Peroxisomal Malate Dehydrogenase Is Not Essential for Photorespiration in Arabidopsis But Its Absence Causes an Increase in the Stoichiometry of Photorespiratory CO 2 Release

Peroxisomes are important for recycling carbon and nitrogen that would otherwise be lost during photorespiration. The reduction of hydroxypyruvate to glycerate catalyzed by hydroxypyruvate reductase (HPR) in the peroxisomes is thought to be facilitated by the production of NADH by peroxisomal malate dehydrogenase (PMDH). PMDH, which is encoded by two genes in Arabidopsis ( Arabidopsis thaliana ), reduces NAD + to NADH via the oxidation of malate supplied from the cytoplasm to oxaloacetate. A double mutant lacking the expression of both PMDH genes was viable in air and had rates of photosynthesis only slightly lower than in the wild type. This is in contrast to other photorespiratory mutants, which have severely reduced rates of photosynthesis and require high CO 2 to grow. The pmdh mutant had a higher O 2 -dependent CO 2 compensation point than the wild type, implying that either Rubisco speciﬁcity had changed or that the rate of CO 2 released per Rubisco oxygenation was increased in the pmdh plants. Rates of gross O 2 evolution and uptake were similar in the pmdh and wild-type plants, indicating that chloroplast linear electron transport and photorespiratory O 2 uptake were similar between genotypes. The CO 2 postillumination burst and the rate of CO 2 released during photorespiration were both greater in the pmdh mutant compared with the wild type, suggesting that the ratio of photorespiratory CO 2 release to Rubisco oxygenation was altered in the pmdh mutant. Without PMDH in the peroxisome, the CO 2 released per Rubisco oxygenation reaction can be increased by over 50%. In summary, PMDH is essential for maintaining optimal rates of photorespiration in air; however, in its absence, signiﬁcant rates of photorespiration are still possible, indicating that there are additional mechanisms for supplying reductant to the peroxisomal HPR reaction or that the HPR reaction is altogether circumvented.

The oxygenation of ribulose-1,5-bisphosphate by Rubisco initiates the photorespiratory metabolic pathway (Badger, 1985). This reaction competes with photosynthetic carbon assimilation, producing CO 2 and NH 3 while consuming ATP and reducing equivalents, thus minimizing the efficiency of net CO 2 assimilation (Tolbert, 1997;Wingler et al., 2000;Reumann and Weber, 2006). Rates of photorespiration are about 20% of the rate of net CO 2 assimilation in C 3 plants under current ambient CO 2 concentrations at 25°C and thus consume a large portion of the solar energy absorbed by a leaf (Ort and Baker, 2002). The oxygenation reaction yields 3-phosphoglycerate (3-PGA) and 2-phosphoglycolate (2-PG). The 3-PGA is assimilated into the Calvin-Benson cycle, whereas 2-PG cannot enter this cycle directly, nor can it be converted to carbohydrates. To minimize carbon loss and to prevent the depletion of intermediates of the Calvin-Benson cycle, three-quarters of the carbon in 2-PG is recycled by the photorespiratory pathway into 3-PGA, typically releasing one CO 2 per two 2-PG molecules.
The recycling of carbon by the photorespiratory pathway involves at least 11 enzymatic steps that occur in three leaf organelles (chloroplast, mitochondria, and peroxisomes; Reumann and Weber, 2006). The peroxisomes oxidize glycolate, produced in the chloroplast, to glyoxylate by transferring electrons to O 2 , thus consuming O 2 and forming hydrogen peroxide (H 2 O 2 ; Reumann, 2000;Mano and Nishimura, 2005). The glyoxylate is further metabolized by two aminotransferases in the peroxisomes, producing Gly, which is subsequently processed by the mitochondrion, releasing CO 2 and NH 3 while producing NADH and Ser. It is generally accepted that the Ser returns to the leaf peroxisomes, where the amino group is transferred to glyoxylate, forming Gly and hydroxypyruvate (Reumann and Weber, 2006). The hydroxypyruvate is thought to be reduced to glycerate by hydroxypyruvate reductase (HPR), with NADH being provided by the peroxisomal malate dehydrogenase (PMDH; Reumann and Weber, 2006). The glycerate is phosphorylated within the chloroplast and enters into the Calvin-Benson cycle as 3-PGA.
Mutants of the photorespiratory cycle have provided much to the understanding of this biochemical pathway (Somerville and Ogren, 1982;Leegood et al., 1995;Somerville, 2001;Reumann and Weber, 2006). These mutations generally show chlorotic phenotypes when grown under ambient CO 2 concentrations but are not affected under growth conditions that suppress photorespiration, such as elevated CO 2 (Somerville and Ogren, 1982;Reumann and Weber, 2006). The observed phenotypes have been assumed to be due to the depletion of photosynthetic carbon and nitrogen cycle intermediates and perhaps to the accumulation of toxic photorespiratory intermediates. However, recently, it has become apparent that photorespiratory mutants exposed to air show severe photodamage to PSII due to an inhibition of protein repair processes, linked perhaps to increased oxidative damage to stromal protein synthesis machinery (Takahashi et al., 2007). The original mutant screens searching for chlorosis under ambient CO 2 concentrations primarily identified conditional lethal mutations (plants not viable under long-term exposure to ambient CO 2 concentrations), and more subtle mutations in the photorespiratory pathway were not easily resolved.
However, some subtle mutations in photorespiration do exist; an example of this is peroxisomal HPR, in which a mutant in barley (Hordeum vulgare) was selected by screening plants that grew poorly in air. The HPR mutant showed reduced levels of the NADHdependent peroxisomal HPR activity, but rates of photosynthesis were only reduced by 25% compared with wild-type plants when measured under ambient CO 2 concentrations (Murray et al., 1989). Related mutants lacking PMDH have also been reported in Arabidopsis (Arabidopsis thaliana) as growing slower in air, but surviving (Pracharoenwattana et al., 2007). It was proposed that an alternative pathway may metabolize hydroxypyruvate during photorespiration in these mutants, such as a cytosolic HPR or being linked to simultaneous b-oxidation in the peroxisome (Murray et al., 1989;Kleczkowski et al., 1990;Pracharoenwattana et al., 2007). This alternative pathway may provide an overflow metabolic route of hydroxypyruvate reduction if the commonly recognized reaction becomes limiting when rates of photorespiration are high (Murray et al., 1989).
Perturbing the flow of carbon through an alternative photorespiratory pathway has important implications for CO 2 efflux following the Rubisco oxygenation reaction. Evidence strongly indicates that the photorespiratory pathway releases 1 mol of CO 2 for every 2 mol of 2-PG entering the cycle during the conversion of Gly to Ser (Farquhar et al., 1980). However, disruption of the photorespiratory pathway may lead to the accumulation of intermediate compounds, allowing for alternative enzymatic and spontaneous reactions to occur. For example, in barley Gly decarboxylase (GDC) mutants lacking the ability to decarboxylate Gly, the levels of glyoxylate and formate under photorespiratory conditions increased in comparison with those in wild-type plants (Wingler et al., 1999). The production of formate can come from glyoxylate reacting with H 2 O 2 nonenzymatically, and the formate can be further incorporated into Ser via the C1tetrahydrofolate synthase/Ser hydroxymethyl transferase pathway, partially compensating for the lack of GDC activity in these plants (Wingler et al., 1999).
The role of PMDH and the HPR reaction are unclear in photorespiration. In Arabidopsis, two PMDH genes have been identified and inactivated either individually or together (Pracharoenwattana et al., 2007). While the PMDH1 gene is preferentially expressed in seedlings and the PMDH2 gene is preferentially expressed in leaves, they do appear to be partially redundant (Pracharoenwattana et al., 2007). Mutants containing T-DNA insertions in both PMDH genes demonstrate that this enzyme plays an important role in NADH oxidation during fatty acid b-oxidation in germinating seeds (Pracharoenwattana et al., 2007). However, as these plants appeared to grow well in air, this study cast doubt on the role of PMDH during photorespiration and challenged the notion that PMDH provides the NADH required for the reduction of hydroxypyruvate by HPR (Reumann and Weber, 2006). Here, we reanalyzed the Arabidopsis single and double pmdh mutants and discovered that these mutations cause subtle changes in the rates of photosynthesis under photorespiratory conditions (ambient CO 2 and O 2 concentrations) and significantly influence the flow of carbon during photorespiration, leading to increased CO 2 released relative to the rates of Rubisco oxygenation reactions.

CO 2 Assimilation Rate, Growth Analysis, and Leaf Characteristics
Rates of net CO 2 assimilation measured at approximately air CO 2 concentrations (373 mbar) in both the wild-type and pmdh1 plants were significantly higher than in the pmdh2 and double pmdh1pmdh2 lines (Table  I). However, when rates of net CO 2 assimilation were measured at elevated CO 2 (1,515 mbar), there was no significant difference between the lines (Table I). Stomatal conductance was similar in all genotypes except for the pmdh2 lines, which had lower stomatal con-ductance values then the other lines under both measurement conditions ( Table I). The CO 2 compensation point (G), inferred from the initial slope of a CO 2 response curve at lowest CO 2 concentrations, was significantly higher in the pmdh1pmdh2 plants compared with the other three genotypes (Table I). These data were consistent with variation in compensation points measured for leaf discs in a mass spectrometer closed cuvette compensation point measurement system (data not shown).
The pmdh1pmdh2 lines grew more slowly in air (as would be expected), as indicated by plant size, and growth analyses showed that growth rate on a leaf area basis was 30% to 40% less than in the wild type (Table II). However, under elevated atmospheric CO 2 concentrations (2.8 mbar), the growth rates of the wildtype and pmdh1pmdh2 lines were not significantly different (Table II). The growth analysis measurements were conducted during the first 31 d after germination, before the plants were too big for the imaging system. Additionally, there were no significant differences in Rubisco content (mmol site m 22 ), total leaf nitrogen per area (mmol m 22 ), and leaf mass per area (g m 22 ) in the wild-type and pmdh1pmdh2 lines grown under elevated atmospheric CO 2 (Table II).

The G in Response to Oxygen Concentrations
The G, as determined from measurements made with the closed leaf chamber coupled to the mass spectrometer ( Fig. 1), had a positive relationship with increased oxygen concentration, as predicted from Equation 3 (Fig. 2). Values of G were similar in the wild-type and pmdh1pmdh2 lines at low oxygen (10%), when the rates of photorespiration are expected to be low (see below and Fig. 2); however, G was greater in the pmdh1pmdh2 plants compared with the wild-type plants when measured at the higher oxygen levels (Fig. 2).
O 2 Exchange and Rubisco Reactions at the G The rates of 16 O 2 evolution, 18 O 2 uptake, and net CO 2 exchange were also measured with the leaf discs in a closed system attached to the mass spectrometer ( Fig. 1). The rates of CO 2 exchange and 18 O 2 uptake in the dark were similar between the wild-type and pmdh1pmdh2 plants in response to changing oxygen levels (data not shown). Additionally, measurements of 16 O 2 evolution and 18 O 2 uptake at the compensation point, where net CO 2 exchange is zero, were similar in Table I. Leaf gas-exchange measurements The rates of net CO 2 assimilation and stomatal conductance in air (373 mbar) and elevated (1,515 mbar) CO 2 concentrations and the G in wild-type plants, peroxisomal NAD + -malate dehydrogenase 1 and 2 plants (pmdh1 and pmdh2, respectively), and double pmdh1pmdh2 T-DNA mutants of Arabidopsis. Gasexchange measurements were made at a pO 2 of 200 mbar, irradiance of 500 mmol quanta m 22 s 21 , and leaf temperature of 25°C. Values of G were inferred from the initial slope of a CO 2 response curve at lowest CO 2 concentrations. Shown are means 6 SE of measurements made on three to five leaves from separate plants. Statistical analysis was conducted using ANOVA, and different letters indicate significant differences between plants at P , 0.05.

Variable
Wild Type pmdh1 pmdh2 pmdh1pmdh2 Net CO 2 assimilation rate (mmol CO 2 m 22 s 21 ) Air 10.6 6 1.1 a 9.9 6 1.3 a 8.2 6 1.3 a,b 6.9 6 0.4 b Elevated CO 2 17.5 6 1.4 a 17.2 6 2.6 a 14.5 6 2.9 a 15.2 6 1.  Table II. Growth analysis and leaf characteristics The growth rate under ambient CO 2 and elevated atmospheric CO 2 (2.8 mbar) in wild-type and pmdh1pmdh2 plants. Rubisco content, total leaf nitrogen per area, and leaf mass per area in both genotypes were determined in plants grown under elevated atmospheric CO 2 only. Values shown are means 6 SE of measurements made from three to five leaves from separate plants. Statistical analysis was conducted using Student's t test, and different letters indicate significant differences between plants at P , 0.05.

Variable
Wild Type pmdh1pmdh2 Growth rate in air CO 2 (cm 2 cm 22

Postillumination Burst and Photorespiratory Release of CO 2
The postillumination burst (PIB), determined by turning off the light and monitoring the rapid efflux of 12 CO 2 in the dark after an initial 10-min period in the light, was significantly higher in the pmdh1pmdh2 plants compared with the wild-type plants at 20% and higher oxygen levels (Fig. 4A). The values of PIB measured at low oxygen (10%) were similar in the wild-type and pmdh1pmdh2 plants (Fig. 4A). The photorespiratory release of CO 2 was determined by injection of CO 2 enriched in 13 C into the leaf chamber to saturate the rates of the Rubisco carboxylation reaction in the light and subsequently monitoring the rates of photorespiratory 12 CO 2 release. The release was greater in the pmdh1pmdh2 plants compared with the wild-type plants at 20% and higher oxygen levels but not at 10% oxygen (Fig. 4B). The ratio of PIB to the rate of v o as well as the ratio of photorespiratory CO 2 release to v o were higher in the pmdh1pmdh2 plants compared with the wild-type plants at 20% and 40% oxygen but not at 10% oxygen (Fig. 4, C and D).

Metabolite Analysis
Leaf discs were harvested after 10 min at the G and extracts subjected to gas chromatography-mass spectrometry (GC-MS) analysis. Relative amounts of numerous primary metabolites that were identified reliably were compared (Supplemental Table S1). Several photorespiratory intermediates were detected by this method, but differences in relative amounts between pmdh1pmdh2 and the wild type were no more than 1.6-fold. Glycolate, glyoxylate, Glu, and glycerate were unchanged in amount, while Gln and Asp were marginally lower and malate, Gly, and Ser were marginally higher. The method employed did not detect hydroxypyruvate or oxaloacetate.

The Role of PMDH
It is apparent that PMDH activity is not essential for photosynthesis in air in Arabidopsis. However, it is necessary for optimal rates of photosynthesis and growth in air (Tables I and II). The explanation for how the photorespiratory pathway functions without , and CO 2 concentration (dashed line) of a leaf disc in a closed chamber coupled to a mass spectrometer. A leaf disc was placed in the chamber and flushed with nitrogen gas, and a volume of pure oxygen highly enriched in 18 O was injected to obtain an O 2 concentration of 10%, 20%, or 40%. After a 10-min dark period, the leaf disc was illuminated (300 mmol quanta m 22 s 21 ) for 10 min, then darkened for 5 min and again illuminated for 10 min. Black arrows indicate lights off, and white arrows indicate lights on. The PIB was determined after the initial 10-min light period, and the parameters relating to the G were determined at the end of the final light period. CO 2 enriched in 13 C was injected into the leaf chamber to saturate photosynthesis in the light, and the rates of photorespiratory CO 2 release were estimated from the rate of 12 CO 2 released during this time period. Figure 2. The G at different O 2 partial pressures in the wild type and the pmdh1pmdh2 double mutant. Measurements were made in a closed leaf chamber coupled to a mass spectrometer as described for Figure 1. The lines represent the theoretical relationship between G and O 2 partial pressure: where , a is the ratio of CO 2 released per oxygenation reaction (0.5 for the solid line and 0.8 for the dotted line), O is the O 2 partial pressure, K c = 258 mbar and K o = 171 mbar are the Michaelis-Menten constants of Rubisco carboxylation and oxygenation, V omax /V cmax = 0.255, and R d /V cmax = 0.01 (Ruuska et al., 2000). Data are means of measurements of three to five different plants, and error bars represent SE.
PMDH activity in air may lie in two areas: first, the potential sources of reductant for the HPR reaction; second, the ability of hydroxypyruvate to be processed by alternative reactions either in the peroxisome or the cytosol.
It is widely accepted that the PMDH provides reductant in the form of NADH to facilitate the reduction of hydroxypyruvate by HPR (Reumann and Weber, 2006). The peroxisomal membrane is assumed to be impermeable to NADH (Donaldson, 1982;Vanroermund et al., 1995;Reumann and Weber, 2006), and the NADH required within the peroxisomes must be synthesized in situ. Traditionally, the malate/ oxaloacetate shuttle across the peroxisomal membrane is believed to supply malate derived from either the chloroplast or the mitochondrion to facilitate the reduction of NAD + to NADH via PMDH. Additionally, b-oxidation of fatty acids results in the reduction of NAD + to NADH within the peroxisomes (Pracharoenwattana et al., 2007). However, under ambient air conditions, the NADH produced by b-oxidation would presumably be insufficient to meet the requirement of photorespiratory NADH consumption in a typical C 3 plant. Here, we show that of the two genes known to encode PMDHs, it appears that PMDH2 has a more dominant role in photorespiration (Table I), as predicted by RNA coexpression analysis (Reumann and Weber, 2006). However, in the double mutant pmdh1pmdh2 lacking the expression of both PMDH genes, the rate of net CO 2 assimilation was still 70% of the wild-type rate under ambient CO 2 and O 2 concentrations during the measurement conditions.
Previous studies of a barley HPR mutant (Murray et al., 1989) also found that although the peroxisomal HPR step was apparently blocked, photosynthesis at air CO 2 concentrations was only reduced by 20% to 30%. This study concluded that an alternative NADPH-dependent HPR pathway based in the cyto-sol may metabolize a portion of the hydroxypyruvate (Fig. 5). However, preliminary studies indicate that extracts prepared from leaves of an Arabidopsis hpr knockout mutant have negligible rates of HP reduction with either NADH or NADPH (I. Pracharoenwattana, unpublished data). In addition to this possibility, the oxidation of hydroxypyruvate to glycolate by nonenzymatic reaction with H 2 O 2 (see below and Fig. 5) may also play a role.

Growth and Leaf Analysis
As in other photorespiratory mutants, the growth rates at high CO 2 were similar between the wild-type and pmdh1pmdh2 plants (Table II). Additionally, the amount of Rubisco, the leaf mass per area, and the total leaf nitrogen were similar in wild-type and pmdh1pmdh2 plants grown under elevated CO 2 (Table II). These similarities between wild-type and pmdh1pmdh2 plants under high CO 2 growth conditions are important because they indicate that the differences in photosynthetic and photorespiratory parameters in air between the genotypes are based on metabolic differences during the measurement conditions and are not due to prior growth conditions. This is particularly important with regard to Rubisco content, as will be further discussed below.

G and the Response to Oxygen Availability
The G, in the presence of mitochondrial respiration, increased with oxygen concentration in both wild-type and pmdh1pmdh2 plants (Fig. 2). However, G was higher in the pmdh1pmdh2 plants compared with the wild-type plants at ambient and higher oxygen concentrations, where the rates of photorespiration would be high (Table I; Fig. 2). There are several factors that can potentially influence G, as shown in Equation 3 in Figure 2. For example, G is influenced by the Rubisco Michaelis-Menten constants for CO 2 and O 2 , K c and K o , respectively, and the maximal rates of Rubisco carboxylation (V cmax ) and oxygenation (V omax ) reactions, as shown in Equation 3. These parameters determine the CO 2 /O 2 specificity of Rubisco (S c/o ), the ratio of the carboxylase to oxygenase rate at equal CO 2 and O 2 concentrations, as described by S c/o = (V cmax /K c )(K o /V omax ) . These kinetic values, and subsequently S c/o , vary between species but are generally considered constant within an individual organism even when grown under different environmental conditions (Jordan and Ogren, 1981). The ratio of respiration (R d ) in the light to V cmax (R d /V cmax ) can also influence values of G, as demonstrated in tobacco (Nicotiana tabacum) plants, which had reduced amounts of Rubisco but similar rates of R d (Ruuska et al., 2000). Here, the wild-type and pmdh1pmdh2 plants had similar rates of respiration in the dark (data not shown) and similar amounts of total Rubisco (Table II). Additionally, the rates of v o and v c were similar at 20% and lower oxygen concentrations, as estimated from the leaf O 2 exchange at the G (Fig. 3, C and D). Assuming that respiration in the dark is similar to R d , these data indicate that the increased values of G in pmdh1pmdh2 at 20% O 2 were not related to changes in S c/o or alterations in the ratio R d /V cmax .  In addition to the above parameters, values of G are also influenced by G*, the G in the absence of respiration (Fig. 2, Eq. 3). Values of G* are estimated as G* = a(V omax K c O)/(V cmax K o ), where a is the ratio of CO 2 released per oxygenation reaction and O is the oxygen concentration. As described in the following section and in Figure 2, changes in a can have a significant effect on G, especially when rates of photorespiration are high.

Photorespiratory CO 2 Release per Oxygenation Reaction
During photorespiration, 1 mol of 2-PG is assumed to release 0.5 mol of CO 2 during the conversion of Gly to Ser; therefore, the value of a = 0.5 (Farquhar et al., 1980;von Caemmerer and Farquhar, 1981). As shown in Figure 2, changes in a can have a large effect on the predicted G values. For example, at 20% O 2 , changing a from 0.5 to 0.8 increases the predicted value of G from 38 to 62 mbar, and the effect is greater at higher O 2 concentrations (Fig. 2). The predicted G values in Figure 2 when a = 0.5 (solid line) describes the response of G to increasing O 2 concentration in the wildtype plants but underestimates G in the pmdh1pmdh2 plants. However, when changing a to 0.8 in Equation 3 (Figure 2, dashed line), the theoretical estimates of G in response to O 2 fit the measured values of G in the pmdh1pmdh2 plants (Fig. 2).
The rates of O 2 uptake and evolution, measured at the G with the mass spectrometer, were similar in wildtype and pmdh1pmdh2 plants at 20% and lower O 2 concentrations (Fig. 3, A and B). At 40% O 2 , at which photorespiration would be higher, the rates of O 2 uptake and evolution were significantly lower in the pmdh1pmdh2 plants compared with the wild-type plants (Fig. 3, A and B). This is likely because at 40% O 2 , the unidentified alternative photorespiratory pathway utilized by the pmdh1pmdh2 plants is insufficient to keep up with the recycling of photorespiratory intermediates of the Calvin-Benson cycle, and Rubisco activity either becomes ribulose-1,5-bisphosphate limited or inhibited by the increased levels of photorespiratory intermediates (Wingler et al., 2000). The lower rates of v o and v c in the pmdh1pmdh2 plants compared with wild-type plants at 40% O 2 further support the idea that Rubisco activity becomes inhibited at the highest O 2 concentration in the pmdh1pmdh2 plants (Fig. 3, C and D).
We used two techniques to investigate the relationship of the ratio of photorespiratory CO 2 released per oxygenation reaction. In the first method, we monitored the PIB of CO 2 in the dark from the leaf tissue after measuring O 2 and CO 2 exchange at the G for 10 min (Fig. 1). These experiments were conducted on the mass spectrometer, where net CO 2 exchange, O 2 evolution, O 2 uptake, and O 2 concentrations were simultaneously determined (Figs. 1 and 4). The PIB represents the CO 2 released by the photorespiratory cycle after the oxygenation and carboxylation reactions have ceased after turning off the light. Figure 4A shows that the PIB was higher in the pmdh1pmdh2 plants compared with the wild-type plants at the higher O 2 concentrations, indicating that there is a greater release of photorespiratory CO 2 per glycolate molecule in the pmdh1pmdh2 plants compared with the wild-type plants. The photorespiratory pathway will continue to function after the light has been turned off until the pools of photorespiratory intermediates produced in the light period have been consumed (Laisk and Edwards, 1997;Hoefnagel et al., 1998;Atkin et al., 2000). Therefore, the peak rate of CO 2 released during the PIB is an indicator of the rate of CO 2 produced during photorespiration prior to darkening. In the traditional view of photorespiration, the CO 2 released is primarily produced from the conversion of Gly to Ser by GDC within the mitochondria, but it is not a specific measure of a. However, in the pmdh1pmdh2 plants and other photorespiratory mutants, such as the GDC barley mutant, the PIB will also measure other decarboxylating reactions associated with the altered photorespiratory pathways.
Additionally, the export of Ser and/or Gly via the vascular system (Madore and Grodzinski, 1984;Wingler et al., 2000) or the involvement of Ser/Gly in one-carbon metabolism and protein synthesis (Hanson and Roje, 2001) would alter the flow of carbon through the photorespiratory cycle. If large quantities of Gly were lost from the photorespiratory cycle, due to alternative metabolism or export to vascular tissue, this would decrease a and thus G, since the amount of CO 2 released per Rubisco oxygenation reaction would approach zero. Alternatively, the catabolism of Gly through C1 metabolism could potentially increase a if the stoichiometry was greater than 0.5 CO 2 released per Gly molecule. However, Gly does not appear to be used directly for C1 metabolism but is first converted to Ser through GDC and Ser hydroxymethyltransferase reactions (Mouillon et al., 1999). Alternatively, if large quantities of Ser were diverted out of the photorespiratory cycle, there would be no effect on a, as the release of CO 2 due to the conversion of Gly to Ser would be conserved. Although the potential for Ser and/or Gly loss from the photorespiratory cycle cannot be excluded, it does not fit with the increase in G seen in the pmdh1pmdh2 mutant.
The second method of estimating the release of CO 2 during photorespiration was to take advantage of the mass spectrometer's ability to measure CO 2 isotopologues, 12 CO 2 and 13 CO 2 , simultaneously. After the rates of O 2 and 12 CO 2 exchange were measured at the G for 10 min (Fig. 1), a volume of 13 CO 2 was injected into the leaf chamber to saturate the rates of v c and eliminate the v o reaction. By monitoring the release of 12 CO 2 directly after the injection of 13 CO 2 , the rates of CO 2 released by photorespiration prior to the 13 CO 2 injection can be estimated in the light without darkening. The rates of CO 2 released during photorespiration determined from this method were also greater in the pmdh1pmdh2 plants compared with the wild-type plants (Fig. 4B).
Both the PIB and 13 CO 2 techniques show that the rates of photorespiratory release of CO 2 were greater in the pmdh1pmdh2 plants compared with the wildtype plants (Fig. 4, A and B). This indicates that either the pmdh1pmdh2 plants had greater rates of photorespiration or that the amount of CO 2 released per oxygenation reaction was greater compared with that in the wild-type plants. At 20% oxygen, the measured rates of v o were similar in both genotypes (Fig. 3C), and the ratio of photorespiratory CO 2 release per Rubisco oxygenation reaction in the pmdh1pmdh2 plants was higher compared with the wild-type plants using both methods of determining photorespiratory CO 2 (Fig. 4, C and D). This supports the predicted higher a values in the pmdh1pmdh2 plants compared with the wild-type plants, as presented in Figure 2.

Photorespiratory Metabolites
Overall, the metabolite analysis did not provide any evidence for major perturbation of photorespiratory metabolism (Supplemental Table S1). The small increases in Gly and Ser are consistent with a block downstream at the PMDH step, while decreases in Gln and Asp are consistent with less efficient ammonia reassimilation. It is intriguing 2-ketoglutaric acid and isocitrate increased, by 3.6-and 7.5-fold, respectively, in the pmdh1pmdh2 plants compared with the wildtype plants (Supplemental Table S1). This may suggest that the altered consumption of reductant (NADH) in the peroxisomes due to the lack of PMDH caused the NADH/NAD ratio in the mitochondria to increase. An increase in mitochondrial NADH/NAD has been demonstrated to inhibit the NAD-dependent isocitrate dehydrogenase (Igamberdiev and Gardestrom, 2003) and may explain the relatively large increase in isocitrate in the pmdh1pmdh2 plants. The increase in 2-oxoglutarate could potentially be attributed to an alteration in the Glu requirement for recycling nitrogen through the GS/GOGAT-catalyzed reactions. However, a full evaluation of the metabolic basis for the greater CO 2 release in the absence of PMDH will require more detailed analysis in the future.
How Is the Stoichiometry of CO 2 Release Altered?
The data reported here with the pmdh1pmdh2 mutant clearly show that some restrictions to the photorespiratory pathway can lead to changes in the pathway of carbon and photorespiratory CO 2 release. Indeed, the results here represent the most dramatic demonstration of this reported in the literature. Studies on various photorespiratory mutants have not consistently agreed on the way that potential CO 2 release steps may be altered. Studies with a barley mutant deficient in Ser:glyoxylate aminotransferase demonstrated significantly elevated release of labeled CO 2 from exogenously supplied glyoxylate (Murray et al., 1987). In contrast, a HPR barley mutant (Murray et al., 1987) showed little evidence for an increased CO 2 release, even though there was considerable photosynthesis in air. A barley mutant deficient in catalase (Kendall et al., 1983) showed no evidence for increased CO 2 release, while an Arabidopsis transgenic plant with antisense reduction in the levels of the 2-oxoglutarate/ malate translocator (Schneidereit et al., 2006) showed a significant increase in G.
Based on previous studies, the most obvious explanation for an increased CO 2 release stoichiometry would be an alternative oxidative decarboxylation of accumulating pools of glyxoylate and hydroxypyruvate through their nonenzymatic interactions with H 2 O 2 to produce formate and glycolate, respectively, plus CO 2 (Halliwel, 1974;Halliwel and Butt, 1974;Grodzinski, 1979;Walton and Butt, 1981;Wingler et al., 1999). In addition, the resulting formate can also be decarboxylated in the peroxisomes by its oxidation via catalase and H 2 O 2 (Halliwel and Butt, 1974) or can be oxidized to CO 2 by formate dehydrogenase in the mitochondrion (Hourton-Cabassa et al., 1998). These potential reactions are shown in Figure 5. The potential for these alternative decarboxylations to occur would increase under conditions in which both glyoxylate and hydroxypyruvate accumulate, as would be expected for the pmdh1pmdh2 mutant described here. Only mutations inactivating PMDH, HPR, or glycerate processing steps would be expected to elevate hydroxypyruvate. Most of the previously studied photorespiratory mutants are blocked in earlier steps, and it is possible that the large increased in CO 2 release stoichiometry observed here may be due to the greater potential for hydroxypyruvate decarboxylation and glycolate recycling in the peroxisome (Walton and Butt, 1981).

CONCLUSION
The peroxisomal NAD + -malate dehydrogenase 2 (pmdh2) appears to play a more prominent role than the peroxisomal NAD + -malate dehydrogenase 1 (pmdh1) during photorespiration. Rates of photosynthesis in air were reduced in the double pmdh1pmdh2 mutant, but by only 30%. The rates of gross O 2 evolution and uptake were similar in pmdh1pmdh2 and wild-type plants, indicating that chloroplast linear electron transport and photorespiratory O 2 uptake by Rubisco were similar between genotypes. The CO 2 PIB and the rate of CO 2 released during photorespiration were both greater in the pmdh1pmdh2 mutant compared with the wild-type plants, suggesting that the ratio of photorespiratory CO 2 release to Rubisco oxygenation was altered in the mutant. In summary, PMDH is essential for maintaining optimal rates of photorespiration in air; however, in its absence, significant rates of photorespiration are still possible, indicating that there are additional mechanisms for supplying reductant to the peroxisomal HPR reaction or that the HPR reaction is altogether circumvented.

Growth Conditions
Arabidopsis (Arabidopsis thaliana) seeds were surface sterilized, stratified, and allowed to imbibe on wet filter paper. Seeds were germinated on halfstrength Murashige and Skoog medium with 1% (w/v) Suc and then transferred to potting soil (Debco seedling raising mix). For the gas-exchange measurements and the high-CO 2 growth analysis, plants were grown under 2.8 mbar of CO 2 in a controlled-environment growth cabinet at an irradiance of 200 mmol quanta m 22 s 21 at plant height, and air temperature of 25°C during the day and 15°C at night, with a daylength of 14 h. Wild-type and pmdh1pmdh2 plants used for the growth analysis under ambient CO 2 concentrations were grown during the southern hemisphere's autumn month of May in a glasshouse under natural light conditions (25°C day and 15°C night temperatures). For all plants, 0.9-L pots were used, and the potting soil was mixed with 2.4 to 4 g Osmocote/L soil (15:4.8:10.8:1.2 nitrogen:phosphorus: potassium:magnesium + trace elements: boron, copper, iron, manganese, molybdenum, and zinc; Scotts Australia) and watered daily.

Measurements of Net CO 2 Assimilation Rates
Individual plants were brought from the growth chambers into the laboratory, and a single leaf was dark adapted for 30 min in the LI6400 (LiCor Biosciences) leaf chamber (LI6400-40). Gas-exchange measurements were made at pO 2 of 200 mbar, irradiance of 500 mmol quanta m 22 s 21 , and leaf temperature of 25°C. The G measurements shown in Table I were inferred from the initial slope of a CO 2 response curve at lowest CO 2 concentrations. All other values of G were measured on the mass spectrometer as described below.

Mass Spectrometric Measurements
The exchange of O 2 and CO 2 was measured from leaf discs placed into a closed leaf chamber attached to a mass spectrometer (Isoprime Micromass) as described previously (Maxwell et al., 1998;Ruuska et al., 2000). After the leaf disc was placed in the chamber, the chamber was flushed with nitrogen gas and a volume of pure oxygen enriched in 18 O was injected to obtain an O 2 concentration of 10%, 20%, or 40%. After a 10-min dark period, the leaf disc was illuminated (300 mmol quanta m 22 s 21 ) for 10 min, then darkened for 5 min, and again illuminated for 10 min (Fig. 1). The PIB was determined after the initial 10-min light period, and the parameters relating to the G were determined at the end of the final light period. CO 2 enriched in 13 C ( 13 CO 2 ; 99% enrichment) was injected into the leaf chamber to saturate the rates of the Rubisco carboxylation reaction in the light. The PIB and the photorespiratory CO 2 release in the light were estimated from the rate of 12 CO 2 released. The rate of CO 2 release was calculated from the 12 CO 2 concentration signal by measuring the slope of a moving five-point window (10-s width) over the data during this time period, and the peak values for rates were determined. To correct for the contamination of 12 CO 2 during the injection of 13 CO 2 , several injections were conducted with an empty leaf chamber throughout each day that measurements were conducted. The background 12 CO 2 increase during the blank injection was subtracted from the data collected during the measurement with the leaf disc enclosed in the chamber.

Growth Analysis
Growth analyses were done on a leaf area expansion basis by nondestructively measuring the displayed leaf area of individual plants at time intervals. Total leaf area (including petioles) was measured by imaging the chlorophyll fluorescence (CF Imager; Technologica) and using the leaf area function available for total fluorescent pixels of each plant. The area was calibrated against a known image standard. Leaf area expansion rates are calculated from the slope of ln(area) versus time in units of cm 2 cm 22 d 21 , and means and SE values are reported for four to five separate plants imaged over a 7-to 10-d time course during the exponential growth phase of each seedling. Imaging commenced at approximately 22 d after sowing.

Metabolite Analysis
Plant tissues (10-30 mg) were snap frozen and pulverized in liquid nitrogen. The enzymes in the tissue were inactivated by heating at 70°C for 15 min in 500 mL of extraction solution (methanol:water = 20:3, with adipic acid as internal standard). Cells debris was removed by centrifugation. For each sample, 60 mL of supernatant was transferred to a new tube and dried using a SpeedVac. The residue was dissolved in 20 mL of pyridine with methoxylamine hydrochloride (20 mg mL 21 ) and incubated at 30°C for 90 min with shaking. Metabolites were converted to tert-butyldimethylsilyl derivatives by reacting at 70°C for 30 min with 30 mL of N-methyl-N-[tert-butyldimethylsilyl]trifluoroacetimide plus 1% tert-butyldimetheylchlorosilane. The metabolite derivatives were analyzed by GC-MS (Agilent 6890 gas chromatograph coupled with a 5973N mass selective detector). GC-MS data were processed by AMDIS. Metabolite derivatives were identified by comparison of the retention time and mass spectra of authentic standards, and the amount of metabolites was calculated by total ion current signal of each metabolite peak normalized to the internal standard and tissue weight.

Calculations of v o and v c from MS Measurements
The calculations used were described previously (Ruuska et al., 2000). For each oxygenation reaction of Rubisco, 1 mol of O 2 is consumed, and a further 0.5 mol of O 2 is consumed during photorespiration by glycolate oxidase in the peroxisomes. We assume that O 2 uptake by PSI via the Mehler reaction is negligible and that O 2 uptake by mitochondria was the same in the light and dark. Therefore, v o was calculated as: If the whole chain electron transport results in NADPH production for photosynthesis and photorespiration, then v c can be calculated as: Statistical Analysis ANOVA was conducted using Student's t test in Statistica (version 6.0; StatSoft). Tukey's LSD test was used for post hoc comparisons.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Table S1. Relative amounts of numerous primary metabolites that were identified reliably with GC-MS. Leaf material was harvested after 10 min at the G.