INTRODUCTION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

The oxygen isotope composition of plant organic material is known to reflect that of source water and the leaf evaporative conditions at the time the material was formed (Ferhi and Letolle, 1979; Sternberg et al., 1989; Yakir et al., 1990a; Aucour et al., 1996; Saurer et al., 1997; Farquhar et al., 2000). In both wheat (Barbour et al. 2000) and cotton (Barbour and Farquhar, 2000) strong correlations were found between stomatal conductance (g_s) and the enrichment in ¹⁸O of whole leaf material above source water (_l). Current theory describing the level of enrichment in leaf tissue is well supported by these data. However, the theory contains a number of untested elements, one of which (the assumption that Suc reflects water somewhat less enriched than that at the evaporating sites within the leaf) we have attempted to address in this paper.

Water at the sites of evaporation is enriched because the heavier H₂¹⁸O molecule diffuses more slowly and has a lower vapor pressure than H₂¹⁶O. The isotope effects caused by these properties are kinetic isotopic fractionation (_k) and fractionation associated with the proportional depression of water vapor by H₂¹⁸O (*). Farquhar and Lloyd (1993), following Craig and Gordon (1965), relate the steady-state enrichment of water at the evaporation sites above source water (_e) to _k, *, the isotopic composition of atmospheric water vapor relative to source water (_v), and the ratio of ambient to intercellular water vapor pressures (e_a/e_i) by:

(1)

Many studies have reported finding lower enrichment in bulk leaf water (_L) than that predicted by Equation 1 (e.g. Allison et al., 1985; Bariac et al., 1989; Walker et al., 1989; Walker and Brunel, 1990; Yakir et al., 1990b; Flanagan et al., 1991a, 1991b, 1993, 1994; Wang et al., 1998), and there have been two main approaches taken to explain this discrepancy. The first, proposed by White (1983), suggested that leaf water is divided into at least two pools, only one of which is exposed to evaporation. Yakir et al. (1989, 1990b, 1993, 1994) extended this idea, suggesting that only water in the intercellular spaces and cell walls is able to become enriched by evaporation, and that water in the symplastic pool is strongly buffered against short-term environmental changes. The two-pool model gained support recently in a paper looking at variation in leaf water from poplar and cottonwood (Roden and Ehleringer, 1999), where 10% of leaf water was assumed to be unenriched by evaporation.

However, the "pools of water" models do not predict an increase in the discrepancy between _L and _e with increasing transpiration rate, first observed by Walker et al. (1989) and investigated further by Flanagan et al. (1991b, 1994). From these observations, G.D. Farquhar (unpublished results; cited by White [1989]; presented in full by Farquhar and Lloyd [1993]), suggested that the difference between bulk leaf water and that at the sites of evaporation was due to gradients of isotopes within the leaf. The gradients may form because diffusion of enrichment away from the sites of evaporation is opposed by convection of unenriched water via the transpiration stream. The ratio of convection to diffusion over a length (l, in meters) is described by the Péclet number ():

(2)

where v (in meters per second) is the velocity of water movement and D is the diffusivity of H₂¹⁸O in water (2.66 × 10⁹ m² s¹). Velocity is proportional to transpiration rate, which can be expressed as a slab velocity E/C (where E [moles per meter per second] is the transpiration rate and C is the density of water [5.55 × 10⁴ mol m³]) by v = kE/C. The scaling factor (k, the constant of proportionality) is thought to be of the order 10² to 10³. Thus,

(3)

The unknown parameters k and l are combined to give an effective length (L = kl), which is 10² to 10³ times the actual length. L is many times the length of l because the v through a porous medium, such as a leaf, is many times greater than if water moved as a slab (E/C) from the vein to the stomata. The average leaf water enrichment over the scaled effective length (_L) at isotopic steady state is (Farquhar and Lloyd, 1993):

(4)

Equation 4 implies that as E and  increase, _L/_e decreases. The divergence is well represented by 1  _L/_e, as presented by Flanagan et al. (1994).

g_s is the most important plant-mediated influence on leaf water ¹⁸O. Conductance affects a number of parameters in Equation 1: Decreasing g_s decreases * through increased in leaf temperature (T_l); T_l increases as stomata close also drive increases in e_i; and decreasing g_s increases _k. The overall result is that at lower g_s, _e is higher (all other things being equal). Conductance also affects the extent to which enrichment at the evaporating sites diffuses back into the leaf. As stomata close the transpiration rate decreases, resulting in a lower  (Eq. 3), and so higher _L (Eq. 4). In summary, higher enrichment at the sites of evaporation within the leaf due to lower g_s is reinforced by a higher bulk leaf water enrichment predicted by the Péclet effect.

Evaporative enrichment of leaf water is passed on to organic material formed in the leaf by exchange of carbonyl oxygen, with an equilibrium fractionation factor (_wc) resulting in organic oxygen being about 27 more enriched than water (Sternberg and DeNiro, 1983; Sternberg et al., 1986). Insofar as gradients in enrichment away from the evaporating sites are expected, organic material formed in different parts of the leaf or within different organelles within a cell is expected to reflect the isotopic composition of the local water. This means that the very first products of photosynthesis (such as triose phosphates) should reflect chloroplast water ¹⁸O, whereas products formed in the cytoplasm (such as Suc) should reflect cytoplasmic water. Barbour and Farquhar (2000) assume that Suc is in full isotopic equilibrium with cytoplasmic water, which they suggest has the same isotopic composition as _L.

Of the several uncertainties in the above approach, the one of interest in this paper is the assumption that the Péclet effect causes bulk leaf water enrichment to be lower than predicted by the Craig/Gordon-type model. The relevance of the Péclet effect within leaves is not strongly supported in data so far. For example, a large data set has been collected from Phaseolus vulgaris by Flanagan et al. (1991b, 1994), but over the combined data the correlation coefficient of the relationship between E and 1  _L/_e is only 0.26. A more convincing relationship may be found for wheat leaves in Walker et al. (1989) (r = 0.85) but these data consist of just four points, the relationship is strongly dependent on a single point, and the data do not extrapolate to the expected Craig-Gordon value at E = 0. The level of ¹⁸O enrichment in leaf water may be indirectly assessed by measuring ¹⁸O of Suc because Suc will reflect ¹⁸O of the water in which it is formed plus _wc. The portion of leaf water that influences Suc ¹⁸O is that with which biochemical intermediates exchange during Suc synthesis, and it is unclear whether this is water close to _e or _L or water of some other isotopic composition.

Measurement of Suc composition in leaves is destructive and the effects of changes in evaporative conditions on the isotope composition of Suc are likely to be somewhat noisy when measured this way. Recent work by Pate et al. (1998) and Pate and Arthur (1998) show that phloem sap may be bled directly from the trunk of Eucalyptus globulus and that the isotope ratio of phloem sap carbon yields useful information.

In the experiments described in this paper a well-controlled and measured environment surrounding the leaf allows well-constrained predictions of _e to be made. Phloem sap Suc samples were taken from the leaf, which allows an assessment of the extent to which variation in leaf water enrichment is reflected in Suc, and whether the Péclet effect is relevant to Suc.

	RESULTS

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

Leaf water is known to take up to 3 h to reach isotopic steady state, but usually comes to within 1 of the steady-state value within about 35 min, depending on the rate of evaporation (Wang and Yakir, 1995). It is expected that Suc will take longer than leaf water to reach steady state, as the pool of Suc must turn over completely. To determine the length of time required for the isotopic composition of Suc to reach steady state after a stepwise change in VPd, a pilot experiment was run with a leaf environment changed from 16 to 8 mbar VPd.

The phloem sap Suc from the leaf exposed to a decrease in vapor pressure differences (VPd) was found to decrease in ¹⁸O, as expected. Figure 1 shows it took about 3.5 h for Suc to reach isotopic equilibrium after the step change. Based on this pilot observation, in subsequent experiments between two and five Suc samples were taken under overnight conditions and every hour for 6 h after the change in VPd. The leaf used in the pilot experiment was not kept under constant conditions overnight, so it is excluded from further analysis.

View larger version (11K): [in this window] [in a new window]   Figure 1.   The measured change in suc over time with a step change in VPd from 16 to 8 mbar.

Figures 2 to 5 summarize the gas exchange data for each full experiment. The step change in VPd was cleanly achieved for all experiments by changing the flow rate of dry air through the cuvette, with the new VPd reached within 60 min in each experiment.

View larger version (22K): [in this window] [in a new window]   Figure 2.   Change in gas exchange parameters and suc with time for experiment 1. A, VPd; B, ea/ei; C, E; D, gs; E, assimilation rate (A); F, measured suc.

The change in VPd also resulted in changes in other gas exchange parameters. The e_a/e_i decreased with the increase in VPd, as expected. g_s also decreased initially, but in all experiments increased slowly after the decrease and in experiment 1 ended up at the same value as before the step change (Figs. 2D, 3D, 4D, and 5D). Accompanying the reduction in g_s was a slight reduction in the assimilation rate. The increase in VPd also resulted in an increase in E for all experiments except experiment 2. E generally increased sharply to a peak at about 25 min after the change in VPd, then settled to a new, fairly constant rate. VPd, e_a/e_i, and E were stable for at least 1.5 h before the end of the experiment for all cases except experiment 2. 

View larger version (23K): [in this window] [in a new window]   Figure 3.   Change in gas exchange parameters and suc with time for experiment 2. A, VPd; B, ea/ei; C, E; D, gs; E, assimilation rate (A); F, measured suc.

View larger version (23K): [in this window] [in a new window]   Figure 4.   Change in gas exchange parameters and suc with time for experiment 3. A, VPd; B, ea/ei; C, E; D, gs; E, assimilation rate (A); F, measured suc.

View larger version (22K): [in this window] [in a new window]   Figure 5.   Change in gas exchange parameters and suc with time for experiment 4. A, VPd; B, ea/ei; C, E; D, gs; E, assimilation rate (A); F, measured suc.

In all experiments Suc ¹⁸O increased with the increase in VPd. As expected, the greatest increase occurred for experiment 4, when the largest increase in VPd was made (Fig. 5F). From experiments 3 and 4 (Figs. 4F and 5F) it can be seen that no change in _suc is found until at least 30 min after the change in VPd.

As predicted by theory, there was a strong negative correlation between e_a/e_i and average _suc measured during the first and last set of samples. Variation in e_a/e_i explained 96% of variation in _suc across the all four experiments and the pilot experiment (Fig. 6). _suc was found to decrease by 17.0 for 1 mbar mbar¹ increase in e_a/e_i. Figure 6 also shows the discrepancy between measured _suc and that modeled by adding _wc (27) to the Craig-Gordon predicted _e when air temperature (T_a) = 25.5°C. The line fitted to the data and that of the Craig-Gordon prediction intersect at e_a/e_i = 1. 

View larger version (16K): [in this window] [in a new window]   Figure 6.   The relationship between suc and ea/ei. Error bars represent SE of each mean (n = 2-4). The solid line represents a least squares regression; suc  = 52.90-17.04 ea/ei, r = 0.98, P = 0.0005. The dashed line represents the predicted suc from the Craig-Gordon model (Eq. 1) where suc = e + wc, and wc = 27; suc = 65.45-29.57 ea/ei. Note that the lines intersect when ea/ei is 1 and suc  36.

	DISCUSSION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

As expected, _suc increased with increasing VPd. The effects of changes in VPd are 2-fold, but both in the same direction. As VPd increases, e_a decreases resulting in more enriched _e (Eq. 1), and so _suc. VPd increases also result in decreases in g_s. As described in the introduction, decreasing g_s results in increases in _L directly through increases in e_i, but also due to a lower .

The Craig-Gordon model (Eq. 1) predicts a negative linear relationship between ¹⁸O of water at the evaporating sites and e_a/e_i. If Suc is in equilibrium with _e (_suc = _e + _wc, where _wc = 27) and _v = 0 (as in this experiment, see "Materials and Methods"), then a decrease of 29.6 per 1 mbar mbar¹ increase in e_a/e_i is expected (Fig. 6). As predicted, _suc did decrease with an increase in e_a/e_i, but only by 17.0 per 1 mbar mbar¹ increase in e_a/e_i. The difference in slope implies that the discrepancy between measured _suc and _e + _wc decreased as e_a/e_i increased, as expected with a Péclet effect. When e_a/e_i is equal to 1, and with _v = 0 as in this case, Equation 1 becomes _e = *. When e_a/e_i is 1 there will be no transpiration so that the Péclet effect will not occur, and _L = _e. Bottinga and Craig (1969) have shown that * is 9.2 at 25°C so that from Equation 1 at e_a/e_i = 1 _suc should equal 36.2%. As predicted, the Craig-Gordon line and the line fitted to the data intersect when e_a/e_i is 1 and _suc is 36 (Fig. 6).

Recalculating relative humidity (RH) for each experiment from VPd and T_a, a decrease in _suc of 16 per unit increase in RH (i.e. 0.16 per 1% increase in RH) is found (not shown) and the intercept of the line is 52.7. This compares to the relationship between RH and ¹⁸O of growth ring cellulose from the desert tree Tamarix jordanis, which had a slope of 14 and an intercept of 44.4 (Lipp et al., 1996), and the relationship between ¹⁸O of oak cellulose and RH found by Switsur and Waterhouse (1998) where the slope was 12 and the intercept 36.4. The slope of the relationship and the intercept are expected to be higher in Suc because Suc should reflect leaf water evaporative enrichment more closely than cellulose (Farquhar et al., 1998). The slope of the _suc/RH relationship is less negative in this experiment (where _v = 0) than would have been found if Suc had been collected when the plants were growing in the greenhouse (where _v  *). From Equation 1, when _v  * the slope would be about 20, rather than 16 as found here. Assuming a dampening factor between Suc and cellulose of 0.4 (Switsur and Waterhouse, 1998), this would give a 13 decrease in cellulose ¹⁸O per unit increase in RH, which is close to that reported by Switsur and Waterhouse (1998) and Lipp et al. (1996).

Theory presented in the introduction assumes that leaf water has reached isotopic steady state, and that the Suc pool has turned over, so that all Suc collected is formed under the new conditions. The time taken for Suc to reach isotopic steady state after a step change in VPd was about 4 h in these experiments, and was found in all cases except experiment 2. Figure 3F shows that _suc continued to increase in experiment 2 6 h after the step change in VPd. This was probably due to slow increases in g_s and E, which may be the result of condensation within the gas exchange system (S.C. Wong, personal communication). For this reason, experiment 2 is excluded from subsequent comparison with modeled values.

An Assessment of the Relevance of the Péclet Effect to _suc

As described in the introduction, oxygen atoms in Suc exchange with leaf water during Suc synthesis, although so far it has not been clear what the isotopic composition of this water should be. Although the Péclet effect theory is supported by the close relationship between measured and modeled ¹⁸O of whole leaf tissue in cotton (Barbour and Farquhar, 2000), a direct test of the theory is required.

To test whether the proportional discrepancy between _L and _e increases with an increase in E, _L is estimated by subtracting _wc from average _suc for each equilibrium condition (the first and last set of samples for each experiment). _suc  wc (_wc = 27) is divided by modeled _e, and the resulting value is presented as a deviation from unity following Flanagan et al. (1994). Figure 7 shows the strong positive relationship between 1  [(_suc  wc)/_e] and E, as predicted by Péclet effect theory. Theoretical relationships between the fractional difference and E at different effective lengths are also shown in Figure 7. Values from each experiment are joined, and it seems that L is slightly lower for the leaf used in experiment 1 (at about 11 mm) than leaves used in experiments 3 and 4 (both at about 14 mm). The data tend to follow the theoretical curves, strongly supporting the idea of a Péclet effect. As discussed in the introduction, the alternative to the Péclet effect model, the pools of water model used by Roden and Ehleringer (1999), does not predict the increasing proportional discrepancy between _L and _e with increasing E clearly shown in Figure 7.

View larger version (22K): [in this window] [in a new window]   Figure 7.   The relationship between E and the fractional difference between modeled oxygen isotope composition of leaf water at the sites of evaporation (e) and estimated isotope composition of the water with which Suc exchanged (suc  wc). Error bars represent SE of each mean (n = 2-4). Values from the same experiment are joined and labeled for reference. The predicted relationships at different L are plotted as dashed lines.

Comparison of Modeled and Measured _suc

Figure 8 shows the relationship between average measured _suc for each equilibrium condition and the modeled value from average gas exchange parameters during that period. The input parameters for the model were T_l, T_a, g_s, leaf boundary layer conductance (constant at 5 mol m² s¹), e_a/e_i, and _v (taken as zero, see "Materials and Methods") to calculate _e (Eq. 1), and E and a single fitted value of L to calculate _L (Eq. 3). Modeled _suc was calculated from _L by adding _wc (27).

View larger version (17K): [in this window] [in a new window]   Figure 8.   The relationship between mean measured and modeled suc under each equilibrium condition when L = 13.5 mm and wc = 27. Error bars represent SE of each mean (n = 2-4). The line fitted to the data () represents a least-squares regression using the error bars as a weight, and is not significantly different from 1:1. Measured suc  = 0.35 + (0.99 × Modeled suc), r = 0.98.

The best fit of measured on modeled _suc (using average leaf water enrichment, Eq. 4) was found when L was 13.5 mm and _wc was 27, as shown in Figure 8. This gave a slope of 0.99 and an intercept of 0.35. Ninety-six percent of variation in measured _suc was explained by modeled leaf water ¹⁸O, and by noise in _suc measurement (weighting the regression by error bars includes measurement noise in the analysis).

The fractionation between water and carbonyl oxygen, _wc, has been reported to be between 25 and 30, based on comparison of cellulose and the water in which it formed (Sternberg and DeNiro, 1983). Acetone, a compound containing a single exchangeable carbonyl oxygen, was found to be 28 more enriched than the water with which it exchanged (Sternberg and DeNiro, 1983). When looking at a substance containing more than one oxygen that has gone through a carbonyl group, such as Suc and cellulose, an average value for _wc is applicable. This is the case even though slight differences in _wc may occur for different oxygen atoms in Suc, depending on the proximity of other atoms. Although enrichment during exchange of carbonyl oxygen is likely to be the most important process in determining the isotope ratios of individual oxygen atoms within organic molecules, fractionation associated with biochemical reactions cannot be discounted. Any reaction that has more than one possible product may result in enrichment or depletion of particular oxygen atoms. Any such fractionations are included in the average _wc. An average _wc for Suc of 27 was found to give the closest fit of modeled to measured _suc in these experiments.

The single fitted value of L (13.5 mm) is within the range in L estimated for individual leaves (Fig. 7) and compares well to other values of L in the literature. In P. vulgaris Flanagan et al. (1994) found that values of 8.5 and 6.25 mm fitted the data best, whereas a value of 8 mm was fitted to data from wheat (Barbour et al. 2000) and cotton (Barbour and Farquhar, 2000) leaves. These values (all from relatively large-leafed C₃ species) are well within the range estimated from single measurements of _L by Wang et al. (1998) of between 4 and 166 mm for a variety of species, with broadleaf species tending to have lower L.

An interesting question raised by this experiment is whether the water with which the oxygen in Suc exchange has the same isotopic composition as bulk leaf water? Or, to restate the question, whether L for leaf water is the same as L for Suc? This question could be addressed if the experiment were to be repeated with leaf water samples taken at each VPd, and measured _L compared to modeled _e and measured _suc.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

Castor bean (Ricinus communis L.) plants were grown in 9-L pots in potting mix with a slow-release fertilizer (Scotts Osmocote Plus, Sierra Horticultural Products, Heerlen, The Netherlands; approximately 3 g per pot) in a greenhouse with natural light (30°C/20°C day/night temperatures, and at about 75% RH) and watered twice daily. Plants were placed under continuous light in the laboratory (500 µmol m² s¹ at leaf level) for 48 h before the experiment commenced. At the same time the three youngest, fully expanded leaves were trimmed to about 0.01 m². The youngest fully expanded leaf was trimmed to fit inside the gas exchange chamber (described below), and the second and third expanded leaves were trimmed to decrease the leaf area supplying photosynthate to the sinks and therefore increase Suc export from the leaf of interest. Leaf area of the youngest trimmed leaf was measured and the leaf was placed in the gas exchange chamber at 500 µmol m² s¹ irradiance, 350 ppm CO₂, 21% (v/v) O₂, a T_a of 25.5°C, and a VPd of between 8 and 13 mbar.

The gas exchange system used is that described previously by Boyer et al. (1997), but modified to include a bypass drying loop to facilitate rapid change of the vapor pressure of air in the leaf chamber. As described by Boyer et al., the glass-lidded chamber was large enough to allow a leaf of 0.16 × 0.2 m, and was cooled by circulating water from a temperature-controlled water bath through a water jacket at the underside of the chamber. Air within the chamber was stirred with a fan, which produced a boundary layer conductance to water vapor of 5 mol m² s¹. T_l was measured by two thermocouples pressed against the leaf and an average of the two was used to calculate VPd and e_a/e_i. The CO₂ partial pressures and vapor pressures of inlet and outlet air were measured with infrared gas analyzers (model LI-6251, Li-Cor, Lincoln, NE, and Binos 1, Leybold Heraeus, Hahau, Germany). g_s to water vapor, evaporation, and photosynthetic rates were calculated from these measurements as described in von Caemmerer and Farquhar (1981). The leaf was kept in the cuvette from 5 PM to 8 AM under constant conditions, then a step change in VPd was imposed after the first set of phloem sap samples had been taken. The constant conditions from 5 PM to 8 AM were as close as possible between different leaves to allow comparison of _suc under similar conditions from different leaves. The constant condition chosen was one of low VPd so necessarily the step change in conditions was to a higher VPd, except in the pilot experiment.

The oxygen isotope compositions of all samples were measured on a Carlo-Erba preparation system coupled to a Micromass Isochrom mass spectrometer, as described by Farquhar et al. (1997). Isotope compositions () were measured as deviations from Vienna Standard Mean Ocean Water (VSMOW) so that:

(5)

where R_sample and R_reference are the ratios of ¹⁸O/¹⁶O for the sample and the laboratory reference, respectively, and the laboratory reference has been calibrated to VSMOW.

The oxygen isotope composition of substances within a plant is presented as enrichment above source water (_s); this in effect makes _s the "standard." For example, the enrichment in Suc compared to source water (_suc) is given by:

(6)

but is well approximated by _suc = _suc  s.

Samples of Canberra tap water, used in this experiment to water the plants, were found to have an isotopic composition (versus VSMOW) of 7.3 ± 0.2, so this became the standard to which all plant isotope compositions are compared.

The gas exchange system described above was set up so that air entering the cuvette was dried (vapor pressure < 0.001 mbar), meaning that all water vapor in the cuvette came from transpiration. When this is the case, at isotopic steady state, water vapor must have the same isotopic composition as source water for conservation of mass. That is, the isotopic composition of evaporated water (_E versus VSMOW) is equal to _s, and because all water vapor is at _s, _v = 0.

Suc in phloem sap was collected for isotopic analysis by bleeding, as described by Hall et al. (1971). A light score was made with a sharp razor on the petiole of the leaf in the cuvette to allow the phloem to bleed freely without damaging the xylem. Phloem sap was collected in 5-µL capillaries held steady against the cut in a small clamp stand. A single cut bled freely for up to 60 min, with each capillary taking between 1 and 8 min to fill. The phloem sap, the solid component of which is mostly Suc (Hall and Baker, 1972), was then placed in tin capsules and evaporated in a 60°C oven overnight before oxygen isotope analysis.

Relationships between parameters were established from standard least squares regression and product-moment correlations, using error bars as a weight where appropriate, in Origin 5.0 (Microcal Software, Northampton, MA). With 4 degrees of freedom available, Pearson correlation coefficients greater than 0.81 and 0.92 are statistically significant at the 5% and 1% levels, respectively.