Drought and Oxidative Load in the Leaves of C₃ Plants: a Predominant Role for Photorespiration?

Abstract

Although active oxygen species are produced at high rates in both the chloroplasts and peroxisomes of the leaves of C₃ plants, most attention has focused on the potentially damaging consequences of enhanced chloroplastic production in stress conditions such as drought. This article attempts to provide quantitative estimates of the relative contributions of the chloroplast electron transport chain and the glycolate oxidase reaction to the oxidative load placed on the photosynthetic leaf cell. Rates of photorespiratory H₂O₂ production were obtained from photosynthetic and photorespiratory flux rates, derived from steady‐state leaf gas exchange measurements at varying irradiance and ambient CO₂. Assuming a 10 % allocation of photosynthetic electron flow to the Mehler reaction, photorespiratory H₂O₂ production would account for about 70 % of total H₂O₂ formed at all irradiances measured. When chloroplastic CO₂ concentration rates are decreased, photorespiration becomes even more predominant in H₂O₂ generation. At the increased flux through photorespiration observed at lower ambient CO₂, the Mehler reaction would have to account for more than 35 % of the total photosynthetic electron flow in order to match the rate of peroxisomal H₂O₂ production. The potential signalling role of H₂O₂ produced in the peroxisomes is emphasized, and it is demonstrated that photorespiratory H₂O₂ can perturb the redox states of leaf antioxidant pools. We discuss the interactions between oxidants, antioxidants and redox changes leading to modified gene expression, particularly in relation to drought, and call attention to the potential significance of photorespiratory H₂O₂ in signalling and acclimation.

INTRODUCTION: OXIDATIVE LOAD AND THE ANTIOXIDANT SYSTEM

It is generally accepted that stress‐induced deregulation of plant metabolism leads to the enhanced production of active oxygen species (AOS), the cellular titre of which is policed by the antioxidant system (Noctor and Foyer, 1998). Both AOS and soluble antioxidants are involved in signalling processes in plants: the picture that is emerging suggests that relatively stable oxidants (H₂O₂) and antioxidants (ascorbate, glutathione) act as sensors of the ‘oxidative load’ on the cell (Noctor et al., 2000). According to this view, mild increases in oxidative load trigger events that lead to acclimation and enhanced resistance, while more severely increased and sustained loads tip the developmental balance towards senescence and death.

A modified balance between AOS generation and the antioxidant system affects the expression of antioxidative enzymes, and is also involved in processes such as the hypersensitive response to pathogen attack (Smirnoff, 1993; Levine et al., 1994; Noctor and Foyer, 1998; Veljovic‐Jovanovic et al., 2001). Both hydrogen peroxide (H₂O₂) and glutathione have been implicated in signalling cascades and in the control of gene expression (Levine et al., 1994; Foyer et al., 1997; Willekens et al., 1997). Drought is an example of an abiotic stress in which components of the antioxidative system are perturbed or up‐regulated (Smirnoff, 1993). It can therefore be inferred that enhanced production of AOS, a compromised capacity to remove AOS or both, elicits acclimatory events during drought (Smirnoff, 1993). Most discussion has centred on accelerated AOS production through side‐reactions in the chloroplast, in particular the formation of superoxide and H₂O₂ linked to auto‐oxidation of components associated with photosystem I (PSI; Mehler reaction). Less attention has been paid to the influence of photorespiration.

PHOTORESPIRATION AND OXIDATIVE LOAD

Photorespiration occurs at high rates in the leaves of C₃ plants (Foyer and Noctor, 2000). Probably the best accepted ‘function’ of this pathway is that of an alternative electron sink. The considerable energy used in photorespiratory C and N recycling lowers the quantum yield of photosynthesis, thereby making light utilization in CO₂ fixation less efficient. This effect could be physiologically advantageous in conditions such as drought stress, where stomatal closure may decrease the availability of CO₂ to the photosynthetic apparatus. Labelling studies by Lawlor and colleagues were among the earliest indications that as water becomes scarce the photorespiratory pathway accelerates relative to net photosynthesis (Lawlor, 1976; Lawlor and Pearlman, 1981). Increased allocation of energy to photorespiration could mitigate deleterious effects such as photoinhibition by allowing metabolism to continue using the products of photosynthetic electron transport (Osmond and Grace, 1995). However, while the operation of photorespiration may decrease the probability of photoinhibition and attenuate AOS production in the chloroplast, the photorespiratory C recycling pathway involves the obligatory production of H₂O₂ in the peroxisomes through the action of glycolate oxidase. Increased photorespiratory flux during drought could, therefore, significantly exacerbate the oxidative load on the photosynthetic cell. While considerable attention has been paid to the significance of photorespiration as an alternative sink for light energy, little or none has focused on the potential importance of the attendant high rates of H₂O₂ generation.

ESTIMATION OF PHOTORESPIRATORY RATES BY DERIVATION OF THE RATE OF RUBP OXYGENATION (V_O) FROM GAS EXCHANGE MEASUREMENTS

Photorespiration is masked by photosynthetic O₂ evolution and CO₂ assimilation. Even where labelling experiments enable O₂ consumption or CO₂ evolution to be monitored in the light, accurate quantification of photorespiration is complicated by the concurrent operation of other processes that also consume O₂ and/or release CO₂. One convenient method that is largely able to overcome these difficulties is derivation of the rate of photorespiration from measured rates of CO₂ uptake through modelling (Sharkey, 1988; von Caemmerer, 2000), which has the advantage that photorespiratory flux can be derived solely from measurements of leaf gas exchange. This approach requires estimation of chloroplastic CO₂ and O₂ concentrations (C_c and O_c, respectively), as well as the rate of non‐photorespiratory CO₂ release. The first two parameters can be calculated (C_c) or assumed (O_c) with reasonable accuracy (see legend to Fig. 1). To calculate C_c from the intercellular CO₂ con centration, C_i, it is necessary to use a value for the mesophyll transfer conductance, g_i. This value has been estimated at around 0·3–0·6 mol m^–2 s^–1 for species with photosynthetic rates similar to those of wheat and potato (Loreto et al., 1992; von Caemmerer et al., 1994). We used a value of 0·3 since at ambient CO₂ this gave C_c/C_i close to 0·7 (von Caemmerer and Evans, 1991). To estimate non‐photorespiratory CO₂ release in the light it was assumed for the data of Figs 1–5 that this value was equal to 50 % of the rate measured in the dark. The assumption concerning non‐photorespiratory CO₂ release only significantly affects the derived values of rate of ribulose‐1,5‐bisphosphate (RuBP) oxygenation (v_o) and rate of RuBP utilization (v_RuBP) at very low ambient CO₂ (Table 1).

Fig. 1. Rate of net photosynthesis and derived parameters plotted against irradiance (A, C, E) and ambient CO₂ concentration (B, D, F) for attached leaves of wheat (circles) and potato (triangles). A and B, Net CO₂ uptake; C and D, rate of RuBP oxygenation (v_o); E and F, rate of total RuBP utilization (v_RuBP). Wheat (Triticum aestivum ‘Cannon’) and potato (Solanum tuberosum ‘désirée’) were grown in soil and slow‐release fertilizer to the age of 4–5 weeks (wheat) or 6–9 weeks (potato) in controlled‐environment glasshouses (day length 14 h, day/night temperature 22/18 °C, irradiance 250 µmol quanta m^–2 s^–1). Plants were transferred to the laboratory for measurements of steady‐state photosynthesis. CO₂ and H₂O exchange were measured in a multi‐chamber system designed and developed at Rothamsted by Lawlor and colleagues (Paul et al., 1990). The measuring systems consisted of four or six temperature‐ and humidity‐controlled chambers, illuminated from above by floodlamps and connected to a gas‐mixer and infra‐red gas analyser (IRGA). All experiments were conducted at 20 °C, 50 % relative humidity and 21 % O₂. CO₂ composition was controlled by a gas mixer and irradiance by neutral density sheets. For wheat, middle sections of the fourth leaf were used; for potato, the half of a fully expanded leaf distal from the petiole was introduced into the chamber. For irradiance curves, CO₂ was 360 µl l^–1; for CO₂ curves, irradiance was 650–750 µmol m^–2 s^–1 at the leaf surface. Leaves were incubated for 30 min in darkness, before illumination at each condition until a steady‐state rate of photosynthesis was reached (30–40 min). To calculate v_o and v_RuBP, the Rubisco specificity factor (S_rel) was taken to be 110 (Keys, 2000) and the chloroplastic oxygen concentration (O_c) was assumed to be that of water in equilibrium with air at 20 °C (276 µm). The chloroplastic CO₂ concentration (C_c) was derived from C_i by taking a CO₂ transfer conductance through the mesophyll (g_i) of 0·3 mol m^–2 s^–1 (von Caemmerer et al., 1994) and assuming that the rate of CO₂ uptake affects C_c relative to C_i as in Ruuska et al. (2000): C_c = C_i – A/g_i. C_c was converted to a molar concentration by applying a CO₂ solubility constant at 20 °C of 0·0392 mol l^–1 (von Caemmerer, 2000). The ratio of oxygenation to carboxylation was calculated as O : C = (1/S_rel)(O_c/C_c) and v_o was derived according to Sharkey (1988): v_o = (A – R)/(1/O : C – 0·5), where R is non‐photorespiratory CO₂ release in the light (negative value). The ‘real’ or gross rate of carboxylation at Rubisco was derived as v_c = A + 0·5 v_o – R, and the total rate of RuBP utilization as v_RuBP = v_o + v_c.

Table 1.

Effect of the rate of non‐photorespiratory CO₂ evolution on parameters derived from measurements of net CO₂ uptake (A) by attached wheat leaves at different C_a

		vo		vc		vRuBP
Ca (µl l–1)	A	1	2	1	2	1	2
61	2.3	15.5	12.0	9.37	8.29	24.9	20.3
166	10.5	12.0	11.2	15.8	16.1	27.8	27.4
243	15.8	10.6	10.2	20.4	20.9	31.0	31.0
368	20.9	7.32	7.09	23.9	24.5	31.2	31.5
578	24.6	4.35	4.23	26.1	26.7	30.4	30.9
704	25.0	3.34	3.25	26.0	26.7	29.4	29.9
960	21.1	1.81	1.76	21.4	22.0	23.2	23.8

n = four leaves on different plants, standard errors 15 % of value or less. Values in column 1 are calculated assuming that non‐photorespiratory CO₂ evolution in the light is equal to the rate of dark respiration [after 30 min darkness, mean rate (n = 4) was –0·68 ± 0·05 µmol CO₂ m^–2 s^–1]. Values in column 2 are calculated assuming that non‐photorespiratory CO₂ evolution is zero in the light. At each C_a value, CO₂ uptake was monitored for 30 min at an irradiance of 650–750 µmol m^–2 s^–1. All rates are expressed in µmol m^–2 s^–1.

Figure 1 shows typical irradiance and CO₂ response curves of wheat and potato leaves. For both plants, net CO₂ uptake by leaves responded similarly to increasing irradiance and CO₂ (Fig. 1A and B). Values of v_o increased with irradiance in a similar fashion to net CO₂ uptake (Fig. 1, compare A and C). When photosynthesis was manipulated by CO₂ concentration, v_o changed antagonistically to net CO₂ uptake in wheat, with the highest v_o occurring at the lowest net CO₂ uptake and vice versa (Fig. 1, compare B and D). The response was similar in potato except that v_o decreased at the lowest CO₂ concentration (approx. 60 µl l^–1). We have also observed this effect in some experiments with wheat (see Fig. 5) and the effect of changing CO₂ availability on absolute photorespiratory flux is discussed below. While the light response of v_RuBP was similar to that of net CO₂ uptake and v_o (Fig. 1E), v_RuBP was generally much less responsive to CO₂ than these processes (Fig. 1F). The broad optimum illustrates the capacity of oxygenation to substitute for carboxylation in the utilization of RuBP and hence photosynthetic energy. This is reflected by the increase in the ratio of RuBP oxygenation to RuBP carboxylation (O : C) as the ambient CO₂ concentration (C_a) falls (Fig. 2).

Fig. 5. Changes in H₂O₂ production with ambient CO₂ concentration in attached wheat leaves. The production in photorespiration (A), O₂ reduction in the chloroplast (B) and the total amount generated (C) were calculated as described for Fig. 4. Different symbols show three independent experiments. Circles show values that are the means ± s.e. of three different plants, the same set of plants being measured at each CO₂ concentration. Squares and triangles show means ± s.e. of four different plants, with a different set of plants being measured at each CO₂ concentration.

Fig. 2. Modelled data derived from measurements of net CO₂ uptake and intercellular CO₂ concentration (C_i) in attached wheat leaves at different CO₂ concentrations. C_a and C_c, ambient and chloroplastic CO₂ concentration, respectively; O : C, ratio of RuBP oxygenation to RuBP carboxylation. Data are shown from three independent experiments (methods as in legend to Fig. 1). For each experiment, values are means ± s.e. of three (circles) or four (triangles, squares) leaves. Circles, Experiment in which four leaves, each attached to a different plant, were measured in the steady‐state (30–40 min illumination at the C_a values indicated). Triangles and squares, Experiments in which each measurement at each C_a value was carried out with different leaves, each attached to a different plant. Irradiance was 675–838 µmol m^–2 s^–1, temperature was 20 °C and gas composition was 21 % O₂, CO₂ concentration (C_a) as indicated and balance N₂.

DO THE MODELLED DATA PROVIDE AN ACCURATE MEASURE OF PHOTORESPIRATION?

During drought stress, the calculated value of C_c was predicted to decline sharply at relatively mild relative water deficits (Tourneux and Peltier, 1995), and a good estimate of this parameter is key to accurate modelling of the rate of photorespiration (see legend to Fig. 1). In the data shown in Fig. 2, C_c/C_a values were highest at low CO₂, where net CO₂ utilization is relatively low, and at supra‐atmospheric CO₂, where external CO₂ is high relative to net CO₂ uptake (Fig. 2A). Intermediate values of C_a gave somewhat lower values (Fig. 2A). Using data from ¹⁸O₂ uptake measurements, Tourneux and Peltier (1995) calculated a C_c/C_a value of 0·56 at atmospheric CO₂, assuming that non‐photorespiratory CO₂ evolution is not inhibited by light. In agreement with these data, analysis of the gas exchange data obtained in the three experiments of Fig. 2 gave C_c/C_a as 0·61 ± 0·03 (circles), 0·58 ± 0·03 (triangles) and 0·56 ± 0·06 (squares) at C_a = 345–368 µl l^–1 (Fig. 2A).

Oxygenation and carboxylation are often considered the major sinks for photosynthetic reducing power, and this notion is supported by comparisons of gas exchange analysis with chlorophyll fluorescence analysis (Ruuska et al., 2000). If this is so, the relative rate of linear electron transport [Φphotosystem II (PSII) × irradiance] should correlate with the total demand of oxygenation and carboxylation for electrons, denoted J_e. We calculated J_e from values of net CO₂ uptake measured under widely varying conditions of irradiance, C_a, or ambient O₂ concentration, and compared the values obtained to the relative rate of linear electron transport calculated from chlorophyll fluorescence measurements (for details, see legend to Fig. 3). The data show that the two parameters are related by a straight line that passes close to the origin and that has a reasonable correlation coefficient, both for wheat and potato (Fig. 3C and D). As well as the linear relationship observed, the calculations of J_e gave values close to those predicted by chlorophyll fluorescence analysis, even where the measured rates of net CO₂ uptake were very different. For example, wheat leaves at 230 µl l^–1 CO₂ gave a measured rate of net CO₂ uptake of approx. 15 µmol m^–2 s^–1; in wheat leaves at 2 % O₂ the measured rate was 28 (Fig. 3A). However, fluorescence analysis predicted almost identical rates of electron transport and this was borne out by the J_e values calculated from gas exchange (Fig. 3C). These data suggest that the modelled data provide a reasonably good estimate of the rate of photorespiration.

Fig. 3. Relationship between net CO₂ uptake (A and B), the derived rate of electron flow associated with carboxylation and oxygenation (C and D), and the relative rate of linear electron transport, calculated from chlorophyll fluorescence quenching. A and C, Attached leaves of wheat; B and D, attached leaves of potato. Chlorophyll fluorescence was measured simultaneously with gas exchange using an oxy‐blot fluorometer. The fluorescence excitation beam and emission signal, as well as saturating light flashes, were passed down a fibre‐optic held in fixed position for each chamber by mountings built at Rothamsted. The optic was held at 45° to the leaf surface and monitored that part of the leaf which corresponded to the central third of the leaf portion in the chamber. All leaves used gave F_v/F_m > 0·8 after 30 min dark incubation. Photosynthesis was then induced by illumination under conditions of varying light (circles), CO₂ (triangles) or O₂ (squares). On attainment of the steady‐state rate of net CO₂ uptake (30–40 min illumination), the fluorescence of leaves in each chamber was monitored sequentially for 3 min, during which time F_m′ was measured by applying two saturating flashes (4000 µmol quanta m^–2 s^–1, 2 s duration, 2 min between flashes). The photochemical yield of PSII (ΦPSII) was calculated according to Genty et al. (1989) and J_e was calculated as 4 (v_o + v_c), assuming that each carboxylation event gives rise to 2 PGA and that each oxygenation event produces 1·5 PGA + 0·5 NH₃, whose reassimilation in the chloroplast involves ferredoxin‐dependent glutamate synthase (2 e^– per NH₃). All points represent the means of two or four different plants (standard errors are shown where values are means of four). For wheat, black circles show data at ‘control’ conditions (irradiance = 750 µmol m^–2 s^–1, CO₂ = 382 µl l^–1, O₂ = 21 %). White and grey circles, Gas composition as for ‘controls’ but irradiance = 112 and 340 µmol m^–2 s^–1, respectively. Triangles show data at the same light and O₂ tension except CO₂ = 123 (white), 232 (grey) and 1055 (black) µl l^–1. Squares show data at control light and CO₂ but O₂ = 2 % (white) and 7 % (grey). For potato, ‘control’ conditions were as for wheat and are denoted by grey squares. White and black circles, Control gas composition but irradiance = 235 and 1250 µmol m^–2 s^–1, respectively. Triangles show data at control light and O₂ but CO₂ = 60 (white), 96 (grey with cross), 150 (grey) and 220 (black) µl l^–1. Squares, Conditions as for squares in wheat data.

DOES PHOTORESPIRATORY FLUX INCREASE IN ABSOLUTE TERMS WHEN CO₂ AVAILABILITY FALLS?

The above data show the increased allocation of energy to photorespiration as CO₂ availability decreases. Similar effects are known to occur during drought. The combined use of chlorophyll fluorescence and gas exchange showed that the total electron transport rate decreased much less than net CO₂ uptake with decreasing C_i or during drought stress (Cornic and Briantais, 1991). Mass spectrometric measurements of ¹⁸O₂ uptake and net O₂ evolution demonstrated that during drought, gross O₂ release at PSII was maintained much higher than net O₂ evolution in potato (Tourneux and Peltier, 1995) or net CO₂ uptake in wheat (Biehler and Fock, 1996). Lastly, electron transport rates derived from chlorophyll fluorescence analysis decreased less than net CO₂ uptake in droughted barley leaves (Wingler et al., 1999).

Although it is clear that photorespiration increases relative to net CO₂ uptake during drought, it is less evident whether flux through the photorespiratory pathway increases in absolute terms. This question is relevant to the oxidative effect mediated by photorespiration during drought. Neither chlorophyll fluorescence nor mass spectrometric analysis using ¹⁸O₂ uptake can provide an unequivocal answer to this question because of possible concomitant changes in the rate of the Mehler reaction. Biehler and Fock (1996) approached the problem by estimations of ¹⁴C‐glycolate labelling, which was reported to decrease during drought. A recent study by the same group quantified photorespiration in tomato leaves by measuring ¹²CO₂ evolution in the first few seconds after switching to supplying ¹³CO₂ only, and subtracting estimated rates of non‐photorespiratory CO₂ evolution (Haupt‐Herting et al., 2001). It was reported that although photorespiration increased markedly compared with net CO₂ uptake, and remained high in droughted leaves, decreased water potential led to decreased photorespiratory flux (Haupt‐Herting et al., 2001). On the other hand, studies of several barley mutants with decreased amounts of enzymes involved in the photorespiratory pathway suggest that flux does increase to some extent during mild drought stress (Wingler et al., 1999).

The modelled data shown in Fig. 1 suggest that photo respiration will increase in absolute terms as chloroplastic CO₂ concentration becomes restricted. In most of our experiments, we have observed an optimum for calculated rates of v_o at C_a equal to 100–200 µl l^–1. Whether or not v_o declines at very low CO₂ probably depends on factors such as irradiance (Cornic and Briantais, 1991), since one of the factors that could restrict photorespiratory flux at low CO₂ is a limiting rate of RuBP regeneration (von Caemmerer, 2000). Additional evidence for absolute increases in photo respiratory flux at CO₂ concentrations below atmospheric comes from other experiments in which we measured leaf amino acids by freeze‐clamping leaves during steady‐state photosynthesis (Novitskaya et al., 2002). In these experiments, the Gly/Ser ratio correlated with the calculated value of v_o, whether this was manipulated by CO₂, irradiance or O₂. In wheat leaves, Gly/Ser was 4·8 ± 1·1, 4·8 ± 1·3, 8·0 ± 1·2, 8·3 ± 0·8 and 7·3 ± 0·9 at C_a values (µl l^–1) of 350, 210, 167, 99 and 66, respectively. In potato leaves, the ratios were 4·3 ± 1·1, 5·7 ± 1·0 and 7·5 ± 0·3 at C_a = 300, 218 and 105 µl l^–1, respectively (all values means ± s.e. of four different plants; Novitskaya et al., 2002).

HOW MUCH H₂O₂ IS PRODUCED THROUGH PHOTORESPIRATION AND THE MEHLER REACTION?

The stoichiometry of the glycolate pathway dictates that H₂O₂ generation in photorespiration equals v_o. The rate of H₂O₂ production in the chloroplast is difficult to estimate, but a value of 10 % is often cited for the proportion of electrons flowing to O₂ at PSI (Foyer and Noctor, 2000). Some authors have estimated higher values (e.g. Miyake and Yokota, 2000), while studies at very high concentrations of CO₂ and on plants with low Rubisco suggest that the proportion is probably considerably less than 10 % (Ruuska et al., 2000). If the ATP yields of linear electron transport are sufficient to satisfy almost all the requirements of stromal metabolism, then it remains as yet unclear how the photosynthetic system could allow a high proportion of electron flow to O₂ in the Mehler reaction (Noctor and Foyer, 2000). In view of these considerations and the incisive data of Ruuska et al. (2000), we consider here the value of 10 % as a likely maximum value for the proportion of electrons allocated to the Mehler–peroxidase reaction.

Assuming this constant value, then the rate of H₂O₂ formation via both the Mehler reaction and photorespiration increases more or less in proportion with net CO₂ uptake, when the latter is manipulated by changes in irradiance (Fig.). At all irradiances, photorespiratory H₂O₂ production accounts for about 70 % of total H₂O₂ formed (Fig. 4). When chloroplastic CO₂ concentration falls, as is likely during drought, photorespiration becomes even more dominant in the production of H₂O₂ (Fig. 5). The data represented by circles in Fig. 5 show that the calculated rate of H₂O₂ generation increases gradually as CO₂ is decreased, so that maximum rates are observed at the lowest C_a value (60 µl l^–1). The other two curves suggest an optimum CO₂ concentration for photorespiratory H₂O₂ generation, below which photorespiration and the attendant H₂O₂ production decline (Fig. 5; compare data for potato leaves in Fig. 1D). It should be noted that the measurements at very low CO₂ are those on which the largest error is to be expected: first, it is under these conditions that the assumption concerning the rate of non‐photorespiratory CO₂ release is most influential (Table 1); and secondly, small changes or inaccuracies in the measurement of net CO₂ uptake and/or C_i can produce large changes in the calculated value of v_o. Whatever the response of photorespiratory flux at very low CO₂, all three curves suggest that photorespiratory H₂O₂ production will increase (by 1·5‐ to 2‐fold) at C_c concentrations below those maintained in well‐watered plants in air (Fig. 5). Our C_a curves were carried out at 650–750 µmol m^–2 s^–1, which at air CO₂ concentrations drives a rate of photorespiratory H₂O₂ generation that is about 85 % of the value predicted at the near‐saturating irradiance of 1100 µmol m^–2 s^–1 (Fig. 4). The data of Fig. 4 predict that at an irradiance of 650–750 µmol m^–2 s^–1, mild drought may drive photorespiratory H₂O₂ generation faster than saturating irradiance in non‐drought conditions (compare Figs 4 and 5).

Fig. 4. Effect of irradiance on modelled rates of H₂O₂ production in photorespiration and following superoxide production at PSI. Triangles, H₂O₂ production in the Mehler reaction; cicles, H₂O₂ production in photorespiration; squares, total H₂O₂ production. Steady‐state photo synthesis was measured in attached leaves of wheat at C_a = 350 µl l^–1, 212 and irradiance as indicated. Photorespiratory H₂O₂ production is equal to v_o, derived as in Fig. 1. H₂O₂ production in the chloroplast assumes that 10 % of total electron flow through the photosynthetic electron transport chain is linked to the Mehler–peroxidase reaction. Calculation of total electron flow assumes no sinks other than the Mehler–peroxidase reaction, photorespiration and CO₂ fixation. The utilization of electrons associated with photorespiration and CO₂ fixation was calculated as described for Fig. 3. If the Mehler reaction accounts for 10 % of photosynthetic electrons, then total electron flow, J_II, is 40/9 (v_o + v_c), and steady‐state H₂O₂ production in the chloroplast = 0·025 J_II (O₂ + 2e^– + 2H⁺ = H₂O₂, plus two electrons required in the peroxidatic conversion of H₂O₂ to H₂O = four electrons per H₂O₂ generated).

Low C_a values drive up the total oxidative load on the photosynthetic cell (Fig. 5A). It is clear that under most conditions, and particularly when CO₂ is scarce, most of the H₂O₂ produced is formed by glycolate oxidation. The only condition in which H₂O₂ is produced faster in the chloroplast than in the peroxisome is at artificially high C_a (Fig. 5B). It is possible that in contrast to our assumption of a constant electron allocation to the Mehler reaction, the allocation increases as C_c drops during drought, although the data shown in Fig. 3, particularly for wheat, provide little evidence of large changes in sinks other than CO₂ fixation and photorespiration. Although the Mehler reaction could increase during drought, we can calculate from our data that this reaction would have to account for approx. 38–30 % of total electron flow in order to equal the rate of photorespiratory H₂O₂ production at C_a = 61–166 µl l^–1. In a recent study of watermelon leaves at different CO₂ concentrations, a considerable residual electron flow, which could not be accounted for by electron flow linked to CO₂ fixation and photorespiration, was present at all ambient CO₂ concentrations (Miyake and Yokota, 2000). This residual flow included a component that required O₂ higher than 1·7 % and which the authors attributed to a physiological Mehler reaction linked to oxidation of fer redoxin and/or monodehydroascorbate reductase (Miyake and Yokota, 2000). Expressed as a fraction of total electron flow, this component increased about two‐fold as C_i dropped to around 50 µl l^–1 though the increase in absolute rate was less marked (Miyake and Yokota, 2000). Using a combined approach of ¹⁸O₂ measurements and estimations of photorespiratory flux through glycolate labelling, Biehler and Fock (1996) concluded that the Mehler reaction accounted for less than 15 % of photosynthetic electrons in unstressed wheat and increased to approx. 29 % in wheat subjected to drought. It therefore remains possible that the Mehler reaction increases during drought, but exactly how this could happen remains unclear. Our current understanding suggests that the two factors which most strongly influence the rate of the Mehler reaction in vivo will be irradiance (Fig. 4) and ATP demand (Foyer and Noctor, 2000). While photorespiration has a slightly higher ATP : reductant demand than CO₂ fixation, fairly small increases in the Mehler reaction would be sufficient to support the rise in O : C during drought, even assuming the complete absence of other ATP‐generating reactions (e.g. cyclic electron transport through PSI, nitrite reduction). Drought‐induced uncoupling of the thylakoid electron transport chain from photophosphorylation cannot be completely excluded, though this idea receives little support from the well known increases in non‐photochemical quenching observed in droughted leaves (Demmig et al., 1988; Cornic and Briantais, 1991; Biehler and Fock, 1996; Wingler et al., 1999).

ANTIOXIDANT DEFENCES AND H₂O₂: POTENTIAL INFLUENCE IN DROUGHT‐ASSOCIATED SIGNALLING

Earlier concepts that active oxygen species (AOS) exert their effects through physicochemical damage have been replaced by one that recognizes the active role of AOS in signalling. In particular, drought‐induced increases in superoxide and H₂O₂ by the thylakoid membranes are well documented (e.g. Bartoli et al., 1999). The extent of accumulation of these oxidants is determined by the capacity of the major redox buffers of the plant cell, ascorbate and glutathione, and the associated enzymes of antioxidant defence. Moreover, many studies have documented the responses of these antioxidants to drought and the general enhancement of antioxidant defence that accompanies prolonged exposure to water deficits (Smirnoff, 1993; Iturbe‐Ormaetxe et al., 1998). Several of these have used transformed plants with increased expression of antioxidant enzymes to increase drought tolerance. In particular, overexpression of either superoxide dismutase (SOD) or ascorbate peroxidase in the chloroplast (McKersie et al., 1996; Yan et al., 2002) has been shown to confer a degree of extra protection against water deficits. Similarly, overexpression of Fe‐SOD in poplar chloroplasts was found to protect the photosynthetic electron transport system from over‐reduction at low CO₂ partial pressures (Arisi et al., 1998). This is a particularly interesting observation as SOD is already present at high activities in chloroplasts. Moreover, SOD catalyses one of the fastest reactions known to biology and the enzyme‐catalysed reaction is limited only by the rate of diffusion. It is therefore unlikely that enhanced protection of PSII activity is due simply to increased enzyme capacity. An alternative explanation is that Fe‐SOD acts as a metabolite channel at PSI, not only catalysing superoxide dismutation, but also channelling H₂O₂ along the thylakoid membrane surface, directly to ascorbate peroxidase (APX), which then reduces H₂O₂ to H₂O. It has previously been suggested that SOD overexpression facilitates increased stress tolerance via enhancement of tissue H₂O₂ contents, which signal changes in gene expression leading to general increases in defence responses. Although this has never been demonstrated, the systemic accumulation of H₂O₂ is associated with the expression of defence genes in the hypersensitive response and with wounding or mechanical stimulation (e.g. Alvarez et al., 1998). H₂O₂ can act as a local signal, for example, as a second messenger of hormone action (Orozco‐Cardenas et al., 2001) or, in the extreme case, as a signal for hypersensitive cell death.

It was initially suggested that an oxidative burst is involved in the triggering of drought stress responses (Shinozaki and Yamaguchi‐Shinozaki, 1997). Emerging evidence is confirming this notion. H₂O₂ and ABA interact in controlling the stomatal aperture, and both are involved in the regulation of Ca²⁺ channel function (Pei et al., 2000). The ATMPK3 gene, encoding a mitogen‐activated protein kinase (MAPK), is involved in responses to changes in osmolarity, and is activated by H₂O₂. The subsequent H₂O₂‐induced MAPK cascade acts to repress auxin responses (Kovtun et al., 2000). ABA synthesis is regulated by redox factors: antioxidants block drought‐induced ABA accumulation (Jia and Zhang, 2000). The pools of the endogenous antioxidants ascorbate and glutathione could therefore be regulators of ABA‐linked signalling in vivo. It is important to note that the apoplast contains large amounts of ascorbate, but little or no glutathione. In addition, there is little capacity for ascorbate regeneration in the apoplast, and monodehydroascorbate and dehydroascorbate have either to be re‐reduced at the membrane surface or returned to the cytosol for re‐reduction. This means that large redox changes resulting from hormone or stress‐induced H₂O₂ production are more likely to occur in the apoplast than in the cytoplasm because redox buffering capacity is relatively low in the first compartment. There is therefore potential for multiple interactions between H₂O₂, antioxidants and drought‐induced hormones. Mutants are likely to be invaluable in uncovering such interactions, as they enable the effects of perturbations in compartment‐specific components to be elucidated.

PERTURBATION OF ANTIOXIDANT POOLS BY PHOTORESPIRATION

Extra‐chloroplastic antioxidant capacity in C₃ plants is dominated by peroxisomal catalase. This is undoubtedly because of the need to cope with the abundant production of photorespiratory H₂O₂ at high light and/or low CO₂ (Figs 4 and 5). Catalase is energy‐efficient, since it catalyses a dismutation reaction that does not consume reductant. But the enzyme has a low affinity for H₂O₂, and this may explain the presence of other antioxidative enzymes in the leaf peroxisome (Foyer and Noctor, 2000). However, the isolation of a barley mutant deficient in catalase via a ‘photorespiratory’ screen underlined the indispensability of the enzyme (Kendall et al., 1983). This mutant can grow and set seed under conditions where photorespiration is suppressed, but transfer to air in the light causes characteristic lesions, loss of vigour and eventually leaf death. Smith et al. (1984) reported that prior to the appearance of these symptoms, the leaf glutathione pool became both very oxidized and increased in size by about six‐fold (within 4 d). These effects were confirmed in transformed tobacco into which a construct containing an antisense catalase coding sequence had been introduced (Willekens et al., 1997). In contrast, maize catalase mutants show no marked phenotype (Scandalios, 1994), reinforcing the tight link between the major leaf peroxisomal form of this enzyme and photorespiration.

We have reinvestigated the kinetics of the perturbation of the leaf glutathione pool on transfer of the mutant from 0·6 % CO₂ to air (Fig. 6). In addition, we developed a reliable method to assay leaf H₂O₂ (Veljovic‐Jovanovic et al., 2002) and also measured the leaf ascorbate redox state, to ascertain whether changes in glutathione redox state reflect a generalized cellular oxidation. In both the mutant and the wild‐type, H₂O₂ was increased by transfer to air. It should be noted that although photorespiration was increased after transfer from 0·6 % CO₂, the irradiance in the chamber (approx. 250 µmol m^–2 s^–1) was not sufficient to drive very high rates of photorespiratory H₂O₂ production. A rate of about 5 µmol H₂O₂ m^–2 s^–1 can be predicted at an irradiance of 250 µmol m^–2 s^–1. Assuming a leaf chlorophyll content of 100 mg m^–2, this rate would entail a total production of photorespiratory H₂O₂ of around 5 mmol mg^–1 chl in 48 h (2 d of 14 h light). Even though transfer to air caused a measurable increase in H₂O₂ in both types of plant, the increase was no more than 50 nmol mg^–1chl within this time (Fig. 6). It is therefore clear that even in the mutant, 99·999 % of the H₂O₂ produced was metabolized. Although the mutant showed a rate of photosynthesis that was 30 % lower than that of the wild‐type, rates were maintained in both types of plants for the duration of the experiment (data not shown). Therefore, within the time period of the experiment H₂O₂ metabolism did not occur through routes that caused appreciable cellular damage.

Fig. 6. Changes in leaf H₂O₂ and the redox states of the principal soluble leaf antioxidants, ascorbate and glutathione, in wild‐type barley (open circles) and a mutant deficient in catalase (closed circles). Barley (Hordeum vulgare, var. Maris Mink) and the catalase‐deficient mutant (RPr 79/4; Kendall et al., 1983) were grown under the same conditions as wheat (Fig. 1), except that ambient CO₂ was artificially maintained at 0·6 %. Suppression of photorespiration is necessary for healthy growth of the mutant (Kendall et al., 1983). Six weeks after seeds were sown, plants were transferred to a chamber in which conditions were identical to that in which they had been grown, except that the ambient CO₂ concentration was 400 µl l^–1. Leaf samples were taken at the indicated times after transfer, for determination of H₂O₂, glutathione and ascorbic acid. H₂O₂ was extracted and determined by a modified peroxidase‐coupled assay (Veljovic‐Jovanovic et al., 2002). Reduced and total ascorbate, and total and oxidized glutathione, were assayed as described in Foyer et al. (1995).

The ability of photorespiration to perturb leaf antioxidant redox states is evidenced by the transient oxidation of ascorbate and glutathione in both mutant and wild‐type (Fig. 6). Only in the case of the glutathione pool in the mutant is the oxidation sustained, however, and this occurs despite no appreciable increase in leaf H₂O₂ or ongoing oxidation of the ascorbate pool. The oxidation of glutathione is akin to observations in barley leaves subjected to severe drought, where the oxidized form of glutathione increases from a typical value of 5 % or less to around 40 % of the total pool (Smirnoff, 1993). Several differences exist, however, between the changes in the glutathione pool in the catalase mutants and those observed during drought. First, unlike the changes in glutathione in the catalase mutants, drought‐induced perturbation of the glutathione redox state does not always increase the total pool of glutathione. Secondly, several reports suggest that leaf ascorbate pools become smaller and also, sometimes, more oxidized during drought (Smirnoff, 1993). In plants with low catalase activity, however, both the redox state (Fig. 6) and total content (data not shown) of ascorbate were unaffected by transfer to photorespiratory conditions. In tobacco with low catalase, an increased ascorbate pool was associated with faster photorespiration, but this may well have been an irradiance effect since, in this study, accelerated photorespiration was achieved by transfer from low to high light (Willekens et al., 1997). We are currently analysing the effects of drought on leaf ascorbate and glutathione in wheat.

Can it be assumed that in wild‐type plants catalase capacity is so high that drought‐induced increases in photorespiratory H₂O₂ production will never be perceived as an oxidative stress? The existence of peroxisomal isoforms of ascorbate peroxidase may be rationalized by the need to prevent leakage of H₂O₂ into the cytosol (Foyer and Noctor, 2000). It cannot be discounted that the operation of peroxisomal APX, together with other enzymes of the ascorbate–glutathione cycle, could be an important route through which high rates of photorespiration could impact on the redox state of glutathione. The experiment of Fig. 6 might be considered of doubtful physiological relevance: the transient perturbation of the glutathione pool in wild‐type plants could be attributable to a lag period before catalase becomes fully synthesized or active after transfer from high CO₂. Nevertheless, the data do underline the potential of high rates of photorespiratory H₂O₂ production to produce specific perturbations in important components of leaf redox homeostasis. Moreover, while catalase is present in the leaf peroxisomes of C₃ plants in very high amounts, the enzyme turns over rapidly and must be continually resynthesized (Hertwig et al., 1992). In his review of the relationship between drought and AOS production, Smirnoff (1993) noted that ‘it would be instructive to know if turnover of catalase is influenced by water deficit’. This remains an important question. Salt stress engendered decreases in catalase activity in rye leaf pieces, as well as net oxidation of both ascorbate and glutathione and loss of chlorophyll (Streb and Feierabend, 1996). In these experiments, however, leaf pools of glutathione and ascorbate were not significantly increased. In contrast, inhibition of catalase with aminotriazole led to a net accumulation of both reduced and oxidized forms of glutathione, whereas ascorbate oxidation did not occur and chlorophyll loss was less significant (Streb and Feierabend, 1996). There were therefore clear differences between salt stress and catalase inhibition in these experiments. As in our experiments with catalase‐deficient barley, appreciable increases in leaf H₂O₂ were not observed in either salt stress or as a result of catalase inhibition (Streb and Feierabend, 1996). Using wheat, we are currently investigating the response to drought of transcripts encoding catalase and APX.

CHLOROPLASTIC AND PEROXISOMAL AOS PRODUCTION: ROLES IN DAMAGE AND SIGNALLING

Most studies suggest that drought‐induced decreases in net CO₂ uptake are due to stomatal limitations, and the data we present are consistent with this notion. Modelling photorespiratory rates as a function of C_a suggests that photorespiratory flux will increase during drought. Because of the difficulties of relating C_i to leaf water status (Sharkey and Seemann, 1989), it is difficult to know at what point during the progression of drought increased photorespiratory flux will occur. The analysis of Wingler et al. (1999) in barley suggests that absolute flux began to increase at leaf water potentials as high as –1 MPa (which were reached after 8–10 d of drought in their experiments) and continued to increase until a water potential of –1·5 MPa was reached. This may well correspond to relatively mild decreases in C_i, perhaps equivalent to those observed here at C_a values in the range 100–200 µl l^–1, where photorespiratory flux was maximal. Thereafter, further decreases in C_i may diminish absolute rates of RuBP oxygenation. Our data suggest that when RuBP oxygenation is maximal, the total oxidative load posed on the cell will owe more to photorespiratory H₂O₂ production than to chloroplastic events. From the point of view of physicochemical damage, however, AOS produced in the thylakoids might be locally very dangerous. Although the chloroplast has robust antioxidant defences, production of the hydroxyl radical through iron‐catalysed reduction of H₂O₂ by superoxide or ascorbate may pose a serious threat. Another dangerous species for photosynthetic function could be singlet oxygen produced in the thylakoid membrane. It is clear, however, that the increased oxidative load that accompanies most, if not all, stresses exerts its effect as much through modifications in gene expression as through physicochemical damage. Particular attention has been paid to oxidative events occurring at the apoplast, but we also call particular attention to the significance of photorespiratory H₂O₂ in signalling and acclimation. Photosynthetic events can modify nuclear gene expression (Karpinski et al., 1997; Pfannschmidt et al., 2001) but the intermediaries are ill‐defined. In C₃ plants at least, the photorespiratory pathway enables the light‐dependent deposition of H₂O₂ outside the chloroplast where this oxidant, or antioxidants with which it interacts, may be involved in triggering acclimatory events in abiotic stresses such as drought.

ACKNOWLEDGEMENTS

The authors thank BBSRC for financial support. S.V.J. was the recipient of two Royal Society fellowships. L.N. was the recipient of a Rothamsted International fellowship.

Received: 18 July 2001; Returned for revision: 1 November 2001; Accepted: 12 November 2001.