Electrostatic interaction between redox cofactors in photosynthetic reaction centers

Intramolecular electron transfer within proteins is an essential process in bioenergetics. Redox cofactors are embedded in proteins and this matrix strongly influences their redox potential. Several cofactors are usually found in these complexes and they are structurally organized in a chain with distances between electron donor and acceptor short enough to allow rapid electron tunneling. Amongst the different interactions that contribute to the determination of the redox potential of these cofactors, electrostatic interactions are important but restive to direct experimental characterization. The influence of interaction between cofactors is evidenced here experimentally, by means of redox titrations and time-resolved spectroscopy, in a chimeric bacterial reaction center (Maki et al . (2003) J. Biol. Chem. , 278(6):3921-3928) composed of the core subunits of Rubrivivax gelatinosus and the tetraheme cytochrome of Blastochloris viridis . The absorption spectra and orientations of the various cofactors of this chimeric reaction center are similar to those found in their respective native protein, indicating that their local environment is conserved. However, the redox potentials of both the primary electron donor and its closest heme are changed: the former is downshifted in the chimeric reaction center when compared to the wild type, conversely the latter is upshifted. We propose a model in which these reciprocal shifts in the midpoint potentials of two electron transfer partners are explained by an electrostatic interaction between them.


Introduction
Proteins exert a fine electrochemical tuning of the redox potential of the cofactors they bind in order to perform the various electron transfer reactions that are involved in biological processes. As a famous example, the redox potentials (E m ) * of c-type cytochromes are tuned by more than 500 mV (see (1) and references therein). The physicochemical basis of such wide range modulations have been rationalized by various authors who all agree that the protein medium has the unique property of providing a dielectric environment in which the redox cofactors are embedded with an intricate charge or dipole distribution (1)(2)(3). From a general standpoint, the free energy difference between the oxidized and reduced forms of any redox cofactor in a protein is the sum of several terms. Amongst these are ∆G conf and ∆G el .
∆G conf accounts for any conformational change that the change in the redox state of the cofactor may induced (including proton or ion binding or release). ∆G el results from the electrostatic potential at the cofactors resulting from the individual charged groups and the permanent dipoles within the protein. Estimating the respective values of these different terms is a difficult task, yet their sum is readily accessible experimentally since it can be obtained by comparing the absolute values of the E m in solution and in the protein. The latter may be obtained by two different methods. The most commonly used one is equilibrium redox titration. The other relies on the determination, by a functional analysis, of the free energy change associated with an electron transfer reaction between two cofactors. Such a change in free energy is generally considered as equal to the difference in E m 's between the electron donor and acceptor. Thus, provided one of these two redox potentials is known, the other one is readily inferred. However, these two methods sometimes yield different results and this has led to the distinction between 'equilibrium redox potential' and 'operating redox potential'.
These differences arise because the two methods do not probe the same state of the redox 4 cofactors. Two types of phenomena may account for these differences. One comes from the distinct time-domain involved in equilibrium redox titration and functional analysis. Whereas redox titrations require thermodynamic equilibrium between the sample and the solution poised at a given potential, the functional analysis allows one to probe transient states whose free energy may differ significantly from that of the equilibrated states. Indeed, in response to the change in the redox state of a given cofactor, the protein environment may undergo energetic relaxation (such as proton transfer, conformational changes) which may be slower than the lifetime of the transient oxidized (or reduced) cofactor. Armstrong and colleagues have nicely illustrated this point with the 'fast-scan electrovoltammetric' technique (4). Other examples are found in photosynthetic reaction centers (RC) (see e.g. (5,6)). An alternative but non exclusive explanation of the different E m 's yielded by redox titration and functional analysis relies on the fact that most of the membrane proteins involved in electron transfer reactions bind several cofactors which are usually located at less than 15 Å one from the other. Such short distances may result in significant electrostatic interactions between the different cofactors. Thus, for a given cofactor, ∆G el includes the electrostatic contributions of the nearby electron carriers. Such a contribution has been nicely illustrated in the case of the tetraheme cytochrome of Blastochloris (formerly Rhodopseudomonas) viridis (7,8). It is noteworthy however that throughout a redox titration, all the cofactors undergo an identical charge change in terms of sign (i.e. all are either reduced or oxidized), whereas in an electron transfer chain, two nearby cofactors involved in an electron transfer reaction will undergo charge changes of opposite sign (one will be oxidized at the expense of the other).
Consequently, if the electrostatic interaction between the two is significant, the difference between their equilibrium E m 's will be greater than the free energy change associated with the electron transfer between them. In this paper we will illustrate the importance of such electrostatic interactions in electron transfer chains. Although electron transfer chains 5 embedded in a single protein are found in many biological pathways, the photosynthetic chains are ideally suited for such studies. Indeed not only do they allow redox titration of the various redox cofactors but also the kinetics of electron transfer reaction may be characterized with an unequalled time-resolution. In particular, the bacterial photosynthetic RC of Blc.
viridis and its associated tetraheme cytochrome have been intensively studied. Its three dimensional structure has been solved (9) and the spectroscopic or redox properties of the various cofactors are known (see (10) for a review). Further, the different electron transfer steps have been characterized (11). In membrane fragments as well as purified RC of Blc.
viridis the reduction of the oxidized primary electron donor P + by the closest heme, c 559 , is multiphasic (12). Interestingly, this feature has been interpreted along the lines of either of the two phenomena which has just been described: a conformational heterogeneity yielding a distribution of substates (11) or a low equilibrium constant of this electron transfer reaction (13). To reconcile this latter hypothesis with the equilibrium constant of ~100 expected from the difference in midpoint potentials of the P + /P and c 559 + /c 559 , Baymann and Rappaport proposed that an electrostatic interaction between P and its closest heme raises the redox potential of the P + /P couple (13). As discussed below, the present results support this hypothesis.
From an experimental standpoint Blc. viridis has the drawback of being unable to grow heterotrophically making the mutagenesis approach uncertain, despite few successful attempts. Recently, Maki et al. (14) succeeded in transferring the membrane bound cytochrome of Blc. viridis to another bacterium, Rvi. gelatinosus, in which mutagenesis strategy may be planned (15,16).

Experimental procedures
Interspecific replacement of the gene coding for the RC-bound cytochrome subunit in Rvi.
gelatinosus, used to produce VC-F mutant, is described in (14). The site directed substitution, by a two-step PCR method, of Arg204 for a Leu in the cytochrome subunit of the VC-F strain is described in (17). 7 Cells of the VC-F mutant of Rvi. gelatinosus were grown for 24 h, in the light and in anaerobic conditions in Hutner medium with 50 µg.mL -1 kanamycin and 20 µg.mL -1 ampicilin. For membrane preparations, cells were harvested by centrifugation at 4000 g for 10 min, resuspended in 20 mM Tris-HCl (pH 7), and disrupted by French Press at 50 MPa.
The remaining intact cells were separated from the membrane supernatant by centrifugation at 10000 g for 10 min. Then, the membrane fragments were collected after centrifugation at Redox titrations were performed as in (13,19)   in Blc. viridis could be satisfyingly fitted with a Nernst curve with E m = +500 mV. The P + /P couple in Rvi. gelatinosus titrates with an E m of ~ +400 mV, consistent with previous findings (22), but, interestingly, the titration curve showed a slight deviation to a one-electron Nernst equation (dotted line). In the VC-F chimeric RC the results were significantly different from that in Rvi. gelatinosus WT: i) the deviation to the one-electron Nernst equation (dotted line) was more pronounced, ii) the overall midpoint potential was lower by about 50 mV. µs after the flash, this fraction of long lived P + was only 3 % when the low potential hemes were reduced. Under these latter conditions the free energy change associated with P + reduction is expected to be large owing to the low midpoint potentials of the electron donor (see figure 1). Thus, after this equilibrium is reached (i.e. in the hundreds of µs time range, see figure 1), the amount of P + remaining should be too low to be detectable. Conversely, in the eventual RC's devoid of the tetraheme subunit, P + is expected to decay via charge recombination i.e. in the tens of ms time range (23). Thus the amount of P + still detectable at   Figure 4B). Polar plots of EPR signal amplitudes allowed one to orient the g z of 3.1 along the 0° axis, whereas the maximum of the g z = 3.3 line was found at 45°

Redox characteristics of the P + /P couple in the WT and chimeric RCs
( Figure 4C). The g z vector of a heme being perpendicular to the porphyrin ring, we concluded that the highest potential heme c 559 was almost perpendicular (90°) to the membrane plane, as for Bcl. viridis (24) (compare Figure 4C to Figure 7 in (24)). Moreover, the absorption spectra of the four hemes bound to the tetraheme subunit are conserved in the chimeric RC with respect to the WT Blc. viridis RC (see Figure 5A for a comparison of absorption changes associated with the oxidation of the c 556 and c 559 hemes in both strains). Furthermore, the E m 's of the three lower potential hemes embedded in the cytochrome subunit are remarkably conserved when compared to those found in Blc. viridis (14). Based on the sensitivity of both the spectroscopic and redox properties of a redox cofactor to any significant alteration of its protein environment, we take these results as a strong indication in favor of a conserved structure.
As a third hypothesis which would account for the singular titration curve of the P + /P couple in the VC-F strain, we would now like to consider the existence of a significant electrostatic interaction between the BChl dimer and its nearest redox-active neighbor: the c 559 heme. Such an interaction between two close redox cofactors implies that the E m of either one of the two cofactors depends on the redox state of its neighbor. In the present framework, one should thus consider the following scheme † † : where E P(c + ) and E P(c) are the E m of the P + /P couple in the presence of the oxidized and reduced cytochrome, respectively, ∆Ψ is the electrostatic interaction between both cofactors and E c(P + ) and E c(P) are the E m of the cytochrome in presence of P + and P, respectively. The equation used to fit the titration curve is then derived from both following points: 1. The total amount of oxidized P + or of reduced P can be written as :   12 This equation allows one to distinguish several cases: i) each of the two extreme ones where E P(c) << E c(P) or E c(P) << E P(c+) yields a redox titration which follows a one electron Nernst curve (with E m =E P(c) and E m =E P(c) + ∆Ψ , respectively). This reflects the fact that the redox changes of each cofactor occur in well-separated redox potential ranges, they thus do not interfere. A third, and experimentally more interesting, case arises when E P(c) E c(P) or E c(P) E P(c + ) since, according to the above equation, the titration curve should be markedly different from a classical Nernst curve. This latter case applies to the VC-F strain. The data presented in figure 2 could be satisfyingly fitted with equation 12 (Figure 2 solid lines), with E c(P) , E P(c) and ∆Ψ as varying parameters. In the VC-F strain the best fit parameters were: E c(P) = +380 mV, E P(c) = +350 mV and ∆Ψ= +50 mV. In the Rvi. gelatinosus RC, the fit with equation 12 yielded: E c(P) = +300 mV, E P(c) = +350 mV and ∆Ψ= +50 mV. In Blc. viridis, because the E m 's of the two interacting cofactors belong to more distinct redox potential regions, the deviation to a one electron Nernst curve was, as expected, less pronounced. The data could be satisfyingly fitted either with a one electron Nernst curve with E P = +500 mV or according to equation 12 with E c(P) = +380 mV, E P(c) = +450 mV and ∆Ψ= +50 mV. It is noteworthy that the values found for the E m of the P + /P couple as well as for ∆Ψ were similar in the chimeric and 'parent' RC from Rvi. gelatinosus. Yet, one could argue that the value found here for the E m of the P + /P couple in Rvi. gelatinosus WT is 50 mV smaller than the previously reported one of 400 mV. This is not unexpected since, in the WT, the E m of the closest heme to P is significantly lower than that of the P + /P couple so that the respective oxidation (or reduction) of the two cofactors occurs in distinct redox potential domains.
Consequently, the most significant fraction of P should be oxidized in the presence of the oxidized cytochrome and the titration curve should yield, as a first approximation, a E m of E P(c) +∆Ψ = 350+50 = +400 mV, in good agreement with previous reports (22). 13 To further test this interpretation, we have measured the equilibrium redox titration of the P + /P couple in a site-directed mutant of the VC-F strain in which the E m of the c 559 heme is expected to be decreased, thereby increasing the gap between the E m 's of P and its closest heme. According to the calculation of Gunner and Honig, the presence of a positively charged arginine residue (R264) in the vicinity of the c 559 heme of the Blc. viridis cytochrome subunit contributes to raise the midpoint potential of this redox center (7). This was confirmed by Chen et al. who substituted, by site-directed mutagenesis, this residue for a lysine, and found a downshifted E m for the c 559 by about 110 mV (26). We have also targeted this Arg and substituted it for a Leu in the gene coding for the tetraheme subunit in the VC-F strain (17).  Nernst that witnesses the electrostatic interaction between two close redox centers becomes less prominent when increasing the difference between the E m of the two centers.
As a further support to our model model, the E m 's of the hemes that were found to interact electrostatically with P when fitting the data with equation 12 (+300 mV for Rvi. gelatinosus WT and +380 mV for the VC-F strain) are consistent with the figures found in the literature for the closest hemes in the Rvi. gelatinosus or Blc. viridis cytochrome subunit, respectively (see Figure 1 for the data on the redox centers in the RC of these two species). We thus 14 propose that the peculiar redox titration curve observed for the P + /P couple reflects a significant electrostatic interaction between P and the c 559 heme. One also expects the redox titration of the c 559 + /c 559 couple to be reciprocally affected by this interaction. To test this hypothesis we performed the dark equilibrium titrations of both c 559 and c 556 hemes in Blc.
viridis and VC-F RC's.  To check the involvement of the c 559 in the oxidation of the c 556 heme by P + , we measured the kinetics of absorption changes poising the redox potential at +380 mV ( Figure 6B). Under these conditions, owing to their respective E m 's, a significant fraction of RC's should be in the Pc 559 c 556 + state (~32 %) before the actinic flash. In those RC's the electron transfer step between P + and the c 559 heme should be clearly detectable. At 100 and 200 ns, the broad absorption increase reflects the photo-oxidation of P (see (13)). In the 500 ns to 5 µs time 16 range, a trough is observed with a minimum at 559 nm, suggesting the oxidation of the c 559 heme by the P + . From 10 µs to 50 µs the trough shifts to a lower wavelength, consistent with inter-heme c 556 -c 559 electron transfer. The kinetics monitored at various wavelengths was globally fitted with two exponentials. The half times of the fast and slow phases were found to be 700 ns and 11 µs, respectively.

Redox characteristics of the exogenous cytochrome
One expectation that ensues from the redox titration is that the oxidation of the c 559 by P + should be energetically uphill. This prediction is supported by the present data. The fact that the c 559 oxidation step cannot be easily discriminated from the c 556 oxidation step suggests that the fraction of centers in the Pc 559 + state in equilibrium with the P + c 559 state is small. In the likely kinetic scheme describing the overall electron transfer reaction between P + and the c 556 heme, one can consider a two step process, where k 1 , k-1 , k 2 and k-2 are the forward and backward absolute rate constants of steps 1 and 2, respectively: According to the E m of the three redox cofactors, the first step is energetically uphill and the second one is downhill. If, in addition, as in the case of Blc. viridis, k 2 <k 1 (see figure 1), one expects the transiently formed Pc 559 + c 556 to be hardly detectable. Interestingly, decreasing the relative amplitude of the second step by e.g. increasing the redox poise in order to oxidize the c 556 heme should enhance the relative amplitude of the first step, facilitating its detection in good agreement with the present data.
The unusually slow oxidation of the c 556 heme found in the VC-F RC with respect to the parents RC also supports the occurrence of an uphill step in the overall electron transfer from the c 556 to P + via the c 559 heme. In Blc. viridis the apparent rate of the oxidation of the c 556 heme is ~ 2.7 10 5 s -1 (11,13), to be compared with the 6.3 10 4 s -1 value found in the chimeric RC. As a first approximation, the apparent rate of oxidation of the c 556 heme is, according to the above scheme, k app = k 2 /(1+k -1 /k 1 ). Assuming that the absolute rate of electron transfer between the two hemes (k 2 ) is unchanged in the chimera relative to Blc. viridis (an assumption which is supported by the above conclusion on the structural close similarities between both RC's), and taking k-1 /k 1 to 10 30/60 = 3.2 (corresponding to the ∆E m between P and c 559 of -30 mV found here); one expects the apparent c 556 oxidation rate to be k app = 6.4 10 4 s -1 , in agreement with the value of 6.3 10 4 s -1 found here.

Discussion
Two main features arise from the combined study of the chimeric RC by either equilibrium redox titration or kinetic analysis: whereas the physico-chemical properties of the hemes embedded in the cytochrome subunit are remarkably conserved with respect to the parent species Blc. viridis, the E m of the P + /P couple is down-shifted in the chimeric RC. Further, an electrostatic interaction of 50 mV raises the E m of the P + /P couple when the c 559 is oxidized.
Although the situation where this interaction can be directly observed, as in the present case, by the deviation to a one-electron Nernst curve is singular, we would like to argue that the existence of electrostatic interaction between nearby redox centers is likely to apply as well in other multi redox cofactors membrane proteins. Amongst these, the "parent" RC's of the chimera should be first considered. Interestingly, the redox titration of the P + /P couple in WT Rvi. gelatinosus also displayed a significant deviation to a n=1 Nernst curve. As discussed above, the deviation can be accounted for by an electrostatic interaction of 50 mV between P and the closest high-potential heme. Further it is noteworthy that the less pronounced deviation in the WT Rvi. gelatinosus with respect to the chimeric RC is not an indication of a smaller electrostatic interaction. On the contrary, it can be rationalized by the combined effect of a larger difference in the E m of P and its closest heme and a significant electrostatic interaction between these two redox centers. This point is illustrated by the disappearance of 18 the deviation to a one electron Nernst curve and a higher apparent E m for P in the R264L mutant of the VC-F RC where the E m of the closest heme to P has been downshifted. The E m of the c 559 heme being lower in the R264L mutant than in the native VC-F, the redox changes of P occur in a potential range where the closest heme is fully oxidized. As a consequence the electrostatic interaction between these two redox centers contribute to upshift the apparent E m of P when compared to the native VC-F RC. In Blc. viridis, the resolution of the three dimensional structure (9) and the determination of the order of the hemes in the cytochrome subunit (24,27,28) grounded several studies which emphasized the role of inter-cofactors electrostatic interaction. Based on the finding that the oxidation of the outermost solventexposed hemes was moderately electrogenic, Gao et al. estimated the interaction between P + and the c 559 heme in Blc. viridis to be 30 mV (29), in qualitative agreement with our finding.
When their three dimensional structure is known, proteins are liable to electrostatic calculation. The Poisson-Boltzmann equation may be solved numerically and yield electrostatic interaction between cofactors (see e.g. (1,30,31)). Gunner and Hönig performed such calculations with the Blc. viridis RC and calculated the shift in E m 's undergone by the four hemes relative to solution (7). Electrostatic interactions between the four hemes of the tetraheme subunit ranged from 5 to 77 mV (7). The P + /P couple was not treated in this study, likely because of the lack of reliable redox titration of this cofactor in solution owing to its peculiar dimeric character. Yet, based on a kinetic analysis of the reduction of P + by the c 559 in whole cells, Baymann and Rappaport concluded that the equilibrium constant of this reaction was significantly lower than predicted from the difference in E m of the two couples.
The rationale for this apparent discrepancy was proposed to be an electrostatic interaction of 80 mV between these two cofactors (13).
As an example pertaining to respiratory chains the Succinate Quinol Reductase binds three iron sulfur clusters which are thought to act as the entry path for electron. Interestingly, as 19 determined by redox titration, the second cluster in the chain has a remarkably low redox potential (-250 mV) when compared to those of its two neighbors (-25 and -60 mV, (32)). If this E m corresponds to the operating potential of this cofactor, it would make its reduction highly endergonic (see (33) for a review). Although, an energetically uphill step in an overall exergonic chain is not, per se, to be excluded, it has been proposed that this highly reducing E m reflects an anticooperative electrostatic interaction between redox centers rather than the "true operating" midpoint potential of this singular cluster (34).
The suggestion reported here that the E m of a redox cofactor depends on the redox state of its neighboring redox cofactor raises the question of the dielectric properties of the protein medium. This question has been extensively studied and the emerging picture is that the screening of the electrostatic interaction between charges is best described by an effective distance-dependent dielectric constant. This phenomenological description was detailed by Warshel and coworkers (see e.g. (35)) and experimentally supported by various groups (see (36) for the Rhodobacter sphaeroides reaction center case and references therein). By studying the shift in E m of the P + /P couple in Rhodobacter sphaeroides reaction center induced by mutations of ionizable groups at selected sites, Johnson and Parson concluded that, although proteins are usually thought as low dielectric medium, electrostatic interactions resulting from a change in the charge distribution around P are efficiently screened with an average screening factor of about 40 (36). With an electrostatic interaction of 50 mV between P and its closest heme located at 20 Å (center to center) we get, by applying the Coulomb's law, an average screening factor of 16 which is significantly larger than the value of 2-4 often used in electrostatic calculations but yet significantly lower than the value found by Johnson and Parson, suggesting that the screening of buried charges is less efficient in the present case than in their work. Among the various reasons which may concur to this smaller screening factor are: i) the fact that we are dealing here with electrostatic interaction between redox 20 cofactors and not between ionizable side chains and a redox cofactor, ii) the likely larger shielding from the bulk resulting from the binding of the cytochrome subunit to the RC in the present case at variance with the Rhodobacter sphaeroides RC and/or, iii) the preservation of the membrane environment in the present study. In line with this, Maki et al. measured the redox potential of the c 559 heme in RC from the VC-F strain either embedded in their native membrane or solubilized in detergent and found a lower value in this latter case than in the former (14).
Obviously, the question of electrostatic interaction between cofactors should not be restricted to photosynthetic RC's. In order to achieve electron transfer reactions with rates compatible with biological catalysis, the electron donor and acceptor are usually located within 15 Å apart (37), an edge-to-edge distance similar to the one between P and the closest highpotential heme, here. This principle is likely to apply to all the membrane bound proteins involved in bioenergetic processes.  Midpoint redox potentials (E m ) are indicated for each species (see (10,13,25,38) for Blc. viridis and (10,22) for Rvi. gelatinosus). Electron transfer (E.T.) rates between cofactors (12,13), as well as orientation of the hemes relative to the membrane plane (determined by EPR spectroscopy (24)) are indicated for Bcl. viridis.