The Response of Carbon Metabolism and Antioxidant Defenses of Alfalfa Nodules to Drought Stress and to the Subsequent Recovery of Plants 1,2

proteins, and the decrease of N 2 ase proteins and N 2 ase–linked respiration, reveal that both impairment of bacteroid function and oxidative stress take place in alfalfa nodules before any detectable effect on SS expression. Further studies are required to determine if

There was also upregulation (mRNA level) of cytosolic ascorbate peroxidase and downregulation of SS, homoglutathione synthetase, and bacterial catalase A.
Drought stress did not affect nifH mRNA level or leghemoglobin (Lb) expression, but decreased MoFe-and Fe-proteins. Rewatering of plants led to a partial recovery of the activity (75%) and proteins (>64%) of N 2 ase, a complete recovery of Suc, and a decrease of malate (-48%) relative to control. The increase in O 2 diffusion resistance, the decrease in N 2 ase-linked respiration and N 2 ase proteins, the accumulation of respiratory substrates and oxidized lipids and proteins, and the upregulation of antioxidant genes reveal that bacteroids have their respiratory activity impaired and that oxidative stress occurs in nodules under drought conditions prior to any detectable effect on SS or Lb. We conclude that a limitation in metabolic capacity of bacteroids and oxidative damage of cellular components are contributing factors to the inhibition of N 2 ase activity in alfalfa nodules.
Drought is a major factor limiting crop production and has a particularly negative impact on symbiotic N 2 fixation (Sprent, 1972;Zahran, 1999). However, the causes for the drought-induced inhibition of nitrogenase (N 2 ase) activity are still uncertain.
Studies on soybean (Glycine max), common bean (Phaseolus vulgaris), and pea (Pisum sativum) have shown that the inhibitory effects may be mediated by a decrease in nodule O 2 permeability (Durand et al., 1987;Serraj and Sinclair, 1996;Ramos et al., 1999) and metabolic activity (Díaz del Castillo and Layzell, 1995). The finding that the activity of sucrose synthase (SS), but not other carbon and nitrogen metabolism enzymes, rapidly declined in nodules upon imposition of stress provided strong evidence for a major role of SS in the inhibition of N 2 fixation (González et al., 1995;Gordon et al., 1997;Ramos et al., 1999). A detailed biochemical analysis of the rug4 mutant of pea, which displays severely reduced SS activity, further demonstrated an essential role of SS in symbiosis (Gordon et al., 1999). These, and subsequent results (Gálvez et al., 2005;Marino et al., 2006), led to the conclusion that SS is a critical regulatory enzyme in nodule carbon metabolism and in the early response of N 2 fixation to drought. The authors suggested that the inhibition of SS activity restricts the availability of malate and other dicarboxylic acids for bacteroid respiration (Fig. 1A) and that this is responsible for the inhibition of N 2 ase activity (Gordon et al., 1997;Gálvez et al., 2005).
Another mechanism that could play a role in the drought-induced inhibition of N 2 fixation, but has received much less attention, is oxidative stress. In plant cells, this occurs when the generation of reactive oxygen species (ROS) overwhelms the antioxidant defenses (Fig. 1B). Indeed, drought induces oxidative stress in pea nodules (Gogorcena et al., 1995) and other types of abiotic stress also lead to general decreases of antioxidant activities which are associated with nodule senescence (Swaraj et al., rooting medium of pea plants inhibits SS activity of nodules prior to N 2 ase activity, suggesting that SS is an early target of oxidative stress in nodules (Marino et al., 2006).
However, in all of these studies the response of nodule antioxidants was not analyzed at the molecular level and in most of them N 2 ase activity was not monitored, making it difficult to establish a relationship between the decrease of antioxidant protection and the loss of nodule function.
Alfalfa is a perennial forage legume of great agronomical interest which produces indeterminate nodules upon infection of roots with Sinorhizobium meliloti. Previous agronomical, physiological, and biochemical studies strongly suggest that alfalfa is more drought tolerant than pea (Moran et al., 1994) and ureide-producing grain legumes (Sinclair and Serraj, 1995). For example, photosynthesis was inhibited by 77% in pea leaves having a water potential (Ψ w ) of -1.3 MPa (Moran et al., 1994) but only decreased by 28% in alfalfa leaves at a Ψ w of -1.8 MPa (Rubio et al., 2002). This inhibition was accompanied by consistent decreases in antioxidant activities and soluble protein in pea leaves, but by either minor or no changes in alfalfa leaves. On the basis of this relatively high drought tolerance, we have selected alfalfa as plant material to test the hypotheses that the SS activity and the antioxidant defenses of nodules are involved in the drought-induced inhibition of N 2 ase activity. and expressed relative to values of control plants ( Fig. 2A). Following the same criteria for significant gene upregulation (ratio >2) or downregulation (ratio <0.5) in the qRT-PCR analysis as those generally used in cDNA array studies (El Yahyaoui et al., 2004), our results show that the nifH mRNA level in alfalfa nodule bacteroids did not appreciably change under moderate or severe stress but decreased after the recovery period. Immunoblots revealed a significant decrease in the contents of the two proteins of N 2 ase (MoFe-protein and Fe-protein), especially in nodules from severely stressed plants, and a partial (MoFe-protein) or complete (Fe-protein) recovery in nodules from reirrigated plants (Fig. 2B). The apparent N 2 ase activity was measured as H 2 evolution in intact alfalfa plants cv Aragón using an open flowthrough system to minimize plant disturbance (Minchin et al., 1986). In alfalfa cv Aragón, this activity consistently decreased upon application of moderate (-43%) and severe (-82%) drought stress, and recovered up to 75% of the control values after rewatering of plants (Fig. 2C).
An open-flow through system was also used to measure simultaneously H 2 and CO 2 evolution, and hence to determine N 2 ase activity and some related parameters in control and drought stressed plants of alfalfa cv N4 (Table I). A moderate drought stress induced a sharp decline (-81%) in total N 2 ase activity, which was accompanied by less pronounced decreases in total root respiration (-54%) and in N 2 ase-linked respiration (-66%). The discrepancy between the extent of inhibition of N 2 ase activity and its associated respiration can be explained by major increases in the carbon cost of N 2 ase (+126%) and in the O 2 diffusion resistance of nodules (+158%).

Effect of Drought on Organic Acids, Sugars, and Associated Enzymes
The effects of drought on the major dicarboxylic acids (succinate, α−ketoglutarate [αKG], and malate) and sugars (Suc) of nodules were examined ( Fig. 3A), as they are used by the host cells and bacteroids for the production of the energy and reducing power required for N 2 fixation and other metabolic reactions (Temple et al., 1998). The application of a moderate drought caused an accumulation of Suc (+58%), no changes in malate, and a decrease (-23%) in αKG, relative to values of control nodules. Intensification of stress had no further negative effect on the αKG content of nodules, but led to a consistent accumulation of succinate (+68%), malate (+42%), and Suc (+435%). Rewatering of plants caused the return of succinate and Suc to control values, whereas it sharply decreased (-48%) the nodule contents of αKG and malate.
Drought also affected the activities of some carbon and nitrogen metabolism enzymes in nodules (Fig. 3B), albeit the effects were less intense than those observed for carbon metabolites (Fig. 3A). In fact, the activities of SS, isocitrate dehydrogenase (ICDH), alkaline invertase (AI), and glutamate synthase (GOGAT) in nodules of moderately stressed plants did not significantly differ from those of control plants. Application of severe drought stress only led to a minor decrease (-12%) in ICDH activity but caused a significant decline (-30%) in SS, AI, and GOGAT activities. Upon rewatering of plants, SS activity returned to control values, ICDH and AI activities remained higher than the control (+20%), and GOGAT activity did not recover (Fig. 3B).

Effect of Drought on Antioxidant Enzymes of Nodule Host Cells
The high specificity and sensitivity afforded by qRT-PCR was central for the quantification of mRNAs encoding various isoforms of antioxidant enzymes, such as the superoxide dismutases (SODs), from the nodule host cells (Fig. 4A). Genespecific primers (Supplemental Table S1) were designed for RT-PCR amplification of the cDNAs coding for CuZnSODc, CuZnSODp, plastidic FeSOD, and mitochondrial MnSOD. According to the criteria mentioned above for changes in gene expression to be considered significant, there was upregulation of CuZnSODc and FeSOD under moderate drought and of CuZnSODp during recovery, but no changes in the expression of MnSOD for any of the treatments (Fig. 4A).
The expression of other antioxidant enzymes (Fig. 4A), as well as of other important proteins (Fig. 4B which was approximately 0.12 mg mg -1 protein for all treatments, and on the activities of CuZnSODc, CuZnSODp, MnSOD, FeSOD, APXc, GR, and catalase (data not shown). However, the protein level of the CuZnSODc isoform increased slightly in nodules of plants exposed to a moderate or severe (+25%) drought stress and more markedly in nodules of plants recovered from drought (+148%).

Effect of Drought on Antioxidant Enzymes of Bacteroids
Free-living and symbiotic S. meliloti contain two SOD and three catalase isoforms. The MnSOD (sodA) is located in the cytosol (Santos et al., 1999), whereas the CuZnSOD (sodC) is located in the periplasm (Ampe et al., 2003). Catalase A (katA) is mainly expressed in the bacteroids, catalase C (katC) in the bacteria within the infection threads, and catalase B (katB) in both the bacteria and bacteroids (Jamet et al., 2003). Using specific primers (Supplemental Table S1), we examined the effect of drought stress on the expression of the five genes in the bacteroids (Fig.   5). Application of a moderate water deficit to plants led to a modest, yet significant, increase in the sodA mRNA level, which returned to control values under severe stress and consistently decreased in the recovery treatment. In contrast, no changes were observed in sodC expression for any treatment. Interestingly, the kat genes were differentially expressed with drought stress, although the changes were modest in many cases (Fig. 5). Whereas katB and katC were upregulated in nodules of moderately stressed plants, the mRNA level of katA markedly decreased during drought stress and remained low upon rewatering of plants. Also, in nodules of severely stressed plants, the katB gene remained upregulated, but the mRNA level of katC did not differ significantly from the control.

Effect of Drought on Antioxidant Metabolites
The major redox metabolites of nodules were quantified using enzymatic methods and high-performance capillary electrophoresis (CE). This enabled us to determine not only the total amounts of ascorbate and thiol tripeptides, but also the proportions of their oxidized forms (Fig. 6). The content of total ascorbate (ascorbate + dehydroascorbate) in control alfalfa nodules was 5.9 µmol g -1 dw (0.9 µmol g -1 fresh weight), which is in the range of that reported for other legume nodules (Dalton et al., 1986;Gogorcena et al., 1995;1997). Ascorbate decreased with moderate (-30%) and severe (-58%) drought stress and partially recovered (79% of control) upon rewatering of plants, whereas dehydroascorbate did not significantly vary with any of the treatments (Fig. 6). Dehydroascorbate accounted for approximately 19% of total ascorbate in nodules of control and moderatelystressed plants and increased to 33% during severe stress and recovery, proportions below the range of 40%-80% observed by Groten et al. (2006) and Marino et al. (2007) in pea nodules of a comparable developmental stage. In this repect, we should note that, in our hands, the proportion of dehydroascorbate increased up to 50%-60% if alfalfa or common bean nodules were not harvested directly into liquid nitrogen but left instead to stand on ice for less than one hour (J. Loscos, M.A. Matamoros, and M. Becana, unpublished data).
Mature alfalfa nodules have been reported to contain approximately two thirds of GSH (γGlu-Cys-Gly) and one third of hGSH (γGlu-Cys-βAla), a structurally related homolog that is present exclusively in some legume species and tissues (Matamoros et al., 2003). This was confirmed by our determinations in control nodules of 4 µmol g -1 dw (0.7 µmol g -1 fresh weight) of GSH and 2.1 µmol g -1 dw (0.3 µmol g -1 fresh weight) of hGSH ( Fig. 6). Moderate or severe drought stress had no effect on the contents of GSH or hGSH, whereas, during the recovery of plants, GSH significantly decreased (-30%) but hGSH remained constant. As a result, the proportion of GSH declined from 67% in control plants to 52% in droughtrecovered plants, and the proportion of hGSH increased concomitantly from 33% to 48% (Fig. 6). The oxidized forms of the two thiol tripeptides remained fairly constant and at a low level (<4%) during drought and subsequent recovery.

Effect of Drought on Lipid Peroxidation and Protein Oxidation
The accumulation of oxidatively damaged lipids and proteins is a marker of oxidative stress in plant and animal tissues. These products arise, among other mechanisms, by the oxidative attack of lipids and proteins by ROS (Stadtman, 1992;Halliwell and Gutteridge, 1999) and can be conveniently detected and quantified by HPLC and immunological methods. To determine if alfalfa nodules were experiencing oxidative stress under drought conditions, malondialdehyde (MDA), a product of membrane lipid peroxidation, was quantified by reaction with thiobarbituric acid (TBA) and subsequent HPLC analysis of the corresponding colored adduct (Fig. 7A). Moderate and severe drought stress caused similar consistent increases (60%-73%) in the nodule content of MDA, which remained at that high level upon rewatering of plants. The pattern of oxidized proteins in nodules was also examined in gels upon derivatization of the protein carbonyl groups with 2,4-dinitrophenylhydrazine (DNP) using a commercial antibody raised against the hydrazone derivatives (Fig. 7B). Such immunoblots revealed that, relative to controls, the levels of oxidatively modified proteins were higher in nodule extracts from drought-stressed plants and similar in nodule extracts from plants recovered from drought. Taken together, these results provide conclusive evidence that drought induces oxidative stress in alfalfa nodules, a situation previously reported for pea nodules (Moran et al., 1994) and leaves (Iturbe-Ormaetxe et al., 1998).

DISCUSSION
The results of this study with alfalfa reveal that the decline of N 2 ase activity with drought does not involve an inhibition of SS activity but a metabolic limitation of bacteroids. This may be caused by alterations in the O 2 availability and respiratory capacity of bacteroids and by the oxidative damage of nodule cell components. Two major differences in the response of C metabolism to drought were found with respect to previous results with other legumes. First, in alfalfa nodules, a moderate drought stress inhibited N 2 ase activity by 43% ( Fig. 2) but had no effect on SS mRNA (Fig. 4B) and activity (Fig. 3), whereas concomitant decreases in SS and N 2 ase activities were observed in common bean (Ramos et al., 1999), pea (Gálvez et al., 2005), and soybean (González et al., 1995). Second, in alfalfa nodules subjected to severe drought stress, when N 2 ase was inhibited by 82%, the concentrations of malate and succinate increased by 40% and 68%, respectively (Fig. 3), whereas they declined progressively with drought in pea nodules (Gálvez et al., 2005). Consequently, in alfalfa nodules under stress conditions, SS activity was sufficient to sustain the production of both organic acids from Suc. The accumulation of Suc under drought stress was also reported for other legume nodules and was mainly attributed to the inhibition of SS activity (González et al., 1995;Ramos et al., 1999;Gálvez et al., 2005). Our results, especially with moderately stressed nodules, are in sharp contrast with this explanation and suggest that the accumulation of Suc in alfalfa nodules is caused by a still active import of Suc from the shoot, together with a limitation of Suc consumption in the nodules due to impairement of respiratory activity. Furthermore, the accumulation of dicarboxylic acids and Suc with drought stress (Fig. 3) indicates that nodule metabolism, and in particular N 2 ase activity and respiration, is not limited by the provision of reduced carbon.
The existence of metabolic limitations underpinning the drought-induced decline of N 2 ase activity was initially proposed on the basis that it was only partially restored by raising external O 2 concentration (Diaz del Castillo and Layzell, 1995;Serraj and Sinclair, 1996). A potential limiting factor is Lb degradation (Irigoyen et al., 1992;Guerin et al., 1995), but there was no effect of drought on the Lb content of alfalfa nodules, in agreement with earlier studies on pea (González et al., 1998) and soybean (Gordon et al., 1999). Another metabolic constraint for nodule activity under drought stress conditions could be a decrease in the respìratory capacity of the bacteroids (Díaz del Castillo and Layzell, 1995). This conclusion is supported by the accumulation of malate and succinate (Fig. 3), the main respiratory substrates of bacteroids (Lodwig and Poole, 2003), and by the reduction of N 2 ase-linked respiration (Table I). In accordance with this hypothesis, the increase in resistance to O 2 diffusion would be in response to an increase in O 2 concentration in the infected zone, due to reduced bacteroid respiration. The decline of organic acids in nodules of drought-recovered plants would then be explained by a reactivation of both N 2 ase activity and respiration in the bacteroids. Although our results cannot discern whether an alteration of bacteroid respiration is a cause or an effect of the inhibition of N 2 ase activity, the molecular analysis of N 2 ase expression provides some insight (Fig. 2). Moderate and severe drought stress did not significantly affect the mRNA levels of nifH, but did decrease, to different extents, the contents of MoFe-protein and Fe-protein in bacteroids. These observations suggest that the loss of N 2 ase activity is, at least in part, caused by protein degradation, in addition to other probable factors such as a decrease in ATP and reducing power. Interestingly, rewatering of plants decreased the nifH mRNA level but allowed the partial (MoFeprotein) or total (Fe-protein) recovery of the N 2 ase proteins, which suggests increased translation of the nifH transcript or a higher stability of the Fe-protein in the drought-recovered plants. In any case, N 2 ase activity was not completely restored, which suggests that nodule metabolism remains affected even after 2 d of plant rewatering. This metabolic limitation is evident by the significantly lower content of MoFe-protein in bacteroids and by the declines in malate content and GOGAT activity in nodules of drought-recovered plants (Fig. 3). In fact, GOGAT, which is a primary enzyme for nitrogen assimilation in alfalfa nodules (Temple et al., 1998), may be limited by a lower availability of αKG. This enzyme activity is associated with nodule development and N 2 fixation (Groat and Vance, 1981) and appears to be particularly sensitive to drought stress (Ramos et al., 1999).
Another potential, and frequently overlooked, metabolic constraint of droughtstressed nodules is the oxidative damage of nodule components. Our results reveal that this occurs in alfalfa nodules. The upregulation during drought of a number of genes involved in antioxidant protection, namely, CuZnSODc, FeSOD,APXc,GRc,sodA,katB,and katC (Figs. 4A and 5), indicate that both the host cells and bacteroids are probably experiencing oxidative stress. This was confirmed by the accumulation of peroxidized lipids (estimated as MDA) and oxidatively modified proteins (estimated as carbonyl groups) in nodules during drought stress and subsequent recovery (Fig. 7). At least two factors may contribute to ensuing oxidative stress in alfalfa nodules. First, although the activities of APX and associated enzymes, as assayed in vitro, remained fairly constant with drought, a large decrease in the concentrations of ascorbate (Fig. 6) and probably of NAD(P)H (Gogorcena et al., 1995) may compromise H 2 O 2 detoxification through the ascorbate-GSH cycle in vivo. Second, nonenzymatic formation of cytotoxic aldehydes and protein carbonyl derivatives (Fig. 7) requires the generation of highly oxidizing ROS, such as the hydroxyl radical, which in turn depend on trace amounts of metal ions (Stadtman, 1992;Halliwell and Gutteridge, 1999). Indeed, catalytic iron (Gogorcena et al. 1995;Evans et al., 1999) and hydroxyl radical production (Becana and Klucas, 1992) have been found to increase in senescing nodules.
In summary, we conclude that a decrease in SS expression (mRNA and activity) is not the cause of the drought-induced loss of N 2 ase activity in alfalfa.
Interestingly, a similar response of nodule metabolism to drought was found in M. truncatula (R. Ladrera, E.M. González, and C. Arrese-Igor, unpublished data). This raises the possibility that the decrease in SS activity observed in other legumes is not directly responsible for the inhibition of N 2 ase, but rather that the negative effects of drought on both activities are concomitant and mediated through another factor, possibly oxidative damage. Alternatively, it is possible that, depending on legume species, two models exist concerning the role of carbon metabolism in the inhibition of N 2 ase activity by drought. In soybean, pea, and common bean, the inhibition would be mediated by SS, leading to a decrease in the organic acid levels of nodules. In alfalfa and M. truncatula, the inhibition would not involve SS activity and organic acids would accumulate in nodules. In any case, the accumulation of respiratory substrates, lipid peroxides, and oxidized proteins, and the decrease of N 2 ase proteins and N 2 ase-linked respiration, reveal that both impairment of bacteroid function and oxidative stress take place in alfalfa nodules before any detectable effect on SS expression. Further studies are required to determine if oxidative stress and effects on N 2 ase protein content occur prior to, and therefore may be responsible for, the observed reduction in bacteroid respiration.  , 1997). Plants of alfalfa cv Aragón were grown for 50 to 55 d and were then separated at random into four groups. One of them (control) was kept under optimal water conditions and two other groups were subjected to drought stress by withholding irrigation until the plants reached a leaf Ψ w (mean ± SE) of -1.3 ± 0.1 MPa (moderate drought; 5-7 d) and -2.1 ± 0.2 MPa (severe drought; 8-10 d). The fourth group of plants was also subjected to severe drought and then allowed to recover by re-irrigation for 2 d (recovery). Control and recovery plants had a similar leaf Ψ w of -0.6 ± 0.2 MPa.

Biological Material and Plant Treatments
Leaf Ψ w was measured 1 h after the beginning of the photoperiod in representative leaves, situated in the upper third of the shoot, with a pressure bomb (Soil Moisture Equipment, Santa Barbara, CA). Nodule Ψ w was measured in the same plants as leaf Ψ w using C52 sample chambers coupled to a HR-33T microvoltmeter (Wescor, Logan, UT). Values of nodule Ψ w (mean ± SE) were -0.8 ± 0.1, -1.5 ± 0.1, -2.5 ± 0.1, and -0.7 ± 0.1 MPa for the control, moderate stress, severe stress, and recovery treatments, respectively. Plants of alfalfa cv N4 were subjected only to a moderate drought stress (Ψ w of -1.5 ± 0.1 MPa) and used to assess the respiration and carbon cost associated with N 2 ase activity. Nodules were harvested into liquid nitrogen and stored at -80°C, except for samples to be used for dw determination, which were dried for 48 h at 80°C.

N 2 ase Activity and Root Respiration
Total N 2 ase activity and nodulated root respiration were measured on intact plants using a flow-through gas system that incorporated H 2 detectors (Witty and Minchin, 1998). Root systems of alfalfa cv N4 were sealed in their growth pots, allowed to stabilise for 18h in a stream of air enriched with 500 µl CO 2 L -1 , and then exposed to a gas stream of 79% (v/v) Ar and 21% (v/v) O 2 . Respiratory CO 2 production was measured using an IR gas analyzer and N 2 ase activity was measured as H 2 production using a electrochemical hydrogen sensor (City Technology Ltd., Portsmouth, UK).
After steady-state conditions had been reached following exposure to Ar/O 2 (within 65 min), the O 2 concentration in the gas stream was increased over the range 21 to 50% (8.55 to 20.45 mmol O 2 L -1 ). N 2 ase-linked respiration was calculated from the linear relationship between changes in total root respiration and H 2 production during the O 2 stepping period (Witty et al., 1983).
For alfalfa cv Aragón, the apparent N 2 ase activity was measured in intact plants, as indicated above, by measuring H 2 evolution with an electrochemical sensor (Qubit Systems, Canada) in an open flow-through gas system under a stream of 79% (v/v) N 2 and 21% (v/v) O 2 (Witty and Minchin, 1998). The H 2 sensor was calibrated with high purity gases flowing at the same rate as the sampling system (500 mL min -1 ).
However, the GR activity assayed corresponded to the sum of the cytosolic and plastidic isoforms.
Soluble protein was quantified by the dye-binding microassay (Bio-Rad) using bovine serum albumin as the standard. Lb concentration was determined by the pyridine-hemochrome method, using an extinction coefficient (556 nm minus 539 nm) of 23.4 mM -1 cm -1 for the difference spectrum between the reduced (+dithionite) and the oxidized (+ferricyanide) hemochromes (Appleby and Bergersen, 1980).

Carbon and Antioxidant Metabolites
Organic acids and sugars were extracted from nodules in 5% (w/v) trichloroacetic acid, and samples were processed as described by Wilson and Harris (1966) with minor modifications (Gálvez et al., 2005). Succinate, malate, and αKG were quantified by ion chromatography in a DX-500 system (Dionex) by gradient separation with an IonPac AS11 column (Dionex) according to the method recommended by the supplier. Suc was analyzed by CE in a Coulter PACE system 5500 (Beckman) coupled to a diode array detector (Marino et al., 2006).
Ascorbate and dehydroascorbate were measured as described by Bartoli et al. (2000) with some modifications (Matamoros et al., 2006) using nodules harvested directly into liquid nitrogen to avoid artifactual oxidation of ascorbate. Extracts were made with 1 M HClO 4 , cleared by centrifugation, neutralized with 1 M K 2 CO 3 to pH 5.6, and centrifuged again. Two aliquots were made and one of them was treated with 0.4 mM dithioerythritol for 15 min at room temperature. The aliquots were incubated with 0.05 units of ascorbate oxidase (Sigma) and the decrease in absorbance at 265 nm was monitored until stable and used to calculate ascorbate concentration based on an extinction coefficient of 14.3 mM -1 cm -1 . Ascorbate was quantified by direct analysis of the aliquots not treated with dithioerythritol, and dehydroascorbate as the difference in ascorbate concentration between the treated and untreated aliquots.
Thiol compounds were extracted from nodules with 2% (w/v) metaphosphoric acid and 1 mM EDTA. The extracts were cleared by centrifugation and treated with 65 mM dithiothreitol for 15 min at room temperature. The concentrations of the thiol tripeptides (reduced plus oxidized forms) were measured by CE (Marino et al., 2006) using GSH (Sigma) and hGSH (Bachem, Weil am Rhein, Germany) for calibration.
The proportion of oxidized thiols was determined in the same extracts, prior to dithiothreitol treatment, by an enzymatic recycling procedure using yeast GR to reduce the disulfide forms and vinylpyridine as thiol blocking agent (Law et al., 1983).

Oxidative Damage of Lipids and Proteins
The nodule content of MDA, a major product formed from decomposition of lipid peroxides, was quantified by HPLC essentially as described elsewhere (Iturbe-Ormaetxe et al., 1998). The (TBA) 2 -MDA adduct was resolved on a C 18 column (4.6 x 250 mm, 5 µm; Baker), eluted at 1 mL min -1 with 5 mM potassium phosphate (pH 7.0) containing 15% acetonitrile and 0.6% tetrahydrofurane, and detected at 532 nm.
Calibration curves were made with 1,1,3,3-tetraethoxypropane (Sigma) as the standard, which is stoichiometrically converted into MDA during the acid-heating step of the TBA reaction. The purity of the peak corresponding to the (TBA) 2 -MDA adduct was routinely monitored by scanning between 400 and 600 nm with a photodiode-array detector. The peaks in the samples and standard showed identical spectra, with a maximum of absorbance at 532 nm and a shoulder at 495 nm (Iturbe-Ormaetxe et al., 1998).
The oxidative damage of proteins was measured by derivatization of carbonyl groups with DNP and subsequent separation of proteins on SDS-gels, using the OxyBlot protein oxidation kit (Chemicon, Temecula, CA) according to the supplier's recommendations. The DNP-hydrazone derivatives of proteins were detected on membranes using rabbit specific anti-DNP as the primary antibody (1:150 dilution), a peroxidase conjugate of goat anti-rabbit IgG as the secondary antibody (1:3000 dilution), and the SuperSignal West Pico chemiluminescence kit (Pierce).
Densitometric semi-quantitative analysis was performed with the Quantity One software (Bio-Rad).

Transcript Levels
Total RNA was isolated from nodules using the RNAqueous kit (Ambion), treated with DNaseI (Roche) at 37ºC for 30 min, and reverse-transcribed using Moloney murine leukemia virus reverse transcriptase (Promega). qRT-PCR analysis was carried out with the iCycler iQ System (Bio-Rad) using iQ SYBR-Green Supermix reagents (Bio-Rad) and specific primers for genes expressed in the nodule host cells and bacteroids (Supplemental Table S1). The PCR program consisted of an initial denaturation and Taq activation step of 5 min at 95ºC, followed by 50 cycles of 15 s at 95ºC and 1 min at 60ºC. A melting curve analysis was performed after every PCR reaction to confirm the accuracy of each amplified product. All reactions were set up in duplicate. The mRNA levels of alfalfa genes were normalized against EF1-α (El Yahyaoui et al., 2004) and those of S. meliloti genes against smc00324 and smc02641 (Becker et al., 2004). Values of treated plants were expressed relative to those of control plants using the 2 -∆∆Ct method (Livak and Shmittgen, 2001). The absence of contamination with genomic DNA was tested by qRT-PCR in all RNA samples, after the DNase treatment but prior to reverse transcription, using the primers of the housekeeping genes.

Statistical Analysis
Each measurement of N 2 ase activity and associated parameters was made with a pool of 3 to 5 intact plants. Each biochemical assay was made with nodules harvested from plants growing in different pots. Data were obtained from two (N 2 ase activity, carbon metabolism, and antioxidant enzymes and metabolites) or four to six (mRNAs of antioxidant enzymes) series of plants that were grown independently. The total numbers of replicates are stated in each figure. For each treatment, the measurements from the different series of plants were pooled for statistical analysis. For each parameter, the mean values of the drought and recovery treatments were compared with the controls by ANOVA and the Dunnett's t test.

Supplemental Data
The following material is available in the online version of this article. Table S1. Primers used for qRT-PCR.