plastochron data were not collated in this study, but the positive association of D with TL (Tables 2 and 3) appears to support the conclusion that leaf development and Rubisco turnover are generally uncoupled.

The correlation of high peak Rubisco concentration in younger leaves and higher residual levels in old leaves observed in this study add to the complexity of the Rubisco–photosynthesis–plant productivity relationship. Many studies (e.g. Evans, 1989; Pooter et al., 1990; Makino, 2003) have pointed to a positive correlation of photosynthesis and leaf N (largely attributed to the fact that 25% of leaf N is contained in Rubisco). Also, accumulation of large amounts of Rubisco by higher plants, especially C₃ plants, has been regarded as a way to compensate for the enzyme’s inefficiency in CO₂ assimilation (see review by Stitt and Schulze, 1994). That the resulting increase of Rubisco in the leaf itself serves to down-regulate its activity in photosynthesis (Ray et al., 2003; Suzuki et al., 2009) is paradoxical. Plant N productivity could benefit if excess and inactive Rubisco was degraded to release all or most of its N for re-allocation to new growth. From the results of this study, it appears that this benefit is not realized, possibly because senescence in high leaf N plants occurs rapidly, cutting short the N remobilization process. This finding, if proven in further studies, adds another layer of complexity to the high plant leaf N criterion that has been previously used in yield improvement by breeding programmes.

Examination of trait means for genotypes with high and low scores for PC1 (which reflected significant trait associations identified by correlation analysis, Table 2) highlighted that the PC coefficients reflect tangible differences among genotypes, or between groups of genotypes when those groups are selected according to the PC score. Further experimentation is needed to confirm if low D-high yield plants are also efficient at N absorption from the soil. The lower-order PCs showed greater contribution of Rubisco turnover parameters (D, F to PC2, F and G to PC3, and g to PC4), but the contribution of the morphological component diminished. This possibly reflects the weak association of these two sets of traits by correlation analysis (Table 2).

The spacing of plants in this study was such that most leaves were well illuminated. A typical perennial ryegrass sward forms a closed canopy and leaves at different positions are exposed to different light intensities (Gastal and Lemaire, 2002). Leaves at lower positions of the tiller get shaded by growing upper leaves. Studies have shown that a decrease in the level of incident radiation could trigger or accelerate senescence and Rubisco degradation (Hikosaka, 1996). This indicates potential of Rubisco dynamics to differ between leaves of the same age on different tillers, and this is bound to affect productivity of the sward. Also, previous work with rice plants has shown the amount of leaf N and leaf Rubisco content to correlate with the rate of N influx into the leaf (Makino et al., 1984; Imai et al., 2005), indicating that the availability of the substrate might serve as a determining factor of the rate of Rubisco synthesis. In grasses, Lattanzi et al. (2005) reported that leaf senescence was accelerated by N limiting nutrition. Therefore, leaf senescence and Rubisco turnover may be influenced by the nutrition status of plants, and further investigations should include N nutrition treatment.

Genetic control of Rubisco turnover, plant morphology, and yield

Under the environmental conditions of this study, and in the evaluated mapping population, genetic control of the measures of leaf Rubisco turnover was widely spread across the perennial ryegrass genome. The qD-5 and qTN-5 overlap within the LG5 cluster was the only evidence of genetic co-regulation of Rubisco turnover and yield parameters, but appears potentially important for marker-based perennial ryegrass breeding. All three herbage yield-related QTL (qTN-5, qPI-5, and qDM-5) in the cluster had opposite allelic effect to qD-5 (Table 4), implying that an increase in D would result in a decrease of TN, PI, and DM, and vice versa. This finding indicates that selection for increased TN via QTL on LG5 may work mechanistically to reduce D, hence reducing the plant’s energy overload, and allowing greater DM through increased tillering. Another region of interest was the QTL cluster around marker pps0113 on LG2. The alignment of allelic effects involving DM, TN and PI QTL, while opposite for LL QTL indicates either: (i) pleiotropic action by a single gene resulting in increased DM and PI by increasing TN but reducing LL at the same time; or (ii) linkage; a gene responsible for an increase in LL is tightly linked to the gene controlling DM, PI, and TN. Marker pps0113 would be of interest for simultaneous improvement of TN, LL, PI, and DM.

Some QTL identified in this study were in common with published studies in the same mapping family. For example, the QTL for LL on LG2 identified in this study and Sartie et al. (2011) was flanked by the same marker (pps0113). The QTL for DM identified in the two studies had support interval overlaps on LG1 and were in close proximity on LG2. These QTL co-locations or proximity may indicate QTL stability across environments.

Comparative genomic analysis indicated clear positional correspondence between QTL controlling Rubisco concentration detected in the ryegrass population and those reported in rice (Ishimaru et al., 2001). This indicates conservation of key genetic loci controlling Rubisco concentration across grass genera. Markers or candidate genes from the rice genome at these positions could be explored for breeding for improved productivity outcomes in perennial ryegrass and other grass species.

Agronomic application

From the animal nutrition point of view, the association of low leaf Rubisco concentration with increased plant productivity may be beneficial to perennial ryegrass-based livestock systems. Ruminants have low protein utilization efficiency and ingestion of high protein pasture results in increased urinary N excretion (Castillo et al., 2000). N from urine patches is of environmental concern because it can be released into the atmosphere as ammonia and nitrous oxide or leach as nitrate into the soil and ground water (Tamminga, 1992). This low D- productivity association may be useful in tailoring forage crop breeding to meet animal nutritional needs.

Future work

Genetic analysis of leaf Rubisco turnover and plant morphological characteristics was based only on one cross. It is therefore important to confirm the observed inter-trait and trait-marker associations in other perennial ryegrass crosses. Further evaluations of Rubisco turnover and plant morphology should be conducted under contrasting N nutrition and total leaf N measurements taken because the proportion of N accumulation in leaves and in Rubisco as well as plant growth and productivity are strongly affected by N nutrition.

Conclusions

This study confirmed the existence of genetic variation in the time course of leaf Rubisco turnover during the life span of leaves in a perennial ryegrass population. While it was a starting hypothesis that such variation, if found, may affect N use efficiency and hence plant yield, unexpectedly it was found that plants with lower leaf Rubisco concentration tended to have higher yield. The negative correlation of leaf Rubisco concentration with herbage yield indicates that perennial ryegrass growth could be limited more by the effect of the energy cost of Rubisco synthesis rather than photosynthetic capacity. Although there is need for confirmation of this finding using a larger range of genotypes, the region of QTL clustering on LG5 appears to be valuable for simultaneous improvement of yield and nutritive value of perennial ryegrass. Comparative genomic analysis indicated conservation of the genetic loci controlling Rubisco concentration between rice and perennial ryegrass, providing an opportunity for cross-species genetic information transfer that can enhance implementation of marker-assisted selection in perennial ryegrass.

Acknowledgements

This project was funded by the T. R. Ellett Agricultural Research Trust. The authors express their thanks to Mike Hickey, Mark Osborne, Peter Jessop, James Slater, Julia Collins, Steven Ray, Lindsay Silva, Lesley Taylor, and Scott Avery for their technical contribution. They are grateful to ViaLactia Biosciences (Auckland, NZ) for the provision of the GeneThresher SSR markers used on the genetic linkage map. LJI was the holder of a Massey University post-doctoral Scholarship.