Introduction 

In the context of climate change and fluctuating precipitation patterns ([20]), water use efficiency (WUE, ratio between biomass accumulated and water transpired) has become an interesting trait for breeding new genotypes ([9], [41], [46]). Isotopic discrimination against ¹³C (Δ¹³C) during CO₂ diffusion through stomata and photosynthesis is an indirect indicator of intrinsic water use efficiency (iWUE, ratio between net CO₂ assimilation rate, A, and stomatal conductance to water vapour, g_s - [14], [15], [13]). Indeed, this trait was found to be under tight genetic control in several tree species like poplar, pines, chestnut and oaks among others ([22], [5], [4], [40]). Furthermore, given the lack of correlation observed between WUE and productivity or, in some instances, the positive correlation between both traits ([46]), selecting tree species or genotypes based on their ability to express a high WUE is desirable. Such a strategy has been explored in some crops in a context of improving water management ([9], [41]) and might thus be particularly suitable for sustaining wood production under the incertitude of water availability.

Poplar genotypes are known to display large differences in growth performance and biomass production ([39]). A large genotypic variability of Δ¹³C was found among Populus deltoides × P. nigra genotypes in controlled environment ([26]) and open field experiments ([29], [10]). Using direct measurements, this variability in Δ¹³C is also found tightly related to variability in the whole plant water use efficiency in Populus deltoides × P. nigra and in Populus nigra ([39], [40]). Interestingly, no correlation was found between productivity and Δ¹³C ([29], [10]), giving opportunity to breed for enhanced WUE without compromising productivity in poplar genotypes. Nevertheless, previous studies on poplar genotypes were conducted on seedlings of 1-2 years and there are few studies investigating the stability of genotype ranking for WUE at tree scale.

Previous studies has demonstrated that Δ¹³C (recorded from tree ring cellulose and bulk wood) is severely affected by age, usually decreasing by a 1-3 per mil with increasing tree age ([16], [1], [12], [28], [6]). Such an age-related effect may be related to: (i) assimilation of respired CO₂ near the forest floor which is already depleted in ¹³C ([45]); (ii) re-assimilation of respired CO₂ retained within the plant canopy ([16]); (iii) decreasing hydraulic conductivity of xylem conduits with age ([31]); (iv) reduced total soil-to-leaf hydraulic conductance with tree size ([27]); or (v) tree height and light availability ([6]). Records of Δ¹³C in leaves and phloem sap of poplars with different ages revealed a rather stable ranking of the genotypic differences with tree age ([3]). More recently, using a diachronic approach, several genotypes of Populus deltoides × P. nigra growing in common gardens were compared and the effects of ageing (from 5 to 20 years) on the genotype ranking were studied using the Δ¹³C signals in tree rings as a surrogate for WUE ([38]). In the latter study, small changes in Δ¹³C with tree age and no significant interaction between genotype and tree age were found, reinforcing the pertinence of ranking genotypes using Δ¹³C. However, in a such diachronic study, the influence of long-term environmental variability and annual effects related to rainfall on Δ¹³C is assumed similar as the sampled trees were grown in common gardens that experienced the same environmental conditions.

In the present study we used a synchronic approach where Δ¹³C was recorded and compared in tree rings with same formation date, but from tree of different ages and contrasting environmental conditions. This sampling technique was used to minimise long-term environment effect and maximise age effect. Three different genotypes of Populus deltoides × P. nigra were used and carbon isotopic composition (δ¹³C) of the bulk wood from the year 2009 was measured and Δ¹³C was computed. Since cellulose and lignin are the major components of bulk wood, and due to different biochemical pathways involved in their formation, the carbon isotopic signals of these two components are also different, i.e., lignin is ~3‰ depleted than cellulose ([24], [27]). Any variation in cellulose:lignin ratio can potentially affect the overall δ¹³C signals of whole wood. Therefore, the variation in carbon percentage in the whole wood was investigated, that potentially elucidates the variation of lignin in the bulk wood ([25]). Moreover, nitrogen percentage in the bulk wood that can give insight regarding the tree response to the variation of available soil nitrogen resource ([17], [2]) was also measured.

The main difficulty of the synchronic approach is the lack of large-scale common garden grouping a sufficient number of genotypes with different ages. We therefore used 41 small common gardens (Fig. 1) maintained in different locations in France by the Institut du Développement Forestier (IDF-CNPF) and gathered samples for three genotypes of Populus deltoides × P. nigra. Using this approach, we tested the hypotheses that: (i) water use efficiency estimated from Δ¹³C in wood is affected by tree age; and (ii) genotype ranking based on Δ¹³C does not change with tree age.

Fig. 1 - Location of 41 sampling sites across France. Each site is represented by a closed star.

  Enlarge/Shrink   Download   Full Width  Open in Viewer

  Material and methods 

The aim of this study was to evidence the effect of age on stability of genotype ranking for Δ¹³C. However, studying long-term physiological response in trees can be tricky because of age and environmental signals overlapping each other. Therefore, maximising age-related signals through appropriate sampling technique becomes a prerequisite. In this regard a synchronic technique, where carbon isotope discrimination was compared among tree rings with same formation date but from trees of different ages and environments, thereby maximising the age signal. A major constraint while applying such technique is to find a large number of poplar genotypes of different ages in a common garden. To obtain a satisfying sample of trees of different ages in different genotypes and environments, we sampled trees across a large number of plantations in France (Fig. 1). These small plantations were established by the Centre National de la Propriété Forestière - Institut pour le Développement Forestier (CNPF-IDF) to compare the local performance of genotypes.

Study sites

The genotypes were planted in plots of 25 trees at a distance of 7 × 7 m, i.e., at a density of 204 stems ha^-1. Forty-one sites were selected (Fig. 1) which were located all over France (see Tab. S1 in Supplementary material for site information and spatial coordinates). Sites covered a large range of climates with mean annual temperature from 10 to 14 °C and precipitation from 600 to 1200 mm. Sampling sites were subsequently categorised as water available (WA, annual rainfall from 1000 to 1200 mm) or water limiting (WL, annual rainfall from 600 to 800 mm; precipitation record maintained by the Centre National de la Propriété Forestière - Institut pour le Développement Forestier, CNPF-IDF).

Sample collection and preparation

We focused on 3 Populus deltoides × P. nigra genotypes: Koster, I-214 and Dorskamp, as these genotypes are known to display contrasting Δ¹³C at leaf level ([29]). Moreover, they were present in a large number of the surveyed plantations and covered the different age classes. The last annual ring corresponding to year 2009 was sampled in February 2010 with an increment borer (0.5 cm²) on ~5 trees site^-1 genotype^-1, yielding 376 samples. All cores were divided into seven age classes, i.e., 4, 7, 9, 11, 13, 15 and 18 years. In order to minimise the environmental effect on Δ¹³C: (i) date effect was discarded as all cores corresponded to year 2009; and (ii) site effect was averaged in each age class by ensuring the representation of trees from both site types, i.e., WA and WL sites.

The ring formed in 2009 was carefully separated from the adjacent one and from bark with a sharp razor blade. After drying at 70 °C for 48h, each ring was ground separately into fine homogeneous powder using a ring grinder (SODEMI, CEP Industries Department, Cergy-Pontoise. France). One mg of the resulting wood powder was weighed in tin capsules for δ¹³C analysis.

Carbon and nitrogen percentage and carbon isotope analysis

Wood powder was combusted at 1050 °C in sealed evacuated quartz tubes containing cobalt oxide and chromium oxide as catalyst, and an amount of pure oxygen. The gases produced during combustion, CO₂, H₂O and NOx were passed through a reduction tube where N₂ was produced and excess of oxygen was removed. Water was trapped by using anhydrous magnesium perchlorate and after reversible absorption, carbon and nitrogen percentages were measured in an elemental analyser (NA 1500-NC^®, Carlo Erba, Milan, Italy - [38]). Finally, combusted products were separated by gas chromatography and the CO₂ was delivered to an isotope ratio mass spectrometer (Delta-S^®, Finnigan, Bremen, Germany). Carbon isotope composition was expressed as δ¹³C (eqn. 1):

where R_sample and R_standard are the ¹³C/¹²C ratios in a sample and the standard (Vienna- Pee Dee Belemnite) respectively. Accuracy of the measurements was ± 0.1‰. Carbon isotope discrimination between atmosphere and wood was calculated as (eqn. 2):

where δ¹³C_air is the carbon isotope composition of CO₂ in the atmosphere and δ¹³C_wood is the carbon isotope composition of the wood powder. Given that δ¹³C_air was -8.07‰ during 2003 and an annual decrease of 0.0281‰ ([27]), δ¹³C_air was estimated at -8.24‰ for 2009. We assumed that the mean value of δ¹³C_air was similar across sites and did only marginally change with the season.

Basal area increment

The data corresponding to annual circumference (cm) of each individual tree was acquired through the annual growth records from the Centre National de la Propriété Forestière-Insitut pour le Développement Forestier. Assuming trunks were circular in section, circumference was used to calculate the radius and eventually basal area for the years 2008 and 2009. BAI (cm² y^-1) for year 2009 was calculated as the difference between the total area corresponding to years 2009 and 2008.

Statistical analysis

Normality and homoscedasticity of data were checked graphically with residual vs. predicted and normal quantile-to-quantile plots. The data set covering the seven age classes was analysed using linear mixed models fitted with genotype, age and their interaction as fixed effects and site as a random effect. Multiple comparison tests (post-hoc Tukey HSD test) were used to evaluate pairwise differences between age classes within each genotype. Furthermore, each age class was tested for genotype effect and post-hoc Tukey HSD test were used to evaluate pair-wise differences between genotypes within each age class. All tests were performed with R ([37]) and R packages “nlme” ([35]) and “multcomp” ([19]). All tests and correlations were declared significant at P < 0.05.

  Results 

Carbon and nitrogen percentage

Variance between sites was smaller than residual variance (intercept = 0.731; residuals = 1.84) and model R² resulted at 0.547 for carbon percentage. No genotype-age interaction was detected in the carbon percentage of bulk wood. It differed significantly among genotypes (P < 0.001 - for means, see Tab. 1). A non-significant trend with age was detected due to the slightly lower carbon percentage values at the age of 13 yrs. Nevertheless, carbon percentage did not display any significant trend related to age (Fig. 2). Variance between sites was smaller than residual variance (intercept = 0.009; residuals = 0.071) and model R² resulted at 0.525 for nitrogen percentage. For the nitrogen percentage, no interaction between genotype and age was found. Nitrogen percentage did not differ among genotypes but declined significantly with age (P < 0.001); it dropped from 0.19 down to 0.11% over the tested period (Fig. 3).

Fig. 2 - Age-related variation of Carbon percentage in the bulk wood of the year 2009, sampled during winter 2010 in three Populus deltoides × P. nigra genotypes over seven age classes. Each point represents the mean of several sites. Error bars represent ± SE.

  Enlarge/Shrink   Download   Full Width  Open in Viewer

Fig. 3 - Age-related variation of mean nitrogen percentage in the bulk wood of the year 2009, sampled during winter 2010. The curve represents the means of three Populus deltoides × P. nigra genotypes over seven age classes. Error bars represent ± SE and different letters indicate significant differences between age classes after Tukey HSD test (P < 0.05).

  Enlarge/Shrink   Download   Full Width  Open in Viewer

Age and genotype effects for Δ¹³C

Variance between sites was smaller than residual variance (intercept = 0.38; residuals = 0.51) and model R² resulted at 0.99. Δ¹³C variation was significantly explained by an interaction between genotype and age (P < 0.001 - Tab. 1). A detailed post-hoc analysis revealed that an age effect was detectable in all three genotypes (Fig. 4). There was a significant increase of Δ¹³C by 2‰ from 4 to 9 yrs in Dorskamp and Koster (Fig. 4a, Fig. 4c) and a less visible one in I-214 (Fig. 4b) where it increased from 4 to 7 yrs and remained almost stable afterwards.

Fig. 4 - Age-related variation in carbon isotope discrimination (Δ¹³C) in the 2009 year-ring for three Populus deltoides × P. nigra genotypes (a) Dorskamp, (b) I 214 and (c) Koster. Each point represents a mean value from several sites. Error bars represent ± SE and different letters indicate significant differences between age classes for each genotype after Tukey HSD test (P < 0.05).

  Enlarge/Shrink   Download   Full Width  Open in Viewer

Genotype effect within each age group was found significant and was assessed with post-hoc Tukey HSD test (Tab. 2). In the young age classes (4 to 9 yrs), Koster displayed the lowest mean Δ¹³C followed by Dorskamp and I-214 that displayed similar values. At higher ages, Koster and I-214 displayed the lowest and Dorskamp displayed the highest mean Δ¹³C, with the exception of age class of 13 yrs, where no significant difference could be detected.

Inter annual and genotype correlation between Δ¹³C and BAI

The average BAI of the year 2009 was homogenous among the three genotypes over the tested period (Tab. 1), and therefore the genotypic differences in Δ¹³C were independent from any difference in radial growth rate. The inter annual variation of BAI due to the age effect was positively correlated to Δ¹³C in the three genotypes (Fig. 5)

Fig. 5 - Correlation between carbon isotopic discrimination (Δ¹³C) and basal area increment (BAI) in three Populus deltoides × P. nigra genotypes (a) Dorskamp: Δ¹³C = 0.0161BAI + 18.9; (b) I-214: Δ¹³C = 0.0087 BAI + 19.0; and (c) Koster: Δ¹³C = 0.0138 BAI + 18.0. Each point represents values measured at a given site.

  Enlarge/Shrink   Download   Full Width  Open in Viewer

  Discussion 

The present research was aimed to assess the stability of genotype differences in water use efficiency (WUE) with tree age; we compared three cultivars of Populus deltoides × P. nigra for this trait in plantation in situ across France and used a synchronic approach, based on Δ¹³C recorded from the year 2009 of individuals with different ages.

Genotype and age effect on Δ¹³C and relationship with BAI

Whole wood consists of complex mixtures of molecules like cellulose, hemicellulose and lignin with variable isotopic signatures. Early studies used whole wood for the isotopic analysis on tree rings. However, the isotopic composition of individual wood components differs largely ([47]). Thus, we suspected that the observed significant genotypic difference in mean Δ¹³C could potentially be due to changes in cellulose vs. lignin ratios with contrasting isotopic signals. Although previous studies have shown that using whole wood is better than inducing extraction error while using cellulose to measure δ¹³C ([18], [38]), in this study we minimised this effect by using the last annual ring (2009, with same cambial age) for isotope analysis.

Our results showed the existence of an age effect on Δ¹³C, where Δ¹³C increased to a larger extent during first age classes of 4-11 years in two genotypes (Dorskamp and Koster). This positive trend in Δ¹³C with age is contrary to several previous studies reporting age-related decreases in Δ¹³C using different dendrochronological approaches ([1], [12], [28], [34]). Δ¹³C has been expected to be an estimator of water use efficiency ([15]). Several studies have shown a clear negative correlation between Δ¹³C and A/g_s (intrinsic water use efficiency, iWUE) and WUE (whole-plant water use efficiency - [43], [7], [39]). Thus, increase in Δ¹³C with age in our study reflects a decrease in iWUE with age. This decrease in iWUE (A/g_s) can either be due to decreased CO₂ assimilation rate (A) or increased stomatal conductance (g_s - [14], [9]). Furthermore, Δ¹³C and tree growth are often correlated depending upon the growth environment. This correlation was found to be positive for some species, i.e., Eucalyptus globulus Labill ([33]. [36]), Fagus sylvatica L. ([11]) and Pinus radiata ([44]). On the contrary, some other conifer species displayed a negative correlation between Δ¹³C and growth, i.e., Larix occidentalis Nutt. ([49]), Pinus pinaster ([32], [5]), Picea mariana Mill ([21]) and Pinus caribaea Morelet ([48]). However, no correlation between Δ¹³C and tree growth was found in P. × euramericana in common garden ([29]) and poplar plantation ([3]). Within each genotype, we found a positive correlation between BAI and Δ¹³C with age, which is in line with previous findings in different Populus deltoides × P. nigra genotypes ([38]). Similar positive correlation between BAI and Δ¹³C has been reported in Pinus radiata ([44]). The positive correlation between Δ¹³C and BAI suggests that the inter-annual variation of Δ¹³C is controlled to a larger extent by stomatal conductance than by photosynthetic capacity ([21], [48]). In poplar, mostly variation in stomatal conductance rather than photosynthetic capacity seems to control carbon gain and growth ([8], [30], [40]). Therefore, based on the significant positive correlation between BAI and Δ¹³C across tested genotypes, we may conclude that the observed increase in Δ¹³C with age and interannual variation of Δ¹³C was largely controlled by the variation in stomatal conductance rather photosynthetic capacity.

Stability of the genotypic ranking for Δ¹³C

Checking genotype rank stability with age is of central importance for selecting genotypes for higher water use efficiency. Therefore, genotype ranking in Populus deltoides × P. nigra was tested for Δ¹³C in controlled vs. open field conditions ([29]) and subsequently, in well irrigated vs. water stress under field conditions ([30]). In both studies, genotypic ranking remained stable with no correlation found between Δ¹³C and productivity traits. However, these studies were done on young plants. Parallel to that, many previous studies evidenced an age effect on Δ¹³C and demonstrated that duration and extent of age effect is variable according to many species ([23]). In this context, genotype ranking made on young plants were susceptible to change with age. Our results showed that genotype ranking shuffled across first age class (4 years) to seventh age class (18 years), i.e., I-214 displayed highest and Koster the lowest Δ¹³C values in the first age class (4 years), whereas I-214 was ranking second in the seventh age class (18 years). In spite of rank shuffling with age, genotypic ranking for mean Δ¹³C values matched with that found by Marron et al. ([26]) and Monclus et al. ([29]), where Koster had the lowest Δ¹³C and Dorskamp had the highest, and with Rasheed et al. ([38]) in Begaar for Dorskamp and I-214. Thus, we may conclude that based on a relatively small number of cultivars, a stability of genotype ranking was observed with age in two cultivars, at least during the approx. 20 years from planting to harvest. This conclusion is based on a limited range of cultivars. Unfortunately, extension to a larger number of genotypes was impossible given the lack of suitable even-aged common garden plantation of poplar cultivars at suitable age, not to speak of common gardens with different tree ages.

  Conclusion 

In this study, Δ¹³C assessed from the year 2009 increased with tree age in the three genotypes, indicating a decrease in water use efficiency with age at whole tree level. Furthermore, inter-annual variation of Δ¹³C was found positively correlated to BAI in all genotypes which shows that Δ¹³C was largely controlled by stomatal conductance rather photosynthetic capacity. Significant genotypic effect was detected for mean Δ¹³C over the tested period. The genotypic ranking of Δ¹³C was: (i) maintained among the three Populus deltoides × P. nigra genotypes, despite of crossing detected with tree age; (ii) found consistent with the previous ranking for Δ¹³C. Finally, Koster genotype was found highly water use efficient and productive with age among the tested genotypes. However, further studies are required to test other genotypes as well.

  Acknowledgements 

The authors are very grateful to Cyril Buré (INRA, UMR EEF) for helping in arranging the field visits and in collecting samples together with the different technicians of CRPF. Christian Hossann and Claude Bréchet (INRA, UMR EEF), for spending time doing all isotope analyses. Finally, we thank WFJ Parsons (CEF) for English editing. We are thankful to two anonymous reviewers for their suggestions to improve the manuscript. The research was supported by the European Union FP7 Project “Novel Tree Breeding Strategies” Project no. FP7 - 211868. F. Rasheed was supported by a grant from the Higher Education Commission of Pakistan.

  References

  Supplementary Material

Tab. S1 - Complete list of 376 tree cores sampled for three genotypes (Dorskamp, I-214 and Koster) under each age class along with information of each site condition (Water available, WA and Water limiting, WL) and their GPS coordinates.
Tab. S2 - Coefficients of predictor variable for Carbon, Nitrogen and along with variance explained by sites, which was taken as random factor.
Tab. S3 - Coefficients of predictor variable for Nitrogen along with variance explained by sites, which was taken as random factor.
Tab. S4 - Coefficients of predictor variable for Carbon isotope discrimination along with variance explained by sites which was taken as random factor.