Structure of the Complex between Plastocyanin and Cytochrome f from the Cyanobacterium Nostoc sp. PCC 7119 as Determined by Paramagnetic NMR

The complex between cytochrome f and plastocyanin from the cyanobacterium Nostoc has been characterized by NMR spectroscopy. The binding constant is 16 mm–1, and the lifetime of the complex is much less than 10 ms. Intermolecular pseudo-contact shifts observed for the plastocyanin amide nuclei, caused by the heme iron, as well as the chemical-shift perturbation data were used as the sole experimental restraints to determine the orientation of plastocyanin relative to cytochrome f with a precision of 1.3 Å. The data show that the hydrophobic patch surrounding tyrosine 1 in cytochrome f docks the hydrophobic patch of plastocyanin. Charge complementarities are found between the rims of the respective recognition sites of cytochrome f and plastocyanin. Significant differences in the relative orientation of both proteins are found between this complex and those previously reported for plants and Phormidium, indicating that electrostatic and hydrophobic interactions are balanced differently in these complexes.

In oxygen-evolving photosynthetic organisms, light-driven ATP synthesis requires the participation of cytochrome b 6 f complex (1,2), which couples proton translocation across the thylakoid membrane to the electron transport between PSI 1 and PSII (3). In the cytochrome b 6 f complex, cytochrome f (Cf) transfers electrons from the Rieske iron sulfur cluster to a soluble metalloprotein that acts as the electron donor for the P700 cofactor of PSI. The Cf subunit consists of a ϳ28-kDa N-terminal soluble part anchored to the membrane by a Cterminal helix (4). It represents an atypical c-type cytochrome because of both its ␤-sheet secondary structure and the unusual heme axial coordination (5). The long axis of the soluble part is tilted relative to the membrane normal, and the heme is oriented appropriately for approach of the Rieske protein from the membrane side and of plastocyanin (Pc) from the luminal side (6,7).
Pc is the most ubiquitous electron carrier between Cf and P700 (8). It is a type I cupredoxin (9) that consists of an anti-parallel ␤-sandwich structure with a single copper atom (10 -12) that is coordinated by two nitrogen atoms and two sulfur atoms from highly conserved residues.
In addition to its physiological relevance, the electron transfer reaction between Cf and Pc represents an excellent case to study the transient nature of protein interactions in electron transfer chains (13). The lifetime of this kind of complexes is on the order of 1 ms or less. Due to the large amount of functional data available, this reaction has become a very useful model to test theoretical approaches for the prediction of structures of protein-protein complexes (14 -20).
In plants this reaction shows fast kinetics at 100 mM ionic strength (Ͼ10 8 M Ϫ1 s Ϫ1 ) despite both its modest binding constant (ϳ7 mM Ϫ1 ) under these conditions and the small difference in redox potential (20 mV) between donor and acceptor (21,22). The mechanism of this electron transfer reaction has been studied with several techniques (22). Such studies support the importance of electrostatic interactions involving the acidic patches ("site 2") on Pc and the basic residues of Cf for binding under in vitro conditions (23)(24)(25)(26) and the essential role of specific residues in the hydrophobic patches of both Cf (27) and Pc (26). Notably, the relevance of the electrostatic interactions could not be confirmed in vivo (15). In the system from Phormidium laminosum, the only cyanobacterium for which the kinetics of the reduction reaction have been analyzed so far, the electrostatic effects appear to be weaker and less optimized compared with plants (28,29).
The solution structures of two Pc-Cf complexes have been obtained by dissecting the diamagnetic and paramagnetic contributions to the chemical-shift perturbations of Pc resonances upon Cf binding. The first one (PDB entry 2PCF) corresponds to the complex between spinach Pc and turnip Cf (30), and the second corresponds to the proteins from P. laminosum (31). Both structures show modest interface areas (600 -850 Å 2 per protein). Moreover, in both cases the hydrophobic patch of Pc ("site 1") lies near Tyr-1 of Cf, thus providing an appropriate environment for efficient electron transfer toward the copper atom through the exposed copper-coordinating His residue. In addition to this, chemical-shift perturbation data have been reported for several heterologous plant and cyanobacterial systems (32)(33)(34). Despite their similarities, significant differences are found between the cyanobacterial and the plant complexes. In Phormidium, Pc binds Cf in a "head-on" conformation in which the hydrophobic patch accounts for the whole recognition interface in Pc, contrary to the "side-on" interface that also involves the acidic patches, which is found in the plant complex. Both kinds of complexes have been predicted in theoretical studies using the co-ordinates from plant proteins (15)(16)(17). The ionic strength dependences of the structures suggest that in the plant complex electrostatics play a dominant role, whereas in Phormidium complex formation is governed by the hydrophobic effect.
It is known that Phormidium is a thermophilic organism (35). A higher ambient temperature could influence the balance between electrostatic forces and hydrophobic effects, making this complex unusual and different from that in other cyanobacteria. Hence, it is unknown if differences between the reported plant and Phormidium complexes are applicable to all cyanobacteria. Here, the structure of the complex between Pc and Cf from another cyanobacterium, Nostoc (formerly Anabaena), has been determined. Interestingly, the results herein presented are consistent with a single conformation in the transient complex between Pc and Cf that resembles the characteristic side-on binding mode present in plants yet has an interface similar to that found in the Phormidium complex.

EXPERIMENTAL PROCEDURES
Protein Preparation-Uniformly (99%) 15 N-labeled Nostoc sp. PCC 7119 Pc was produced in Escherichia coli JM109 transformed with pEAP-WT (36). The culture conditions and purification methods will be published elsewhere (64). Cadmium substitution of the copper in plastocyanin was performed as published (37) except that a PD-10 column (Amersham Biosciences) pre-equilibrated with a solution of 50 mM HEPES, pH 7.0, containing 1 mM CdCl 2 was used instead of a Sephadex G25 gel filtration column.
The soluble part of Nostoc sp. PCC 7119 Cf was produced in E. coli DH5␣ transformed with both pEC86, containing the c-type cytochrome maturation cassette (38), and an expression vector for Cf, pEAF-WT, obtained by insertion of a chimeric petA gene in pBluescript II (Stratagene). This chimeric gene coded for a fusion protein with Cf truncated at the C terminus (at position 253) and the signal peptide of cytochrome c 6 (36). Cells were grown in LB medium with 100 g/ml ampicillin, 12 g/ml chloramphenicol, and 6 mg/ml Fe(NH 4 ) 3 citrate under semianaerobic conditions (39) at 35.5°C, 150 rpm for 32 h up to an A 600 of 1.3. Protein yields up to 1.5 mg/liter were obtained in this manner. The purification procedure used for Cf expressed from pEAF-WT will be described elsewhere. 2 NMR Sample Preparation-Pc and PCd protein solutions were concentrated to the required volume by ultrafiltration methods (Amicon, YM3 membrane) and exchanged into 10 mM sodium phosphate, pH 6.0, H 2 O/D 2 O 95:5 solutions. Protein concentrations were determined by absorption spectrophotometry using a ⑀ 598 of 4.5 mM Ϫ1 cm Ϫ1 for the oxidized form of Pc and a ⑀ 278 of 5.5 mM Ϫ1 cm Ϫ1 for PCd. The PCd ⑀ 278 was estimated using protein concentration values from Bradford assays. A A 278 /A 598 ratio of 1.0 of the oxidized Pc indicated sufficient purity for characterization by NMR. The stock concentrations were 2.0 mM 15 N-labeled Pc and 2.7 mM 15 N-labeled PCd.
The soluble domain of Cf was concentrated using Amicon YM10 membrane and exchanged into 10 mM sodium phosphate, pH 6.0, 3 mM sodium ascorbate, H 2 O/D 2 O 95:5 solutions. The concentration determination was based on optical spectroscopy using an ⑀ 556 of 31.5 mM Ϫ1 cm Ϫ1 for the reduced Cf (30). A 3.7 mM ferrous Cf stock solution with a A 278 /A 598 ratio of 0.9 was used. Cf was kept in a reduced form with a few equivalents of sodium ascorbate and was stable in this form for days.
The ferric form was prepared by the addition of a 5-fold excess of potassium ferricyanide (K 3 [Fe(CN) 6 ]) followed by gel filtration (Amersham Biosciences Superdex G75) to remove ferrocyanide. Complete oxidation was verified by the disappearance of the absorption band at 556 nm. Then, a 2.0 mM ferric Cf stock solution was prepared.
NMR Spectroscopy-All NMR experiments were performed on a Bruker DMX 600 NMR spectrometer operating at 298 K. The 1 H and 15 N assignments of reduced Nostoc Pc assignments were taken from Badsberg et al. (40). For sequence-specific assignment of the backbone amide resonances of PCd (Supplemental Table S2), a two-dimensional 1 H, 15 N HSQC (41), two-dimensional 1 H, 15 N HSQC nuclear Overhauser enhancement spectroscopy with 150 ms mixing time, and two-dimensional 1 H, 15 N HSQC total correlation spectroscopy with 80-ms mixing time spectra were recorded.
The effects of complex formation on PCd were followed by acquiring two-dimensional 1 H, 15 15 N HSQC spectra of free Pc, the oxidized complex and the reduced complex were acquired, always on the same sample. Ferric Cf from a stock solution was added to a 15 N-labeled PCd sample with final concentrations of 0.35 and 0.50 mM, respectively. Cf was reduced with 10 mol eq of a concentrated sodium ascorbate solution. Given the final Cf concentration and the binding constant, the percentage of Pc bound was calculated to be 55%.
All data processing was performed with AZARA (www.bio.cam.ac.uk/ azara), and analysis of the chemical-shift perturbations (⌬␦ Bind ) with respect to the free protein was performed in Ansig (42)(43)(44). The spectra were calibrated against the internal standard [ 15 N]acetamide (0.5 mM).
Binding Curves-Titration curves were obtained by plotting ⌬␦ Bind against the molar ratio of Cf II/III :PCd for the most strongly affected signals. Non-linear least squares fits to a 1:1 binding model (21) were performed in Origin 6.0 (Microcal Inc.). This model accounts for the dilution effect of both proteins during the titration, with the ratio of Cf and PCd and ⌬␦ Bind as the independent and dependent variables, respectively. The binding constant (K a ) and the maximum chemical shift change (⌬␦ max ) were the fitted parameters. A global fit of the data was performed in which the curves were fitted simultaneously to a single K a value, whereas the ⌬␦ max for each resonance was allowed to vary.
Chemical Shift Mapping-The shifts observed in the complex PCd-Cf II with 3 eq of Cf were extrapolated to 100% bound for all residues using the K a obtained from the fits. The average chemical-shift perturbation (⌬␦ avg ) of each amide was calculated using the following equation (45), in which ⌬␦ N is the change in the 15 N chemical shift, and ⌬␦ N is the change in the 1 H chemical shift when the protein is 100% bound to Cf.
Restraints Classes-Details of the restraints definitions are provided in the supplemental material. Briefly, four groups of restraints were defined. The interface restraints represent the chemical-shift perturbation data for Pc nuclei (Supplemental Table S1). These are satisfied when the nuclei are close the Cf surface. Additional interface restraints were defined to serve as a weak van der Waals repel function. PCS were used to define pseudo-contact restraints and angle restraints according to the procedure described in Ubbink et al. (30), and minimal distance restraints were defined for amide groups that did not experience a PCS.
Electrostatic restraints based on kinetic rather than NMR data were used previously (30) to represent the electrostatic attraction between PCd and Cf. In the Nostoc complex, these were not used because the NMR experimental data were sufficient to obtain a well defined structure.
A summary of the restraint groups is listed in Table I. The product between the number of restraints and the scaling factor used in the calculations indicates the importance of each restraint group. Note that the pseudo-contact restraints, which give quantitative information, are dominant.
Nostoc Cf Homology Model-A homology model of Cf was built using the COMPOSER (46) module of SYBYL 6.5 (Tripos Inc.) using X-ray diffraction data from Brassica rapa, PDB entry 1CTM (resolution 2.30 Å (5)) and PDB entry 1HCZ (resolution 1.96 Å (47)) as templates. Including the structure of Phormidium Cf (PDB entry 1CI3 (48)) as the template did not improve the model. Three sequence stretches (residues 1-9, 13-183, and 198 -254) were considered as conserved. These regions showed identities of 66.7, 64.9, and 54.4%, respectively. Two loops corresponding to residues Trp-4 -Gln-6 and Ala-184 -Val-197 were simulated to allocate 1-and 3-residue insertions, respectively, using the TWEAK option. On average, the r.m.s.d. of backbone atoms of the model of Cf with respect to the above structures was 0.56 Å. The structure of Cf from Mastigocladus (made available only after completion of our calculations, PDB entry 1VF5 (6)) shows a similar extension of the small domain of Cf. The Nostoc Cf model shows a 1.16-Å r.m.s.d. with this structure. The largest differences correspond to residues 184 -197, which show high B-factors in the Mastigocladus Cf crystal structure, suggesting that this loop may be flexible.
Structure Calculations-Structure calculations were performed using XPLOR-NIH Version 2.9.1 (49,50). The structures of Nostoc Pc (PDB entry 1NIN (40)) and the homology model of Cf were treated as rigid bodies, and the co-ordinates of Cf were fixed. PCd was placed at a random position and allowed to move in a restrained rigid-body molecular dynamics calculation. None of the standard energy terms was used. Only the groups of experimental restraints described above were applied to dock the proteins. Five thousand cycles (see supplemental material) of calculations were performed (9 h on a dual processor Pentium IV PC running under LINUX). Only structures with a total restraints "energy" (E tot ) below a threshold were saved, yielding ϳ90 structures. To assure sufficient sampling of the orientation space, a large random displacement of Pc occurred when a (local) minimum had been found, as judged from a total restrained energy that had not changed for more than 50% during 10 cycles. About 200 of such displacements occurred in a representative run. As an illustration, E tot has been plotted against the cycle number in Fig. S2 for a sector of one trajectory, corresponding to 200 cycles.
The resulting structures were ranked according to total restraint energy and the top ten structures, with total restraint energy values from 28 to 29 arbitrary units subjected to restrained energy minimization of the side chains followed by a short restrained rigid body energy minimization, both using the XPLOR-NIH repulsive van der Waals term with reduced scaling. This largely removed the collisions between Pc and Cf atoms while maintaining the low total restraint energy value. The ten best structures have been deposited in the Protein Data Bank under entry 1TU2. Buried surface areas have been calculated using NACESS (51).
Electron Transfer Pathways-To determine the residues that could be involved in the electron transport, the best electron transfer pathway for each of the energy-minimized complex structures was calculated using Greenpath Version 0.971 (52). This program performs a Green function analysis based on two-state super-exchange model (53). No enhanced coupling was used for aromatic rings. For representation purposes, we selected all the coordinates of the residues that appear in any of these paths instead of just representing the bonds and jumps involved.

RESULTS AND DISCUSSION
Binding Affinity-To characterize the complex of Pc and Cf from Nostoc, Cf was titrated into a solution of 15 N-labeled Cd-substituted Pc, PCd. The copper in Pc was replaced by the redox inactive substitute cadmium to allow for studies with both oxidized and reduced Cf without interference from electron transfer reactions. The effects of the titrations were followed with two-dimensional HSQC experiments. Shifting of resonances during the titration indicated that binding and dissociation were fast on the NMR timescale (Ͼ100 s Ϫ1 ). In Fig.  1, the chemical-shift perturbations due to binding to reduced Cf (Cf II ) are plotted for several residues of PCd. The curves clearly illustrate that the chemical-shift perturbations increase as a function of the Cf II concentration. A global fit of the data to a 1:1 binding model (21)  , and the same affinity was obtained with oxidized Cf (K A ϭ 16 Ϯ 2 ϫ 10 3 M Ϫ1 ; data not shown). This value is slightly lower than that obtained for native (Cu(I)) Nostoc Pc (26 Ϯ 1 ϫ10 3 M Ϫ1 (64)). Hence, it can be concluded that the binding affinity is independent of the oxidation state of Cf but appears to vary slightly between Pc with a singly charged metal and a doubly charged one.
Interface Map-In Fig. 2, the size of the chemical-shift perturbations for residues affected upon titration with Cf II are color-coded onto the surface of Pc. Perturbed residues map in sequence stretches 7-17, 32-44, 63-72, and 88 -100. These stretches form a large area comprising residues from both classical binding sites (10,25). Three proline residues, at positions 37, 38, and 91 (gray), are located in the middle of the interface, close to the copper ligand His-92. In addition to the main interaction area, three isolated residues (Lys-51, Asp-54, and Leu-55) undergo a significant perturbation. These residues are located in a region below site 2, comprising mainly charged residues that have an important role in the Nostoc complex structure, as is explained below. The perturbation map of PCd is similar to that found for Pc(Cu(I)) (64).
The overall sizes of the perturbations and the localized nature of the binding map indicate that the complex between PCd and Cf is well defined according to the classification for well defined versus dynamic complexes, suggested by Worrall et al. (55) and Prudêncio and Ubbink (56). That is, PCd adopts a single predominant conformation during most of the lifetime of the complex. This is also supported by the observation of intermolecular pseudo-contact effects, as explained below.
Experimental Restraints and Structure Calculation-When comparing the perturbations of Pc resonances observed with reduced and oxidized Cf, it is striking that in the latter case many Pc nuclei experience an additional shift that is of similar size for 1 H and 15 N. This is illustrated in Fig. 3 Interface  Distance  41  5  205  Pseudo-contact  Distance  81  20  1620  Minimal distance  Distance  90  10  900  Angle Angle 81 a a Scaling of the angle restraints is not comparable with that of distance restraints.
paring Pc in the presence of oxidized and reduced Cf (⌬␦ oxidized -⌬␦ reduced ). In this case there is a clear correlation of the shifts, suggesting that the additional shifts are caused by intermolecular PCS from the ferric heme iron onto Pc nuclei because these should be similar (in ppm) for the proton and nitrogen nuclei of a given amide. The strongest PCS are found in the hydrophobic patch region, suggesting that this region comes closest to the heme iron in the complex. The binding effects and the PCS can be used to determine the orientation of Pc relative to Cf in the complex. Both types of shifts were translated into restraints for a rigid-body docking calculation. The binding shifts were used in a qualitative manner; any shift larger than the threshold yields a restraint that requires that the amide is brought close to the surface of Cf (interface restraints, Supplemental Table S1). The PCS are used quantitatively; during the calculations, the PCS are calculated for the given position of Pc and compared with the experimental values. If the difference is larger than the error margins, the restraint is violated. The position of Pc is changed, and the PCS are calculated again. Thus, in an iterative fashion, the optimal position, with minimal restraint violations is found. These are called the pseudo-contact restraints (Supplemental Table S1). The sign of the PCS gives information about the angle between iron-nucleus vector and the orientation of the magnetic susceptibility tensor, thus providing angle restraints. The absence of a PCS can be used to define a minimaldistance restraint, which is violated when the nucleus gets too close to the heme iron. Also, these restraints are evaluated iteratively during the calculations. All these restraints were used in a rigid-body calculation to obtain the orientation of Pc relative to Cf with minimal violations of the experimental restraints. For this purpose the solution structure of Pc (40) and a homology model of Cf, based on its conserved amino acid sequence, were used. Apart from the experimental restraint terms, no other forces, such as electrostatics, were used, except for a weak van der Waals repel function to avoid extensive collisions between the atoms of both proteins. Further details about the definitions and numbers of restraints as well as the calculations are given under "Experimental Procedures" and the Supplemental Material.
Results of the Structure Calculation-Two types of complexes were obtained from the above restrained rigid body calculations. All the structures with the lowest sum of violations (28.5 Ϯ 0.5 arbitrary units) cluster in one orientation, with an average positional root mean square deviation of 1.3 Ϯ 0.6 Å ("positional r.m.s.d."; see below), with respect to their average structure (Fig. 4). The iron-cadmium distance was 16.2 Ϯ 0.1 Å. In addition, an alternative set of structures with another conformation was found (Supplemental Fig. S3). These have a higher sum of violations (38.9 Ϯ 0.3 arbitrary units) and, thus, a much worse fit to the NMR data, in particular for the interface restraints, because the interaction surface is small and does not involve the hydrophobic patch. These structures have a significantly larger degree of variability with a positional r.m.s.d. of 3.2 Å and larger iron-cadmium distance (17.4 Ϯ 0.1 Å). Consequently, these alternative complexes with larger violations, a small interface, and lower predicted efficiency in the electron transfer reaction are considered to be a non-physical solution of the calculations and are not considered further.
Quality of the Structure; Accuracy and Precision-The quality of the 10 best structures was evaluated on the basis of the violations between the calculated and the experimentally observed values for each group of restraints (see "Experimental Procedures"). With regard to interface restraints, systematic violations only appear for a few amide nuclei, with the largest found for Lys-20. The amides of these residues experience small chemical-shift perturbations upon binding that may be attributable to indirect effects (30). For the minimal distance restraints, there are a few small violations that correspond to Ile-101 in most of the 10 structures. This amino acid is at the boundary of residues that experience PCS.
The violations of the PCS restraints can be evaluated from the largest deviations for the two His residues, which are very close to the iron. It is likely that the necessary assumptions on the size, axiality, and orientation of the magnetic susceptibility tensor (see supplemental material) are the limiting factors in the accuracy of the structures. The angles between the assumed zz direction and the iron-nucleus vector, as calculated from the ten best structures, are plotted in Fig. 5B. Positive PCS correspond to an angle smaller than 54°, and negative PCS have an angle larger than that. The angles observed in the structures are all in agreement with the positive sign of the PCS. The PCd position relative to that Cf is the most relevant feature in the precision of the complex structure. After rigidbody and side-chain energy minimizations of the 10 structures, the positional r.m.s.d. is determined by aligning the Cf molecules in each structure and calculating the r.m.s.d. of backbone heavy atoms of Pc compared with the average structure. The average backbone positional r.m.s.d. for PCd is 1.3 Ϯ 0.6 Å for the 10 best structures. This precision is achieved without the input of electrostatic restraints in structure calculations as was done in the case of the complex of plant proteins (30) in which the positional r.m.s.d. was 1.05 Å. In the present case it can be concluded that the number of experimental restraints, which is larger than in previous studies, is sufficient to resolve a well defined structure.
Description of the Structure-The complex is shown in Fig. 4 in ribbon and space-filling representations. The binding interface involves the hydrophobic areas close to the metal centers in both proteins. In Pc it comprises 14 residues from the hydrophobic patch (Leu-14, Val-36, Pro-37, Pro-38, Leu-64, Met-66, Pro-68, Pro-91, His-92, and Ala-95), the neighboring Lys-35, and from the adjacent region of site 2 (Lys-62, Gln-63, and Glu-90). The buried area in this recognition patch on Pc is 572 Å 2 . A second, minor recognition site on Pc corresponds to residues Asp-54 and Lys-57. It accounts for 41 Å 2 of the buried surface. This patch is clearly involved in electrostatic interactions with charged residues at the small domain region of Cf. On Cf, the recognition site measures 525 Å 2 , comprising 20 residues that are at least partially buried during binding. Three of the five aromatic residues in this patch, namely Tyr-1, Phe-3, and Tyr-102, represent ϳ31% of the recognition surface of Cf. Most residues at the recognition site are hydrophobic or polar, with charged residues lying at the rim of the interface.  Several aspects of the amino acid composition of the interface can be related to the transient nature of the complex. Charge interactions are observed at the edge of the interface, in accord with the general findings for transient complexes (33,56). Residues Lys-57 and Lys-62 of Pc are close to Glu-189 and Asp-64 from Cf, respectively. Several others (Asp-10, Lys-11, Lys-35, Asp-44, Lys-51, Asp-54, Asp-90, and Arg-93 from Pc and Asp-100, Glu-108, Glu-165, Asp-190 and the heme propionates in Cf) are further away from the other protein but may be close enough to contribute to the electrostatic interactions. The charge interactions are expected to be favorable with negative charges on Cf and predominantly positive ones on Pc.
Six proline residues are located in the interface, Pro-37, Pro-38, Pro-68, and Pro-91 from Pc and Pro-118 and Pro-120 from Cf representing 17% of the residues at the interface and contributing ϳ30% to the buried surface area. Proline is a residue with very low propensity in protein-protein interfaces of tight complexes (3.8% (57)(58)(59)(60)). However, proline is more abundant in transient protein complex interfaces (61). The inability of Pro to form hydrogen bonds may be employed in this way to limit the affinity in such complex. Four glutamine residues are found in the interface, representing 13.7% of the buried area (4.3% in recognition sites of tight complexes).
These polar, uncharged residues surrounding the hydrophobic patch may enhance dissociation by facilitating resolvation of the interface, as suggested previously by Crowley and Ubbink (33).
The buried surface area of the interface has a standard size for non-obligate protein complexes (57)(58)(59)(60). Both sites (1 and 2) of Pc make contact with the Cf surface, burying a total of 1100 -1200 Å 2 , similar to the 1200 -1400 Å 2 in the Phormidium complex (31) and smaller than the 1720 Å 2 found in the plant complex (62).
Recent kinetic data using site-directed mutants of Nostoc Pc have revealed a large overlap between the recognition patches of Pc for PSI (63) and Cf, 2 in particular for Pc substitutions affecting electrostatic interactions. However, there are some differences, according to the effects observed for several substitutions. For instance, Arg-93 mutants show a much larger effect on the electron transfer to PSI. Arg-93 belongs to the proximal patch of site 2, but it does not make contact with Cf in the complex. The closest charged residue of Cf (Glu-165) is 9 Å away. Thus, the electrostatic interaction between them may be moderate, and the kinetic effect of replacing Arg-93 could be attributable to more general electrostatic interactions, specifically during the encounter phase of the complex formation process, as described for the Arg-93 of P. laminosum Pc (28,29). The mutation L14A has been shown to have a large effect on the reactivity of Pc toward PSI (63). On the other hand, this mutation has only minor effects (a 2-fold decrease in reaction rates) on the reduction of Pc by Cf. 2 This is in agreement with the location of Leu-14 at the rim of the recognition site. Indeed, only a minor part (2 Å 2 ) of its solvent-accessible surface is buried in the complex interface.
Electronic Coupling-The structure reported herein is consistent with fast electron transfer. However, the average distance from iron to cadmium is the largest (16.2 Ϯ 0.1 Å) found so far, with 11.0 and 13.9 Å in two plant complexes (30, 65) and 15.0 Ϯ 2.0 Å for the complex of Phormidium (31).
To get information about the residues that may be involved in coupling of electron transfer between both proteins, the best theoretical paths in the ensemble of the energy-minimized structures of plant, Phormidium, and Nostoc were analyzed using Greenpath Version 0.971 (52). Putatively important residues are shown in Fig. 6. In all spinach structures Tyr-1 and His-87 (His-92 in Nostoc) are important for the electronic coupling, with a "path length" value (including bonds and throughspace couplings) of 14.2 Ϯ 0.4 Å. Similar results are found using the structure of the complex of poplar Pc with turnip Cf complex (65), although in this case the path length is larger (19 Ϯ 1 Å). In Phormidium the path length (19.4 Ϯ 0.5 Å) is similar to that in poplar, but all the best pathways found involve the aromatic ring of Phe-3 in Cf. In Nostoc, two sets of pathways are found in the final ensemble of structures; one (20.  0.6 ϫ 10 Ϫ9 and 1.6 Ϯ 0.3 ϫ 10 Ϫ9 ) are somewhat smaller than those found for the plant and Phormidium structures (5.3 Ϯ 0.8 ϫ 10 Ϫ7 and 7 Ϯ 2 ϫ 10 Ϫ8 , respectively).
A combined Montecarlo and molecular dynamics calculation on a heterologous plant Pc-Cf complex (17) suggested two electron pathways. The first involves Tyr-83 (in plant Pc). In all the solution structures reported including this one, the complex conformation is very different from that supporting this pathway (configuration E in Ref. 17). The Nostoc Pc residue equivalent to Tyr-83 (Tyr-88) is 12.5 Å away from the closest Cf atom, and mutation of Tyr-88 to either Ala or Phe does not affect the reduction of Pc by Cf in Nostoc. 2 The second pathway was described for the theoretical conformation D, which is the most similar to the Nostoc structure. It involves Pro-86 in plant Pc (Pro-91 in Nostoc) that should receive electrons from a Cf heme propionate. The relative coupling for the theoretical complex (1.8 ϫ 10 Ϫ9 ) is similar to those found for the Nostoc proteins. However, in the Nostoc structure the distance between Pro-91 and the heme is ϳ7 Å, too far for efficient direct electron transfer, and instead, the suggested pathway involves Tyr-1.
The Role of Electrostatic Interactions-Perhaps the most remarkable aspect of the Nostoc structure is the side-on binding of Pc. This conformation resembles the orientation found in the plant complex rather than the one found for the other cyanobacterial complex from Phormidium (Fig. 7). To establish the significance of the different orientations of Pc in the Nostoc and Phormidium complexes, the sum of violations was calculated for the Nostoc proteins, with Pc in the orientation of that found for Phormidium. The sum is significantly larger than those of the best structures, with larger minimum-distance and interface terms. The pseudo-contact restraints show a smaller violation for His-92, but the number of violations in the rest of the protein increases. To test whether this orientation corresponded to a local minimum in the conformational space, it was used as the input structure for a restrained rigid-body docking calculation. The calculation showed that Pc changed its orientation toward that of the 10 best structures (sum of violations, 30.8 arbitrary units; r.m.s.d. with the average of the 10 best structures, 1.5 Ϯ 0.5 Å). It is concluded that the experimental data are sufficient to make a clear distinction between the different orientations.
The structure of the Phormidium complex does not exhibit a strong ionic strength dependence (31), and also the electron transfer kinetics show only a weak salt dependence (28,29). Both the plant and the Nostoc complexes employ electrostatics but with reverse charges. In plant, positive charges on Cf interact with the negative charges on Pc, whereas in Nostoc the opposite is observed. Such charge interactions are in accord with the kinetic studies. These have demonstrated a strong ionic strength dependence of the reduction reaction of Pc by Cf in vitro for the plant system (21, 23-26, 28, 29), although the physiological significance of electrostatic interactions is not clear (15). Also for the Nostoc proteins the reaction is ionic strength-dependent. 2 Thus, it appears that a side-on binding of Pc is observed in complexes that involve electrostatic interactions. It is noteworthy that in the present calculations, electrostatic restraints were not used in any way (contrary to the plant case), and the present structure was based entirely on NMR spectroscopy-derived restraints.
It is also important to note that in all structures determined of Pc and Cf, the orientation of Pc is such that a short electron transfer chain is created between the Q o site of cytochrome b 6 f and plastocyanin (involving the QH 2 , Rieske FeS cluster, Cf heme, and Pc copper). Fig. 8 shows a model of the cytochrome b 6 f-Pc complex based on the structure of cytochrome b 6 f (6) and the present structure of Pc-Cf, obtained after alignment of the Cf molecules. It is clear that Cf faces the lumen with the region around heme ligand Tyr-1, readily forming a complex with Pc.