l-Galactono-1,4-lactone Dehydrogenase Is Required for the Accumulation of Plant Respiratory Complex I*

Mitochondrial NADH-ubiquinone oxidoreductase (complex I) is the largest enzyme of the oxidative phosphorylation system, with subunits located at the matrix and membrane domains. In plants, holocomplex I is composed of more than 40 subunits, 9 of which are encoded by the mitochondrial genome (NAD subunits). In Nicotiana sylvestris, a minor 800-kDa subcomplex containing subunits of both domains and displaying NADH dehydrogenase activity is detectable. The NMS1 mutant lacking the membrane arm NAD4 subunit and the CMSII mutant lacking the peripheral NAD7 subunit are both devoid of the holoenzyme. In contrast to CMSII, the 800-kDa subcomplex is present in NMS1 mitochondria, indicating that it could represent an assembly intermediate lacking the distal part of the membrane arm. l-galactono-1,4-lactone dehydrogenase (GLDH), the last enzyme in the plant ascorbate biosynthesis pathway, is associated with the 800-kDa subcomplex but not with the holocomplex. To investigate possible relationships between GLDH and complex I assembly, we characterized an Arabidopsis thaliana gldh insertion mutant. Homozygous gldh mutant plants were not viable in the absence of ascorbate supplementation. Analysis of crude membrane extracts by blue native and two-dimensional SDS-PAGE showed that complex I accumulation was strongly prevented in leaves and roots of Atgldh plants, whereas other respiratory complexes were found in normal amounts. Our results demonstrate the role of plant GLDH in both ascorbate biosynthesis and complex I accumulation.

The respiratory chain includes five enzymatic complexes embedded in the inner mitochondrial (mt) 2 membrane, ensuring oxidative phosphorylation. In a number of organisms, complexes I, III, and IV have been shown to be associated in a supercomplex called "respirasome" (1). The implications of this structural organization in terms of electron transport efficiency and/or complex assembly/stability are still under discussion (2,3). Although only marginal amounts of "respirasomes" seem to be present in plant mitochondria, stable mt supercomplexes containing complex I (CI) and dimeric complex III have been described in several species (4 -6), including tobacco (7).
In most eukaryotes, complex I (NADH:ubiquinone oxidoreductase; EC 1.6.5.3) catalyzes the oxidation of NADH and couples the transfer of electrons to ubiquinone with the translocation of protons from the matrix compartment to the intermembrane space. Eukaryotic complex I (CI) is around 1,000 kDa in size and includes more than 40 subunits (8,9), of which several (10) are mt-encoded (named NAD in plants). Besides the subunits directly involved in NADH and ubiquinone binding and in electron transfer (Fe-S cluster bearing subunits) (11), the function of most CI subunits is still unknown. They are distributed in two large domains, a matrix domain and a membrane domain containing highly hydrophobic polypeptides which, in animal and fungi, include all NAD subunits (12). Although it shares more than 30 conserved subunits with all mt CI (13), the plant enzyme displays several peculiarities. First, the hydrophobic arm is unusually long and possesses specific protuberances (14). Second, two subunits of the peripheral arm, NAD7 and NAD9, are mt-encoded. Third, the plant complexes have been reported to contain specific subunits, among which are three to five ␥-carbonic anhydrase (␥-CA) or ␥-CA-like isoforms (15,16) and the L-galactono-1,4-lactone dehydrogenase (GLDH), which catalyzes the last step in the plant ascorbate synthesis pathway (17,18). Whereas ␥-CA(like) isoforms have been demonstrated to be integral proteins of the CI membrane arm (16), the exact localization of GLDH is still unclear, and the protein has been proposed to be associated with a minor low molecular mass complex in Arabidopsis cell cultures (19).
As for other plant organelle enzymes involved in electron transport chains that include subunits of dual origin, CI assembly is a multiple-step intricate process with many levels of regulation and requires protein assembly factors. Models of CI assembly have recently emerged in fungi, mammals, and the algal Chlamydomonas (20), but little information is currently available in land plants, because of the scarcity of respiratory mutants.
In this study, we give new insights on the composition of CI assembly intermediates in plants, and we re-evaluate the association of GLDH with this complex by genetic and biochemical approaches using Nicotiana sylvestris and Arabidopsis thaliana mutants. The N. sylvestris CMSII mtDNA mutant, devoid of the mt nad7 gene (21), and the nuclear NMS1 mutant, impaired in the processing of the mt nad4 transcript (22), lack significant CI activity (23,24). Both holo-CI and a minor complex, around 800 kDa in size, were previously shown by blue native analyses to be absent in CMSII mt membranes (7). Here we show that although holo-CI is similarly lacking in NMS1, the 800-kDa complex is present, indicating that it could be an assembly intermediate. GLDH was bound to the subcomplex only. To further evaluate possible relationships between GLDH and CI content, we characterized an Arabidopsis mutant interrupted in the GLDH gene. Homozygous Atgldh plants developed only when supplemented with ascorbate, and they were impaired in CI content, whereas other respiratory complexes were in normal amounts. These results show that GLDH expression is necessary for CI accumulation in plants.

EXPERIMENTAL PROCEDURES
N. sylvestris plants were grown in soil, in a greenhouse under a 16-h photoperiod, at a day/night temperature regime of 23/17°C, and under natural illumination supplemented with artificial lighting when necessary to maintain a minimum of 200 mol m Ϫ2 s Ϫ1 at the leaf surface.
Arabidopsis seeds of the SALK_060087 line carrying a T-DNA insertion in the At3g47930 gene were obtained from the Nottingham Arabidopsis stock center (25). Seeds were sterilized and sown under aseptic conditions on agar plates containing Gamborg B5 salt and vitamins, supplemented with 0.5% sucrose. The plated seeds were kept at 4°C for 5 days, and transferred to an illuminated, temperature-controlled growth chamber. Plants were grown at 20°C under a 16-h day/8-h night photoperiod of white light (ϳ80 mol m Ϫ2 s Ϫ1 ).
Preparation of Mitochondria of N. sylvestris Leaves-Mitochondria were purified from 100 g of homogenized leaves by differential centrifugation and gradient centrifugation using two layers of 26 and 46% Percoll in 0.5 M mannitol, 10 mM MOPS-KOH, pH 7.2, 0.1% bovine serum albumin (7). Mitochondria were washed in 0.5 M mannitol, 10 mM MOPS-KOH, pH 7.2, and aliquots were frozen in liquid nitrogen and stored at Ϫ80°C when not used immediately.
Preparation of Crude Membrane Extracts from A. thaliana Leaves and Roots-Two-hundred mg of plant material were harvested and briefly pound at 0°C, in a conical glass homogenizer in 2 ml of 75 mM MOPS-KOH, pH 7.6, 0.6 M sucrose, 4 mM EDTA, 0.2% polyvinylpyrrolidone 40, 8 mM cysteine, 0.2% bovine serum albumin (26). The lysate was filtrated across one layer of miracloth and centrifuged first at 1300 ϫ g for 4 min and the supernatant at 22,000 ϫ g for 20 min. The resultant sediment, which contained most of the thylakoid and mt membranes, was resuspended in 200 l of 10 mM MOPS-KOH, pH 7.2, 0.3 M sucrose.
Crude Arabidopsis membranes suspensions were washed with 600 l of water, sedimented at 22,000 ϫ g for 20 min, and resuspended in 150 l of 50 mM BisTris/HCl, pH 7, 0.5 M 6-aminohexanoic acid, 1 mM EDTA, 0.3 M sucrose, and 0.5 mM Pefabloc SC. After addition of 30 l of 10% ␤-dodecyl maltoside or 60 l of 10% digitonin at 4°C for 15-30 min, respectively, samples were centrifuged at 22,000 ϫ g for 8 min. The supernatants were supplemented with 8 or 16 l of 5% Coomassie Blue G-250 (prepared in 0.020 M BisTris/HCl, 0.5 M 6-aminohexanoic acid) and an aliquot equivalent to 12 mg (leaves) or 20 mg (roots) of fresh weight material were loaded to each lane of a BN gel.
Protein complexes from N. sylvestris mt membranes or from Arabidopsis crude membrane extracts were separated by BN-PAGE using 4 -13 or 3.6 -12% gradient acrylamide gels in Bis-Tris buffer as described previously (7). Blue native gels (thickness 1.5 mm) were made as described previously (27,28) with minor modifications. For analysis of crude membrane extracts, the gel buffer consisted of 0.05 M BisTris, 0.5 M 6-aminohexanoic acid, pH 7, 0.01% detergent and the cathode buffer contained 0.005% Blue G-250 and 0.01% detergent.
Determination of In-gel Enzymatic Activities-After BN-PAGE, the NADH dehydrogenase activity of CI was revealed by incubation of the gel in the presence of 1 mM nitro blue tetrazolium (NBT) and 0.2 mM NADH in 0.05 M MOPS, pH 7.6 (29), after washing for 20 min in 0.1 M MOPS, pH 7.6. The cytochrome oxidase activity was revealed after washing the gel in 50 mM sodium phosphate, pH 7.4, by incubation in the same buffer but containing 0.22 M saccharose, 0.5 mg/ml bovine heart cytochrome c, 0.5 mg/ml 3.3Ј-diaminobenzidine, and 24 units/ml catalase (30).

RESULTS
Presence of a Discrete 800-kDa CI Subcomplex Containing GLDH in the N. sylvestris NMS1 Mutant-As reported previously for CMSII (7), no signal around 1,000 kDa, corresponding to the CI molecular mass, could be detected in BN-PAGE profiles of dodecyl maltoside-solubilized NMS1 mt membranes (Fig. 1A), whereas complexes III (around 500 kDa) and V (around 600 kDa) were readily visible. Similarly, the major NADH/ NBT signal revealing NADH dehydrogenase activity was lacking in NMS1 (Fig. 1A). However, a weak NADH/NBT signal around 800 kDa was present in NMS1 as in the WT (Fig. 1A), suggesting the presence of a minor CI subcomplex in both lines. The polypeptide composition of the mt complexes was resolved by two-dimensional BN/SDS-PAGE (Fig. 1B). In the WT, the 1,000-and 800-kDa CI complexes shared many subunits, including the previously characterized 75-kDa subunit, belonging to the NADH dehydrogenase module (8) as well as NAD7 (7). In NMS1, these polypeptides were associated with the 800-kDa subcomplex only (Fig. 1B), in good agreement with its NADH dehydrogenase activity (Fig. 1A).
The composition of the subcomplex was further investigated by immunological studies. Using anti-␥-CA antibodies, a major signal at 1,000 kDa was observed in the WT profile only. In addition, a weak signal was observed in both WT and NMS1 profiles at around 800 kDa (Fig. 1C), indicating that the subcomplex contains a part of the membrane arm. In contrast, no signal was seen at the level of holo-CI in the WT profiles using antibodies raised against GLDH ( Fig. 2A). Likewise no signal was detected either at the level of CI or at the level of the supercomplex I-III when mt membrane proteins were solubilized by digitonin ( Fig. 2A), a treatment that preserves CI/CIII association (5, 7). However, a clear signal was observed below CI in WT and NMS1 only, at the same mobility as the 800-kDa subcomplex characterized by in-gel NADH/NBT staining. A second signal was detected around 500 kDa ( Fig. 2A). In addition, a diffuse band above the position of CI appeared only in WT profiles, although some discrete bands were detectable below the 500-kDa band in all profiles. Two-dimensional BN/SDS-PAGE showed that both 800-and 500-kDa immunosignals were associated with a polypeptide of about 60 kDa (Fig.  2B), in good agreement with the GLDH molecular mass (32). Whether the 500-kDa signal represents a still unidentified low molecular weight CI subcomplex, devoid of the NADH-oxidizing module, remains to be determined. These results show that GLDH is not associated with holo-CI in N. sylvestris but to a subcomplex accumulating both in the WT and in the NAD4deficient NMS1 mutant, but not in the NAD7-deficient CMSII mutant.
Complex I Accumulation Is Impaired in an A. thaliana gldh Mutant-A possible implication of GLDH in CI formation was further investigated by examining the accumulation of respiratory complexes in an A. thaliana gldh insertion mutant. The  Arabidopsis nuclear genome contains a single GLDH gene (At3g47930) (33), and the SALK_060087 line was identified as a putative mutant containing a T-DNA insertion within this gene. SALK_060087 seeds were germinated in vitro on gelose medium, and roughly a quarter (22.2%) of them (as expected from a single and recessive mutation) exhibited a delayed germination, developed chlorotic cotyledons (Fig. 3A), and ultimately died at the cotyledon stage. However, all the chlorotic seedlings could be rescued by 10 mM ascorbate supplementation (Fig. 3B). PCR amplification using gldh internal and T-DNA left border primers showed these plants to be indeed homozygous for the T-DNA insertion in the last intron of the gldh gene (supplemental Fig. 1, A and B). In the offspring of heterozygous gldh/GLDH plants, ascorbate-dependent plantlets were selected and checked for the genotype and GLDH expression. No GLDH transcript could be detected by semiquantitative reverse transcription-PCR (supplemental Fig. 1C). Taken together, these results indicate that these plantlets are homozygous gldh mutants, deficient in ascorbate synthesis.
A protocol was developed to determine the composition of mt complexes from minute amounts of material without previous mt purification (see under "Experimental Procedures"). Dodecyl maltoside-or digitonin-solubilized proteins from leaf or root crude membrane extracts were resolved by BN-PAGE and BN/SDS-PAGE. In one-dimensional BN-profiles of leaves, main bands represented complexes originating from thylakoid membranes, which were identifiable from their constitutive spots appearing in the second dimension (Fig. 4, A and B), in agreement with the previous reports of proteomic investigation of plant photosystems (34,35). The presence of mt CI was clearly revealed by in-gel NADH/ NBT staining of the BN profiles from dodecyl maltoside-solubilized proteins of Col-0 leaf membrane extracts (Fig. 4A), whereas no significant NADH/NBT staining at 1,000 kDa could be detected in gldh leaf extracts. Similarly, NADH/NBT staining was seen at the levels of CI (1,000 kDa) and supercomplex I-III (1,500 kDa) in the profiles of digitonin-solubilized proteins from Col-0 leaf membrane extract (Fig. 4B) but not in profiles of gldh extract. In contrast, cytochrome oxidase in-gel staining gave identical responses in Col-0 and gldh mutant (Fig. 4B). The two-dimensional BN/SDS-PAGE profiles from dodecyl maltoside-solubilized proteins of Col-0 and gldh leaf extracts were comparable, and similar amounts of spots originating from the mt complexes III and V on both sides of PSI were detected (Fig. 4C). The extracts contained several major or minor complexes having their mobility close to that of CI, as PSII-LHCII, or vacuo-   NOVEMBER 21, 2008 • VOLUME 283 • NUMBER 47 JOURNAL OF BIOLOGICAL CHEMISTRY 32503 lar ATPase, a known contaminant of N. sylvestris (Fig. 1, B and  C) or Neurospora crassa mt preparations (36). Thus only the 75-kDa spot seen above the PSII-LHCII trace represents an unambiguous marker for CI. It was seen in the Col-0 but not in the gldh two-dimensional profiles (Fig. 4C) in good agreement with the data of NADH/NBT stain obtained following the first dimension.

Complex I and L-Galactono-1,4-lactone Dehydrogenase
The results obtained from leaf extracts were confirmed by analyses of root extracts. Spots originated from dodecyl maltoside-solubilized proteins of Col-0, and gldh mt membranes represent main spots on two-dimensional profiles (Fig. 4D) fitting well with the profiles obtained from mt preparations of N. sylvestris leaves (Fig. 1). Comparable amounts of spots derived from complexes III and V were observed in the profiles of both lines, whereas spots derived from CI were lacking in the gldh mutant. Because this deficiency could not be explained by a decrease of mt membranes in the sample as shown by the comparable levels of complexes III, IV, and V, these results indicated that CI is lacking in gldh plants.

DISCUSSION
As in other organisms, the elucidation of complex I assembly in plants depends on the availability of mutants of either peripheral or membrane domains. Here we bring new insights to this process using available N. sylvestris mutants, the CMSII mutant lacking the peripheral NAD7 subunit (21) and the NMS1 mutant lacking the membrane NAD4 subunit (22), as well as a newly characterized A. thaliana gldh mutant. Atgldh seedlings did not develop beyond the cotyledon stage in the absence of ascorbate supplementation, confirming that GLDH activity is an obligatory step in the guanosine diphosphate mannose pathway for ascorbate biosynthesis in plants (17,18). A strict ascorbate requirement has already been shown for the development of Arabidopsis mutants impaired in the synthesis of GDP-L-galactose phosphorylase, an enzyme involved upstream in the ascorbate biosynthesis pathway in plants (37). However, if ascorbate supplementation to gldh plants allowed greening and development, their growth rate remained very low. Reduced growth rates were also observed in antisense gldh tomato plants (38).
Holo-CI could not be detected in leaf mt membranes of the N. sylvestris NMS1 mutant lacking the membrane NAD4 subunit, as is the case in the CMSII mutant devoid of the peripheral NAD7 subunit (7). However, in contrast to CMSII, a minor 800-kDa form previously observed in the WT (7) was detected in NMS1. This subcomplex contains subunits characteristic of the peripheral domain (e.g. NAD7 and the 75-kDa component of the dehydrogenase module) and displays in-gel NADH dehydrogenase activity (Fig. 1). The presence of this subcomplex in NMS1 mitochondria despite the lack of holo-CI indicates that it is probably an assembly intermediate and not a degradation product. Existence of a subcomplex comprising matricial subunits has been previously reported in the maize NAD4-deficient NCS2 mutant (39). However, the presence in the N. sylvestris subcomplex of the ␥-CA subunit, an integral membrane subunit of plant CI (16), shows that it also contains a part of the membrane arm. In contrast to ␥-CA, GLDH, the enzyme catalyzing the last step in ascorbate synthesis (17), was only associ-ated with an 800-kDa subcomplex and not with the holo-CI in N. sylvestris mitochondria. Ascorbate synthesis carried out by the sub-CI-bound GLDH form of N. sylvestris leaves is not essential, as previously suggested for the low molecular mass CI of Arabidopsis cell cultures (40), because ascorbate levels are not markedly altered in the CMSII mutant lacking this subcomplex (41,42). Interestingly, GLDH was also associated with a 500-kDa mt complex (Fig. 2), but whether it corresponds to a CI assembly intermediate remains to be shown.
Taken together, these results support the following model for the CI assembly process in plants: the peripheral arm including the N and Q modules (8) would bind to a large membrane segment, resulting in an assembly intermediate around 800 kDa that includes GLDH. The formation of mature CI would require the elimination of GLDH and the addition of a set of subunits, including NAD4, the analogous subunit (NUOM) being located at the tip of the membrane arm in Escherichia coli (43). Similarly, the 700-kDa subcomplex that accumulates in a Chlamydomonas reinhardtii NAD4-deficient mutant includes subunits of both the matrix and the membrane arm, but it is devoid of ND5, located in the membrane arm tip (44). This 700-kDa subcomplex in the C. reinhardtii ND5-deficient mutant is deprived of a set of subunits located at the distal domain of the CI membrane arm (45). Lack of the membrane arm distal segment does not prevent the integration of ␥-CA into the CI subcomplex of either N. sylvestris NAD4-deficient (this work) or C. reinhardtii NAD5-deficient mutants (45). The subcomplexes accumulating in Chlamydomonas have been proposed to be assembly intermediates (44), and the analysis of N. sylvestris mutants strongly supports common points between the late steps of plant and C. reinhardtii CI assembly processes.
Alternatively, it can be hypothesized that fixation of GLDH to the assembly intermediate impairs further attachment of the distal membrane domain, and therefore the holoenzyme can only derive from a GLDH-free subcomplex and the presence of NAD4. In this hypothesis, the GLDH-associated subcomplex would not be a productive intermediate. However, such a model is not supported by the study of the Atgldh insertion mutant, which shows that the absence of GLDH expression inhibits the accumulation of the holoenzyme. Indeed, complex I and its associated NADH dehydrogenase activity were lacking in Atgldh leaf and root membranes, although the other respiratory complexes were unaffected. Lack of complex I in the gldh mutant demonstrates that GLDH is involved in the process leading to the formation of CI, although the exact mechanism remains to be elucidated. GLDH may be involved in the synthesis/stability of individual subunits and/or assembly intermediates. To date, no plant CI assembly factors have been characterized. In fungi and mammals, several protein factors associated with high molecular weight CI subcomplexes but not with the holoenzyme have been shown to be directly involved in the CI assembly process. In N. crassa, CIA30 and CIA84, two proteins that do not belong to the mature holoenzyme, were shown to be associated with a large membrane segment accumulating in the nuo21 mutant unable to assemble holo-CI (46). Disruption of CIA genes resulted in the inability to form the membrane segment. Recent studies of human CI assembly have led to the characterization of discrete subassembly complexes (most of them in the range of 800 and 450 -500 kDa) and the identification of at least three different assembly factors, B17.2L (47,48), NDUFAF1, a CIA30 homologue (49,50), and Ecsit (51). The deficiency of each of these three factors leads to a severe reduction in human CI levels and activity. Although further biochemical studies are needed to elucidate the kinetic steps of the CI assembly process, our results demonstrate the dual function of GLDH in plants, i.e. ascorbate synthesis and holocomplex I accumulation.