Carbon isotope compositions (δ13C) of leaf, wood and holocellulose differ among genotypes of poplar and between previous land uses in a short-rotation biomass plantation

Authors


Abstract

The efficiency of water use to produce biomass is a key trait in designing sustainable bioenergy-devoted systems. We characterized variations in the carbon isotope composition (δ13C) of leaves, current year wood and holocellulose (as proxies for water use efficiency, WUE) among six poplar genotypes in a short-rotation plantation. Values of δ13Cwood and δ13Cholocellulose were tightly and positively correlated, but the offset varied significantly among genotypes (0.79–1.01‰). Leaf phenology was strongly correlated with δ13C, and genotypes with a longer growing season showed a higher WUE. In contrast, traits related to growth and carbon uptake were poorly linked to δ13C. Trees growing on former pasture with higher N-availability displayed higher δ13C as compared with trees growing on former cropland. The positive relationships between δ13Cleaf and leaf N suggested that spatial variations in WUE over the plantation were mainly driven by an N-related effect on photosynthetic capacities. The very coherent genotype ranking obtained with δ13C in the different tree compartments has some practical outreach. Because WUE remains largely uncoupled from growth in poplar plantations, there is potential to identify genotypes with satisfactory growth and higher WUE.

Introduction

In face of the current concerns about climate change and energy security, short-rotation plantations have gained considerable interest as a promising and relatively inexpensive source of bioenergy (Njakou Djomo et al. 2011, 2013). These plantations are typically managed at high planting densities (∼1000–40 000 ha−1) and harvested on a 2–10 year rotation, for a total lifespan of up to 30 years. Among suitable hardwood species, poplar (Populus spp.) is commonly used in Europe (Herve & Ceulemans 1996; Aylott et al. 2008), primarily because of its high juvenile productivity, re-sprouting ability and easy vegetative propagation (Dillen et al. 2011). Typical yields of poplar short-rotation plantations fall in the range of 8–10 oven-dry tons ha−1 year−1 (odt ha−1 year−1) (see King et al. 2013 for a review), but maxima of more than 15 odt ha−1 year−1 were already reported (Scarascia-Mugnozza et al. 1997; Liberloo et al. 2006). However, because of its riparian origin and pioneering behaviour, the fast growth of poplar is intimately linked to relatively high transpiration rates (Hall et al. 1998; Allen et al. 1999). This has raised questions about the large-scale deployment of such plantations (Perry et al. 2001; Lasch et al. 2010). Understanding the physiological and environmental controls on the efficiency with which water is used to produce biomass is therefore central if we are to achieve productive, reliable and sustainable systems while balancing the need for water with other uses (King et al. 2013).

Water use efficiency (WUE) describes the rate of CO2 uptake or plant dry matter production per unit plant water loss. Therefore, WUE stands as a crucial link between carbon and water fluxes, and as such can be approached on different biological and time scales ranging from the leaf to the individual and/or to the ecosystem (e.g. Niu et al. 2011; Rasheed et al. 2013). Because of its integrative nature, the stable carbon isotope ratio (δ13C) of plant organic matter has long been used to assess variations in plant WUE (Dawson et al. 2002; Cernusak et al. 2013). The δ13C of leaves (δ13Cleaf) is predicted to be linearly and negatively related to intrinsic WUE (WUEi; the ratio of leaf net CO2 assimilation rate to stomatal conductance) (Farquhar et al. 1982). This relationship has been confirmed experimentally in both agricultural (Hubick & Farquhar 1989) and woody species (Ripullone et al. 2004; Fichot et al. 2011). There is, however, now evidence that factors as variation in mesophyll resistance (Warren & Adams 2006), respiratory processes or post-photosynthetic fractionation (Werner et al. 2012) can affect the δ13C–WUE relationship. This does not necessarily result in a 1:1 correspondence. δ13C values as proxies for WUE should therefore be used with caution and confirmed by gas exchange measurements.

The isotopic signature of plant material and WUE are under both genetic and environmental control (Cernusak et al. 2013). Marked differences have been identified among C3 woody species (e.g. Cernusak et al. 2007a, 2008). External factors influencing photosynthesis and transpiration can also readily affect δ13C values, the best documented being water (e.g. Diefendorf et al. 2010) and nutrient (e.g. Livingston et al. 1999; Cernusak et al. 2007b) availability.

Growth and δ13C are controlled by numerous underlying physiological and morphological drivers, for example, leaf net CO2 assimilation rate, transpiration or carbon allocation among tree compartments. Studying the interdependence between δ13C and growth can therefore provide insight into which drivers are most important to both processes. In addition, for species of interest for plantations and agroforestry such as poplar, a lack of trade-off (i.e. no antagonism between traits) is of particular interest as it may be seen as an opportunity for optimizing tree WUE when selecting for productive genotypes (Monclus et al. 2006). Most studies relating δ13C to growth performance have used δ13Cleaf. This could be problematic if the time frame considered for growth differs from the course of leaf tissue synthesis and it might partly explain the variable relationships observed (Toillon et al. 2013). The δ13C recorded in tree rings may provide valuable and complementary information considering the longer time scale of integration.

In contrast to δ13Cleaf that provides a proxy for photosynthetic activity over the limited course of leaf formation, the 13C signal recorded in tree rings can provide an estimation of WUE weighed over the period of radial growth (McCarroll & Loader 2004). Wood is, however, a composite tissue comprising several chemical components, primarily cellulose, hemicelluloses and lignin. Each of these components has a distinct isotopic signature because of their specific biosynthesis processes (Wilson & Grinsted 1977; McCarroll & Loader 2004). Therefore, differences in the proportion of different wood constituents may blur the signal when comparing contrasting individuals. Isolating a single component of the wood can in turn remove this variation, and cellulose is usually preferred (Loader et al. 2003). Cellulose generally reflects the δ13C of photosynthetically fixed CO2, although post-photosynthetic fractionations may alter the δ13C of primary photosynthates when converted into cellulose (Gessler et al. 2008; Tcherkez et al. 2011). Cellulose also has the advantage of being deposited during the year of formation only, contrary to lignins.

In this study, we determined the 13C isotopic signature of different compartments (bulk leaf, bulk wood and holocellulose extracted from bulk wood) of six poplar genotypes grown for biomass. We investigated how this relates to other traits of carbon uptake and of growth performance [including leaf structure, canopy development, growth phenology, relative growth rate (RGR) and above-ground biomass]. To our knowledge, this is the first in-depth study in poplar considering different scales of measurements for both δ13C and growth. Measurements were performed in the largest short-rotation bioenergy plantation installed in the Benelux (18.4 ha) during the second growing season of the first rotation. Interestingly, this plantation was set-up on a set of parcels with two different former land uses (cropland versus pasture) already known to differ in soil nutrient concentrations (Broeckx et al. 2012; Verlinden et al. 2013b). Therefore, the questions addressed were as follows: (1) How do the δ13C signals measured in different plant compartments (δ13Cleaf, δ13Cwood and δ13Cholocellulose) relate mutually? (2) Does the relationship between δ13C and growth-related traits depend on the time scale of integration and the compartment considered? (3) Does former land use influence 13C signals of different compartments and does it impact the ranking of genotypes for 13C in different compartments?

Materials and Methods

Site description and plantation design

The operational POPFULL (Ceulemans 2010) site is located in Lochristi, province East Flanders, Belgium (51°06′44′N, 3°51′02′E) in a flat landscape at an elevation of 6.25 m above sea level. The site is subjected to an oceanic climate, with a mean annual temperature of 9.5 °C and a mean annual precipitation of 726 mm (Royal Meteorological Institute of Belgium). The 18.4 ha site was a former farmland consisting of an adjacent set of parcels of pasture and croplands with corn (Zea mays L.) as the most recent cultivated crop in rotation. During previous crop production, fertilization was applied at a rate of 200–300 kg ha−1 year−1 of nitrogen (N) as liquid animal manure and chemical fertilizers. An extensive soil survey before plantation establishment revealed that land use type influenced the upper soil layer composition (up to 15 cm): C and N mass fractions were significantly higher and bulk density was significantly lower in pasture as compared with cropland (see Broeckx et al. 2012; Verlinden et al. 2013b). According to the Belgian soil classification, the area is situated in the sandy soil region of Flanders with poor natural drainage (Van Ranst & Sys 2000). This was evidenced from granular analyses, which characterized the soil as a sandy texture and a clay-enriched deeper soil layer at 75 cm (Broeckx et al. 2012).

On 7–10 April 2010 an area of 14.5 ha (excluding the headlands) was planted with 12 selected poplar (Populus spp.) genotypes, representing different species and hybrids of Populus deltoides Bartr. (ex Marsh.), P. maximowiczii Henry, P. nigra L. and P. trichocarpa Torr. & Gray (ex Hook.). The plantation was designed such that each genotype was represented by two to four (replicate) monoclonal blocks of eight double rows wide, with block lengths varying from 90 to 340 m. Dormant and unrooted cuttings were planted in a double-row planting scheme with alternating inter-row spacings of 0.75 and 1.50 m and a mean distance of 1.10 m between trees within a row, yielding a planting density of ± 8000 trees ha−1. More details on site conditions, on poplar materials and on the plantation layout can be found in Broeckx et al. (2012) and Verlinden et al. (2013a).

Measurements

For this study, six poplar genotypes were chosen to encompass the different parentages present in the plantation, that is, Bakan, Skado (both P. trichocarpa × P. maximowiczii, T × M), Wolterson (pure P. nigra, N), Koster, Oudenberg (both P. deltoides × P. nigra, D × N) and Grimminge [P. deltoides × (P. trichocarpa × P. deltoides), D × (T × D)]. For each genotype, measurements were performed at eight sampling plots located within the monoclonal blocks and spread over the plantation area (6 genotypes × 4 plots × 2 land use types). All measurements were performed during the second growing season of the first 2 year rotation (2011); all trees were therefore single stems.

Leaf traits

Leaf gas exchange was measured repeatedly during the 2011 growing season in seven measurement campaigns to span most of the growing season: 4–6 May, 18–20 May, 4–8 July, 27–29 July, 16–19 August, 5–9 September, 26–30 September 2011. For the six genotypes, measurements were performed on the same four replicate trees with a LI-6400 open path photosynthesis system (Li-Cor, Lincoln, NE, USA) equipped with a leaf chamber fluorometer (LI-6400-40; Li-Cor). Measurements were taken in the upper canopy, on the first fully mature sunlit leaf of the current year main axis. All measurements were performed at a constant block temperature (25 °C) and at a controlled leaf-to-air vapour pressure deficit (VPD) close to 1 kPa (1.2 ± 0.04, mean ± SE). Leaves were first acclimated for 10 min in the chamber at a CO2 concentration of 400 ppm and under a photosynthetic photon flux density (PPFD) of 1500 μmol m−2 s−1. Preliminary test experiments showed that this PPFD was enough to ensure saturating light conditions for all genotypes. Light-saturated assimilation rate at atmospheric CO2 concentration (Asat, μmol m−2 s−1) and stomatal conductance (gs-sat, mol m−2 s−1) were then recorded. WUEi (mmolCO2 molH2O−1) was calculated as the ratio between Asat and gs-sat. The leaves sampled were oven dried for 24 h at 60 °C and ground to a fine powder for the determination of bulk leaf carbon isotope composition (δ13Cleaf).

Other leaf traits were determined in early August 2011 for five randomly selected trees per plot (n = 40 per genotype, 240 in total). For each tree, the first fully mature leaf on the main stem was collected. The individual leaf area was determined by scanning the lamina and processing the scans with the ImageJ software (Rasband 1997, Bethesda, MD, USA). Depending on the size of the leaf, 4–12 leaf discs were subsampled from the lamina with a punch of 20 mm in diameter, avoiding the lamina midrib. Leaf discs and petioles were then oven dried for 24 h at 60 °C to determine the specific leaf area (SLA). SLA was calculated by dividing the area of the n discs per leaf by their total dry mass. Subsequently, the oven-dried leaf discs were ground to a fine powder for the determination of C and N concentrations and for the analysis of δ13Cleaf. The N concentration of the same samples used for δ13C analyses was obtained with a Flash elemental analyser (Thermo Fisher Scientific, Bremen, Germany) and was expressed on an area basis (NA, g cm−2).

Wood traits

Wood samples were collected at the end of the 2011 growing season during the dormant period (January 2012) on the same trees as those previously sampled for δ13Cleaf determination (n = 40 per genotype). Approximately 2 cm long samples were cut at the base of the current year (2011) main stem and used for the assessment of the carbon isotope composition of wood and holocellulose (δ13Cwood and δ13Cholocellulose, respectively). Bark and pith were first removed before the samples were oven dried at 60 °C and subsequently ground milled. The extraction of holocellulose was carried out using 100 mg of wood powder following a slightly modified protocol of Brendel et al. (2000). Briefly, the material was transferred to autoclavable borosilicate USP type 1 V-vials with PTFE-coated screw caps (Supelco/Sigma-Aldrich, Bellefonte, PA, USA), and gently suspended in 2 mL 80% acetic acid and 0.2 mL 69% nitric acid. Samples were gently mixed and subsequently boiled at 120 °C for 20 min in a vertical tabletop autoclave (PBI International, Milan, Italy). After cooling to room temperature, 2.5 mL 99% ethanol was added and the samples were briefly vortexed and let to settle for 15 min. The pellet was washed twice with 5 mL 99% ethanol, twice with 5 mL de-ionized water, once with 4 mL ethanol and finally once with 4 mL acetone. All samples were dried in an oven at 50 °C. The extraction of holocellulose was confirmed by micro-pyrolysis.

Carbon isotope analysis

All carbon isotope analyses were performed at the Stable Isotope Laboratory of the James Hutton Institute (Invergowrie, Dundee, UK). Subsamples of ground material (leaf, wood and holocellulose) were enclosed in tin capsules and combusted in a Delta V isotope ratio mass spectrometer (IRMS) (Thermo Fisher Scientific). The CO2 produced by combustion was separated and its 13CO2/12CO2 ratio was analysed by the IRMS. Scale normalization of the measured δ13C values followed the internationally accepted IAEA standard procedure using two international standard reference materials as scale anchors. The carbon isotope composition (δ13C, ‰) was expressed relative to the Pee Dee Belemnite (PDB) standard and was calculated as:

display math(1)

where Rsa and Rsd are the 13CO2/12CO2 ratios of the sample and the standard, respectively (Farquhar et al. 1989). The accuracy of the measurements was assessed by repeated measurements of laboratory standards and was ± 0.11‰ (standard deviation).

Growth measurements, leaf area index and bud phenology

On 15 July 2011 (3 weeks before leaf sampling), the stem diameter at 22 cm height was measured for the 240 trees using a digital caliper (CD-15DC; Mitutoyo, Aurora, IL, USA; 0.01 mm precision). In addition, the fifth leaf (foliar rank counting from the first top leaf exceeding 20 mm in length) of each tree was marked with a label. Three weeks later, diameter measurements were repeated. The number of newly produced leaves was counted from the position of the label, and the leaf increment rate was calculated as the number of leaves produced per day. Diameter measurements at the two dates were converted to above-ground woody biomass estimates using genotype-specific allometric relationships relating stem diameter to above-ground woody tree biomass (see details in Broeckx et al. 2013). The biomass estimates at the two dates were then used to compute the RGR (g kg−1 day−1) of the above-ground woody tree compartment as:

display math(2)

where m2 and m1 represent the estimated dry masses at dates 2 and 1 (t2 and t1, respectively) (Hoffmann & Poorter 2002). Diameter measurements were finally repeated at the end of the growing season to estimate the final above-ground woody biomass.

To monitor canopy development, leaf area index (LAI) was measured in each plot monthly throughout the growing season from April to November (Broeckx et al. 2013; Verlinden et al. 2013a). The LAI-2200 Plant Canopy Analyzer (Li-Cor Biosciences) was used to measure LAI indirectly by comparison of above- and below-canopy readings with a 45° view cap. The LAI measured closest to the date of leaf sampling corresponded to the maximal LAI of the growing season (LAImax). Leaf area duration (LAD) was calculated as the area below the 2011 seasonal LAI curve per plot by integration over time.

The phenological onset and ending of the growing season were monitored by observing the apical buds of four trees per plot during spring and autumn 2011 (meteorological data in Supporting Information Fig. S1). The timing of spring bud flush was defined at a stage according to ‘Bud sprouting, with a tip of the small leaves emerging out of the bud scales, which couldn't be observed individually’ (based on UPOV 1981). The timing of bud set, accompanied by the end of leaf production, was set at the time when the ‘apical bud was present but not fully closed, bud scales were predominantly green and no more rolled-up leaves were present’ (Rohde et al. 2010). Subsequently, the length of the growing season was defined as the period in between these two phenological stages.

Soil analysis

Soil characteristics were estimated at each of the 48 measuring plots. In December 2011, soil samples were taken up to 30 cm depth using a handmade gouge auger. At each plot, six samples (3 cm in diameter) spatially distributed over the 5.5 × 6.75 m plot area were taken. The six samples were blent and treated further as a mixed sample per plot. Preceding the analyses, the soil was oven dried for 3 d at 70 °C. Total C and N mass fractions were analysed by dry combustion (NC-2100; Carlo Erba Instruments, Milan, Italy). The pH was measured with a glass electrode in a KCl solution. Mass fractions of phosphorus (P) and potassium (K) were measured using inductively coupled plasma (ICP) analysis, based on the method of Egnér et al. (1960).

Statistical analyses

A Kolmogorov–Smirnov test was applied for testing the data for normality. The influence of genotype, former land use type and plot on the measured traits was then examined using a nested model of analysis of variance (ANOVA), with genotype and land use type as fixed factors, and plot (nested within genotype) as a random factor. When a significant genotype effect was observed, the Tukey's HSD post hoc test was further used to discriminate genotypes. Relationships between pairs of continuous variables were examined using linear regression analysis and Pearson's correlation coefficients (r). The slope of the relationship between δ13Cwood and δ13Cholocellulose was tested against the test value of 1 using the ‘smatr’ package in the R software (version 3.0.1, a language and environment for statistical computing and graphics). The homogeneity of slopes among genotypes was further tested using analysis of covariance where δ13Cholocellulose was considered as the covariate, δ13Cwood as the dependent variable and genotype as the fixed factor. δ13Cwood and δ13Cleaf were compared in a paired t-test. Statistical tests were considered significant at P ≤ 0.050. Otherwise specified, statistical analyses were performed in SPSS (version 20; SPSS Inc., Chicago, IL, USA) and Microsoft Excel 2010 (version 14.0.6112.5000).

Results

δ13Cholocellulose versus δ13Cwood

As expected, the values of δ13Cholocellulose were higher than those of δ13Cwood, but both were strongly correlated (r = 0.948, P < 0.0001) (Fig. 1). When all genotypes were combined, the slope of the relationship was not significantly different from the 1:1 line (P = 0.228), with an average offset of δ13C between holocellulose and bulk wood of 0.92 ± 0.29‰ (Fig. 1). At the genotype level, the relationships were all significant (P < 0.002). Analysis of covariance revealed that there was no significant interaction with genotype (P = 0.115), indicating that the slopes did not differ significantly among the six genotypes studied. There were, however, significant differences in the δ13C offset among genotypes (Table 1). The fraction of holocellulose extracted from bulk wood also varied significantly among genotypes from 33.6 to 37.7% for the genotypes Wolterson and Skado, respectively (Table 1). However, these differences did not appear to explain the variations observed in the δ13C offset.

Figure 1.

Relationship between carbon isotope composition of bulk wood (δ13Cwood) and extracted holocellulose (δ13Cholocellulose) for the six poplar genotypes. The dashed line represents the 1:1 line. The full line represents the linear regression on the pooled dataset (n = 240), given with the Pearson's linear correlation coefficient (r) and the P-value. Statistical analyses indicated that the slope of the relationship was not significantly different from 1. The offset (±standard deviation) refers to the average difference between δ13Cholocellulose and δ13Cwood.

Table 1. Significance (P-values) of the mixed model analysis of variance for the different physiological traits measured on the six poplar genotypes
 StatisticsGenotypeLand use type
(P-value ANOVA)
Genotype effectLand use effectInteractionSkadoBakanWoltersonKosterOudenbergGrimmingeCroplandPasture
  1. Land use type and genotype were considered as fixed factors. Mean values are given with standard deviation between brackets. P-values indicating a significant effect (P < 0.05) are highlighted in bold. Different capital letters in superscript indicate significant differences (P < 0.05) following Tukey's post hoc test.
  2. aLAImax and LAD were sampled at the plot level (n = 8 per genotype; n = 24 per land use type); other variables were measured at the tree level (n = 40 per genotype; n = 120 per land use type).
  3. ANOVA, analysis of variance; LAD, seasonal leaf area duration; LAImax, maximal seasonal leaf area index; N, area-based nitrogen content; offset, difference between δ13Cholocellulose and δ13Cwood; RGR, relative growth rate; SLA, specific leaf area.
Carbon isotope composition            
δ13Cleaf<0.001<0.0010.089−26.81 (0.37)A−27.05 (0.66)A,B−27.56 (0.47)B−28.34 (0.76)C−26.96 (0.50)A,B−28.36 (0.95)C−27.83 (0.97)−27.19 (0.69)
δ13Cwood<0.001<0.0010.703−26.69 (0.53)A−26.98 (0.61)A−27.77 (0.37)B−27.92 (0.62)B−27.98 (0.47)B−28.68 (0.62)C−27.93 (0.86)−27.41 (0.77)
δ13Cholocellulose<0.001<0.0010.148−25.85 (0.46)A−25.97 (0.66)A−26.92 (0.31)B−26.94 (0.59)B−26.98 (0.46)B−27.75 (0.61)C−26.91 (0.87)−26.51 (0.73)
Offset (δ13Cholocellulose – δ13Cwood)0.0410.0410.0190.79 (0.31)A1.01 (0.20)C0.83 (0.37)A,B0.95 (0.28)A,B,C1.00 (0.29)B,C0.94 (0.22)A,B,C0.97 (0.27)0.87 (0.31)
Leaf            
Leaf increment rate# day−1<0.0010.6690.0030.327 (0.041)B0.311 (0.050)B0.352 (0.065)B0.436 (0.051)A0.422 (0.054)A0.332 (0.043)B0.365 (0.062)0.362 (0.077)
SLAm2 kg−10.0150.5830.03411.86 (1.20)A,B11.96 (1.46)A,B11.51 (1.29)A,B12.92 (1.90)A11.38 (1.10)B12.57 (1.49)A,B12.11 (1.62)11.95 (1.42)
Individual leaf areacm2<0.0010.8200.310404 (105)A402 (141)A63 (8)B89 (27)C113 (40)C235 (79)B215 (155)219 (172)
Leaf Ng m−2<0.001<0.0010.0361.89 (0.40)B1.59 (0.51)B2.60 (0.32)A1.58 (0.43)B2.03 (0.61)B1.64 (0.50)B1.62 (0.58)2.16 (0.46)
LAImaxam2 m−2<0.0010.9650.1113.34 (0.79)A1.94 (0.60)B,C1.57 (0.55)C2.63 (0.39)A,B2.26 (0.32)B,C2.77 (0.71)A,B2.42 (0.57)2.42 (1.00)
LADam2 day m−2<0.0010.5980.102560 (152)A298 (98)B,C183 (57)C370 (74)B290 (32)B,C378 (95)B353 (109)640 (177)
Wood            
RGRg kg−1 day−10.3100.0030.6516.07 (3.91)4.07 (1.73)6.43 (1.90)5.41 (2.65)5.63 (2.59)6.03 (3.38)4.66 (2.00)6.65 (3.29)
Final above-ground biomassg tree−1<0.0010.1330.1041891 (610)A1358 (423)A,B748 (292)C1191 (531)B,C1345 (638)A,B1493 (592)A,B1258 (510)1417 (715)
Ratio holocelluloseg (100 g)−1<0.0010.2750.40436.8 (3.2)A,B37.7 (3.1)A33.6 (3.2)C34.6 (2.7)B,C34.7 (2.4)B,C34.7 (2.9)B,C35.6 (2.7)35.0 (3.6)

There were significant differences among genotypes in terms of δ13Cwood and δ13Cholocellulose (P < 0.001; Table 1). Interestingly, the ranking of genotypic means was identical for δ13Cwood and δ13Cholocellulose, and was dependent on parentage (Fig. 2; Table 1). Both T × M hybrids (Skado and Bakan) showed the least negative δ13C values with no significant difference between them. The pure P. nigra genotype (Wolterson) was ranked third, closely followed by and not significantly different from the D × N genotypes (Koster and Oudenberg). The backcrossed D × (T × D) genotype (Grimminge) exhibited the most negative δ13C values, being ranked sixth and significantly different from the five other genotypes (Fig. 2; Table 1).

Figure 2.

Box plots representing biological variation in δ13C across clonal replicates (n = 40) for holocellulose (left panel), bulk wood (middle panel) and August leaves (right panel) for the six poplar genotypes. Each box represents the quartile below (Q1) and above (Q3) the median value. Vertical bars represent minimum and maximum values except when the latter are away 1.5 times from the top of the interquartile (Q3–Q1) range. Values beyond this range (outliers) are represented as circles. The dashed lines within the boxes represent the mean values; the full lines represent the median values. Numbers below the boxes show the genotype ranking based on the mean values. Different letters above the boxes indicate significant differences between genotypes following Tukey's post hoc test. Parentages are given above the genotype names. T = Populus trichocarpa, M = P. maximowiczii, N = P. nigra, D = P. deltoides.

δ13Cleaf and how it is related to WUEi and δ13Cwood

WUEi, as interpreted by the ratio of Asat and gs-sat, was linearly correlated to δ13Cleaf for the six genotypes over the whole growing season (r = 0.592, P < 0.0001) (Fig. 3). As predicted by theory (Farquhar et al. 1989), this positive and significant relationship makes δ13C a potential surrogate of WUE. A positive and linear relationship was evidenced between δ13Cleaf and δ13Cwood (Fig. 4). This relationship was also significant for all six genotypes considered separately. There was, however, no consistency among genotypes neither for the slope of the relationship nor for the offset of δ13C between leaf and wood. As the slope of the pooled dataset was significantly smaller than 1 (Fig. 4, P < 0.001), five of the six genotypic relationships displayed slopes ranging from 0.44 to 0.70. The exception was genotype Grimminge with a slope of 1.24. For half of the genotypes (Bakan, Skado and Koster), δ13Cleaf was on average more negative than δ13Cwood, although only for Koster the difference was significant (P < 0.001). For the three other genotypes the opposite was observed, that is, δ13Cleaf was less negative than δ13Cwood (all P ≤ 0.010). The most prominent difference of 1.03‰ was recorded for genotype Oudenberg (Table 1; Fig. 2). As a consequence, the ranking of genotypic means for δ13Cleaf was slightly different as compared with that for wood. Genotype Oudenberg showed the secondly least negative values and shifted the ranks of the other genotypes (Fig. 2).

Figure 3.

Relationship between carbon isotope composition of leaves (δ13Cleaf) and intrinsic water use efficiency measured by gas exchange. Measurements were performed repeatedly during the 2011 growing season. The microclimate during the gas exchange measurements differed significantly from the ambient one inducing 13C composition in the leaves. Data points represent genotypic means at one date over four trees (n = 4). The full line represents the linear regression on the pooled dataset (n = 37), given with the Pearson's linear correlation coefficient (r) and the P-value.

Figure 4.

Relationship between δ13Cleaf and δ13Cwood for the six poplar genotypes. The dashed line represents the 1:1 line. The full line represents the linear regression of the pooled dataset (n = 240), with the Pearson correlation coefficient (r) and its significance level. Pearson r values for correlations per genotype are also given in the legend. Significance levels: ***P < 0.001; **P < 0.010; *P < 0.050.

Relationships with (soil) nutrients, previous land use and growth parameters

Total soil N concentrations in the upper 30 cm layer recorded at the specific measurement plots were significantly higher in pasture than in cropland. This confirmed the findings of a previous soil survey conducted over the complete plantation (Broeckx et al. 2012; Verlinden et al. 2013b). Potassium concentrations were significantly lower in pasture than in cropland (Table 2) while pH, C and P concentrations were not different between previous land use types (Table 2). Values of δ13Cleaf, δ13Cwood and δ13Cholocellulose were significantly higher for the trees grown in previous pasture than for those grown in previous cropland for about 0.5‰ (Table 1). Besides, former land use significantly affected the leaf N contents. Trees growing on previous pasture displayed higher leaf N concentrations as compared with trees growing on former cropland (2.16 versus 1.62 g m−2, respectively) (Table 1). The effect was however genotype dependent (Table 1). There was no significant effect of former land use for other leaf traits, but significant genotype × land use interactions were recorded in most cases (Table 1). Following the leaf N pattern, RGR was significantly higher for trees grown in previous pasture (6.65 g kg−1 day−1) than for those grown in cropland (4.66 g kg−1 day−1) (Table 1). Although land use affected several traits, no significant correlations were found between individual soil characteristics and other tree-related traits at the plot level.

Table 2. Mean values with standard deviation between brackets of selected soil characteristics and significance (P-values) of the difference between previous land use types
  StatisticsCroplandPasture
(P-value)
  1. Soil was sampled in the upper 30 cm layer at the plot level (n = 24 per land use type). P-values indicating a significant effect (P < 0.05) are highlighted in bold.
  2. C, carbon; K, potassium; N, nitrogen; P, phosphorus.
pH 0.5285.33(0.45)5.24(0.62)
Kmg (100 g)−10.00413.88(3.10)10.63(4.84)
Pmg (100 g)−10.07425.62(7.78)22.46(6.14)
Cg (100 g)−10.4151.64(0.23)1.57(0.31)
Ng (100 g)−10.0330.171(0.017)0.182(0.017)

Values of δ13Cleaf were significantly and positively correlated with leaf N, both for the pooled dataset and the six genotypes separately (Fig. 5). The same trend was observed for δ13Cwood (r = 0.271, P < 0.001 for the pooled dataset) and δ13Cholocellulose (r = 0.239, P < 0.001 for the pooled dataset). This suggested a higher WUEi with higher N concentrations in leaves. When all genotypes were combined, the final above-ground biomass was positively, but weakly, correlated with δ13Cwood and δ13Cleaf (Table 3; Supporting Information Fig. S2). At the genotype level, the correlation was significant for only two genotypes (Oudenberg and Skado). Relationships between δ13C values and other traits related to biomass production were overall weak and variable too (Table 3; Supporting Information Fig. S2). Relationships with leaf increment rate were only significant when pooling all data (r = −0.14, P < 0.05 and r = −0.22, P < 0.01 for δ13Cleaf and δ13Cwood, respectively). No correlation was detected with LAImax, neither at the genotype level nor when pooling data (Table 3; Supporting Information Fig. S2), and the same was observed for LAD (data not shown). In contrast, bud phenology indices and growing season length were strongly related to δ13Cleaf and δ13Cwood (Fig. 6). Less negative δ13C values were observed for genotypes starting their growing season earlier as well as for those ending their growing season later (Fig. 6). This resulted in a significant and positive correlation (P < 0.001) between the length of the growing season and δ13C values (Fig. 6).

Figure 5.

Relationship between δ13Cleaf and leaf area-based nitrogen content for the six poplar genotypes. The lines represent the linear regressions per genotype (n = 40), with Pearson correlation coefficients (r) given in the legend. Significance levels: ***P < 0.001; **P < 0.010; *P < 0.050. The correlation of the pooled dataset (n = 240) was also significant (r = 0.432 and P < 0.0001, not shown in the graph). Inset panel shows the relationship between δ13Cleaf and mass-based nitrogen.

Table 3. Correlations of stable carbon isotopes ratios in leaf (δ13Cleaf) or wood (δ13Cwood) with growth-related traits, performed on the pooled data and per genotype
  Individual leaf areaSLARGRJuly–AugustFinal above-ground biomassLAImax*
  1. Significant correlations (P < 0.05) are highlighted in bold. Correlations performed at plot level for parameters indicated with*. The P-values of the correlations are given below the Pearson r coefficients.
  2. δ13C, carbon isotope ratio; final aboveground biomass, standing above-ground woody dry mass after the growing season; LAImax, maximal leaf area index (in the period of leaf sampling); RGRJuly−August, relative growth rate between July and August; SLA, specific leaf area.
δ13Cleaf
 All pooledr = 0.322−0.2750.1230.2520.025
P < 0.001<0.0010.062<0.0010.866
 Skado0.478−0.0070.4860.4460.164
0.0020.9680.0020.0040.697
 Bakan−0.2140.3250.168−0.039−0.497
0.1910.0430.3210.8120.210
 Wolterson−0.243−0.285−0.2910.2680.363
0.1310.0740.0720.0940.376
 Koster0.532−0.5570.3200.1080.450
<0.001<0.0010.0470.5080.264
 Oudenberg0.4660.1470.1600.3400.312
0.0020.3660.3240.0320.452
 Grimminge0.1380.0070.3360.2720.192
0.4020.9680.0390.0950.648
δ13Cwood
 All pooled0.484−0.1430.1160.2690.008
<0.0010.0280.079<0.0010.956
 Skado0.277−0.0910.6190.351−0.227
0.0920.586<0.0010.0310.588
 Bakan−0.005−0.120−0.0240.138−0.657
0.9760.4660.6890.3950.077
 Wolterson−0.206−0.015−0.118−0.100−0.508
0.2080.9290.4760.5440.199
 Koster0.423−0.3350.1110.136−0.070
0.0070.0270.5020.4100.869
 Oudenberg0.4650.3950.3770.4930.562
0.0020.0120.0160.0010.147
 Grimminge0.0950.0790.4650.1760.011
0.5660.6310.0030.2780.980
Figure 6.

Relationship between δ13Cwood and phenology parameters (dates of bud burst and bud set, and growing season length). The Pearson correlation coefficient (r) is given with the significance of the linear regression for the pooled data at the plot level (n = 48).

Discussion

Relationships between δ13Cwood, δ13Cholocellulose and δ13Cleaf

Because of the presence of lignin in bulk wood, which is 3–5‰ more depleted in 13C than cellulose (Benner et al. 1987; Loader et al. 2003), the offset in δ13C between bulk tissue and cellulose is generally in the range of 0.5–2.0‰ (Borella et al. 1998; Loader et al. 2003; Verheyden et al. 2005; Harlow et al. 2006; Rasheed et al. 2011; Szymczak et al. 2011). The average δ13C offset of 0.92‰ found in our study fitted within this range, although near the lower end. The fact that the cellulose extracted in our study corresponded to holocellulose, not to purified α-cellulose, may partly explain this.

The δ13C offset between whole-wood and holocellulose was not constant across the six genotypes, genotypic averages falling within the range of 0.79–1.01‰. Although these differences were not sufficient to alter the genotype ranking, our results suggest that δ13Cwood may not provide a reliable and consistent signal for inter-genotype comparisons, even for juvenile trees such as those studied here. This is only partly confirmatory of the findings reported by Rasheed et al. (2011) on several hybrid poplars. In their study, the δ13C offset significantly differed among genotypes at only one study site while no difference was detected at the two other sites. However, their data further indicated that there was a significant temporal trend with ageing in the offset of several genotypes. This reinforces the idea that cellulose extraction may be a prerequisite in poplar when one wants to draw conclusions on genotypic differences in WUEi from wood, even from closely related genotypes.

The most plausible reason for the differences observed in the δ13C offset relates to genotypic variations in the relative proportion of the different wood constituents. Surprisingly, genotypic differences in the δ13C offset did not follow any obvious parentage pattern. This was clearly visible from the genotypes Bakan and Skado (full-sibs, T × M parentage), which had the highest and the lowest offset values, respectively. It is known that substantial variation in wood composition and properties exists among poplar genotypes, including lignin and cellulose contents (Studer et al. 2011; Porth et al. 2013).

Values of δ13Cleaf were significantly and positively related to δ13Cwood (and thus δ13Cholocellulose). This suggested that overall a non-negligible part of the 13C signal carried by the whole wood was actually captured by leaves sampled in August. A similar relationship has already been reported across different woody species (Panek & Waring 1997; Guehl et al. 1998; Schulze et al. 2006), but to the best of our knowledge this is the first time reported in poplar. On average, wood was 13C enriched as compared with leaves in only three genotypes (Bakan, Skado and Koster; positive offset from 0.08 to 0.44‰). This pattern is consistent with literature surveys that have revealed that heterotrophic tissues from C3 plants are generally 13C enriched as compared with autotrophic tissues because of the existence of post-photosynthetic fractionation processes (see Badeck et al. 2005 and Cernusak et al. 2009 for reviews). In contrast, the opposite was observed in the other three genotypes (Oudenberg, Grimminge and Wolterson; negative offset from −0.21 up to −1.03‰). For these same genotypes, the weakest correlation coefficients between δ13Cwood and δ13Cleaf were recorded. The δ13C imprinted on August leaves was therefore less representative of the seasonal integrated WUEi, for these three genotypes. Differences in leaf developmental stage (Cernusak et al. 2009) are unlikely to explain this as all leaves sampled for all genotypes were sunlit mature leaves originating from the current year main stem. However, besides integrating the growing season, the 13C signal recorded in wood also carries a weighted average of the whole canopy. Genotypic differences in canopy architecture and relative contributions of main leaves versus branch leaves in the photosynthetic process may thus be one possibility.

Our results indicate that some caution should be taken when drawing conclusions on genotypic differences in WUEi from the δ13C signals of different compartments. However, δ13Cwood and δ13Cholocellulose were still efficient at discriminating the different poplar parentages. δ13Cleaf values proved to be less relevant, mainly because of the discrepancy observed between δ13Cleaf and δ13Cwood for genotype Oudenberg. Based on the δ13C values recorded, the two T × M genotypes (Bakan and Skado) clearly displayed the highest WUE integrated over the season, followed by the N (Wolterson) and the D × N (Koster and Oudenberg) genotypes. Finally, the lowest WUE integrated over the season was observed for the D × (T × D) (Grimminge) genotype.

Relationship between growth and δ13C

Overall, relationships between δ13C estimates and final above-ground biomass production or other growth-related traits were weak and highly variable. However, it is important to note that no trade-off could be observed, confirming previous findings reported for P. deltoides × P. nigra (Monclus et al. 2005, 2006; Voltas et al. 2006; Bonhomme et al. 2008; Fichot et al. 2010; Toillon et al. 2013), P. deltoides × P. trichocarpa (Rae et al. 2004; Monclus et al. 2009) or P. nigra (Chamaillard et al. 2011). Interestingly, the use of δ13Cwood or δ13Cholocellulose (which integrate the whole growing season) as a complement to δ13Cleaf (which integrates several weeks) did not fundamentally change the observed patterns as both were partly related (see above discussion). Therefore, we conclude that growth performance and WUE were to a large extent uncoupled in our study. Whether this will remain true with tree ageing remains to be analysed. Recent data on 15- to 17-year-old poplar genotypes have indicated that tree ring δ13C and growth based on annual basal area increment remained unrelated over time (Rasheed et al. 2011).

Although δ13C estimates were overall poorly correlated with traits directly related to growth and carbon uptake, strong correlations (r > 0.79) were observed with leaf phenology, that is, with bud burst and bud set dates, as well as with the length of the growing season. A longer growing season does not necessarily translate into a higher investment in growth in poplars (Toillon et al. 2013), and this was also observed in our study. Although the two most productive genotypes (Bakan and Skado, both T × M) displayed the longest growing season, both traits were indeed partly uncoupled in the other four genotypes (see Verlinden et al. 2013a). This probably explained why phenology correlated better with δ13C estimates than growth parameters per se. The relationship between leaf phenological events and WUE may be informative for strategies regarding resource use efficiency. The patterns reported in the literature are however conflicting (Hall et al. 1994; Aletà et al. 2009). In our case, the genotypes with the earliest growth onset were also the most efficient to use water based on δ13C values. Because bud burst and bud set dates were closely, but negatively related, genotypes exhibiting a longer growing season displayed a higher WUE. The nature of the relationship between phenology and WUE remains, however, correlative and is based on a limited number of genotypes, all the more with different parentage. Thus, it is likely that the relationship observed was primarily driven by the contrasting life history of the different poplar species considered. Yet, from a practical perspective, the two genotypes Bakan and Skado warrant attention for short-rotation plantations given their extended growing season, high biomass production and high δ13C.

Effect of soil characteristics and previous land use type on δ13C

There was a clear effect of previous land use on δ13C estimates. Leaf and stem tissues were systematically more 13C enriched (0.5‰ on average) for trees growing on former pasture as compared with former cropland. Interestingly, soil analyses revealed a higher N mass fraction in pasture, in agreement with previous results obtained on the same plantation (Broeckx et al. 2012; Verlinden et al. 2013b). This suggested that increased N availability resulted in enhanced tree WUE. Nitrogen supply increases WUE (Lajtha & Whitford 1989; Harvey & van den Driessche 1999; Livingston et al. 1999; Ripullone et al. 2004), although opposite trends have also been reported (Guehl et al. 1995; Korol et al. 1999; Hubbard et al. 2004). This latter observation is most likely because of species-specific responses to nutritional status. In that respect, it is noteworthy that the influence of other elements, as, for example, Fe and Mg, might not be marginal. Variation in mineral absorption may thus affect the results. According to Ripullone et al. (2004), increased WUE in response to higher N availability may arise from (1) increased stomatal control without influence on assimilation rate (Toft et al. 1989); (2) increased assimilation rate because of increased N investment in the photosynthetic apparatus without effect on stomatal conductance (Liu & Dickmann 1996; Harvey & van den Driessche 1999; Welander & Ottosson 2000); or (3) a combination of both (Wang et al., 1998; Livingston et al. 1999). The positive relationship between leaf N content and δ13Cleaf observed for all genotypes suggests that spatial differences in tree WUE among land use types were mainly mediated by increased photosynthetic capacities. A similar relationship between leaf N and δ13C has already been reported in poplar (Ripullone et al. 2004; Monclus et al. 2006; Toillon et al. 2013). Gas exchange measurements have confirmed that variations in photosynthetic capacities and assimilation rate can contribute at least partly to variations in WUE depending on growing conditions (Ripullone et al. 2004; Monclus et al. 2006).

Along with N, the elements P and K are two of the most common nutrients limiting plant growth. In our study, previous land use also differed in K concentration, with higher values recorded in the intensively fertilized cropland (Verlinden et al. 2013b). P concentrations were slightly higher in cropland too, although the differences were not significant. Increased K supply resulted in decreased transpiration and increased leaf level WUE in several greenhouse-grown poplar genotypes (Harvey & van den Driessche 1999) as well as in Sitka spruce (Bradbury & Malcolm 1977). Similarly higher foliar K and P concentrations have been associated with increased WUEi (Xu et al. 2000; Cao et al. 2011). However, these trends do not fit with the δ13C differences observed in our study for the soil K and P concentrations. We conclude therefore that variations in δ13C among land use types were primarily caused by differential N availability.

Local variations in water availability may play an important role in driving spatial δ13C variations in large-scale plantations. However, although not directly addressed, we conclude that this was unlikely the case for the following reasons. Firstly, the plantation was installed on a flat land with very minor differences in elevation, leading to a more or less constant water table across plots. Secondly, during the year of sampling no differences in soil C, soil texture and bulk density were detected between the two former land uses (Verlinden et al. 2013b). These three factors are the most important determinants of the water-holding capacity of a soil (Vereecken et al. 1989). Thirdly, the leaves used for δ13Cleaf determination were sampled in early August, when soil water availability had no longer been a limiting factor anymore for the past several weeks. Local but continuous measurements indeed revealed that soil water potential in the first 40 cm was returned to zero from mid-July onwards due to considerable rainfall events (138 mm cumulated over the 4 weeks before sampling, see Broeckx et al. 2013).

Although δ13C estimates were significantly influenced by previous land use, we found no significant link between plot soil characteristics and average plot δ13C values. The impact of soil characteristics on tree growth performance had been previously investigated for different poplar genotypes grown under short-rotation coppice, and no clear relations could be drawn either (Laureysens et al. 2004). However, in contrast to the latter study significant variation in individual soil characteristics occurred among the plots covering the whole plantation area. This suggests that a lack of relationship with δ13C may not be attributable to the low variation in soil properties. Another possible explanation for the lack of relationship may be that important soil elements were unaccounted for, although the most common ones known to influence poplar physiology (i.e. pH and mass fractions of nutrients N, P and K) were included in this study (Heilman & Fu-Guang 1993; Harvey & van den Driessche 1997, 1999).

In conclusion our results indicate that substantial spatial variations in leaf-level WUE occur in large-scale plantations as a result of differences in former land use. Whether this pattern remains true after several rotations is unknown and requires additional assessment. Values of δ13Cwood and δ13Cholocellulose were tightly related, but the offset between the two was genotype dependent. Although this did not directly affect genotype ranking, δ13Cwood may not provide a reliable signal to infer on WUE differences among poplar genotypes of contrasting parentage. Genotypes with the highest WUE displayed the longest growing season. However, irrespective of the δ13C estimate considered, our results showed that WUE was largely uncoupled from growth performance. This leaves hope to identify genotypes with satisfactory growth and higher WUE in poplar bioenergy plantations.

Acknowledgments

This research has received funding from the European Research Council under the European Commission's Seventh Framework Programme (FP7/2007-2013) as ERC grant agreement no. 233366 (POPFULL), as well as from the Flemish Hercules Foundation as Infrastructure contract ZW09-06. Further funding was provided by the Flemish Methusalem Programme, the Research Council of the University of Antwerp and the UGent Multidisciplinary Research Partnership ‘Biotechnology for a Sustainable Economy’ (01MRB510W). We gratefully acknowledge the excellent technical assistance of Joris Cools and Bart Ivens, the logistic support of Kristof Mouton at the field site, the laboratory analyses of Gerrit Switsers and Nadine Calluy as well as Lorenz Gerber for the analysis of the purity of the holocellulose samples by micropyrolysis.

Ancillary