Summary
Quantification of leaf respiration is of great importance for the understanding of plant physiology and ecosystem biogeochemical processes. Leaf respiration continues in light (RL) but supposedly at a lower rate compared to the dark (RD). Yet, there is no method for direct measurement of RL and most available methods require unphysiological measurement conditions.
A method based on isotopic disequilibrium quantified RL (RL 13C) and mesophyll conductance of young and old fully-expanded leaves of six species compared RL 13C to RL values determined by the Laisk method (RL Laisk).
RL 13C and RL Laisk were consistently lower than RD. Leaf ageing negatively affected photosynthetic performance, but had no significant effect on RL or RL/RD as determined by both methods. RL Laisk and RL 13C were measured successively on the same leaves and correlated positively (r2=0.38), but average RL Laisk was 28% lower than RL13C. Using A/Cc curves instead of A/Ci curves, a higher photocompensation point Γ* (by 5 μmol mol-1) was found but the correction had no influence on RL Laisk estimates.
The results suggest that the Laisk method underestimated RL. The isotopic disequilibrium method is useful for assessing responses of RL to irradiance and CO2, improving our mechanistic understanding of RL.
INTRODUCTION
Foliar respiration is a major component of the global carbon cycle, releasing more than three times the amount of CO2 liberated by anthropogenic emission each year (Le Quere et al., 2009; Beer et al., 2010), if it is assumed that foliar, i.e. plant leaf respiration constitutes 50-80% of plant respiration globally (Atkin et al., 2007; Lehmeier et al., 2010). Thus knowledge of the drivers and controls of leaf respiration is essential for understanding plant physiology and the global carbon budget, and that knowledge is required for improving the representation of leaf respiration in climate-vegetation models (Atkin et al., 2007; Heskel et al., 2013). The fact that leaf respiration rate is lower in light (RL, also termed day respiration) compared to the dark (RD) – when normalized to the same temperature – has long been recognized and demonstrated in leaf- (Brooks & Farquhar, 1985; Atkin et al., 2000; Gong et al., 2015), stand- (Schnyder et al., 2003; Gong et al., 2017a), and ecosystem-scale (Wehr et al., 2016) studies. The inhibition of respiration by light is underpinned by the light-induced down-regulation of the activity of several enzymes of respiratory metabolism (Tcherkez et al., 2005; Tcherkez et al., 2012a). Yet, the quantification of RL is technically challenging and the mechanism controlling its variation is uncertain.
In practice, RL cannot be directly measured using conventional gas exchange measurements because RL is masked by other concurrent, major fluxes: photosynthetic CO2 uptake and photorespiratory CO2 release. Net CO2 assimilation rate can be expressed as: A = Vc - 0.5 Vo - RL, where Vc is the rate of carboxylation and Vo is that of oxygenation, and 0.5 Vo is the rate of photorespiration (F). A can be further expressed as: where Γ* is the CO2 compensation point in the absence of day respiration and Cc is the chloroplastic CO2 mole fraction. At a Cc that equals to Γ*, A is equal to –RL. Based on Eqn 1, RL can be estimated from the common intersection of curves of net CO2 assimilation rate (A) vs. intercellular CO2 concentration (Ci) measured under low CO2 and sub-saturating levels of irradiances (defined as RL Laisk here), as described in Laisk (1977) and further extended by Brooks & Farquhar (1985). This is based on the notion that at RL is insensitive to light intensity. The Laisk method uses A/Ci curves instead of A/Cc curves to determine the common intersection, and thus gives the apparent, Ci-based CO2 compensation point and RL at . Although it has been widely used as a standard method for determining RL and Γ* (in assuming ) (Farquhar et al., 1980; von Caemmerer, 2000; Walker & Ort, 2015), uncertainties and limitations of the Laisk method have been intensively discussed. First, ignoring the influence of mesophyll conductance (gm) might lead to errors in estimates of RL and Γ*, as (Brooks & Farquhar, 1985; von Caemmerer et al., 1994; Walker & Ort, 2015). Second, the measurement must be performed at very low CO2 that generally contrast with growth conditions (Villar et al., 1994; Yin et al., 2011). Experimental evidence has indicated a CO2 effect on respiration rate in light (Gong et al., 2017a) and on the abundance of transcripts encoding enzymes of the respiratory pathway in both long-term (Leakey et al., 2009) and short-term (Li et al., 2013) treatments. These observations raise the concern that RL measured by the Laisk method might differ from actual RL under growth conditions. Similarly, other methods, such as the Kok method (Kok, 1948) and a method based on chlorophyll fluorescence (Yin et al., 2011), generally must be performed at low CO2 levels or low irradiance levels and require manipulation of CO2 assimilation rate (for a review see Yin et al., 2011). Furthermore, during both Kok and Laisk measurements, variations in Cc are critical but have not been accounted for, potentially leading to errors in RL estimates (Farquhar & Busch, 2017; Tcherkez et al., 2017a b).
Techniques that allow measuring RL without requiring modifications of environmental conditions such as CO2 mole fraction or irradiance typically use carbon isotopes. The principle of deconvoluting CO2 flux components by artificially created isotopic disequilibrium (i.e. labelling) has been widely explored for i.e. photorespiration (Ludwig & Canvin, 1971) or stand- (Schnyder et al., 2003; Gong et al., 2017a) or ecosystem-scale (Ostler et al., 2016) autotrophic respiration. This type of labelling method exploits the fact that CO2 flux components have distinct dynamics of tracer incorporation during the labelling. Abrupt changes to a 13CO2 atmosphere were used to monitor the liberation of 12CO2 by respiration in the first minutes following the isotopic changeover (Loreto et al., 2001; Pinelli & Loreto, 2003). However, when using pure 13CO2 this technique is relatively costly and requires a 13C-sensitive infrared gas analyzer. Gong et al. (2015) described a leaf-level isotopic disequilibrium method to quantify RL using CO2 sources of natural 13C abundance, which is based on concurrent measurements of photosynthetic gas exchange and 13C/12C isotope composition (denoted as δ, definition see methods) of CO2 fluxes, i.e. online 13C discrimination by net photosynthesis (online Δ). In other words, the δ-value of gross fixed CO2 (associated with the flux Vc) responds instantaneously at the onset of labelling (i.e. abrupt change of δ of CO2 fed to leaf), with the δ-value of the photorespired CO2 (flux 0.5 Vo) following with only a short delay (half-life in the order of a few minutes (Ludwig & Canvin, 1971)). By contrast, the δ-value of respired CO2 responds rather slowly (half-life in the order of one to a few days (Schnyder et al., 2003; Lehmeier et al., 2008; Tcherkez et al., 2012b; Gong et al., 2017a). This approach requires two sets of online Δ measurements on similar leaves (or the very same leaves, as in this study), so as to examine the isotopic mass balance at the photosynthetic steady-state (Gong et al., 2015). This method has the following advantages: (i) RL measurements can be done at any setting of environmental parameters, e.g. identical to growth conditions; (ii) it measures RL at the photosynthetic steady-state without manipulation of the photosynthesis rate; (iii) it simultaneously provides a reliable measurement on mesophyll conductance (gm), another important parameter. As it relies on the measurements of δ-values and CO2 exchange rates, diffusive leaks across the gasket of leaf cuvette must be minimized (Gong et al., 2017b) or accounted for (Gong et al., 2015).
Here, we use the isotopic disequilibrium method (presented by Gong et al. 2015) to measure RL (RL 13C) on single leaves, and compare the results with the Laisk method (RL Laisk) applied to the very same leaves. Thus, our objectives were to answer the following questions. (i) Does the isotopic disequilibrium method also show an inhibition of leaf respiration by light? (ii) Do RL estimates from isotopic disequilibrium agree with those from the Laisk method for different species and leaves of different age effects? Or (iii) is there any consistent offset in RL estimates obtained with the two methods, and if yes, is the offset correlated to leaf age or simply due to assumptions on internal/mesophyll conductance? To this end, 13CO2/12CO2 exchange of leaves from plants grown with ambient CO2 with a δ13C of CO2 (δ12CCo2) near −10‰ was measured sequentially in the presence of CO2 with a δ13CCO2 of −31.2‰ and −6.3‰, and RL of leaves was solved using isotopic mass balance equations. These measurements were immediately followed by determinations of RL Laisk. The comparison of RL 13C and RL Laisk was performed on both young and old mature leaves of two grass and four legume species. Villar et al. (1995) have reported that ageing of leaves of an evergreen shrub led to a reduction of RL Laisk/RD from 0.5 to 0.2. This is the reason why we included young and old leaves, since it might increase the variation range of RL and thus enhance the method comparison. In addition, we estimated gm of every leaf, so that A/Ci curves could be converted to A/Cc curves to estimate Γ* and RL Laisk based on the common intersection of A/Cc curves.
MATERIALS AND METHODS
Plant material and growth conditions
Six herbaceous plant species were used, namely barley (Hordeum vulgare), wheat (Triticum aestivum), castor bean (Ricinus communis), French bean (Phaseolus vulgaris), soybean (Glycine max) and broad bean (Vicia faba). Plants were grown from seed in plastic pots filled with quartz sand, placed in a growth chamber (PGR15, Conviron, Winnipeg, Canada) and supplied with a modified Hoagland nutrient solution with 7.5 mM nitrate (cf. Gong et al., 2017b) every two to three days. Environmental conditions during plant growth were: a photosynthetic photon flux density (PPFD) of 700 μmol m-2 s-1 during the 12 h-long photoperiod per day, ambient CO2 concentration ([CO2]) of about 400 μmol mol-1, air temperature of 22 °C during photo- and dark- periods, relative humidity of 50% during photoperiod and 60% during dark period. The density of plants in the growth chamber was rather low, thus leaves were not shaded. Young leaves, defined as the youngest fully expanded leaves, were measured when plants reached a stage of having 3-4 mature leaves per branch/tiller. Old leaves, defined as two age categories older than the measured young leaves, were measured about 10 days later. At that time plants had 5-7 mature leaves per branch/tiller. For dicots, the fully expanded terminal leaflets were measured. Young leaves were measured for all species, while old leaves of G. max and V. faba were not measured.
13CO2/12CO2 gas exchange facilities
13CO2/12CO2 gas exchange and labelling were performed using the protocols and facilities described in Gong et al. (2015) with modifications and advancements as follows. The approaches in Gong et al. (2015) provided a mean leak coefficient and a RL/A for a group of similar leaves (same species and age, treated as replicates). In this study, leak coefficients were measured for each leaf and used for the correction of its gas exchange data, using the equations in Gong et al., (2015). To quantify RL, the two components of A must be separated, as A = P – RL, where P is the apparent photosynthesis rate (P = Vc – F). Briefly, we switch the CO2 source supplied to leaf photosynthesis to create isotopic disequilibrium between P and RL, namely, P will be immediately labelled while RL is fed by substrate formed during plant growth (old carbon) (Gong et al., 2015), thus RL can be solved by isotopic mass balance (see below).
The leaf-level 13CO2/12CO2 gas exchange and labelling system included a portable CO2 exchange system (LI-6400, LI-COR Inc., Lincoln, USA) housed in a gas exchange mesocosm (chamber 1, cf. Gong et al., 2015; Gong et al., 2017b), and another gas exchange mesocosm (chamber 2) for the purpose of providing labelling CO2. The air supply to both mesocosms and the LI-6400 was mixed from CO2-free, dry air (with 21% O2) and CO2 of known δ13CCO2 (cf.(Schnyder et al., 2003), with δ13C denoting the 13C composition of a sample defined as the relative deviation of its 13C/12C ratio (ℜsample) to that of the international VPDB standard (ℜVPDB): δ13C = ℜsample/ℜVPDB − 1. [CO2] inside chamber 1 was monitored with an infrared gas analyzer (LI-6262, LI-COR Inc., Lincoln, USA). During leaf gas exchange measurements, the plants to be measured and the sensor head of the LI-6400 were placed inside the chamber 1. Using this setup,we separately controlled the CO2 concentration and δCco2 in the leaf cuvette and growth chambers. The growth chamber and leaf cuvette systems were coupled to a continuous-flow stable isotope ratio mass spectrometer (IRMS; Deltaplus Advantage equipped with GasBench II, ThermoFinnigan, Bremen, Germany) for 13C analysis of the sample air. The whole-system precision of repeated measurements on δ13C was 0.09‰ (SD, n=50). For further details of the method see Gong et al. (2015) and Gong et al. (2017b).
Determinations of KC02, RD
Measurements of each leaf started with the determination of the cuvette leak coefficient for CO2 (KCO2) with the leaf present in the cuvette during these measurements (Gong et al., 2015). Each leaf was held in the leaf cuvette of the LI-6400 for more than 20 min in the dark, at a constant [CO2] of 488 ± 9 (SD) μmol mol-1 in the leaf cuvette (Cout) and 400 μmol mol-1 in the chamber 1 (CM) that housed the LI-6400 measurement head (detailed measurement conditions are shown in Table S1). When gas exchange had reached a constant rate, gas exchange parameters, including [CO2] and the δ13C of the incoming (Cin and δin) and outgoing cuvette air (Cout and δout) were measured with the LI-6400 and the online IRMS. Thereafter, CM was reduced to about 200 μmol mol-1 and the same gas exchange parameters were measured at steady-state. Since manipulating CM should only affect the diffusive leak between the chamber 1 housing the leaf gas exchange equipment and the internal space of the leaf cuvette but not RD, KCO2 was determined as the slope of the observed net CO2 exchange rate in the dark (ND) and (CM − Cout)/s relationship as: where s is the leaf area (Gong et al. 2015). Knowing the KCO2 of each intact leaf, CO2 exchange data were corrected as shown in Gong et al. (2015) and RD determined. Since leak coefficients for 12CO2 and 13CO2 were virtually the same (Gong et al., 2015), KCO2 was used to correct both 12CO2 and 13CO2 flux data. Before all calculations, data of δ and rates of CO2 fluxes were corrected for leak artefact using KCO2 of individual leaves and equations in Gong et al. (2015).δ13C of RD was calculated as: with δin and δout are δ measured at inlet and outlet air stream, respectively.
13C labelling
After measurements of KCO2, RD and δRD, the light source of the LI-6400 was switched on (PPFD 700 μmol m-2 s-1) to measure the online Δ using CO2 sources with different δ13C of CO2 (−6.3‰ and −31.2‰). The 12C/13C discrimination associated with net photosynthesis, ΔA, was calculated according to Evans et al. (1986): ΔA = ξ(δout − δin)/(1+ δout − ξ(δout − δin)), where ξ = Cin/(Cin − Cout). Here, ξ was below 15 during ΔA measurements. Measurements of ΔA were done in the photosynthetic steady-state: after about 30 min of stabilization in the conditions similar to that of plant growth average Cout of 394 ± 34 (SD) μmol mol-1, average relative humidity of 76 ± 10 %, block temperature of 22 °C (mean leaf temperature was 23.3 ±0.2 °C, Table S1). Online Δ was firstly measured using the depleted CO2 source (−31.2‰), then measured with the enriched CO2 source (−6.3‰) on each leaf. Chamber 2 was used to mix the labelling air containing the enriched CO2 with the targeted [CO2]. When labelling start, well mixed air in Chamber 2 was supplied to the inlet of LI-6400 with a peristaltic pump. Using this setup, the labelling air can completely flush out the air in the LI-6400 system within 8 min. The second online Δ was measured within 15 min after the start of labelling (i.e. switching of CO2 sources), and all photosynthetic gas exchange rates are not influenced by labelling, as only δ13C of CO2 fed to leaf was changed (Gong et al., 2015).
Calculations of RL
Substituting the relationship giving the photosynthetic assimilation in the absence of day respiration P (= Vc − F) into equation (1) gives:
Applying isotopic mass-balance to equation (4) gives: where δP, δA, δRL are the δ13C of P, A and RL, respectively. With the two sets of online Δ measurements we have: where subscripts “d” and “e” indicates parameters measured with the 13C-depleted and 13C-enriched CO2 sources, respectively. Since 13C discrimination in P (ΔP), is independent of the δ13C of the CO2 source (Farquhar et al., 1989):
Combining the rearranged Eqn 6-8 we have:
Equation (9a) includes the isotope composition of day-respired CO2 (both under a 13C-enriched and 13C-depleted atmosphere) in the denominator. Under the assumption that day respiration reacts very slowly to photosynthetic input (see Introduction), δRL d = δRL e and equation (9a) rearranges to:
In practice, the approximation δRL d = δRL e is not critical: if some C atoms photosynthetically fixed under the 13C-depleted atmosphere were channelled to respiratory metabolism and liberated as CO2 under the 13C-enriched atmosphere, this would lead to a change of a few per mils only in the denominator and the change in RL 13C would be very small. In fact, during the first measurement phase (≈ 20 min) under the 13C-depleted atmosphere, we expect at most 10% turnover in leaf respiratory pools (measured by Nogues et al. (2004) for dark respiration) meaning a maximal putative change in δRL of about 0.6‰ (Table S2, the denominator in equation 9a would thus be equal to 0.975 instead of 1).
In Gong et al. (2015), the approximation that 1+ΔP=1 was used. Here, we applied a different approximation that ΔP = ΔA e, which was shown to be an acceptable approximationwhen the enriched CO2 source (−6.3‰) was close to that of the growth environment (−10‰) (Gong et al., 2015).
Thus, RL was calculated as follows:
The δ-value of net assimilated CO2 was calculated as: δA = (δin Cin − δout Cout)/(Cin − Cout). RL 13C/RD was calculated with a (small) correction accounting for the temperature difference between light and dark, using a Q10 of 2 (see Gong et al., 2015).
Calculation of mesophyll conductance
Mesophyll conductance (gm) is defined as gm = A/(Ci − Cc) (cf. Evans et al., 1986), where Cc is the CO2 mole fraction at the site of carboxylation in the chloroplast. Estimation of Cc was based on the photosynthetic 12C/13C discrimination model of Farquhar et al. (1989) (cf. Gong et al., 2015). In fact, a modified equation of 13C/12C discrimination that includes both mesophyll resistance and ternary effects (Farquhar & Cernusak, 2012) is: while the simplified equation that excludes mesophyll resistance (or assumes infinite gm) can be written as:
Therefore, the subtraction (11a) - (11b) gives: where a = 4.4‰, b = 28.9‰, am combines dissolution and diffusion in the liquid phase so that am = 1.8‰ (Evans et al., 1986) and f = 11‰ (Ghashghaie et al., 2003; Lanigan et al., 2008). Γ*, was approximated to be equal to measured by the Laisk method (see below). t represents the ternary correction factor (Farquhar & Cernusak, 2012): where E is the transpiration rate and gsc is the stomatal conductance to CO2. Here, we ignored the boundary layer resistance because air was well mixed in the leaf cuvette of the LI-6400 (Kromdijk et al., 2010). Δp can be calculated using Eqn 6-8, assuming δRL = δRD (Gong et al. 2015). Each leaf had two measurements of Δp using the two CO2 sources (Δp e and Δp d, and theoretically they should be very similar and can be treated as technical replicates, see also Fig. 3), thus the mean of Δp e and Δp d was used to calculate Cc using Eqn 12.
It should be noted that equation (11a) simply represents the model of photosynthetic fractionation where the term associated with day respiration has been omitted. That is, the full model following Farquhar et al. (1989) notations is: where e is the isotope fractionation by day respiration, with respect to net fixed photosynthates which are assumed to represent the respiratory substrates. However, considering that respiratory substrate pool turn-over is slow and mostly disconnected from photosynthesis at time scales less than the duration of the measurements (30-45min), day respiration is fed by a distinct carbon source and thus equation (14) has to be changed to (Tcherkez et al., 2011):
In Eqn 15, e is still expressed relative to net fixed CO2 (i.e. e = (δA − δRL)/(δRL + 1)). Under our conditions, t is very small (< 0.1‰), thus Eqn 15 can be rearranged as
Measurement of RL and using the Laisk method
After online Δ measurements, each single leaf was measured for RL using the Laisk method (Laisk, 1977; Brooks & Farquhar, 1985) with the LI-6400 open system. Briefly, A/Ci curves were obtained at three levels of PPFD, 50-70, 100-150, and 250 μmol m-2 s-1, and Cout was decreased from 110 to 50 μmol mol-1 step-wise at each PPFD. Average relative humidity was 77±9% and block temperature 22 °C (meaning that leaf temperature was 22.4±0.2 °C, Table S1). Again, the observed A and Ci values were firstly corrected for leak artefacts. The coordinates of the common intersection of A/Ci curves provided the estimates of RL Laisk and (Fig. S4). We also tested the slope-intersection regression approach suggested by Walker & Ort (2015), a modified Laisk method, but it yielded very similar results (data not shown) as the common intersection approach in the original Laisk method.
Using Laisk measurements, we estimated RL Laisk CC and Γ* from the A/Cc curves (cf. Fig. S4). For this purpose, we established the relationship between gsc and gm across the measured leaves, and gsc/gm was plotted against A or Cout to check whether the gsc - to - gm ratio was independent of photosynthesis rate or CO2 mole fraction. Using the gsc/gm relationship, gm along A/Ci curves was estimated from measured gsc, and thus A/Ci curves could be converted into A/Cc curves.
RESULTS
RLacross species and leaf age
As expected, both RL Laisk and RL 13C were consistently lower than RD, demonstrating that the labeling technique generally also shows an inhibition of leaf respiration in the light compared to the dark (Table 1). Further, leaf age had no effect on RL Laisk or RL 13C (Table 1). Also, both methods showed similar species effects: H. vulgare and P. vulgaris had higher RL Laisk and RL 13C than the other species; T. aestivum had the lowest RL Laisk and RL 13C of young leaves and R. communis the lowest RL Laisk and RL 13C of old leaves. Pooling over all RL 13C and RL Laisk paidata, a significant positive correlation was found (r2 =0.38, p<0.001, Fig. 1). Importantly, however, RL Laisk was systematically smaller than RL 13C by 28% (averaged over all leaves), and this effect was similar for the different species and age classes (Fig. 1). As a result, the ratio of respiration in light to that in darkness at the same temperature (RL/RD) was higher for the isotopic disequilibrium method than the Laisk method: RL 13C/RD ranged between 0.6 and 1.3 with a mean of 0.9, and RL Laisk/RD ranged between 0.4 and 0.9 with a mean of 0.7 (Fig. 2). Both measurements showed a tendency of increasing RL/RD with leaf ageing; however, a significant age effect on RL 13C/RD was detected in P. vulgaris while a clear age effect on RL Laisk/RD was found in P. vulgaris, T. aestivum and R. communis (Fig. 2). RD was not significantly different between age classes, but differed between species, with T. aestivum having the smallest RD value of all species.
Photosynthetic parameters
Leaf ageing had clear effects on many gas exchange parameters (Table 1) when averaged across species. Old leaves had an approx. 30% lower net CO2 assimilation rate (A), 58% lower stomatal conductance to water vapour (gsw), 47% lower mesophyll conductance (gm), 11% lower ratio of internal-to-atmospheric CO2 mole fraction (Ci/Ca), and a 19% lower ratio of chloroplastic-to-atmospheric CO2 mole fraction (Cc/Ca), as compared to young leaves. On the other hand, old leaves had a 13% higher and 46% higher intrinsic water-use efficiency (WUEi = A/gsw) compared with young leaves, averaged across species (Table 1). Nevertheless, A and gm in P. vulgaris did not differ significantly between the two age classes. Across individual leaves of all species and age classes, RL was not significantly correlated to A, gsc or gm (r2 <0.1, p>0.05), but was significantly correlated to RD, with both methods .
Isotope fractionation and mesophyll conductance
Carbon isotope discrimination during net CO2 assimilation (ΔA) showed clear differences during the two sets of online Δ measurements (Fig. 3), that is, the observed discrimination was influenced by the isotope composition of inlet CO2. This was due to the isotopic disequilibrium between respiratory (RL) and photosynthetic (P) CO2 fluxes. By contrast, Δp was not influenced by CO2 sources in any species (Fig. 3) supporting the accuracy of flux partitioning of P and RL.Furthermore, the calculation using Eqn 16 yielded estimates of e of −16.5% with 13C-depleted inlet CO2 and +11.2% with 13C-enriched inlet CO2 (averaged across species). Those estimates were close to values that could be simply computed from the δ13C difference between growth CO2 source and outlet CO2 (that is, e = δout − δgrowth CO2 where δgrowth CO2 = −10% and δout denotes the isotopic composition of CO2 in the leaf cuvette during measurements in light), assuming there was no fractionation between photosynthates and respired CO2 (Wingate et al., 2007): e obtained in this way was −18.3% and + 6.1% with 13C-depleted and 13C-enriched inlet CO2, respectively. The agreement between the two calculations of e again indicates that our flux partitioning of P and RL was performed properly.
gm was calculated from carbon isotope discrimination during apparent photosynthesis (Δp) using equation (12). As measured under conditions similar to growth conditions using our isotopic disequilibrium method, gm and gsc showed a strong linear correlation across young and old leaves of all species (gsc=0.67gm+0.01, r2=0.82, p<0.001, Fig. S1). Meanwhile, gsc/gm showed no significant correlation with A (p>0.05, r2 <0.1) or CO2 mole fraction in the leaf cuvette (Cout, p>0.05, r2<0.1). The gm-gsc relationship was used to calculate gm of each leaf during Laisk measurements (A/Ci curves) and thus to calculate Γ* and RL Laisk cc using A/Cc curves (cf. Fig. S4).This established that was generally lower than Γ* with a mean absolute difference of 5 μmol-1 for both young and old leaves (Fig. 4a), while RL Laisk cc (obtained from A/Cc courves) was not different from RL Laisk (obtained from A/Ci curves; Fig. 4b). An example of the offset in the common intersection point is given in Fig. S4.
DISCUSSION
In this work, RL was measured using both an isotopic disequilibrium method and the classical Laisk method on single leaves of different species, and values obtained therefrom were compared.
Reliability of RLvalues derived from isotopic disequilibrium
The present results showed a positive correlation between the two sets of RL measurements across all species and age classes, while on average RL Laisk estimates were 28% smaller than RL 13C. To our knowledge, this is the first comparison of RL estimated from the Laisk method and an isotopic disequilibrium method that does not require manipulation of photosynthetic gas exchange rates using non-physiological environmental conditions. It is not totally unexpected that the two methods provided consistently different RL estimates, given that the measurements were performed with contrasting environmental conditions and different theoretical bases. The isotopic disequilibrium method measures CO2 efflux that is not labelled (i.e. respiration fuelled by old carbon) during leaf photosynthesis. An important assumption involved is that after a short period of labelling, no tracer (new carbon) has been incorporated into respiration. Any contribution of new carbon to the respiratory CO2 efflux will lead to an underestimation of RL. The potential error seems to be negligible, since our calculations using Eqn 8 (Table S2) showed that this assumption might have led to a 2.5% underestimation of RL only, thus cannot explain the offset between RL estimates measured by the two methods. Also, in perennial ryegrass, no new carbon was observed in shoot dark respiration for about 2 h following a 1 h-long labelling period (Lehmeier et al., 2008), again suggesting insignificant underestimation of RL by short-term labelling (30-45min). The labelling dynamics in shoot respiration should be similar to that of single leaves considering that leaf respiration contributes to about half of total plant respiration (Atkin et al., 2007). However, information on labelling dynamics of single leaves is currently very limited, thus the kinetics of label appearance in day respired CO2 and its putative environmental dependence should be studied in a greater number of species.
Does RL Laiskrespond to environmental conditions imposed during measurement?
Estimates of RL differ between methods (Villar et al., 1994; Yin et al., 2011), and this effect is likely related to the different measurement conditions. Importantly, the response of RL to environmental conditions like irradiance and CO2 concentration is not well understood to date, mainly due to methodological limitations. Light has long been recognized to inhibit RL so that RL is believed to be higher at very low light, a phenomenon that is possibly also at the origin of the Kok effect (Brooks & Farquhar, 1985; Villar et al., 1994; Atkin et al., 2000; Yin et al., 2011). However, the effect of light at higher levels is not well documented. It is notable that both the Laisk and Kok method require manipulation of PAR, so the effect of PAR on RL cannot be quantified with these methods. Also, uncertainty remains as to whether there is a short-term response of RL to CO2 mole fraction. Early reports of a decrease of leaf RD with short-term increase of CO2 (see the discussion by Amthor (2000) and Yin et al. (2011)), were suggested to be largely attributable to CO2 diffusive leaks during gas exchange measurements (Amthor, 2000; Jahnke & Krewitt, 2002; Long et al., 2004; Gong et al., 2015). Results of the short-term CO2 response of day respiration are scarce. However, using 13C-labelling, it was shown that respiratory metabolism (TCA pathway) increased as CO2 mole fraction decreased (Tcherkez et al., 2008), while there seemed little effect on RL assessed with the Kok method (Tcherkez et al., 2012b). CO2 mole fraction can potentially impact on RL via changes in nitrogen assimilation caused by altered rates of photorespiration (Tcherkez et al., 2012a; Abadie et al., 2016). On the one hand, increased photorespiration at low CO2 is believed to cause high mitochondrial NADH levels and thus inhibit TCA decarboxylases. On the other hand, the increased demand for carbon skeletons to assimilate nitrogen at high photorespiration should stimulate day respiratory metabolism (Abadie et al. 2016). However, the contribution of TCA decarboxylations to total respiratory CO2 production in the light is rather small when compared to pyruvate dehydrogenation (Tcherkez et al., 2008). Therefore, the net effect of CO2 on RL itself may be modest. Still, a short-term change in CO2 mole fraction may in principle influence RL, and thus the possibility that RL is misestimated by the Laisk method cannot be excluded. This could contribute to explaining why Laisk estimates of RL are smaller than 13C-derived estimates, as shown here.
Further, the low CO2 conditions used with the Laisk method may provoke a diffusive leak as the (non-controlled) CO2 concentration outside the cuvette is higher than inside. That would increase the estimate RL if not accounted for properly, further affecting the relationship between RL Laisk and RL 13C. In the present work, however, the leak effect was accounted for. Also, the leak coefficients of intact leaves (KCO2) measured here were generally very low, much lower than the producer-suggested value of 0.44 μmol s-1. Nevertheless, we found a clear leak artefact on RL Laisk of V. faba. The diffusive leak had no significant effect on estimates of of young leaves of V. faba and R. communis (Fig. S2). Importantly, leak artefacts on RD are also not ignorable given that measurements of RD of small leaves are quite close to the detecting limit of currently available infra-red gas analysers. Since KCO2 may vary substantially between species and leave age classes, leak effects should be minimized (cf. Gong et al. 2017b) or accounted for by the measurement of the leak coefficient for every single leaf, as done here.
Does gm influence RL estimates?
Another potential uncertainty associated with the Ci-based Laisk method is the assumption on mesophyll conductance. The compensation point in the absence of day respiration, Γ*, is a Cc-based value and thus should be determined from A/Cc curves rather than A/Ci curves. In other words, the use of A/Ci curves to estimate (as a proxy of Cc) involves the assumption that gm is infinite. Consequently, assuming an infinite gm might lead to errors in the estimated Γ* and RL by the Laisk method (von Caemmerer, 2013; Walker & Ort, 2015). Here, gm of each leaf was quantified using online Δ measurements, and demonstrated that gm of older leaves was 47% smaller than that of young leaves, in agreement with studies using both online Δ or florescence methods (reviewed in Flexas et al. (2008). The estimates of gm obtained here were not very sensitive to errors in Γ*. In fact, the difference between species and age classes was not influenced by changes in Γ* within 20 μmol mol-1 (Fig. S3). Other methods like the constant J method were suggested to be sensitive to errors in Γ* (Harley et al., 1992). Furthermore, the robust relationship between gsw and gm across all species found here was similar to that reported in tree leaves (Whitehead et al., 2011). Knowing the relationship between gsc and gm allowed us to estimate gm and thus convert A/Ci curves into A/Cc curves in the Laisk method. That way, we were able to derive the parameters of interest (Γ* and RL) from A/Cc curves (cf. Fig. S4). These calculations assumed that the gsc-gm relationship was the same under the measurement condition of the Laisk method and normal growth condition, which is supported by the fact that gsc-to-gm ratio showed no significant correlation with net assimilation rate or CO2 mole fraction in the leaf cuvette. Furthermore, analyses of published data also showed a strong gsc − gm relationship across species and growth conditions (Flexas et al., 2013). Importantly, however, our results show that the Laisk method based on A/Ci curves systematically underestimated Γ* (by 5 μmol-1) but not RL (i.e. RL determined from A/Ci curves and A/Cc curves were identical).
Although statistical significance was found in T. aestivum only, the age effect on both and the Cc-based value of Γ* suggested that there was some error in the Laisk method. In fact, Γ* is given by [O2]/2Sc/o (where [O2] is oxygen mole fraction at carboxylation sites and Sc/o is Rubisco specificity) and is thus not expected to change with leaf age. Assuming a single conductance term from intercellular spaces (Ci) to the site of carboxylation (Cc) is perhaps not completely realistic, as some authors suggested that there is some resistance of the chloroplast envelope to intracellular CO2 movement (von Caemmerer, 2000), thereby leading to a lack of common intersection in Laisk curves (Tholen et al., 2012). According to the model of Tholen et al. (2012), , with total mesophyll conductance subdivided into conductance associated with cell wall and plasmalemma (gwp) and chloroplast envelope and stroma (gch).Under such an assuption, the offset of and apparent Γ* between age classes can be explained by a small increase in photorespiration F (according to the difference of Cc) and a small decrease in gch with ageing. Improving the representation of mesophyll conductance in the Laisk method is beyond the scope of the present paper, but our results suggest that the estimates of Γ* or RL obtained via the Laisk method are not precise enough (Gu & Sun, 2014), and should be viewed as approximations of actual Γ* and RL.
Conclusions and perspectives
This study showed a high variation in RL of similar leaves measured by both methods, and RL was positively correlated to RD, but not to net CO2 assimilation rate or other parameters. These observations do not support the assumption that leaf RL is a fixed proportion of photosynthesis or maximum Vc as used in many models (cf. De Kauwe et al. (2016)), but suggest that scaling RL to RD is a more reliable approach for the modelling purpose. We found a tendency for RL/RD to increase during leaf aging, and this finding is not in agreement with that reported by Villar et al. (1995). The average age difference between young and old mature leaves was about 16-20 days in our study, much shorter than that of tree leaves (about 2 years) in the study of Villar et al. (1995). Taken as a whole, our results show that RL estimates obtained using the isotopic disequilibrium method and the Laisk method were positively correlated, but RL estimated by the isotopic disequilibrium method was generally higher than that measured by the Laisk method. Both methods captured the difference in RL between species but found no effect of leaf ageing. Although RL estimates differed between measurement techniques, most leaf-level studies (including the present study) support the notion that RL is lower than RD (Villar et al., 1994; Yin et al., 2011; Gong et al., 2015; Tcherkez et al., 2017a). Mesocosm-scale 13C labelling study also showed that stand RL is inhibited by light (Schnyder et al., 2003; Gong et al., 2017a). Previous comparisons between Laisk and Kok methods showed a systematic difference between RL estimates, with RL estimated by the Kok method being generally lower than that measured by the Laisk method (Villar et al., 1994; Yin et al., 2011). Also in the case of the Kok method, it has been recently suggested that the apparent inhibition of respiration by light is at least partially explained by considerable changes in Cc during the manipulation of irradiance (Farquhar & Busch, 2017), in addition to other changes such as that in photochemical yield (for a review, see Tcherkez et al. 2017a b). Thus, our study suggests that common methods (Laisk or Kok) likely provide underestimated RL values and thus overestimated inhibition of day respiration by light. For the mechanistic understanding of day respiratory metabolism, the response of RL to light and CO2 mole fraction should be assessed in further studies, and the isotopic disequilibrium method is suitable for such a purpose since it does not require irradiance and CO2 alterations.
ACKNOWLEDGMENTS
We thank The New Phytologist Trust for supporting the 18th New Phytologist Workshop ‘The Kok effect: beyond the artefact, emerging leaf mechanisms (KOALA)’ Angers, France, July 2016. We also thank all participants of this workshop for enlightening discussion and comments. This research was supported by the Deutsche Forschungsgemeinschaft (DFG SCHN 557/7-1).
AUTHOR CONTRIBUTIONS
R.S. and X.Y.G. designed and planned the research; J.W. and R.S. performed the gas exchange measurements and isotope analyses; J.W., R.S. and X.Y.G. analyzed the data; G.T., R.S., H.S., and X.Y.G. discussed the results and implications; X.Y.G. wrote the first draft, and all authors contributed to the revision.