Suppression of Nda-type Alternative Mitochondrial Nad(p)h Dehydrogenases in Arabidopsis Thaliana Modifies Growth and Metabolism, but Not High Light Stimulation of Mitochondrial Electron Transport

The plant respiratory chain contains several pathways which bypass the energy-conserving electron transport complexes I, III and IV. These energy bypasses, including type II NAD(P)H dehydrogenases and the alternative oxidase (AOX), may have a role in redox stabilization and regulation, but current evidence is inconclusive. Using RNA interference , we generated Arabidopsis thaliana plants simultaneously suppressing the type II NAD(P)H dehydrogenase genes NDA1 and NDA2. Leaf mitochondria contained substantially reduced levels of both proteins. In sterile culture in the light, the transgenic lines displayed a slow growth phenotype, which was more severe when the complex I in-hibitor rotenone was present. Slower growth was also observed in soil. In rosette leaves, a higher NAD(P)H/ NAD(P) + ratio and elevated levels of lactate relative to sugars and citric acid cycle metabolites were observed. However, photosynthetic performance was unaffected and microarray analyses indicated few transcriptional changes. A high light treatment increased AOX1a mRNA levels, in vivo AOX and cytochrome oxidase activities, and levels of citric acid cycle intermediates and hexoses in all genotypes. However, NDA-suppressing plants deviated from the wild type merely by having higher levels of several amino acids. These results suggest that NDA suppression restricts citric acid cycle reactions, inducing a shift towards increased levels of fermentation products, but do not support a direct association between photosynthesis and NDA proteins.


Introduction
The respiratory chain is highly branched in plants as compared with non-photosynthetic eukaryotes (Rasmusson et al. 2008).In particular, type II NAD(P)H dehydrogenases (DHs) and the alternative oxidase (AOX) are energy bypass proteins, which form alternative electron transport pathways.These circumvent the proton-pumping protein complexes I, III and IV, the latter two constituting the cytochrome pathway of oxygen consumption.Given that they bypass the proton-pumping respiratory complexes and do not pump protons themselves, the energy bypass proteins reduce the efficiency of respiratory ATP production (Fernie et al. 2004, Rasmusson et al. 2004).Pairwise co-expression of genes for specific AOX isoforms and corresponding type II NADH DHs has suggested that they may work as cooperative functional units (Clifton et al. 2005, Escobar et al. 2006, Ho et al. 2007, Rasmusson et al. 2009, Yoshida and Noguchi 2009).Under conditions with a large electrochemical proton gradient caused by a limiting ADP supply, the energy bypass proteins can allow removal of excess reductants, which otherwise may induce the formation of reactive oxygen species (Møller 2001).Energy bypass enzymes may also regulate cellular NAD(P)(H) levels, as suggested for an NADPH-specific type II DH (Liu et al. 2008, Liu et al. 2009).
To date, all investigated plant NDB proteins have been localized to the external surface of the inner mitochondrial membrane, whereas potato NDA1 and A. thaliana NDA1, NDA2 and NDC1 reside on the matrix-facing surface of the inner membrane (Rasmusson et al. 1999, Elhafez et al. 2006).It has been proposed that NDA1 and NDA2 are also localized in peroxisomes, but a peroxisomal functional role was not suggested (Carrie et al. 2008).Correlations between gene expression patterns and enzyme activities in potato and A. thaliana mitochondria have strongly suggested that NDA1 oxidizes NADH (Svensson and Rasmusson 2001, Svensson et al. 2002, Moore et al. 2003).Although the substrate specificity of NDA2 has not been determined, the close similarity between A. thaliana NDA1 and NDA2 and the sequence of their active sites suggested that both oxidize NADH (Michalecka et al. 2004).
Internal type II NADH DHs are thought to function only when concentrations of NADH in the matrix are high.This is based on the higher K m (NADH) for internal rotenone-insensitive NADH oxidation compared with complex I-mediated, rotenone-sensitive activity (Rasmusson et al. 2004).Light induction of NDA1 has suggested a possible function in the oxidation of photorespiratory NADH (Svensson and Rasmusson 2001, Michalecka et al. 2003, Escobar et al. 2004).This is consistent with the involvement of AOX and rotenone-insensitive NADH oxidation during glycine oxidation in isolated mitochondria and protoplasts (Dry andWiskich 1985, Igamberdiev et al. 1997).NDA2, on the other hand, is not regulated by light and was suggested to function in heterotrophic metabolism (Michalecka et al. 2003, Escobar et al. 2004, Elhafez et al. 2006).
Mutant plants lacking functional respiratory complexes I or IV display severe phenotypes when grown under standard conditions, including decreased growth and photosynthesis and changes in morphology, metabolites and gene expression (Newton et al. 2004, Noctor et al. 2007).In contrast, genetic modifications of energy bypass pathways have resulted in substantially milder phenotypes.For example, A. thaliana lacking AOX1a displayed growth defects only under non-optimal growth conditions (Fiorani et al. 2005, Giraud et al. 2008).The suppression of the external type II NADPH DH in Nicotiana sylvestris caused delayed bolting in high light, but did not affect biomass accumulation (Liu et al. 2008, Liu et al. 2009).Furthermore, an A. thaliana T-DNA mutant for the NDA1 gene did not display a growth phenotype (Moore et al. 2003), and suppression of the external NADH DH gene NDB4 in A. thaliana mainly induced temporal changes in leaf area (Smith et al. 2011).The relatively subtle phenotypes associated with decreased energy bypass capacities make such modifications appealing models to study mild redox changes in plant cells.From a systems biology point of view, subtle changes that avoid network-level perturbations are advantageous for studies of metabolism (Stitt et al. 2010).
Arabidopsis thaliana NDA1 and NDA2 are paralogs displaying 82% amino acid sequence identity (Michalecka et al. 2003, Moore et al. 2003), and may have at least partially overlapping functions.To investigate the physiological role(s) of type II NAD(P)H DHs, while avoiding redundancy problems, we suppressed both NDA1 and NDA2 using a single RNA interference (RNAi) construct.We found that NDA1,2-suppressed plants grown in soil had decreased biomass and a higher NAD(P)H/ NAD(P) + -ratio.Specific changes in the metabolome included changes in fermentation products, sugars and citric acid cycle metabolites in leaves.However, small or no effects were observed on in vivo respiratory and photosynthetic parameters, and on transcript profiles.

A single RNAi construct suppresses the expression of both NDA1 and NDA2
To target specifically both NDA1 (At1g07180) and NDA2 (At2g29990) by RNAi, we selected a highly conserved cDNA region that is not shared with NDB and NDC genes.For this region, a hybrid NDA1/NDA2 segment was made by PCR (Supplementary Fig. S1).An inverted repeat construct was transformed into A. thaliana Col-0 and, after segregation analysis, plants homozygous for the RNAi construct were obtained in the T 3 generation.
As shown in Fig. 1, three RNAi lines displayed decreased transcript levels for NDA1 [9-36% of wild-type (WT) levels] and NDA2 (56-78% of WT levels).For protein analyses, peptide antisera were generated against A. thaliana NDA1 and NDA2.In Western blots, affinity-purified NDA1 and NDA2 antibodies were specific for Eschericha coli-produced NDA1 and NDA2, respectively (Supplementary Fig. S2).For mitochondria purified from shoots, an NDA1 signal was specifically detected in the WT but was hardly discernible in lines 17.11 and 26.6 (Fig. 2).The NDA2 antibodies detected a band in the WT which was approximately half as strong in the RNAi lines (Fig. 2).The signals were consistent with the size of approximately 48 kDa previously observed in A. thaliana using potato NDA antibodies (Michalecka et al. 2003).After purification, the potato NDA antibody almost exclusively detected NDA2 (Supplementary Fig. S2).A malate DH antiserum detected a single band of approximately 35 kDa.This is consistent with the size of the mitochondrial isoform (35-37 kDa), but not the smaller (approximately 32 kDa) glyoxysomal/peroxisomal isoform (Gietl et al. 1996, Liu et al. 2009).This suggests that the mitochondrial preparations were devoid of significant peroxisomal contamination.

Transgenic lines displayed a light-dependent growth inhibition in sterile culture
All three RNAi lines displayed significantly smaller rosette diameters than the WT when grown in light on sterile medium containing sucrose or glucose (Fig. 3A).With mannitol, a significant growth decrease was observed in one RNAi line, and consistent, though not significant (P = 0.06), changes were observed in the other two RNAi lines, as compared with the WT.The average growth decreases in the RNAi lines, as compared with the WT, were 30 ± 1.5%, 20 ± 1% and 22 ± 4%, on media containing sucrose, glucose and mannitol, respectively.Normalized to the WT, the growth of each of the three transgenic lines was significantly different between sucroseand glucose-supplemented cultures (results not shown).In contrast, all lines displayed similar growth in darkness on medium containing sucrose (Fig. 3B).
Using sucrose as the carbon source, relative effects of respiratory inhibitors were analyzed.The cellulose synthesis inhibitor isoxaben, to which an A. thaliana complex I-deficient mutant was insensitive (Nakagawa and Sakurai 2006), was also investigated.All inhibitors decreased the growth of all genotypes (Fig. 4).However, the complex I inhibitor rotenone had a stronger effect on the transgenic lines than on the WT.The AOX inhibitor n-propyl gallate (n-PG) had a slightly less detrimental effect on the growth of all three transgenic lines than on the WT, although the difference was not significant for individual lines.Antimycin A (a complex III inhibitor) and isoxaben did not show consistent effects segregating the genotypes (Fig. 4).The seed germination ratio was similar in all genotypes, or slightly higher in the transgenic lines (data not shown).
NDA gene suppression causes slower growth in soil and elevated NAD(P)H/NAD(P) + ratios Plants were grown in soil under a relatively short light period (12 h) and normal growth light (80 mmol m À2 s À1 ).A significant growth defect was observed for the RNAi lines, as seen in the rosette size and dry weight (Fig. 5A, B).Ratios of fresh weight and leaf area to dry weight were unaffected, and other phenotypes were not observed (results not shown).In 10 h daylength, which allows analysis during a longer vegetative growth phase, a growth defect phenotype was also observed (Fig. 5C).Relative growth rates in older plants (days 20-43), calculated from data in Fig. 5C, did not show significant differences between the genotypes (results not shown).Thus, the results of the growth experiments on sterile medium and in soil jointly suggest that there is a growth difference between the WT and the NDA1,2-suppressed lines, which is established at a relatively early stage of development.
The NDA1,2-suppressed lines displayed dramatically increased NADH levels and decreased NAD + levels in rosettes (Fig. 6).Consequently, the NADH/NAD + ratio had increased to 3 and 3.6 times the WT level in lines 17.11 and 26.6, respectively.The NADPH levels were unchanged, but the NADP + level was significantly decreased in both transgenic lines, giving an NADPH/NADP + increase of 1.7-1.8times in the transgenic lines (Fig. 6).Total levels of both NAD(H) and NADP(H) were significantly decreased in both transgenic lines (Fig. 6).Overall, the results suggest that the NDA1,2 suppression strongly affects the cellular NAD(P)(H) redox balance of the leaf.
In vivo alternative and cytochrome oxidase activities are increased by high light but unaffected by NDA gene suppression We wanted to test whether NDA proteins and AOX have functional associations directly connected to photosynthetic  metabolism.We therefore designed a system for parallel analysis of gene expression, in vivo respiration and photosynthesis, which allowed comparisons of normal growth light (80 mmol m À2 s 1 ) and a 2 h high light treatment (800 mmol m À2 s À1 ).Chl fluorescence was measured by clamping leaves on intact plants, respiration and oxygen isotope discrimination were analyzed in darkened detached leaves, and rosettes were sampled for transcript analyses.
The high light treatment did not substantially affect the NDA1 and NDA2 transcript levels in the WT, or the suppression of the NDA genes in the RNAi lines (Fig. 7).The decreased levels of NDA1 and NDA2 transcript in the transgenic lines of the T 4 generation in the growth light (Fig. 7) were consistent with the data for the T 3 generation (Fig. 1), demonstrating that the suppression was stable over two plant generations.AOX1a and NDB2 transcript levels generally increased in response to high light, whereas the transcript levels of NDB1, NDC1 and the 28.5 kDa subunit of complex I were essentially unchanged (Fig. 7).After the 2 h high light treatment, leaf NAD(P)(H) levels were less affected by the transgenic modification than in the growth light.In the high light, only a small increase in NADH was observed in the RNAi lines as compared with the WT (Supplementary Fig. S3).NADP + levels were significantly lower in the RNAi lines than in the WT in the high light, but less so than observed in the growth light (Fig. 6; Supplementary Fig. S3).Overall, differences observed between the genotypes were thus reduced by the high light treatment.
Under the growth light conditions, the in vivo complex IV activity was twice that of AOX in all genotypes (Fig. 8).After the high light exposure, a general increase was seen in the total dark respiration rate and the activities of complex IV and AOX, as compared with plants remaining in growth light.The results thus suggest that high light leads to a larger input of electrons into the respiratory chain during the following dark period, but the NDA genes are not essential for mediating the increased electron flux.The capacity of the alternative pathway did not change in response to the high light treatment in any genotype, nor did it differ between the WT and the transgenic lines in growth light or high light (Table 1).These results indicate that the induced AOX1a expression was not connected to an elevated level of activated enzyme, consistent with a report where the AOX1a mRNA level, but not the total AOX protein level, was increased after 2 h of high light (Yoshida and Noguchi 2009).
For plants grown in the normal growth light, net photosynthesis and stomatal conductance at saturating light were invariant between the WT and the NDA-suppressed lines (data not shown).The maximum quantum efficiency of PSII (F v /F m ) measured at growth light conditions was also not different between genotypes (Table 2).The high light treatment induced a similar decrease of F v /F m in the WT and the transgenic lines (Table 2), indicating a similar level of photoinhibition.These results indicate that the transgenic modification does not restrict photosynthetic capacity, and does not support an involvement of NDA proteins in photosynthetic metabolism in high light conditions.

The suppression of NDA genes induces light-dependent changes in metabolite levels
To investigate further the growth decrease, metabolic profiling was performed on rosettes from the control and high   S1, S2).In growth light, the transgenic lines displayed decreased levels of several hexoses (fructose, galactose and glucose) and citric acid cycle intermediates (citrate and fumarate), as compared with the WT.In contrast, lactate levels were substantially elevated in the transgenic lines.
The high light treatment induced massive increases in a large set of metabolites in all genotypes (Fig. 9).The changed metabolites included several amino acids, especially glycine, which increased 22-fold in the WT, but also phenylalanine (6-fold) and alanine (4-fold).In addition, the citric acid cycle metabolites isocitrate, cis-aconitate and citrate and the hexoses glucose, galactose and fructose were elevated >1.5-fold in the WT.This is consistent with the expected effects of high light on photosynthesis, photorespiration and osmotically active components.
Different sets of metabolites varied between the genotypes in high light and in growth light.In high light, the transgenic lines displayed higher levels of malate and several amino acids (e.g.alanine, glycine and glutamate) compared with the WT (Fig. 9; Supplementary Table S2).We recalculated the metabolite data into ratios, which are independent of potential errors in tissue amounts used.Specific to growth light, the ratios of lactate to several central sugars and to most citric acid cycle  metabolites were significantly higher (1.9-2.4 times) in the transgenic lines compared with the WT, whereas the ratio of alanine to most citric acid cycle metabolites was moderately higher in both light conditions used (Supplementary Fig. S4).In contrast, the ratios of major redox-linked metabolite couples (glycine/serine, malate/aspartate, glutamate/2-oxoglutarate, 2-oxoglutarate/citrate and dehydroascorbate/ascorbate) were similar in all genotypes, independently of the light regime.This suggests that these redox couples can be unaffected by large changes in NADH/NAD + , whereas the increased level of fermentation products (lactate) is highly responsive to the deficiency in NDA gene expression.

Transcript profiling of RNAi lines reveals subtle changes in gene expression
A microarray analysis was performed on plants grown in normal growth light.Lines 17.11 and 26.6 were compared with the WT individually and after pooling the data from both transgenic lines.As expected, the NDA1 and NDA2 signals were substantially reduced, by 90% and 54%, respectively (Supplementary Table S3).This is a slightly larger reduction than observed by real-time reverse transcription-PCR (RT-PCR) (Fig. 7).Consistent with the real-time RT-PCR data for AOX1a, NDB1, NDB2, NDC1 and the 28.5 kDa subunit of complex I (Fig. 7), their microarray signals deviated <15% between the WT and any of the transgenic lines (data not shown).Apart from the signals for the silenced NDA genes, a total of 30 probes displayed statistically significant signal changes of !1.4-fold when comparing the pooled data for the transgenic lines with the WT (Supplementary Table S3).The majority of observed changes constituted decreases in transcript abundance relative to the WT, including a 4-fold signal decrease for GLUCOSE-6-PHOSPHATE DH 4 (At1g09420) and smaller effects on the COPT2 gene for a copper transporter (At3g46900) and on two genes specifying 2-oxoglutarate and iron-dependent oxygenase superfamily proteins (At3g46490 and At3g50210).Among the probes displaying increased levels, some were connected to pathogen-related responses, e.g.disease resistance protein (At3g44630) and pathogenesisrelated gene 5 (At1g75040).
For an overview of the transcriptomic effects, the profiles were analyzed using MapMan (Table 3).This revealed significant differences in expression patterns for gene bins specifying central cellular functions, including changes in RNA processing and transcriptional regulation, and a general increase in the gene expression for cytosolic ribosomal proteins in the transgenic lines.Additionally, a decrease was seen for the bin jointly containing the alkaloid biosynthesis enzymes nitrilases, nitrile lyases, berberine bridge enzymes, reticuline oxidases and troponine reductases.No general association with central carbon metabolism or stress responses could be observed.The transcript profiling thus indicates that the observed NDA1,2 suppression and associated metabolic changes have moderate influences on transcription.Growth light 0.13 ± 0.02 0.16 ± 0.02 0.12 ± 0.02 High light 0.14 ± 0.02 0.13 ± 0.01 0.11 ± 0.01 Plants were grown and treated as in Fig. 7.
Values are means ± SE from four replicates (three rosettes in each replicate).

Discussion
Suppression of NDA genes causes growth retardation under standard growth conditions In this study, a single RNAi construct suppressed mRNA levels for both NDA1 and NDA2 in A. thaliana plants, causing corresponding decreases in protein levels in isolated leaf mitochondria.Numerous studies have reported the targeting of NDA proteins to mitochondria, and their presence and activity in the same organelle (Rasmusson et al. 1999, Rasmusson and Agius 2001, Michalecka et al. 2003, Moore et al. 2003, Rasmusson et al. 2004, Elhafez et al. 2006, Carrie et al. 2008).
Additionally, the 11-mer C-terminal peptides of NDA1, NDA2 and NDB1 have been found to target green fluorescent protein to peroxisomes in Arabidopsis (Carrie et al. 2008), and several homologs from rice and moss display similar results upon the  S2).The WT level in growth light is set to unity for all data.Error bars denote the SE for 6-7 biological replicates, each containing three rosettes.Cit., citrate; Fru, fructose; Fum., fumarate; Gal, galactose; Glc, glucose; Lac., lactate; Mal., malate; Male., maleate; Shiki., shikimate.
same targeting analysis (Xu et al. 2013).Western blots using antisera made against NDA1 have detected signals of approximately 50 kDa in isolated peroxisomes from potato and A. thaliana (Rasmusson andAgius 2001, Carrie et al. 2008).However, in potato, this was interpreted as a putative weak interaction with catalase (Rasmusson and Agius 2001), which can constitute up to 50% of the peroxisomal protein level (Struglics et al. 1993), and has a mass similar to NDA proteins ($50 kDa).By Western analyses, NDB1 was detected in potato mitochondria, but not in peroxisomes (Liu et al. 2009).In contrast to a targeting function of the C-terminus, the recent structural determination of the yeast homolog Ndi1p showed that the C-terminus, which is conserved among eukaryotes, is part of a membrane-binding domain of amphiphilic helices (Feng et al. 2012).It is therefore presently unclear whether NDA and/or NDB proteins are present in peroxisomes in vivo.Considering that NAD(P)H:ubiquinone oxidoreductase or quinol oxidase activity has not been reported for peroxisomes, it is furthermore not possible to predict a functional pathway containing a putative peroxisomal homolog, nor to speculate on the consequences for cellular NAD(H) homeostasis.We have therefore focused our analyses on the known mitochondrial function of the NDA proteins.A previously described A. thaliana NDA1 T-DNA mutant displayed normal growth (Moore et al. 2003), whereas the NDA1,2-suppressed lines had a slower growth in sterile culture and in soil (Figs. 3, 5).The phenotypic difference between the NDA1 mutant and the NDA1,2-suppressed lines indicates a functional redundancy between NDA1 and NDA2, consistent with the high sequence conservation, with both being matrixfacing NADH DHs (Michalecka et al. 2003, Moore et al. 2003, Elhafez et al. 2006, Rasmusson et al. 2008).This is also consistent with the highly elevated NADH/NAD + ratio in the NDA1,2suppressed lines in growth light (Fig. 6).While type II NAD(P)H DHs bypass one site of energy conservation (complex I), AOX bypasses two (complex III and IV).We therefore expected the phenotype of the NDA-suppressed plants to be less severe than the stress-dependent phenotypes of AOX-deficient plants (Vanlerberghe et al. 1994, Ordog et al. 2002, Fiorani et al. 2005, Giraud et al. 2008).However, the slow growth of the NDAsuppressed plants indicates a metabolic perturbation under non-stressful growth conditions.Importantly, it also shows that NDA proteins are not merely supplementary pathways to complex I, which has been a general belief (Rasmusson et al. 2004).
In sterile culture, a slow growth phenotype was only observed in light (Fig. 3).This is consistent with the previously reported light induction of the NDA1 transcript, NDA protein accumulation and internal rotenone-insensitive NADH oxidation (Svensson andRasmusson 2001, Michalecka et al. 2003).Specific sugar effects were observed as all three transgenic lines grew significantly faster on glucose than on sucrose, relative to the WT.This suggests interactions between sugar metabolism and the respiratory chain [e.g. by shifts between different NAD(P)-linked catabolic pathways induced by different carbon sources].This is further supported by the decreased level of leaf hexoses in the transgenic lines in growth light (Fig. 9).Previous investigations have suggested interactions between mitochondrial activities and sucrose metabolism, especially in relation to cell wall synthesis (Kro ¨mer 1995, Carrari et al. 2003, Nakagawa and Sakurai 2006, van der Merwe et al. 2010).The present lack of mechanistic models, however, emphasizes the need to identify the components that link these functional domains.

The relative importance of different electron transport pathways in promoting photosynthetic metabolism
High light metabolism involves several mitochondrial processes, including photorespiratory glycine oxidation, reoxidation of surplus chloroplast reductants, and citric acid cycle provision of carbon for N assimilation (Raghavendra and Padmasree 2003).The increased NADH level in the WT in response to high light, and the relatively small effect of the transgenic perturbation on the NADH level in high light (Fig. 6; Supplementary Fig. S3), is consistent with a dominance of the photosynthetic metabolism for the total leaf NAD(P)H levels in high light.Roles for NDA1 and AOX1 genes in photosynthetic metabolism have been postulated based on their diurnal regulation and light responsiveness (Svensson and Rasmusson 2001, Escobar et al. 2004, Elhafez et al. 2006, Rasmusson and Escobar 2007, Yoshida and Noguchi 2009), the negative effects of AOX inhibitors on photosynthesis (Raghavendra and Padmasree 2003) and the elevated plastoquinone reduction level in AOX1a mutants after high light treatments (Yoshida et al. 2011).Rotenone-insensitive NADH DHs and AOX are also active in glycine oxidation in isolated leaf mitochondria and protoplasts, indicating that they facilitate photorespiratory pathway flux (Dry andWiskich 1985, Igamberdiev et al. 1997).Based on this, we expected that high light, via an increased ubiquinone reduction level (Yoshida et al. 2011), would increase AOX1a expression and thus in vivo AOX activity in the WT during the following dark measurement period.If internal type II NADH DHs oxidize NADH formed by the glycine decarboxylase, then a deficiency in NDA proteins should limit the ubiquinone reduction level, and thus AOX activation in the RNAi plants.However, in contrast to this scheme, high light induced both AOX-and complex IVmediated dark respiration in vivo, but independent of the NDA gene suppression (Fig. 8).The AOX1a mRNA level was increased (Fig. 7), but not the AOX capacity (Table 1).Furthermore, the degree of photoinhibition (i.e.decrease in maximum quantum efficiency of PSII) was similar in NDAsuppressed and WT plants (Table 2), indicating that photosynthesis in high light was unaffected by the NDA silencing.In addition, the glycine/serine ratio was similar in all genotypes (Supplementay Fig. S4), indicating that the NDA suppression does not restrict photorespiration.The glycine/serine ratio was also unchanged in the N. sylvestris CMSII mutant and the A. thaliana ndufs4 mutant, both lacking complex I (Dutilleul et al. 2003, Meyer et al. 2009).In contrast, an A. thaliana mutant for UCP1 displayed a restricted glycine to serine conversion (Sweetlove et al. 2006), and an AOX1a mutant showed an elevated glycine/serine ratio in the presence of antimycin A (Strodtko ¨tter et al. 2009).These results, in combination with the generally elevated in vivo AOX and complex IV activities seen here (Fig. 8), stress the importance of not only AOX but also the cytochrome pathway for allowing the reoxidation of reductants from chloroplasts, as recently observed in genotypes with modified AOX1a expression (Florez-Sarasa et al. 2011).To allow elevated AOX and complex IV activities, DHs are necessary.The results thus demonstrate a substantial redundancy between the multiple oxidation pathways for glycine-derived NADH.This would include reductant export and subsequent oxidation by the external NADH DHs (Fig. 10).Such redundancy is also suggested by the more severe, but not lethal, effect of rotenone on growth of NDA-suppressed seedlings, as compared with the WT (Fig. 4).Shuttling of NADH-derived reductant from the glycine decarboxylase, via malate/oxaloacetate exchange, has been suggested for reductant export to the cytosol and for reductant delivery to the electron transport chain (Wiskich et al. 1990, Kro ¨mer andHeldt 1991).
Fermentation products as markers for redox restrictions in the mitochondrial matrix?
Transcript levels for genes encoding proteins central to cellular functioning (e.g.ribosomal proteins) were significantly affected by the NDA suppression, without indicating a direct cause of the growth defect in the transgenic plants (Table 3).Thirty probes were responsive to the transgenic modification, but the only one connected to central metabolism was GLUCOSE-6-PHOSPHATE DH 4 (Supplementary Table S3).This plastidand peroxisome-located homolog is however inactive as a DH and is involved in intracellular transport of GLUCOSE-6-PHOSPHATE DH 1 (Wakao andBenning 2005, Meyer et al. 2011).The restricted number of transcriptional changes in our transgenic lines (Supplementary Table S3) is similar to observations in tomato plants with suppressed fumarase and succinate DH (Nunes-Nesi et al. 2007, Araujo et al. 2011).Additionally, A. thaliana with suppressed AOX1a showed limited changes in respiratory and stress-associated genes (Umbach et al. 2005).In contrast, a T-DNA mutant for AOX1a displayed substantial expressional changes in a range of genes associated with plastids and oxidative stress under normal growth conditions, where no phenotype was observed (Giraud et al. 2008).NDA1,2-suppressed lines displayed increases in lactate and alanine relative to citric acid cycle intermediates under growth light and high light, respectively (Fig. 9; Supplementary Fig. S4).This suggests a restriction of the citric acid cycle at the level of pyruvate, and a resulting shift in equilibrium towards an accumulation of the fermentation products lactate and alanine (Fig. 10).Consistently, increases in alanine were observed in A. thaliana mutants lacking complex I (Nakagawa andSakurai 2006, Meyer et al. 2009), and rotenone treatment of A. thaliana cells increased lactate and alanine levels and decreased the levels of citric acid cycle intermediates (Garmier et al. 2008).An increased NADH reduction level has also been predicted to decrease the activities of pyruvate, 2-oxoglutarate and NAD-isocitrate DHs (Igamberdiev andGardestro ¨m 2003, Noctor et al. 2007).To avoid accumulation of pyruvate, it is probably converted into lactate via the lactate DH or to alanine via the alanine aminotransferase, similar to the reactions that take place under hypoxia (Rocha et al. 2010).Thus, the increases in lactate and alanine associated with the suppression of NDA genes indicate an increase in the matrix NADH reduction level, though not to the extent that it affects the whole-leaf malate/aspartate ratio.The highly elevated NADH/NAD + ratio in the transgenic plants under normal growth light (Fig. 6) suggests that there is also a substantial change in extramitochondrial NAD(H).The effect is substantially smaller in high light, but may be obscured by a general increase in NADH in high light, and by changes in the chloroplast.Consistent with the elevated NADH/NAD + ratio in the RNAi plants in normal growth light (Fig. 6), an increased NADH/NAD + ratio and lactate level were observed in seedlings of A. thaliana mutants for the cytosolic NAD + -glycerol-3-phosphate DH, which was suggested to be a part of a redox shuttle from the cytosol to mitochondria (Shen et al. 2006).However, the metabolic changes in NDA-suppressed lines must also involve the citric acid cycle and export of reductant across the inner mitochondrial membrane.The higher steady-state levels for fermentation products in plants restricted in complex I activity or NDA gene expression suggests that the matrix NADH pool associated with the citric acid cycle cannot be sufficiently reoxidized by reductant export via the malate/oxaloacetate shuttle (Fig. 10).This further indicates a functional redox separation between the glycine cleavage system and the citric acid cycle DHs, for example by metabolic domain formation (Wiskich et al. 1990).
Another path of redox export is the mitochondrial citrate valve, through which cytosolic NADPH may be derived from mitochondrial NADH, driven by NADH oxidation via reverse NAD-isocitrate DH activity (Igamberdiev and Gardestro ¨m 2003).The increased NADPH/NADP + ratio in the NDAsuppressed plants (Fig. 6) thus suggests that the citrate valve may be active in mitochondrial export of surplus reductant.The elevated NADPH/NADP + ratio could additionally provide a link Fig. 10 Metabolic model for separate NAD(H) redox systems in the mitochondrial matrix.Internal NADH dehydrogenases, such as complex I and NDA proteins, not essential for photorespiratory flux.Instead, NADH generated by glycine oxidation is efficiently shuttled out of the mitochondrion for oxidation by external NADH dehydrogenases and by the peroxisomal photorespiratory reactions.Matrix NADH generated by pyruvate dehydrogenase and the citric acid cycle is probably not available for shuttling.Instead, it may restrict flux by feedback inhibition in a plant deficient in internal NADH dehydrogenases, causing an aerobic fermentative metabolism.A redox transfer may take place between the glycine decarboxylase and the internal NADH dehydrogenases (dashed arrows) via the suggested intramitochondrial malate/oxaloacetate shuttle (Wiskich et al. 1990).OAA, oxaloacetate; 2-OG, 2-oxoglutarate; Triose-P, triose phosphate; UQ, ubiquinone.
to the growth phenotype by, for example, causing the observed decrease in shikimate, which has been linked to decreased plant growth (Janacek et al. 2009).A specifically elevated NADPH/ NADP + ratio and decreased shikimate level have also been linked to slower growth in A. thaliana with suppression of the external NADPH DH NDB1 (Wallstro ¨m et al. 2013).Some similar effects are observed in this investigation, mainly regarding gene expression for ribosomal proteins and decreases in sugars and organic acids.In contrast, elevated lactate and alanine levels were specific to this investigation, while amino acid decreases, glucose-specific growth inhibition and glucosinolateand jasmonate-associated gene expression were seen in NDB1suppressed plants (Wallstro ¨m et al. 2013) but not in NDA1,2 RNAi plants.Nevertheless, some of the metabolic alterations observed here (Fig. 9) may be due to the change in NADPH/ NADP + observed at growth light in the NDA1,2-suppressed plants (Fig. 6).It is also important to differentiate the influence of reductant ratios, i.e. in NADH/NAD + and NADPH/NADP + , from changes in total NADP(H) and total NAD(H).An increase in total NADP(H) relative to NAD(H), without effects on the redox ratios, was induced by overexpression of a chloroplastic NAD kinase, and effects included an elevation in several amino acids (Takahashi et al. 2009, Takahara et al. 2010).In contrast, in the NDA1,2 RNAi plants, a decrease in total NADP(H) was not linked to corresponding decreases in these same metabolites (cf.Figs. 6 and 9).However, to elucidate fully the relative importance of the NAD(H) and NADP(H) redox status and absolute amounts for growth and metabolism, parallel studies using transgenic plants with separate effects on NAD(H) and NADP(H) levels will be needed.

RNAi constructs and plant transformation
The NotI site of pSL1180 (Amersham Biosciences) was removed by digestion with PstI and KpnI (bp 2,973-3,047), followed by blunting and ligation to give pSL1180ÁNotI.The A. thaliana NDA1 cDNA clone U51324 (from the Arabidopsis Biological Resource Center) was used as PCR template with primers against NDA2: 5 0 -GATATCTGTGGTTGGTGGTGGACCAACT-3 0 and 5 0 -TGATCATCTTTCACAATCCCTCGCACA-3 0 (underlined parts correspond to NDA2 sequences, flanked by cloning adaptors).The amplified sequence was inserted into pCR4-TOPO (Invitrogen), and then into pSL1180ÁNotI, using EcoRV and BclI.The DNA segment was then transferred into pHANNIBAL (CSIRO Plant Industry) (Wesley et al. 2001), using XbaI and BamHI (for antisense insertion) and XhoI and EcoRI (for sense insertion).The regions from pSL1180ÁNotI containing the amplified NDA1,2 segment (from XhoI to EcoRI and XbaI to BamHI) including 100 bp upstream and downstream of the sequences in pHANNIBAL were used in BLAST searches.There were no matches constituting !21 continuous bases in mRNA sequences from other genes in A. thaliana, suggesting that offtarget effects are highly unlikely.
The RNAi cassette from pHANNIBAL was excised with NotI and inserted into the SmaI site of pCAMBIA3300 (Cambia).Sequencing confirmed an unchanged gene-specific segment.Cloning of pCR4-TOPO containing the NDA1,2 segment and pSL1180ÁNotI for cleavage with BclI were performed in the dam -E. coli strain GM3819 (Parker and Marinus 1988).All other cloning was done in the E. coli DH5a strain (Invitrogen).Agrobacterium tumefaciens strain LBA4404 (Hellens et al. 2000) was transformed with pCAMBIA3300, containing the NDA1,2 RNAi cassette, for subsequent floral dip transformation (Clough and Bent 1998) of A. thaliana Col-0.

Plant material and growth conditions
For all experiments, sown seeds were stratified at 4 C for 2 d.Real-time RT-PCR screening for NDA1,2-suppressed lines was performed with plants grown in soil, with 16 h light, 55 mmol m À2 s À1 and 22 C. Unless denoted otherwise, soilgrown plants for other experiments were grown on a 1 : 1 : 2 ratio of perlite : vermiculite : soil with 10 h light, 80 mmol m À2 s À1 (growth light) and 25 C. Starting the second week, the plants were supplied with 0.5Â Hoagland solution (Epstein 1972) once weekly.For high light treatments, plants were moved to 800 mmol m À2 s À1 and 25 C for 2 h before measurements and sampling.Control plants remained in the growth light for the same time period.For extractions, the treatment started 2 h into the light period.For activity measurements, the treatment was started at a staggered schedule, rotating between the replicates.Dry weights were determined by drying aboveground tissue at 75 C for !72 h.For mitochondrial isolation, plants were grown as previously described (Keech et al. 2005).For sterile culturing, seeds were sterilized for 1 min in 70% ethanol, 5 min in 50% bleach and 0.5% Tween-20, rinsed three times in sterile water and suspended in 0.1% agarose.Seeds were sown on plates, stratified and kept in 16 h light, 65 mmol m À2 s À1 and 24 C.For growth in darkness (20 C), germination was induced with 65 mmol m À2 s À1 light and 24 C for 1 h.Plant growth medium (Somerville and Ogren 1982) with 0.5Â macronutrients and 1Â micronutrients supplied with 0.8% (w/v) plant agar and 58.4 mM of either sucrose, glucose or mannitol was used.For inhibitor treatments, 50 mM n-PG, 2 mM antimycin A, 5 nM isoxaben and 40 mM rotenone was used.NDA-suppressed and WT seedlings were grown in separate sections on the same plates.
For transcript profiling, isolated total RNA was analyzed with ATH1 microarrays at the Nottingham Arabidopsis Stock Centre, where the data set is deposited under the reference number 566.Array data were normalized using robust multiarray average (Irizarry et al. 2003) in the program affylmGUI in the Bioconductor R package (Smyth 2004), where MAS5 present and absent calls were also calculated.P-values were calculated and adjusted for multiple testing (Benjamini and Hochberg 1995).MapMan version 10.0 (Thimm et al. 2004) was used for functional analysis of gene response profiles in the transgenic lines, after deletion of probes with absent calls.Significant differences for bins were determined with the Wilcoxon rank sum test, using the Benjamini-Hochberg correction of P-values.Similar results were observed for expressional changes in the individual transgenic lines and in the pooled data for both transgenic lines.

Mitochondrial preparations and Western blotting
Mitochondria were purified from approximately 30 g of A. thaliana rosettes as described previously (Keech et al. 2005), with the following modifications: (i) extracts were filtered through 50 mm nylon mesh; (ii) a 28/45% Percoll gradient was used in a gradient buffer containing 0.3 M sucrose, 10 mM TES, 1 mM EDTA, 10 mM KH 2 PO 4 , 1 mM glycine and 0.1% (w/v) defatted bovine serum albumin (BSA), pH 7.5; and (iii) the resuspension medium for purified mitochondria was supplemented with 5% (v/v) dimethylsulfoxide (DMSO).Protein and Chl contents were measured by the bicinchoninic acid protein assay (Sigma-Aldrich) and acetone extraction (Arnon 1949), respectively.

Measurements of respiration, oxygen isotope fractionation and photosynthetic parameters
Leaves were placed in the dark for 30 min to avoid lightenhanced dark respiration.Total respiration and oxygen isotope fractionation were measured as previously described (Florez-Sarasa et al. 2007), lasting approximately 60 min.Electron partitioning between the two respiratory pathways was calculated from the oxygen isotope fractionation by complex IV (D c ) and AOX (D a ).The latter was determined in the presence of 10 mM KCN in WT plants (three replicates) giving a value of 30.4 ± 0.2%.The value for D c used for calculations was 20.9%, as previously obtained for A. thaliana leaves (Florez-Sarasa et al. 2007).Leaf area was determined with an AM 300 leaf area meter (ADC Bioscientific Ltd.) after the respiration measurements.The capacity of the AOX pathway was determined as previously described (Florez-Sarasa et al. 2009).
Leaf gas exchange analyses were performed at steady state, after approximately 20 min at saturating light (1,000 mmol m À2 s -1 ), but otherwise as previously described (Flexas et al. 2007).The maximum quantum efficiency of PSII (F v /F m ) was obtained from Chl fluorescence measurements, as previously described (Galle ´et al. 2007) except for using a 30 min dark adaptation period.

Fig. 1
Fig. 1 Real-time RT-PCR analysis of RNAi lines.Levels of transcripts encoding NDA1, NDA2, the 28.5 kDa subunit of respiratory complex I and MZA15.2 (At5g46630) are shown as means ± SE for three independent RNA preparations.Asterisks denote significant differences from the WT.Each RNA extraction was made using leaves from two 20-day-old, soil-grown plants of the T 3 generation.

Fig. 2
Fig.2NDA protein analysis.Leaf mitochondria were purified from T 4 rosettes of lines 17.11, 26.6 and the WT, and analyzed by Western blotting using antibodies against A. thaliana NDA1 and NDA2.Antibodies against the wheat NAD9 subunit of complex I and watermelon glyoxysomal malate DH (MDH) were used as controls.Protein sizes are given in kDa to the left of the blot, and antibodies are specified to the right.

Fig. 3
Fig. 3 Growth of NDA1,2-suppressed lines on sterile culture medium.The graphs show rosette diameters of seedlings grown in light for 10 d (A) and hypocotyl lengths of seedlings grown in darkness for 7 d (B), with the supplements denoted above each graph.Means ± SE for 31-40 T 4 seedlings are shown.Asterisks denote significant differences from the WT.

Fig. 4
Fig. 4 Effects of inhibitors on the rosette diameter of different RNAi lines.For each genotype, the growth inhibition is shown as determined by comparison with the same genotype on control plates (sucrose and control solvent, in light).All seedlings were 10 d old.Means ± SE for 35-49 T 4 seedlings are shown.Significant differences between the responses in WT and individual transgenic lines are denoted with an asterisk.

Fig. 5
Fig. 5 Growth of NDA1,2-suppressed lines in soil.Visual appearance of representative plants (A) and rosette dry weight (B) of T 4 generation plants grown for 30 d with a 12 h photoperiod are shown.(C) Growth analysis of T 4 generation plants grown with a 10 h photoperiod.Means are shown ± SE for 8-10 samples, each containing 1-3 rosettes.The scale bar corresponds to 5 cm.Asterisks denote significant differences from the WT.

Fig. 6
Fig. 6 Nicotinamide nucleotide levels in leaves of NDA1,2-suppressed lines.T 4 plants were grown in soil under growth light for 35 d.Rosettes were harvested 4 h after the start of the light period and analyzed for content of nicotinamide nucleotides.Averages ± SE are shown for 5-7 biological replicates, each containing three rosettes.Asterisks indicate significant differences from the WT.

Fig. 7
Fig. 7 Transcript levels of transgenic lines in growth light and high light.The transcript levels for different alternative respiratory chain components and for the 28.5 kDa subunit of complex I (28.5 kDa) are shown.Real-time RT-PCRs were performed on rosette RNA from soilgrown T 4 plants, 35 d after sowing.The plants either remained at growth light (GL) or were moved to high light (HL) for 2 h before sampling.Transcript levels are means ± SE for three independent RNA preparations, each derived from three rosettes.The WT signal in growth light corresponds to 100%.Different letters denote significant differences as indicated by ANOVA and the Tukey-Kramer method.

Fig. 8
Fig. 8 In vivo activities of respiratory pathways.Total respiration (A), cytochrome pathway activity (B) and alternative pathway activity (C) in the WT and NDA1,2-suppressed lines are denoted relative to the leaf area.Plants were grown and treated as in Fig. 7. Values are means ± SE from five replicates, each consisting of three rosettes.Different letters denote significant differences, as for Fig. 7.

Fig. 9
Fig. 9 Metabolic profiling of rosettes from the WT and the NDA-suppressed lines.Plants of the T 4 generation were grown and treated as in Fig. 7. (A) Heat map showing the high light response in all genotypes.(B and C) Plots of metabolites displaying consistently different levels between the WT and the transgenic lines in growth light (B) and after 2 h of high light (C).The metabolites included showed a significant difference between the WT and the pooled values from both NDA-suppressed lines under the respective light regime (see Supplementary TableS2).The WT level in growth light is set to unity for all data.Error bars denote the SE for 6-7 biological replicates, each containing three rosettes.Cit., citrate; Fru, fructose; Fum., fumarate; Gal, galactose; Glc, glucose; Lac., lactate; Mal., malate; Male., maleate; Shiki., shikimate.

Table 2
The maximum quantum efficiency of PSII photochemistry (F v /F m ) in leaves of WT and NDA1,2-suppressing lines Plants were grown and treated as in Fig.7.Values are means ± SE from 12 leaf replicates (each from a different plant).Asterisks denote significant differences compared with the WT in growth light.

Table 3
MapMan analysis of transcript profile data in NDA1,2-suppressed lines