Ca2+-binding and Ca2+-independent Respiratory NADH and NADPH Dehydrogenases of Arabidopsis thaliana*

Type II NAD(P)H:quinone oxidoreductases are single polypeptide proteins widespread in the living world. They bypass the first site of respiratory energy conservation, constituted by the type I NADH dehydrogenases. To investigate substrate specificities and Ca2+ binding properties of seven predicted type II NAD(P)H dehydrogenases of Arabidopsis thaliana we have produced them as T7-tagged fusion proteins in Escherichia coli. The NDB1 and NDB2 enzymes were found to bind Ca2+, and a single amino acid substitution in the EF hand motif of NDB1 abolished the Ca2+ binding. NDB2 and NDB4 functionally complemented an E. coli mutant deficient in endogenous type I and type II NADH dehydrogenases. This demonstrates that these two plant enzymes can substitute for the NADH dehydrogenases in the bacterial respiratory chain. Three NDB-type enzymes displayed distinct catalytic profiles with substrate specificities and Ca2+ stimulation being considerably affected by changes in pH and substrate concentrations. Under physiologically relevant conditions, the NDB1 fusion protein acted as a Ca2+-dependent NADPH dehydrogenase. NDB2 and NDB4 fusion proteins were NADH-specific, and NDB2 was stimulated by Ca2+. The observed activity profiles of the NDB-type enzymes provide a fundament for understanding the mitochondrial system for direct oxidation of cytosolic NAD(P)H in plants. Our findings also suggest different modes of regulation and metabolic roles for the analyzed A. thaliana enzymes.

The type II NAD(P)H DHs usually possess one non-covalently bound FAD, except for in hyperthermophilic Archaea, where FAD is replaced by FMN (2). The peptide sequence of most enzymes contains two well conserved motifs for dinucleotide binding (1). In many organisms the presence of several type II DH isoenzymes increases the catalytic flexibility of respiratory NAD(P)H oxidation. In bacteria and yeast, diverse roles have been implicated for different homologs. For example, a catalytic function in redox balancing was suggested for homologs in yeast (1,2,6), whereas a role in redox sensing was proposed for the homologs in the cyanobacterium Synechocystis sp (7). In plants, the relative expression of gene homologs varies between tissues (8,9), during development (10), in response to light (8,10,11) and upon several kinds of stress (12)(13)(14). The differential gene expression of plant type II NAD(P)H DH homologs points to diverse physiological roles of the enzymes.
In mitochondria of plants and fungi, type II NAD(P)H DHs are attached to the inner and outer surface of the inner membrane (2,3). External NADH and NADPH oxidation measured in isolated plant mitochondria is generally dependent on Ca 2ϩ (3,15) with NADH oxidation being less sensitive to inhibition by chelators (16,17). NADH oxidation has even been observed in the absence of Ca 2ϩ for several plant materials (14,18,19). There is strong evidence that there are separate DHs, each relatively specific for external NADH and NADPH oxidation in plants (20 -22). However, the absence of specific inhibitors has made it difficult to study the isoenzymes individually in isolated mitochondria. Several proteins showing NAD(P)H oxidation in the presence of artificial electron acceptors have been purified from different plant species (23)(24)(25)(26). However, the requirement for artificial quinones, which can affect substrate specificity and Ca 2ϩ dependence (25,27), has complicated the catalytic characterization of the purified enzymes.
Based on sequence homology to type II NAD(P)H DHs in yeast and E. coli, two genes, nda1 and ndb1, were described in potato, and their gene products were localized to the internal and external side of the inner mitochondrial membrane, respectively (40). Homologs of type II NAD(P)H DHs with high amino acid sequence similarity to the potato NDA1 and NDB1 proteins are also present in rice and Arabidopsis thaliana (8). The seven homologs found in A. thaliana group into three families. These are nda1-2 and ndb1-4, all of which are closely related to fungal homologs, and ndc1, which groups together with cyanobacterial homologs upon phylogenetic analysis (8). The N termini of homologs of all three families target green fluorescent protein to mitochondria (8). Intramitochondrial localization studies suggest that NDB1, NDB2, and NDB4 are external enzymes, whereas NDA-and NDC-type proteins are internally located (9). The plant NDB homologs cluster together with NDE1 of N. crassa, and all of these proteins contain an insertion with more or less degenerate EF hand motifs for Ca 2ϩ binding (8,37). However, Ca 2ϩ binding by the enzymes has not been shown experimentally.
For plants, a substrate has been identified only for the NDB1 homolog of potato, which is an external Ca 2ϩ -dependent NADPH DH, as shown by overproduction in tobacco plants (21). Substantial correlative evidence in potato and A. thaliana indicates that NDA1 is a matrix-facing NADH DH (9,10,13,41). Defining the substrate and Ca 2ϩ specificities of the A. thaliana homologs is essential for interpretation of gene expression profiles and for elucidating the physiological roles of these enzymes.
In this study we have analyzed the Ca 2ϩ binding properties of the A. thaliana type II NAD(P)H DHs and characterized three of the enzymes in terms of substrate specificities and Ca 2ϩ stimulation. NDB1, NDB2, and NDB4 were found to be principally able to oxidize both NADH and NADPH. However, the enzymes showed high substrate specificity at physiologically relevant substrate concentrations and pH.

EXPERIMENTAL PROCEDURES
Plasmid Construction-As templates for gene amplification using PCR, full-length cDNA clones were provided by the Arabidopsis Biological Resource Center (U51324, encoding Atnda1; U12861, encoding Atndb2) and the German Resource Center for Genome Research (U12390, clone MPIZp768J207Q2, encoding Atndb4). For Atndb3, an incomplete EST clone 62A4T7, accession number T41616 (42), was used. For Atnda2, Atndb1, and Atndc1, cDNA synthesized using RNA isolated from A. thaliana seedlings (14) was used for amplification. Attempts to clone the complete Atndb3 by reverse transcription-PCR were unsuccessful, as previously reported by another group (9).
Mutagenesis of pET-T7Atndb1-A single base pair substitution changing the codon GAC (Asp-387) to GCC (Ala) in NDB1 was introduced using PCR. A forward primer, 5Ј-ATC CTT CCT GGC TCA CTG-3Ј, and a reverse mismatch-primer, 5Ј-TCT TCC ATG GTC AAG GTT CCT GAG TTG GCC GCA TC-3Ј, containing the new codon and an NcoI recognition site were used for amplification. The products were cloned into a TOPO vector, cut out with HindIII and NcoI, and then inserted into the pET-T7Atndb1 cut with the same enzymes to replace the wild type segment. The obtained plasmid was denoted pET-T7Atndb1-D387A and confirmed by DNA sequencing of the inserted region.
Bacterial Strains and Growth Conditions-For Ca 2ϩ binding studies, the pET21 derivatives were transformed into E. coli BL21(DE3)/pLysS. Cells were collected from plates and grown in 50 ml of LB containing ampicillin (100 g/ml) and chloramphenicol (34 g/ml) at 37°C and 200 rpm orbital shaking. At an A 600 of ϳ0.5, isopropyl-␤-D-thiogalactopyranoside (IPTG) was added to a final concentration of 1 mM. After 1 h the cells were harvested by centrifugation at 2000 ϫ g for 5 min, washed in 50 mM Tris/Cl pH 8.0, resuspended in 1 ml of high salt medium I (0.5 M NaCl, 20 mM Tris/Cl, 5 mM imidazole, pH 8.0), and frozen in liquid nitrogen.
For complementation studies and in vitro NAD(P)H oxidation assays, the recombinant genes on plasmids were expressed in the E. coli strain MWC008(DE3). The ndh and nuo genes in MWC008 are defective by insertional disruption using kanamycin and tetracycline resistance markers (43). The gene for T7 RNA polymerase was introduced into the MWC008 chromosome using the DE3 lysogenization kit (Novagen). In the presence of 0.1-1 mM IPTG negative effects on growth were observed for the MWC008(DE3) strain harboring pET21a (not shown), and therefore, IPTG was not added to the cultures. Low levels of T7 RNA polymerase are most likely produced in MWC008(DE3) also in the absence of IPTG, and this was sufficient for the transcription of genes cloned in pET21a. The wild type E. coli strain AN387 (44) was used as a control. Antibiotics were used at the following concentrations: 12 g/ml tetracycline, 50 g/ml kanamycin, and 100 g/ml ampicillin. For minimal media, M63 agar (45) was supplemented with 30 mM glucose or mannitol. For liquid cultures, colonies were resuspended from LB plates, and 0.5 or 1 liter of LB medium containing 0.5% glucose and appropriate antibiotics was inoculated to give an A 600 of ϳ0.05. The cultures were grown in 5-liter baffled flasks at 30°C and 80 -90 rpm for 16 -18 h. At an A 600 of 1.2-1.8, cells were harvested by centrifugation at 7,000 ϫ g for 30 min at 4°C and washed in 50 mM Tris/Cl, pH 8.0. Cell pellets were stored at Ϫ20°C until used for membrane isolation.
Protein Extraction-Extraction of protein for the Ca 2ϩ binding assay was done from transformed BL21(DE3)/pLysS cells frozen in high salt medium I. During thawing of the cells, protease inhibitors were added at the following final concentrations: 1 mM phenylmethylsulfonyl fluoride, 2 M L-trans-epoxysuccinyl-leucylamido(4-guanidino)butane, 1 M pepstatin, and 2 M leupeptin. The cell suspensions were sonicated for 5 ϫ 5 s and then centrifuged at 100,000 ϫ g for 30 min at 4°C. Pellets were resuspended in high salt medium I supplemented with 1% (w/v) Triton X-100 and incubated with stirring for 30 min on ice. After centrifugation as above, the pellets constituting insoluble protein fractions were resuspended in high salt medium I containing 6 M urea and incubated as above. Urea-extracted proteins were collected by taking the supernatant after a final centrifugation at 100,000 ϫ g for 30 min at 4°C.
Protein Analyses-Polypeptides were resolved in 8 or 10% SDS-PAGE gels and either stained with Coomassie Brilliant Blue R-250 or wet-transferred to nitrocellulose membranes as described (14). T7-tagged protein was immunodecorated using a T7 monoclonal antibody (Novagen) and horseradish peroxidase-conjugated anti-mouse secondary antibody and detected as described (46). T7 signal intensity was calculated using the Kodak 1D Image Analysis software. Radioactive labeling of electroblotted proteins was carried out as previously described (47) with 1.6 M 45 CaCl 2 as the probe and H 2 O as the washing solution. Radioactive label was detected with phosphor screens using a Personal Molecular Imager FX and the image-processing program Quantity One (Bio-Rad). As molecular mass markers in SDS-PAGE, the low molecular weight marker kit (Amersham Biosciences) and the PageRuler Prestained Protein Ladder (Fermentas) were used. The prestained molecular marker proteins bound 45 Ca 2ϩ , as previously observed (47).
Membrane Preparation-MWC008(DE3) cells harboring different plasmids were resuspended to 1 ⁄ 30 of the culture volume in high salt medium II (0.5 M NaCl, 20 mM Tris, 0.1 mM CaCl 2 , and 1 mM MgCl 2 , pH 8.0) supplemented with 4 mM MgSO 4 , 2 g/ml DNase I, and 0.1 mM phenylmethylsulfonyl fluoride. The suspension was passed twice through a French press cell operated at 18,000 p.s.i. Unbroken cells and debris were removed by centrifugation at 5000 ϫ g for 10 min. The supernatant was centrifuged at 200,000 ϫ g for 90 min at 4°C. The obtained pellet containing membranes was homogenized and diluted to 1 ⁄ 180 of the original culture volume in high salt medium II and centrifuged at 200,000 ϫ g for 1 h at 4°C. The pellet was finally homogenized in 1 ⁄ 450 of the original culture volume of high salt medium II, frozen in liquid nitrogen, and stored at Ϫ80°C. For high EGTA conditions, high salt medium II contained 10 mM EGTA instead of 0.1 mM CaCl 2 and additionally 4 mM MgCl 2 during washing and resuspension steps. For initial screening purpose, membranes were washed under low salt conditions in 50 mM Tris, pH 8.0, and resuspended in 20 mM Mops, 2.5 mM MgCl 2 , pH 7.2. Enzyme activities were also tested on fresh preparations to certify that freezing and thawing did not have any major effects on the activities. Membrane protein concentrations were determined using the bicinchoninic acid protein assay (Sigma-Aldrich) with bovine serum albumin as standard.
Enzyme Assays-O 2 consumption of isolated membranes was measured at 25°C using an O 2 electrode (Rank Brothers, Cambridge, UK). NAD(P)H oxidation of bacterial membranes was measured at room temperature (21°C) using an Aminco DW2a or an Olis DW2 Conversion dual wavelength spectrophotometer at 340 -400 nm. NAD(P)H and succinate oxidation was measured in medium A (20 mM Mops/KOH, 2.5 mM MgCl 2 , 0.5 mM EGTA, pH 7.2). Additions and substrate concentrations were as indicated in the figure legends. When using EGTA at 10 mM, the concentration of Mops in medium A was increased to 100 mM, and CaCl 2 was added to 10

Production of A. thaliana Type II NAD(P)H DHs as T7-tagged
Fusion Proteins in E. coli-Plasmids were constructed to contain cDNAs for seven A. thaliana type II NAD(P)H DHs and denoted pET-T7Atnda1, -T7Atnda2, -T7Atndb1, -T7Atndb2, -T7Atndb3, -T7Atndb4, and -T7Atndc1. Gene amplification was done from full-length sequences, except for the lowly expressed NDB3 (8,9), which was derived from a partial cDNA clone. It has been reported for yeast NDI1 that only the mature, but not the full-length protein, can be functionally produced with a T7 tag in E. coli (32). Also, when the potato NDA1 polypeptide is imported into mitochondria, N-terminal processing has been observed (40). Therefore, the inserts derived from the six full-length A. thaliana cDNAs were designed to lack the parts coding for unconserved N termini and to start at a position aligning to the E. coli NDH-2 start codon. All plasmids were constructed to encode fusion proteins with a T7 tag at the N-terminal end (supplemental Fig. 1).
For investigation of Ca 2ϩ binding properties, the different A. thaliana NAD(P)H DHs were produced in E. coli BL21(DE3)/ pLysS transformed with the engineered plant genes cloned in pET21a. Gene expression was induced by the addition of IPTG to the growth medium. In total cell lysates the NDB-and NDCtype proteins were visible upon SDS-PAGE analysis and Coomassie staining (not shown). All fusion proteins were visibly enriched in urea extracts of insoluble cell material (Fig. 1A). This fraction was, therefore, used for Ca 2ϩ binding analysis.
The seven fusion proteins and C-terminal-truncated products were detected by Western analysis using an antibody against the N-terminal T7 tag (Fig. 1B). After initial analyses, the gels were loaded with unequal amounts of E. coli extracts to achieve similar T7-tag Western band intensities for the different plant proteins. The standardized gels and blots are the ones shown in Fig. 1. The apparent sizes of the fusion proteins were consistent with those calculated from the cDNA sequences.
NDB1 and NDB2 Proteins Bind Ca 2ϩ -The urea extracts of E. coli containing A. thaliana fusion proteins were Westernblotted, and the nitrocellulose membranes were incubated with 45 Ca 2ϩ as probe. Signals were obtained for NDB1 and NDB2 (Fig. 1C). The signal intensities were similar to 0.5 g of calmodulin, which was used as a positive control. No signal was detected in E. coli extracts containing the other homologs, except for a faint background derived from small amounts of E. coli proteins present in the extracts. To investigate the role of the EF hand domain of NDB1 in Ca 2ϩ binding, the Asp at position 387 of the full-length sequence was replaced by an Ala. Analyses of total cellular protein from transformed cells by SDS-PAGE showed that pET-T7Atndb1-D387A and pET-T7Atndb1 in BL21(DE3)/pLysS yielded similar levels of NDB1 and NDB1-D387A proteins, respectively (Fig. 2B). The total protein fraction was also used for Ca 2ϩ -overlay analyses. Binding of 45 Ca 2ϩ was detected for NDB1, whereas the NDB1-D387A variant did not bind 45 Ca 2ϩ (Fig. 2A). The signal was similar in strength to 45 Ca 2ϩ binding by E. coli cellular proteins, which are present in the total cell protein fraction but much less in the urea extract used in Fig. 1. The results demonstrate that the EF hand domain is essential for Ca 2ϩ binding by NDB1.

Functional Complementation of an E. coli Strain Deficient in Type I and Type II NADH DHs by NDB2 and NDB4-
The functionality of the A. thaliana fusion proteins in the bacterial system was tested in vivo under conditions where the limiting factor for growth is the capacity to reoxidize NADH. E. coli strain MWC008 is deficient in both type I and type II respiratory NADH DHs. The strain is, therefore, unable to grow with mannitol as the only carbon source (43). MWC008 was equipped with the gene for T7 RNA polymerase and then transformed with the various pET21a derivatives containing genes for the different A. thaliana NAD(P)H DHs. Wild type E. coli grew to a higher cell density than MWC008(DE3) containing the different plasmids on all media (Fig. 3). All strains grew on minimal medium plus glucose. On mannitol, however, only MWC008(DE3)/pET-T7Atndb2, -T7Atndb4, and the wild type showed growth (Fig. 3). This indicates that the NDB2 and NDB4 proteins are produced and can functionally substitute for the respiratory NADH DHs lacking in the E. coli mutant.
Substrate Concentration Affects Apparent Specificity and Ca 2ϩ Dependence of NDB-type Enzymes-Lack of complementation of E. coli MWC008(DE3) for growth does not exclude in vitro activity of the A. thaliana fusion proteins. For example, the amount of enzyme protein produced might be too low for sufficient complementation, or the enzyme could be inactive in E. coli cells due to a lack of cofactors. To screen for type II NAD(P)H DH activities and to study enzyme properties, membranes were isolated from MWC008(DE3) containing the dif-ferent plasmids after growth in LB plus glucose and in the absence of IPTG. Western blot analyses of isolated membrane fractions showed single bands of expected sizes for the NDB1, NDB2, NDB4, and NDB1-D387A fusion proteins (not shown). Quantification of the T7-tag signal indicated 5-10 times higher concentrations of NDB2 and NDB4 antigens as compared with NDB1 and NDB1-D387A, which showed similar signal intensities. No antigen signal was seen for NDA1, and only very faint signals were detected for NDA2 and NDC1 (not shown). NAD(P)H oxidation with O 2 as final electron acceptor was detected in isolated membranes containing NDB1, NDB2, NDB4, and NDB1-D387A proteins, and each of them displayed distinct catalytic profiles (Fig. 4). At high nucleotide substrate concentrations (0.8 mM), NDB1 oxidized both NADPH and NADH, and the activities were highly dependent on Ca 2ϩ (Fig.  4A). The NDB1-D387A protein showed Ca 2ϩ -independent NADPH oxidation similar to NDB1 but without the Ca 2ϩ -dependent component. Virtually no NADH oxidation activity was detected for NDB1-D387A either in absence or presence of Ca 2ϩ (Fig. 4A). NDB2 showed a clear preference for NADH over NADPH. Ca 2ϩ had no significant effect on the steady-state rate with NADH but induced a low rate of NADPH oxidation. The NDB4 fusion protein was found to oxidize NADH and to a lesser extent NADPH, and both activities were unaffected by Ca 2ϩ (Fig. 4A). No NAD(P)H oxidation was detected in membranes of MWC008(DE3)/pET21a under any condition (not shown). Membrane preparations had a succinate oxidase activity similar to vector control (38 Ϯ 9 nmol O 2 min Ϫ1 mg Ϫ1 ) irrespective of plasmid expressed by the cells (not shown).
To investigate if the substrate concentration affects the specificity profiles of the NDB-type enzymes, activities in the membrane preparations were also measured using 10ϫ lower NAD(P)H concentrations (80 M). Under these conditions, NDB1 oxidized NADPH at about 5-fold higher rates than NADH and in a completely Ca 2ϩ -dependent manner (Fig. 4B). NADPH oxidation by NDB1 under low substrate conditions reached 60% of the activity measured at high substrate concentration, as calculated by converting rates of NAD(P)H oxidation into O 2 consumption using a factor of two. For the NDB1-D387A mutant most of the Ca 2ϩ -independent NADPH oxidation seen at higher substrate concentration was absent at the lower substrate level. NDB2 oxidized exclusively NADH at low nucleotide concentrations. The steady-state rate was stimulated by Ca 2ϩ with a statistically significant difference in a paired t test at p Ͻ 0.01. Also NDB4 displayed a strong specificity for NADH under low substrate concentrations (Fig. 4B). It can be concluded that the A. thaliana NDB-type enzymes bound to the E. coli membranes are highly specific to single nucleotide substrates at an NADH or NADPH concentration of 80 M.
The enzyme activity of NDB4 was found to be sensitive to the membrane isolation procedure. Washing and resuspending the membrane fraction of MWC008(DE3)/pET-T7ndb4 at low salt reduced O 2 consumption in membranes by about 80% as compared with high salt conditions (not shown). Neither NDB1 nor NDB2 was affected by the low salt conditions during preparation. Immunodetection revealed similar amounts of NDB4 protein in membrane fractions prepared under low and high salt conditions (not shown). Therefore, it is likely that the absence of ions during membrane isolation inactivated NDB4.
Effect of pH on Substrate Specificity and Ca 2ϩ Dependence of NDB-type Proteins-Oxidation of 0.8 mM NAD(P)H by membranes containing NDB-type fusion proteins was measured as O 2 consumption over a physiologically relevant pH range (6.8 -7.8) in medium B. Activities at pH 7.2 in this buffer (Fig. 5A) were similar to those observed in medium A (Fig. 4A). This excludes any significant buffer-specific effects on the enzyme activities. The NDB1 enzyme activities were considerably affected by pH. At pH 6.8 and 7.2, the maximal oxidation rates for NADH and NADPH by NDB1 were similar. The Ca 2ϩ dependence of NADPH oxidation was 80% at pH 7.2 but only 25% at pH 6.8 (Fig. 5A). A somewhat lower Ca 2ϩ dependence at pH 6.8 was also seen for NADH oxidation by NDB1. At pH 7.5, oxidation of NADPH was 3-fold higher than of NADH, thereby substantially increasing the NADPH specificity of NDB1 (Fig. 5A). At pH 7.8, the enzyme was completely NADPH-specific, albeit less active. In the NDB1-D387A mutant, the Ca 2ϩ -dependent rates seen for NDB1 were abolished, but not the Ca 2ϩ -independent rates (Fig. 5A). NADH oxidation by NDB2 and NDB4 was little affected by pH over the measured range, but a Ca 2ϩstimulation was evident for NDB2 at pH 7.5 and 7.8. The difference was significant at p Ͻ 0.05 in a paired t test of the unnormalized data. For both NDB2 and NDB4, the observed NADPH oxidation activity was negligible at pH 7.8 but increased at lower pH (Fig. 5A). For all enzymes, NADH oxidation was completely inhibited by KCN. With NADPH, residual rates of up to 60 nmol O 2 min Ϫ1 mg Ϫ1 were detected for NDB2 and NDB4 at lower pH (not shown), which indicates a low NADPH oxidase activity.
Intact plant mitochondria oxidize NADPH directly to the quinone analog DcQ with high Ca 2ϩ dependence (14,21). Isolated E. coli membranes containing the different NDB-type enzymes were used to investigate if NAD(P)H oxidation rates to DcQ are consistent with those to O 2 . Using DcQ, possible restrictions in the bacterial respiratory chain, such as insufficient terminal oxidase capacity or low quinone availability, are circumvented. The pH curves for NDB1 obtained by oxidation of 120 M NAD(P)H to DcQ (Fig. 5B) are highly similar to those seen with 0.8 mM NAD(P)H measured to O 2 (Fig. 5A). The only difference was a higher ratio of NADPH: NADH oxidation at pH 7.2 for NDB1 in the DcQ assay. Membranes containing NDB2 or NDB4 fusion protein displayed similar pH profiles for NADH oxidation with both terminal electron acceptors. However, NADPH oxidation was virtually absent with DcQ for both enzymes. The steadystate NADH to DcQ activity by NDB2 was stimulated by Ca 2ϩ at pH 7.5 and 7.8 (Fig. 5B), with a significant difference for p Ͻ 0.05 in a paired t test of the unnormalized data, confirming the results for O 2 consumption (Fig. 5A). All investigated NDB-type enzymes accepted DcQ efficiently, as the rates for the main substrates were in all cases 1.5-2 times higher than with O 2 as final electron acceptor.
Ca 2ϩ and pH Affect NDB2 Initial Catalytic Rates-Membranes containing NDB1 and NDB4 proteins generally displayed immediate linear rates of O 2 consumption after the addition of NAD(P)H. However, for NDB2, the addition of NADH in the absence of Ca 2ϩ always resulted in a lag phase before the linear oxidation rate was reached ( Table 1). The lag phase lasted several minutes at higher pH, whereas at lower pH the lag phase was shorter and less pronounced (supplemental Fig. 2). The addition of Ca 2ϩ before NADH resulted in that the maximum rate was reached from start. Also, complete oxidation of a small amount of NADH (20 M) before starting the measurement (by the addition of 0.8 mM NADH) shortened the lag time considerably (not shown). To investigate if the lag phase may be caused by mobilization of residual Ca 2ϩ bound to E. coli membranes, the sample was preincubated with 10 mM EGTA in the reaction mixture for 1 h at 4°C before the assay was performed. Also, membranes were isolated with 10 mM EGTA present in all buffers. However, a similar lag phase for NADH oxidation was observed in both cases (not shown). The results, thus, show that Ca 2ϩ has a distinct stimulating effect on NADH oxidation by NDB2, which is especially pronounced during the pre-steady state.

DISCUSSION
From the perspective of enzyme function in vivo, substrate and affector specificities must be placed in their metabolic context. External mitochondrial NAD(P)H DHs are directed toward the intermembrane space, which is metabolically and ionically connected to the cytosol via porin channels (48). The pH of the plant cytosol is estimated to be 7.2-7.5 (49) but can decrease by 0.3-0.6 units upon light changes or hypoxia (50 -53). The estimated concentration of free Ca 2ϩ in the plant cytosol is below 1 M (15, 49) but can  Fig. 4. A, O 2 consumption was measured in medium B but otherwise in the same way as for Fig. 4. B, NAD(P)H:DcQ activity was measured in Medium B in the presence of 40 M DcQ and 120 M NADH or NADPH. 1 mM CaCl 2 was present where indicated. Enzyme assays were started by the addition of bacterial membranes. Steady-state rates are presented as percent of maximum rate for each construct and experiment. In A, rates corresponding to 100% were 188 Ϯ 1, 441 Ϯ 100, 100 Ϯ 9, and 98 Ϯ 2 nmol O 2 min Ϫ1 mg Ϫ1 for NDB4, NDB2, NDB1, and NDB1-D387A, respectively. In B, the control rates were 570 Ϯ 36, 1809 Ϯ 488, 342 Ϯ 0, and 335 Ϯ 53 nmol of NADH min Ϫ1 mg Ϫ1 for NDB4, NDB2, NDB1, and NDB1-D387A, respectively. Error bars denote S.D. as for Fig. 4. increase dramatically in response to stress (54 -56). The concentrations of free and total NADH in the plant cytosol are estimated to be around 0.5 and 55 M, respectively (57,58). NADPH concentrations are in both cases estimated to be around 150 M (57,58). The catalytic parameters of the studied NDB proteins will be discussed in relation to these conditions. NDB4 was able to complement an E. coli type I and type II NADH DH-deficient strain (Fig. 3). Heterologous complementation by type II NADH DHs has previously been observed with homologs of A. tumefaciens (34) and Synechocystis sp (7). NDB4 is a Ca 2ϩ -independent NADH DH when assayed as a fusion protein in E. coli membranes using low substrate concentrations ( Fig. 4 -5). NADPH oxidation was only detected in the lower pH range at high NADPH concentrations. A similar pH profile has been determined by mutation analyses for N. crassa NDE2, which oxidizes NADH and to a lesser extent NADPH with pH optima of 7.2 and 6.2, respectively (38). At pH 7.2-7.5, both N. crassa NDE2 and A. thaliana NDB4 are, therefore, likely to act as NADH-specific DHs.
The presence of a completely Ca 2ϩ -independent homolog is consistent with observations that Ca 2ϩ dependence can vary markedly between tissues in a plant. Mitochondria from fresh sugar beet roots show a strongly Ca 2ϩ -dependent NADH oxidation, whereas the rates in mitochondria from cold-stored roots are mainly Ca 2ϩ -independent (19). The transcript levels for ndb4 in A. thaliana are usually low (8,9) but can increase up to 10-fold in response to different stress treatments (12). External Ca 2ϩ -independent NADH oxidation has been measured in mitochondria from A. thaliana seedlings but was stable upon up-regulation of ndb4 transcripts (14). The reason for this is presently unclear.
Like NDB4, NDB2 complemented the E. coli double mutant, and the activity of the enzyme was virtually specific for NADH (Figs. 4 and 5). The steady-state oxidation rates were stimulated by Ca 2ϩ , especially at low substrate concentrations or higher pH. The activity lag phase observed in the absence of Ca 2ϩ ( Table 1) further emphasizes that NDB2 is affected by Ca 2ϩ . It also confirms that lower pH decreases the Ca 2ϩ requirement of NDB2. This is consistent with a lower sensitivity to chelators observed for external NADH oxidation by plant mitochondria at lower pH (59). The complete oxidation of a small amount of NADH also shortened the lag phase, a phenomenon previously observed in Jerusalem artichoke mitochondria (60). Thus, NADH oxidation by NDB2 in the E. coli membranes displays characteristics of external NADH oxidation by plant mitochondria. The lag phase before reaching full activation may reflect a state transition phase for NDB2.
A correlation of up-regulated ndb2 transcript levels and an increase in external Ca 2ϩ -dependent NADH oxidation of mitochondria has been observed in A. thaliana (14). In the present study Ca 2ϩ affected the NDB2 activity and bound to the enzyme but not to NDB3 or NDB4. Thus, NDB2 is most likely the mitochondrial Ca 2ϩ -dependent external NADH DH, unique to plants.
NDB1 is a Ca 2ϩ -dependent NADPH DH when analyzed as a fusion protein in E. coli membranes under low substrate conditions (Fig. 4B). This is in line with a previous characterization of potato NDB1 overproduced in transgenic tobacco (21), where, however, a background of NADH oxidation could have masked a low NADH oxidation rate by NDB1. At high substrate concentrations, A. thaliana NDB1 also oxidized NADH in a fully Ca 2ϩ -dependent manner (Fig. 4). Thus, Ca 2ϩ -dependent NADH oxidation measured in purified plant mitochondria at pH 7.2 (15) could be due to both NDB2-and NDB1-type enzymes. The higher NADPH:NADH oxidation ratio for NDB1 at low substrate concentrations indicates that NADH oxidation is a low affinity component. Similar characteristics, however, reversed for the substrates are found for AtuNDH-2 of A. tumefaciens (34). Considering that cytosolic concentrations of total NADH and NADPH can be up to 55 and 150 M, respectively (58), NDB1 most likely acts as an NADPHspecific enzyme in vivo.
NDB1 transcript is present in several tissues of A. thaliana (8,9), and the levels in both potato and A. thaliana remain remarkably stable under different conditions (10 -14). NDB1 substrate specificity and Ca 2ϩ dependence were highly influenced by small pH changes around neutral (Fig. 5). This could indicate that NDB1 is regulated at activity level in vivo. NDB1 might be active only for short periods of time during signalinduced Ca 2ϩ oscillations (61) or during conditions that decrease cytosolic pH (49).
The observed substrate specificities of NDB1, NDB2, and NDB4 test previous hypotheses based on sequence similarity in the putative NAD(P)H binding region of eukaryotic and bacterial NAD(P)H DHs (1,21). In the NADH-specific type II DHs from bacteria and fungi as well as in the NADH-specific NDB2 and NDB4, the second ␤ sheet of the ␤␣␤ motif for dinucleotide binding (62) ends with a Glu which can form a hydrogen bond to the adenine ribose moiety of NAD(H) while possibly rejecting the phosphate group of NADPH. The N. crassa NDE1 and the NDB1 of potato and A. thaliana, which are all NADPHspecific (21,37) (Fig. 4), contain an uncharged Gln at this position, which could facilitate binding of the NADP(H) molecule. The presented results for three NDB-type enzymes are consistent with this hypothesis and lend further support to the importance of the terminal residues of the nucleotide binding motif (21). The effects by substrate concentration and pH, however, demonstrate that specificities are not absolute and may involve other residues or secondary structures, as previously shown for glutathione reductase (63). The Ca 2ϩ effects on enzyme activities of NDB1 and NDB2 were consistent with their ability to bind Ca 2ϩ ions in a 45 Ca 2ϩ overlay assay (Fig. 1). An insertion containing EF hands and EF hand-like motifs is present in NDE1 of N. crassa and NDB-type homologs of A. thaliana, potato, and rice (3,21,37) (Fig. 6). The N-terminal EF hand in plants and the C-terminal EF hand in N. crassa conform with the classical EF hand sequence, whereas the other EF hand-like sequences are degenerate (1,21). Six critical positions for co-ordination of the Ca 2ϩ ion have been identified in canonical EF hand sequences (64,65). An exchange of Asp for Ala in the second coordinating residue of the NDB1 EF hand abolished Ca 2ϩ binding to the protein and the Ca 2ϩ -dependent component of the NAD(P)H oxidation (Figs. 2, 4, and 5). This indicates that Ca 2ϩ binding to the EF hand domain is not essential for activity per se but shifts the pH curve in the alkaline direction, possibly by stabilizing the active state of the enzyme at higher pH. On the whole, the results suggest that only NDB1 and NDB2 possess functional EF hand domains with high affinity for Ca 2ϩ .
NDB3 deviates from NDB1 only in the second coordinating position, which is occupied by a Glu instead of an Asp (Fig. 6). Glu is present at this position in 2% of Ca 2ϩ binding EF hands (64), and thus, a low affinity binding of Ca 2ϩ to NDB3, not detectable by the 45 Ca 2ϩ overlay assay, cannot be excluded. NDB4 neither bound 45 Ca 2ϩ (Fig. 3) nor was stimulated by Ca 2ϩ in vitro (Figs. 4 and 5). Consistently, the N-terminal EF hand-like sequence contains positively charged Lys residues not found in Ca 2ϩ binding domains (64) (Fig. 6). These charges should repel a Ca 2ϩ ion and may even mimic a bound Ca 2ϩ and thereby promote a constitutively active conformation.
There are indications that potato NDA-and NDB-type proteins reside as high molecular mass forms in the inner mitochondrial membrane (66). In yeast, all three type II NADH DHs are part of a supramolecular complex with other DHs and citric acid cycle enzymes (67). It was also shown that external NADH oxidation via NDE1 and NDE2 inhibits the mitochondrial glycerol 3-phosphate DH in the same complex under high NADH concentrations (68). The data described here show that the analyzed plant NDBtype enzymes are independent of other proteins for activity. However, it remains possible that other mitochondrial proteins modulate the activities, as in yeast. This may explain the relatively small Ca 2ϩ effect on NDB2 in E. coli membranes, which partly contrasts the strong Ca 2ϩ dependence of external NADH oxidation often seen in isolated mitochondria (3,69). We report here a qualitative enzymatic characterization of the external A. thaliana type II NAD(P)H DHs NDB1, NDB2, and NDB4 produced with an N-terminal T7 tag in E. coli. The enzymes resided in isolated membranes and reduced the quinones of the bacterial respiratory chain. In previous investigations, NAD(P)H oxidation by membrane-bound DHs in plants has always been measured in materials (e.g. isolated mitochondria) containing a mix of enzyme homologs. Both Ca 2ϩ and pH were known to affect NAD(P)H oxidation by type II DHs with species-specific variations (3,70), but the responses of individual homologs to these parameters have not been studied. Our results on the individual enzymes clearly demonstrate that plant mitochondria contain at least three separate external DHs specific for NADH and NADPH and with different Ca 2ϩ dependences.