C2 photosynthesis generates about 3-fold elevated leaf CO2 levels in the C3–C4 intermediate species Flaveria pubescens

Summary Photorespiration raises cellular CO2 levels about 3-fold in leaves of C3–C4 intermediate Flaveria species. This was shown by using 14C-based fluxomics to determine the Rubisco in vivo carboxylation-to-oxygenation ratios.


Introduction
Land plants form three major classes characterized by specific modes of photosynthetic CO 2 assimilation. In C 3 plants, CO 2 enters metabolism directly via ribulose 1,5-bisphosphate (RubP) carboxylase/oxygenase (Rubisco). In the mesophyll of C 4 plant leaves and in CAM (crassulacean acid metabolism) plants, CO 2 is initially fixed by phosphoenolpyruvate carboxylase. The resulting four-carbon (C 4 ) compounds are decarboxylated in the Rubisco-containing bundle-sheath of C 4 plants (Hatch and Slack, 1970) or become stored in the vacuoles of CAM plants for daytime decarboxylation and refixation of the released CO 2 by Rubisco (Lüttge, 2004). Both modifications to the C 3 mode of CO 2 assimilation are adaptations to specific environmental conditions such as low CO 2 or water availability. While C 4 plants represent only about 3% of all land plant species, they dominate nearly all grasslands in the tropics, subtropics, and warm temperate zones (Sage, 2004). They also include highly productive crops, such as corn and sugar cane, and there is much interest to introduce yield-relevant features of C 4 photosynthesis into C 3 crops.
Given the ecological and agricultural significance of C 4 plants, it is important to understand how they evolved and what were the crucial steps in this process. A number of studies have shown that the evolution of C 4 photosynthesis was not a unique event but occurred at least 66 times during the past 35 million years (Sage, 2004;Sage et al., 2012). Among these plant lineages, the small genus Flaveria (Yellowtops) has received particular attention because it includes species with CO 2 assimilation modes ranging from C 3 via a broad range of C 3 -C 4 intermediate species to C 4 (Powell, 1978;Apel and Maass, 1981;Ku et al., 1983;Bauwe, 1984). Notably, extant Flaveria C 3 -C 4 intermediate species represent true evolutionary intermediates between C 3 and C 4 photosynthesis (Kopriva et al., 1996;McKown et al., 2005). Major physiological features of such plants are low apparent photorespiration (Apel and Maass, 1981;Holaday et al., 1982 in combination with an enhanced refixation of photorespiratory CO 2 (Holbrook et al., 1985;Bauwe et al., 1987) and high glycine accumulation Chollet, 1983, 1984).
Mechanistically, corresponding to the distribution of the photorespiratory enzyme glycine decarboxylase (GDC) in leaves of C 4 plants (Ohnishi and Kanai, 1983), these specific characteristics are closely related to a confinement of GDC activity to the leaf bundle sheath (Hylton et al., 1988;Moore et al., 1988). Based on these and other data, it was hypothesized that C 3 -C 4 intermediate species reduce apparent photorespiration by an efficient refixation of photorespired CO 2 in the bundle sheath (Monson et al., 1984;Edwards and Ku, 1987;Rawsthorne, 1992). This initial focus on the importance of CO 2 refixation was later extended by the hypothesis that the confinement of glycine decarboxylase could result in a concentration of CO 2 in the bundle sheath of C 3 -C 4 intermediate plants (von Caemmerer, 1989;Monson and Rawsthorne, 2000). Today, such a mechanism, in which photorespiratory glycine serves as a vehicle to move 'CO 2 ' from the mesophyll to the GDC-containing bundle sheath, is seen as a crucial step during the evolution of C 4 photosynthesis (Bauwe, 2011;Sage et al., 2012). In other words, the multiple evolution of C 4 photosynthesis might have been triggered by and possibly even required the preceding presence of a much simpler CO 2 concentration system than the C 4 cycle, based on relatively small alterations to the high-flux photorespiratory glycine metabolism.
This hypothesis is now widely accepted and the genetic alterations necessary to restrict photorespiratory GDC activity to the bundle sheath are being unravelled (Wiludda et al., 2012;Schulze et al., 2013). On the other hand, it is not known how efficient this photorespiratory CO 2 pump could be. Here, 14 CO 2 incorporation studies designed to obtain an estimate of the in vivo rates of the two Rubisco-catalysed reactions in the C 3 -C 4 species Flaveria pubescens relative to the control C 3 species Flaveria cronquistii are reported. The ratio of these reactions, carboxylation versus oxygenation of RuBP, is co-determined by kinetic parameters of Rubisco and by the CO 2 /O 2 concentration ratio (Laing et al., 1974;Peisker, 1974;Farquhar et al., 1980). Hence, a higher in vivo carboxylation/ oxygenation ratio in F. pubescens relative to a control C 3 species would not only indicate an elevated CO 2 /O 2 concentration ratio but also allow quantifying the efficiency of the photorespiratory CO 2 pump.

Materials and methods
Plant growth and 14 C labelling Flaveria cronquistii A.M. Powell (C 3 ), Flaveria pubescens Rydberg (C 3 -C 4 ), and Flaveria trinervia (Spreng.) C. Mohr (C 4 ) were grown in soil in a controlled environment chamber at 28/22 °C (day/night) and 250-300 µmol photons m -2 s -1 at a photoperiod of 16 h. Fully expanded leaves excised from 40-60-d-old plants were fixed by thin wires in a frame positioned in a purpose-built fast-acting 14 CO 2 labelling device . Leaves were pre-illuminated at 30 Pa 12 CO 2 and 210 kPa O 2 for 10-15 min at about 1200 µmol photons m -2 s -1 and 25 °C to ensure maximum stomata opening and achievement of the steady-state rate of photosynthesis. Plants were then exposed to 14 CO 2 (2000 MBq mmol -1 ) for 0.6, 1.2, 2.4, 5, 15, 60, 120, and 360 s at the same concentrations of CO 2 and O 2 , temperature and light as applied during pre-illumination. At the given time points, within 0.1 s, the leaf samples were automatically transferred into boiling 80% ethanol. 14 CO 2 incorporation was linear over the whole experiment. All experiments were performed in triplicate (three individual plants in three consecutive days, resulting in three leaf samples per time-point for each species).

Metabolite analysis
All leaf samples were individually extracted as described before (Värk et al., 1968) with slight modifications. After 2 min in boiling 80% ethanol, the samples were extracted for 15 min at 86 °C with 5 ml of 80% ethanol (twice) and 20% ethanol (once). All four ethanolic fractions were combined. The remaining samples were then further extracted for 15 min at 86 °C with 5 ml 96% ethanol acidified with 3 drops of 3 N HCl. The two extracts were separately (to avoid the hydrolysis of disaccharides) dried at 37 °C, individually re-dissolved in 5 ml H 2 O each and cleared by centrifugation. The supernatants were combined, dried as above, and the metabolites re-dissolved in 1 ml H 2 O. This final extract was used to determine total extractable radioactivity, radioactivity in amino acids (AAA 339 analyzer, Mikrotechna, Czech Republic), and other metabolites by using two-dimensional paper chromatography. Residual radioactivity in the fully extracted, dried, and triturated leaf samples was determined by using a non-aqueous scintillation cocktail. These analytical methods including the protocol used for starch analysis were described in more detail elsewhere (Keerberg et al., 2011).

Photosynthetic-photorespiratory gas exchange
Rates of net and true photosynthesis, photorespiratory CO 2 evolution from the leaf, intracellular decarboxylation of early photosynthates, and rates of reassimilation of photorespiratory CO 2 were determined during steady-state photosynthesis by using standard gas-exchange measurement techniques in combination with a radiogasometric method described before Keerberg, 1995, 2007). In short, this method is based on the analysis of time curves of 14 CO 2 evolution from labelled photosynthates in leaves previously exposed to 14 CO 2 . Photorespiration (210 kPa O 2 ) and day respiration (15 kPa O 2 ) were distinguished by measurement under different O 2 concentrations. Re-fixation ratios (D) of photorespiratory CO 2 were calculated from 14 CO 2 evolution at the very high concentration of 3 kPa 12 CO 2 , where re-fixation of 14 CO 2 evolved inside the cell is close to zero, relative to 14 CO 2 evolution at air levels of 12 CO 2 .

Modelling and data analysis
From the radioactivity values for individual metabolites in combination with the specific radioactivity of the 14 CO 2 fed to leaves, the amounts of carbon incorporated at the selected time points were calculated and plotted against the duration of feeding with 14 CO 2 . The amounts of carbon fixed in individual compounds were expressed in absolute (µmol C m -2 ) and relative (per cent of total carbon fixed) units. These experimental labelling curves contain the information about rates of all relevant carbon fluxes and corresponding metabolite pool sizes.
To extract this information on Flaveria photosynthetic-photorespiratory metabolism, the model shown in Fig. 1 was used. The model allows CO 2 incorporation into the reductive pentose phosphate cycle (RPPC) either directly with rate R 1 or via the C 4 cycle with rate R 6 . Total carbon flux through the photorespiratory cycle is denoted R 2 . R 5 is the export rate of phosphorylated sugars into other pathways, for example, sucrose biosynthesis. R 7 denotes the rate of carbon efflux from the RPPC to the C 3 skeleton of C 4 acids, while R 8 describes the rate of accumulation of C 4 -acids. In order to simplify calculations, metabolites were grouped into four pools: (i) pool 'SP' with sugar phosphates plus 3-phosphoglycerate, (ii) pool 'Gly' with metabolites of the two-carbon branch of the photorespiratory cycle, (iii) pool 'Ser' with metabolites of the three-carbon branch of the photorespiratory cycle, and (iv) pool 'C 4 ' with malate and aspartate. Each of these four pools comprises two metabolic sub-pools with different labelling kinetics, for example, photorespiratory pools with rapid turnover in peroxisomes and mitochondria (Gly-I and Ser-I with pools C 2 and C 3 , respectively) or less mobility in the cytosol and chloroplasts (Gly-II and Ser-II with pools C 4 and C 5 , respectively). At steadystate photosynthesis, these pools are in diffusional equilibrium with exchange rates R 3 and R 4 , respectively. At the glycine-intoserine conversion step, one molecule of CO 2 is released per serine molecule formed, corresponding to a glycine decarboxylation rate of R 2 /4. The resulting CO 2 is re-fixed in the RPPC or the C 4 cycle or escapes from the leaf. The extent of re-fixation is described by the re-fixation coefficient D, which was experimentally determined as described above.
Formally, the metabolic model is described by the four analytical functions shown as equations 1-4, one for each major metabolite pool (similar to Keerberg et al., 2011). To determine individual pool sizes C i and carbon fluxes R i , the experimental values of the radioactivity of sugar phosphates, metabolites of the glycine and serine branches of the photorespiratory pathway, and of C 4 -acids were simultaneously fitted to these functions by multi-component non-linear regression analysis. These functions also consider the time-dependent dilution of the applied tracer CO 2 by unlabelled photorespiratory CO 2 , which is important particularly at the start of tracer feeding under steady-state photosynthesis. A more detailed explanation of these functions is provided in the Supplementary data at JXB online.

Results and discussion
The analysis of in vivo Rubisco carboxylation and oxygenation rates is not trivial. Potentially, such data can be extracted from gas exchange experiments (Pärnik and Keerberg, 1995), but this approach is biased by limited knowledge of the internal diffusion pathways for CO 2 and O 2 . Bias becomes even stronger at a varying intercellular distribution of photosynthetic tasks, such as the operation of CO 2 -concentrating mechanisms. Assuming that there is no large variation in the plastidial O 2 concentrations (Tolbert et al., 1995), it should be possible approximately to assess the efficiency of the photorespiratory CO 2 pump in C 3 -C 4 intermediate plants by the quantification of carbon fluxes through the individual routes of the photosyntheticphotorespiratory biochemical network. Speed and complexity of the biochemical processes involved require fast and, consequently, sensitive labelling techniques using 14 CO 2 as a tracer in combination with model-based data analysis.
For our study, three Flaveria species were used, F. cronquistii (C 3 ), F. pubescens (C 3 -C 4 intermediate), and F. trinervia (C 4 ). These species have previously been examined for their photosynthetic types (Apel and Maass, 1981;Ku et al., 1983;Rumpho et al., 1984), kinetic properties of Rubisco (Bauwe, 1984;Wessinger et al., 1989;Kubien et al., 2008), and phylogenetic position within the genus (Powell, 1978;Kopriva et al., 1996;McKown et al., 2005). These studies include the observation (Bassüner et al., 1984;Monson et al., 1986) that C 3 -C 4 intermediate Flaveria species fix a small fraction of CO 2 via the C 4 pathway (R 6 in the model shown in Fig. 1) while most of the CO 2 enters metabolism directly via the RPPC (R 1 in Fig. 1). It was not our intention to perform a comprehensive re-analysis of photosynthetic-photorespiratory carbon Fig. 1. Model of major photosynthetic-photorespiratory carbon fluxes in Flaveria including the reductive pentose phosphate cycle (RPPC) with the attached photorespiratory pathway and the C 4 photosynthetic pathway. R 1 , rate of CO 2 fixation in RPPC; R 2 , rate of carbon flux through the glycolate cycle; R 3 , rate of carbon exchange between different pools of glycine; R 4 , rate of carbon exchange between different pools of serine; R 5 , rate of transport of sugar phosphates out of the RPPC; R 6 , rate of CO 2 fixation by the C 4 pathway; R 7 , rate of carbon flux from RPPC into 'C 3 skeletons' of C 4 acids, R 8 , rate of accumulation of C 4 acids; C 1 , total pool of sugar phosphates in the RPPC; C 2 , active pool of the glycine branch of the photorespiratory pathway; C 3 , active pool of the serine branch of the photorespiratory pathway; C 4 and C 5 , corresponding non-photorespiratory metabolite pools; C 6 , extra-cyclic pool of sugar phosphates; C 7 , total pool of C 4 acids. D (reassimilation coefficient) describes the fraction of refixed relative to total photorespiratory CO 2 . A 1 and A 2 are the partition coefficients describing the relative contributions of the RPPC and the C 4 pathway to refixation of photorespiratory CO 2 . Note that Gly-I and Ser-I also include all other metabolites from the respective branches of the photorespiratory pathway. Gly-II and Ser-II represent less mobile (cytosolic, plastidial, vacuolar) pools of these metabolites. metabolism of these species. Instead, we wanted to focus on the quantification of key fluxes including control data confirming adequate fidelity of our approach.
Building upon previous studies Keerberg et al., 2011), the model schematically shown in Fig. 1 was developed which embraces, in a generalized form, all the relevant information that is necessary to determine Rubisco carboxylation/oxygenation ratios in vivo. It considers time-and flux-dependent changes in the tracer's specific radioactivity at all nodes of the network and allows the separation of high-and low-turnover pools of key metabolites of photosynthetic CO 2 and photorespiratory O 2 fixation. In order to simplify the model and make it as robust as possible, the metabolically related metabolites of the four major pathways were combined into four pools, each of which is described by a labelling function P(t,C i ,R i ) shown as equations 1-4.   (2); Gly-I plus Gly-II] and the serine branch [equation (3); Ser-I plus Ser-II] of the photorespiratory pathway, and the C 4 pathway [equation (4); C 4 ]. S S and S C are time-dependent functions that describe changes in the specific radioactivity of CO 2 fixed in the RPPC and the C 4 pathways, respectively. Functions P(SP), P(Gly), P(Ser), and P(C 4 ) were simultaneously fitted to experimental data points collected over a time scale from 0.6 to 360 s during steady-state photosynthesis. Quantitative values for carbon fluxes Ri between the sub-pools directly involved in photosynthetic CO 2 fixation and photorespiration, for example, from SP-I (pool size C 1 ) via Gly-I (pool size C 2 ) to Ser-I (pool size C 3 ), were calculated by multi-component non-linear regression analysis. Figure 2 demonstrates that the model approximations for all four major metabolite pools represented by the model fit Fig. 2. Time-courses of CO 2 incorporation into sugar phosphates, C 4 acids, and intermediates of the two branches of the photorespiratory pathway. Shown are time-courses relative to true photosynthesis, which was set to 100% for easier comparison. Symbols represent mean values from three data points (biological replicates). Solid (F. cronquistii), dashed (F. pubescens), and dotted lines (F. trinervia) are best fits to the labelling functions (Equations 1-4) and were calculated by multi-component non-linear regression analysis. very well to the experimental data points. This includes initial CO 2 fixation by the C 4 pathway in F. trinervia in combination with final refixation of CO 2 released from C 4 acids by the RPPC as well as the 'glycine anomaly' of the C 3 -C 4 intermediate plant F. pubescens. As mentioned in the Introduction, the specific alterations to glycine metabolism of C 3 -C 4 intermediate plants are due to a specific distribution of photorespiratory GDC activity (Rawsthorne, 1992), which represents the enzymatic backbone of the photorespiratory CO 2 pump.
Another apparent feature is the overlap of primary and secondary labelling kinetics, which is best seen with the C 4 acids but also within the glycine and serine branches of the photorespiratory pathway (Keerberg et al., 2011). In the case of the C 4 acids, the complex labelling kinetics results from direct CO 2 fixation (R 6 in Fig. 1), secondary labelling of carbons 1-3 by the synthesis of phosphoenolpyruvate from RPPC intermediates (via phosphoglycerate mutase and enolase; R 7 ), and export as a metabolically less mobile pool (probably to the vacuole; R 8 ). Also, two metabolic pools with different labelling kinetics exist in both branches of the photorespiratory pathway. This is because one fraction each (Gly-II and Ser-II with pools C 4 and C 5 , respectively) is present in cellular compartments that do not directly contribute to photorespiratory reactions. These fractions show a lower turnover than the photorespiratory most active pools (Gly-I and Ser-I with pools C 2 and C 3 , respectively). At steady-state photosynthesis, the pools equilibrate pairwise with exchange rates R 3 and R 4 . To consider such effects, and specifically calculate fluxes between metabolite pools directly involved in CO 2 fixation and photorespiration, the model allows overlapping pools with different labelling kinetics to be separated by component analysis. Figure 3 provides examples of how the sequestration of metabolites into different pools was quantified and how the separation of primary and secondary labelling was achieved in the case of F. pubescens. The Fig. 3. Examples for the model-based separation of fast-and slow-turnover pools in the 'Gly' and 'Ser' branches of the photorespiratory pathway and for primary versus secondary labelling and accumulation of C 4 acids. All data are for F. pubescens. example data display carbon incorporation into high-(Gly-I and Ser-I) and low-turnover (Gly-II and Ser-II) pools within the glycine and serine branches of the photorespiratory pathway. They also demonstrate the quantitative separation of the 'active' C 4 carbon pool of C 4 acids from label appearing in carbon atoms 1-3 and in C 4 acids exported to the vacuole. Collectively, these data show that the chosen model is an adequate tool for the calculation of fluxes through the major routes of photosynthetic CO 2 fixation from quantitative 14 CO 2 labelling data.
The relevant fluxes are summarized in Table 1 and complemented by results from radiogasometric measurements performed in parallel with the same set of plants. These independent data show rates of true photosynthesis, total decarboxylation, and photorespiratory CO 2 evolution. They allowed calculating the extent to which photorespiratory CO 2 is re-fixed.
CO 2 can become incorporated into the RPPC either directly with rate R 1 or indirectly via the C 4 pathway with rate R 6 . The sums R 1 +R 6 then represent total CO 2 incorporation from external sources and show an increasing contribution by the C 4 cycle, very low in F. cronquistii, low in F. pubescens, and, as expected, very high in F. trinervia. These total influx rates correspond reasonably well to directly measured rates for true photosynthesis P T , which provides a strong argument for the soundness of all other flux calculations. Higher values for P T (C 3 <C 3 -C 4 <C 4 ) go together with increased rates of sucrose formation (R 5 ; directly measured in Table 1) and C 4 acid accumulation as end-products (R 8 ). Moreover, the photosynthetically active pools of C 4 acids (C 7 ; not listed in Table 1) increased from 13 ± 1 (C 3 ) via 57 ± 19 (C 3 -C 4 ) to 161 ± 39 µmol C m -2 (C 4 ). It is important to note that the increase of C 4 cycle activity from F. cronquistii to F. pubescens (5.8% to 8.3% of P T , calculated as R 6 -R 8 ) is only very small in comparison with the activity of the C 4 cycle in F. trinervia (81.7% of P T ). This suggests that CO 2 accumulation occurs mainly by glycine-shuttling and less by C 4 cycle activity in the bundle sheath of F. pubescens.
Carbon flux through the glycolate cycle, R 2 , is stoichiometrically related to the rate of RuBP oxygenation, R 2 /2. As a result of the operation of CO 2 -concentrating mechanisms in F. pubescens and in F. trinervia, photorespiration-related fluxes become distinctly lower from C 3 towards C 4 metabolism. To determine the true rates of RuBP carboxylation, in addition to the sum of R 1 and R 6 , it was necessary to consider the refixation of CO 2 generated from internal sources. In C 3 and C 3 -C 4 plants, photorespiration is the dominating internal source of CO 2 during photosynthesis. R 2 is stoichiometrically related to photorespiratory glycine decarboxylation as R 2 /4, because one molecule of CO 2 is released per one molecule of serine formed from two glycine molecules. The extent to which refixation occurs must be separately determined. This was done by radiogasometric measurements Keerberg, 1995, 2007), which allowed direct quantification of the sum DEC of photorespiratory glycine decarboxylation plus C 4 acid decarboxylation plus minor CO 2 releasing processes. It is reasonable to assume that all fractions of internally generated CO 2 are re-assimilated with the same efficiency. In combination with the rate R P of CO 2 losses from the leaf (simplifying referred to as photorespiratory CO 2 evolution), this assumption allows assessing the partitioning D between re-fixation and loss of CO 2 from the leaf. The calculated total rates with which Rubisco fixes CO 2 arriving by diffusion from the stomata (R 1 ), from decarboxylation in the C 4 cycle (R 6 ), and from photorespiration (D*R 2 /4) were related to RuBP oxygenation rates (R 2 /2). The comparison Table 1. Carbon fluxes in photosynthetic-photorespiratory carbon metabolism of Flaveria species Values marked with an asterisk represent means ±SE from three measurements on different plants by using a radiogasometric method (Pärnik and Keerberg, 2007). All other values were calculated as means ±SE by multi-component non-linear regression analysis from the time-course of 14 C-incorporation (simultaneous fit to equations 1-4; labelling data from three independent experiments). shows that the resulting in vivo carboxylation-to-oxygenation ratio of Rubisco is more than three times higher in F. pubescens relative to F. cronquistii under the same experimental conditions. Rubisco from C 4 Flaveria species has a somewhat lower affinity to CO 2 , but it is also known that Rubisco from C 3 and C 3 -C 4 Flaveria species show more or less identical kinetics (Bauwe, 1984;Wessinger et al., 1989;Kubien et al., 2008). Since the oxygen compensation point of C 3 plants is only slightly above air levels (Tolbert et al., 1995), plastidial oxygen concentrations are probably close to air oxygen concentrations in F. cronquistii and F. pubescens but presumably also in the C 4 species F. trinervia. Therefore, in a comparison of these species, measurement of in vivo carboxylation-tooxygenation ratios allows the calculation of the relative CO 2 concentration in chloroplasts. Considering the reported K m values of Rubisco for CO 2 , which are even somewhat higher than steady-state internal CO 2 levels, our data suggest that the photorespiratory CO 2 pump elevates the mean intraplastidial CO 2 concentration during steady-state photosynthesis about 3-fold in leaves of the C 3 -C 4 intermediate species F. pubescens relative to the C 3 species F. cronquistii. This is considered to be a sound estimate because small contributions from C 4 photosynthesis are balanced by the operation of a significant fraction of Rubisco at non-elevated CO 2 levels in the mesophyll of F. pubescens.

Supplementary data
Supplementary data can be found at JXB online.
Supplementary data. An explanation of the labelling functions of the model shown in Fig. 1 used for the quantitative analysis of the labelling kinetics.