Tracing fresh assimilates through Larix decidua exposed to elevated CO2 and soil warming at the alpine treeline using compound-specific stable isotope analysis


Author for correspondence:

Rolf T. W. Siegwolf

Tel: +41 56 310 2786



  • How will carbon source–sink relations of 35-yr-old larch trees (Larix decidua) at the alpine treeline respond to changes in atmospheric CO2 and climate? We evaluated the effects of previously elevated CO2 concentrations (9 yr, 580 ppm, ended the previous season) and ongoing soil warming (4 yr, + 4°C).
  • Larch branches were pulse labeled (50 at% 13CO2) in July 2010 to trace fresh assimilates through tissues (buds, needles, bark and wood) and non-structural carbon compounds (NCC; starch, lipids, individual sugars) using compound-specific isotope analysis.
  • Nine years of elevated CO2 did not lead to increased NCC concentrations, nor did soil warming increase NCC transfer velocities. By contrast, we found slower transfer velocities and higher NCC concentrations than reported in the literature for lowland larch. As a result of low dilution with older carbon, sucrose and glucose showed the highest maximum 13C labels, whereas labels were lower for starch, lipids and pinitol. Label residence times in needles were shorter for sucrose and starch (c. 2 d) than for glucose (c. 6 d).
  • Although our treatments showed no persistent effect on larch carbon relations, low temperature at high altitudes clearly induced a limitation of sink activities (growth, respiration, root exudation), expressed in slower carbon transfer and higher NCC concentrations.


Increasing atmospheric CO2 concentrations triggering ongoing global warming have an impact on carbon (C) cycling in terrestrial ecosystems (IPCC, 2007). Free air CO2 enrichment (FACE) experiments have been carried out to understand the effects of elevated CO2 on tree ecosystems (see reviews by Körner, 2006; Norby & Zak, 2011). Although most FACE experiments showed an overall increase in tree photosynthesis as a result of elevated CO2 (Handa et al., 2005; Crous et al., 2008; Bader et al., 2010), this additionally acquired C could often not be used for growth (Körner, 2006). Yet, continuously increased photosynthesis is only possible if trees maintain their C source–sink balance. Carbon can be transferred on many possible pathways to diverse sinks, but phenological stages as well as environmental factors limit the activity of these sinks (Körner, 2006).

At the alpine treeline, sink activities, such as growth, respiration and root exudation, are temperature limited, and no persistent tree growth is thought to be possible below a seasonal mean air temperature of 6°C (Körner, 1998). Nevertheless, treeline species generally show positive net photosynthesis even below 0°C (Tranquillini, 1979), and pine trees growing in the treeline ecotone have been found to store surplus C in the form of non-structural carbon compounds (NCC) (Hoch et al., 2002; Hoch & Körner, 2003). Recently, a consistent increase in starch reserves has been reported across 14 tree species from the lowland to the treeline world-wide (Hoch & Körner, 2011). Thus, at sites at which temperature is thought to limit sink activity, elevated CO2 could further increase already high NCC concentrations. Conversely, warming could increase sink activity, resulting in faster NCC transfer. However, how the changing climate is affecting C allocation within trees still remains unclear, because most studies have focused on growth responses to environmental changes and not on the transport pathways for C (Körner, 2006).

13C pulse labeling is a powerful tool for tracing fresh assimilates (sugars, starch and lipids) through trees (Kagawa et al., 2006; Barthel et al., 2011; Epron et al., 2012; Keel et al., 2012). Effects of elevated CO2 and warming on C allocation can be assessed by comparing the temporal dynamics of the 13C label in NCC and by determining concentrations of NCC in different tissues. Compound-specific isotope analysis of single neutral sugars and sugar alcohols allows differentiation between the various temporal and spatial scales in which these substances are mixing with 13C-labeled fresh assimilates (Schneider & Schmitz, 1989; Kuhns & Gjerstad, 1991; Nogues et al., 2006).

For this study at a Swiss alpine treeline site, we labeled larch branches (Larix decidua) with highly enriched 13CO2 at the beginning of July 2010. The trees were growing at a FACE site (CO2 concentration c. 580 ppm, from 2001 to 2009), where an additional soil warming experiment (4°C above ambient temperature, 2007 to ongoing) was carried out. The aim of this study was to investigate the effect of 9 yr of elevated CO2 and 4 yr of soil warming on C allocation within larch branch tissues (needles, bark and wood) and NCC (single sugars, starch and lipids) at the treeline. Furthermore, we compared our results with previously published data from lowland trees of the same species to investigate differences in sink activity between these ecosystems. We hypothesized the following: (1) larch trees that were exposed to elevated CO2 until October 2009 and that showed increased photosynthetic uptake throughout the nine FACE years would still show increased NCC concentrations in the following growing season (2010) in bark and wood (not in needles as larch trees are deciduous); (2) an increase in soil temperature would increase sink activity (growth, respiration or root exudation), resulting in an increased transfer velocity of NCC; and (3) the combined treatment (elevated CO2 concentration until autumn 2009 and warmed soil) would result in an increased transfer velocity of NCC as well as increased NCC concentrations.

Materials and Methods

Site description

The study was carried out in a 35-yr-old afforestation site at Stillberg near Davos, Switzerland at an altitude of 2180 m asl, which is just above the current treeline (Barbeito et al., 2012). The mean annual precipitation is 1100 mm, and the mean annual air temperature is 2°C, with February as the coldest (Tmean = −5°C) and July as the warmest (Tmean = 10°C, average from 1975 to 2009, climate station of the Swiss Federal Institute for Forest, Snow and Landscape Research WSL at Stillberg, 2090 m asl; Dawes et al., 2011a) month. The slopes facing north-east are inclined by 25–30°. The soils are mainly sandy Ranker and weakly developed Podzols.

The main tree species at the study site are Larix decidua Mill. and Pinus mugo ssp. uncinata DC. They form an open canopy with a dense understory layer consisting mostly of ericoid dwarf shrubs (Vaccinium myrtillus L., Vaccinium gaultherioides Bigelow (group V. uliginosum agg.) and Empetrum nigrum ssp. hermaphroditum (Hagerup) Böcher). The 2010 growing season, defined here as the period when daily mean soil temperatures (10 cm depth) were above 3.2°C (Körner & Paulsen, 2004), lasted from 4 June to 26 September.

Experimental set-up

In June 2001, 20 L. decidua trees were selected in the upper part of the afforestation for a FACE experiment. They were divided into 10 groups of two neighboring trees (Hättenschwiler et al., 2002). Half of the groups were randomly assigned to a treatment with elevated CO2 (576 ± 15 ppm, mean ± 1SE from 2001 to 2009), whereas the rest received no additional CO2 (c. 380 ppm; Dawes et al., 2011b). Pure CO2 was added to the trees during daylight hours of the snow-free period through drip irrigation tubes woven into the tree crowns. The FACE experiment lasted until the beginning of October 2009.

A soil warming treatment was added in half of the FACE plots (five CO2-exposed plots and five ambient CO2 plots) after snow melt at the end of May 2007. In each warmed plot, 26 m of 420-W heating cables were placed on the ground surface below the dwarf shrubs, thus covering a plot area of c. 1.1 m2. Heating was accomplished by switching the power supply on and off in 1-min intervals (Hagedorn et al., 2010). The experimental warming resulted in a mean increase in soil temperatures at a soil depth of 5 cm by 3.7 ± 0.3°C in 2010, whereas the air was only 0.9 ± 0.1°C warmer at 20 cm above the ground (mean ± 1SE in 2007; Hagedorn et al., 2010). Trees were exposed to warming treatments during the snow-free period from 2007 to 2010.

Pulse labeling experiment

The pulse labeling experiment was carried out on 16 larch trees within the experimental set-up. At the time of labeling, larch trees were, on average, c. 2.5 m in height and had a basal area of c. 25 cm2 at a height of 50 cm (Dawes et al., 2011b). As the labeling could not be carried out in 2009, because of limited resources, we started the labeling on the earliest date possible in 2010 (5 July), after needles on existing shoots were fully expanded but when the elongation of new shoots had just started. We chose one sun-exposed branch per tree and labeled it once for c. 2 h. The labeling restricted to only one branch was justifiable at this phenological stage, as branches were mainly functioning as autonomous systems in C allocation (Kozlowski, 1992), because of the C sink strength of the start of shoot growth. All 16 larch trees were labeled during four consecutive sunny days between 08:00 and 17:00 h. Just before the labeling started, branches were enclosed in transparent polyethylene bags with a volume of 40 l that were sealed around the branch using putty (Therostat, Terson Henkel GmbH, Dü sseldorf, Germany). The bag was fully inflated with ambient air using a piston pump (10 l min−1) to ensure the same defined volume of air for each labeling and to avoid direct branch contact. The air inside was stirred with a fan to provide a homogeneous CO2 distribution within the bag and also circulated through the piston pump, thus creating a closed air loop. Part of the air in the loop circulated through a scrubber tube filled with silica gel, keeping the relative air humidity at 80 ± 2.5% (measured with an HMP35 Humidity Probe, Vaisala, Helsinki, Finland). The mean air temperature within the bag was 23 ± 1.5°C.

To monitor CO2 concentrations and temperatures, a CO2 infrared gas analyzer (GMP343, Vaisala) was mounted on an aluminum rod and enclosed in the bag. This set-up was kept in place with a tripod. The infrared gas analyzer was calibrated for 12CO2 with an absorption maximum at a wavelength of 4.26 μm. Therefore, its sensitivity to 13CO2 (absorption maximum at 4.36 μm) was reduced to c. 30% (as described by McDermitt et al., 1993 for Li6250). Our GMP343 is equipped with comparable filters with the same absorption range. As we labeled with c. 50 at% 13CO2, the actual CO2 concentration was c. 50% higher than the measured CO2 concentration in the bag. The labeling was initiated by injecting 20 ml of 98.9 at% enriched 13CO2 into the air circulation loop, resulting in an initial actual in-bag CO2 concentration of 780 ppm with c. 50 at% 13C (400 ppm 13CO2 and 380 ppm ambient CO2). New label was added as soon as the measured CO2 concentration dropped to 200 ppm. For label repetitions, we mixed 20 ml of 98.9 at% 13CO2 with 20 ml of pure CO213C = −29.9‰), resulting in an additional 800 ppm of CO2 with c. 50 at% 13CO2 and 50 at% 12CO2. The bag was removed after the label additions had been repeated three times. The average labeling time was 130 ± 13 min per tree.

Sampling of plant material

The total length of the selected branch on each tree was measured before the labeling experiment started. To determine the label strength immediately after labeling, needles were taken from different parts of the labeled branch immediately after the bag was removed (Endrulat et al., 2010). The average δ13C value of these needles was 359 ± 42‰ (δ notation relative to the international reference standard Vienna-Pee Dee Belemnite: V-PDB). Shoots formed in 2009 (bark and wood) with needles and buds from 2010 were sampled from the branch before labeling (4 July 2010), 4 h after labeling (0.17 d), after 1 d, after 2 wk and on 15 September 2010 just before the needles started to change color (c. 71 d after label application). Shoots were kept cool to minimize metabolic processes until the different tissues (needles, buds, bark and wood) were separated in the evening and heated in a microwave oven (60 s at 600 W) to stop enzymatic and metabolic activities (Wanek et al., 2001). All sampled tissues were subsequently dried for 24 h at 60°C and ground to fine powder.

Extraction of sugars, lipids and starch

The water-soluble fraction was extracted from 100 mg of dried plant powder using 1.5 ml of Milli-Q water (18.2 MΩ, total organic C < 5 ppb). The samples were heated at 8°C for 30 min, centrifuged at 10 000 g for 2 min, and 800 μl of the supernatant were transferred to 2-ml reaction vials. Neutral carbohydrates were purified from ionic compounds using anion and cation exchange cartridges (OnGuard II H 1cc, OnGuard II A 1cc, Dionex Corporation, Sunnyvale, CA, USA; Wild et al., 2010). To preserve the HPLC column, phenols were removed (OnGuard II P 1cc, Dionex Corporation). (See Rinne et al. (2012) for details about sugar purification.) Lipids and starch were extracted from the plant material used previously for sugar extraction. We used the method described by Wanek et al. (2001) and modified after Richter et al. (2009). For the lipid extraction, a monophasic mixture (12 : 3 : 5, v/v/v) of methanol, chloroform and water (MCW) was used. To separate the lipophilic and water phases, additional water and chloroform were added. The lower lipophilic phase was transferred to glass vials. Before starch extraction, the plant material was washed three more times with MCW and with water to remove all soluble compounds. Starch from this plant material was then gelatinized at 99°C and broken down to glucose by enzymatic hydrolysis through α-amylase (heat stable form, Sigma Aldrich, St. Louis, MO, USA). Enzymes were removed from the glucose solution by centrifugal filter units (Vivaspin 500, Sartorius Stedim Biotech GmbH, Göttingen, Germany) at 12 000 g for 50 min. All extracted solutions were stored at −20°C for further analysis.

Bulk and compound-specific measurement of δ13C values

δ13C (‰) values were determined with an elemental analyzer (EA: EA1110 CHN, Carlo Erba, Milan, Italy) connected to an isotope ratio mass spectrometer (IRMS: Delta S, Finnigan MAT, Bremen, Germany). Laboratory standards with known δ13C values were measured as every tenth sample of the analysis sequence (SD < 0.2‰). We analyzed δ13C values from 0.6–0.8 mg of homogenized dried material of all tissues (needles, buds, bark and wood) from all sampling times, and 50 μl of the extracted solutions of sugars, starch and lipids. The concentrations of sugars, starch and lipids were measured by drying aliquots of extracts. Starch concentrations were additionally determined enzymatically by converting starch from aliquots of plant material to glucose (Clarase, Enzyme Solutions Pty Ltd, Crydon South, Australia) and quantifying it photometrically (Körner et al., 1995).

Compound-specific isotope analysis was carried out for sugar extracts. δ13C values were measured on a Delta V Advantage IRMS coupled with high-performance liquid chromatography (HPLC) using a Finnigan LC Isolink interface (Thermo Fisher Scientific Inc., Bremen, Germany) with a CarboPac PA20 anion exchange column (3 × 150 mm, Dionex) for chromatographic separation (Rinne et al., 2012). The mobile phase was 1 mM NaOH (Boschker et al., 2008). The column temperature was set to 35°C. Sugar extracts contained pinitol/myo-inositol, sucrose, glucose, raffinose and fructose, in the order of their retention times. Pinitol and its precursor myo-inositol, both sugar alcohols, were co-eluting and were therefore treated as one compound. We used an external standard dilution row (400–1600 ng per injection) containing the compounds in our samples with known δ13C values to calculate concentrations and to correct δ13C values (SD < 0.5‰). As the concentrations of raffinose were below the detection limit (1 mg g−1 dry matter (DM)) during the growing season, it was not possible to obtain reliable δ13C values for this compound. During fructose analysis at 35°C, we experienced isomerization and decided not to use these δ13C values. To create an isotope dilution series, we used D-glucose with two 13C atoms (99 at%) (Cambridge Isotope Laboratories Inc., Andover, MA, USA) which we diluted with D-glucose (δ13C = −11.42‰), creating glucose with δ13C values ranging from −11 to 1600‰. These samples were measured with both EA-IRMS and HPLC-IRMS. We corrected the δ13C values measured by HPLC-IRMS for the offset, which increased slightly with increasing δ13C values, following a polynomial function (x, δ13C values measured with HPLC-IRMS; y, offset = ax2 + bx, a = 1.65E-06 ± 1.54E-07, < 0.01, b = 3.49E-03 ± 2.20E-04, < 0.01, r2 = 1.00).

Data analyses

The δ13CSample (‰) notation was converted to an absolute isotope ratio:

display math(Eqn 1)

with RStandard = 0.0111802 (V-PDB standard; Zhang et al., 1990). RSample was converted to the percentage of heavy atoms ASample (at%):

display math(Eqn 2)

To account for differences in the percentage of heavy atoms already present before labeling (natural abundance and FACE), we calculated the percentage of labeled atoms in excess of natural background, Excess 13CSample (%):

display math(Eqn 3)

(ASample, percentage of 13C of a certain tissue at a certain sampling time, compound and tree; ABackground, percentage of 13C of this tissue, compound and tree before labeling). The 13C label of the needles immediately after labeling varied greatly among trees (Excess 13CInitialNeedles = 0.42 ± 0.05%). This was mostly a result of differences in the amount of labeled tissue (labeled branch length, 260 ± 25 cm). We found a significant negative correlation between branch length and 13C label in needles immediately after labeling (F1,14 = 8.1, = 0.01, r = −0.61). To correct for this difference, we expressed Excess 13CSample of all tissues and compounds relative to the 13C label of bulk needles immediately after labeling of each tree (Endrulat et al., 2010):

display math(Eqn 4)

(Excess13Cinit, 13C excess for needles immediately after labeling). Hence, Excess 13Crelative of the 13C label of needles immediately after labeling was defined as 100%, whereas the maximum 13C label of needle, bark or wood compounds (such as sucrose) could well exceed this value.


The effects of elevated CO2, soil warming and time of sampling on all measured parameters were tested by linear mixed effect models using R version 2.14.1 (R Development Core Team, 2008) and the NLME package (Pinheiro et al., 2008). One model was fitted with restricted maximum likelihood for each compound and tissue type to test Excess 13Crelative for differences among treatments and sampling times. Random effects were included to account for the hierarchy of treatment set-up, that is, warming treatment nested within group (groups of two neighboring trees, half of which were randomly assigned to the treatment with elevated CO2; = 8). To correct for the heterogeneity in variance of residuals at different sampling times, we applied a heterogeneous residual structure. Taking into account the violation of independence of residuals between different sampling times, we applied a residual autocorrelation structure for individual trees (auto-regressive model of order 1 (corAR1); Pinheiro et al., 2008).

To describe the temporal dynamics of Excess 13Crelative in the different tissues and in NCC, curves were fitted to the time vs Excess 13Crelative data. These curves were all special cases of one equation, which included an exponential growth and an exponential decrease (label decay studies often use exponential decrease functions; e.g. Ruehr et al., 2009):

display math(Eqn 5)

(kG and kD, fitting parameters for the exponential growth and decrease functions, respectively; t, time in days; a and c, scaling factors). The function starts at = 0 with Excess 13Crelative = y0 + c and goes towards t→∞ with Excess 13Crelative = y0 + a. All needle compounds (except lipids) and buds were maximally labeled immediately after labeling; therefore, we set a = 0. Bark photosynthesis of larch shoots is much lower than bark respiration (Keller, 1973); for bark and wood, we therefore set y0 + c = 0.

Parameters were determined and curves fitted using Sigma Plot 12 (Systat Software Inc., Chicago, IL, USA). The mean residence time was calculated according to Högberg et al. (2008):

display math(Eqn 6)

For bark, wood and needle lipid samples, the tracing time of 71 d was too short; thus, kD was not significantly different from zero. Therefore, τ could not be calculated and, instead, we calculated the time in days until 99% of the maximum label was reached as:

display math(Eqn 7)


Concentrations of individual NCC

The previous CO2 enrichment and ongoing soil warming treatments did not affect the concentrations of individual NCC for any of the investigated tissues (Table 1). Averaged over all 16 larch trees and two sampling times (before labeling and after 71 d), needles (203 ± 7 mg g−1 DM) and bark (204 ± 11 mg g−1 DM) showed the highest total NCC concentrations (total sugar concentrations + starch concentrations + lipid concentrations, mg g−1 DM), whereas wood contained less NCC (86 ± 7 mg g−1 DM). Relative to the total NCC concentrations, sugars were most abundant in needles (71%), whereas their contribution was lower in bark (53%) and wood (47%). By contrast, lipid concentrations were relatively high in bark (20%) and wood (16%) compared with needles (8%).

Table 1. Average concentrations (mg g−1 dry matter) of non-structural carbon compounds extracted from needles, bark and wood of Larix decidua at Stillberg (Switzerland) for different treatments (control, warmed soil, elevated CO2 and elevated CO2 + warmed soil)
TissueCompoundControlWarmingCO2CO2 + warming
  1. Mean ± 1SE (= 4) values for all replicates were averaged over two sampling times (before and 71 d after labeling). No significant treatment effects (< 0.05) were found.

NeedlesSugars and sugar alcohols149.2 ± 19.1135.2 ± 8.8142.8 ± 8.4147.2 ± 8.2
Pinitol/myo-inositol39.0 ± 5.034.1 ± 1.935.5 ± 1.637.2 ± 2.4
Glucose19.4 ± 4.215.1 ± 1.617.3 ± 1.717.2 ± 1.8
Sucrose75.2 ± 8.872.2 ± 8.074.9 ± 5.777.7 ± 5.8
Raffinose2.2 ± 0.61.7 ± 0.62.4 ± 0.71.9 ± 0.6
Fructose13.3 ± 2.312.0 ± 3.312.6 ± 2.113.3 ± 1.7
Starch51.8 ± 20.155.5 ± 22.434.9 ± 7.836.0 ± 6.5
Lipids16.4 ± 2.419.4 ± 2.915.6 ± 1.416.5 ± 1.9
BarkSugars and sugar alcohols106.9 ± 14.494.9 ± 13.2119.0 ± 20.0106.9 ± 15.9
Pinitol/myo-inositol23.1 ± 4.223.5 ± 4.225.5 ± 5.524.7 ± 3.8
Glucose37.4 ± 4.333.5 ± 5.043.1 ± 7.633.5 ± 6.8
Sucrose24.9 ± 2.526.8 ± 2.929.1 ± 2.225.7 ± 3.4
Raffinose4.1 ± 1.33.6 ± 0.93.2 ± 1.13.7 ± 1.3
Fructose22.2 ± 8.911.6 ± 6.118.8 ± 8.823.7 ± 9.5
Starch59.3 ± 9.255.7 ± 7.354.6 ± 6.262.6 ± 8.1
Lipids44.6 ± 7.642.9 ± 7.540.0 ± 5.736.8 ± 3.9
WoodSugars and sugar alcohols44.9 ± 6.137.9 ± 6.540.0 ± 4.937.9 ± 4.5
Pinitol/myo-inositol12.2 ± 0.812.0 ± 3.511.1 ± 1.711.3 ± 1.6
Glucose3.2 ± 0.62.9 ± 0.72.8 ± 0.53.1 ± 0.8
Sucrose25.5 ± 5.819.7 ± 5.322.3 ± 4.619.5 ± 4.2
Raffinose2.7 ± 0.91.8 ± 0.72.4 ± 0.62.9 ± 0.4
Fructose2.4 ± 0.41.8 ± 0.42.2 ± 0.52.6 ± 0.5
Starch48.0 ± 6.636.2 ± 11.230.9 ± 5.540.7 ± 9.2
Lipids11.7 ± 1.515.3 ± 3.712.3 ± 2.115.2 ± 3.0

On average, needle sugars consisted of 52% sucrose, 25% pinitol/myo-inositol, 12% glucose, 9% fructose and only very small contributions of raffinose. In bark, the average sucrose contribution to total sugars was smaller than that in needles, whereas the contribution of glucose was higher (25% sucrose, 34% glucose, 22% pinitol/myo-inositol, 18% fructose and small raffinose contribution). The opposite pattern was observed for wood (54% sucrose, 29% pinitol/myo-inositol and small contributions of glucose, fructose and raffinose). The average starch concentrations were highest in bark, intermediate in needles and lowest in wood (bark, 58 ± 4 mg g−1 DM; needles, 45 ± 8 mg g−1 DM; wood. 38 ± 4 mg g−1 DM). The same was true for the mean lipid concentrations, which were much smaller in needles and in wood than in bark (needles, 17 ± 1 mg g−1 DM; wood, 14 ± 1 mg g−1 DM; bark, 41 ± 3 mg g−1 DM).

Between the sampling times of 4 July (pre-labeling) and 15 September (last sampling), significant changes in total NCC concentrations, as well as in the relative composition of NCC, were found in all tissues (Fig. 1). The total amount of NCC decreased with time in needles and bark (needles, F1,8 = 3.2, = 0.11; bark, F1,9 = 30.0, < 0.01), whereas it increased in wood (F1,6 = 76.7, < 0.01). The most obvious change in the relative composition of NCC was the increase in sucrose concentrations towards the end of the growing season in all tissues (needles, F1,11 = 12.0, < 0.01; bark, F1,11 = 23.6, < 0.01; wood, F1,11 = 172.0, < 0.01). In needles, there was a significant decrease in starch concentrations (F1,8 = 10.0, = 0.01), whereas starch concentrations increased in wood between 4 July and 15 September (F1,4 = 26.0, = 0.01) and remained unchanged in bark (F1,10 = 1.6, = 0.25).

Figure 1.

Concentrations of non-structural carbon compounds: lipids, starch, sugars (fructose, raffinose, sucrose and glucose) and sugar alcohols (pinitol) extracted from needles, bark and wood of Larix decidua at Stillberg (Switzerland). Pinitol includes myo-inositol, which was co-eluting. Sugar and sugar alcohol bars are hatched. Compounds were extracted from samples taken before labeling (4 July 2010) as well as 71 d after labeling (15 September 2010). Mean values for all trees are shown (all treatments = 16).

Tracing the 13C label through tissues

For reasons of simplicity, we use ‘13C label’ when referring to Excess 13Crelative throughout the rest of the text. The 13C label of bulk samples and NCC immediately after labeling is the calculated starting point of Eqn (Eqn 4) and is termed ‘initial 13C label’ throughout the text. All values calculated from Eqn (Eqn 4) are given in Table 2.

Table 2. Parameters describing the curves fitted to time after labeling vs 13C label (Excess 13Crelative) data for all trees of all treatments (= 16), given separately for bulk tissues of needles, buds, bark and wood, as well as for non-structural carbon compounds extracted from needles of Larix decidua
Tissue/compoundExcess 13Crelative (%)Time (d)Curve fit
StartEndτ mean residence timeTime99% r 2
  1. Start, calculated initial 13C label after labeling (Eqn (Eqn 4)); End, remaining 13C label as time approaches ∞; τ, mean residence time of the label in a certain pool; Time99%, duration before 99% of the maximum label was reached. Time99% for needle lipids is tainted with a high uncertainty, as y0 was not significantly different from zero. The 13C label is expressed relative to the initial needle bulk label (Excess 13Crelative). Mean ± 1SE (= 16) values for all trees. ‘NA’, no values were available; ‘None’, maximal label was reached at the beginning.

Needles100 ± 322 ± 21.7 ± 0.2None0.82
Sugars and sugar alcohols435 ± 44 ± 143.6 ± 1.0None0.88
Pinitol/myo-inositol20 ± 8NANANone0.40
Glucose345 ± 471 ± 206.1 ± 3.4None0.76
Sucrose729 ± 341 ± 222.3 ± 0.3None0.88
Starch191 ± 314 ± 102.2 ± 0.9None0.63
Lipids9 ± 2NANA8.5 ± 4.40.46
Buds66 ± 625 ± 45 ± 3.3None0.58
Bark0NANA2.7 ± 1.30.53
Wood0NANA6.2 ± 1.10.71


Following the pulse labeling, the decrease in 13C label in bulk needle samples (Fig. 2a), as well as in most needle NCC (sugars and starch, but not lipids; Fig. 3a), and in individual sugars (sucrose, glucose and pinitol; Fig. 3a), followed an exponential function. Neither the previous CO2 treatment nor the ongoing soil warming treatment influenced significantly the temporal dynamics of the 13C label in needle compounds or bulk material (Fig. 2a, Table 3). Needle sugars showed a higher initial 13C label compared with starch and lipids, mainly as a result of highly labeled sucrose (Table 2). The 13C label in needle sucrose decreased rapidly and was already very low 2 wk after labeling (3 ± 0%; Fig. 3a). Accordingly, the mean residence time of the 13C label in sucrose was short (Table 2). By contrast, the 13C label of needle glucose had a rather long mean residence time, although its initial 13C label was lower than that for sucrose (Table 2). Needle pinitol/myo-inositol was initially only marginally labeled with 13C and the 13C label hardly changed during the experiment; thus, neither the mean residence time nor the remaining 13C label in pinitol/myo-inositol could be calculated using Eqn (Eqn 4). The initial 13C label in needle starch was lower than that in needle sugars and decreased quickly, resulting in a mean residence time of c. 2 d (Table 2). Needle lipids showed a different pattern, with a 13C label that increased until the eighth day after labeling, followed by a decrease (Fig. 3a). The maximum 13C label in lipids was 23%. Whether the pulse labeling took place in the morning or in the afternoon did not affect the 13C label dynamic in needle sucrose (F1,7 = 0.0, = 0.89) or needle starch (F1,7 = 0.8, = 0.39).

Table 3. Temporal dynamics of 13C label in needle, bark and wood tissues of the non-structural carbon compounds extracted from different tissues of Larix decidua at different sampling times throughout the tracing period compared for all treatment combinations
 Time (d)Excess 13Crelative (%)
ControlWarmingCO2CO2 + warming
  1. Mean ± 1SE (= 4) values for the treatments: control, warmed soil, elevated CO2 and elevated CO2 + warmed soil. The 13C label is expressed relative to the initial needle bulk label (Excess 13Crelative).

  2. a

    Significant treatment effects (< 0.05) at the corresponding sampling time.

NeedleSugars and sugar alcohols0.17371 ± 40427 ± 45470 ± 70393 ± 59
1293 ± 45312 ± 36387 ± 101310 ± 17
1415 ± 28 ± 113 ± 312 ± 2
716 ± 14 ± 05 ± 15 ± 1
Pinitol/myo-inositol0.1718 ± 316 ± 123 ± 519 ± 1
118 ± 117 ± 221 ± 419 ± 2
1417 ± 114 ± 219 ± 320 ± 2
7111 ± 19 ± 112 ± 29 ± 1
Glucose0.17359 ± 79480 ± 94281 ± 41224 ± 45
1298 ± 62303 ± 27288 ± 47276 ± 29
1441 ± 1512 ± 336 ± 1340 ± 6
716 ± 24 ± 15 ± 03 ± 1
Sucrose0.17677 ± 68703 ± 30735 ± 132601 ± 86
1409 ± 59425 ± 41580 ± 173444 ± 27
145 ± 42 ± 04 ± 13 ± 0
710 ± 00 ± 00 ± 00 ± 0
Starch0.17123 ± 45196 ± 37203 ± 50188 ± 39
177 ± 23133 ± 3179 ± 91118 ± 20
1415 ± 316 ± 422 ± 713 ± 3
7110 ± 112 ± 311 ± 111 ± 2
Lipids0.1710 ± 19 ± 111 ± 311 ± 1
115 ± 214 ± 215 ± 217 ± 2
1421 ± 119 ± 126 ± 622 ± 4
7117 ± 115 ± 216 ± 117 ± 3
BarkSugars and sugar alcohols166 ± 573 ± 383 ± 883 ± 10
71a9 ± 16 ± 19 ± 19 ± 2
Pinitol/myo-inositol11.4 ± 0.41.7 ± 0.62.0 ± 0.82.1 ± 0.2
719 ± 07 ± 111 ± 212 ± 2
Glucose134 ± 425 ± 222 ± 247 ± 7
71a14 ± 38 ± 111 ± 213 ± 3
Sucrose1292 ± 34316 ± 10380 ± 52343 ± 57
711 ± 01 ± 01 ± 01 ± 0
Starch16 ± 15 ± 19 ± 39 ± 2
718 ± 05 ± 18 ± 18 ± 2
Lipids12 ± 13 ± 22 ± 12 ± 0
719 ± 17 ± 19 ± 214 ± 6
WoodSugars and sugar alcohols716 ± 05 ± 16 ± 17 ± 1
Pinitol/myo-inositol7111 ± 18 ± 112 ± 212 ± 1
Glucose7110 ± 35 ± 16 ± 18 ± 2
Sucrose711 ± 01 ± 01 ± 01 ± 1
Starch718 ± 17 ± 110 ± 110 ± 2
Lipids71a17 ± 29 ± 117 ± 221 ± 2
Figure 2.

Temporal dynamics of 13C label after pulse labeling in bulk tissues of (a) needles, (b) buds, (c) bark and (d) wood of Larix decidua shoots collected at Stillberg (Switzerland) between 5 July 2010 and 15 September 2010 for different treatments. Warming treatment: unwarmed (gray lines and open symbols); warmed (black lines and symbols). CO2 treatment: ambient CO2 (dashed lines and circles); elevated CO2 (solid lines and triangles). The 13C label is expressed relative to the initial needle bulk label (mean for all trees of a single treatment ± 1SE,= 4). All curves were fitted using an exponential increase/decrease function (Eqn (Eqn 4)). Note that there is a different y-scale for upper and lower panels.

Figure 3.

Temporal dynamics of 13C label in different non-structural carbon compounds, such as sugars (including sugar alcohols), lipids and starch (upper panels), as well as in single low-molecular sugars (glucose and sucrose) and sugar alcohols (pinitol) (lower panels) extracted from (a) needles and (b) bark of Larix decidua at Stillberg (Switzerland). The 13C label is expressed relative to the initial needle bulk label (Excess 13Crelative; mean for all trees from all treatments ± 1SE,= 16). The curves are fitted using an exponential increase/decrease function (Eqn (Eqn 4)). Zoomed in sections of the graphs are displayed in the upper right corner of the corresponding panels.


The maximum 13C label in bark amounted to only 13% of the initial needle label during the experiment. The 13C label in bulk bark followed an exponential increase to a maximum, followed by an exponential decrease (Fig. 2c). Trees exposed to warmed soil but ambient CO2 showed a faster decrease of the 13C label in bark sugars relative to control trees. Thus, the 13C label in this treatment combination was significantly lower after 71 d (warming: F1,6 = 11.1, = 0.02; Table 3). Although it was not statistically significant, this general pattern occurred in bulk bark (Fig. 2c) as well as in all single compounds of bark (Table 3). The 13C label in bark sucrose after 1 d (330 ± 20%) was comparable with the 13C label in needle sucrose (459 ± 49%). For glucose and sucrose, the 13C label in bark decreased throughout the growing season, whereas it increased in pinitol/myo-inositol and lipids (Fig. 3b). For bark starch, the 13C label did not change over time (Fig. 3b).


As in bark, the temporal dynamics of the 13C label in bulk wood followed an exponential increase. However, relative to bark, the time until 99% of the maximal label was reached in bulk wood was longer (Table 2). The maximum 13C label in bulk wood (13%) was similar to that in bulk bark (13%). Significantly less 13C label was found in bulk wood from larch trees exposed to warmed soil after 71 d relative to control trees, indicating a faster decrease in the 13C label in the wood of these trees (warming: F1,6 = 10.9, = 0.02; Fig. 2d). The 13C label in wood compounds after 71 d was comparable with that in the compounds of bark and needles. In wood lipids, we again found a significantly faster decrease in 13C label for trees exposed to warmed soils at ambient CO2 than for other treatment combinations (CO2 × warming: F1,6 = 22.2, < 0.01; Table 3).


The 13C label in buds followed an exponential decrease similar to that seen in needles, although starting from a lower value (Table 2). However, at the end of the experiment, more 13C label remained in bulk buds than in needles (Table 2). In addition, the mean residence time was longer in buds than in needles (Table 2). The treatments did not influence the temporal dynamics of 13C label in bulk buds (Fig. 2b). No NCC were extracted for buds.

Between tissues

We used significant correlation analyses of 13C labels in NCC for two tissues at a specific sampling time as an indication for the transfer of this NCC between these two tissues (variation given by individual trees). For sucrose, evidence of a fast C transfer between tissues was given by a strong correlation of the 13C label 1 d after the labeling between bark and needles (F1,12 = 18.9, < 0.01, = 0.78, including all trees = 16). Pinitol/myo-inositol had approximately the same 13C label in all tissues at the end of the tracing period (needles, 10%; bark, 10%; wood, 11%), and it correlated significantly between tissues (needles vs bark, F1,14 = 9.4, < 0.01, = 0.63; bark vs wood, F1,14 = 34.5, < 0.01, = 0.84), indicating a slow transfer of pinitol/myo-inositol among tissues.


Concentrations of NCC

In all tissues, the relative concentrations of the different NCC and their changes during the experiment (Fig. 1) were similar to those reported in previous studies, which were mainly carried out on evergreen conifers (Kandler et al., 1979; Hansen et al., 1997; Höll, 1997; Wiemken & Ineichen, 2000), but also on L. decidua growing in lowland Switzerland (Hofstetten 500 m asl; Hoch et al., 2003). NCC concentrations in all larch tissues were higher in the current study than in lowland Switzerland (Hoch et al., 2003). Pinitol/myo-inositol concentrations in larch branch wood at Stillberg were about four times higher than in larch branch wood in lowland Switzerland (Hoch et al., 2003). This could be a result of stress induced by low temperatures, as pinitol indicates a long-term adaptation response to high levels of hydroxyl radicals (Adams et al., 1992; Orthen et al., 1994). In addition, we found lipid concentrations about twice as high in larch branch wood at the treeline as in lowland branch wood of the same species (Hoch et al., 2003). Increasing accumulation of lipids with altitude is an indication for sink limitation, as discussed previously for pine trees (Hoch & Körner, 2003). Compared with lowland larch, we noted an apparently negligible carbohydrate remobilization from needles in autumn at our treeline site. Although Hoch et al. (2003) found a decrease in non-structural carbohydrate concentrations from June to October (including glucose, sucrose, fructose and starch) in larch needles, from 160 to 75 mg g−1 DM, we only observed a minor decrease in the same sugars and starch in larch needles from 160 to 143 mg g−1 DM (from July to September; in both studies, the final sampling took place c. 2 wk before the end of the growing season). This is surprising, as larch trees were previously thought to retract most of their non-structural carbohydrates from needles in autumn (Gower & Richards, 1990). However, non-structural carbohydrate concentrations in branch wood in autumn were higher at Stillberg (81 mg g−1 DM) than in larch branch wood in lowland Switzerland (50 mg g−1 DM; Hoch et al., 2003), thus indicating a highly reduced capacity for storing additional C. This interpretation is supported by a study of Körner et al. (2005), in which excess C of deciduous trees exposed to elevated CO2 concentration was allocated to the decomposer pathway by non-structural carbohydrate-enriched litter.

Carbon transfer between tissues

Lower sink activity at the treeline was further reflected in slower C transfer rates relative to lowland larch trees. For L. decidua (at the same site as Hoch et al., 2003), Keel et al. (2007) found the following transfer velocities from needles to bark and wood: after 9 h, 60% of the initial 13C label remained in needles (Stillberg, 84%; calculated from Eqn (Eqn 4)), whereas 10% of the 13C label appeared in bark (Stillberg, 7%) and 2% in wood (Stillberg, 3%). Different stages of phenological development between these sites in July (time of labeling for both sites) may partly explain the faster C transfer in the lowland site. Although needle expansion on existing larch shoots was complete at both sites, the shoot expansion of lowland larch trees was most likely further advanced than at Stillberg at this time of the year, thus inducing higher basipetal transport of photosynthates (Schneider & Schmitz, 1989; Kozlowski, 1992; Kagawa et al., 2006). For Larix gmelinii saplings growing north of Yakutsk (Russia), Kagawa et al. (2006) observed a slower C transfer from needles in the middle of June: after 4 d, 64% of the initial 13C label remained in needles (Stillberg, 29%). However, in June, needles in Yakutsk were just foliating, probably incorporating 13C label into their structural tissue. In fact, buds at Stillberg showed a similar temporal development of the 13C label: after 4 d, 64% of the initial 13C label remained in buds. Although these comparisons indicate that C transfer velocities in larch trees growing in Yakutsk and Stillberg were similar and slower than in lowland Switzerland, they also reveal the importance of phenological development on C transfer.

13C label in NCC

Combining 13C labeling with subsequent compound-specific isotope analysis allowed us to quantify the mean residence times and the time for label appearance in different NCC (Table 2). Although we found no earlier study investigating these parameters on the NCC level for adult trees, our findings are in line with current physiological knowledge (Hansen et al., 1997; Dennis & Blakeley, 2000; Magel et al., 2000; Brüggemann et al., 2011). In general, the initial 13C labels of individual needle NCC were dependent on their dilution with older unlabeled NCC, whereas the mean residence times of the 13C labels were dependent on the transfer velocities to other tissues or allocations to other NCC or sinks (growth, respiration or root exudates). We found the following order of 13C label dilution with older NCC: sucrose had the lowest dilution with old sucrose, followed by glucose, starch, pinitol/myo-inositol and lipids (highest mixing with old lipids).

Our data revealed a fast transfer of sucrose between needle and bark, which can be explained by the function of sucrose as the main transport form of carbohydrates in larch (Dennis & Blakeley, 2000), and a slow transfer of pinitol between all tissues. For 5-yr-old lowland L. decidua, Schneider & Schmitz (1989) found faster transfer velocities of needle sucrose than in our treeline trees. They labeled branches with 14C in June and found that the percentage of label in needle sucrose, compared with total needle sugar, dropped from 55% to 8% during the first day after labeling, whereas, at Stillberg, it dropped from 91% to 77% within 1 d. The slower transfer of sucrose in the current treeline study further indicates lower sink activities (growth, root exudates and respiration) at the treeline than in the lowland.

In needle starch, there was still a substantial 13C label left 71 d after the pulse labeling. The combination of a fast decline but persistent presence of the label in starch is probably a result of the dual purpose of starch for long-term and transitory storage in needles. When photosynthesis exceeds growth and maintenance demands, long-term storage starch is produced as an overflow mechanism (Rasse & Tocquin, 2006). Transitory starch builds up during the day when sucrose transfer cannot keep up with production (Harley et al., 1992; Hansen et al., 1997). At night, this starch is hydrolyzed to sucrose, which is transferred from the needles to other tissues. This close link between transitory starch and sucrose is revealed by very similar mean residence times of these two compounds (starch, 2.2 ± 0.9 d; sucrose, 2.3 ± 0.3 d), and was further reflected in the mean residence time for bulk needles (1.7 ± 0.2 d) mostly defined by the sucrose 13C label.

Neutral lipids are also thought to be an overflow mechanism at times of high assimilation (Hoch & Körner, 2003). They are not remobilized during normal growth (Hoch et al., 2002). Thus, the apparent increase in 13C label in lipids observed over time possibly reflects such an oversupply of photosynthates at the treeline, and the consequent conversion of osmotically active sugars to (osmotically inactive) neutral lipids.

Treatment effects on carbon allocation

Elevated CO2 caused a consistent increase in photosynthetic C uptake of larch trees at the Stillberg site by over 40% during nine growing seasons (Dawes et al., 2013). Before the FACE treatment stopped at the beginning of October 2009, larch trees were exposed to 580 ppm CO2 for over one-quarter of their lifetime. In contrast with our hypothesis, neither NCC concentrations in any measured tissue nor C transfer were affected by the previous FACE treatment. As NCC concentrations in wood and bark of Stillberg larch trees were not measured during the FACE experiment, it remains unclear whether elevated CO2 ever had an immediate effect on NCC concentrations in these tissues. Sugar and starch analyses of larch needles during the FACE experiment revealed no long-term effect of elevated CO2 on NCC concentrations (Dawes et al., 2013), whereas trees exposed to CO2 enrichment showed an increase in tree ring growth which accumulated to 33% over the 9 yr of FACE (Dawes et al., 2011b). These growth responses were more pronounced in years with relatively high spring temperatures and early snowmelt, suggesting that trees were only able to use extra C assimilated under elevated CO2 in years when growing season temperatures were favorable (Dawes et al., 2011b).

However, there seems to be an apparent contradiction in our findings. Although growth stimulation by the CO2 enrichment suggests that C availability was limiting larch growth at Stillberg, we found higher NCC concentrations and slower C transfer relative to lowland larch, indicating that larch trees at the treeline have low sink activity and therefore experience C saturation. This contradiction can probably be resolved by considering the temporal temperature pattern throughout the growing season. Although growth was probably stimulated by increased CO2 concentration only at times when temperature was not limiting (Tmean = 10°C in June–July; average from 1996 to 2010), larch trees were most probably not capable of storing additional C at times of low sink activity towards the end of the growing season (Tmean = 5°C in September–October; average from 1996 to 2010 at WSL climate station, Stillberg). This assumption is supported by our findings that larch trees on Stillberg retract carbohydrates only partially before shedding their needles in autumn (Fig. 2), thus allocating excess C to litter.

As a result of the importance of temperature for C use and storage in trees at the alpine treeline (Körner, 2006), we expected a strong response of sink activity to soil warming. However, we found no general warming-induced increase in C transfer. One possible explanation could be that warming only increased soil temperature without affecting canopy temperature. Thus, warming did not induce a significant above-ground growth effect for larch trees at Stillberg between 2007 and 2009 (Dawes et al., 2011a). However, we found a faster depletion of the 13C label in larch wood and bark NCC exposed to warmed soil, but not to elevated CO2, leading to a significantly reduced label after 71 d. This was probably a result of slightly higher shoot growth in this treatment group, as these trees also had longer annual lateral shoot lengths at the end of the growing season in 2010 (see Fig. 4; method in: Dawes et al., 2011b). In support of this explanation, we found significant negative correlations for mean lateral shoot length vs 13C label magnitude after 71 d for bark sugars, bark starch, bulk wood and wood lipids (Fig. 4). Our results indicate that temperature fluctuations between years and throughout the season have a greater effect on C transfer through larch trees growing at the treeline than our soil warming experiment.

Figure 4.

Correlations of 13C label of (a) bark sugars, (b) bark starch, (c) wood bulk and (d) wood lipids (samples taken on 15 September 2010) against mean lateral shoot length (measured in September 2010). Points indicate individual Larix decidua trees exposed to one of the possible treatment combinations. Warming treatment: unwarmed (open symbols); warmed (black symbols). CO2 treatment: ambient CO2 (circles); elevated CO2 (triangles). Linear regressions over all trees (all treatments; = 16) are shown as black lines: bark sugars (F1.14 = 5.3, = 0.04, r = −0.53); bark starch (F1.13 = 5.4, = 0.04, = −0.54); bulk wood (F1.14 = 5.5, = 0.03, r = −0.53); wood lipids (F1.14 = 9.0, < 0.01, r = −0.62). The 13C label is expressed relative to the initial needle bulk label (Excess13Crelative).


The use of compound-specific isotope analysis provided new insights into C source–sink relations of treeline L. decidua on NCC resolution, but did not reveal persistent effects of elevated CO2 and soil warming on NCC transfer and concentrations. Slower C transfer and considerably higher NCC concentrations of larch trees at the treeline relative to the lowland clearly indicated that the physiology of these treeline larch trees is temperature limited, and thus their sink activity is generally low. Low remobilization of carbohydrates from larch needles before needle fall in autumn indicated a highly reduced capacity for the storage of additional NCC during cold periods. Only during warm periods in summer were these trees able to utilize additionally acquired C under elevated CO2 for slightly increased tree ring growth.


We thank C. Lötscher for her support in the laboratory and for her encouraging spirit, L. Läubli for grinding and weighing plant material, and the colleagues of the Laboratory of Atmospheric Chemistry (PSI) and the Grassland Sciences Group (ETH) for numerous discussions. Special thanks go to the Laboratory of Chemical Ecology and Ecosystem Research at the University of Vienna (A. Richter, W. Wanek and B. Wild) for providing the protocols for compound-specific C isotope analysis on sugars. Thanks go to the Institute for Snow and Avalanche Research for contributing to the successful operation of the FACE system and to S. Hättenschwiler and I. T. Handa for establishing the soil warming experiment. This research was supported by the Swiss Secretariat for Educational Research, COST Action 639 ‘Gas budget of soils’, Contract Nr. C07.0033, and by the Swiss National Science Foundation, REQUIP, Contract Nr. 206021_128761.