JXB Advance Access originally published online on July 30, 2004
Journal of Experimental Botany 2004 55(407):2461-2472; doi:10.1093/jxb/erh200
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
RESEARCH PAPER |
Dealing with the genotypexenvironment interaction via a modelling approach: a comparison of QTLs of maize leaf length or width with QTLs of model parameters
INRA – ENSAM1, Laboratoire d'Ecophysiologie des Plantes sous Stress Environnementaux, 2, Place Pierre Viala, F-34060 Montpellier cedex 1, France
To whom correspondence should be addressed. Fax: +33 467 522116. E-mail: francois.tardieu{at}ensam.inra.fr
Received 8 December 2003; Accepted 14 May 2004
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Quantitative genetics of adaptive traits is made difficult by the genotypexenvironment interaction. A classical assumption is that QTLs identified in both stressed and control conditions correspond to constitutive traits whereas those identified only in stressed treatments are stress-specific and correspond to adaptive traits. This hypothesis was tested by comparing, in the same set of experiments, two ways of analysing the genetic variability of the responses of maize leaf growth to water deficit. One QTL detection was based on raw phenotypic traits (length and width of leaf 6) of 100 recombinant inbred lines (RILs) in four experiments with either well-watered or stressing conditions in the field or in the greenhouse. Another detection followed a method proposed recently which consists of analysing intrinsic responses of the same RILs to environmental conditions, determined jointly over several experiments. QTLs of three responses were considered: (i) leaf elongation rate per unit thermal time in the absence of stress, (ii) its response to evaporative demand in well-watered plants, and (iii) its response to soil water status in the absence of evaporative demand. The QTL of leaf length differed between experiments, but colocalized in seven cases out of 13 with QTLs of the intrinsic leaf elongation rate, even in experiments with stressing conditions. No colocalization was found between QTLs of leaf length under water deficit and QTLs of responses to air or soil water status. By contrast, QTLs of leaf width colocalized in all experiments, regardless of environmental conditions. The classical method of identifying the QTL of constitutive versus adaptive traits therefore did not apply to the experiments presented here. It is suggested that identification of the QTL of parameters of response curves provides a promising alternative for dealing with the genetic variability of adaptive traits.
Key words: Genotypexenvironment interaction, leaf growth, QTL, water deficit
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
The concepts of adaptive versus constitutive characters differ according to the scientific communities which use them. When applied to gene expression or protein amounts, a constitutive character is supposed to be independent of environmental conditions, while the opposite applies to an adaptive character (Seki et al., 2002
). When applied to phenotypic traits, the definition usually shifts towards the existence or non-existence of a genotypexenvironment (GxE) interaction on the considered trait with a positive effect on yield (Blum, 1996). An adaptive trait is then defined as an alteration in plant structure or function which improves the behaviour under stress of the considered genotype (e.g. reduction in transpiration rate, allowing the plants to conserve water through to the end of the crop cycle). Conversely, a constitutive trait is either unaffected by environmental conditions, or is affected by similar amounts in all studied genotypes (no GxE interaction). Although it does not respond to water stress, a constitutive trait can bring a comparative advantage under water deficit (e.g. transpiration efficiency under well-watered conditions, deep root system, or early vigour; Richards et al., 2002). Breeding for constitutive traits has yielded several ‘success stories’. QTLs of deep rooting colocalize with QTLs of yield under water deficit (Tuberosa et al., 2002); improving water-use efficiency of well-watered plants increases wheat yield under severe water deficit (Condon et al., 2002). By contrast, breeders are often reluctant to consider adaptive traits associated with a large ‘built-in’ GxE interaction which lowers heritability.
A classical way to identify QTLs of adaptive traits is to set up a network of experiments with contrasting environmental conditions and identify QTLs in all experiments separately. It is then assumed that QTLs identified in both stressed and control conditions will correspond to constitutive traits, whereas those identified in stressed treatments only will be stress-specific and will correspond to adaptive traits (Prioul et al., 1997; Ribaut et al., 1997; Fracheboud et al., 2002). This is based on the hypothesis that no QTLs are detected in stressed treatments, but are missed in control treatments for statistical reasons. This hypothesis cannot be considered as likely, thereby leading to an overestimation of the frequency of adaptive QTLs.
An alternative method has recently been proposed (Reymond et al., 2003), based on the fact that although an adaptive trait changes with environmental conditions, it frequently follows a reproducible behaviour. For instance, leaf elongation rate changes with meristem temperature, but follows a tight relationship with it provided that plants experience no water or nutrient deficits, and no evaporative demand. Under these circumstances, this relationship applies to different experimental conditions (Ben Haj Salah and Tardieu, 1995, for maize; Granier et al., 2002, for Arabidopsis thaliana). In the same way, the responses of maize leaf elongation rate to evaporative demand and to soil water status are stable characteristics of a genotype, which apply to field as well as controlled conditions (Tardieu et al., 2000). An adaptive trait, with a GxE interaction, can therefore be linked to stable underlying characteristics of genotypes, independent of experimental conditions. The authors of this study have proposed that a genetic analysis could be carried out on these stable characteristics that describes the responses of genotypes to environmental conditions (Reymond et al., 2003).
The purpose of the present paper was to compare the two above-mentioned methods on the same data set, in the case of leaf growth. First, QTLs of final leaf length and width were determined individually in four experiments, three in the greenhouse with contrasting soil water status and one in the field with well-watered plants subjected to a high evaporative demand. Second, QTLs of the response of leaf elongation rate to environmental conditions were determined in the same experiments, taking advantage of short-term variations of elongation rate within and between experiments. The first method therefore considered adaptive traits independently in several experiments, while the second considered that the variation of these traits with environmental conditions were intrinsic characteristics of each genotype. In particular, the aim was (i) to test whether QTLs of parameters of response curves colocalized with QTLs of raw phenotypic traits, and (ii) to test the validity of the hypothesis that QTLs of adaptive traits are specific to experiments in stressed conditions, while QTLs detected in both control and stress conditions would correspond to constitutive traits.
|
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Genetic material
The mapping population used in this study consisted of recombinant inbred lines (RILs) with six generations of self pollination, derived from a cross between two parental lines, F-2 (an early French flint) and Io (a late North American semi-dent considered as drought-tolerant). A total of 145 RILs was produced and 152 RFLP probes were used for mapping these RILs (Causse et al., 1996
). QTLs were identified in a subset of 100 randomly chosen RILs.
Experiments
Field experiment:
One experiment was carried out near Montpellier, southern France (FC2, Table 1). Seeds were sown in pairs on 5 July 1999 under a mobile shelter that allowed air temperature and air vapour pressure deficit to be modified. Plants were thinned when leaf 3 emerged, leaving five plants per RIL. RILs were spatially organized in sequences of 10 plants of different RILs, each sequence being distributed at random in the field. The soil was watered twice a week, with amounts greater than the Penman evapotranspiration during the same periods so plants experienced no water deficit.
|
Air temperature and relative humidity were measured every 20 s (HMP35A Vaisala Oy, Helsinki, Finland). The temperature of the meristematic zone was measured with a fine copper-constantan thermocouple (0.4 mm diameter) located inside the stem in the meristematic zone. Light was measured continuously using a PPFD sensor (LI-190SB, Li-Cor, Lincoln, NB, USA). All temperatures referred to hereafter are meristem temperatures. All data of temperature, PPFD, and relative humidity were averaged and stored every 600 s in a data logger (Campbell Scientific, LTD- CR10 Wiring Panel, Shepshed, Leicestershire, England). Vapour pressure difference between meristem and air (VPD, kPa) was calculated as the difference between saturation vapour pressures at meristem temperature and at air dew point temperature. An equivalent VPD was calculated for 1 d periods by cumulating measured VPDs after correction for changes in stomatal conductance due to diurnal variations of PPFD (Reymond et al., 2003
for detail). Briefly, the mean VPDs at each time step of 600 s were multiplied by a coefficient (ki) which was 0 and 1, respectively, at PPFDs of 0 and 500 mmol m–2 s–1, proportional to PPFD between these two values, and equal to 1 for PPFDs above 500 mmol m–2 s–1.
 | (1) |
 | (2) |
The vertical position of the tip of the sixth leaves was measured twice a day on four plants per RIL, in the morning (05.00–07.00 h solar time) and in the evening (14.00–16.00 h) during the period elapsed between appearances of leaves 6 and 8, during which the elongation rate of leaf 6 is constant with time under constant temperature. The position of the leaf tip was measured using a ruler attached to a 2.5 m horizontal bar, itself fixed on vertical metal sticks permanently left in the soil. Leaf elongation rate was calculated individually for each of the four replicates of each RIL, during days and nights. Each elongation rate, corresponding to one plant for one day or one night, was considered as an individual data point, and was analysed as a function of measured environmental conditions measured on the same plant during the same period. The final length and width of the sixth leaf were measured on four plants per RIL, using a ruler after the ligule had appeared.
Greenhouse experiment, well-watered plants:
One experiment was carried out in the greenhouse in well-watered conditions (Experiment GC2, Table 1). On 10 June 2000, seeds were placed at 0.025 m depth in columns (0.15 m diameter, 0.4 m height) containing a 40:60 mixture (v/v) of a loamy soil and an organic compost. Six seeds per RIL were sown in pairs and thinned to three plants when leaf 3 emerged. Each column contained three different RILs and was located at random in three blocks. The soil was maintained at retention capacity, by daily watering with a modified one-tenth strength Hoagland solution corrected with minor nutrients. Columns were individually weighed every third day in order to check that the soil water content was 35–40% of dry soil, corresponding to a predawn leaf water potential higher than –0.1 MPa.
Meristem temperature was measured and VPDeq was estimated as in the field experiment. Ranges of each environmental variable were obtained by manipulation of climate inside the greenhouse. On two nights, plants were covered by a 4.0x4.7 m plastic shelter and air temperature was lowered by two air conditioners, allowing meristem temperature to be decreased to 18.7 °C. The same shelter was used on the other two nights, but the air was heated so that meristem temperature reached 26.1 and 27.4 °C, respectively. Finally, all plants were moved on two non-consecutive nights into a growth chamber in order to measure leaf elongation rate at low meristem temperature (15.1 and 14.4 °C on the two nights). Daytime VPD was varied, either by turning off the water of the air-cooling system (2–3.5 kPa) or by leaving the water circulating (0.6–1.5 kPa). Low VPDs were obtained by spraying water on the soil (0.2–1 kPa). An example of the time-course of environmental conditions is presented in Fig. 1.
|
The position of the tip of leaf 6 was measured at least twice a day (at 05.00–07.00 h and 16.00–18.00 h solar time) on three plants per RIL, during the period that elapsed between the appearances of leaves 6 and 8, with a ruler fixed on the top of the columns. Leaf elongation rate was calculated individually for each plant during days and nights and processed as in the field experiment. The final length and width of the sixth leaf were measured on three plants per RIL, as in the field experiment.
Greenhouse experiments, water-deficient plants:
Two experiments were carried out with soil water deficit (Experiments GS1 and GS2, Table 1). Plants were sown with the same procedure as in Exp GC2, on 10 March and 6 October 2000. While filling columns, a soil sample was taken from every second column to determine the initial soil water content. It was confirmed that soil water content was similar in all columns and homogeneous within each column (not shown). Soil water content was determined afterwards by weighing columns every day. Differences in weight were attributed to changes in soil water content, as plant weight was negligible. A water-release curve was obtained by coupling mean soil water content of the column to predawn leaf water potential in the range of water potentials from –0.5 to –1.5 MPa. This procedure allowed the predawn leaf water potential of every plant in the experiments to be estimated daily. Irrigation was stopped when leaf 5 appeared. A soil water potential of –0.3 MPa was reached in 3–6 d, depending on the leaf area of the RIL under consideration (Fig. 2). Soil water status was then controlled in such a way that each plant experienced a variety of predawn leaf water potentials ranging from –0.03 to –0.25 MPa in Experiment GS1. This range was larger in Experiment GS2, from –0.10 to –0.6 MPa. Columns were partially rewatered at the end of the growing period of leaf 6 and a second sequence of dehydration was followed. Light, meristem or air temperatures, and meristem-to-air VPD were measured, as in Experiment GC2. Leaf elongation rate final length and width of the sixth leaf were measured in three plants per RIL, as in Experiment GC2.
|
Genetic analysis
QTL detections were performed with two methods, either as presented in Reymond et al. (2003)
by using the Statistical Analysis System software (SAS Institute Inc., Cary, NC, USA), or by using the PLABQTL software (Utz and Melchinger, 2000). In the first method, QTLs were detected by composite interval mapping with investigation of epistatic interactions. Cofactors were first chosen using a stepwise regression between the studied trait and the allele value at each marker. A backward regression was then carried out in each chromosome to limit the number of cofactors. The presence of main-effect QTLs was tested every 5 cM between the 152 markers (445 positions on the genome) using a multiple regression with the cofactors. For that, the allele value was determined every 5 cM as the probability of occurrence of allele F-2 at this position according to the allelic value at flanking markers. When the tested position on the genome was close to a cofactor (±10 cM), the effect of this cofactor was removed. Because the theoretical distribution of the test statistics (the F of the linear regression) was unknown for multiple regression with cofactors, the threshold value was determined by 1000 permutations (Churchill and Doerge, 1994). In the last step, a regression was carried out between the studied trait and all combinations of two positions on the genome, taking into account the main-effect QTLs previously determined. This allowed determination of epistatic QTLs. In order to strengthen the genetic analysis, and to test low-effect, epistatic QTLs, a second detection was carried out with an independent method. This method (PLABQTL) was based essentially on the same principles, with a composite interval mapping with cofactors either chosen by the software, or imposed externally. In order to allow better comparison between methods, the same cofactors as in the first method were used. The main difference with the first method is that epistatic interaction between QTLs was determined by PLABQTL only with QTLs that had first been detected as main-effect QTLs.
The two methods of QTL detection were first applied to leaf length or width in individual experiments, by taking into account the mean value of the three or four plants of one RIL. They were also applied to the parameters of the relationships between individual elongation rates corresponding to one plant for one day or night, and environmental conditions measured on the same plant during the same period (see Results). In both methods, the partial r2 of each was estimated after accounting for the effects of all other QTLs found for the same trait. Broad-sense heritability was estimated according to Gallais (1990):
 | (3) |
is the genetic variance and 
the environmental variance. It was estimated on the three plants of one RIL. Analysis of variance and statistical analyses were carried out by using the SAS and R softwares (R Development Core Team, 2003). |
Results |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Variability between experiment of final leaf length and width, GxE interaction
The length of leaf 6 had a large variability in the whole set of data, ranging from 17 to 80 cm (Figs 3, 4). An analysis of variance showed significant effects of experiments, genotypes, and the interaction between them. Maximum lengths were observed in Experiment GC2, in which plants were well-watered in the greenhouse (Figs 3, 4b). The two levels of water deficit decreased leaf length, by 10 and 29 cm on average in Experiments GS1 and GS2 (Figs 3, 4c). The field experiment GC2 with well-watered plants but high VPD showed a leaf length intermediate between Experiments GS1 and GS2, on average 25 cm shorter than in Experiment GC2 (Figs 3, 4b). Differences were therefore larger between the two experiments with well-watered plants in the field and in the growth chamber than between growth-chamber experiments with two levels of water deficit. Consistent with a significant experimentxgenotype interaction, the ranking of RILs according to leaf length differed between experiments, with increasing differences when the difference in environmental conditions increased (Fig. 5). It was well conserved between greenhouse experiments with no or low water deficit (GC2 and GS1), or between experiments with well-watered plants in field or greenhouse (GC2 and FC2, r2=0.53, not shown). It was looser between the greenhouse experiments with or without severe water deficit (GC2 and GS2, GS1 and GS2) and independent between the field experiment with high VPD and the greenhouse experiments with soil water deficit (FC2 and GC2).
|
|
|
The three individual values of leaf length corresponding to one RIL in one experiment were closely related to their mean value (Fig. 4). This resulted in high heritabilities in individual experiments with no or moderate water deficits (h2=0.75, 0.78, and 0.73 in experiments FC2, GC2, GS1, respectively), suggesting that reproducibility was high in these experiments. In the same way, the correlations between mean values of lengths measured on leaves 6 and 7 were high in experiments GS2 and FC2 (r2=0.84 and 0.93). Heritability was lower in experiment GS2, with more severe water deficit (h2=0.39, Fig. 4c), and the correlation between lengths of leaves 6 and 7 was also lower than in other experiments (r2=0.69). This difference was probably due to the fact that, during a sequence of water deficit, each plant replicate depleted soil water at a different rate, depending on its leaf area and that of its neighbours in the same column. Plants were therefore subjected to different soil water status. This can be visualized in Fig. 2, in which differences in predawn leaf water potential larger than 0.1 MPa were observed between individual plants of RILs at a given time.
Leaf width was less affected by the tested environmental conditions than leaf length. It only differed significantly in Experiment GS2 compared with the other three treatments (Figs 3, 4e, f). The effects of genotype, experiments, and genotypexexperiment interaction were significant, although the latter had a relatively low contribution to the total variance compared to the case of leaf length. The correlations between leaf widths measured in Experiments GS2, GS1, and FC2 were consistently higher than those corresponding to leaf length, especially in the comparison between Experiments GS2 and FC2 (Fig. 5). Heritability was high in all individual experiments (Fig. 4e, f), confirmed by the correlation between the widths of leaves 6 and 7 (r2=0.87, 0.72, and 0.58 in Experiments GC2, GS2, and FC2, respectively).
Differences in the mean values of leaf length and rankings of RILs between experiments can be predicted from differences in environmental conditions
In the examples presented in Fig. 6a, the differences in final length observed between greenhouse and field experiments with well-watered plants ranged from 12 to 31 cm, depending on the RIL. Consistent with earlier studies (Ben Haj Salah and Tardieu, 1997; Reymond et al., 2003), differences in length may be caused by differences in evaporative demand during leaf elongation. VPDs, averaged over the whole period of leaf growth, were 0.6 and 1.6 kPa in the greenhouse and field experiments, respectively. This corresponded to VPDs higher than 3 kPa in most afternoons of the field experiment. According to the model of Reymond et al. (2003), applied to the three presented RILs with their respective measured sensitivities to VPD, this difference in evaporative demand would result in differences in leaf length of 11–24 cm. Thus, the effect of evaporative demand accounted for a large part of the difference between greenhouse and field experiments.
|
Differences between RILs in the responses of final leaf length to evaporative demand were therefore of the same order of magnitude as differences in final length between RILs in a given experiment. This changed the ranking of RILs leaf length between experiments and explained the significant genotypexexperiment interaction. The same change in ranking of genotypes was observed on shorter timescales. When plants were subjected to changing evaporative demand (Fig. 1), all RILs had a leaf elongation rate which decreased during the day and increased during the afternoon, following changes in VPD. The ranking of RILs was similar during the two consecutive nights, but was not conserved during the day period with highest evaporative demand (11.00–16.15 h), due to different sensitivities to evaporative demand.
A similar analysis was carried out for the effect of soil water status on final leaf length (Fig. 6b). Leaf lengths were compared in Experiments GC2, GS1, and GS2, in which mean predawn leaf water potentials averaged during the period of leaf growth were –0.03, –0.11, and –0.23 MPa, respectively, with some differences between RILs, and mean VDP was low throughout the experiments (Table 1). As mentioned above, predawn leaf water potential had an appreciable variability between plants of a given RIL because plant transpiration scenarios differed between plants (Fig. 2). In spite of this variability, it can be seen in Fig. 6b that the studied RILs had appreciable differences in sensitivity of leaf length to soil water deficit, with effects ranging from 16.7–40.8 cm, i.e. similar to the differences between RILs in a given experiment.
Taken together, these results suggest (i) that individual experiments were reproducible in terms of final leaf length, although with a lower heritability in the experiment with high water deficit due to differences in water status between plants of an RIL at a given time; (ii) that the ranking of RILs on leaf length differed between experiments, both in well-watered and water-deficit conditions; (iii) that differences between experiments in rankings of RILs were at least in part due to contrasting sensitivities to evaporative demand and to soil water status.
QTLs of leaf length differed in the four experiments while QTLs of width were common
Fourteen QTLs of length were detected with the first method of QTL detection (6, 1, 3, and 4 in Experiments GC2, FC2, GS1, and GS2, respectively, Table 2). Two were detected as main effects, and 12 as interactions between two loci (‘epistasis’). The second method of QTL detection confirmed the two main-effect QTLs (LOD scores >3). Among the other 12 QTLs, it detected 4 QTLs as main-effect (LOD scores >3), 4 as tendencies (LOD scores of 2–3), and 4 were not confirmed. In the following, only QTLs detected with both methods were kept. Three QTLs are also displayed because they had LOD scores higher than 3.5 with the second method.
|
|
This double analysis provided 13 QTL positions (Fig. 7), with one cluster on chromosome 2 corresponding to three experiments with no or low soil water deficit (Experiments GS1, GC2, and FC2) and two clusters corresponding to experiments with low or severe soil water deficit (Experiments GS1 and GS2), one on chromosome 6 and one on chromosome 3. Clusters of QTLs therefore occurred for groups of experiments in which the ranking of QTLs was relatively well conserved (Fig. 5). The other six QTL positions corresponded to individual experiments. QTLs were also detected on the differences in length observed between two experiments (GC2–GS2 and GC2–FC2). Analyses of both raw and normalized difference (divided by the length in Experiment GC2) provided no new information, with three QTLs already detected on chromosomes 2, 3, and 8.
|
Leaf width had a common QTL in the four experiments, located on chromosome 5 between 125 and 135 cM (Table 2; Fig. 7). Other QTLs were detected by the first method as interactions between loci, but were not confirmed by the second method. This common position of one QTL for the four experiments is consistent with the relatively low GxE interaction observed in Fig. 5.
An alternative method: QTLs of parameters of responses to soil water deficit and evaporative demand
The rationale for this method (Reymond et al., 2003) is the desire to take advantage of all sources of variation of both soil water deficit and VPD, between experiments, between consecutive days of an experiment, and between plants of a RIL. The method consists in building response curves of leaf elongation rate to environmental conditions at a timescale of several hours. Because maize leaf elongation rate is stable for 5–7 d after leaf appearance, it is possible to consider changes in leaf elongation rate with environmental conditions as reproducible events, on which response curves are built (Tardieu, 2003). Fluctuations of leaf elongation rate were synthesized by responses to three environmental conditions, namely meristem temperature, leaf-to-air VPD, and predawn leaf water potential.
The response of leaf elongation rate to meristem temperature was considered during night periods in experiments without soil water deficit, when evaporative demand was null and leaf elongation rate only depended on meristem temperature. This resulted in linear relationships which were common, for each RIL, to several experiments and allowed the calculation of leaf elongation rate per unit thermal time (Fig. 8a, d; see Tardieu et al., 2000 for detail). Expressed in this way, leaf elongation rate becomes independent of temperature in the studied range, and is common to several experiments. Elongation rate per unit thermal time, termed parameter a hereafter, therefore defines the intrinsic capacity of the considered RIL to elongate in the absence of stress (expressed in cm °Cd–1). It was calculated here with data from Experiment GC2, and from data collected at the beginning of Experiments GS1 and GS2 while predawn leaf water potential was still higher than –0.08 MPa. Data originating from another greenhouse experiment fitted in the same regression.
The response to evaporative demand was considered during day periods without soil water deficit, while meristem temperature, light intensity, and evaporative demand fluctuated with time (Fig. 1). It was shown earlier that light intensity has no direct effect on leaf elongation rate at this timescale, and essentially acts via its effect on leaf-to-air VPD (Ben Haj Salah and Tardieu, 1996). The sensitivity to evaporative demand was estimated via the slope of the response curve of leaf elongation rate, expressed per unit thermal time, to meristem-to-air VPD. Common linear relationships applied to the four experiments in the field and the greenhouse. For each RIL, the sensitivity to evaporative demand could therefore be considered as common to the four considered experiments, as shown for two RILs in Fig. 8b, e. It was calculated by a common regression over the whole data set, and termed parameter b hereafter (cm °Cd–1 kPa–1).
The response to soil water deficit was analysed during night periods. The slope of the response curve of leaf elongation rate to predawn leaf water potential was calculated jointly in Experiments GC2, GS1, and GS2. A common linear relationship applied to the experiments carried out with or without soil water deficit, as shown for two RILs in Fig. 8c, f. The slope of this relationship defines the sensitivity of the considered RIL to soil water deficit over the whole set of data, and is termed parameter c hereafter (cm °C d–1 MPa–1).
It was shown earlier (Tardieu et al., 2000; Reymond et al., 2003) that the three responses can be combined in a model with three parameters :
 | (4) |
QTLs of parameter a of the model colocalize with QTLs of final leaf length
The ability of a RIL to elongate in optimum conditions (parameter a) was accounted for by nine significant QTLs in the first method of QTL detection (Table 2, Fig. 7). Three were detected as main-effect QTLs and confirmed by the second method of detection (two on chromosome 2 and one on chromosome 4). Six QTLs were detected by the first method as interactions between two loci, among which one was confirmed as a main-effect QTL by the second method (chromosome 4) and two as tendencies. The other three QTLs in interaction could not be confirmed by the second analysis. As above, only QTLs detected by both methods were kept in the analysis.
Among the six resulting QTLs of parameter a, five were located at less than 15 cM from a QTL of leaf length (Fig. 7). Reciprocally, among the 13 QTLs of leaf length, seven were located at less than 15 cM from a QTL of parameter a, two other QTLs of length on chromosome 2 were located at a greater distance, and four had no colocalization with a QTL of parameter a. In all cases, the allele for longer leaf was also the allele for faster elongation rate.
Sensitivity to evaporative demand (parameter b) had six significant QTLs in the first method (Table 2; Fig. 7), among which two were main effect QTLs, also detected by the second method. The four QTLs detected as interactions between two loci by the first method were not confirmed. The two confirmed QTLs located on chromosomes 1 and 8 colocalized with significant QTLs of the x-intercept of the response of leaf elongation rate to VPD, termed b0, which represents the VPD that would stop elongation. One QTL of parameter c was detected as a main effect in the first method, and was confirmed by the second method. Another QTL was confirmed, of the x-intercept of the response of leaf elongation rate to predawn leaf water potential, termed c0, which represents the water potential that would stop leaf elongation. Although some QTLs detected by the first method for parameters b and c colocalized with QTL of leaf length (especially on chromosome 5), this was not the case for any of the QTLs detected by both methods. Further, no colocalization was detected between differences in leaf length between experiments and QTLs of parameters b and c.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Adaptive versus constitutive traits
The concepts of adaptive versus constitutive traits can be discussed from the data presented here, and somewhat change the definition by Blum (1996)
given in the Introduction. Leaf length had the characteristics of an adaptive trait, with a GxE interaction and an instability of QTLs which both increased with differences in environmental conditions between experiments (Figs 5, 6, 7). However, this instability was determined by responses of genotypes which were stable across environments (Fig. 8). The responses of leaf elongation rate to evaporative demand (in the absence of soil water deficit) and to soil water status (in the absence of evaporative demand) were common to several experiments for each genotype. They were therefore stable characteristics of each genotype, without GxE interaction nor changes in ranking of genotypes between experiments. Differences in rankings of RILs, which resulted in an instability of QTLs of leaf length, were linked to intrinsic differences in the sensitivities of RILs to evaporative demand or soil water deficit.
The case of parameter a is particular. It can be viewed as a response to an environmental condition, namely meristem temperature, or as a constitutive trait, the ability of the leaves of a given genotype to elongate in the absence of stress. This second view is based on the fact that elongation rate, expressed in thermal time, is independent of temperature. It is also based on the fact that the change with temperature of the rates of processes such as tissue elongation or cell division are matched by those of the duration of the same processes, thereby having a negligible effect on the final size or number of cells (Ben Haj Salah and Tardieu, 1995; Tardieu et al., 2000; Lafarge and Tardieu, 2002). Consistently, QTLs of parameter a colocalized with QTLs of leaf lengths in several experiments. Leaf width also had several characteristics of a constitutive trait although it was affected by water deficit. In effect, it had a relatively low GxE interaction and stable QTLs across four experiments with contrasting environmental conditions.
The fact that a trait with high GxE interaction can be predicted from intrinsic characteristics of genotypes may open the way to a modelling of the interaction (Tardieu, 2003). Recently, the authors have combined the QTL models of the three parameters a, b, and c with an ecophysiological model which predicts the leaf elongation rate of plants subjected to any climatic scenario. This allowed prediction of the leaf elongation rate of new RILs of the same mapping population, which were only known by their alleles at QTLs, under new climatic scenarios in a growth chamber (Reymond et al., 2003). RILs differed by their intrinsic elongation rate and by their sensitivity to evaporative demand, thereby having changed rankings under different environmental conditions. Predicted and observed leaf elongation rates were in good agreement, so the GxE interaction could be predicted over short timescales. The test remains to be done over a longer period, in particular, taking into account the duration of elongation.
Comparison of QTLs of final leaf size and of parameters of the model of elongation rate
The detection of QTLs with two different methods, both based on multiple regression, provided converging results. The main difference between the two methods is that the first allowed calculation of QTLs of interaction between loci (‘epistasy’) even if these QTLs did not appear as main effect. This method was also slightly more stringent than the second one (PLABQTL), which only calculated main-effect QTLs. Overall, the decision to keep only the QTLs common to both methods in the analysis was adapted to the objective of this paper, a comparison of QTLs. It results in a fewer number of QTLs than those published earlier (Reymond et al., 2003), in which the main objective was to get the best genetic model, which was then checked against independent data.
The colocalization of QTLs of final leaf length with those characterizing the model of leaf growth is reassuring in terms of the ability of both methods to analyse the genetic variability of leaf growth. However, the study presented here casts doubt on the method that classifies QTLs of raw phenotypic traits on the basis of environmental conditions during experiments. More than half of the QTLs of final leaf length, determined in different experiments, colocalized with parameter a, the intrinsic leaf elongation rate, even in experiments with atmospheric or soil water deficits. By contrast, parameters of sensitivity to evaporative demand or to soil water deficit had no clear colocalization with final leaf length in experiments with either high evaporative demand (FC2) or water deficit (GS2). A first interpretation of this result could be that the study failed to detect QTLs of parameters of response curves in loci where QTLs of leaf length of stressed plants were identified. This is possible, due to the relatively low number of RILs taken into account in this analysis, and to the stringent analysis with two different methods which reduced the number of QTLs. However, it is still the case that the clusters of QTLs of responses which were determined on chromosomes 1 and 8 did not correspond to QTLs of leaf length in stressed experiments, suggesting that detection of QTLs of raw phenotypic traits (or even of their differences between experiments) may not be an appropriate method for identification of the zones of the genome involved in responses to environmental conditions.
The presence of a strong QTL of leaf width, stable between experiments and unrelated to QTLs of leaf length or of parameters of the model, is an unexpected result of this study. It suggests that leaf width and length have different genetic determinants. Responses to environmental conditions also differed between these two variables, with a low effect of a moderate soil water deficit and a low difference between greenhouse and field experiments. This confirms an earlier study which suggested a difference in response to incident light (Muller et al., 2001): leaf elongation rate and final leaf length were unaffected by a change in incident light, while leaf width had a clear response to it.
Although identification of QTLs of responses to environmental conditions is more time-consuming than analysis of raw phenotypic traits such as leaf length, leaf area, or yield, it has several advantages. (i) The main advantage is probably that this method brings results which characterize a genotype per se, and not the behaviour of a genotype in a given environment. It is therefore possible to compare results of experiments carried out at different times and in different experimental conditions (field, controlled conditions). (ii) The analysis presented here suggests that this method might be more reliable than others for identifying QTLs of mechanisms of adaptation. (iii) The authors have shown earlier that it allows the behaviour of virtual RILs in a series of environmental conditions (Reymond et al., 2003) to be predicted, including conditions which do not exist yet but can be forecast in climate changes scenarios. This could help selecting in silico the combinations of alleles of interest for such scenarios.
|
Acknowledgements |
|---|
This work was supported by the French programme of genomics Génoplante. The authors thank Christelle Bencivenni for help in data processing and Claude Welcker for critically reading the manuscript. Philippe Naudin and Philippe Hamard contributed to measurements.
|
Footnotes |
|---|
* Present address: CEA – LBDP. Laboratoire de Biologie et du Développement des Plantes. Cadarache, France.
|
References |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Ben Haj Salah H, Tardieu F. 1995. Temperature affects expansion rate of maize leaves without spatial distribution of cell length. Plant Physiology 109, 861–870.[Abstract]
Ben Haj Salah H, Tardieu F. 1996. Quantitative analysis of the combined effects of temperature, evaporative demand and light on leaf elongation rate in well-watered field and laboratory-grown maize plants. Journal of Experimental Botany 47, 1689–1698.[ISI]
Ben Haj Salah H, Tardieu F. 1997. Control of leaf expansion rate of droughted maize plants under fluctuating evaporative demand. Plant Physiology 114, 893–900.[Abstract]
Blum A. 1996. Crop responses to drought and the interpretation of adaptation. Plant Growth Regulation 20, 135–148.
Causse M, Santoni S, Damerval C, Maurice A, Charcosset A, Deatrick J, de Vienne D. 1996. A composite map of expressed sequences in maize. Genome 39, 418–432.
Churchill CA, Doerge RW. 1994. Empirical threshold values for quantitative trait mapping. Technical report of Biometrical Unit. Ithaca, NY: Cornell University.
Condon AG, Richards RA, Rebetzke GJ, Farquhar GD. 2002. Improving intrinsic water-use efficiency and crop yield. Crop Science 42, 122–131.
Fracheboud Y, Ribaut JM, Vargas M, Messmer R, Stamp P. 2002. Identification of quantitative trait loci for cold tolerance of photosynthesis in maize (Zea mays L). Journal of Experimental Botany 53, 1967–1977.
Gallais A. 1990. Théorie de la sélection en amélioration des plantes. Paris: Masson.
Granier C, Massonnet C, Turc O, Muller B, Chenu K, Tardieu F. 2002. Individual leaf development in Arabidopsis thaliana: a stable thermal-time-base program. Annals of Botany 89, 595–604.
Lafarge T, Tardieu F. 2002. A model coordinating the elongation of all leaves of a sorghum cultivar, applies to Mediterranean and Sahelian conditions. Journal of Experimental Botany 53, 715–725.
Muller B, Reymond M, Tardieu F. 2001. The elongation rate at the base of a maize leaf shows an invariant pattern during both the steady-state elongation and the establishment of the elongation zone Journal of Experimental Botany 52, 1259–1268.
Prioul JL, Quarrie S, Causse M, de Vienne D. 1997. Dissecting complex physiological functions through the use of molecular quantitative genetics. Journal of Experimental Botany 48, 1151–1163.[ISI]
R Development Core Team. 2003. R: a language and environment for statistical computing. Vienna: R Foundation for Statistical Computing. http://www.R-project.org
Reymond M, Muller B, Leonardi A, Charcosset A, Tardieu F. 2003. Combining quantitative trait loci analysis and an ecophysiological model to analyze the genetic variability of the responses of maize leaf growth to temperature and water deficit. Plant Physiology 131, 664–675.
Ribaut JM, Jiang C, Gonzalez-de-Leon D, Edmeades GO, Hoisington DA. 1997. Identification of quantitative trait loci under drought conditions in tropical maize. II. Yield components and marker-assisted selection strategies. Theoretical and Applied Genetics 94, 887–896.[CrossRef]
Richards RA, Rebtzke GJ, Condon AG, van Herwaarden AF. 2002. Breeding opportunities for increasing the efficiency of water use and crop yield in the temperate cereals. Crop Science 42, 111–121.
Seki M, Narusaka M, Ishida J et al. 2002. Monitoring the expression pattern of 7000 Arabidopsis genes under drought, cold and high-salinity stresses using a full length cDNA microarray. The Plant Journal 31, 279–292.[CrossRef][ISI][Medline]
Tardieu F, Reymond M, Hamard P, Granier C, Muller B. 2000. Spatial distributions of expansion rate, cell division rate and cell size in maize leaves: a synthesis of effect of soil water status, evaporative demand and temperature. Journal of Experimental Botany 51, 1505–1514.
Tardieu F. 2003. Virtual plants: modelling as a tool for the genomics of tolerance to water deficit. Trends in Plant Science 8, 9–14.[CrossRef][ISI][Medline]
Tuberosa R, Sanguineti MC, Landi P, Giuliani MM, Salvi S, Conti S. 2002. Identification of QTLs for root characteristics in maize growing in hydroponics and analysis of their overlap with QTLs for grain yield in the field at two water regimes. Plant Molecular Biology 48, 697–712.[CrossRef][ISI][Medline]
Utz HF, Melchinger AE. 2000. PLABQTL, a computer program to map QTLs. University of Hohenheim, Stuttgart: Institute of Plant Breeding, Seed Science and Population Genetics.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  R. Tuberosa, S. Salvi, S. Giuliani, M. C. Sanguineti, M. Bellotti, S. Conti, and P. Landi Genome-wide Approaches to Investigate and Improve Maize Response to Drought Crop Sci., December 18, 2007; 47(Supplement_3): S-120 - S-141. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. Hirel, J. Le Gouis, B. Ney, and A. Gallais The challenge of improving nitrogen use efficiency in crop plants: towards a more central role for genetic variability and quantitative genetics within integrated approaches J. Exp. Bot., July 1, 2007; 58(9): 2369 - 2387. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C Welcker, B Boussuge, C Bencivenni, J-M Ribaut, and F Tardieu Are source and sink strengths genetically linked in maize plants subjected to water deficit? A QTL study of the responses of leaf growth and of Anthesis-Silking Interval to water deficit J. Exp. Bot., January 1, 2007; 58(2): 339 - 349. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. Muller, G. Bourdais, B. Reidy, C. Bencivenni, A. Massonneau, P. Condamine, G. Rolland, G. Conejero, P. Rogowsky, and F. Tardieu Association of Specific Expansins with Growth in Maize Leaves Is Maintained under Environmental, Genetic, and Developmental Sources of Variation Plant Physiology, January 1, 2007; 143(1): 278 - 290. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. HILLIER, D. MAKOWSKI, and B. ANDRIEU Maximum Likelihood Inference and Bootstrap Methods for Plant Organ Growth via Multi-phase Kinetic Models and their Application to Maize Ann. Bot., July 1, 2005; 96(1): 137 - 148. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||