Multi-isotope labelling of organic matter by diffusion of 2H/18O-H2O vapour and 13C-CO2 into the leaves and its distribution within the plant

. Isotope labelling is a powerful tool to study elemental cycling within terrestrial ecosystems. Here we describe a new multi-isotope technique to label organic matter (OM).We exposed poplars ( Populus deltoides ⇥ nigra ) for 14 days to an atmosphere enriched in 13 CO 2 and depleted in 2 H 182 O. After 1 week, the water-soluble leaf OM (  13 C = 1346 ± 162 ‰) and the leaf water were strongly labelled (  18 O =  63 ± 8,  2 H =  156 ± 15 ‰). The leaf water isotopic composition was between the atmospheric and stem water, indicating a considerable back-diffusion of vapour into the leaves (58–69 %) in the opposite direction to the net transpiration ﬂow. The atomic ratios of the labels recovered ( 18 O / 13 C, 2 H / 13 C) were 2–4 times higher in leaves than in the stems and roots. This could be an in-dication of the synthesis of more condensed compounds in roots and stems lignin vs. cellulose) or be the result of and H and fractionation during transport and biosynthesis. We demonstrate that the three major OM elements (C, O, H) can be labelled and traced simultaneously within the plant. This approach could be of interdisciplinary interest in the ﬁelds of plant physiology, palaeoclimatic reconstruction or soil science. sampling dates, while in the results section we present the data of single sampling dates or average values of the whole labelling experiment (environmental conditions, equilibrium fractionation factors).


Introduction
Artificial labelling with stable isotopes facilitates the observation of bio(geo)chemical cycling of elements or compounds with minor disturbance to the plant-soil systems. It has provided many insights into plant carbon allocation patterns (e.g. Simard et al., 1997;Keel et al., 2006;Högberg et al., 2008), water dynamics (e.g. Plamboeck et al., 2007;Kulmatiski et al., 2010) and soil OM processes (e.g. Bird and Torn, 2006;Girardin et al., 2009) in terrestrial ecosystems. Only a few studies have used labelling approaches with more than one stable isotope, for example, to study the interactions between the carbon and nitrogen cycle (e.g. Bird and Torn, 2006;Schenck zu Schweinsberg-Mickan et al., 2010). However, to our knowledge, isotopic labelling of organic matter (OM) with its three major elements, carbon (C), oxygen (O) and hydrogen (H), has never been done in ecosystem studies before, even though combined 13 C, 18 O and 2 H analyses have been widely used to study plant physiological processes and to reconstruct past climatic conditions (Hangartner et al., 2012;Roden and Farquhar, 2012;Scheidegger et al., 2000;Werner et al., 2012). Similarly, artificial labelling with those isotopes would be useful not only to clarify basic mechanisms related to the plant water-use efficiency or the oxygen and hydrogen signals in tree rings but also to study other OM dynamics in the plant-soil system such as OM decomposition in the soil.

M. S. Studer et al.: Multi-isotope labelling of organic matter
The C, O and H contents of OM have been applied to distinguish major groups of compounds by plotting the atomic ratios O / C and H / C in a van Krevelen diagram (Kim et al., 2003;Ohno et al., 2010;Sleighter and Hatcher, 2007). This approach is based on the distinct molecular composition of organic compounds. For example, the glucose molecule (C 6 H 12 O 6 ) is characterized by high O / C (= 1) and H / C (= 2) ratios and is the precursor of other compounds, such as cellulose ((C 6 H 10 O 5 ) [n] O / C = 0.8, H / C = 1.7; Fig. 3a). Condensation or reduction reactions during biosynthesis lead to other compound groups with lower atomic ratios (e.g. lignin) or similar H / C but lower O / C ratios (e.g. lipids, proteins) compared to glucose. Following the logic of the van Krevelen diagram, we wanted to test whether we can use the isotopic ratios 18 O / 13 C and 2 H / 13 C of the labels recovered in plant-soil bulk materials after labelling the fresh assimilates with those stable isotopes in order to detect the utilization of the labelled assimilates for the synthesis of different OM compounds. With this multi-labelling approach we would gain information about the characteristics of the OM formed by simple isotopic analysis of bulk material. This has several advantages compared to compound-specific analysis, such as being much less laborious and less expensive and yielding integrated information on the bulk OM sampled.
In this study we added the 13 C, 18 O and 2 H labels via the gaseous phase in the plants' atmosphere (CO 2 , water vapour). Pre-grown plants were exposed to the labelled atmosphere continuously for 14 days under laboratory conditions, and the labels added were traced in different plant compartments (leaves, petioles, new stems, stem cuttings, roots) and soil OM at different points in time. We applied a simple isotope mixing model to estimate the fraction of 18 O and 2 H that entered the leaf by diffusion from the atmosphere into the leaf intercellular cavities and plotted the atomic and isotopic ratios of the OM formed in van Krevelen diagrams to test whether the multi-isotope labelling approach can be used to detect changes in the OM characteristics.

Plants and soil
The soil, a Cambisol, was sampled from the upper 15 cm in a beech forest (8 33 0 E, 47 23 0 N; 500 m elevation), coarsesieved (2.5 ⇥ 3.5 cm), and large pieces of hardly decomposed organic material were removed. The soil had a clay-loam texture, a pH of 4.8, an organic C content of 2.8 % and a C / N ratio of 11. The plant pots (volume = 8.2 dm 3 ) were filled with 3018 ± 177 g soil (dry weight equivalent). Fifteen poplar seedlings (Populus deltoides ⇥ nigra, Dorskamp clone) were grown indoors from 20 cm long stem cuttings for 5 weeks before they were transferred into labelling chambers (described below). They were kept in the chamber for acclimatization for 1 week prior to labelling. At the begin-ning of the labelling experiments, the average dry weight of fresh plant biomass (without the original stem cutting) was 3.3 ± 0.1 g and the average total leaf area was 641 ± 6 cm 2 per plant. At the end of the experiment (last sampling) the dry weight was 5.4 ± 1.1 g and the total leaf area was 1354 ± 161 cm 2 . The leaf area was measured with a handheld area meter (CID-203 laser leaf area meter, CID Inc.).

Labelling chamber, procedure and environmental conditions
The labelling chambers (MICE -Multi-Isotope labelling in a Controlled Environment -facility) provide a hermetical separation of the shoots (leaves, petioles and new stems) from the roots, rhizosphere and the soil. The plant shoots are enclosed by one large polycarbonate cuboid (volume 1.2 m 3 ) with a removable front plate and five 2 cm wide gaps in the bottom plate to slide in three plants in each row. Small polycarbonate pieces, Kapton tape and a malleable sealant (Terostat IX, Henkel AG & Co.) wrapped around the stem cuttings were used to seal off the upper from the lower chamber. The belowground compartments (soil and roots) are in 15 individual pots, which are hermetically sealed from the laboratory and aerated with outdoor air. This setup ensures that all plants receive the same labelling treatment and prevents the diffusion of labelled atmospheric gases into the soil.
The environmental conditions in the MICE facility are automatically controlled and monitored by a software programme (based on LabVIEW, National Instruments Switzerland Corp.), The software switches the light sources (xenon lamps, HELLA KGaA Hueck & Co) and different valves (3/2 way, Bürkert-Contromatic AG) on/off. The valves include or exclude instruments in a gas circuit attached to the upper chamber to regulate the CO 2 and H 2 O concentration, which is measured by an infrared gas analyser (LI-840, LI-COR Inc.). The chamber air is fed by a vacuum pump (N 815, KNF Neuberger AG) through perforated glass tubes within a water reservoir to humidify the air or through a Peltier-cooled water condenser to dry the air (Appendix Fig. A1). Further, the chamber air can be fed through a Plexiglas tube filled with soda lime to absorb the CO 2 , or CO 2 is injected from a gas cylinder.
The isotope labels ( 13 C, 18 O and 2 H) were added continuously for 14 days via gaseous phase to the plant shoots. We used CO 2 enriched in 13 C (10 at.% 13 C-CO 2 , Cambridge Isotope Laboratories Inc.) as well as water vapour depleted in 18 O and 2 H ( 18 O = 370 and 2 H = 813 ‰, waste product from enrichment columns at the Paul Scherrer Institute). Thus the labelled gases added were enriched by 8.90 at.% 13 C and depleted by 0.07 at.% 18 O and 0.01 at.% 2 H relative to the ambient air.
The soil moisture was maintained at 100 % field capacity and the relative air humidity was 74 % in order to promote the back-diffusion of water into the leaves. The light intensity was low (80 ± 25 µmol m 2 s 1 photosynthetic active radia-tion), and the CO 2 concentration was kept at 508 ± 22 ppm in order to maintain a high atmospheric carbon supply. The day-night cycles were 12 h and the temperature within the labelling chamber was 31 ± 3 C throughout the experiments.

Sample collection
The plant-soil systems were destructively harvested at five sampling dates (three replicates each) to detect the dynamics of the labelling over time. The first sampling was done 1 day before the labelling experiment started (unlabelled control, referred to as t = 0). Subsequently, plant-soil systems were sampled after 1, 2, 8 and 14 days of continuous labelling.
At each sampling date the plant-soil systems were separated into leaves, petioles, stems, cuttings, roots (washed with deionized water and carefully dabbed with tissue) and bulk soil (visible roots were removed with tweezers). A subsample of six leaves was sampled all along the stem (homogeneously distributed). The uppermost leaves, that were newly formed and completely labelled during the experiment, were excluded, since we wanted to study the tracer uptake and translocation dynamics in already existing leaves prior to the treatment. In one out of the three plant replicates we took two leaf sub-samples from distinct positions along the shoot. We sampled six leaves from the upper half of the shoot and six leaves from the lower half hereafter referred to as "top" and "bottom", respectivelyfvials in a water bath. Leaves, stems, roots and bulk soil were collected in airtight glass vials and frozen immediately at 20 C for later cryogenic vacuum extraction of the tissue water. Cuttings and petioles were dried for 24 h at 60 C.
The tissue water was extracted with cryogenic vacuum extraction by heating the frozen samples within the sampling vials in a water bath at 80 C under a vacuum (10 3 mbar) for 2 h. The evaporating water was collected in U vials submersed in a liquid nitrogen cold trap. After thawing (within the closed U vials), the water samples were transferred into vials and stored frozen at 20 C for later 18 O and 2 H analysis. To study the water dynamics, additional water vapour samples from the chamber air were collected by Peltier-cooled water condensers in external air circuit connected to the plant labelling chamber and analysed for 18 O and 2 H.
The dried plant residues of the cryogenic vacuum extraction were used for isotopic bulk analyses as described below. The water-soluble OM was extracted from the leaves by hot water extraction. Sixty milligrams of milled leaf material was dissolved in 1.5 mL of deionized water and heated in a water bath (85 C) for 30 min. After cooling and centrifugation (10 000 g, 2 min), the supernatant was freeze-dried and analysed for 13 C. 2 H analyses were not possible on the hot water extracts, that represent mainly sugars, due to incomplete equilibration with ambient water vapour (Filot, 2010).

Isotopic and elemental analyses
All samples were milled to a fine powder with a steel ball mill and weighed into tin ( 13 C analyses) or silver ( 18 O and 2 H analyses) capsules and measured by isotope-ratio mass spectrometers (IRMS). The 13 C samples were combusted at 1700 C in an elemental analyser (EA 1110, Carlo Erba) and the resulting CO 2 was transferred in a helium stream via a variable open-split interface (ConFlo II, Finnigan MAT) to the IRMS (Delta S, Thermo Finnigan; see Werner et al., 1999). The samples for 18 O analyses were pyrolysed at 1040 C in an elemental analyser (EA 1108, Carlo Erba) and transferred via ConFlo III interface (Thermo Finnigan) to the IRMS (Delta plus XL, Thermo Finnigan). The samples for 2 H analyses were equilibrated with water vapour of known signature prior to the IRMS measurements in order to determine the isotopic signature of the non-exchangeable hydrogen (as described in Filot et al., 2006;Hangartner et al., 2012). After equilibration the samples were pyrolysed in a thermochemical elemental analyser (TC/EA, Thermo-Finnigan) at a temperature of 1425 C and the gaseous products were carried by a helium stream via a ConFlow II open split interface (Thermo Finnigan) into the IRMS (Isoprime, Cheadle). The amount of exchangeable hydrogen (25-27 %) and oxygen (2-3 %) was measured for the leaf, stem and root tissue using depleted water vapour to equilibrate the samples. The measurement precisions of the solid sample analyses were 0.12 ‰ 13 C, 0.54 ‰ 18 O and 1 ‰ 2 H and were assessed by working standards measured frequently along with the experimental samples. The precisions were lower than reported for measurements of natural abundance, since highly labelled sample material was analysed.
Elemental C, H and N content of solid samples was analysed in an elemental analyser (CHN-900, Leco Corp.), as was the elemental O content (RO-478, Leco Corp.).
The liquid samples from the cryogenic vacuum extraction (tissue water) were pyrolysed in an elemental analyser (TC/EA, Thermo Finnigan) and the evolving CO and H 2 gases were transferred via the ConFlo III interface (Thermo Finnigan) to an IRMS (Delta plus XL, Thermo Finnigan) for oxygen and hydrogen isotope ratio analysis (Gehre et al., 2004). The precision of the liquid sample measurement was ± 0.75 ‰ 18 O and ± 1.59 ‰ 2 H.

Calculations
Isotopic ratios were expressed in delta ( ) notation as the deviation (in ‰) from the international standards Vienna Pee Dee Belemnite (V-PDB, 13 C / 12 C = 1.11802 ⇥ 10 2 ) and Vienna Standard Mean Ocean Water (V-SMOW, 18 O / 16 O = 2.0052 ⇥ 10 3 and 2 H / 1 H = 1.5575 ⇥ 10 4 ). The significance of changes in isotopic signature between the sampling dates and the unlabelled control (t = 0) were statistically tested via t tests performed by R software (R Core Team, 2014).
In the following paragraphs we describe first the calculations for the leaf water source partitioning (Eqs. 1-4). These equations are given for the oxygen isotope ( 18 O), but they also apply for hydrogen ( 2 H). Then we describe the calculations for the relative recovery of the isotopes ( 18 O / 13 C and 2 H / 13 C) in the bulk OM (Eqs. 5-7).
The leaf water isotopic signature (at steady state) can be described by a model of Dongmann et al. (1974) to calculate leaf water H 18 2 O enrichment, a derivative of Craig and Gordon (1965) (Eq. 1). According to this model, the isotopic signature of the leaf water (L) is the result of kinetic (" k ) and equilibrium (" ⇤ ) fractionation processes during evaporation of the source water (S) within the leaves and the backdiffusion of atmospheric water vapour (V) into the leaves as affected by relative air humidity (h).
We used a two-source isotope mixing model (Eq. 2, principles described in Dawson et al., 2002) to assess the contribution of the two main water pools (soil and atmospheric water) to the leaf water based on its isotopic signatures. An overview on the input data for the mixing model is given in Appendix A (Fig. A1).
where 18 O leaf,water is the isotopic signature (in ‰) of water extracted from the leaves at a specific sampling date and 18 O source,1 and 18 O source,2 are the theoretical isotopic signatures of the leaf water if all of the water were to originate from either the soil (source 1) or the atmospheric (source 2) water pool.
The first source, hereafter referred to as "evaporating source", represents the water taken up from the soil by the roots, which is transported via the xylem to the leaf, where it evaporates. The isotopic signature of the evaporating source (Eq. 3) is estimated by the maximum leaf water enrichment that would occur at 0 % relative air humidity, i.e. by the first part of the Dongmann approach (solving Eq. 1 with h = 0).
where 18 O stem,water is the isotopic signature (in ‰) of the water extracted from the stem tissue (approximating the xylem water) and " k and " ⇤ atm are the kinetic and equilibrium fractionation terms, respectively, at the specific sampling date.
The second source, hereafter named "condensation source", refers to the water vapour that diffuses from the atmosphere into the leaves and condenses on the cell walls. The contribution of this source would be maximal at 100 % relative humidity, which results in Eq. (4) when solving Eq. (1) with h = 1.
where 18 O atm,vap is the isotopic signature of the water vapour of the chamber atmosphere and " ⇤ atm is the equilibrium fractionation inside the chamber at the specific sampling date. The signature of the atmospheric water vapour was measured on its condensate ( 18 O atm,cond ) collected in the Peltier water trap, which was therefore corrected with the equilibrium fractionation during condensation inside the Peltier-cooled water condenser (" ⇤ pelt ). The kinetic fractionation due to the difference in molecular diffusivity of the water molecule species (" k = 20.7 ‰ 18 O and 10.8 ‰ 2 H) was estimated according to Cappa et al. (2003) for a laminar boundary layer (Schmidt number q = 2/3; Dongmann et al., 1974). The equilibrium fractionation due to the phase change during evaporation and condensation at different temperatures was calculated as in Majoube (1971) with the conditions present at the specific day. The condensation (dew point) temperature inside the Peltiercooled water condenser (T pelt,DP ) was determined based on the remaining humidity and the air pressure of the air leaving the condenser (details on the calculation are given in Appendix B). The equilibrium fractionation factors during the labelling experiment were on average " ⇤ atm = 8.9 ± 0.2 ‰ for 18 O and 72.7 ± 2.7 ‰ for 2 H at T = 31.3 ± 2.7 C inside the labelling chamber and " ⇤ pelt = 11.1 ± 0.2 ‰ for 18 O and 103.3 ± 3.3 ‰ for 2 H at T pelt,DP = 6.0 ± 2.5 C inside the water condenser.
We compared the distribution of the assimilated labels ( 13 C, 18 O, 2 H) in the leaf, stem and root tissue by its isotopic ratios. Therefore we converted the notation to atom fraction (Eq. 5) according to Coplen (2011).
where 13 C t=x is the isotopic signature (in ‰) of the bulk tissue at sampling date x and R is the ratio of the heavier to the lighter isotope ( 13 C / 12 C) of the international standard V-PDB. The atom fraction of 18 O and 2 H was calculated accordingly but using R V-SMOW as a reference and neglecting the 17 O isotope amount.
For the van Krevelen approach we calculated the elemental ratios. The relative label distribution within the plant organic matter (tissue, OM), hereafter referred to as 18 O / 13 C and 2 H / 13 C ratio, was calculated based on the excess atom fraction measured in each tissue (Eq. 6).
where x E ( 18 O) t=x/t=0 and x E ( 13 C) t=x/t=0 are the excess atom fractions of the labels detected at a specific sampling date (t = x), relative to the unlabelled control (t = 0). Equations (6) and (7) were analogously calculated for the 2 H / 13 C ratio.
In a second step we corrected the isotopic ratios ( 18 O / 13 C and 2 H / 13 C) with the maximum label strength of the precursor of the plant OM, i.e. the maximum label strength of fresh assimilates (Eq. 7). The maximum label strength was approximated by the excess atom fraction relative to the unlabelled control (x E ) measured in the leaf water-soluble organic matter (wsOM) and the leaf water for the 13C and the 18 O and 2 H label strength, respectively. The leaf water is the direct source of hydrogen in assimilates and the indirect source of oxygen via the atmospheric CO 2 dissolved in water (Schmidt et al., 2001(Schmidt et al., , 2003. The oxygen isotope composition of dissolved CO 2 equilibrates immediately with the leaf water signature, whereby carbonic anhydrase catalyses this process and induces a temperature-dependent kinetic 18 O fractionation (Gillon and Yakir, 2000;Uchikawa and Zeebe, 2012). The fractionation was assumed to be constant in this experiment with controlled temperature and was thus omitted by the calculation of the excess atom fraction.

Labelling of the leaf water and water-soluble OM
The 18 O and 2 H label added as water vapour to the chamber atmosphere ( 18 O = 370, 2 H = 813 ‰) was mixed with transpired water, which was isotopically enriched compared to the added label (Fig. 1) 2 H signature (in ‰) of the depleted water label added as water vapour to the atmosphere (solid line), the water added to the soil (dashed line), the resulting water vapour in the chamber atmosphere (black circles) and the extracted leaf (grey circles) and stem water (white circles). Error bars on the leaf water indicate ± 1 standard deviation of three plant replicates.
158 ± 13 ‰ for 2 H after 2 days of labellin with the depleted water vapour (Fig. 1). The leaf water was thus on average depleted by 63 ± 7 ‰ for 18 O and 126 ± 14 ‰ for 2 H compared to the unlabelled leaf water signature and was between the signature of the atmospheric water vapour and the water added to the soil ( 18 O = 9 ± 0, 2 H = 74 ± 2 ‰). This indicates that a substantial amount of the leaf water originated from the atmospheric water pool, suggesting that it entered the leaf via diffusion through the stomata. The depletion of the water within a leaf was dependent on its position on the shoot (Fig. 2c, e). The leaf water of the leaves sampled in the upper half of the shoot was 7 ± 2 and 18 ± 8 ‰ less depleted in 18 O and 2 H than the leaves sampled at the lower half. The isotopic signature of the stem water ( 18 O = 10 ± 0 and 2 H = 74 ± 4 ‰), as well as the root ( 18 O = 6 ± 1 and 2 H = 58 ± 4 ‰) and the soil water ( 18 O = 6 ± 1 and 2 H = 63 ± 3 ‰), was not signifi-   cantly depleted and reflected the signature of the water added to the soil (Fig. 1).
At the second sampling date, the leaf water seemed to be more depleted than the water vapour within the chamber air (Fig. 1). This is the result of different sampling procedures. The leaf sampling was performed at one point in time (3 h after the light switched on), while the atmospheric water vapour collected by condensation represents an average for the previous 24 h. Therefore the depletion of the water vapour is underestimated before the equilibrium of the isotopic signature in the atmosphere was reached. In the following the average values of signatures detected after the equilibrium was reached are given (t = 8 and t = 14). We estimated the contribution of the atmospheric water vapour that enters the leaf by diffusion to the leaf water isotopic signature with a two-source mixing model (Table 1). The results were obtained for the two water isotopes 18 O and 2 H separately. Both indicated a substantial contribution of the atmospheric water vapour to the leaf water isotopic signature, whereby the estimates based on the oxygen isotope yielded a higher contribution (69 ± 7 %) than the hydrogen estimates (58 ± 4 %). The estimates for the leaves sampled at different position on the shoot varied by 5 %, whereas the contribution of atmospheric water to the leaf water was higher in the leaves sampled at the bottom (71 ± 4 % based on 18 O and 60 ± 2 % based on 2 H) than in the leaves at the top (66 ± 2 and 55 ± 0 %, respectively) of the shoots.
The 13 C-CO 2 added (8938 ‰ 13 C) was presumably also strongly diluted by respired 12 C-CO 2 , but we did not measure the isotopic signature of the CO 2 within the chamber air. The leaf water-soluble OM was significantly enriched as early as 1 day after labelling and levelled off towards the end of the experiment. On the last two sampling dates its isotopic signature was on average 1346 ± 162 ‰ 13 C.

Labelling of the bulk OM
All three applied labels could be detected in the plant bulk material ( Table 2). We measured the isotopic signature of the non-exchangeable hydrogen, which was estimated to be 74 ± 1 % of the total OM. After 14 days of continuous labelling, the leaves, petioles, stems and roots were enriched by 650-1150 ‰ in 13 C, and depleted by 4-17 ‰ in 18 O and 6-31 ‰ in 2 H. Thus the plant biomass was significantly labelled even under the extreme environmental conditions (high temperature and low light availability) that were critical for net C assimilation (increasing tissue respiration and reducing photosynthesis, respectively). However, the labelling was not strong enough to trace the OM within the large OM pools of the cuttings and soil OM, in which the change in isotopic signature was close to the detection limit or could not be detected. The measured depletion in 18 O of the bulk soil can be accounted for natural variability, since the same effect has been observed in non-treated soil (data not shown here).
The labelling of the leaf bulk OM occurred in parallel to the labelling of the leaf water and water-soluble OM (Fig. 2). The leaf OM was enriched in 13 C after 1 day (Fig. 2b) and depleted in 18 O and 2 H after 2 days (Fig. 2d, f). The incorporation of the label into the leaf OM was, like the labelling of the leaf water, dependent on the position on the shoot. The biomass of the leaves at the top was more enriched in 13 C (by up to 673 ‰) than the biomass of the leaves at the bottom of the shoots and, in contrast to the leaf water, more depleted in 18 O and 2 H (by up to 9 and 21 ‰, respectively) at the top than at the bottom. This indicates a higher overall assimilation in the leaves at the top of the shoot.

Atomic and isotopic ratios to characterize OM
The atomic ratios of the plant bulk OM were in the range of 13.7-115.4 C / N, 0.70-0.83 O / C and 1.56-1.72 H / C (Table 3). The leaf OM was characterized by the lowest C / N and O / C ratios and concurrently by highest H / C ratios (Fig. 3a). The other plant tissues indicated a linear trend in decreasing O / C and H / C and increasing C / N ratios in the order of stems, petioles, roots and cuttings.
The recovery of the three isotopes varied between the leaf, stem and root tissue, while they were similar between the sampling dates (Fig. 3b). The isotopic ratios of the excess atom fractions were 3.5 ± 0.4 ⇥ 10 3 18 O / 13 C and 5.3 ± 0.5 ⇥ 10 4 2 H / 13 C in the leaves, 1.4 ± 0.1 ⇥ 10 3 Table 1. Diffusion of atmospheric water vapour into the leaf water. 18 O and 2 H signatures of leaf water and its two sources: (i) the evaporating source (Eq. 3), estimated by the stem water signature plus kinetic and equilibrium leaf water enrichment (assuming full evaporation without back-diffusion), and (ii) the condensation source (Eq. 4), assessed by the atmospheric water vapour signature plus equilibrium fractionation to account for the gas-liquid phase change. The contribution of the second source (diffusion and condensation of atmospheric water vapour) to the leaf water (f source,2/leaf, water ) was estimated by a two-source isotope mixing model for 18  18 O / 13 C and 2.9 ± 0.6 ⇥ 10 4 2 H / 13 C in the stems and 1.0 ± 0.2 ⇥ 10 3 18 O / 13 C and 1.0 ± 1.4 ⇥ 10 4 2 H / 13 C in the roots after the equilibrium in the leaf water and watersoluble OM labelling was reached. Thus the 18 O / 13 C ratios were on average 2.6 (±0.2) times lower in the stems and 3.8 (±0.7) times lower in the roots than in the leaves (Table 3) and the 2 H / 13 C ratios were 1.9 (±0.2) and 3.1 (±0.6) times lower in the stems and roots, respectively, than in the leaves. The isotopic ratios were in the range of 0.17-0.43 18 O / 13 C and 0.14-0.23 2 H / 13 C after the correction for the maximum label strength, approximated by the 18 O, 2 H and 13 C excess atom fraction within the leaf water and the water-soluble OM, respectively. The normalized isotopic ratios were thus in the order of magnitude of the atomic ratios reported for OM compounds (Table 3, Fig. 3c) but they were in the range characteristic for condensed hydrocarbons and thus lower than expected for fresh OM.

Diffusion of atmospheric water vapour into the leaf
The strong depletion in 18 O and 2 H observed in the leaf water indicates a high back-diffusion of labelled water vapour from the atmosphere into the leaf. The diffusion is dependent on the gradient between atmospheric and leaf water vapour pressure and the stomatal conductance (Parkhurst, 1994). The higher the atmospheric water vapour pressure (the smaller the gradient), the more water molecules diffuse back into the leaf. The latter is further enhanced the larger the stomatal conductance is (Reynolds Henne, 2007). Here we maintained the atmospheric vapour pressure constant at a high level, ensuring a high back-diffusion at a given stomatal conductance. In our experiment the leaf water 18 O and 2 H signature is determined by (i) the signature and the amount of labelled (depleted) water vapour diffusing into the leaf intercellular cavities; (ii) by the enrichment due to transpiration (kinetic and equilibrium fractionation); and (iii) by the influx of xylem water, which is isotopically enriched relative to the labelled water vapour. The latter is proportionally enhanced by increasing transpiration rates as a result of the diffusion convection process of H 2 O (Péclet effect; Farquhar and Lloyd 1993).
The distinct label signal in the water sampled in leaves at different positions on the shoot indicates differences in the transpiration rate. Meinzer et al. (1997) demonstrated in large poplar trees that shading or lower irradiance leads to lower stomatal conductance and transpiration rates. Thus the back-diffusion in the leaves on the bottom might have been reduced due to lower stomatal conductance. However, the increased transpiration in the leaves at the top led to an even stronger dilution of the isotopic signal in the leaf water due to (i) increased evaporative leaf water enrichment and (ii) the Péclet effect (enhanced influx of xylem water, which was enriched compared to the labelled atmospheric water vapour).
The amount of leaf water that entered the leaf by backdiffusion was estimated to be 58-69 %. This result is in contradiction to the common perception that most of the leaf water is taken up from the soil via roots. However it is Table 2. Multi-isotope labelling of bulk OM. 13 C, 18 O and 2 H signatures (in ‰) of the plant-soil compartments (three replicates ± 1 standard deviation) measured before and after 1, 2, 8 and 14 days of continuous labelling. A significant enrichment ( 13 C) and depletion ( 18 O, 2 H) compared to the unlabelled control (t = 0) is highlighted with * (t test, P < 0.05). The degree of labelling is indicated by the change in the isotopic signature of the last sampling date (t = 14) compared to the control (t=0). in line with the observations made by Farquhar and Cernusak (2005), who modelled the leaf water isotopic composition in the non-steady state and estimated the contribution of atmospheric water to the leaf water to be approximately two-thirds of the total water supply. Althoug, our estimates are based on a modelling approach that does not take into account the Péclet effect or daily fluctuations in the isotopic signatures as described below, our estimates correspond very well to the findings of Farquhar and Cernusak (2005). The model used to estimate the quantitative contribution of the two water sources is based on the measured signature of the leaf water ( 18 O leaf,water ) and the estimated sig- Table 3. Atomic and isotopic ratios of the labelled bulk OM. C/N, O/C and H/C atomic ratios and 18 O/ 13 C and 2 H/ 13 C isotopic ratios (of the excess atom fraction) measured in different plant compartments after the equilibrium in the atmospheric labelling was reached. Indicated are average values of two sampling dates (t = 8 and 14) with three plant replicates each (± 1 standard deviation). natures of the water at the evaporating and condensation site ( 18 O source,1 and 18 O source,2 , respectively). The "dilution" of the (laminar) leaf water with the relatively enriched xylem water through the Péclet effect is included in the 18 O leaf,water . This explains the lower contribution of atmospheric water ( 5 %) estimated in the leaves sampled at the top (due to the Péclet effect resulting from higher transpiration rates) compared to the leaves sampled at the bottom of the shoot.
Some inaccuracy in the two-source mixing model estimates might have been introduced by daily fluctuations in the environmental and labelling conditions. The mixture ( 18 O leaf,water ) was sampled after 3 h of light, whereas the estimation of the two sources ( 18 O source,1 and 18 O source,2 ) is based on daily average values of environmental parameters and the atmospheric water vapour ( 18 O atm,vap ) label strength. In our experiment, fluctuations in 18 O atm,vap were caused by adding the labelled vapour mainly during nighttime, when transpiration was low. Thus the atmospheric label strength was presumably highest before the lights were switched on and gradually diluted during the day by transpired water vapour. Hence the actual 18 O atm,vap at the time of plant sampling was probably more depleted than the measured average signature. Therefore 18 O source,2 and its contribution to the leaf water was slightly overestimated. The effect of the temperature fluctuations (±3 C) via changes in the equilibrium fractionation was minor for the outcome of the mixing model < 1 %.
Nonetheless, the strong depletion of the leaf water in 2 H and 18 O proves that back-diffusion of atmospheric water vapour into the leaf is an important mechanisms for leaf water uptake. This supports the hypothesis that atmospheric water vapour diffusion might be as important as the flux of water from the xylem into the leaf (at least under humid condi-   and (b, c) isotopic ratios of oxygen and hydrogen to carbon within the leaves (black circles), petioles (white circles), stems (black triangles), stem cutting (white triangles) and roots (black squares). The circles overlain on the plots in (a) and (c) indicate atomic ratios characteristic for different compound classes (adapted from Sleighter and Hatcher, 2007). Panel (a) illustrates the atomic ratio of all tissues measured (15 replicates ± 1 standard deviation, panel (b) the isotopic ratios of the 13 C, 18 O and 2 H excess atom fraction (relative to the unlabelled tissues) measured after equilibrium in the labelling (see Fig. 1 and 2) was reached (t = 8 and 14, six replicates ± 1 standard deviation) and panel (c) shows the isotopic ratios after normalization with the maximum label strength detected in the leaf water ( 18 O, 2 H) and water-soluble OM ( 13 C).
tions) and be an important mechanisms for the reversed water flow observed in the tropics (Goldsmith, 2013). Furthermore, these results demonstrate that the leaf water isotopic composition is strongly affected by the atmospheric signature at humid conditions and that thus the applicability of the dual-isotope approach (Scheidegger et al., 2000), e.g. to reconstruct past climate conditions by tree ring analysis, is only valid if the source water and atmospheric vapour 18 O are similar. The back-diffusion of atmospheric vapour at high humidity could be another factor in addition to the evaporative enrichment (as demonstrated by Roden and Farquhar, 2012) to overshadow the effects of stomatal conductance on the leaf 18 O signature.

Tracing OM
The O / C and H / C ratio of the plant bulk material was close to the signature of cellulose (Fig. 3a). The leaves had a lower O / C ratio with a constant high H / C ratio, indicating that its OM contains more reduced compounds such as amino sugars or proteins, which is also supported by its low C / N ratio. The trend of decreasing O / C and H / C ratios observed in the other tissues is in the direction of condensation reactions. This trend most likely indicates the increasing lignification of OM from shoots, to roots, to cuttings. The same trend has been observed in the ratios of the labels added from the leaf, to the stem, to the root OM (Fig. 3b, c). The lower isotopic O / C and H / C ratios in the root and stem tissue compared to the leaf tissue could indicate the utilization of the labelled assimilates for the synthesis of more condensed compounds (e.g. lignin) in those tissues. However, other factors affecting the isotopic ratios of the OM are the maximum label strength, the exchange of hydrogen and oxygen with xylem water during transport and biosynthesis, and the isotopic fractionation during metabolism.
The isotopic ratios (Fig. 3b) were around three orders of magnitudes smaller than the expected atomic ratios of OM (Sleighter and Hatcher, 2007). This is mainly due to the different maximum label strength, which was highest for the 13 C and lowest for the 2 H. After correction for this factor, the isotopic ratios were in the range of the atomic ratios characteristic for condensed hydrocarbons (Fig. 3c). The isotopic ratios might be lower than expected due to inaccurate approximation of the maximum label strength of fresh assimilates (by the leaf water and water-soluble OM) or be the result of 18 O and 2 H label losses during transport and biosynthesis.
One reason for the label loss might be the use of other (more enriched) sources during biosynthesis. For example, O 2 (enriched by 23 ‰ 18 O) has been identified as a further source of aromatic compounds, such as phenols and sterols (Schmidt et al., 2001). However, for hydrogen, water is the only known source (Schmidt et al., 2003) and therefore the use of other O or H sources during biosynthesis can not explain the (major) loss of the 18 O and 2 H label.
Another potential reason could be the kinetic fractionation during biosynthesis that leads to distinct isotopic signatures of different OM compounds (described in Schmidt et al., 2001Schmidt et al., , 2003Badeck et al., 2005;Bowling et al., 2008). However, assuming constant isotopic fractionation during the experiment with constant environmental conditions, the iso-topic ratios would not be affected, since they are based on the excess atom fraction relative to the unlabelled OM.
A third reason for the loss of the 18 O and 2 H label could be the exchange of hydrogen and oxygen atoms with water. O and H exchanges with tissue water during transport and the synthesis of new compounds (as recently discussed for oxygen in phloem sugars and cellulose in Offermann et al., 2011, andGessler et al., 2013). O of carbonyl groups (Barbour, 2007;Sternberg et al., 1986) and H in nucleophilic OH and NH groups or H adjacent to carbonyl groups (Augusti et al., 2006;Garcia-Martin et al., 2001) exchange with water. Thus biochemical reactions lead to different isotopomers of organic compounds (Augusti and Schleucher, 2007). The proportion of O and H exchanged can be considerable; for example, during cellulose synthesis, around 40 % of O and H is exchanged with the tissue water (Roden and Ehleringer, 1999;Yakir and DeNiro, 1990). The exchange with water explains to some extent the stronger relative 18 O and 2 H signal in the leaf OM compared to the stem and root OM, since the leaf water was labelled, while the stem and root water was not. The 18 O / 13 C isotopic ratios in particular were increased in the leaf OM compared to the relations observed in the atomic ratios (Fig. 3a). The leaf OM had the lowest O / C atomic ratios but the highest 18 O / 13 C isotopic ratios of all plant compartments (Table 3). This effect is less expressed for the 2 H / 13 C ratios, since only the fraction of hydrogen that does not exchange with ambient water vapour was measured. The non-exchangeable fraction (74 %) is hydrogen bound to carbon (Filot et al., 2006), which is hardly exchanged with xylem water.

Conclusions
We present a new technique to label OM at its place of formation by the application of labels through the gaseous phase ( 13 CO 2 and 2 H 18 2 O). In this study we were able to show that, in a humid atmosphere, the atmospheric water vapour isotopic signature dominates the leaf water signature due to a strong back-diffusion of water vapour into the leaf. Further, we detected differences in the relative distribution of 13 C, 18 O and 2 H in the leaves, stems and roots. This could indicate the synthesis of different compounds in the particular tissues and thus a change in OM characteristics, but it could also be the result of exchange and fractionation processes during transport and biosynthesis. To further test these two possibilities, a better estimation of the maximum label strength by compound-specific sugar analysis would be needed, which has been further developed for 13 C (Rinne et al., 2012) and for 18 O (Zech et al., 2013) recently but does not yet exist for 2 H analysis.
The multi-isotope labelling technique can be used to assess the amount of vapour diffusing into the leaves and to trace the dynamics of the labelled OM. It could be applied in soil sciences, as for example to track the decomposition pathways of soil OM inputs or in the field of plant physiology and palaeoclimatic reconstruction, for example to further investigate the O and H exchange and fractionation processes during transport and metabolic processes or the importance of the ambient air humidity besides its isotopic composition for the climate signal stored in tree-ring cellulose. Furthermore, the multi-isotope labelling technique has the potential to make changes in OM characteristics visible (e.g. C allocation into the non-structural vs. structural pool), for example after a change in climatic conditions, and to trace the labelled OM during its decomposition within the soil.