Department of Biological Sciences, Macquarie University, North Ryde, NSW, Australia
AXA Chair of Biosphere and Climate Impacts, Department of Life Sciences, Grand Challenges in Ecosystems and the Environment and Grantham Institute for Climate Change, Imperial College London, Ascot, UK
We present a unifying model for isoprene emission by photosynthesizing leaves based on the hypothesis that isoprene biosynthesis depends on a balance between the supply of photosynthetic reducing power and the demands of carbon fixation.
We compared the predictions from our model, as well as from two other widely used models, with measurements of isoprene emission from leaves of Populus nigra and hybrid aspen (Populus tremula × P. tremuloides) in response to changes in leaf internal CO2 concentration (Ci) and photosynthetic photon flux density (PPFD) under diverse ambient CO2 concentrations (Ca).
Our model reproduces the observed changes in isoprene emissions with Ci and PPFD, and also reproduces the tendency for the fraction of fixed carbon allocated to isoprene to increase with increasing PPFD. It also provides a simple mechanism for the previously unexplained decrease in the quantum efficiency of isoprene emission with increasing Ca.
Experimental and modelled results support our hypothesis. Our model can reproduce the key features of the observations and has the potential to improve process-based modelling of isoprene emissions by land vegetation at the ecosystem and global scales.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Isoprene (2-methyl-1,3-butadiene; C5H8) is released into the atmosphere by its main source, terrestrial vegetation. With a total annual emission of c. 0.5 Pg C yr−1 (Guenther et al., 2006, 2012; Arneth et al., 2008), this extremely volatile and reactive molecule is the most important biogenic volatile organic compound (BVOC) produced by plants.
Why do certain plants emit isoprene and others do not? What is the advantage for emitters in losing 2% or more of their assimilated carbon in the form of isoprene? What are the controls over isoprene production and emission? These questions remain largely unresolved. However, some indications have emerged in recent years thanks to advances in diverse fields from cell physiology to phylogeny (Li & Sharkey, 2012; Monson et al., 2013; Niinemets & Monson, 2013; Sharkey, 2013). Isoprene appears to protect the photosynthetic apparatus from heat and oxidative damage by enhancing membrane stability at high temperatures, and by quenching reactive oxygen species (Sharkey & Yeh, 2001; Vickers et al., 2009; Velikova et al., 2011, 2012; Possell & Loreto, 2013). Isoprene is produced in the chloroplast from its immediate precursor dimethylallyl diphosphate (DMADP), which is synthesized via the methylerythritol 4-phosphate (MEP) pathway (Lichtenthaler, 1999; Logan et al., 2000; Sharkey et al., 2008). Isoprene production is therefore controlled by the supply of DMADP, and by the activity of isoprene synthase (IspS) (Rasulov et al., 2009a,b, 2010; Vickers et al., 2010; Li et al., 2011; Monson, 2013). The metabolic controls of the MEP pathway, in relation to isoprene biosynthesis, are just beginning to be understood (Li & Sharkey, 2012; Banerjee et al., 2013; Weise et al., 2013), and the whole pathway controls cannot yet be included in isoprene emission models in a wholly mechanistic manner (Grote et al., 2013; Li & Sharkey, 2013).
In addition to its physiological interest, isoprene has sparked attention in climate science because of its impact on atmospheric chemistry and climate. Because of its abundance and reactivity, isoprene emission substantially affects the atmospheric content of tropospheric ozone, methane and secondary organic aerosols (Poisson et al., 2000; Sanderson et al., 2003; Claeys et al., 2004; Heald et al., 2008; Pike & Young, 2009; Nozière et al., 2011; Paasonen et al., 2013). To investigate the potential impact of isoprene on air quality and climate, models for isoprene emission have been developed (Grote & Niinemets, 2008; Monson et al., 2012; Grote et al., 2013). Many recently published studies have used the MEGAN model (Guenther et al., 2012), which is based on the pioneering work of Guenther and co-workers (Guenther et al., 1991, 1993). In MEGAN, a species-specific standard isoprene emission (Is) is modified by empirical functions that account for the observed variations in isoprene emissions as a result of various environmental controls. Although simple, this approach is vulnerable to model overparameterization because of interactions among environmental drivers (Niinemets et al., 2010; Sun et al., 2012). Other models have been developed based on the available knowledge of the underlying biochemical processes. These include the models of Niinemets et al. (1999) and Martin et al. (2000), and the SIM–BIM model (Zimmer et al., 2000, 2003). Nevertheless, all isoprene emission models remain largely empirical, and the mechanistic content of current models admits considerable scope for improvement (Monson et al., 2012; Grote et al., 2013).
Although often invoked as a potential driver of isoprene production (Niinemets et al., 1999; Rasulov et al., 2010; Li & Sharkey, 2012), few studies have quantitatively explored the impact of leaf energetic status on isoprene emissions. We define the leaf energetic status as the balance (or imbalance) between the supply of photosynthetic induced reducing power and the demands of carbon fixation and photorespiration. Here, we investigate the hypothesis that the rate of isoprene biosynthesis depends on the leaf energetic status. We used observations from Populus nigra grown in full sun (this study) and hybrid aspen (Populus tremula × P. tremuloides) grown at two CO2 concentrations (Sun et al., 2012). For each dataset, the experimental protocol allowed us to study short-term variations in isoprene emission, and associated variations of the electron balance between photosynthetic supply and carbon assimilation requirements. Changes in both isoprene emission and energy balance were obtained by modifying the light and CO2 conditions of the experiments. We used these datasets to test a new modelling framework, in which changes in leaf energetic status are approximated by the difference between the light- and Rubisco-limited electron fluxes for carbon assimilation. We used the same data to test the responses of two of the better known among the published isoprene models: the Guenther et al. (1993) algorithm that underlies MEGAN, and the ‘process-based’ model developed by Niinemets et al. (1999), Niinemets (2004) and modified by Arneth et al. (2007).
Isoprene is produced in the chloroplast by the MEP pathway, in which glyceraldehyde 3-phosphate (G3P) and pyruvate (Pyr) are transformed into DMADP. The process involves reduction steps that require reducing power in the form of NADPH and/or ferredoxin (Fd) (Charon et al., 1999; Hecht et al., 2001; Seemann et al., 2006; Li & Sharkey, 2012). DMADP is further transformed into isoprene by the enzyme IspS. Therefore, isoprene production is co-driven by enzymatic activity and NADPH and/or ATP availability (Lichtenthaler, 1999).
Plastid NADPH is provided by the electron transport flux generated by the light reactions of Photosystem II. As reduction steps in carbon assimilation and photorespiration consume almost all of the NADPH generated, it is common to assume that the total electron flux (Jtot, Fig. 1) is the same as the total electron flux used in carbon assimilation (JCO2+O2). However, in reality, Jtot is always somewhat larger than JCO2+O2. It has to be so in order to supply NADPH for additional redox reactions in the leaf (Niinemets et al., 1999; Singsaas et al., 2001; Niinemets, 2004). The reduction steps along the MEP pathway constitute some of these additional reactions. Thus, Jtot can be expressed as Jtot = JCO2+O2 + Jiso + Jother, where Jiso and Jother represent electron fluxes involved in isoprene production and other redox reactions, respectively, in the leaf. We hypothesize accordingly that the additional reducing power available for isoprene production is dependent on the extent to which the NADPH requirements of the Calvin–Benson and photorespiratory cycles are satisfied (Harrison et al., 2013; Morfopoulos et al., 2013). As illustrated in Fig. 1, the MEP pathway could be envisioned to act like a small branch circuit, with the greatest influx occurring when the demand of carbon assimilation for reducing power is least (Rosenstiel et al., 2004; Owen & Peñuelas, 2005). However, the MEP pathway alone does not have the capacity to absorb all of the excess energy generated. Thus, our hypothesis also suggests that isoprene emissions might co-vary with other, more effective energy quenching processes, including the Mehler reaction and the xanthophyll cycle.
Although the biochemical mechanisms controlling the partitioning of the NADPH fluxes inside the plastid are incompletely understood, the nature of the responses of isoprene emission to different environmental drivers suggests that this hypothesis is well founded (Morfopoulos et al., 2013). Indeed, the literature shows a persistent tendency for plants to increase isoprene emission (and the fraction of assimilated carbon transformed to isoprene) with increasing leaf energetic status. For example:
Isoprene emissions increase with decreasing CO2 concentration (Rosenstiel et al., 2003; Wilkinson et al., 2009; Possell & Hewitt, 2011; Sun et al., 2012).
The fraction of assimilated carbon transformed to isoprene increases with increasing light intensity (Sharkey & Loreto, 1993; Harley et al., 1996; Lerdau & Keller, 1997).
The temperature optimum for isoprene emissions is lower than that of IspS activity, and apparently co-controlled by the temperature dependences of the electron transport rate and IspS activity (Monson et al., 1992, 2012; Rasulov et al., 2010).
Isoprene emissions decrease in plants fed with nitrate (which consumes NADPH in the process of nitrate reduction to ammonia), but increase if fed with ammonia directly (Rosenstiel et al., 2004).
Isoprene emissions increase when light use efficiency decreases (Peñuelas et al., 2013).
These observations all support the hypothesis that isoprene emissions are influenced by the balance of reducing power between what can be produced by light reactions and what is needed for carbon assimilation and other major NADPH sinks.
Ideally, to represent this hypothesis quantitatively, we should model the total electron flux and the dynamics of all relevant electron sinks. However, in reality, process-based models that can simulate the total electron transport rate (Jtot) are in an early stage of development (Ye et al., 2013), the partitioning of the additional reducing power between Jother and Jiso remains enigmatic, and the nanomole scale at which isoprene emission occurs (compared with the micromole scale of electron flux) makes it unrealistic to attempt a full mass balance of the competing processes. Accordingly, our pragmatic approach is to model the energetic status of the leaf using the Farquhar model (Farquhar et al., 1980) for photosynthetic carbon assimilation, thus approximating the energetic status of the leaf as the difference between the light-limited electron flux (J) and the electron flux required to support Rubisco-limited photosynthesis (Jv). J is an approximation of the amount of reductant that light reactions can supply, and Jv represents the capacity of Rubisco to absorb this reducing power. Therefore, energy transfers to processes other than carbon assimilation [Jtot − JCO2+O2 = Jother + Jiso] should be correlated with the magnitude of the difference [J − Jv]. Based on this proxy, we build a model of isoprene emissions that we describe further in the text. We test the model with data on isoprene emission as a function of internal CO2 concentration (Ci) and photosynthetic photon flux density (PPFD).
We further test our hypothesis by examining observed and modelled changes in the fraction of assimilated carbon allocated to isoprene production. The ratio of isoprene emission to gross carbon assimilation (Iso/Agross) is a sensitive indicator of the allocation of reducing power to the MEP pathway vs the Calvin–Benson cycle (Niinemets et al., 2013). Under a constant leaf temperature and CO2 concentration, we would expect the fraction of assimilated carbon re-emitted as isoprene to be constant, if only enzymatic limitations are involved. However, if isoprene production depends on the energetic status of the leaves, Iso/Agross would be expected to increase with increasing PPFD (Niinemets et al., 2013), as carboxylation becomes progressively Rubisco limited, whilst electron transport continues to increase.
Finally, we examine changes in the quantum efficiency of isoprene emission (Φiso). Previous studies have reported changes with environmental conditions (Monson et al., 1992; Logan et al., 2000; Sun et al., 2012). Changes in the quantum efficiency of CO2 assimilation (ΦCO2) cannot explain changes in Φiso. The processes controlling quantum yields for isoprene are not fully understood. We postulate that differences in the quantum efficiency of isoprene emission (Φiso) are driven by the energetic status of the leaves, and can thus be related to the variation in [J − Jv]. Thus, we expect the quantum yield of isoprene emission to be lower when the NADPH demand for carbon assimilation is higher.
We show that our energetic status model is able to reproduce changes in isoprene emission induced by changes in Ci and PPFD, the observed tendency of (Iso/Agross) to increase with increasing PPFD and the observed increase in Φiso with decreasing CO2 concentration.
Materials and Methods
Plant material and growing conditions
In this study, we examine results from experiments conducted on two different species: Populus nigra L. and hybrid aspen (Populus tremula L.× P. tremuloides Michx.).
The first set of experiments was conducted on three saplings of P. nigra, grown in 15-l pots with a substrate composed of peat and sand (2 : 1) in a nursery (Tres Turons S.C.P., Castellar del Vallès, Catalonia, Spain). Plants were grown in a sunny environment under Mediterranean ambient conditions outdoors for 2 months before the measurement (2 May–7 July 2012). The typical Mediterranean climate is characterized by seasonal summer drought with warm temperatures and mild winters. This is reflected by the average monthly temperature of 22.8°C in August and 7.9°C in January. Mean annual precipitation and temperature are 723 mm and 15.1°C (1951–2010), respectively (Ninyerola et al., 2000). As a result of high temperature and low precipitation, the plants were under conditions of high evaporative demand. However, regular irrigation ensured that the substrate was held at field capacity throughout this period. Here, we used data from one leaf of each sapling, giving an overall dataset of three sun-adapted individuals.
The second set of experiments was conducted with 2-yr-old saplings of hybrid aspen (P. tremula × P. tremuloides) grown under two different ambient CO2 concentrations (380 and 780 μmol mol−1). These experiments, together with a full description of the materials and methods used, are reported in Sun et al. (2012, 2013b), and here only a brief summary of the methods is provided. The plants were grown in a custom-made, four-chamber, open gas exchange system. Each individual chamber experienced a 12-h photoperiod at levels of light between 500 and 800 μmol m−2 s−1, day/night air temperature of 28–30/23˚C and air relative humidity of 60%. Two chambers (chambers 1 and 3) were kept at an ambient CO2 concentration of 380μmol mol−1 (HA-G380), whereas the other two chambers were treated with an elevated CO2 concentration of 780 μmol mol−1 (HA-G780). Here, we used data from three leaves in each chamber, giving an overall dataset of six individuals grown at ambient CO2 concentration and six individuals grown at elevated CO2 concentration. For each dataset, the results shown are averaged values across individuals.
Foliage gas exchange analyses and isoprene emission rates
Gas exchange measurements were conducted on individuals of P. nigra using a Li-Cor LI-6400 portable photosynthesis system (an open gas exchange analyser using a 6-cm2 clamp-on leaf cuvette (LI 6400; LI-COR, Inc., Lincoln, NE, USA)). The calibration of the infrared gas analyser (IRGA) was performed by the manufacturer < 1 yr before the measurements.
The exhaust tube of the IRGA measure head was connected to a Proton-Transfer-Reaction Mass Spectrometer (PTR-MS) system (Ionicon Analytik, Innsbruck, Austria), using tubing material made of Siltek-passivated stainless steel (Restek, Bellefonte, PA, USA). Analyses of emission rates for isoprene were performed simultaneously with gas exchange measurements with the PTR-MS. The PTR-MS technique is based on chemical ionization, specifically non-dissociative proton transfer from H3O+ ions to most of the common BVOCs, and has been fully described elsewhere (Lindinger et al., 1998). In our experiment on P. nigra, the PTR-MS drift tube was operated at 2.1 mbar and 60°C, with an E/N (electric field/molecule number density) of c. 130 Td (Townsend) (1 Td = 10−17 V cm2). The primary ion signal (H3O+) was maintained at c. 6 × 106 counts per second. The instrument was calibrated using an aromatic mix standard gas (TO-14A; Restek) and isoprene standard gas with 100 nmol mol−1 isoprene in N2 (Abelló-Linde SA, Barcelona, Spain). Before data acquisition, the leaf cuvette was left empty in order to analyse the background concentrations of isoprene, and thereafter to calculate the foliar emission rates. No significant drift in the background of isoprene was found during the experiments.
Foliage gas exchange analyses and isoprene emission rates on hybrid aspen were obtained using a Walz GFS-3000 portable gas exchange system and a Fast Isoprene Sensor (FIS; Hills Scientific, Boulder, CO, USA). More information on the methods can be found in Sun et al. (2012, 2013b).
Before each experiment, the leaf was enclosed in the gas exchange system and left under baseline conditions until net assimilation (A), stomatal conductance (gs) and Ci stabilized (typically 20–30 min). For P. nigra, baseline conditions were PPFD of 1000 μmol m−2 s−1, leaf temperature of 30°C, relative humidity of 50% (± 10%) and ambient CO2 concentration of the leaf chamber (Ca) of 390 μmol mol−1. For hybrid aspen, baseline conditions were PPFD of 500 μmol m−2 s−1, leaf temperature of 30°C, relative humidity of 60%, Ca of 380 μmol mol−1 for HA-G380 and Ca of 780 μmol mol−1 for HA-G780. After preconditioning the leaf as explained above, two types of response curve were created: (1) the leaf net assimilation vs internal CO2 concentration (A/Ci); and (2) the leaf net assimilation vs PPFD (A/PPFD).
CO2 response curves of net assimilation and isoprene emissions
Ci response curves were obtained at a leaf temperature of 30°C and a quantum flux density of 1000 μmol m−2 s−1 for P. nigra and 500 μmol m−2 s−1 for hybrid aspen. The Ca values used to generate the A/Ci response curve were:
At every Ca, values of A, isoprene emission rate (Iso) and stomatal conductance (gs) were recorded when the gas exchange rates were stable, typically 5–10 min after the change in Ca.
PPFD response curves of net assimilation and isoprene emissions
By applying sequential changes in PPFD, light response curves at different Ca were obtained. Three different Ca values (200, 390 and 1000 μmol mol−1) were applied for P. nigra, and two different Ca values (380 and 780 μmol mol−1) were applied for hybrid aspen.
The waiting time between each light intensity was c. 10 min. The data were logged when the rates of A, gs, Ca and Iso were in the steady state, except for hybrid aspen at PPFD higher than 1500 μmol m−2 s−1, where the values were recorded after 5–8 min to avoid the development of photoinhibition.
Energetic status model
Our isoprene model is modified in one small (but important) way from that introduced by Harrison et al. (2013) and Morfopoulos et al. (2013), and deals with the issue of negative values for isoprene emission generated using the first version of the model (Supporting Information Notes S1; Table S1; Figs S1, S2). In these earlier papers, the isoprene emission rate was assumed to be linearly related to the energy status of the leaf, whereas, here, the fraction of electrons allocated to isoprene biosynthesis is linearly related to the energetic status of the leaf:
where Iso is the isoprene emission, f(Ci) is a function of internal CO2 concentration, f(T) is a function of temperature, taking into account the response of enzymatic activity to temperature, J is the light-limited electron flux, taken to be a non-rectangular hyperbolic function of absorbed PPFD and the maximum electron flux Jmax, following Farquhar et al. (1980), and
which is the electron flux required to support Rubisco-limited carbon assimilation. Γ* is the CO2 compensation point in the absence of mitochondrial respiration in the light, Vcmax is the Rubisco carboxylation capacity and K = Kc(1 + [O2]/Ko), where Kc and Ko are the Michaelis coefficients of Rubisco for CO2 and O2, respectively (Farquhar et al., 1980). The term ε in Eqn (Eqn 2) is not constant, but varies depending on the energetic status of the leaf, estimated by [J – Jv]. The function f(Ci) in Eqn (Eqn 2) is chosen to take the value Ci/Γ* when Ci ≤ Γ* and ‘1’ otherwise, and reflects the common observation that isoprene emission ceases when Ci< Γ* as a result of a minimum supply of carbon chains required for isoprene synthesis and/or the inhibition of the electron transport rate below Γ* (Dietz et al., 1985; Wolfertz et al., 2003; Rasulov et al., 2009b, 2011; Monson et al., 2012; Sun et al., 2012) This fall-off of isoprene at low Ci is not fully understood and is not always observed: the emission of isoprene in CO2-free air has been reported in a few studies (Monson & Fall, 1989; Affek & Yakir, 2003; Li & Sharkey, 2012). However, comparable conditions are not found in natural environments. Using the Ci response curves, changes in the fraction ε of the light-limited electron flux (J) allocated to isoprene production (Eqns (Eqn 1), (Eqn 2)) were plotted against the corresponding difference between light- and Rubisco-limited electron fluxes [J − Jv]. Parameters c1 and c2 were obtained from a linear regression between ε and [J − Jv] when Ci> Γ* (Figs 2a, 3a,b). Because all our experiments were conducted at a leaf temperature of 30°C, we neglect here the temperature dependence caused by IspS activity, and f(T) is accordingly set equal to unity. Quantum efficiencies for isoprene production (Φiso) were calculated as the initial slope of isoprene emission vs PPFD, for PPFD < 250 μmol m−2 s−1. The bounds of uncertainty of the energetic status model displayed in the figures represent uncertainties in the estimated values of Vcmax and Jmax in the Farquhar model.
The G93 algorithm
The algorithm developed by Guenther et al. (1993), which is the basis of the isoprene module of the MEGAN model (Guenther et al., 2012), is the most widely used algorithm for the prediction of isoprene emission by plants. Hereafter, this algorithm is referred to as G93. In G93, the emission rates of isoprene are calculated by multiplying a species-specific standard emission rate (Is) by a set of empirical equations taking into account changes in environmental variables. The standard conditions for Is are a leaf temperature of 30°C and an incident PPFD of 1000 μmol m−2 s−1. Because, in this study, all the experiments were conducted at a constant leaf temperature of 30°C, we consider only changes driven by light intensity:
where CL1 and α are empirical coefficients. For each light response curve, in order to take into account the CO2 effect on the standard emission rates, the value of Is was taken as the observed emission rate at a PPFD of 1000 μmol m−2 s−1, under the CO2 conditions of the experiment.
The Niinemets model
The Niinemets model (Niinemets et al., 1999) is based on the quantification of the NADPH cost for isoprene synthesis. It builds on the Farquhar model of photosynthesis. The general concept is that a temperature-dependent fraction of the electron flux (εN) is used for isoprene production:
where Jiso is the electron flux required in order to produce a quantity of isoprene and Jtot is the total photosynthetic electron flux, approximated by J, using the Farquhar model:
where Aj is the gross assimilation under electron transport-limited conditions, Ci is the internal CO2 concentration and Γ* is the compensation point.
The total NADPH cost for isoprene production per mole CO2 assimilated is 1.17 times higher (2.33 NADPH per CO2) than for sugar synthesis (2 NADPH per CO2), and six molecules of CO2 must be assimilated to produce one isoprene molecule. Drawing a parallel with the Farquhar model, Jiso is thus estimated as:
Because all our experiments were conducted at a leaf temperature of 30°C, we neglect the temperature dependence of εN. The effect of changes in CO2 concentration on εN is adapted from Arneth et al. (2007):
where εNs is the fraction of electrons used for isoprene production under the standard conditions of leaf temperature Ts = 30°C, PPFD = 1000 μmol m−2 s−1 and Ca_s = 390 μmol mol−1. In this study, εNs was estimated from experiment by varying PPFD at a Ca of 390 μmol mol−1 for P. nigra and 380 μmol mol−1 for hybrid aspen.
The Farquhar model
The Farquhar et al. (1980) photosynthesis model describes the limitations on the C3 photosynthetic rate (A) by two main equations representing the limitations imposed by Rubisco-catalysed carboxylation (Vcmax) and RuBP regeneration, which is limited by PPFD and by the maximum electron transport rate (Jmax). Under Rubisco limited conditions, A is expressed as:
Av is the gross assimilation under Rubisco-limited conditions, Rd is the mitochondrial respiration in the light and was assumed to be equal to dark respiration divided by 2 (Niinemets et al., 2005; Misson et al., 2010; St Paul et al., 2012). Under electron transport limitation, A is expressed as:
where J is the potential rate of electron transport. J, in turn, depends on PPFD up to a maximum Jmax (de Pury & Farquhar, 1997). For each Ca, the averaged value of observed Ci was used for the model simulations.
Values of Michaelis–Menten constants, activation and de-activation energies, specificity for Rubisco and their temperature dependences were taken from Bernacchi et al. (2002) and Medlyn et al. (2005) (Notes S2; Table S2).
For the experiment on P. nigra, probably as a result of the growing conditions of the plants (Mediterranean summer sunshine), the plants adapted their maximum Rubisco capacity (Vcmax) to the prevailing high levels of irradiance and temperature. As a result, under most of the experimental conditions (including a large part of the A/Ci curve), the carbon assimilation was found to be limited by electron transport and not by Rubisco capacity. In order to estimate Vcmax, we therefore used the light response curve for assimilation, at a Ca of 200 μmol mol−1 and PPFD ≥ 1500 μmol m−2 s−1, where A was saturating. We calculated Vcmax by minimizing the residual sum of squares (RSS) between the Rubisco-limited equation and the observations. The capacity for photosynthetic electron transport (Jmax) was obtained similarly by minimizing RSS between the light-limited equation and the assimilation data from all experiments. For hybrid aspen, Jmax and Vcmax were estimated from A/Ci curves by minimizing RSS between the Farquhar model and the observations.
Model parameters are summarized in Table 1. Statistical analyses were performed using the software R version 2.15.0 (http://www.r-project.org/).
Table 1. Model parameter values at a leaf temperature of 30°C
Farquhar model uncertainties (in parentheses) were obtained by fitting the model to the maximum and minimum bounds of the assimilation curves.
Parameters not fitted to the data; Jmax, maximum electron flux; Vcmax, maximum Rubisco carboxylation capacity; αCO2, quantum yield of electron transport; θ, curvature parameter of the light response curve; c1 and c2, parameters of our energetic status model; α and CL1, parameters of the G93 algorithm.
For each plant type, isoprene emissions showed a strong negative response to changes in Ci (Figs 2b, 3c,d). For P. nigra, the maximum isoprene emissions were c. 33 nmol m−2 s−1 at low Ci (73–174 μmol mol−1), declining to 8 nmol m−2 s−1 at high Ci (1280 μmol mol−1). Maximum isoprene emission rates (at low Ci) represented up to 2.24% of assimilated carbon (Fig. S3); this percentage dropped to 0.17% at high Ci. For hybrid aspen, averaged isoprene emissions peaked at low Ci (105–140 μmol mol−1) with maxima of c. 21 nmol m−2 s−1 for HA-G380 and 25 nmol m−2 s−1 for HA-G780, declining below 4 nmol m−2 s−1 at high Ci (1400 μmol mol−1). A decline in isoprene emissions for very low values of Ci was observed whatever the growing conditions. As highlighted in Sun et al. (2012), isoprene emissions reached higher rates for individuals grown under elevated CO2 concentrations, contrary to that which is usually assumed. Maximum emission rates represented a loss of assimilated carbon into isoprene of 5.6% for HA-380 and 6.6% for HA-G780; this percentage dropped to 0.09% for high values of Ci.
For all experiments, a very strong linear correlation was found between [J − Jv] and the number of electrons ε engaged in the isoprene production pathway, with r2 > 0.89 (Figs 2a, 3a,b). Yet, the response of ε vs [J − Jv] seems to start to saturate at very negative values of [J − Jv] in each dataset. This behaviour might be caused by an overall saturation of the redox state of QA (the primary acceptor of Photosystem II) associated with a limitation of capacity of Jtot that can be observed under high Ci (Dietz et al., 1985).
With parameters obtained from the linear regression of ε vs [J − Jv], our model simulated isoprene emissions in response to changes in Ci with excellent agreement to the observations (r2 = 0.94, 0.87 and 0.93 for P. nigra, HA-G380 and HA-G780, respectively) (Figs 2b, 3c,d).
We also tested the response vs Ci of the Niinemets model corrected by the empirical CO2 response function proposed by Arneth et al. (2007) (Fig. S4). The Niinemets model reproduced the data reasonably well, but tended to underestimate the isoprene emissions for P. nigra, whereas it tended to overestimate the isoprene emissions for the hybrid aspen experiments. It should also be noted that, without the CO2 response function proposed by Arneth et al. (2007), the Niinemets model would show an increase in isoprene emissions with increasing Ci, imitating the response of Aj.
Experiments varying PPFD
For all experiments, isoprene emission rates increased with increasing PPFD, with observed maxima for isoprene emissions inversely related to Ca (and consequently to Ci) – opposite to the net assimilation rates. Observed isoprene emissions vs J were found to have a quadratic type of response, in line with our model (shown for hybrid aspen in Fig. S5).
For P. nigra at each Ca, our model captured the variations in isoprene emissions extremely well with r2 > 0.99 (Fig. 4, Table 2). For Ca of 200μmol mol−1, however, our model systematically underestimated the observed values. The Niinemets model showed comparable r2 values (Table 2), consistent with the fact that isoprene emission, in both our model and that of Niinemets, is proportional to J. G93 was the only model with a component (Is) fitted directly to the observations, yet G93 performed less well than the other two models. All models underestimated isoprene emission rates at the highest PPFD of 2500 μmol m−2 s−1.
Table 2. Isoprene emissions vs changes in photosynthetic photon flux density (PPFD) at different CO2 concentrations (Ca)
Ca (μmol mol−1)
Energetic status model
For hybrid aspen, all models captured well the variation in isoprene emissions with PPFD with r2 > 0.88. Yet, our model tended to systematically underestimate isoprene emissions for HA-G380 (Fig. 5).
Isoprene: assimilation ratios (Iso/Agross)
Observed mean Iso/Agross increased with increasing PPFD, regardless of Ca, plant type or growth conditions. However, the range of Iso/Agross across individuals is considerable. The fraction of assimilated carbon re-emitted as isoprene was inversely related to the CO2 concentration. The high ratios of Iso/Agross at low Ca were a result of a combination of high isoprene emission rates and low carbon assimilation rates.
Our energetic status model can reproduce an increase in the fraction of carbon allocated to isoprene emission with increasing PPFD (Figs 6, 7). It fails to reproduce absolute values of Iso/Agross; however, it should be noted that the simulated Iso/Agross includes combined uncertainties of the isoprene model and the Farquhar model.
G93 shows versatility in the simulation of carbon allocated to isoprene emission, with simulated Iso/Agross decreasing with PPFD for P. nigra, but increasing for hybrid aspen.
With the exception of hybrid aspen at Ca= 380 μmol mol−1, the Niinemets model failed to capture the changes in Iso/Agross with changing PPFD, showing no relationship between Iso/Agross and PPFD.
Isoprene quantum efficiencies
As predicted by our hypothesis, the observed quantum efficiencies for isoprene production were dependent on the CO2 concentration (Fig. 8). Higher quantum efficiencies correspond to lower Ca, at which the demand for reductant by the Calvin–Benson cycle is lower. Our model captured the observed decrease in Φiso with increasing Ca. However, the model overestimated Φiso at high Ca and underestimated Φiso at low Ca for P. nigra. The model overestimated Φiso for HA-G380 and underestimated Φiso for HA-G780.
The overall performance of each model is illustrated in Figs 9 and 10. Our energetic status model gave excellent results overall (r2 = 0.97 for P. nigra, r2 = 0.94 for hybrid aspen). No major pattern was detected in the residuals, although the model has the tendency to underestimate the observations (Figs S6, S7). Moreover, this model reproduced the following key features of the observations:
A decrease in isoprene emissions with increasing Ci.
An increase in isoprene emissions with increasing PPFD, with maxima inversely proportional to the CO2 concentration.
An increase in the proportion of assimilated carbon diverted to isoprene production (Iso/Agross) with increasing PPFD.
A decrease in the quantum efficiency of isoprene production with increasing CO2 concentration.
With Is adjusted for each experiment, G93 reproduces very well the observed variations in isoprene emission with PPFD, especially for hybrid aspen (Ci experiments are not included for G93). For P. nigra, the bell-shaped pattern observed in the residuals vs fitted values plot (Fig. S6) suggests that the standard light response of G93 is not adapted to fit the observations.
With no empirical adjustment included to account for the CO2 effect, the Niinemets model (r2 = 0.09–0.14) failed to reproduce the observed variations in isoprene emission with PPFD and Ci. Including a CO2 effect in this model, however, caused major improvements (r2 = 0.97–0.89).
We used the Ci and PPFD response curves of assimilation and isoprene emissions for P. nigra (this study) and P. tremula × P. tremuloides (hybrid aspen) (Sun et al., 2012), where changes in balance between electron supply and electron demand for carbon assimilation purposes were driven by different environmental variables. We tested against these data a new model in which isoprene production is a function of the energetic status of the leaves, alongside two widely used isoprene models: the G93 algorithm (Guenther et al., 1993) and the Niinemets model (Niinemets et al., 1999; Arneth et al., 2007). The new model showed excellent results and a visible improvement relative to the original Niinemets model (Figs 9, 10).
Our model finds its origin in the Niinemets model based on the ‘energetic requirements for isoprene synthesis and leaf photosynthetic properties’. It keeps the major advantage of its simplicity and thus the evident potential for its use in large-scale modelling, where excessive complexity is to be avoided wherever possible. Yet, the new model diverges from its prototype in two fundamental ways. First, it links isoprene emission directly to the electron flux (J) rather than to light-limited assimilation. Second, it links isoprene emission to reductant availability, and thus transcribes the original idea of Niinemets et al. (1999) of a ‘competition for electrons between isoprene synthesis and Calvin and photorespiratory cycles’. The component of electron flux generated by Photosystem II and not used for carbon assimilation and photorespiration is extremely hard to investigate experimentally (Singsaas et al., 2001). Nevertheless, our hypothesis is supported by the following: the high positive correlations found between the observations and simulations made with our energetic status model; the fact that measured Iso/Agross increases with increasing PPFD; the fact that observed Φiso is inversely proportional to Ca; strong linearity between the flux of electrons engaged in isoprene production and [J − Jv]; and a quadratic type of response of isoprene emission to J.
In fact, the first derivation of the model of Niinemets et al. (1999) and Niinemets (2004) predicted that the fraction of electrons channelled into isoprene synthesis varies with CO2 concentration, but this variation was not explicitly formalized. In the later development of this model, Arneth et al. (2007) included this effect empirically in the emission model. Nevertheless, the reduction in isoprene emissions at intercellular CO2 concentrations between 0 and 150 μmol mol−1 (Loreto & Sharkey, 1990; Rasulov et al., 2009b, 2011; Sun et al., 2012) was not considered. Wilkinson et al. (2009) also included the CO2 dependence of isoprene emission, but did not consider the declining part of the isoprene emission at low CO2 concentrations. It has been shown that this reduction is associated with reduced availability of DMADP and is suggested to indicate limited NADPH or ATP availability (Rasulov et al., 2009b, 2011). Here, the model based on NADPH limitation described well the entire CO2 response curve (Figs 2, 3), in line with the experimental observations of the variation of DMADP pool size with [CO2].
A limitation of the present study is that experiments were conducted under constant temperature. This has the advantage of decoupling effects related to NADPH production from effects of enzyme kinetics. However, isoprene emissions also respond strongly to temperature, both instantaneously and over longer periods (Guenther et al., 1991; Pacifico et al., 2009; Laffineur et al., 2011; Sun et al., 2013a). Therefore, an improved understanding of the controls on isoprene emission for global or regional modelling purposes also requires that the hypothesis presented here be tested and analysed under variations of temperature, as well as PPFD and Ci.
Following the logic of the G93 algorithm, many studies (including ours) have examined isoprene emission under the standard conditions of a leaf temperature of 30°C and a PPFD of 1000 μmol m−2 s−1. This might be a limitation, as interactions between different drivers are then neglected. As an example of the importance of this limitation, the recent study of Sun et al. (2013a) showed cancellation of the isoprene response to rapid changes in Ci at higher temperature. Thus, there is a need for more complete experimental studies focusing on the interactions between the effects of simultaneous changes in temperature, PPFD and Ci.
In future model development, it will also be important to consider the adaptation of model parameters to long-term variations in temperature and CO2, and effects of changes caused by leaf ontogeny – all of which could modify the expression of the IspS gene (Monson, 2013; Rajabi Memari et al., 2013; Rosenkranz & Schnitzler, 2013) and the pool size of DMADP (Sun et al., 2012; Rasulov et al., 2013). The consideration of such changes is needed to allow the inclusion of acclimation in isoprene emission on time scales from days to months, and thus eventually to allow the responses of isoprene emissions to global change to be modelled in a more explicitly process-based manner than has been possible so far.
The research leading to these results has received funding from the European Community's Seventh Framework Programme (FP7 2007-2013) under grant agreement no. 238366, and from the Estonian Ministry of Science and Education (institutional grant IUT-8-3) and Estonian Science Foundation (Grant 9253).