Ferredoxin:NADP+:reductase (FNR) catalyzes one terminal step of the conversion of light energy into chemical energy during photosynthesis. FNR uses two high energy electrons photoproduced by photosystem I (PSI) and conveyed, one by one, by a ferredoxin (Fd), to reduce NADP+ to NADPH. The reducing power of NADPH is finally involved in carbon assimilation. The interaction between oxidized FNR and Fd was studied by crystallography at 2.4 Å resolution leading to a three‐dimensional picture of an Fd–FNR biologically relevant complex. This complex suggests that FNR and Fd specifically interact prior to each electron transfer and disassemble upon a redox‐linked conformational change of the Fd.
Ferredoxin:NADP+:reductase (FNR), a flavine‐adenine‐dinucleotide (FAD) flavoenzyme, uses high energy electrons photoproduced by photosystem I (PSI) and protons to reduce nicotinamide‐adenine‐dinucleotide phosphate (NADP+) to NADPH following the reaction 2Fdred + NADP+ + 2H+ <=> 2Fdox + NADPH + H+. The reducing power of NADPH is finally involved in carbon assimilation (Carrillo and Vallejos, 1987; Arakaki et al., 1997). The two electrons involved in NADP+ reduction are transferred, one at a time, from PSI to NADP+ by a [2Fe–2S] ferredoxin (Fd). Each electron transfer requires the formation of a transient complex between a reduced Fd (Fdred) and a preformed FNR–NADP+ complex (Batie and Kamin, 1984b; Sancho and Gómez‐Moreno, 1991). More recently, it has been proposed that a larger transient complex involving PSI, Fd and FNR could be formed during linear electron transport (van Thor et al., 1999). We have previously determined the structures of the uncomplexed oxidized (Serre et al., 1996) and reduced (R. Morales, unpublished data) FNR and the oxidized and reduced Fd (Morales et al., 1999) from the cyanobacterium Anabaena PCC7119. To gain further structural insights on this important step of photosynthesis, we have studied by crystallography the interaction between oxidized FNR and Fd, both extracted from the same organism.
Results and Discussion
The two proteins co‐crystallized only with a molar excess of FNR over Fd. This results in a crystallographic asymmetric unit that consists of two reductases, named FNR1 and FNR2, and one Fd molecule. FNR1 and FNR2 interact with each other and ‘grasp’ the globular Fd molecule like a pincer (Figure 1).
In the FNR1–Fd–FNR2 crystallographic moiety, the Fd [2Fe–2S] cluster is 7.4 Å distant from the exposed C8‐isoalloxazine methyl (C8M) of FNR1 (Figure 1) thought to be involved in electron transfer in flavoproteins (Fritz et al., 1973). Such a close distance between the two redox centers might account for the fast electron transfer between Anabaena Fd and FNR in solution (6000 s−1; Walker et al., 1991). On the whole, the interactions between Fd and FNR1 are far more specific than those between Fd and FNR2 in terms, for example, of the number of hydrogen bonds and ion pairs (10 versus 3) (Table 1), van der Waals contacts [VDW (<4.1 Å) = 54 versus 21] and buried area upon complex formation (1600 A2 versus 1100 A2). These characteristics of interactions are quite comparable to those observed in other crystallographic complexes (for review see Lo Conte et al., 1999). Moreover, the molecular dipoles of Fd and FNR1, as calculated by GRASP (Nicholls et al., 1991), are nearly colinear and orientated in the same direction, resulting in the negative pole of Fd being close to the positive pole of FNR1. The biological relevance of the association between Fd and FNR1 (Fd–FNR1) is also supported by a wealth of studies in solution, mostly on spinach leaves and Anabaena: chemical cross‐linking (Zanetti et al., 1988), differential chemical modifications (De Pascalis et al., 1993), limited proteolysis (Gadda et al., 1990) and joint protein engineering and spectroscopy (e.g. Aliverti et al., 1997; Hurley et al., 1997; Martínez‐Júlvez et al., 1998, 1999 and citations therein). Nearly all the interactions identified in solution by these various approaches are found at the Fd–FNR1 interface. This strongly suggests that the binding of a second reductase (FNR2) to Fd–FNR1 within the crystal does not perturb the association between Fd and FNR.
Moreover, the association of Fd and FNR2 could not represent a catalytically efficient electron transfer complex because the distance between the Fd cluster and the FNR2‐FAD is too long for a fast electron transfer (14.5 Å). The crystallographic moiety FNR1–Fd–FNR2 cannot mimic a complex in vivo mainly because the interactions between the reductases FNR1 and FNR2 interfere with their respective NADP+ binding sites (Karplus et al., 1991; Serre et al., 1996). However, the FNR1–FNR2 interaction probably has no effect on the Fd–FNR association since NADP+ and Fd bind to distinct sites on FNR, whereas they act in a pattern of negative cooperativity (Batie and Kamin, 1984a). Moreover, such a heterotrimeric moiety would also impair the formation of the transient productive complex PSI–Fd–FNR, which has recently been described (van Thor et al., 1999). Finally, the role of FNR2 could be to stabilize crystal packing by contacts with neighboring proteins.
In Fd–FNR1 (Figure 2), Fd binds to the N‐terminal FAD binding domain (residues 12–137) and the C‐terminus structural sub‐domain (residues 237–302) on the concave side of the FNR1 molecule, which contains the solvent‐exposed di‐methyl edge of the FAD isoalloxazine. These two domains do not include the NADP+ binding site already known (Karplus et al., 1991; Serre et al., 1996). By comparison with the uncomplexed oxidized molecules (Serre et al., 1996; Morales et al., 1999), Fd undergoes a conformational change in binding to FNR1 at the level of the loop 60–67 [the root mean square difference (r.m.s.d.) between the Cα carbons of residues 60–67 is 2.5 Å]. This change in particular could be explained by the strong interaction of the Fd Phe65 aromatic side chains with FNR1 (11 VDW contacts). By contrast, the FNR structural changes upon complexation concern only side chains interacting with surface side chains of the Fd (Table 1). To sum up, the molecular interface between Fd and FNR1 consists of a core of hydrophobic side chains including Fd Phe65 and FNR1 Leu76, Leu78 and Val136, negatively charged groups of the Fd and positively and some negatively charged groups of FNR. These groups interact with each other either directly or through ‘bridging’ water molecules. In addition, the Fd–FNR1 interface is stabilized by hydrophobic forces either involving long carbonyl side chains like FNR Arg264 (11 VDW with Fd) or resulting from the loss of ∼10 water molecules upon complex formation. The latter point is in accordance with calorimetry experiments on the spinach complex (Jelesarov and Bosshard, 1994).
Comparison with hypothetical Fd–FNR models
Some hypothetical models of Fd–FNR complexes have been proposed (De Pascalis et al., 1993; Karplus and Bruns, 1994; Fukuyama et al., 1995). Analysis of one of these models (Fukuyama et al., 1995) compared with Fd–FNR1 shows that the core of the interface also consists of an intermolecular patch of the hydrophobic side chains, Fd Phe65 and FNR Leu76, Leu78 and Val136. However, the orientation of the Fd towards FNR, in both the crystallographic and the hypothetical complex, differs by ∼30°. This is due to the involvement of either the Fd acidic patch Asp67–Asp68 in our crystallographic model, or Asp26–Glu29–Glu30 (Anabaena numbering) in all the hypothetical models, in the molecular interaction. As shown by differential chemical modification studies on spinach (De Pascalis et al., 1993) both interactions are possible although the protection of Asp67–Asp68 is stronger than that of Asp26–Glu29–Glu30 upon complexation.
Electron and proton transfer
The arrangement of the redox centers of Fd and FNR1 (Figure 3) is similar to that of the flavin mononucleotide and [2Fe–2S] cluster observed in phthalate dioxygenase reductase (PDR), an NADH oxidizing enzyme that contains one FNR‐like and one Fd‐like domain in a single polypeptide chain (Correll et al., 1992). However, in PDR, the iron–sulfur cluster atom that is the closest to the FAD C8M is Fe1 (7.9 Å) instead of S2 (7.4 Å) in Fd–FNR1. Density functional theory calculations (Noodleman et al., 1995), together with the X‐ray structures of the uncomplexed oxidized and reduced Fd, have strongly suggested that upon reduction the change in net charge on Fe1 spreads partially over the S2 sulfur atom of the [2Fe–2S] cluster (Morales et al., 1999). Therefore, the close proximity of S2 to C8M in a dissociable Fd–FNR complex might aim at a more efficient electron transfer. The respective redox centers of Fd and FNR1 are only separated by an Fd main chain segment of limited steric hindrance, Ala43–Gly44–Ala45 (Figure 3), which should allow a ‘direct’ electron transfer between them. However, our model suggests that, as previously proposed (Hurley et al., 1993), the π orbitals of the Phe65 aromatic ring could also be involved in the electronic coupling between the two cofactors.
It has been proposed that proton transfer from the external medium to the FNR isoalloxazine N5 atom could be mediated by the solvent‐exposed Glu301 and Ser80 in Anabaena FNR (Serre et al., 1996), similar to Glu312 and Ser96 in spinach FNR (Bruns and Karplus, 1995). This mechanism has since been debated on the basis of site‐specific mutagenesis studies (e.g. Medina et al., 1998). In our Fd–FNR1 crystallographic association, the carboxylic group of Glu301 is no more exposed to solvent but is hydrogen‐bonded to the hydroxyl group of Fd Ser64 (Ser64‐Oγ), which is in turn exposed to solvent (Figure 3). This is, maybe, not coincidental and puts forward again a possible proton transfer pathway between the external medium and the FNR isoalloxazine N5 through Fd Ser64 and FNR Glu301 and Ser80 side chains.
After reduction of Fd by PSI (e.g. Barth et al., 1998; van Thor et al., 1999), the first step of the catalytic cycle leading to the reduction of NADP+ to NADPH requires the binding of this first Fdred molecule to a preformed FNRox–NADP+ complex (Batie and Kamin, 1984b). The FNRox–NADP+ crystal structure of a pea FNR mutant (Y308W) suggests that the FNR's invariant C‐terminal tyrosine (Tyr303 in Anabaena) is displaced into the solvent so that the NADP+ nicotinamide could stack on to the FAD isoalloxazine of FNR (Deng et al., 1999). Molecular modeling shows that the ’displaced' C‐terminal Tyr303 might be accommodated either by a cavity between Fd and FNR1 or by the solvent itself without changing our Fd–FNR1 crystallographic complex in accordance with biochemical evidence (Batie and Kamin, 1984b; Sancho and Gómez‐Moreno, 1991).
Upon electron transfer to FNR, the resulting Fdox dissociates from the FNRsemiquinone(sq)–NADP+ complex (Batie and Kamin, 1984b). How can this event be triggered? On the one hand, it has been observed, by crystallography, that the peptide bond linking Cys46 and Ser47 points its carbonyl oxygen away from the nearby cluster S2 sulfur atom in the reduced Fd and toward it in the oxidized Fd (Morales et al., 1999) (Figure 4). In any redox state the molecular surface is nearly the same and the 46–47 peptide link lies in close contact with the Fd Phe65 aromatic ring. On the other hand, as shown in this paper, Fd Phe65 interacts directly with FNR1. Taken together, these observations strongly suggest that the dissociation of the FNRsq–Fdox complex could be triggered by the redox‐linked 46–47 peptide ‘flip’ of the Fd, which transiently moves the aromatic side chain of Fd Phe65. It is also of interest to note that such a redox‐dependent mechanism has been proposed for the dissociation of the oxidized plastocyanin and PSI (Bendall, 1996).
The second reduced Fd molecule then probably binds to FNRsq–NADP+ similarly to FNRox–NADP+ as we did not observe significant conformational differences between the crystal structures of the oxidized and reduced FNRs (Serre et al., 1996; R. Morales, unpublished data).
The specificity of the molecular interactions and the molecular dipole complementarity suggest that our Fd–FNR1 crystallographic model represents a highly probable conformational state of an electron transfer complex between Fd and FNR during the last step of photosynthesis. Nevertheless, the hydrophobic patch centered on FNR Leu76, Leu78 and Val136 is sufficiently extended to accommodate other hydrophobic interactions with Fd so that the two redox centers are still close enough to each other for fast electronic exchanges. This should allow a relatively flexible mode of docking and favor turnover, in accordance with kinetics and theoretical studies (Walker et al., 1991). Moreover, our model of a redox‐sensitive association between Fd and FNR could be envisaged between Fd and other functional partners such as PSI (e.g. Barth et al., 1998; van Thor et al., 1999), sulfite reductase (Akashi et al., 1999) or thioredoxin reductase (Dai et al., 2000).
Oxidized FNR and Fd were co‐crystallized at 20°C under conditions similar to those used to grow FNR crystals (Serre et al., 1996), but only in the presence of an excess of FNR over Fd (1.4/1 or 2/1, mol/mol). It is of interest to note that the use of this ‘non‐stoichiometric’ approach has also been crucial to obtain crystals of several intermolecular complexes such as cytochrome c peroxidase–cytochrome c (Pelletier and Kraut, 1992). The orange‐red color of our crystals gave the first indication that they contained FNR and Fd since the crystals of oxidized FNR and Fd are bright yellow and dark red, respectively (Morales et al., 2000).
X‐ray data collection, processing and structure determination.
X‐ray diffraction data of a single crystal were collected at 100 K at the ID2 beamline of the European Synchrotron Radiation Facility (ESRF) (Table 2). The crystal cell parameters are: a = 63.73, b = 63.72, c = 158.02 Å and the space group is P212121, with two FNR and one Fd molecules in the asymmetric unit. The two FNR molecules were positioned by molecular replacement with AMoRe (Navaza, 1994) (correlation coefficient = 0.75 and Rcryst = 0.30) with the Anabaena FNR (Serre et al., 1996) as a model. The Fd [2Fe–2S] cluster iron atoms were located using the anomalous signal and the molecule itself finally positioned by Rmin search techniques. At this stage, crystal packing analysis showed that the crystal could be described as a twin. Interleaving cycles of electron density map calculations, model building and refinement were finally carried out with O (Jones et al., 1991) and the SHELX97 ‘twin option’ (Sheldrick et al., 1993). The final electron density map accounts for the FNR and Fd amino acid sequences. The final R and Rfree factors were 0.22 and 0.29, respectively (Table 2) (Morales et al., 2000).
The atomic coordinates of the complex (1EWY) will be released at the Protein Data Bank upon publication.
We thank Dr Bjarne Rasmussen and Xavier Vernède for their help in collecting Synchrotron data at ESRF, Professor Keiichi Fukuyama for the coordinates of his theoretical model and Professor Hans Bosshard, Dr Jean‐Marie Mouesca and Professor Gordon Tollin for helpful suggestions.
- Copyright © 2000 European Molecular Biology Organization