• We examined the relationships among productivity, water use efficiency (WUE) and drought tolerance in 29 genotypes of Populus × euramericana (Populus deltoides×Populus nigra), and investigated whether some leaf traits could be used as predictors for productivity, WUE and drought tolerance.
• At Orléans, France, drought was induced on one field plot by withholding water, while a second plot remained irrigated and was used as a control. Recorded variables included stem traits (e.g. biomass) and leaf structural (e.g. leaf area) and functional traits [e.g. intrinsic water use efficiency (Wi) and carbon isotope discrimination (Δ)].
• Productivity and Δ displayed large genotypic variability and were not correlated. Δ scaled negatively with Wi and positively with stomatal conductance under moderate drought, suggesting that the diversity for Δ was mainly driven by stomatal conductance.
• Most of the productive genotypes displayed a low level of drought tolerance (i.e. a large reduction of biomass), while the less productive genotypes presented a large range of drought tolerance. The ability to increase WUE in response to water deficit was necessary but not sufficient to explain the genotypic diversity of drought tolerance.
Poplars are among the fastest growing trees in temperate latitudes. Their high productivity is associated with large water requirements. As a consequence, productivity depends closely on water availability (Ceulemans et al., 1988; Tschaplinski & Blake, 1989; Tschaplinski et al., 1994; Zsuffa et al., 1996). In Europe, the most frequently used poplars are Populus × euramericana (Populus deltoides × Populus nigra) hybrids, and their cultivation is restricted to flood plains and bottomlands where water availability is not a limiting factor for growth. Consequently, commercial genotypes have been selected primarily on criteria such as high productivity, adequate wood properties and tolerance to pathogens such as foliar rusts. The global warming expected at the end of the 21st century will produce an increased probability of drought episodes, larger vapour pressure deficits and in general more frequent and more severe extreme climatic events (Saxe et al., 2001). The counterbalancing effects of increasing water deficit and rising CO2 concentration are the main factors expected to modify forest production (Loustau et al., 2005). The effects of climate change are predicted to be the most severe in short rotation forestry such as poplar plantations, and the severity of the effects is expected to decrease for less intensively managed forests with longer revolutions and low productivities (Loustau et al., 2005). Furthermore, to sustain the extension of poplar cultivation from flood plains and bottomlands to uplands, where soil water availability is subject to seasonal shortages, would require hybrids with greater water use efficiency (WUE). Thus, it seems to be increasingly important to add traits of drought tolerance and WUE to the panels of criteria already used in poplar selection.
Δ shows great genotypic diversity among poplar species and is highly heritable (Rae et al., 2004; Marron et al., 2005; Monclus et al., 2005). Nevertheless, relationships between productivity and Δ differ among studies: a positive relationship has been found for Populus davidiana (Dode) Schneider (Zhang et al., 2004) while no link has been shown for Populus × interamericana Brockh. (Rae et al., 2004). Thus, selection for high WUE may result in lower or higher productivity according to the predominant factor influencing variations of WUE for a considered species (Condon et al., 2002). Previous studies with P. × euramericana hybrids did not show any relationship between productivity and Δ (Marron et al., 2005; Monclus et al., 2005). Productive genotypes were characterized by a large total leaf area and water use efficient genotypes were characterized by a large leaf nitrogen content (NM), at least in the field (Marron et al., 2005; Monclus et al., 2005). The lack of correlation between productivity and Δ opens a way to select genotypes combining high productivity and high WUE. This lack of correlation also suggests that the intergenotype variability for WUE could be controlled by stomatal conductance (Farquhar et al., 1989). However, the negative relationship between Δ and NM could reflect the fact that the diversity for Δ may also be partly driven by photosynthetic capacity (Brugnoli & Farquhar, 2000).
The results described for P. × euramericana hybrids were obtained from experiments performed under well-watered conditions. In this context, the objectives of the present study were: (1) to test the impact of a moderate water deficit on productivity and on carbon isotope discrimination on a panel of P. × euramericana hybrids; (2) to highlight the relationships among Δ, net CO2 assimilation, stomatal conductance, and Wi, and (3) to test the relationships among productivity, WUE and drought tolerance. These objectives have been addressed using a field trial comprising two plots with randomized blocks established during 2001 at 15-m spacing. Water deficit was induced during spring 2004 by withholding irrigation from one plot while the second plot, which was regularly irrigated, acted as a control. The second plot had already been analysed in detail in 2003 (Monclus et al., 2005). Thus, a corollary of this study was the possibility to test the temporal stability (2004 vs 2003) of genotypic ranking for productivity, Δ, and leaf traits under well-watered conditions. Stem variables encompassed biomass, length and circumference and leaf traits consisted of maximal individual leaf area, carbon and nitrogen contents, specific leaf area, stomatal density, Δ, net CO2 assimilation rate, photosynthetic capacity, stomatal conductance and Wi.
Materials and methods
Plant material and growth conditions
Homogeneous 25-cm-long woody-stem cuttings of 29 Populus deltoides Bartr. ex Marsh. × Populus nigra L. genotypes were planted during January 2001 in a homogeneous open field at INRA Orléans (47°46′ N, 1°52′ E; 110 m above sea level) on a loamy sand soil (pH = 5.9) without fertilizer addition (see Appendix Table A1 for a detailed list of the genotypes and their origins). Two plots were established at 15-m spacing. Each plot was divided into five randomized complete blocks (one cutting of each genotype per block). Inside each plot, trees were planted at 0.5 × 1.2 m spacing. Two border rows of cuttings from the genotypes I214 and Branagesi surrounded each plot. Cuttings were coppiced at the end of 2001, 2002 and 2003. During 2004, bud flush occurred synchronously for all genotypes during the first week of April. Systematic pruning on 18 May 2004 left only one, 40-cm-high shoot on each stool. New shoots and branches of the main stem (sylleptics) were regularly removed during the experiment in order to produce stable continuous growth and to minimize the variability potentially induced by different numbers of shoots per stool. The two plots were irrigated every second day (20 mm of water) with overhead sprinklers, from bud flush to 15 June 2004, the beginning of the experiment. From 15 June 2004 to the end of the growing season, a water deficit was induced by irrigation cessation on one of the plots while the second was kept irrigated and served as a control.
Environmental conditions and soil water potential
From 15 June 2004 to the end of the growing season, daily environmental conditions, in terms of irradiance (J m−2 d−1), temperature (minimum, mean and maximum, °C d−1), potential evapotranspiration (Penman; mm d−1) and rainfall (mm d−1), were measured with a meteorological recording station (Xaria; Degreane Horizon, Cuers, France) that was located near the experimental plot. Soil water availability was followed by measuring predawn leaf water potential (Ψwp, MPa) on a mature leaf with a Scholander-type pressure chamber. Homogeneity of soil water availability within each plot was tested by checking for the spatial stability of Ψwp in all cuttings from four genotypes: I45-51, Soligo, 2000_verde and Luisa_Avanzo (n = 20 plants per plot).
Stem and leaf traits recorded on all genotypes
From 15 June 2004 to 30 September 2004, stem length (Lstem, cm) and circumference 1 m above ground level (Cir, mm) were measured once a week from all cuttings of each genotype. On 7 January 2005, stems were coppiced and final stem length (LstemF, cm), circumference (CirF, mm) and fresh biomass (BiomF, g FW) were recorded.
On 10 September 2004, one fully illuminated mature leaf, which presented the largest width (Foliar Index 17) (Monclus et al., 2005), was collected on each cutting and photocopied in order to measure maximal individual leaf area (LAmax, cm2). Photocopies were scanned and leaf area was estimated with an image analyser (UTHSCSA image tool program developed at the University of Texas Health Science Center at San Antonio, TX, USA, and available by anonymous FTP at http://ddsdx.uthscsa.edu/dig/itdesc.html). Six calibrated discs of lamina (2 cm2) were cut from this leaf; leaf discs were then dried and weighed, and specific leaf area (SLA, cm2 g−1) was computed. Leaf carbon isotope composition (δ13C) and C and N concentrations were assessed for each cutting and genotype from 1-mg homogeneous dry powder obtained from calibrated discs. A continuous flux isotope ratio mass spectrometer (Delta S; Finnigan MAT, Bremen, Germany) was used for 13C analyses. Carbon isotope composition was calculated relative to the Pee Dee Belemnite standard as in Craig (1957): δ13C = [(Rsa − Rsd)/(Rsd) × 1000 , where Rsa and Rsd are the 13C : 12C ratios of the sample and the standard, respectively. The discrimination between atmospheric CO2 (δair, assumed to be close to −8) and plant material (δplant) was calculated as Δ = [δair − δplant)/(1 + (δplant/1000)] according to Farquhar & Richards (1984). Total carbon and nitrogen contents (CM and NM) were obtained with an elemental analyser on the same samples (NA 1500NC; Carlo Erba, Italy) and were expressed on a dry weight basis (mg g−1 DW).
Leaf functional traits related to WUE
Photosynthetic capacity, net assimilation, stomatal conductance, Wi and stomatal density were measured in two groups of four genotypes selected for their contrasting carbon isotope discrimination (Δ13C, ) based on data collected during 2003 on the same plot in well-watered conditions (Monclus et al., 2005): Eco_28, Robusta, Cima and Luisa_Avanzo showed high Δ values, while Pannonia, Agathe_F, Flevo and I45-51 showed low Δ values.
In order to estimate the maximal carboxylation rate (Vcmax, µmol CO2 m−2 s−1) and maximal light-driven electron flow (Jmax, µmole− m−2 s−1), measurements of net CO2 assimilation rate (A) vs internal CO2 concentration (Ci) were taken on well-watered plants with a portable open infrared gas analyser (CIRAS-2; PP System, Hitchin, UK). Photon flux was maintained at 1500 µmol m−2 s−1 and temperature at 25°C. Before measurements, photosynthesis was induced and the stomatal aperture was maximized by decreasing the CO2 concentration in the leaf chamber (200 µmol mol−1) for about 30 min. A was measured at the following CO2 mole fractions: 200, 400, 300, 250, 150, 100, 50, 500, 750, 1000, 1250, 1500 and 1800 µmol mol−1. Vcmax and Jmax were computed from the A/Ci relationships (Dreyer et al., 2001).
Net CO2 assimilation, stomatal conductance and Wi
These parameters were estimated both under well-watered and under water deficit conditions on a clear day (8 September 2004), near midday, on a fully illuminated mature leaf. Net CO2 assimilation rates (A, µmol CO2 m−2 s−1) and leaf conductance (gs, mmol H2O m−2 s−1) were measured using a portable open infrared gas analyser (CIRAS-2) fitted to a leaf chamber operated at 200 ml min−1. The leaf chamber covered an exposed area of 2 cm2. The CO2 concentration inside the chamber was kept at 400 ppm and stomatal distribution was fixed to be equal between the adaxial and abaxial sides. Ambient conditions in terms of radiation, temperature and evaporative demand were recorded. Readings were taken after steady-state conditions had developed (c. 1 min). Carbon dioxide and water vapour concentration differences between inlet and outlet gas circulating through the leaf chamber, as well as leaf temperatures obtained from energy balance equations, were used to calculate A, transpiration (T) and gs on a leaf area basis using von Caemmerer & Farquhar's equations (von Caemmerer & Farquhar, 1981). Wi was computed as the ratio between A and gs.
On 10 September 2004, two calibrated discs of lamina (2 cm2) were cut from one fully illuminated mature leaf (Foliar Index 17) of each cutting, frozen in liquid nitrogen and stored at −80°C. Samples were washed in a mixed tetrahydrofurane/toluene [50/50 volume/volume (v/v)] solution for 30 s and were stuck to aluminium stubs on a Peltier stage (−50°C). They were then examined under a variable pressure scanning electron microscope (model 1450VP; Leo, Cambridge, UK). Backscattered secondary electron images were observed at an accelerating voltage of 20 kV, a probe intensity of 1 nA and a working distance of 10 mm, at ×180 magnification. Digital images of each sample (abaxial and adaxial sides) were captured from two areas on the leaf discs.
Analysis of stomatal density was carried out using the visilog 6.3 program (Noesis, Courtaboeuf, France) executing a macro command. The threshold to distinguish between stomata and leaf matrix in the image was obtained by an interactive image-segmentation based on selected image grey-level ranges that resulted in a binary image (stomata black and background white). Binary filter operations were performed on the recorded images with a single erosion step to remove pixel noise; finally, a dilatation operation was applied to compensate for the erosion steps before the stomatal number was calculated. The total number of stomata within an image was determined by counting, using the convention of excluding parts of stomata overlapping the margins of the image, and was converted to stomata per mm2. Total stomatal density (TSd) was computed by summing adaxial and abaxial stomatal densities.
Results were evaluated by linear regression, linear correlation (Pearson's coefficients), rank correlation (Spearman's coefficient), and analysis of variance (ANOVA) using the SPSS statistical software package (SPSS, Chicago, IL, USA). For each variable, the normality of the distribution was tested by a Shapiro–Wilk test. All statistical tests were considered significant at P≤ 0.05. Genotypic means are expressed with their confidence interval (α = 0.05) or their standard deviation.
In order to study the genotypic variability of traits on all genotypes in this experiment under well-watered conditions, we performed multivariate analyses using principal components analysis (PCA). The variables measured in well-watered conditions (BiomF, LstemF, CirF, LAmax, NM, CM, SLA and Δ) were standardized and orthogonal factors (= F1 and F2 axes) were successively built as linear combinations of these variables to maximize the part of the variability explained by these factors. Variables were first represented on the plane defined by the two main factors of the PCA; their coordinates were their linear correlation coefficients (Pearson's coefficient) with these factors. Variables measured in water deficit conditions were projected as supplementary variables in the main plane F1 × F2 defined by variables measured in well-watered conditions. Coordinates of variables measured in water deficit conditions were obtained by computing linear correlations between these variables and scores F1 and F2 obtained from PCA measured in well-watered conditions.
Genotypic diversity and correlations among variables under well-watered conditions
Substantial genotypic diversity was recorded in the well-watered plot for all tested variables (Table 1). BiomF differed between genotypes along a 3-fold range. NM of mature leaves, LAmax and SLA ranged from 23 to 33 mg g−1, 133.6 to 461.3 cm2 and 107 to 145 cm2 g−1, respectively (Fig. 1). Δ differed by 2.7 among genotypes.
Table 1. Abbreviations, descriptions and values of variables used in the study
WW vs WD
General mean (ICα=0.05)
General mean (ICα=0.05)
The variables A, Vcmax, Jmax, gs, Tsd and Wi were recorded on eight genotypes selected for their contrasting Δ measured during 2003. The other variables were recorded on the full set of 29 genotypes. All variables were recorded on trees in the two treatments [except for Vcmax and Jmax, which were not determined (nd) in water deficit conditions]. Variables with the subscript F were recorded on 7 January 2005. For each variable and each treatment, well-watered (WW) vs water deficit (WD) conditions, general means, confidence intervals (CI; α = 0.05), and genotypic diversity (ANOVA, PWW and PWD) are indicated. The effects of drought on each variable and on genotypic ranking were estimated by ANOVA (P) and rank correlations (Spearman's coefficient, rs), respectively. Levels of significance are indicated by asterisks: *P 0.05; **P 0.01; ***P 0.001 and by ‘ns’ for nonsignificant values. DW, dry weight.
A general PCA was performed with the genotypic means of the variables recorded on all genotypes (Fig. 2a, Table 1). The main plane of the PCA (F1 × F2) explained 66% of the genotypic diversity, with 44% for F1 alone, while the F3 axis did not differentiate the genotypes (data not shown). Two independent groups of variables were identified from the F1 × F2 plane. The first group included BiomF, LstemF, CirF and LAmax and the second group included SLA and Δ. In each group the variables were positively correlated (Table 2). The F1 axis of the PCA was only defined by the first group (BiomF, LstemF, CirF and LAmax), while the F2 axis opposed NM to SLA and Δ. There was no clear grouping among genotypes in the F1 × F2 plane (Fig. 2b). Genotypes NL-3972, Lambro and Mella displayed high productivity, Soligo and Mellone_Caro had high values of Δ, and I45-51 and Agathe_F showed both high productivity and low values of Δ. Interestingly, we were unable to detect any correlation between productivity (as estimated from BiomF, LstemF or CirF) and Δ under these conditions (Table 2).
Table 2. Linear correlations (Pearson's coefficients) computed for well-watered (below the diagonal line) and for water deficit (above the line) conditions between variables from 29 (bold, n = 29) or eight (not bold, n = 40) genotypes
Vcmax and Jmax were not determined in water deficit conditions (nd). The level of significance is indicated by asterisks: *P 0.05; **P 0.01; ***P 0.001. See Table 1 for definitions of variables.
The data recorded during the present experiment (2004) were compared with those recorded on the same experimental plot under well-watered conditions during 2003 (Monclus et al., 2005). Genotypic diversity was of the same magnitude during the two years for leaf Δ, leaf CM and SLA, while it was lower during 2004 than during 2003 for NM and LAmax (Fig. 1). Rank correlations (Spearman's coefficient) for leaf traits (LAmax, NM, CM, SLA and Δ) showed that the genotypic ranking remained stable from 2003 to 2004 for all variables except SLA (Fig. 1). Values of SLA and Δ were significantly higher in 2004 than 2003, while values of LAmax were significantly lower in 2004 than 2003; these differences can be ascribed to differences in climate, 2003 having been warmer, with larger vapour pressure deficits, although the poplars were optimally irrigated and no drought stress occurred on the control plots in either year (Fig. 1a, Table 3). The differences could also have been related to a seasonal effect (earlier collection during 2003) or, most probably, to a combination of both seasonal and climatic effects. NM was lower during 2004, possibly because of decreased nitrogen availability in the soil (Fig. 1). The genotypic ranking remained stable between 15 July 2003 and 7 January 2005 for biomass (rs = 0.58, P 0.01) and circumference (rs = 0.70, P 0.001), but not for stem length. During both years, biomass was positively correlated with stem length and circumference and maximal individual leaf area, but not with Δ (r = –0.23; P = 0.23) (Table 2). The negative correlation observed between Δ and NM during 2003 was again detected during 2004, but was less significant (P = 0.08). Δ and SLA scaled positively during 2004 only (Table 2).
Table 3. Climate (air temperature, Penman evapotranspiration and daily irradiance) during the expansion of the leaves collected during 2003 and 2004
25 June to 15 July 2003
22 August to 10 September 2004
Significant differences (ANOVA) are indicated by asterisks: **P 0.01; ***P 0.001.
Impact of a moderate water deficit on genotypic diversity and correlations among variables
Withholding irrigation during the whole summer of 2004 induced significant decreases in Ψwp as compared with controls (Fig. 3a), with peak differences at the beginning of September. Ψwp remained high in controls (above −0.30 MPa) while it declined to −0.60 MPa in drought-stressed individuals. The spatial and genotypic stabilities of Ψwp were checked by comparing values from four different genotypes and individuals spread over the whole experimental area; no spatial or genotypic differences were detected. We may state from these observations that the drought was moderate but long-lasting (several weeks), with a peak during September, and that it was of similar intensity and duration in all genotypes over the whole experimental plot. The impact of water deficit on stem length and circumference became significant from 15 July 2004 and led to a significant decrease in biomass production for most of the genotypes as recorded on 7 January 2005 (Figs 3b,c,4a, Table 1). While the reductions in Ψwp were comparable among cuttings, the impact on biomass differed widely among genotypes (Fig. 4a). For example, I45-51 and Soligo displayed the same level of productivity under well-watered conditions and were subjected to similar water deficit intensities, as attested by similar maximal decreases in Ψwp (−0.50 and −0.52 MPa, respectively). Nevertheless, they differed substantially in reduction in biomass in response to drought (Fig. 4a): I45-51 showed similar productivities in the two treatments, while a 50% decrease was observed for Soligo. It is noteworthy that most of the productive genotypes were very sensitive to drought (i.e. showed a large reduction in biomass), while the less productive genotypes showed a large range of biomass decreases, i.e. of drought tolerance.
In general, leaf variables such as LAmax, SLA, NM and Δ were affected by drought, and the expected changes occurred (i.e. decreases in LAmax, SLA, NM and Δ) (Fig. 1, Table 1). Although no correlation has been previously reported between leaf traits measured in well-water conditions and the level of drought tolerance, genotypic variability in the decrease in LAmax was strongly correlated with genotypic variability in drought tolerance (Fig. 4b). However, a link has not been shown between the decrease of LAmax in response to drought and both LAmax and above-ground biomass measured under well-watered conditions (data not shown). Interestingly, 23 genotypes among the 29 studied decreased or tended to decrease their Δ in response to drought (Fig. 4b). Of the six genotypes that tended to increase Δ, none was drought tolerant (triangles, Figs 4a,c). Conversely, the whole range of drought tolerance levels was observed among those for which Δ decreased or tended to decrease (Fig. 4c). Although most of the variables were affected by drought, the genotypic diversity remained comparable to that observed under well-watered conditions (Fig. 1, Table 1). Furthermore, the moderate drought constraint did not induce any significant change in genotypic ranking for variables related to productivity and for leaf variables such as LAmax, NM and Δ. SLA was the only variable for which genotypic ranking was significantly modified by drought (Table 1). It is noteworthy that genotypic ranking for SLA did not differ between well-watered conditions during 2003 and moderate water deficit conditions during 2004 (Fig. 1).
Under both control and drought conditions, Δ did not correlate with biomass, while it was positively correlated with SLA (Table 2). The projection of variables measured after water deficit on the main plane of the previous PCA resulted in the same grouping of variables as in controls, and, within each group, variables remained positively correlated (Fig. 2a, Table 2). Taking into account the finding that drought led to decreases in productivity and in Δ, the relative positions of variables of drought plants vs control plants allow us to identify an axis related to genotypic productivity and another related to Δ. In this context, the F1 axis can be regarded as the productivity axis, but F2 differs slightly from a Δ axis, for which an inclined axis would be more appropriate.
Relationship between Δ and other leaf functional variables
Under well-watered conditions, great genotypic diversity was recorded for all functional variables (Table 1): Wi (= A/gs), A and gs ranged from 0.028 to 0.045 (mmol mol−1), from 14.4 to 27.0 (µmol m−2 s−1) and from 515 to 712 (mmol m−2 s−1), respectively. Vcmax, Jmax and TSd differed among genotypes by more than a 2-fold range. Under drought, significant genotypic diversity was observed for A and TSd only.
Under moderate drought conditions only, a linear and negative relationship was found between Δ and Wi (Table 2). Δ scaled positively with gs under moderate drought conditions. A and gs were positively correlated, but Wi scaled positively with A in controls, and negatively with gs under drought (Table 2). TSd scaled positively with Δ whatever the water conditions and negatively with Wi in controls. Under well-watered conditions only, Jmax scaled positively with A, Vcmax and NM and negatively with SLA. No correlation was observed between Wi and variables of productivity. This observation corroborates the lack of relationship described, at the genotypic level, between productivity and Δ. Jmax and Vcmax scaled positively with LstemF, and Vcmax tended to be positively linked with biomass in well-watered conditions (P = 0.053). TSd scaled negatively with BiomF, CirF and NM whatever the water availability.
Genotypic diversity for productivity, Δ and leaf traits under well-watered conditions
Great genotypic diversity was found among Populus ×euramericana genotypes for all tested variables. This confirms similar observations in a glasshouse and in the field for the same panel of genotypes under optimal water supply (Marron et al., 2005; Monclus et al., 2005). Although genotypic ranking differed between glasshouse and field experiments (Monclus et al., 2005), it did not differ for most of the variables when measurements were performed in the field during two successive years (2003 and 2004). Genotypic ranking did not differ for stem biomass and circumference, while it varied for the length of the stem, confirming that stem circumference is the more reliable and early index of genotypic productivity. Inter-annual differences in leaf Δ, LAmax, leaf NM and CM and SLA may be imputed, at least in part, to differences in environmental growth conditions (Niinemets et al., 1998; Barbour et al., 2002; Monclus et al., 2005). Nevertheless, genotypic ranking was modified for SLA only, suggesting that the plasticity of leaf density/thickness to environmental changes is genotype dependent (Monclus et al., 2005).
As previously described for P. × interamericana (Rae et al., 2004) and for P. × euramericana growing in a glasshouse and in the field (Marron et al., 2005; Monclus et al., 2005), biomass did not correlate with Δ. This result, combined with the finding that Δ and productivity showed broad-sense heritabilities above 0.5 in P. × interamericana (Rae et al., 2004) and P. × euramericana (Monclus et al., 2005), suggests that there is potential for improving Wi in poplar without reducing overall productivity. Stomatal density, which was negatively correlated with biomass and circumference and positively correlated with Δ, could be used as one index for the selection of genotypes combining high productivity and WUE.
The lack of correlation between productivity and Δ suggests that the genotypic variability for WUE is controlled by stomatal conductance (Farquhar et al., 1989). Nevertheless, the negative correlation observed between Δ and NM during 2003 and 2004 suggests that genotypic diversity of photosynthetic capacity could also contribute to the intergenotype variability for WUE (Monclus et al., 2005). This negative correlation is usually explained in terms of increases in concentration of chlorophyll and Rubisco in leaves resulting in a higher photosynthetic capacity per unit leaf area, a smaller concentration of carbon dioxide at sites of carboxylation for the same stomatal conductance and lower discrimination against 13C (Anderson et al., 2000; Macfarlane et al., 2004). Δ and SLA scaled positively during 2004, suggesting that the variability of internal conductance to CO2 transfer into the leaves could also contribute to intergenotype variability for WUE. This link, already demonstrated by Vitousek et al. (1990), has been cited as evidence of the effect on Δ of the internal conductance for CO2 transfer. However, genotypic ranking for SLA differed significantly between 2003 and 2004, while it was conserved for Δ, indicating a limited contribution of SLA to intergenotype variability for WUE in P. × euramericana.
Impact of a moderate water deficit on productivity, Δ and leaf traits
Withholding of irrigation induced a moderate but long-lasting drought, as shown by the significant reduction in predawn leaf water potential. Moreover, the intensity of drought was homogeneous over the whole experimental plot and no genotypic effect was visible on leaf predawn water potential. Although the pruning process has produced differences between genotypes in terms of rootstock over the years, we therefore assume that all genotypes were subjected to similar stress during a large fraction of the growth season. As shown by the early deviations of both stem length and circumference increments between controls and drought-stressed plants (15 July), the water deficit persisted over a large fraction of the growth season. Because of this homogeneity in drought within and between blocks, we may state that the impact of drought on biomass production was to a large extent genotype dependent. Except for I45-51, the most productive genotypes displayed large drought-induced decreases in biomass, i.e. a low water deficit tolerance. The less productive genotypes displayed a large range of tolerance. Productive genotypes probably had higher rootstocks than less productive ones, but, because most of the productive genotypes were very sensitive to drought, we can hypothesize that the effect of the rootstock did not reduce the impact of withholding irrigation in our study.
Drought also induced significant decreases in Δ, LAmax, NM and SLA but did not modify the genotypic ranking for productivity, Δ, LAmax and NM. The genotypic ranking obtained for SLA under water deficit differed significantly from that obtained under irrigation in the same year but was comparable to that measured during 2003 under well-watered conditions. Taking into account the fact that SLA shows high plasticity to soil water content and climatic variations (Abrams et al., 1990; Niinemets et al., 1998; Marron et al., 2002; Wright et al., 2004), and that temperatures and cumulated irradiance were significantly higher during July 2003 than September 2004, we conclude that each genotype is characterized by a different range of plastic responses of SLA. Many other studies have found for woody species a similar impact of indices of water availability such as soil water potential and vapour pressure deficit on SLA and Δ (Abrams et al., 1990; Damesin et al., 1997; Niinemets et al., 1998; Korol et al., 1999; Miller et al., 2001; Barbour et al., 2002; Marron et al., 2002; Wright et al., 2004). However, similar variations of environmental conditions modified genotypic ranking for SLA only, suggesting that the response of Δ to water availability is less genotype dependent than that of SLA.
Drought did not modify the relationships between variables and no correlation was found between productivity and Δ under these conditions. Productive genotypes were very sensitive to drought (as seen from the large reduction in shoot biomass) and showed, simultaneously, a large range of Δ values. It is noteworthy that, of the genotypes for which Δ tended to increase in response to drought, none was a drought-tolerant genotype. This result suggested that the genotypic ability to increase WUE is necessary to produce a high level of drought tolerance. However, the whole range of drought tolerance levels was observed among those genotypes for which Δ decreased or tended to decrease, suggesting that the ability to increase WUE is necessary but not sufficient to explain genotypic diversity of drought tolerance among P. × euramericana hybrids. It is clear that drought tolerance it is not only governed by WUE but probably includes a lot of other traits that also contribute to the overall drought tolerance capacity of the genotype (Liu & Dickmann, 1992; Gebre et al., 1994; Chen et al., 1997; Ibrahim et al., 1997; Tschaplinski et al., 1998; Marron et al., 2002, 2003). Although no direct link was detected between leaf traits measured in well-watered conditions and the level of drought tolerance of the genotypes, it is noteworthy that the magnitude of leaf area reduction is a good indicator of the magnitude of above-ground biomass reduction and so of the drought sensitivity of the genotype. The lack of correlation observed between the magnitude of leaf area reduction and leaf area or above-ground biomass measured in well-watered conditions supported the idea that there is potential for improving concomitantly productivity and drought tolerance. One of the genotypes (I45-51) displayed not only a high productivity and a low Δ, but also a rather high level of tolerance to a moderate drought constraint. This genotype clearly warrants additional attention from this point of view.
Carbon isotope discrimination (Δ) scaled negatively with Wi during drought only. This negative correlation is in agreement with the model of Farquhar (Farquhar & Richards, 1984; Farquhar et al., 1989) and with previous experiments with woody species such as Quercus robur (Ponton et al., 2002) and Populus davidiana (Zhang et al., 2004). The lack of correlation in well-watered genotypes may be a result of the fact that we measured gas exchange at peak values of gs during diurnal cycles. Taking into account that Δ scaled positively with stomatal conductance (gs), these results show that the diversity for Δ is mainly driven by stomatal conductance and support our previous conclusion regarding the lack of correlation between Δ and productivity. While Wi scaled positively with A and Jmax in well-watered conditions, no direct correlation was found between Δ and photosynthetic capacity; this again supports the important contribution of stomatal conductance rather than net CO2 assimilation to the diversity of Δ measured in leaf tissues. However, in controls, Jmax scaled positively with A, Vcmax and NM, and NM was negatively correlated with Δ, suggesting at least a minor contribution of photosynthetic capacity to the diversity of Δ in P. × euramericana. In this context, additional studies are needed to determine whether these results can be extended to other poplar genetic backgrounds such as P. × interamericana and to evaluate the genetic determinism of traits related to productivity, WUE and drought tolerance in P. × euramericana and P. × interamericana. The results obtained in this study were derived from coppiced shoots that behave as plants in their juvenile growth phase. Thus genotypic diversity for productivity, WUE and drought tolerance must be validated in more mature plants with a physiology typical of older trees.
This study demonstrated great diversity in the levels of drought tolerance, assessed from the reduction of biomass accumulation under moderate drought, in P. × euramericana genotypes. We observed that, of the most productive genotypes, many were drought susceptible, and that the less productive genotypes displayed a large range of drought tolerance. We have shown that the magnitude of leaf area reduction under water deficit is a good indicator of the drought susceptibility of the genotype. We have also shown that the decrease of Δ in response to moderate drought is necessary but not sufficient to explain the diversity of drought tolerance in P. × euramericana. Because Δ scaled negatively with Wi and positively with stomatal conductance in moderate drought conditions, these results support the idea that diversity for Δ is mainly driven by stomatal conductance. This study confirms that productivity and leaf Δ display large genotypic variability among P. × euramericana hybrids and that genotypic differences remain stable over time in field experiments. It also confirmed the lack of correlation between productivity and Δ that was previously detected in the glasshouse and in the field under different growth conditions (Marron et al., 2005; Monclus et al., 2005) and thus supports the conclusion that there is potential for improving WUE in poplar without necessarily reducing overall productivity. The operational deployment of productive, water-use-efficient genotypes will reduce the cost of plantation management by reducing the need for irrigation. Interestingly, stomatal density could be used to identify genotypes combining productivity and WUE. We have identified among P. × euramericana hybrids at least one genotype combining high productivity, high WUE and drought tolerance, the cultivar I45-51; this cultivar warrants additional attention to determine its potential for afforestation in stations with moderate soil water deficits.
RM was supported by a PhD grant from the Conseil Régional, Région Centre, France and from Institut National de la Recherche Agronomique. Genotypes were kindly provided by O. Forestier from the State Forestry Nursery of Guéméné-Penfao (Loire Atlantique, France). The authors gratefully acknowledge the excellent technical assistance of René Blanluet (UE Amélioration des Arbres Forestiers, INRA Orléans), Isabelle Le Jan and Gilles Moreau (Laboratoire de Biologie des Ligneux et des Grandes Cultures, Université d’Orléans), and their survey of the experimental field. We also acknowledge Micaël Paganotto for writing the macro command in the visilog program, and Nathalie Ningre for performing all scanning electron microscope analyses of stomatal density.
Table A1. List of abbreviations and origins of the 29 Populus x euramericana genotypes studied