Author for correspondence: O. Merah Tel: +33 4 99 61 23 87 Fax: +33 4 67 04 54 15 Email:firstname.lastname@example.org
• Dry mass per unit of leaf area (LDM) and ash content were evaluated as alternative criteria for carbon isotope discrimination (Δ) in durum wheat (Triticum durum) flag leaves and grains.
• Using correlation analysis the relationships between the three parameters (LDM, Δ, ash content) and productivity were determined over three consecutive years in 37 field-grown durum wheat genotypes under contrasting drought conditions.
• Highly significant differences were found between years and among genotypes for all measured traits. Grain Δ and ash content, and LDM and flag leaf Δ were negatively correlated under nondroughted conditions. Positive correlations were found between grain yield, harvest index and both Δ and ash content of the flag leaf under drought. No significant correlations were found between LDM and both Δ and grain yield.
• Differences in LDM do not predict variations in Δ, whereas ash content of grain and flag leaf (under droughted conditions) might be useful in predicting Δ and grain yield. Ash content might provide an alternative screening method in the improvement of drought tolerance and yield stability in durum wheat.
Durum wheat (Triticum durum) is mainly cultivated in the Mediterranean basin. In these areas wheat production is frequently subjected to inadequate and erratic rainfall, leading to substantial yield reductions (Loss & Siddique, 1994). Therefore improvement of drought tolerance and yield stability is an important aim for breeders in these regions (Monneveux & Belhassen, 1996). Several morphophysiological traits have been proposed as screening criteria for drought tolerance (Turner, 1997). Transpiration efficiency (TE, the ratio of dry matter produced to water transpired) is a useful correlate of plant growth in dry areas. However, its direct measurement is tiresome, expensive and time consuming, especially in field trials, limiting its use in breeding programmes. The emergence of carbon isotope discrimination (Δ), which is easily measured, has changed the problem. During photosynthesis, plants discriminate against the heavy isotope of carbon (13C) leading to a depletion of the plant dry matter in 13C. Carbon isotope discrimination is a measure of the 13C : 12C ratio in plant dry matter compared with the value of the same ratio in the atmosphere (Farquhar & Richards, 1984). In C3 species, Δ was found to be positively correlated with pi : pa (i.e., the ratio of internal leaf CO2 partial pressure to ambient CO2 partial pressure) and negatively associated with TE (Farquhar & Richards, 1984; Ehdaie et al., 1991; Johnson & Bassett, 1991; Acevedo, 1993).
The relation between Δ (and TE) and grain yield is well documented in barley and bread wheat (Condon et al., 1987; Craufurd et al., 1991; Acevedo, 1993; Morgan et al., 1993; Sayre et al., 1995). The relationships between Δ, grain or biomass production and harvest index (the ratio of grain yield to total above ground biomass) are less well known in durum wheat under Mediterranean conditions. In addition, carbon isotope analysis, using isotope mass spectrometry, is very expensive, especially for the screening of large collections of genetic resources. Attempts have, therefore, been made to develop alternative screening methods. Any physiological trait which strongly correlates with Δ could be used for an indirect selection of TE and grain yield. Ash content (ma) and dry mass per unit of leaf area (LDM) have been proposed (Masle et al., 1992; Wright et al., 1993) as alternative selection criteria for TE and yield. Voltas et al. (1998) found a negative correlation between Δ and ma both measured in the grain of barley grown under semiarid conditions. The relationships between ma and grain yield were found to be significantly negative in rainfed conditions by these authors. The associations between LDM and TE are more variable. A positive correlation was found by Wright et al. (1993) in peanut and by Brown & Byrd (1996) in pearl millet. However, no correlation was noted between these traits in cowpea by Ismail & Hall (1993). Araus et al. (1997) found a negative correlation between Δ and LDM in barley, in the absence of drought.
In a recent study we investigated the physiological basis of the relationships between Δ, ma, carbon and silicon contents in a collection of 19 durum wheat varieties (Merah et al., 1999a). The differential responses of genotypes to annual climate variations distinguished Middle East genotypes from those of the West Mediterranean region. The objectives of the present study follow on from the previous work: to confirm the association between Δ and ma on a wider and more diversified set of durum wheat genotypes, comprising intra- and interspecific advanced lines; to examine the relationship between LDM and Δ under contrasting water supplies, and to define the relationships between grain yield, biomass production and harvest index and the three former traits.
Materials and methods
Thirty-seven durum wheat genotypes (Triticum durum Desf.) were used in this study (Table 1). This collection included landraces (19), improved varieties (13) from diverse geographical origins and five CIMMYT/ICARDA advanced lines.
Table 1. List of the 37 durum wheat accessions used in this study
Casablanca No. 7580
Rubiao 9053 VA49 No. 7600
Santa Marta 2442 V76 No. 7615
MRB5 (Om Rabi 5)
Scoflag/T. carthlicum IC 9811//Omguer-4
Site and Crop Management
Trials were carried out under rainfed conditions, at ENSA-INRA in Montpellier (South of France) during three successive seasons (1994/95, 1995/96, 1996/97). The soil is a sandy-loam (organic matter content 2.1%, pH 7.8) with a depth of c. 0.6 m. A randomized complete block design was used for the three trials with 2 replicates per genotype. Seeds were sown in two 1.50 m rows per plot (25 cm spacing row and 3 cm interplant spacing). Sowing were made on 24th, 17th and 8th of November 1994, 1995 and 1996, respectively. Anthesis occurred between the last week of April and the beginning of May and maturity at the end of June in all 3 yr.
The first year (1995) was characterized by a low total rainfall, with a strong water deficit occurring from February until the end of the growing cycle (Table 2). The second year (1996), by contrast, was wetter (Table 2). Decreasing rainfall and increasing evapotranspiration occurred through the season leading to a mild terminal drought (water deficit of 58 mm in May and 105 mm in June). The third year (1997) was intermediate. A water deficit established as soon as in 1995, but the rainfall of the first 3 months of the growth cycle was much greater (549 mm and 186 mm in 1995). In addition, the water balance was positive in June due to later rainfall events (Table 2). The 3 yr can, therefore, be characterized as three environments corresponding to an intensive and early drought (1995), a mild terminal drought (1996) and a moderate intermittent drought (1997). The ratios of total water to pan evapotranspiration during the 3 yr were 0.26, 1.75 and 0.46. More detailed information is reported in Merah et al. (1999b).
Table 2. Monthly averages for mean daily temperature (Temp), Penman evapotranspiration (PET), and rainfall during the three cropping seasons at Montpellier, France
For each genotype, 20 flag leaves were randomly detached at anthesis and immediately oven-dried at 80°C for 48 h. At maturity, a 10 g grain sample was collected. Leaf and grain samples were ground in a fine powder. Carbon isotope composition (δ13C) was determined with the isotope mass spectrometer (Micromass, Villeurbanne, France) and defined as δ13C(‰) = [(R sample/R reference − 1) × 1000], R being the 13C/12C ratio. The standard error of determination was 0.1‰. The discrimination (Δ) was calculated using the following formula (Farquhar et al. 1989): Δ (‰) = [a − (p)/(1 + (p)] × 1000, where (p) is the δ13C of the samples a, the δ13C of atmospheric CO2, −8‰. Carbon discrimination of the grain and the leaf are indicated as ΔG and ΔL.
Ash content (ma, expressed in mg g−1 d. wt) was determined after complete combustion of the powder aliquot (5 g for grain milling and 20 mg for leaf powder) in a muffle furnace at 900°C and weighing the residue. Each sample was divided into a minimum of three replicates, at least.
Dry mass per unit of leaf area (LDM) was assessed at anthesis. Four leaves were sampled per genotype. Their area (LA) was measured using a leaf area meter (LI-3000, Li-Cor, Lincoln, NB, USA). Leaves were then dried in a forced oven at 80°C for at least 48 h, and weighed to estimate the dry mass (DM) per unit of leaf area, according to Araus et al. (1997) as LDM = DM/LA.
Date of heading was recorded when half of the spikes had emerged and earliness was expressed as the number of days from sowing to heading date (days to heading, DH). Above ground biomass (AGB) and grain yield (GY) were recorded at maturity and then harvest index (HI) was calculated as HI = GY/AGB.
All the data were subjected to variance analysis using the GLM procedure of SAS (SAS Institute, 1987, Cary, NC, USA). The mean pairwise comparison were based on the Duncan test. Correlation analysis was performed to determine the relationship between the traits using the SAS CORR procedure.
Highly significant differences were found between years for Δ, ma and LDM (Table 3). Differences in GY, AGB and HI were also noted between years (Table 3). The difference in the mean flag leaf carbon isotopic discrimination (ΔF) between 1996 (the wettest year) and 1995 (the driest year) was 2.3‰ (Table 3). For grain carbon isotope discrimination (ΔG) this difference was 2.7‰. The ash content of flag leaf (maF) and grain (maG) values were nearly 35% and 22% lower in 1995 than in 1996 (Table 3). By contrast, LDM values were higher in 1995 than in 1996 and 1997. The mean grain yield for all genotypes was c. 20% higher in 1997, and nearly 68% higher in 1996 than in 1995 (Table 3).
Table 3. Carbon isotope discrimination of flag leaf (ΔF) and mature grain (ΔG), ash content of mature kernel (maG) and of flag leaf (maF), dry mass per unit of leaf area (LDM), grain yield (GY), above-ground biomass (AGB), harvest index (HI) and days to heading (DH) of 37 genotypes of durum wheat grown under rainfed conditions at Montpellier (South of France) during three successive years with contrasting rainfall regimes
Values presented are mean and standard deviation (SD) values. Mean indicated by a different letter are significantly different (P ≤ 0.05) by the Duncan comparison test. LSD: least significant difference. For each trait F-value and degrees of freedom (df) of the genotype, environment and their interactions (G × Y), as well as the covariate effect of number of days from sowing to heading (DH) are also displayed.
*** significance at 0.01 and 0.001 probability levels, respectively.
A pooled analysis of variance indicated that both year and genotype had highly significant effects on all the measured traits (Table 3) and Genotype–Year interactions (G × Y) were significant for all these measurements. The three different years therefore exerted different effects on the set of genotypes tested.
ΔG and maG were negatively correlated in all years (Fig. 1a–c). ΔF was positively correlated with maF in 1995 and 1997 (Fig. 1d,f), the 2 yr characterized by early drought conditions. Negative correlations were noted between ΔF and maG (r = − 0.336; P < 0.05) in 1995 and between ΔF and LDM (r = − 0.425; P < 0.01) in 1996. ΔG and maF were positively correlated (r = 0.495; P < 0.01) in 1995 only.
ΔG and maG were highly correlated to both grain yield and harvest index in the 3 yr (Table 4). Positive correlations were noted between ΔF and both GY and HI under severe and early drought (1995). Similar relations were recorded between maF and both GY and HI in 1995 and 1997 (Table 4). Above-ground biomass was negatively correlated with maG in 1997 (Table 4). No correlation was found between LDM and the other traits (Table 4).
Table 4. Correlations between above-ground biomass (AGB), grain yield (GY), harvest index (HI), carbon isotope discrimination of flag leaf (ΔF) and of grain (ΔG), ash content of flag leaf (maF) and of grain (maG) and dry mass per unit of leaf area (LDM) of durum wheat.
*** significance at 0.05, 0.01 and 0.001 probability levels, respectively.
There were differences for Δ and for production among years with a decrease of the values of these traits from 1995 (the driest year) to 1996 (the wettest year) (Table 3). This decrease is probably related to the difference between the 3 yr in growing season rainfall (Acevedo, 1993; Merah et al., 1999a,b). A strong early drought period was noted from February to May in both 1995 and 1997 and ΔF and maF were similar in both years. However, terminal drought was more pronounced in 1995 than in 1997 (Table 1), and this may explain the differences between 1995 and 1997 in ΔG, GY and AGB. The large difference in Δ-values observed between grain and flag leaf (> 2‰) reveals that the plants were subjected to drought during grain filling, even in 1996.
Significant G–Y interactions were found for all measured traits. The magnitude of G × Y is low for Δ compared with that for grain yield (Table 2), suggesting that genotypes ranking remained rather constant across years. The relative change in this ranking was probably and largely attributed to variation in water availability during the crop cycle (Craufurd et al., 1991; Acevedo, 1993; Merah et al., 1999b). Our results support the hypothesis that under Mediterranean conditions, genotypes which sustain greater stomatal conductance during grain filling (and so higher ΔG), can maintain carbon assimilation and a greater yield in a wide range of environments, as previously emphasised in barley (Acevedo, 1993; Voltas et al., 1998).
Influence of earliness on measured traits
In this study the number of days from sowing to heading (DH) differed, within each trial, by > 2 wk between extreme genotypes, probably causing differences in measured traits (Acevedo, 1993). ANOVA, using DH as covariate, was performed in order to test if DH could be at the origin of traits variation. No significant DH effect was noted for either Δ or ma. Even when the DH effect was significant on LDM, AGB and GY, differences between genotypes remained highly significant (Table 3). Therefore, it appears that the genotypic variability for Δ, ma, LDM, GY and AGB, as reported for other cereals (Acevedo, 1993; Sayre et al., 1995), was not only attributable to differences in phenology.
The negative correlation between ΔG and maG (Fig. 1a–c) is in agreement with work on barley (Voltas et al., 1998). These results suggest that mineral accumulation in the grain is primarily regulated by processes other than transpiration, and consequently not directly related to TE. In fact, minerals accumulate in grains by phloem transport rather than passively from the transpiration stream (Jones & Handreck, 1965). Thus, mineral accumulation in the grain (maG) primarily depends on translocation from the leaves and on minerals removed from the lower parts of the plant after the onset of senescence (Wardlaw, 1990). Under severe drought, photosynthesis is more affected than translocation (Loss & Siddique, 1994). As a consequence, genotypes unable to maintain high rates of stomatal conductance and photosynthesis during grain filling (and then exhibiting low ΔG) would mainly use retranslocation of photoassimilates from preanthesis reserves, and minerals from senescent tissues, for their grain filling. The concentration of minerals in the grain would be then higher (leading to negative correlation between ΔG and maG) and increases more rapidly than ΔG in the transition from mild to strong drought.
Relationships between Δ and dry mass per unit of leaf area
Large differences in LDM were observed between years in our study. Higher values of LDM were observed in 1995 (the driest year) whereas the lowest ones were noted in 1996 (the wettest year). A LDM increase has also been reported in durum wheat at anthesis as a response to drought (Rascio et al., 1990). The increase of LDM from favourable conditions (1996) to limiting water conditions (1995) probably reflects the adaptation to the droughted conditions typical of the rainfed Mediterranean regions (Araus et al., 1997).
In our study, ΔF and LDM were only negatively correlated (r = −0.425; P < 0.01) under favourable conditions. Negative correlations between LDM and Δ have also been observed in wheat and barley under favourable conditions (López-Castañeda et al., 1995; Araus et al., 1997). In contrast, weak or no correlation between Δ and LDM was found by Pooter & Farquhar 1994 and Ray et al. (1998) under drought . The absence of association between Δ and LDM is fully confirmed in our study, since no significant relation between both traits was found in 1995 and 1997 (early drought). In 1996 (wettest year), the water supply was less limiting than in 1995, stomatal conductance was likely to be high and genotypes with a great assimilation capacity would have a lower ΔF (resulting from a lower pi/pa). Thus, variation in ΔF would be relatively more influenced by photosynthetic capacity differences than in a dry year where stomatal conductance is likely to be the primary limiting factor for Δ (Morgan et al., 1993). No significant correlation was found between ΔG and LDM. The lack of the correlation between both traits is probably due to the different conditions having occurred during grain filling and flag leaf formation and sampling for LDM measurement (Table 2). According to Brown & Byrd (1996) and Ray et al. (1998), because of the genotype ranking for LDM change due to environmental and ontogenetic factors, LDM can not be considered as a good predictive criterion of TE and Δ. Our results support this conclusion as, no significant correlations were found between LDM value measured in the 3 yr (Table 3).
Relations of the studied traits with biomass, grain yield and harvest index
Grain Δ (in the 3 yr) and flag leaf Δ (in 1995) were positively correlated with both GY and HI (Table 4). These results agree with those reported for bread wheat and barley under both droughted and well watered conditions (Craufurd et al., 1991; Ehdaie et al., 1991; Acevedo, 1993; Morgan et al., 1993; Sayre et al., 1995; Voltas et al., 1998), suggesting that Δ is a good predictor for grain yield, particularly under marked drought. Similarly, maG was correlated with grain yield in the 3 yr, and maF was related to grain yield only in 1995 (Table 4) when drought influences most of the growth period and when limitation in leaf stomatal conductance probably determined TE. The positive association between GY and both ΔF and ΔG (Table 4) suggests that variation in water used for transpiration determines genotype differences in Δ and GY (Morgan et al., 1993). The significant correlation between maF and ΔF in 1995 and 1997 also supports this explanation (Fig. 1d,f).
The negative correlation registered here between maG and GY (Table 4) fully confirmed results from Voltas et al. (1998) and Merah et al. (1999a) suggesting that grain ma could be used as alternative criteria for ΔG to predict grain yield in a large range of climatic conditions, including under drought. Flag leaf ash content is an alternative criterion for ΔF only under water limited conditions (Fig. 1d–f).
Harvest index (HI) was found to be strongly correlated with ΔG in the 3 yr, whereas ΔF and HI were correlated only in 1995 (Table 4). A positive association between Δ and HI was found in peanut (Wright et al., 1993), lentil (Johnson et al., 1995) and cowpea (Menendez & Hall, 1996). In cereals, the relationship between Δ and HI has been scarcely studied. Ehdaie & Waines (1993) found a positive correlation between Δ and HI in bread wheat, suggesting that higher water use efficiency values may result in reduced dry matter partitioning to the grain. Our results suggest that genotypes which were able to maintain higher transpiration losses (and thus high Δ) were more efficient in carbon partitioning to the grain. Harvest index and maG were negatively correlated in the 3 yr. maF was positively related to HI only under droughted conditions (Table 4). As pointed out in the section headed Relationships between Δ and ash content, genotypes which cannot maintain high rates of stomatal conductance and photosynthesis during grain filling (lower Δ) would fill their grains through retranslocation of photoassimilates from preanthesis reserves, and of minerals from early senescent vegetative tissues (Wardlaw, 1990). The ash concentration in mature grain could indicate the importance of the retranslocation processes during grain filling (Merah et al., 1999a). These results suggest that grain ash content is higher (ΔG being thus lower) in genotypes more affected by drought during grain filling.
A broad genotypic variation in LDM, Δ and ma was found in the durum wheat collection. Grain Δ and ma were correlated in the 3 yr studied, whereas ΔF and maF were correlated only under drought conditions. Variation in Δ was related to differences in LDM only under favourable water conditions.
Both grain yield and harvest index were positively correlated to Δ and ma, especially during drought, suggesting that these traits are strongly dependent on stomatal conductance. The positive correlation between grain yield and Δ in this study on 37 genotypes, of diverse origins and having contrasting responses to water supply, confirms that Δ is a good predictor of grain yield in durum wheat under Mediterranean conditions. The positive correlation between HI and both Δ and ma suggests that genotypes which sustain greater transpiration losses (and thus high Δ) during grain filling, are more efficient in dry matter partitioning to the grain, and therefore can produce higher yield in a wide range of contrasted environments.
In the light of these results, the grain ash content (maG) appears as a possible alternative criterion for ΔG, whereas the use of maF for indirect selection of Δ and yield appears to be restricted to severe droughted environments. The extend to which ma may be useful as predictive criterion of Δ in durum wheat, depends upon the consistency of the ranking of genotypes for ma, its heritability and correlation with other desirable and undesirable traits. Neither grain yield nor harvest index were related to LDM in the three contrasted conditions. Moreover, LDM was not related to Δ during drought. Consequently, LDM can not be used as predictive criterion of TE and Δ. Further investigations are needed to investigate the relationship found between Δ and LDM under more favourable water conditions.
O. Merah was supported by a French-Algerian fellowship. Financial support was provided by the French Ministry of Foreign Affairs, in the framework of the INRA-UPS Orsay-ICARDA joint programme ‘Biotechnology and Durum Wheat Breeding’. Authors thank Dr MM Nachit (CIMMYT/ICARDA Programme, Aleppo, Syria) for providing seeds, and the anonymous referees for their constructive comments.