INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSION REFERENCES

Plastocyanin (Pc) is a small (10 kDa) protein, which functions as a shuttle of electrons from cytochrome b₆f to photosystem I (PSI) in the light reactions of photosynthesis. Pc is a -sheet type 1 blue copper protein, and the copper ion is located in the "northern region" of the protein with the ligand His⁸⁷ (poplar numbering) protruding into the solvent. This residue, implicated in the transfer of the electron from copper to PSI, is surrounded mainly by small hydrophobic amino acids, usually referred to as the Pc hydrophobic patch. The solution structure of Prochlorothrix hollandica Pc was solved recently (Babu et al., 1999). It was shown that it exhibits a unique hydrophobic patch; instead of conserved Gly and Leu residues at positions 12 and 14 (corresponding to positions 10 and 12 in the poplar protein), it has Tyr and Pro, respectively.

PSI is a multisubunit pigment-protein complex that provides the reduction potential necessary for conversion of oxidized form of nicotinamide adenine dinucleotide to reduced nicotinamide adenine dinucleotide. Crystallographic studies indicate that the lumenal site of PSI is essentially flat except for a 10-Å wide hydrophobic cleft between two major transmembrane subunits of PSI, PsaA and PsaB (Schubert et al., 1997). This cleft formed by two  helices, A/B-ij(2) of loops A/B-ij was suggested as a binding site for Pc (Fromme et al., 1994). Site-directed mutagenesis studies suggest that this region is involved in interaction with Pc (Sun et al., 1999; Sommer et al., 2002).

The docking complex is a reversible, specific assembly of the two proteins. During diffusion, proteins guided by long-range electrostatic forces tend to align their dipole moments in favorable orientation, steering two proteins toward the correct encounter complex (Gabdoulline and Wade, 2001). This steering effect, although not very specific, significantly enhances rates of diffusional collision (Zhou, 1993). Upon formation of a loose encounter complex, proteins undergo rotational and vibrational motions exploring conformational space (Northrup et al., 1988). During this process, at the optimal configuration of nuclear coordinates, electron transfer occurs. This electron transfer complex can be predicted based on available biological information. After formation of the encounter complex, short-range electrostatic forces act to enable formation of a more specific complex. These short-range forces include hydrophobic interactions, hydrogen bonds, dipole-dipole interactions, and salt bridges. Thus, the specificity of the association depends on structural properties of protein-protein interfaces, which should be geometrically and chemically complementary. To have favorable free energy of interaction, the enthalpic contribution attributable to desolvation of amino acids, formation of novel H-bonds, and van der Waals and electrostatic interactions should offset the decrease in entropy from the loss of translational and rotational degrees of freedom upon binding.

Depending on the type of the organism these forces have different contribution to the Pc and PSI docking.

Based on the laser-flash kinetic analysis, Hervas et al. (1995) proposed three different kinetic mechanisms for the Pc-PSI association (Scheme 1). The type I collisional mechanism exhibits a high rate constant such that the transient complex can not be measured kinetically. It is driven by long-range electrostatic forces and is ionic strength-dependent. This is the simplest mechanism of rigid body association and is observed for Pc/PSI interactions in many cyanobacteria, such as Anabaena. It was shown that a single arginine residue at position 88 in Anabaena Pc plays an important role in electrostatic steering of Pc to PSI (Molina-Heredia et al., 2001). By contrast, the more evolved type II mechanism, observed in P. hollandica, includes the formation of a detectable transient complex. The kinetics of complex formation are independent of ionic strength, suggesting that hydrophobic forces drive the protein association (Navarro et al., 2001). Finally, the reduction of PSI in chloroplast systems can be described by a type III mechanism that involves formation of a specific transient complex and its rearrangement before electron transfer. It was shown that salt bridges to the PsaF subunit of PSI play a role in this mechanism (Hippler et al., 1998).
To understand better the kinetics of the functional interactions between Pc and PSI, it is necessary to have the structure of their docking complex. The determination of the docking complex by x-ray crystallography and nuclear magnetic resonance (NMR) techniques is a formidable task because of the transient nature of the complex of the electron transfer proteins. Though the structure of the complex of Pc with the soluble part of cytochrome f was thoroughly studied by NMR (Ubbink et al., 1998) and molecular simulations (Ullmann et al., 1997), the high-resolution crystal structure of PSI was unavailable until recently (Fromme et al., 2001). The huge size of water-insoluble PSI makes the application of NMR and crystallography for structural determination of the docking complex far more difficult. In this work we use computational approaches for this purpose.

There are several methods available for the study of protein docking (Janin, 1995). Because molecular recognition consists of geometrical and chemical aspects, the computational algorithms for molecular recognition can be separated into geometry-based docking procedures that attempt to find the best steric fit between two molecules (GRAMM, Vakser and Aflalo, 1994; FTDOCK, Gabb et al., 1997), and approaches based on the minimization of the energy of interaction (AUTODOCK, Morris et al., 1998). The former perform exhaustive six-dimensional search of all possible conformations, the latter carry out statistical sampling of the conformational space. Some programs (DOT, Mandell et al., 2001) attempt to combine both approaches. All these algorithms have good predicting ability, as a root mean standard distance of the obtained complexes was within 2.5 Å from known crystal structures. The free energies of interaction were calculated using a novel molecular mechanics (MM)/Poisson-Boltzman, surface area (MM/PBSA) approach (Srinivasan et al., 1998a, b).

This paper provides an initial attempt to determine the computational structure of the Pc-PSI complex competent for electron transfer. In addition, we used several docking methods to obtain reliable docking structures. Overall, this enabled us to evaluate the efficacy of each method.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSION REFERENCES

Structure of PSI

Gene fragments encoding for PsaA and PsaB subunits of PSI were isolated by PCR with the following primers: for PsaA: 5'CTACACCGCTTGGCTATCGC, 3'GGACTCAATCAGCTCTTGCC; for PsaB: 5'CCCAAGGCCGTGCAGTGGGTG, 3'CCGTTGATCAACTGGGCC. The gene fragments were cloned into the pCRT7/CT TA cloning vector (Invitrogen, Carlsbad, CA) and then sequenced by automated methods (Cleveland Genomics, Cleveland, OH). The gene sequences were submitted to GenBank under accession number AY026898. The sequence alignment of the lumenal loops responsible for docking of Pc was carried out using the ClustalW program (Thompson et al., 1994). Based on the high degree of sequence identity (Fig. 1), it can be assumed that P. hollandica PSI has the same structure as the PSI in other related cyanobacteria. The 2.5-Å resolution crystal structure of S. elongatus was used as a basis for our homology modeling (Jordan et al., 2001). The A/B-ij helices were used as an input for docking programs. P. hollandica amino acid residues varying from the S. elongatus structure were manually replaced using Swiss-PDB Viewer. GRAMM, FTDOCK, DOT, and AUTODOCK were used to obtain reliable structures of the docking complex, and the parameters for each computation are provided below (see Results and Discussion). The best structure was used for free energy calculations.

View larger version (38K): [in this window] [in a new window]   FIGURE 1   Sequence alignment of the A/B-ij loops of the PsaA and PsaB subunits for the following cyanobacteria: S. elongates (high-resolution crystal structure available), P. hollandica, Synechocystis sp. 6803, Anabaena variabilis, and Nostoc. The Asn633 from PsaB involved in hydrogen bonding with Tyr12 Pc is italicized and underlined.

Molecular dynamics

Only the lumenal loop region of PSI (residues 616 to 668 from the PsaA subunit and 603 to 648 from the PsaB subunit) was used in molecular dynamics studies to decrease the number of atoms. The ends were fixed using the BELLY option (Case et al., 2000) to preserve the conformation of the docking site. The structure and charges of Pc were described earlier (Babu et al., 1999). To understand the role of the unique Tyr¹² and Pro¹⁴ residues, molecular dynamics were carried out for the computed complex of PSI with the wild-type (WT) Pc, as well as the Tyr¹²Gly and Pro¹⁴Leu mutants; the resulting coordinates were generated by Swiss-PDB Viewer. All energy minimizations and molecular dynamic simulations were carried out using the SANDER module of AMBER 6.0 package of programs (Case et al., 2000). The complex was solvated with a water box protruding for 10 Å in each direction from the molecule. The long-range electrostatics were treated with the particle mesh Ewald method (Darden et al., 1993). The SHAKE option (Ryckaert et al., 1977) was used to constrain all the bond length allowing for 2.0 fs time step. Nonbonded van der Waals interactions were cutoff beyond 8.5 Å. The complex was minimized with steepest descent and the water box equilibrated for 30 ps at 300 K, keeping the complex fixed. Next, the complex was minimized and the production run for 100 ps at 300 K was for data collection. Snapshots were recorded every 100 fs. Every fifth snapshot of the last 80 ps was used for energy calculations. The free energies of interaction were calculated using the MM/PBSA approach (Srinivasan et al., 1998a,b). The MM energies were calculated with the ANAL module of AMBER 6.0. The electrostatic contribution to solvation free energy was calculated using DELPHI (Honig and Nicholls, 1995; Gilson and Honig, 1998). The interior dielectric constant of 1 was used for the protein because the AMBER force field was parameterized with dielectric constant 1 (Wang and Kollman, 2000). The exterior dielectric constant was set to 80 for the solvent. The solvent-accessible surface area was obtained with MSMS (Sanner et al., 1996) from which the nonpolar contribution to solvation energy was estimated from the following dependence G_nonpolar = 0.00542 * solvent-accessible surface area + 0.92 kcal/mol (Sitkoff et al., 1994).

	RESULTS AND DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSION REFERENCES

Criteria for the docking complex

As there is no evidence that there is any significant conformational change upon binding, Pc and PSI are treated as rigid bodies to allow for the application of docking algorithms. Furthermore, such algorithms currently can not treat large-scale structural changes. It has been shown that electron transfer to PSI occurs through the solvent-exposed nitrogen of P. hollandica Pc His⁸⁵ at the hydrophobic patch, involved in the docking interaction with PSI (Haehnel et al., 1994). The two -helices of the two PsaA and PsaB subunits at the docking pocket of the PSI have two conserved tryptophan residues (PsaA Trp⁶⁵⁵ and PsaB Trp⁶³¹) that come together into van der Waals contact right below the edge of special pair of chlorophylls P700. It was shown that mutation of one of these tryptophans affects the interaction between PSI and the alternative, isofunctional electron donor cytochrome c₆ (Sun et al., 1999). The mutation of Trp⁶²⁷ from PsaB of Chlamydomonas reinhardtii (analogous to Trp⁶³¹ in S. elongatus) completely abolished formation of complex between Pc and PSI and significantly reduced the rate of electron transfer between these proteins (Sommer et al., 2002). These data suggest that these tryptophans lie on the electron transfer path from the copper to P700 and that the Pc should be docked with His⁸⁵ in proximity to these tryptophans to minimize the donor/acceptor distance. This biological information was the main constraint for the complex selection. Additionally, the unique structure of the P. hollandica hydrophobic patch makes the interaction between Pc and PSI more specific. This implies that the docking site should also accommodate bulky Tyr¹² of P. hollandica. Transfer of Pc hydrophobic patch from solvent into the docking site of PSI displaces protein-bound water molecules into the bulk solvent. This induces local rearrangement in water hydrogen-bonding network, resulting in increase of entropy (Head et al., 1996). The amount of liberated water is proportional to the interface accessible surface area. The interface accessible surface area also reflects the steric fit between proteins. The study of protein-protein interfaces (Lo Conte et al., 1999) reveals that the average size of the recognition site is ~1600 ± 400 Å², suggesting that the bigger the area, the more stable the complex. In the context of the current study, it has been shown previously that P. hollandica Pc forms hydrophobic complex with PSI stable enough to be detected by kinetically (Navarro et al., 2001). Thus, our docking complex should possess the following characteristics: the shortest metal-to-metal distance with His⁸⁵ located below P700, a cleft to accommodate the protruding tyrosine residue and the highest area of interface among the obtained complexes.

Comparison of the docking algorithms and resulting structures

All algorithms for the search of the best geometric complementarity used in this work (GRAMM, FTDOCK, DOT) are based on the molecular recognition algorithm developed by Katchalski-Katzir et al. (1992). It estimates surface complementarity between two proteins treated as rigid bodies. The atomic coordinates of the two proteins obtained from PDB files are projected onto a three-dimensional grid, yielding a digital representation. Small positive numbers are assigned to the surface of the bigger molecule (receptor) and large negative numbers are assigned to its interior to penalize for penetration in the core of the protein. Next, the smaller molecule (ligand) is translated and rotated around the receptor searching through all the conformational space in six dimensions. At each rotational step the correlation function using Fourier transformation is calculated. The correlation function evaluates the degree of the geometric match between two molecules. Thus, the best geometric fit yields the highest score, and low scores represent the poor matches, as a result of penetration in the interior. In all the methods the PSI (receptor) was held fixed and the Pc (ligand) was manipulated to explore all possible orientations.

Global range molecular matching (GRAMM)

The GRAMM algorithm is the program for docking of the protein structures of varying accuracy. Although high-accuracy structures provide high-accuracy complexes, a large number of possible conformations between two proteins results in increased number of false-positive matches and increased computational time. Based on the fact that protein-protein interfaces are more hydrophobic (Jones and Thornton, 1996) than the rest of the protein surface, the simplified approach, called hydrophobic docking, was proposed by Vakser and Aflalo (1994). This approach exhibits a higher signal-to-noise ratio and decreases computational time. Also, to overcome the problem of conformational changes of amino acid side chains affecting the docking prediction (Vakser, 1996c), and the problem of inaccuracies in the protein structures, a low-resolution algorithm was developed (Vakser 1995, 1996a; Vakser and Nikiforovich, 1995). It was shown that docking of the molecules lacking high-resolution details (<7 Å) can overcome the multiplicity of the local minima and effectively find the global minimum (Vakser, 1996b, 1997; Vakser et al., 1999). The above mentioned geometry matching algorithm was interpreted in terms of the energy based on long-distance atom-atom potentials (Vakser, 1996a). In our studies we have used the high-resolution generic docking with grid step 2.5 Å, grid size 64 Å. The ligand was rotated with 10°-angle intervals. The energy score for repulsion was 30, the energy score for attraction was 1. The 100 structures that yielded the best score were selected and analyzed visually. Surprisingly, the structure of the docking complex that best incorporated all the above-mentioned criteria was the first one with highest best energy score (146).

FTDOCK

FTDOCK is another method that evaluates not only shape complementarity, but also electrostatic complementarity (Gabb et al., 1997; Aloy et al., 1998). First, it performs the global search using a slightly modified molecular recognition algorithm, which takes coulombic interactions into account. The obtained docking structures are ranked by empirical residue level pair-pair potentials, which reflect observed amino acid contacts between two proteins in nonhomologous complexes obtained from the solved crystal structures (Moont et al., 1999). Next, structures are filtered by the distance constraints using available biological information. To take into account the conformational changes of amino acid side chains, protein interfaces are refined by selecting the most probable position of a side chain from a rotamer library and performing energy minimization (Koehl and Delarue, 1994; Jackson et al., 1998). The PSI and Pc were projected onto 126 × 126 × 126 grid with a 0.985 grid step and a global surface thickness of 1.3 Å. The rotation angle step was made 10° to make it consistent with GRAMM. 15,840 rotations were evaluated in total. The generated complex structures were ranked using pair potentials (RPscore), and the 6-Å distance constraint between His⁸⁵ and the Trp analogous to PsaB Trp⁶³¹ in S. elongatus was used to filter the correct complex. After such screening, only eight structures with the very low pair potentials score of 0.140 corresponded to our criteria.

DOT (Daughter of TURNIP)

DOT is a program that combines both geometry and energy minimization approaches. The van der Waals energy is obtained from the geometric matching algorithm mentioned above (Mandell et al., 2001). The solvent continuum electrostatic model is used to calculate the electrostatic energy by solving the Poisson-Boltzmann equation. The 128 × 128 × 128 potential grid with a 1-Å step was generated for PSI using University of Houston Brownian Dynamics program (Davis et al., 1991; Madura et al., 1995). The solvent dielectric was set to 80, the protein dielectric 4, temperature 300 K, solvent and ionic radius 1.4 Å. From the composite energy term (the sum of electrostatic and van der Waals energies) the partition function was computed to derive the free energy of interaction. The free energy of interaction for the best complex structure was 11.7 kcal/mol, which is in the same order as the experimental values calculated from the binding constant (5.7 kcal/mol) (Navarro et al., 2001).

AUTODOCK3.0

AUTODOCK is an example of a program that attempts to find the complex with minimal interaction energy. Earlier versions used Monte Carlo simulated annealing to sample the docking conformations (Goodsell and Olson, 1990; Morris et al., 1996). Later, a Lamarckian genetic algorithm was introduced, which was shown to be more efficient than the simulated annealing (Morris et al., 1998). This algorithm was used in our simulation of the Pc-PSI association. The Pc was initially placed 5 Å away from the docking site of PSI. In this method, the ligand (Pc) is represented as a chromosome and its translation and orientation toward PSI is represented as genes in that chromosome. The atomic coordinates of the protein in the docking complex represent its phenotype. A random population of 75 individuals was generated and the interaction energy (fitness) of ligand with receptor (PSI) was calculated. The crossovers and mutations introduced randomly in the population and the individuals with the best fitness were selected to produce the next generation. The rate of cross-over was set to 0.8 and the rate of mutation to 0.04. The translational step was equal to 1 Å, and the rotational step equal to 5°. In the Lamarckian algorithm, the best individuals undergo a local search, which is analogous to energy minimization and is based on the Solis and Wets algorithm (1981). The local search finds the local minimum (best fitness) and then the position of the ligand (phenotype) is converted back to translational and orientation values (genotype). Because of the size of the system, the number of energy evaluations was reached faster and the next run was initiated. Total number of runs performed was 20. The results were clustered into several groups with a 1.0-Å cluster tolerance. Of 20 structures obtained, 9 structures clustered in one group, with a mean docked energy of 99.8 kcal/mol, matched our criteria for docking complex.

Analysis of docking complexes

The best structures obtained by above-mentioned algorithms were visualized and compared with one another (Fig. 2). The protein-protein interfaces of the docking complexes were further analyzed by the Protein-Protein Interaction Server (http://www.biochem.ucl.ac.uk/bsm/PP/server; Jones and Thornton, 1995, 1996). The results are presented in Table 1. All complexes have almost the same ratio of polar to nonpolar amino acids at the interface. The prevalence of nonpolar amino acids is the characteristic trait of the docking interface (Sheinerman et al., 2000). The highest interface-accessible surface area (Hubbard, 1992) and the shortest metal-to-metal distance necessary for effective electron transfer was yielded by the structure obtained by the GRAMM algorithm. The high area of the surface contact is probably attributable to interaction with a -sheet region of the PsaA loop (Figs. 1 and 2). Also in this structure, a hydrogen bond of 2.9 Å is identified between Tyr¹² of Pc and Asn corresponding to PsaB Asn⁶³³ in S. elongatus. The Pc His⁸⁵ involved in electron transfer is at the van der Waals distance to the pair of stacked tryptophans homologous to Trp⁶⁵⁵ of PsaA and Trp⁶³¹ of PsaB, suggesting their possible interaction (Fig. 3). These data render the GRAMM-derived structure the best of the four obtained complexes. Thus, this structure was used for the free energy calculations. Last, from root mean standard distance it can be inferred that other structures are 6-8 Å from the GRAMM, mostly because of rotation around C₂ axis of symmetry of the PSI.

View larger version (116K): [in this window] [in a new window]   FIGURE 2   The best structures for the complex of Prochlorothrix Pc with PSI obtained by: (A) GRAMM; (B) FTDOCK; (C) DOT; and (D) AUTODOCK. The side chains are not shown, except for tyrosine at position 12 of the Pc. Pc and residues PsaA Trp655 and PsaB Trp631 are shown in yellow. The copper center, Pc His85, and the PSI core backbone are in red

View larger version (94K): [in this window] [in a new window]   FIGURE 3   Structure of the Pc-PSI complex obtained by GRAMM: front and side view. The van der Waals surface is shown by dots. The His85 of the Pc is right below pair of stacked Trp (analogous to Trp655 for PsaA and Trp631 for PsaB) adjacent to special pair of chlorophylls. The metal centers are in purple. Also shown is the hydrogen bond between Asn633 and Tyr12.

It should be noted that the stability and specificity of complex should be balanced in such a way that will allow efficient electron transfer and rapid dissociation at the same time, because stable complexes can sometimes be trapped in unproductive local minima. From this point of view, the structure obtained by DOT program can be optimal, because it has less surface contact with PSI and is thus less stable. Nevertheless, in DOT, FTDOCK, and AUTODOCK complexes, Pc His⁸⁵ is displaced from the suggested electron transfer pathway (Fromme et al., 2001). For AUTODOCK and FTDOCK complexes there was a gap between van der Waals surfaces of Pc His⁸⁷ and the stacked tryptophans, which would affect electron transfer. These complexes could possibly represent an initial recognition complex.

Free energy of interaction

The evaluation of the free energy of interaction by rigorous approaches, involving free energy perturbation and thermodynamic integration, is computationally intensive. To decrease the computation time a new approach, MM/PBSA, was proposed by Kollman and colleagues (Srinivasan et al., 1998a,b). The free energy (G_bind) is evaluated as a sum of the free energy of interaction in gas phase (G_gas) and free energy of solvation (G_sol). The gas phase free energy is a sum of electrostatic (G) and van der Waals (G) energies calculated from MM. The solvation free energy is composed of nonpolar solvation energy (G) estimated via accessible surface area and electrostatic solvation energy (G) calculated by solving the Poisson-Boltzman equation with DelPhi program. This is represented in the following thermodynamic cycle (Scheme 2; Kollman et al., 2000).
Thus, G_bind = G_gas + G  G  G  TS, where G_gas = G + G, and G_sol = G + G. Because we are interested only in relative binding free energies of WT and mutant proteins, the entropic contribution to free energy is not taken into consideration. Assuming that no conformational change occurs upon docking of Pc to PSI, the energies for complex, ligand, and receptor were evaluated from the trajectory of a complex, instead of using separate trajectories. This approach has proven to yield accurate predictions when compared with experimental data (Massova and Kollman, 1999). The inadequacies of the model are cancelled when the difference of the free energies is calculated (Wang et al., 2001). The results of the free energy calculations for the complex of PSI with WT, Tyr¹²Gly, and Pro¹⁴Leu mutants are presented in Table 2. The free energy of 174.3 kcal/mol for the WT can be compared with those obtained for the complex of Pc with cytochrome f in related studies, 285.3 kcal/mol (De Rienzo et al., 2001) and 355.6 kcal/mol (Ullmann et al., 1997). The relative free energy of the Tyr¹²Gly mutant is 6.7 kcal/mol lower than the WT, which is the result of lower solvation energy (302.4 kcal/mol compared with 308.8 kcal/mol of the WT). The presence of Tyr¹² also increases nonpolar solvation energy. In agreement with this calculation, it has been recently determined experimentally that this mutant has a modestly higher binding constant (Navarro et al., 2001).

View this table: [in this window] [in a new window]   TABLE 2   Free energies of interaction between Pc and PSI calculated by MM/PBSA. WT, wild-type Pc, Y12G  Tyr12Gly12 mutant, P14L  Pro14Leu14 mutant

It is worth noting that the internal electrostatic energy (379.7 kcal/mol) and electrostatic solvation energy (340.2 kcal/mol) are significantly different for the Pro¹⁴Leu mutant than for the WT (363.6 and 323.4 kcal/mol, respectively). This fact can account for the unique reactivity of this mutant in the electron transfer. It was shown by laser-flash photolysis experiments that it has a threefold higher electron transfer constant than the WT (Navarro et al., 2001). It was suggested earlier that the replacement of Pro¹⁴, which has a rigid backbone, with Leu affects the flexibility and geometry of the copper site and the reorganization energy, making Pro¹⁴Leu mutant a better electron donor (Navarro et al., 2001). Thus, the results of the free energy calculations are in reasonable agreement with the experimental observations, further suggesting that the docking structure obtained is an adequate representation of the functional complex.

	CONCLUSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSION REFERENCES

In this work, a combination of docking algorithms, molecular dynamics, and free energy calculations using MM/PBSA was presented. All algorithms have predicted complexes that are very similar, but the structure of a complex that best corresponded to the available biological information was obtained by the GRAMM algorithm, which was the fastest and easiest compared with the others. This work showed that geometric algorithms, which deal mostly with van der Waals interactions, were more effective in prediction than those concerned with energy considerations. This can probably be explained by the hydrophobic nature of the complex of interest. The obtained docking structures are slightly different from one another, but they still can be functionally effective because of the transient nature of the complexes formed by electron transfer proteins. The free energy calculations provided useful insights into experimental data obtained earlier, which in turn speak for the adequacy of the predicted complex. This work also suggests, in concert with others (Fromme et al., 2001), that the pathway of electron from Pc His⁸⁵ to the special pair of chlorophylls in PSI could pass through the pair of tryptophans from PsaA and PsaB subunits stacked at the van der Waals distances. Finally, it should be noted that the prediction of the docking complex was made possible by the unique surface conformation provided by Tyr¹² of the Prochlorothrix Pc hydrophobic patch. This structure can thus be extended to understanding the docking mechanism seen in other Pc/PSI reaction pairs.