Structural and functional characterization of ferredoxin-NADP+-oxidoreductase using knock-out mutants of Arabidopsis


(fax +35 823 33 8075; e-mail


In Arabidopsis thaliana, the chloroplast-targeted enzyme ferredoxin-NADP+-oxidoreductase (FNR) exists as two isoforms, AtLFNR1 and AtLFNR2, encoded by the genes At5g66190 and At1g20020, respectively. Both isoforms are evenly distributed between the thylakoids and soluble stroma, and they are separated by two-dimensional electrophoresis in four distinct spots, suggesting post-translational modification of both isoforms. To reveal the functional specificity of AtLFNR1, we have characterized the T-DNA insertion mutants with an interrupted At5g66190 gene. Absence of AtLFNR1 resulted in a reduced size of the rosette with pale green leaves, which was accompanied by a low content of chlorophyll and light-harvesting complex proteins. Also the photosystem I/photosystem II (PSI/PSII) ratio was significantly lower in the mutant, but the PSII activity, measured as the FV/FM ratio, remained nearly unchanged and the excitation pressure of PSII was lower in the mutants than in the wild type. A slow re-reduction rate of P700 measured in the mutant plants suggested that AtLFNR1 is involved in PSI-dependent cyclic electron flow. Impaired function of FNR also resulted in decreased capacity for carbon fixation, whereas nitrogen metabolism was upregulated. In the absence of AtLFNR1, we found AtLFNR2 exclusively in the stroma, suggesting that AtLFNR1 is required for membrane attachment of FNR. Structural modeling supports the formation of a AtLFNR1–AtLFNR2 heterodimer that would mediate the membrane attachment of AtLFNR2. Dimer formation, in turn, might regulate the distribution of electrons between the cyclic and linear electron transfer pathways according to environmental cues.


Ferredoxin-NADP+-oxidoreductase (FNR; EC is a hydrophilic, FAD-containing enzyme that catalyzes electron transfer between NADP(H) and ferre(flavo)doxin. In higher plants, a small multigene family encodes two distinct forms of FNR (Morigasaki et al., 1993). The root form of FNR functions in non-photosynthetic plastids mediating electrons from NADPH, originating from the cytosolic glucose 6-phosphate via the oxidative pentose phosphate pathway, to the root-specific, nitrate-induced form of ferredoxin (Oji et al., 1985). Reduced ferredoxin, in turn, donates electrons for nitrite reduction by nitrite reductase, or for many other ferredoxin-dependent enzymes involved in assimilation of, for example, nitrogen and sulfur (Neuhaus and Emes, 2000).

The leaf form of FNR (LFNR) has a well described role in linear electron transfer in chloroplasts, where electrons are transferred from ferredoxin to NADP+ for carbon fixation (Ceccarelli et al., 2004). Leaf form FNR has also been implicated in cyclic electron transfer around photosystem I (PSI), which generates a proton gradient across the thylakoid membrane, resulting in ATP production without accumulation of reducing equivalents (Johnson, 2005). In cyclic electron transfer, electrons can be recycled from reduced ferredoxin or NADPH either directly to plastoquinone (Clark et al., 1984; Zhang et al., 2001), or alternatively via the type I NAD(P)H dehydrogenase (NDH-1) complex to plastoquinone and subsequently to the cytochrome (cyt)-b6f complex. The latter route is known to be involved in dark reduction of the plastoquinone pool (Burrows et al., 1998; Joet et al., 2001; Kofer et al., 1998; Shikanai et al., 1998). The change in pH (ΔpH) generated by cyclic electron transfer is important for production of ATP, but also for the induction of thermal dissipation of absorbed energy from PSII antennae (Johnson, 2005). Cyclic electron transfer has recently been shown to be indispensable for the optimal photosynthetic performance of plants (Munekage et al., 2003), although the detailed pathways have remained elusive. Furthermore, LFNR was recently shown to participate in import of proteins into the chloroplast by serving as a possible redox sensor (Kuchler et al., 2002).

It has been shown previously that LFNR is located in two distinct chloroplast compartments, the stroma and the thylakoids. Despite numerous studies, the exact site of thylakoid attachment has remained obscure, although some studies have implicated LFNR in the NDH-1 complex (Guedeney et al., 1996; Quiles et al., 2000), in the cyt-b6f complex (Okutani et al., 2005; Zhang et al., 2001), and also in PSI (Andersen et al., 1992). In Arabidopsis thaliana, the genes At5g66190 and At1g20020 encode the two paralogous forms of LFNR, AtLFNR1 and AtLFNR2, respectively. Hanke et al. (2005) have shown that although both LFNR isoforms are found in the thylakoids and in the stroma, AtLFNR1 is more abundant in the thylakoid fraction. Arabidopsis LFNR1 shows slightly stronger affinity to all ferredoxin isoforms than AtLFNR2. Importantly, the two genes encoding LFNR isoforms have a unique expression pattern in response to different nitrogen growth regimes (Hanke et al., 2004), but the functional specificity of the isoforms still remains elusive. In maize (Zea mays) three LFNR isoforms have been characterized: ZmLFNR1 is found only from the thylakoid membrane associated with the cyt-b6f complex, ZmLFNR3 is an exclusively soluble enzyme and ZmLFNR2 has a dual location (Okutani et al., 2005).

In the present study, we have characterized the knock-out mutant line (ΔFNR1) of the At5g66190 gene encoding AtLFNR1 in Arabidopsis. The homozygote knock-out line of AtLFNR1 is viable, although with reduced capacity for carbon fixation and consequently with reduced biomass accumulation, indicating that At1g20020 can only partially replace At5g66190. Besides marked differences in photosynthetic properties and in the accumulation of thylakoid membrane proteins between the mutant and wild-type (WT) plants, it was also found that the absence of AtLFNR1 prevented the association of AtLFNR2 with the thylakoid membrane, which together with structural modeling suggests the formation of LFNR heterodimers. Contrary to carbon metabolism, nitrogen metabolism was enhanced in ΔFNR1, as evidenced by increased accumulation of nitrate reductase transcripts and high activity of the nitrate reductase enzyme in the leaves of the mutant plants.


Absence of AtLFNR1 resulted in reduced rosette size and pale phenotype

Interruption of the At5g66190 gene encoding the AtLFNR1 protein in the lines SALK_085403 (ΔFNR1) and SALK_067668 (ΔFNR1b) (Figure 1a) resulted in reduced rosette size and pale green leaves (Figure 1b–d). Since the phenotype of ΔFNR1b was identical to the phenotype of ΔFNR1 in all respects, the experiments for further characterization of AtLFNR1 knock-out mutants were performed with the ΔFNR1 line. Most experiments were repeated with the line ΔFNR1b, which gave practically identical results to those from ΔFNR1, and are not separately presented.

Figure 1.

 Genomic constructs and phenotypic characterization of Arabidopsis WT and ΔFNR1 plants.
(a) Site of T-DNA inserts in Arabidopsis thaliana lines Salk_085403 (ΔFNR1) and Salk_067668 (ΔFNR1b). The At5g66190 gene encodes the AtLFNR1 isoform of LFNR. White boxes denote exons, black lines introns and dotted lines 5′ and 3′ untranslated regions (not in scale). The site of a T-DNA insert is presented as triangle.
(b) Phenotype of WT and ΔFNR1 plants grown under standard conditions for 5 weeks.
(c) Chlorophyll a/b ratio and chlorophyll content in WT and ΔFNR1 plants. Values are means from nine independent measurements from three distinct plants grown under standard conditions for 5 weeks.
(d) Dry weight (DW) of the Arabidopsis WT and ΔFNR1 rosettes. The DWs of the rosettes of the plants grown for 4, 5, 6, 7 or 8 weeks under standard conditions were determined. Values are means from 12 measurements. Bars denote the SD.

In accordance with the pale green color of the mutant leaves, a remarkably low chlorophyll content per leaf area was measured in ΔFNR1 (Figure 1c). The chlorophyll a/b ratio, on the other hand, was similar in WT and ΔFNR1 (Figure 1c). Because the visual phenotype and chlorophyll content of ΔFNR1 plants differed markedly from those of WT, we further analyzed the chloroplast ultrastructure by electron microscopy. However, no differences were recorded in the structure or the thylakoid stacking pattern between ΔFNR1 and WT chloroplasts (data not shown).

Protein content of the thylakoid membrane

An overall analysis of thylakoid protein content by one-dimensional (1-D) SDS-PAGE followed by Western blotting revealed a lower accumulation of the PsaD subunit (67%) of PSI, the cyt f subunit (69%) of the cyt b6f complex, the Lhca2 protein (72%) of light-harvesting complex I (LHCI) as well as the Lhcb1 protein (53%) of LHCII (Figure 2a) in the thylakoids of ΔFNR1 when compared with that in WT (Figure 2a). The low content of LHCII complexes was also clear from the Coomassie blue stained gels (Figure 2b). Likewise, the level of the D1 (80%) and D2 (70%) proteins of PSII in ΔFNR1 was somewhat reduced (Figure 2a), whereas the amount of the α subunit of ATP synthase (CF1; 102%) did not differ between the mutant and WT (Figure 2a). The thylakoid membranes of ΔFNR1 and WT were further analyzed by low-temperature fluorescence emission spectra measurements. It is well known that the fluorescence emission peak detected at 731 nm (F731) arises mainly from the chlorophylls associated with PSI, and at 685 nm (F685) from the chlorophylls associated with CP43 of the PSII complex (Krause and Weis, 1991). The fluorescence emission ratio F731/F685 was 1.26 in WT, whereas in the ΔFNR1 plants the ratio was only 1.11 (Table 1), confirming that the PSI/PSII ratio in the mutants was indeed lower than in the WT. Additionally, analysis of the thylakoid protein complexes by Blue Native gel electrophoresis showed a low level of PSI, PSII and LHC complexes in the thylakoid membranes of ΔFNR1 (Figure S1).

Figure 2.

 Protein contents of WT and ΔFNR1 leaf extracts, thylakoids and the soluble fraction.
(a) Thylakoid protein content. After electrophoresis, the proteins were electroblotted on a PVDF membrane and probed with the PsaD, D1, D2, cyt f, Lhcb1, Lhca2 and ATP synthase coupling factor (CFI) α subunit antisera. Ten micrograms of protein were loaded in each well.
(b) Coomassie-stained gel of WT and ΔFNR1 leaf extracts. Bands corresponding to Rubisco and LHCII are depicted. Ten microliters of total leaf extract was loaded in the wells.
(c) Protein content of the thylakoid, soluble and total leaf extract fractions in WT and ΔFNR1. Proteins from the thylakoid and soluble fractions as well as from total leaf extracts were separated by SDS-PAGE, electroblotted on a PVDF membrane and immunolabeled with FNR antibody. The purity of the fractions is demonstrated using Lhcb1 and Rubisco activase (RuA) antibodies. Five micrograms of thylakoid proteins, 15 μg of soluble proteins and 10 μl of total leaf extract were loaded in each well.

Table 1.   Photosynthetic properties of Arabidopsis WT and ΔLFNR1 plants
Photosynthetic capacityWTΔLFNR1
  1. The plants used for measurements were grown under standard growth conditions. Values are means from three to 11 independent measurements from distinct plants, ± denotes for SD. See text for details.

Maximal quantum yield of PSII (FV/FM)0.8090.813
Excitation pressure of PSII (1−qP)0.117 ± 0.020.106 ± 0.01
CO2 fixation (μmol CO2 m−2 sec−1)2.28 ± 0.441.60 ± 0.35
PSI/PSII ratio (F731/F685)1.26 ± 0.031.11 ± 0.05
Size of the electron pool in the intersystem chain (e/P700)17.7 ± 0.6714.4 ± 1.11

Membrane attachment and post-translational modification of LFNR isoforms

Probing of thylakoid proteins separated by 1-D SDS-PAGE with anti-FNR revealed several bands migrating very close to each other in a region of 35 kDa (Figure 2c). In WT, the band pattern corresponding to FNR was similar in the thylakoid and the soluble fractions, suggesting that both LFNR forms are present in both compartments, as was shown previously (Hanke et al., 2005). After two-dimensional (2-D) electrophoresis, four distinct LFNR spots could be detected by silver staining (Figure 3a) and by FNR antibody (data not shown). Two of the spots were identified by matrix-assisted laser desorption/ionization (MALDI) MS analysis as the AtLFNR1 proteins and two as the AtLFNR2 proteins (Figure 3b). In line with the results of Hanke et al. (2005), the isoelectric point of AtLFNR1 was determined to be around 5, whereas AtLFNR2 was significantly more basic. Since both gene products were present in two distinct spots, it seems plausible that both LFNR forms are targets of post-translational modification.

Figure 3.

 Identification of the AtLFNR1 and AtLFNR2 proteins.
(a) Thylakoid membranes were solubilized and proteins separated by isoelectric focusing followed by SDS-PAGE. The silver-stained gel shows the location of the four LFNR spots.
(b) The MALDI-TOF MS analysis of the AtLFNR1 and AtLFNR2 isoforms. The two spots of each isoform gave an identical spectrum, and representative spectra of both isoforms are shown. Dark arrows show the peaks that belong to the peptides of the corresponding protein. Light arrows show peaks of trypsin self-digest used for spectra calibration. m/z is the mass-to-charge ratio.

In contrast to WT thylakoids, no FNR bands could be detected in the thylakoid membranes of ΔFNR1 (Figure 2c) independently of the growth conditions (growth light, high light, long day; see Experimental procedures) of the plants (data not shown). However, when the soluble protein fractions were analyzed, the presence of AtLFNR2 protein could be detected as two separate bands after 1-D SDS-PAGE (Figure 2c), and as two spots after 2-D gel electrophoresis and Western blotting using FNR antibody (data not shown). Thus, in spite of an even distribution of LFNR between the two chloroplast compartments in WT, in the absence of AtLFNR1 the AtLFNR2 protein is dominantly located in the chloroplast stroma. Furthermore, the analysis of the ΔFNR1 leaf extract clearly showed that the total level of LFNR was dramatically decreased in the mutant as compared with the WT (Figure 2c).

Photosynthetic properties

No significant differences in PSII performance could be detected between ΔFNR1 and WT, as deduced from a ratio of the variable to maximum fluorescence (FV/FM) in intact leaves (Table 1). To get an insight into the redox state of the plastoquinone pool, the PSII excitation pressure (1−qP) was monitored. Surprisingly, the excitation pressure of PSII was lower in ΔFNR1 plants than in WT (Table 1). To estimate the plastoquinone pool size of the ΔFNR1 and WT plants the changes in the oxidation curve of P700 upon applying single-turnover (ST) and multiple-turnover (MT) pulses were further analyzed; the measurements are depicted in Figure 4(a). The primary donor of PSI, P700 was oxidized upon illumination with far-red light, and rapidly reached a steady-state level of P700+. On applying a ST saturating flash, P700+ was transiently reduced with a single electron coming from the PSII complex, and became oxidized again by background far-red light. Next, the triggering of saturating MT pulses reduced transiently the whole plastoquinone pool and subsequently also all P700+. Continuous background far-red light returned the system again to the steady-state level of oxidized P700+. According to the results presented in Figure 4(a) and in Table 1, the plastoquinone pool size of the ΔFNR1 plants was smaller than that of the WT plants. Furthermore, the maximum amount of photo-oxidizable P700, measured by excitation of the samples with a saturating pulse of white light, applied on top of the far-red light, was lower in ΔFNR1 (Figure 4a).

Figure 4.

 Redox changes of P700 in WT and ΔFNR1 leaves.
(a) Photo-oxidation and subsequent re-reduction of P700 upon application of single-turnover (ST) and multiple-turnover (MT) pulses on continuous background illumination of far-red (FR) light.
(b) Re-reduction rate of P700 in the dark. P700 was oxidized by far-red (FR) light for 30 sec and after termination of FR illumination, P700+ re-reduction was monitored in the dark. Curves were normalized to the maximal signal. a.u. denotes arbitrary units.

Kinetics of the re-reduction of P700 oxidized by far-red light in darkness was taken as a measure for the capacity of PSI-driven cyclic electron transport. Dark reduction of P700+ was significantly slower in ΔFNR1 than in WT (Figure 4b), suggesting poor electron donation from the intersystem chain and thus a decreased level of cyclic electron flow around PSI. Furthermore, we measured the rate of carbon fixation of the ΔFNR1 and WT plants. The carbon assimilation rate of ΔFNR1 plants was severely impaired, being only 82% of that of WT under growth light conditions (Table 1). Higher carbon fixation activity was characteristic of all plants either grown or shifted for 1–3 h to a higher incident irradiance (300–600 μmol photons m−2 sec−1) than in plants under normal growth conditions, but nevertheless the ΔFNR1 plants always had a lower CO2 assimilation rate than WT (data not shown).

Redox state of stromal components

The impaired photosynthetic activity detected in ΔFNR1 plants (Table 1) prompted us to assay the redox state of the stroma. This was done indirectly by measuring the activity of NADP-malate dehydrogenase (MDH), which is known to be related to the reduction state of stromal thioredoxins (Miginiac-Maslow and Lancelin, 2002; Scheibe, 2004). The NADP-MDH activity measurements indicated that under standard growth conditions the activity was 39% lower in ΔFNR1 than in WT (Figure 5a). Upon shifting to intense illumination (2 h under 400 μmol photons m−2 sec−1) the activity increased both in the WT and in the ΔFNR1 plants, but in ΔFNR1 the activity remained much lower than in WT (Figure 5a). Since the redox state of the intersystem chain and stromal compounds is known to regulate the activity of several chloroplast protein kinases (Aro and Ohad, 2003), we also studied the level of phosphorylation of thylakoid proteins in ΔFNR1 and WT plants. Under growth conditions, neither PSII core phosphoproteins nor LHCII showed distinct differences in the levels of phosphorylation between WT and ΔFNR1 (Figure 5b). In contrast, when the plants grown under standard conditions were shifted to higher irradiances (2 h under 400 μmol photons m−2 sec−1) for short periods, the LHCII kinase in WT was strongly inhibited and LHCII quickly became dephosphorylated, whereas the LHCII of ΔFNR1 remained more phosphorylated compared with WT (Figure 5b). Under the same conditions, no differences could be detected in the phosphorylation of the PSII core proteins between ΔFNR1 and WT (Figure 5b).

Figure 5.

 Redox state of stromal components.
(a) The NADP-malate dehydrogenase (NADP-MDH) activity measured in Arabidopsis WT and ΔFNR1 plants.
(b) Immunoanalysis of thylakoid protein phosphorylation in Arabidopsis WT and ΔFNR1 plants. In vivo phosphorylation of thylakoid proteins was demonstrated by immunoblotting with phosphothreonine antibody. Chlorophyll (0.75 μg) was loaded into each well and Western blotting was performed. For (a) and (b) plants were harvested after 5 weeks’ growth either directly from the standard conditions (growth light; GL) or after an extra illumination of 2 h at 400 μmol photons m−2 sec−1 (high light; HL). a.u. denotes arbitrary units.

Transcript profiles and nitrogen metabolism

A DNA microarray analysis was performed in order to clarify whether the changes in the photosynthetic capacity and stromal redox state of ΔFNR1 (Table 1, Figures 4 and 5) could act as a signal for nuclear gene expression. Transcript profiling of the mutant and WT, however, did not reveal any drastic changes in global gene expression between ΔFNR1 and WT under growth light conditions (Figure 6). In addition to At1g01350, coding for an unknown protein, only the interrupted At5g66190 gene, encoding AtLFNR1, showed clear downregulation. The At5g45410 gene, coding for a zinc finger protein, and At1g77760 and At1g37130, both coding for nitrate reductase enzyme, were the only genes showing slight upregulation. Consequently, we studied whether the increased accumulation of nitrate reductase transcripts is reflected in later steps of nitrogen metabolism by measuring the activity of nitrate reductase enzyme as an accumulation of nitrite. Under growth light conditions, ΔFNR1 plants accumulated 0.53 μmol NOinline image g−1 h−1, whereas WT plants were able to accumulate only 0.30 μmol NOinline image g−1 h−1.

Figure 6.

 Comparison of gene expression between WT and ΔFNR1 plants. Scatter plot of logarithmic signal intensities of the plants grown for 5 weeks under standard conditions.

Structural modeling

Structural models of AtLFNR1 and AtLFNR2 in complex with ferredoxin were made to study the oligomeric structure of AtLFNRs in complexes and to find out the specific structural features of these two isoforms. Leaf form FNRs from several species have been subjected to structural studies and the crystal structures of Anabaena FNR and ZmLFNR1 have been solved in both the free state [Protein Data Bank (PDB) code 1QUE and 1GAW, respectively] and in complex with ferredoxin (1EWY and 1GAQ). Both of these complex structures consist of two FNRs and one ferredoxin molecule (Kurisu et al., 2001; Morales et al., 2000; Serre et al., 1996). In the Anabaena FNR–ferredoxin complex, the FNRs interact with each other and ferredoxin, but only one of the FNRs would participate in the catalytically competent electron-transfer complex cluster. Kurisu et al. (2001) reported that ZmLFNR1 binds ferredoxin-1 as a monomer. There is, however, a second FNR molecule in the asymmetric unit, which is similar to the ‘free-state’ FNR molecule and has no contact with ferredoxin. Based on our structural analysis, another ‘free-state’ FNR molecule actually associates with the FNR–ferredoxin complex when the symmetry mates for the crystal structure are generated. Thus, ZmLFNR1, like Anabaena FNR, forms a dimer in the presence of ferredoxin. The result indicates that FNR dimers appear to exist only in the presence of ferredoxin, since dimers are not seen in the FNR structures without ferredoxin (data not shown).

Structural modeling of the Arabidopsis LFNRs (AtLFNR homodimers and the AtLFNR1–AtLFNR2 heterodimer) was performed using the constructed ZmLFNR dimer in complex with ferredoxin as template since the AtLFNRs share a higher sequence identity (∼75%) with ZmLFNRs than with the Anabaena FNRs (∼45%) (Figure S2). In fact, only six residues are different in the Arabidopsis and maize FNR–ferredoxin interfaces (Figure 7). In addition, there are only two residues that are not conserved in the FNR dimer interface in these species (Figure 7). In ZmLFNR1 Ala165 is substituted by Gly170 and Gln179 in AtLFNR1 and AtLFNR2, respectively (Figure S2). However, this should not hinder dimer formation, since the main chain oxygen of the substituted residue, not the side chain, interacts with the other subunit. The other non-conserved residue in the dimer interface, Arg285 in ZmLFNR, corresponds to Lys290 in AtLFNR1 and Lys299 in AtLFNR2. Thus, the electrostatic interaction (formed by Arg285 and Glu167 in ZmLFNR1) at the dimer interface is conserved in the AtLFNRs. These minor differences do not prevent the formation of similar AtLFNR dimers as seen in the Anabaena and maize FNR–ferredoxin complexes and, furthermore, there are no structural hindrances for the formation of an AtLFNR heterodimer.

Figure 7.

 Structural model of AtLFNr1-AtLFNR2 in complex with ferredoxin.
AtLFNR1 is colored green, AtLFNR2 is colored blue and ferredoxin is colored yellow. The FAD cofactors are shown as pink or purple sticks and the iron–sulfur cluster of ferredoxin as yellow spheres. Residues that differ in the AtFNR1–ferredoxin and ZmLFNR1 interface are shown as yellow sticks, while the two residues in the FNR dimer interface that are not conserved are shown as blue or green spheres. The predicted phosphorylation sites located at the surface of the molecules (Ser66, Tyr238, Thr284 and Ser352 in AtLFNR1 and Ser75, Tyr93, Thr182 and Tyr247 in AtLFNR2) are shown as dark red spheres.


Leaf FNR isoforms are post-translationally modified

In WT Arabidopsis, the two LFNR forms (AtLFNR1 and AtLFNR2) were found in both the thylakoid membrane and the soluble fraction (Figure 2), corroborating the results of Hanke et al. (2005). Furthermore, it is very likely that both forms are subject to post-translational modification that changes the isoelectric point of the protein (Figure 3). Indeed, the separation of the LFNR forms into several bands was also observed by 1-D SDS-PAGE (Figure 2). Previously it was concluded by an in vitro approach that in pea chloroplasts a threonine and/or a serine residue of LFNR is phosphorylated in a light-dependent reaction, whereas the dark phosphorylation of LFNR may occur only on a serine residue (Hodges et al., 1990). Proofs of in vivo phosphorylation of LFNR are, however, missing, and our attempts to find phosphorylated LFNR in Arabidopsis thylakoids or in the soluble fraction by immunolabeling with phosphoantibodies (anti-P-Ser, anti-P-Thr and anti-P-Tyr) remained futile. Nevertheless, the structural modeling of Arabidopsis LFNR isoforms revealed several potential phosphorylation sites on the surface of the proteins (Figure 7), still leaving room for possible in vivo phosphorylation of LFNR.

Arabidopsis LFNR1 is specifically required for the attachment of LFNR to the thylakoid membrane

Our results clearly show that the AtLFNR2 protein cannot bind to the thylakoid membrane if AtLFNR1 is missing (Figure 2). Thus, it is conceivable that AtLFNR1 and AtLFNR2 form a loose heterodimer which binds to the thylakoid membrane. Membrane attachment of AtLFNR2 might require the formation of a heterodimer before binding, or alternatively AtLFNR1 might act as a membrane anchor. Based on structural modeling, Arabidopsis LFNRs are capable of forming hetero- or homodimers in complex with ferredoxin (Figure 7). Structural evidence for the formation of FNR dimers has previously been observed for Anabaena FNR in complex with ferredoxin; additionally, we have now discovered that maize LFNR1 appears to form similar dimers as well. In line with this result, several other studies have also provided experimental evidence for LFNR dimer formation in different plant species (Andersen et al., 1992; Fredricks and Gehl, 1976; Schneeman and Krogmann, 1975; Shin, 2004; Zanetti and Arosio, 1980), but the dimers tend to dissociate during a standard purification procedure (Zanetti and Arosio, 1980).

All amino acid residues that contribute to FNR–ferredoxin intermolecular salt bridges (Kurisu et al., 2001) are conserved among different maize LFNR isoforms. Yet it has been suggested that there may be redox state-dependent changes in the mode of formation of the complex between FNR and ferredoxin (Okutani et al., 2005). When stromal pH increases in actively photosynthesizing chloroplasts, ZmLFNR1 forms a stronger complex with ferredoxin than ZmLFNR2 or ZmLFNR3. Thus, the dimer formation and membrane attachment of LFNR might be a feasible way to regulate the distribution of electrons between the cyclic and linear electron transfer pathways according to environmental cues. This suggestion is in line with the results of van Thor et al. (2000), which showed that in cyanobacteria the redistribution of FNR between the thylakoid and phycobilisome fractions provides a control mechanism for sharing electrons between the linear and cyclic electron flow.

Arabidopsis LFNR1 is essential for efficient photosynthetic performance

Photosynthetic characterization of ΔFNR1 plants clearly demonstrated that AtLFNR2 cannot efficiently replace AtLFNR1. Impaired transfer of electrons from ferredoxin to NADP+ and/or the overall downregulation of the photosynthetic machinery in ΔFNR1 plants resulted in decreased efficiency of carbon fixation, and subsequently in decreased accumulation of biomass (Figures 1 and 2, Table 1). The content of the light-harvesting complexes, both LHCII and LHCI, were significantly decreased in ΔFNR1 (Figure 2), as well as the chlorophyll content (Figure 1) and the PSII and PSI proteins (although to different extents; Figure 2, Table 1). Thus, it seems that the entire photosynthetic apparatus is downregulated in ΔFNR1 to decrease the amount of light energy absorbed, and thus to protect the chloroplasts from oxidative damage.

Electron carriers of the intersystem chain between PSII and PSI complexes were less reduced in ΔFNR1 than in the WT, as evidenced by a lowered excitation pressure of the PSII complex. This seems to be in contradiction with the facts that the activity of the PSII complex did not differ from that of the WT (Table 1) and that the PSI/PSII ratio in the mutant plants was found to be decreased (Figures S1 and S2, Table 1). The low excitation pressure detected in ΔFNR1 could be explained by the small size and slower reduction of the intersystem pool (Table 1). Yet we do not exclude the possibility that alternative electron transport routes, like plastid terminal oxidase, play a role as alternative electron sinks in the thylakoid membrane (Kuntz, 2004). Furthermore, the decay of P700+ in the dark is significantly slowed down in AtLFNR1 as compared with WT (Figure 4), pointing to an important role for FNR in the cyclic electron path. Alternatively, it is also conceivable that the more oxidized stromal redox poise in ΔFNR1 may directly affect the rate of cyclic electron transfer around PSI via the NDH-1 complex.

A bleached phenotype and reduced capacity for carbon fixation together with reduced rosette size in ΔFNR1 are strikingly similar to those of antisense FNR tobacco plants (Hajirezaei et al., 2002). There are, however, also distinct differences between ΔFNR1 and tobacco mutants, for example the excitation pressure of PSII in the transgenic tobacco plants was found to be higher than in the WT (Hajirezaei et al., 2002). It is not currently known how many genes code for LFNR in tobacco, but it seems probable that the LFNR antisense constructs (Hajirezaei et al., 2002) resulted in silencing of all tobacco LFNR genes. Similarly, overexpression of LFNR genes in Arabidopsis probably silenced both LFNR genes, leading to a pale green phenotype and small rosette size (LeClere and Bartel, 2001). Importantly, since only At5g66190 is inactivated in ΔFNR1, the differences from FNR antisense plants in intersystem redox state or in the contents of PSI and LHC proteins are probably, at least partly, due to the different roles of the two LFNR isoforms in cyclic (and linear) electron transfer.

Upon inactivation of At5g66190, resulting in marked changes in the photosynthetic capacity of the mutant plants, it is conceivable that the reducing side of PSI is also modulated. Hence, we took a closer look into the stroma by measuring the activity of NADP-MDH enzyme. The NADP-MDH activity was clearly lower in ΔFNR1 than in WT (Figure 5), implying an oxidized stromal environment, particularly that of thioredoxins (Scheibe, 2004). Imbalance in the thiol redox equilibrium in ΔFNR1 was further supported by studying light intensity-dependent phosphorylation of LHCII proteins. Figure 5 shows that under growth light conditions the LHCII proteins of ΔFNR1 and WT were fully phosphorylated, while under high-light conditions the level of phosphorylation of LHCII remained higher in ΔFNR1, reflecting a smaller accumulation of thiol reductants in ΔFNR1 than in WT (Rintamaki et al., 2000).

ΔFNR1 invests in nitrogen metabolism

To check whether the alterations in physiological properties of ΔLFNR1 are reflected at the level of gene expression, we took a transcript profiling approach on ΔFNR1 and WT plants from standard growth conditions. Analysis of the expression of 8000 different genes did not reveal striking differences between ΔFNR1 and WT. As expected, expression of the inactivated At5g66190 gene was not observed in ΔFNR1, but two genes, At1g77760 and At1g37130, encoding a cytosolic nitrate reductase enzyme, were found to be upregulated in ΔFNR1 (Figure 6). The increased transcript levels of these latter genes were reflected in enhanced activity of the nitrate reductase enzyme as well, and significantly more nitrite was accumulated in ΔFNR1 than in WT. The enhanced activity of nitrate reductase in ΔFNR1 might aim at compensating for the decreased level of NADPH, which is used as an electron donor for nitrate reductase. Alternatively, the increased accumulation of nitrite in ΔFNR1 may result from oxidized stromal redox poise, which might directly hamper the reduction of nitrite to ammonium.

Although several subunits of thylakoid protein complexes accumulated in lower amounts in ΔFNR1 than in WT (Figure 2), no differences in the corresponding nuclear transcript levels could be detected between ΔFNR1 and WT (Figure 6), suggesting changes in translation, assembly and/or degradation of the thylakoid protein complexes in the mutant plants.


In the present study, we have shown that (i) Arabidopsis AtLFNR2 can only partially compensate for the function of AtLFNR1, (ii) AtLFNR1 is involved in linear electron transfer and thereby affects the function of the Calvin cycle, and (iii) that the absence of AtLFNR1 results in reduced accumulation of biomass, which is accompanied by low contents of PSII, PSI and LHC proteins as well as chlorophyll. Additionally, we present evidence that (iv) AtLFNR1 is apparently also involved in cyclic electron transfer around PSI and that (v) AtLFNR1 is a prerequisite for the attachment of AtLFNR2 to the thylakoid membrane, suggesting (vi) a formation of LFNR dimers, which is further supported by structural modeling of the two LFNR isoforms.

Experimental procedures

Plant material

Arabidopsis thaliana ecotype Columbia and T-DNA insertion mutants SALK_085403 (annotated as a homozygote line in SALK) and SALK_067668, with an interrupted At5g66190 gene (Figure 1), were obtained from the SALK collection (Alonso et al., 2003). Gene At5g66190 codes for the AtLFNR1 isoform of FNR. Homozygous plants used for experiments were screened by standard PCR protocols recommended by SIGnAL (SALK) by using primers specific to the 3′ and 5′ flanking sequences of the studied gene together with a T-DNA-specific primer. The plants were routinely grown in a phytotron under a photon flux density of 120 μmol photons m−2 sec−1 in 8-h light/16-h dark cycles at 23°C (growth light). Fully grown mature leaves of 5-week-old plants were used in all experiments. In cases where standard conditions were not used plants were grown under a photon flux density of 300 μmol photons m−2 sec−1 in 8-h light/16-h dark cycles at 23°C (high light) or under 120 μmol photons m−2 sec−1 in 16-h light/8-h dark cycles at 23°C (long day).

Isolation of proteins

For isolation of thylakoids and the soluble protein fraction, leaves were frozen in liquid nitrogen and thereafter homogenized in shock buffer (10 mm HEPES-KOH, pH 7.6, 5 mm sucrose, 5 mm MgCl2 and 10 mm NaF). The suspension was filtered through Miracloth and centrifuged 2500 g, for 4 min at 4°C. After centrifugation, the supernatant (crude fraction of total soluble proteins) was stored for later use, and the thylakoid pellet was resuspended in storage buffer (10 mm HEPES-KOH pH 7.5, 100 mm sucrose, 5 mm NaCl, 10 mm MgCl2). For isolation of total leaf extracts, the leaf disks of equal size were cut from mature Arabidopsis leaves and carefully homogenized in grinding buffer, composed of shock buffer:Laemmli solubilization buffer (Laemmli, 1970; 1:1, v.v).

Determination of protein content

Proteins were determined as described by Bradford (1976).

Determination of chlorophyll content

The chlorophyll content of isolated thylakoid membranes was determined as described by Porra et al. (1989). Chlorophyll content per leaf area was determined according to Inskeep and Bloom (1985).

Sodium dodecyl sulfate-PAGE and Western blotting

Proteins were separated in SDS-PAGE (15% acrylamide, 6 m urea) according to Laemmli (1970). After electrophoresis, proteins were electroblotted to a polyvinylidene fluoride (PVDF) membrane (Millipore, and immunodetection was performed using standard techniques (CDP Star Assay Kit, New England Biolabs, Protein-specific antibodies were purchased from Agrisera (Lhcb1, Lhca2;, Research Genetics (D1 and D2; or were kindly provided by Professor Scheller (FNR, PsaD), Dr Hundal (CF1), Dr Kane (RuA) and Professor Wollman (cyt f). Thylakoid phosphoproteins were immunodetected with a polyclonal phosphothreonine antibody (New England Biolabs). Samples were loaded on gels according to the linearity tests with antibodies (data not shown). Immunoblots were quantified with Fluorchem Image Analyser (Alpha Innotech Corporation,

Two-dimensional electrophoresis

Thylakoid proteins were first subjected to isoelectric focusing. Two hundred micrograms of protein extract was solubilized in RB-buffer (8 m urea, 2 m thiourea, 4% (w/v) 3[(3-cholamidopropyl)dimethylammonio]-propanesulfonic acid (CHAPS), 100 mm dithiothreitol, 0.5% (v.v) biolytes pH 3–10) and applied to a ReadyStrip® IPG Strips (pH 3–10, Bio-Rad Laboratories, Inc, during 12-h active rehydration. The proteins were focused at 150 V for 15 min, at 300 V for 40 min, at 500 V for 1 h, at 1000 V for 2 h followed at 8000 V until 65 000 Vh was reached. After focusing, the strips were incubated in 2% (w/v) dithiothreitol in EB buffer [6 m urea, 0.375 m 2-amino-2-(hydroxymethyl)-1,3-propanediol (TRIS)-HCl pH 8.8, 2% SDS (w/v), 20 % glycerol (v.v)] at room temperature for 15 min followed by incubation in 2.5% (w/v) iodoacetamide in EB buffer at room temperature for 15 min in darkness. Subsequently the gel strip was subjected to SDS-PAGE (14% polyacrylamide gel without urea).

Matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) MS analysis

Identification of proteins by MALDI-TOF MS was performed as described by Herranen et al. (2004).

Quantum efficiency of PSII

The maximum quantum efficiency of PSII photochemistry was monitored as a ratio of variable to maximal fluorescence, FV/FM (FV is the difference between maximal, FM, and initial, F0, fluorescence), measured from intact leaves with a PAM-2000 fluorometer (Walz, after a 30-min dark incubation. FM was measured by applying a 0.8-sec pulse of saturating white light.

Excitation pressure of PSII

Measurements were performed with a PAM 101-Fluorometer (Walz). After transferring a leaf to the sample holder of the PAM fluorometer, steady-state fluorescence (Fs) was monitored within 3 min at a light intensity and temperature corresponding to the growth conditions. Light was provided by a halogen lamp (Schott KL1500, A 0.8-sec pulse of saturating light was given to estimate the maximum level of fluorescence in steady-state light (FM′). Then actinic light was turned off and far-red light was applied for 2 sec for determination of the minimum level of the fluorescence (F0′). The excitation pressure of the PSII complex was calculated as 1−qP, where qP is the coefficient of photochemical quenching, qP = (FM′–Fs)/(FM′–F0′).

The 77-K fluorescence spectra

Low-temperature fluorescence emission spectra of thylakoid membranes were measured at 77 K by using a diode array S2000 Fiber Optic Spectrometer (Ocean Optics, equipped with a reflectance probe. The thylakoid membranes at a concentration of 5 μg chlorophyll ml−1 in a 100 μl volume of storage buffer were excited with the light below 500 nm defined with the filters LS500 and LS700 (Newport Corp.,, placed in the front of a slide projector.

Carbon dioxide fixation

The CO2 assimilation rate of the fully grown mutant and WT plants was determined with the Ciras-1 photosynthesis system (PP Systems, in atmospheric CO2 concentration (360 ppm) under growth conditions using a special Arabidopsis chamber (PP Systems).

Oxido-reduction of P700 and determination of the plastoquinone pool size

A PAM fluorometer (PAM-101/102/103, Walz) equipped with an ED-P700DW-E emitter–detector unit was used to monitor the redox state of P700 by absorbance changes at 810 nm using 860 nm as a reference. Leaves were kept in the dark for 10 min prior to the measurement. The P700 was oxidized by far-red light from a photodiode (FR-102, Walz) for 30 sec, and the subsequent re-reduction of P700+ in the dark was monitored. In order to estimate the size of the intersystem electron pool capable of donating electrons to oxidized P700 in intact leaves, the changes in the oxidation curve of P700 upon applying single-turnover (ST) and multiple-turnover (MT) pulses were analyzed by calculating the ratio of MT area to ST area as described (Asada et al., 1992; Schreiber et al., 1988). A ST flash (14 μsec) was provided by a xenon lamp using a XST-103 single-turnover system (Walz) and a MT pulse (50 msec, 2200 μmol photons m−2 sec−1) was obtained using a slide projector and an electronic shutter.

The NADP-malate dehydrogenase activity

The NADP-malate dehydrogenase (NADP-MDH) activity was measured as described by Scheibe and Stitt (1988).

Determination of nitrate reductase activity

Nitrate reductase activity was assayed as described in Marton et al. (1982).

Complementary DNA microarrays

Arabidopsis cDNA microarray chips are based on the GEM1 clone set from InCyte Genomics ( The chips contain 7942 elements corresponding to approximately 6500 unique genes. Total and poly(A)+ RNA extraction, RNA labeling and hybridization processes were performed according to Piippo et al. (2006). The arrays were scanned with ScanArray 5000 (GSI Lumonics, and the spot intensities were quantified with the ScanArray Express Microarray Analysis System 2.0 (Perkin-Elmer Life Sciences, using the adaptive circle method. The data were analyzed with GeneSpring 7.2 (Agilent Technologies, and the gene annotation was derived from the Arabidopsis Information Resource (TAIR;

Structural modelling of LFNR heterodimers

The amino acid sequence of Arabidopsis LFNR1 (Swiss-Prot accession number Q9FKW6) and LNFR2 (Q8W493) were aligned to the sequences of maize LFNR1 (Q9SLP6), LFNR2 (Q9SLP5) and LFNR3 (sequence derived from Okutani et al., 2005), while Arabidopsis ferredoxin-1 (P16972) was aligned to the maize ferredoxin-1 (P27787) with the program malign (Johnson and Overington, 1993) in the Bodil visualization and modeling package (Lehtonen et al., 2004). The structural models of the Arabidopsis LFNR dimers in complex with ferredoxin were constructed with the program modeller (Sali and Blundell, 1993) based on the alignment and the crystal structure of the maize LFNR dimer in complex with ferredoxin (PDB code 1GAQ). The other subunit in the maize LFNR dimer was derived by applying crystallographic symmetry operators to the published structure (1GAQ) with Pymol (DeLano, 2002). Ten different models were generated and after visual inspection the one with the lowest value for objective function, given by modeller, was chosen. The potential phosporylation sites on Arabidopsis LFNRs were predicted with the NetPhos 2.0 server (Blom et al., 1999).


We thank Virpi Paakkarinen, Maija Holmström and Nina Lehtimäki for excellent technical assistance and Drs T. and E. Tyystjärvi for critically reading the manuscript. Professor Mark Johnson is acknowledged for the excellent facilities at the Structural Bioinformatics Laboratory at the Department of Biochemistry and Pharmacy at Åbo Akademi University. We thank the Salk Institute Genomic Analysis Laboratory for providing the sequence-indexed Arabidopsis TDNA insertion mutants. This work was financially supported by the Academy of Finland (110099, 20536, 107039, 207390) and the National School for Informational and Structural Biology. Funding for the SIGnAL indexed insertion mutant collection was provided by the National Science Foundation.