Small Weak Acids Reactivate Proton Transfer in Reaction Centers from Rhodobacter sphaeroides Mutated at AspL210 and AspM17*

In reaction centers of Rhodobacter sphaeroides, site-directed mutagenesis has implicated several acidic residues in the delivery of protons to the secondary quinone (QB) during reduction to quinol. In a double mutant (AspL210 → Asn + AspM17 → Asn) that is severely impaired in proton transfer capability over a wide pH range, proton transfer was “rescued” by added weak acids. For low pKa acids the total concentration of salt required near neutral pH was high. The ionic strength effect of added salts stimulated the rate of proton-coupled electron transfer at pH < 7, but decreased it at pH > 7.5, indicating an effective isoelectric point between these limits. In this region, a substantial rate enhancement by weak acids was clearly evident. A Brønsted plot of activity versus pKa of the rescuing acids was linear, with a slope of -1, and extrapolated to a diffusion-limited rate at pKaapp ≈ 1. However, the maximum rate at saturating concentrations of acid did not correlate with pKa, indicating that the acid and anion species compete for binding, both with weak affinity. This model predicts that pKaapp corresponds to a true pKa = 4-5, similar to that for a carboxylic acid or QB-, itself. Only rather small, neutral acids were active, indicating a need to access a small internal volume, suggested to be a proton channel to the QB domain. However, the on-rates were near the diffusion limit. The implications for intraprotein proton transfer pathway design are discussed.

In reaction centers of Rhodobacter sphaeroides, site-directed mutagenesis has implicated several acidic residues in the delivery of protons to the secondary quinone (Q B ) during reduction to quinol. In a double mutant (Asp L210 3 Asn ؉ Asp M17 3 Asn) that is severely impaired in proton transfer capability over a wide pH range, proton transfer was "rescued" by added weak acids. For low pK a acids the total concentration of salt required near neutral pH was high. The ionic strength effect of added salts stimulated the rate of proton-coupled electron transfer at pH < 7, but decreased it at pH > 7.5, indicating an effective isoelectric point between these limits. In this region, a substantial rate enhancement by weak acids was clearly evident. A Brønsted plot of activity versus pK a of the rescuing acids was linear, with a slope of ؊1, and extrapolated to a diffusion-limited rate at pK a app ≈ 1. However, the maximum rate at saturating concentrations of acid did not correlate with pK a , indicating that the acid and anion species compete for binding, both with weak affinity. This model predicts that pK a app corresponds to a true pK a ‫؍‬ 4 -5, similar to that for a carboxylic acid or Q B ؊ , itself.
Only rather small, neutral acids were active, indicating a need to access a small internal volume, suggested to be a proton channel to the Q B domain. However, the on-rates were near the diffusion limit. The implications for intraprotein proton transfer pathway design are discussed.
Light absorption by photosynthetic reaction centers (RCs) 2 drives the formation of an electrochemical gradient of H ϩ ions (protons) across the coupling membrane, the thylakoid membrane in chloroplasts and cell membrane in bacteria, and the generation of mobile reducing power. The transfer of electrons in the primary photochemical events generates most of the electrical component, whereas electron-coupled proton uptake and release accompanying the redox reactions of secondary donors and acceptors is largely responsible for the proton concentration gradient (⌬pH) (1).
In Rhodobacter sphaeroides, reducing equivalents are stored in the double reduction of the secondary ubiquinone, Q B , via the primary qui-none, Q A , and quinol is released into the membrane after two lightactivated turnovers of the RC (2,3). Each turnover results in transfer of an electron to the quinones from the primary donor, P, a special pair of bacteriochlorophylls (4 -7). The oxidized primary donor, P ϩ , is rereduced by a secondary donor after each photoactivation, and the events in the acceptor quinone complex can be summarized as (8 -10) in Scheme 1. The proton stoichiometric factors, a, b, etc., indicate the variable influence that the different quinone states have on nearby ionizable amino acid residues (10 -14).
The RC quinones are well buried in the protein, and proton transfer to Q B , which accompanies the second electron transfer to form Q B H 2 , must extend over a distance of 13-15 Å. The delivery pathway has been partially mapped out by site-directed mutagenesis (reviewed in Refs. 9, 10, and 15) and involves several members of a large cluster of ionizable, predominantly acidic, residues that have been identified by inspection of the x-ray structures ( Fig. 1) (16,17). Some are identifiable as likely proton carriers from their proximity to Q B in the structure, and there is general agreement that Glu L212 (18 -20), Asp L213 (19,21), and Ser L223 (22,23), in the L subunit, are terminal members of the pathway. However, in other cases, distinguishing a true proton-carrying role for an amino acid from a less active function, such as setting the local electrostatic potential and pK a values of other residues, is not straightforward. In addition to the terminal members of the pathway, other residues have been identified by site-directed mutagenesis to have highly significant influences on proton-coupled electron transfer to Q B . These include Glu H173 (24), His H126 and His H128 in the H subunit (25), Asp M17 in the M subunit, and Asp L210 (26 -28), Arg L207 , and Arg L217 in the L subunit (29).
In mutant RCs with the mutation Asp L213 3 Asn (mutant L213DN), the second electron transfer rate was inhibited by at least 10 4 -fold at pH Ͼ 7 (21). The proton transfer rate was even more strongly inhibited, because it is not rate-limiting in the wild-type but is rate-limiting in the mutant, leading to estimates that proton transfer is inhibited by Ն10 7fold in the mutant (30). Interestingly, the mutant RC activity could be partially reactivated ("rescued") by small weak acid anions, such as azide, N 3 Ϫ (31). In RCs with the mutation Glu H173 3 Gln (mutant H173EQ), the second electron transfer was also strongly inhibited but could be fully rescued by azide (24). However, it was not clear in these cases if the recovery was due to a proton delivery function of the added acid species, N 3 H (31), or an electrostatic effect of the bound anion, N 3 Ϫ (24), as has been proposed for the mechanism of many second site revertants to this and other primary lesions (32)(33)(34)(35)(36). Paddock and coworkers have studied RCs mutated at a putative H ϩ entry site on the protein surface, which are substantially impaired in proton uptake at elevated pH (25,37). The mutation neutralized two surface histidine residues, His H126 and His H128 , that had been identified in the binding site region of certain divalent transition metal ions that effectively blocked proton uptake (38,39). Both the first and the second electron transfers were inhibited at pH Ն 8, and both were restored to near wild-type rates by addition of the cationic buffer, imidazole (25). Paddock et al. (37) found that the first electron transfer could be rescued by a variety of cationic buffers in a manner that was proportional to the buffer pK a over a wide range of values. The results were analyzed in terms of rate-limiting proton transfer from the protein surface. The analysis also revealed the involvement of a low pK a intermediate in the transfer pathway, consistent with the presence of several carboxylates in the Q B domain.
Here we show that the double mutant Asp L210 3 Asn ϩ Asp M17 3 Asn (L210DN/M17DN), which has greatly impaired proton transfer capabilities over a wide pH range (28), can also be rescued but with a strong and possibly exclusive preference for small, neutral weak acids. A simple model, involving weak binding with competition between the protonated and unprotonated forms of the acid, satisfactorily describes the observed behavior. The implications for wild-type function are discussed in terms of the x-ray structure.

EXPERIMENTAL PROCEDURES
The L210DN/M17DN double mutant of R. sphaeroides was expressed in GaBM, a RC and light-harvesting complex (LH-I and LH-II) deletion strain that lacks the expression of pufB, -A, -L, -M, and -X, as well as pucB and -A. The puc deletion was carried out in a manner similar to Lee et al. (40), except that a streptomycin/spectinomycinresistance cartridge was used in place of the kanamycin-resistance cartridge. Deletion of the puf-operon region was achieved by insertion of a kanamycin-resistance cartridge starting at an engineered BamHI site, just upstream of the pufB, and another engineered BamHI site, just downstream of pufX, in a manner similar to the previously reported pufLM deletion strain (41).
The RC complementation vector, pLMX415, contains pufL, -M, and -X but lacks pufB and -A, which were eliminated using an engineered BamHI restriction site located upstream of pufB, and a naturally occurring BssHII site, located downstream of pufA. The codon changes for L210DN/M17DN double mutation were introduced using a protocol based on the QuikChange mutagenesis method (Stratagene, La Jolla, CA).
Cells were initially grown aerobically in the dark in 2-liter flasks placed in a rotary shaker (320 rpm) with 400 ml of Sistrom's medium supplemented with 10% Luria Broth (42) and in the presence of 2 g/ml tetracycline. Cultures were then allowed to develop full pigmentation under semiaerobic conditions with the addition of 800 ml of Sistrom's medium (43) containing yeast extract (0.4%) and tetracycline (2 g/ml), with 200 rpm agitation. RCs were isolated by detergent fractionation of the membranes with 0.7% lauryl (dodecyl) dimethylamine-N-oxide (LDAO) (Fluka), followed by ammonium sulfate precipitation, and column purification using DEAE-Sephacel (Sigma). The RCs were washed extensively on the column with 60 and 80 mM NaCl (in 0.06% LDAO, 10 mM Tris, pH 8.0), with short, additional washes at 100 -140 mM NaCl, and then eluted from the column with 180 -200 mM NaCl. Isolated RCs typically displayed A 280 /A 800 ratio of 1.25-1.35.
All kinetics assays were performed on samples with 1-2 M RC, 0.02% LDAO, 40 M ubiquinone-10 (Q-10) (Sigma), and 1 mM each of the following buffers, MES, MOPS, Tricine, CHES, and CAPS. The kinetics of the first electron transfer, Q A Ϫ Q B 3 Q A Q B Ϫ , were measured at 397 nm in the absence of donor to P ϩ , and 32 kinetic traces were averaged to achieve the desired signal-to-noise ratio. This wavelength (397 nm) corresponds to a distinctive electrochromic effect of Q A Ϫ on the nearby bacteriopheophytin and is close to isosbestic for P/P ϩ . The kinetics were fit to a single exponential representing the first electron transfer, k AB (1) . This adequately describes the major component, although the spectral response is known to be more complex and still poorly understood (44 -46). The kinetics of the second electron transfer, Q A Ϫ Q B Ϫ 3 Q A Q B H Ϫ , were measured at 450 nm, following the second of two short flashes 0.5 s apart, in the presence of ferrocene as a donor to P ϩ , as previously described (12). The kinetics were analyzed by a two-component exponential fit, representing the second electron transfer, k AB (2) , and the rereduction of P ϩ by the exogenous donor, ferrocene, which is seen at the same wavelength. Because the observed electron transfer rate varied widely with experimental conditions, the ferrocene concentration was adjusted to optimally separate the Q A Ϫ Q B Ϫ 3 Q A Q B H Ϫ electron transfer kinetics from those of P ϩ re-reduction. At 2-4 M ferrocene, P ϩ was reduced with a halftime of Ϸ200 ms, and at 150 -200 M ferrocene the P ϩ reduction halftime was Ϸ2 ms. The second electron transfer kinetics were well fit by a single component. All kinetic measurements were made at 21°C.

The Second Electron Transfer, Q A
electron transfer, with rate constant k AB (2) , is substantially (Ͼ99%) inhibited in L210DN/M17DN mutant RCs (28). We measured k AB (2) Ϸ 14 s Ϫ1 , compared with 1600 s Ϫ1 in wild-type, in low salt (2.5 mM KCl) at pH 7.0. However, the rate in the mutant RCs was greatly enhanced by addition of various salts of weak acids. Because the concentrations required were rather high, we examined the influence of ionic strength using salts of strong acids.
Salt Dependence-To test the effect of weak acid salts independently of any ionic effect, our interest was to find a regime in which the ionic response was minimal. We initially tested various symmetrical univalent and divalent electrolytes, alone or in combination, but found SCHEME 1 Na 2 SO 4 to give the simplest behavior. The Na 2 SO 4 concentration dependence of the measured rate, k AB (2) , is shown in Fig. 2 for pH values from 5.5 to 8.5. At acid pH, the rate was quite sensitive to salt, e.g. at pH 5.5, it increased 4 -5 fold as the concentration of Na 2 SO 4 was raised from 0 to 100 mM. At pH 7.0 and 7.5 the effect of salt was minimal and, after a small increase at concentrations below 20 mM, the rate was almost constant up to 600 mM Na 2 SO 4 . Addition of 300 mM NaCl on top of 600 mM Na 2 SO 4 had no significant effect. At pH 8 and higher, the rate actually decreased with added salt. Over the same concentration range, the rate in wild-type reaction centers was relatively much less sensitive, e.g. it increased from 1600 to 2000 s Ϫ1 at pH 7.0 (not shown).
The pH dependence of k AB (2) in L210DN/M17DN mutant RCs, in the presence and absence of 150 mM Na 2 SO 4 , is shown in the inset to Fig. 2. At pH values below 7.0 the rate was significantly enhanced by added salt, whereas above pH 7.5 addition of salt decreased the rate. At pH 7.0 -7.5, the rate was largely unresponsive to salt, and 150 mM Na 2 SO 4 , pH 7.0, was chosen as the standard condition for testing the activity of weak acid salts; under these conditions, k AB (2) ϭ 23 Ϯ 2 s Ϫ1 . As a control for the effect of additional weak acid salts, most of which were monoprotic, NaCl was titrated on top of the 150 mM Na 2 SO 4 (see Fig. 5A).
Stimulation of k AB (2) by Weak Acid Salts-In contrast to the slight effect of NaCl (up to 600 mM), added on top of 150 mM Na 2 SO 4 , addition of similar concentrations of various, weak monoprotic acids caused a marked acceleration (Fig. 3). The most effective of these was azide (N 3 Ϫ ), which was able to restore a significant fraction of the wild-type rate, at sufficiently high concentrations (Ͼ1 M). Several other weak acids significantly accelerated the rate, and titrations for formate, fluoride, and phosphate are also shown in Fig. 3 (the small effect of NaCl has been subtracted in all panels). A smaller number of experiments were done with 300 mM Na 2 SO 4 as the background salt, with identical results. The stimulation by weak acid salts was determined from the initial slope of the salt concentration dependence, yielding a second order rate constant in terms of total salt added (k (2) Ј). The small effect of NaCl was subtracted, and this slope was multiplied by (1 ϩ 10 pHϪpKA ), where pK A is the pK a of the added weak acid, to express k (2) in terms of the concentration of the protonated form [AH]. On this basis, the efficacy of the acids was clearly correlated with their solution pK a values (see Fig. 6, below), indicating that the protonated weak acid was responsible for overcoming a rate-limiting step in the proton transfer pathway within the RC. However, for those acids that showed curvature, indicating saturation of their effect on the measured rate, k AB (2) , the extrapolated maximum rate did not correlate with the acid pK a . For example, several acids with lower pK a than azide (pK a ϭ 4.72) gave significantly smaller rates at apparent saturation (see, e.g. Fig. 3).
pH Dependence of the Second Order Rate Constant-To strengthen the justification for expressing the second order rate constant in terms of the protonated acid form, we determined k (2) over the usable pH range, 6.5-7.5. At higher pH the rates were too slow for reliable measurement and at lower pH the rate was already quite fast and was more sensitive to the purely ionic effect of added salts. For those acids tested, the second order rate constant, when based on the concentration of the AH species, was independent of pH ( Fig. 4).
The results for all weak acids that were tested are shown in Table 1. Included are three cationic acids, triazole (pK a ϭ 2.3) and pyridine (pK a ϭ 5.4), which were ineffective, and ammonium (pK a ϭ 9.3), which was very weakly effective in stimulating k AB (2) , and two dibasic acids, oxalate (pK a ϭ 1.3 and 4.2) and phosphate (pK a ϭ 2.15 and 7.0). Oxalate gave very weak stimulation, whereas phosphate was effective (see Fig. 3). Nitrite (pK a ϭ 3.45) and cyanate (pK a ϭ 3.45) gave significant stimulation at low concentrations, from which a value of k (2) could be determined, but were less effective at high concentrations (cyanate) or even inhibitory (nitrite). Bicarbonate (pK a ϭ 3.58) could only be tested at low concentrations (up to 20 mM) due to instability of the pH and bubble formation from released CO 2 . Cyanide (pK a ϭ 7.0) severely and irreversibly inhibited RC activity and appeared to destroy the acceptor quinone complex.
Effect of Cd 2ϩ Ions on the Second Electron Transfer Reactivated by Weak Acids-Various divalent transition metal ions have been found to substantially inhibit proton uptake by reaction centers, thereby inhibiting the observed electron transfer (47)(48)(49)(50). We tested the effect of cadmium on RCs of the L210DN/M17DN double mutant in the presence of weak acids to reactivate the second electron transfer to varying degrees.  FEBRUARY 17, 2006 • VOLUME 281 • NUMBER 7

JOURNAL OF BIOLOGICAL CHEMISTRY 4415
Cadmium had no effect on the measured rate in unrescued L210DN/ M17DN mutant RCs or when reactivated with 100 mM azide or 100 mM fluoride (Fig. 5A). However, Cd 2ϩ did partially inhibit the rate when reactivated by 100 mM formate (Fig. 5B). The kinetics were monophasic at all concentrations of Cd 2ϩ , and slowed progressively as the cadmium concentration was raised. This indicates that the unbinding of metal ion is substantially faster than the electron transfer rate in the unbound state, i.e. k off Ͼ Ͼ 80 s Ϫ1 . Half inhibition occurred at 100 M, with a maximum effect of ϳ45% inhibition (after correction for the cadmiuminsensitive baseline rate (23 s Ϫ1 ) of the double mutant RCs). The same degree of inhibition was seen in the initial slope (i.e. k (2) ), when formate was titrated in the presence of 400 M Cd 2ϩ .
In wild-type RCs, in low salt, subsaturating concentrations of Cd 2ϩ induce biphasic kinetics of the second electron transfer, reflecting titration of the site and slow unbinding of the metal ion, and the slow phase amplitude has been used to quantify the binding of metal ion (49,50). In the presence of 150 mM Na 2 SO 4 , however, the affinity was weaker and we found that separation of the two components was less reliable. To ensure that the monophasic kinetic analysis of the mutant reflected the same properties as the assay of slow phase amplitude in the wild-type, we compared the two assays in wild-type RCs. In low salt (2.5 mM KCl), the dissociation constant, K d , was 2 M when assayed by slow phase amplitude, while retardation of the single component (average) rate constant gave K d ϭ 3 M. In 150 mM Na 2 SO 4 , the two assays gave dissociation constants of 8 and 10 M, respectively (data not shown). Thus, we expect the value obtained for the double mutant (K d ϭ 100 M) to be a reliable estimate of the affinity for Cd 2ϩ , which is about 10-fold weaker than in wild-type RCs.
The First Electron Transfer, Ϫ -The first electron transfer to Q B is weakly coupled to proton transfer, with H ϩ uptake driven by relatively small changes in the pK a values of many ionizable residues in and around the quinone binding sites (for review, see Ref. 10). In RCs of the L210DN/M17DN double mutant, the rate of the first electron transfer, k AB (1) , is also greatly inhibited in comparison to the FIGURE 3. Acceleration ("rescue") of the rate of the second electron transfer to Q B in L210DN/M17DN mutant RCs, by added weak acid salts, sodium azide, sodium formate, sodium acetate, and sodium phosphate. The rate (k AB (2) ) was determined from kinetics measured at 450 nm. Conditions: as for Fig. 2, plus 150 mM Na 2 SO 4 . Curves through data points are derived from the Model B described in the Supplementary Materials, with the parameters given in Table 2. All data are corrected for the rate seen with equivalent concentrations of NaCl (see lower curve in Fig. 5A). Error bars for 2-4 separate measurements are shown, or are within the size of the symbols.
wild-type (28), we measured k AB (1) ϭ 320 s Ϫ1 versus 7000 s Ϫ1 , in low salt, at pH 7.0 (data not shown). We observed a very similar response of k AB (1) to added weak acids, with similar values of k (2) calculated from the initial slopes of the concentration dependence. This work will be presented in more detail in a subsequent paper.

DISCUSSION
Salt Effects-The effect of salts like NaCl and Na 2 SO 4 on k AB (2) in RCs of the L210DN/M17DN double mutant is substantial but qualitatively different at low and high pH, accelerating with increased salt at low pH and decelerating at high pH. The existence of a cross-over point at pH 7.0 -7.5 is a clear indicator that this is an ionic effect that reflects the surface charge of the RC, with a functional isoelectric point in this range. The isoelectric point for the whole RC has been reported as pI ϭ 6.1, for R. sphaeroides, strain R26 (51), but a discrepancy between this net pI and the cross-over pH for k AB (2) is not surprising. Proton uptake and transfer coupled to Q B reduction is likely to be under the electrostatic influence of a sub-domain of the cytoplasmic surface rather than the global potential. Furthermore, the mutation of two aspartic acids in the L210DN/ M17DN double mutant is expected to raise the pI of the mutant RCs compared with wild-type. The isoelectric points for proteins can often be estimated from the amino acid composition and the N and C termini. Using standard pK a values, and correcting for the presence of the iron atom (Fe 2ϩ ) and the non-ionizability of the histidine ligands to the iron and to the bacteriochlorophylls, we used the Biology Workbench Protein Tools facility (workbench.sdsc.edu/) to calculate the pI. The value obtained was identical to the measured value (6.1). Although such good agreement is undoubtedly partly fortuitous, it gives confidence to calculations of other pI values. The pI of the cytoplasmic domain of the RC was calculated using only the cytoplasmic loops of the L and M subunits and the globular domain of the H subunit, with the same constraints for . pH independence of k (2) . For each acid, k (2) was determined from the initial slope of the titration of rate versus total salt added, A T , as in Fig. 3, and converted to the free acid, [AH], using the acid pK a and the prevailing pH: k (2) ϭ k (2) Ј ϫ (1 ϩ 10 pHϪpK ). Conditions: as for Fig. 2, with 150 mM Na 2 SO 4 .  (1) and (2) indicate first and second ionization equilibria (phosphate and oxalate). b k (2) Ј values are from experimental initial slopes, corrected for the NaCl control (85 M Ϫ1 s Ϫ1 ). c k (2) ϭ k (2) Ј ϫ (1 ϩ 10 pKϪpH ). d The rates with oxalate could not be considered reliably above the NaCl control. e The rates with triazole and pyridine were less than for the NaCl control.  Table 2. Lower curve is for the salt response, NaCl added on top of 150 mM Na 2 SO 4 . B, Cd 2ϩ titration of k AB (2) partially rescued by 100 mM formate. Conditions for both panels are as those for Fig. 2, with 150 mM Na 2 SO 4 , pH 7.0. the iron ligands. The calculated pIs were 8.0 for wild-type RCs and 8.7 for the L210/M17 double mutant. These values are consistent with the "cis positive rule" for membrane proteins, whereby a positively charged cytoplasmic domain is expected, complementing the negative polarity of the cytoplasmic compartment (52,53). However, they are clearly out of the range indicated by the cross-over pH for k AB (2) , further indicating a more local nature of the electrostatic domain influencing the protoncoupled electron transfer. In the wild-type, with Asp M17 and Asp L210 both ionized, this domain will be much more negative, with a lower pI. The response of k AB (2) for the L210DN/M17DN mutant RCs to the bulk phase ionic conditions is as expected from the pH dependence of the reaction, which slows progressively as the pH is raised (28). The mutations make proton transfer rate-limiting, and any enhancement or depression of proton delivery will affect the measured electron transfer rate. At low pH, when the relevant surface area is positively charged, the local surface pH will be higher than the bulk value. Addition of salt to screen the surface charge (and possibly change it by weak adsorption) will diminish the surface potential and allow higher concentrations of H ϩ at the surface, thereby enhancing the protonation state of all ionizable groups and stimulating the electron transfer rate. Conversely, at higher pH, above the local (effective) pI, the protein is net negatively charged and the pH at the surface will be lower than in the bulk phase.
Addition of salt will again screen the surface charge and diminish the magnitude of the surface potential, but in this case the potential is negative and the surface H ϩ concentration will decrease with increasing ionic strength, resulting in a decrease in the proton limited electron transfer rate. A related effect of salt, on H ϩ uptake kinetics in wild-type RCs, was shown by Maróti and Wraight (54). Salts caused a substantial slowing of H ϩ uptake by the P ϩ Q A Ϫ state, i.e. in the absence of any rate-limiting electron transfer. High salt concentrations were required for this effect at alkaline pH, and this was taken to indicate a substantial charge density, qualitatively consistent with the large number of ionizable residues in the RC cytoplasmic domain.
In the related L210DA/M17DA double mutant (i.e. with Asp 3 Ala), Paddock and coworkers also reported a stimulation of the second electron transfer, k AB (2) , by various salts, including some weak acids (55).
However, they interpreted the effect as of purely ionic origin, and suggested that charge screening allowed penetration of the protein by H ϩ ions, thereby overcoming some limitation in accessibility. Conceivably there is a real difference in the behavior of the two mutants, but it is clear from our studies that, for the L210DN/M17DN double mutant, the effect of many weak acids could be readily distinguished from the general salt effect. For these weak acid anions, strong rate enhancement was seen when added on top of 150 mM Na 2 SO 4 , whereas addition of NaCl had very little influence. Weak Acids-At acid pH, the purely ionic effect of high salt caused a 4-to 5-fold increase in k AB (2) in L210DN/M17DN mutant RCs, thereby compacting the available range for the effect of weak acids. At alkaline pH the rate became so slow that non-physiological routes of Q Ϫ reoxidation (such as via ferrocenium or O 2 ) could obscure the true value of k AB (2) . The range of pH for studying the effect of weak acid stimulation was therefore restricted to 6.5-7.5. Nevertheless, this was sufficient to show that the second order rate constant, based on the concentration of protonated acid species, was pH independent. If the active species were the salt (anion), the correction to protonated species would generate an artificial pH dependence, with a slope of Ϫ1 for log k (2) versus pH. We therefore conclude that the observed rate enhancement is a genuine "rescue" effect of the protonated form of the weak acid.
We tested a variety of acids for stimulatory activity, and found that only a limited selection were indisputably active. These were all small and monoprotic, with the exception of phosphate. Acetic acid was the largest acid that showed sufficient activity to be clearly discriminated above the residual ionic effect. Oxalate exhibited a very marginal ability to rescue k AB (2) , indicating that neither the neutral acid nor the monoanion were significantly active. On the other hand, phosphate was active, and reasonable k (2) values could be calculated for either the neutral or monoanion as active donor (discussed further, below). Two cationic acids with seemingly suitable pK a values, triazole (pK a ϭ 2.3) and pyridine (pK a ϭ 5.4), gave no significant response at all, so the rate was actually less than at equivalent concentrations of NaCl. This is consistent with these species being predominantly neutral at pH 7.0.
Categorizing tested acids as active was done conservatively, because the involvement of the pK a value in determining both the concentration and the intrinsic reactivity of the protonated form presents a strong bias toward apparent activity. Even a tiny enhancement of k AB (2) above the control rate with NaCl can translate into a significant datum in the double logarithmic representation of a Brønsted plot (log k (2) versus pK a ). The potential for an artifactual influence of the pK a is well exemplified by ammonium and oxalate. Comparing a single titration of NH 4 Cl with a single control titration of NaCl yielded a small enhancement (up to 20%), but repetition of these titrations gave variable results, from 0 to 20% above the NaCl control. However, almost any detectable rate increase would place the ammonium data point (pK a ϭ 9.3) quite nicely on the Brønsted plot of Fig. 6. The effect of oxalic acid was also very small, with a slope barely greater than the NaCl control. However, with pK a ϭ 1.3, and hence a very small concentration of protonated species, any finite difference from the background rate yields a datum that fits reasonably well on the trend line. Between these extremes, the FIGURE 6. Relationship between the second order rate constant for rescue (log k (2) ) and acid pK a (Brønsted plot). Acid salts shown: phosphate (pK a (1) ϭ 2.15, pK a (2) ϭ 7.0), fluoride (pK a ϭ 3.16), nitrite (pK a ϭ 3.37), cyanate (pK a ϭ 3.46), formate (pK a ϭ 3.72), bicarbonate (pK a ϭ 3.58), azide (pK a ϭ 4.72), acetate (pK a ϭ 4.76), and ammonium (pK a ϭ 9.25). Dashed line shows diffusion limit at k on ϭ 10 9 M Ϫ1 s Ϫ1 . The fitted line (slope ϭ Ϫ1.01) excludes ammonium and the second pK a of phosphate (pK a (2) ϭ 7.0); see text. For ammonium, the open circle shows the largest experimental value of k (2) obtained, and the arrow indicates the range (see text). k (2) for each acid was determined as described in Fig. 4, using the acid pK a and the prevailing pH 7.0. For phosphate, fluoride, formate, azide, acetate, and ammonium, k (2) values were also obtained by fitting the full titration curve, according to the kinetic model described in the text (closed circles). Points are also shown for H 3 O ϩ (pK a ϭ Ϫ1.74) and H 2 O (pK a ϭ 15.74), calculated assuming concentrations of 10 Ϫ7 M and 55 M, respectively. effects of weak acids were substantial and could be determined with confidence. The data used for our analysis and for the linear fit of the Brønsted plot are therefore limited to those acids that were indisputably active, for which the uncorrected rate enhancement (k (2) Ј, the initial slope with respect to total salt concentration) was at least 5-fold greater than the NaCl control.
pK a Dependence of Recovery-A Brønsted plot (log k (2) versus pK a ) for all active acids was linear (Fig. 6). Although the pK a range was limited by the availability of suitable materials, small, neutral weak acids, it was sufficient to clearly indicate a slope very close to Ϫ1 (the line drawn in Fig. 6 is a least squares fit to only the most conservatively active acids). This is strongly indicative of a direct role for the weak acids as proton donors to the mutant RCs. However, there were some apparent anomalies in the data. In particular, azide and acetate have almost identical solution pK a values, but azide was significantly more active and acetate somewhat less active than predicted by the simple linear relationship. This suggested that size might be a limiting factor and may explain the failure of some candidate acids to exhibit activity (see below).
For a simple process of proton-coupled electron transfer, once the proton donor has sufficient acidity for the proton transfer to be nonrate-limiting, one expects the maximum electron transfer rate (at saturating concentrations of acid) to approach a common value that is independent of pK a . However, the experimental concentration dependences revealed a wide range of maximum rates at saturation that did not correlate with the acid pK a . For example, fluoride (pK a ϭ 3.16) and phosphate (pK a ϭ 2.15) were unable to fully restore the rate, or even to attain the same level as azide (pK a ϭ 4.72). However, this is readily accounted for if we allow for the anion (A Ϫ ) to interact at the same site as the acid (AH) species, i.e. as a competitive inhibitor. k H and k ϪH are the rate constants for forward and backward proton transfer within the RC, and k H /k ϪH ϭ K H ϭ 10 (pKRϪpKA) ϭ 10 ⌬pK , where pK A and pK R are the pK a values for the exogenous weak acid, AH, and the endogenous intermediate, RH, respectively. k ET is the rate constant for electron transfer to Q B H and is determined by the free energy drop in the purely electron transfer step, Q A Ϫ Q B H 3 Q A Q B H Ϫ , and by the distance between the two quinones (see, e.g. Ref. 56). It has been estimated at ϳ10 6 s Ϫ1 in wild-type RCs (57,58) and is largely unaffected by conditions and mutations that induce significant electrostatic changes in the acid cluster of the Q B domain (28,30). We therefore take the wild-type value of k ET ϭ 10 6 s Ϫ1 . In fact, k ET , as used here, could include proton transfer events between RH and Q B Ϫ , but, so long as the proton transfer from RH is fast and not unfavorable, the effective rate constant approaches the true k ET . Alternatively, R could be Q B , itself. The data fits by the model show that AH and A Ϫ both bind very weakly, with K D and K I in the molar range, so k on and k off are of the same order. This is superficially apparent from the concentration range of the rate enhancement. Also, the observed kinetics of electron transfer were monophasic at all concentrations of rescuing acid, indicating that the binding equilibrium is fast (say, 10-times greater) compared with the step that is rate-limiting when proton transfer is restored, i.e. electron transfer, with k ET Ϸ 10 6 s Ϫ1 . The equilibration rate for binding is the sum of on and off rates. Thus, fast equilibrium will be established for k on [AH] ϩ k off Ն 10 7 s Ϫ1 and, because K D Ϸ 1 M, k on [1 M] Ϸ k off Ն 10 7 s Ϫ1 . Furthermore, for the low pK a rescuing acids, the concentration of protonated species is very small, necessitating values of k on close to the diffusion limit. For example, for fluoride (pK a ϭ 3.16) at 20 mM total salt concentration, the concentration of HF is ϳ10 Ϫ6 M at pH 7.0 and k on must be ϳabout 10 8 M Ϫ1 s Ϫ1 to yield a rate that is fast compared with the net rate of the proton-coupled electron transfer (ϳ50 s Ϫ1 under these conditions, including the background rate). For phosphoric acid (pK a ϭ 2.15), the minimum rate constant approaches 10 9 M Ϫ1 s Ϫ1 . Thus, the rate of weak acid binding to the RC is close to the diffusion limit.
The binding of A Ϫ is also weak (K I Ϸ 1 M). The on and off rate constants for anion (A Ϫ ) binding are not explicit in Scheme 2, but the monophasic nature of the observed kinetics indicates that, in the concentration range employed here, [A Ϫ ] Ն 1 mM, the anion binding equilibrium is always fast compared with the other steps in the reaction scheme.
A rate equation for this model can be derived assuming weak binding equilibrium of the anion forms (see Supplementary Materials, Model B). The curves in Figs. 3 and 5A are fitted according to this model. The fitted lines are indistinguishable from those obtained with an even simpler, purely equilibrium, model (Supplementary Materials, Model C), in which acid and anion binding (K D and K I ) and proton transfer (k H / k ϪH ϭ K H ) are in rapid equilibrium. It should be noted that the second order rate constant is not affected by the competitive binding equilibrium between AH and A Ϫ , because it is defined by the slope at low concentrations, when net binding of either species approaches zero.
Implications and Predictions of the Model-The model predicts a linear dependence of log k (2) on the pK a of the weak acid (pK A ), with a slope of Ϫ1. This is well borne out by the experimental data, which extend over a range of pK a from 2 to 9. Because of the potential for systematic bias for acids with extreme pK a values (discussed above), the linear fit in Fig. 6 is only to the data between pK a 2-5, where the measured rate enhancements were 10-to 50-fold greater than the NaCl control. The fitted slope in this range is Ϫ1.0.
The available data did not reach the maximum rate of ϳ10 9 M Ϫ1 s Ϫ1 expected for the diffusion limit, but extrapolation to this limit indicates that such a value would be reached for pK A Ϸ 1. In the simplest case, this would be equal to the pK a of the acceptor species, R Ϫ , within the RC, i.e. pK A ϭ pK R . However, the role of R as an intervening proton carrier between the donor (AH) and the assay function (electron transfer) leads to an offset of the apparent pK a from the true pK a of RH/R Ϫ . The offset is equal to log (k ET /k off ), which is ϳϪ3. Thus, the actual pK a of the acceptor group should be in the ballpark of 4. This is consistent with R Ϫ being a carboxylic acid, but it could also be Q B Ϫ itself, for which a pK a value of 4.5 has been estimated (10, 58). In fitting the data, we have used pK R ϭ 4.3. This value does not enter into the assessment of k (2) . The

SCHEME 2
Chemical Rescue of Proton-coupled Electron Transfer FEBRUARY 17, 2006 • VOLUME 281 • NUMBER 7 precise value of pK R influences the magnitude of the binding affinities, K D and K I , but does not affect the qualitative conclusions to be drawn. The variable maximum rate seen for different weak acids, independently of their pK a and efficacy as proton donors, arises from the competitive binding of the donor acid and inactive base. The two species are present in constant proportion at a fixed pH, and their relative affinities determine the maximum rate that can be recovered at saturating concentrations of added (total) salt (Equation 1) (see Supplementary Materials) as follows.
Note that the pH dependence is identical in form to that expected for a simple model 3 of the wild-type rate, where the relevant pK a is believed to be that for Q B Ϫ (10, 58).
It is noteworthy that, although the protonated forms of the most effective donors (with pK a Ͻ Ͻ 7) are active at low concentrations, they all bind very weakly (K D Ϸ 1 M). This apparent paradox is due to the fast on and off rates (near the diffusion limit) that allow rapid, but low occupancy, binding equilibrium to be established at the low concentrations of free acid prevailing at pH 7.0.
Also somewhat counterintuitively, for weak acids with pK a Ͻ 7 the rescue activity titrates with saturation behavior characterized by the anion (inhibitor) dissociation constant, K I , rather than the acid (donor) dissociation constant, K D . This behavior is predicted by the model (see Supplementary Materials), and can be appreciated as follows. For such acids, at pH 7.0, the concentration of the protonated species is small and the range of concentrations is entirely in the linear region of the hyperbolic binding dependence. Conversely, the anion is the overwhelming species present, and the total salt concentration is very high. Thus, even for the weak binding affinities involved, the RCs become substantially occupied by A Ϫ and the concentration of free RC available to bind AH is progressively limited. The rapid binding equilibria of both AH and A Ϫ yields a rate of electron transfer that is proportional to the fraction of RCs with AH bound, which saturates at K I /K D .
Competition for binding between donor acid and inactive base was not considered by Paddock et al. (37) in their study of acid rescue of the 2xHis mutant, where the donor site was at the protein surface. However, in the 2xHis mutant, only cationic acids were active donors, all with weak affinity for a negatively charged site, i.e. with Asp M17 and Asp L210 ionized. It is likely that the neutral bases bound much more weakly than the acid forms, in which case any effect of competition would be slight.
Properties of the Rescuing Species-In this work, with the L210DN/ M17DN double mutant, almost all rescuing acids shared similar features, small size, neutral acid form, very weak binding of the acid, and slightly stronger (2-to 5-fold) binding of the anion. The possible excep-tions are phosphate and oxalate (negative acid forms for the second pK a ) and ammonium (positive acid form), and a more general statement might be that the more negative (or less positive) species binds more strongly, and we now discuss these.
At least for the most active acids, the relative order of binding strengths (anion Ͼ acid) suggests a local environment that may favor a negative charge. Electrostatic calculations have indicated that the Q B domain in wild-type RCs is designed to accommodate approximately one extra negative charge, i.e. to stabilize the Q B Ϫ semiquinone (and Q B H Ϫ ) (10,59). Furthermore, the protein appears to allow substantial ionization of the acid cluster near Q B , including Asp L210 and Asp M17 (14,16,60,61). Thus, the L210DN/M17DN double mutant should be at least as hospitable, and a preference for the anion species is consistent with this. We might, therefore, expect the monobasic forms of phosphate (pK a (2) ϭ 7.0) and oxalate (pK a (2) ϭ 4.2) to be effective rescuing acids, but this does not appear to be the case. For phosphate, the titrations can be fit for either acid/base pair with the parameters shown in Table 2. Both potential donors, H 3 PO 4 (pK a (1) ϭ 2.15) and H 2 PO 4 Ϫ (pK a (2) ϭ 7.0), are predicted to give good activity, with relative binding affinities that favor the more negatively charged species. However, if both donor forms are indeed active, then the affinity for H 2 PO 4 Ϫ as a competitive inhibitor of H 3 PO 4 (K I ϭ 0.3 M) should be the same as its affinity as a proton donor (K D ϭ 2 M). These values are sufficiently different as to be incompatible with this simple model. For oxalate, the distinction is dramatic, although the activity data are very weak in the first place. We conclude that, if oxalic acid is active at all, it is through the neutral diacid and is very weakly so, perhaps because of its size, which is larger than that of acetic acid (62,63).
To account for the apparent lack of activity by anionic acids, we speculate that the binding site for all anions is subtly distinct from that for the neutral species. Thus, although binding of the conjugate base (anion) blocks the neutral acids from donating the proton, it does not position the anion to be able to donate if it is also an acid. Furthermore, the anion affinities are all quite similar (K I ϭ 0.2-0.4 M), whereas the acid affinities vary more systematically with efficacy.
The relative binding affinities of the neutral acids and anionic bases suggest that cations bind even more weakly. This could account for the low or negligible activity of the cationic acids tested, e.g. no reliable activity could be detected for triazole (pK a ϭ 2.3) and pyridine (pK a ϭ 5.4), but size may also be a contributing, or even the overriding, prohibitive factor. Ammonium (pK a ϭ 9.3) was marginally active in terms of our conservative criterion for activity; it was only slightly greater than the general salt effect of NaCl. Because of the high pK a , however, this translated to an activity that placed it close to the Brønsted line for the neutral acids. Also due to the high pK a of NH 4 ϩ /NH 3 , the shape of the concentration dependence was determined by K D rather than K I (unlike the neutral acids with pK a Ͻ 7), and the apparent affinity of the donor form (NH 4 ϩ ) was comparable with other acids (K D Ϸ 0.5-1 M). It was not possible to estimate K I , except to say that it was not much smaller than K D . 3 The real behavior is not "simple," because the pK a of bound Q B Ϫ is pH-dependent, due to changes in the density and distribution of charges in the protein (Graige et al. (58) and Wraight (10)).

TABLE 2 Dissociation constants (molar) for donor acids and inhibitory bases (from fits to kinetic model B in Supplementary Materials)
For each acid/base pair, K D corresponds to the less negative species, K I to the more negative species. pK a (1) and pK a (2) refer to first and second ionization constants, e.g. of phosphoric acid and oxalic acid. In the absence of any added acids, the inhibited rate in the mutant RCs (Ϸ20 s Ϫ1 ) could reflect donor activity of H 3 O ϩ , with pK a ϭ Ϫ1.74, or H 2 O, with pK a ϭ 15.74. For donation by H 3 O ϩ , at pH 7, the calculated second order rate constant from the single data point would be ϳ2 ϫ 10 8 M Ϫ1 s Ϫ1 . This is significantly lower than the values (10 10 -10 11 M Ϫ1 s Ϫ1 ) normally associated with aqueous proton diffusion. 4 Conversely, if we consider the unrescued rate of this mutant to be due to donation by H 2 O at a concentration of 55 M, the estimated second order rate constant is too large by at least 6 orders of magnitude (see Fig.  6). The intrinsic donor ability of water is evidently so low that almost anything will do better, including H 3 O ϩ , and possibly even buffers, i.e. species we would consider "inactive." Structural Implications-The mutational lesion in the L210DN/ M17DN double mutant is in the middle of the putative proton transfer pathway from the protein surface to Q B and obstructs proton delivery from the entry site (near the surface histidines) to the inner segment consisting of Asp L213 , Ser L223 , and possibly one or more water molecules. To circumvent this block, the exogenous weak acids must either donate to one of these inner sites or find a novel route of access to Q B Ϫ .
In this regard the effect of cadmium is suggestive. Various divalent transition metal ions have been found to substantially inhibit proton-coupled electron transfer and proton uptake by reaction centers (47)(48)(49)64). The binding site for cadmium (Cd 2ϩ ) was identified, by x-ray crystallography, to consist of His H126 , His H128 , and Asp H124 (38). Single mutation of either Asp M17 or Asp L210 caused some decrease in affinity (10-and 4-fold, respectively, at pH 7.7 (27)). However, in addition to the inhibition with relatively high affinity, Cd 2ϩ now showed a further slowing of the rate of the proton-coupled electron transfer that required 100-fold higher concentrations of metal ion. This was interpreted to mean that the observed reaction (k AB (2) ) seen with Cd 2ϩ in the micromolar range proceeded by slow unbinding of the metal, allowing the proton and electron transfer reactions to occur in the uninhibited state. At higher concentrations of metal ion, rebinding of Cd 2ϩ progressively diminished the lifetime of the metal-free RCs, so that the only possible electron transfer path was the much slower reaction of the metal-bound RCs. The limiting rate therefore represented the true rate of the proton-limited reaction in the single mutants (27).
In the L210DN/M17DN double mutant RCs, proton transfer is already fully inhibited and showed no additional sensitivity to Cd 2ϩ . Cadmium also had no effect on the partially recovered kinetics seen in the presence of azide and fluoride. This is consistent with findings in wild-type RCs, where azide could restore second electron transfer activity without affecting the tight binding affinity of Cd 2ϩ (49). However, when the kinetics were rescued by formate, Cd 2ϩ diminished the rate by up to 40%, with a half inhibitory concentration of about 100 M. We do not know for sure that the site of action of Cd 2ϩ indicated here is the same as that identified by crystallography in wild-type RCs, but the low affinity is roughly consistent with the combined effect of the two single mutants (27) and a generally weakened binding due to the high salt concentration used in our work.
We tentatively suggest that these distinct responses to Cd 2ϩ reflect the relative sizes of the acids involved, with the smaller ones (HN 3 and HF) readily able to penetrate past the site of cadmium blockage, while passage of formic acid is partially obstructed. Examination of the x-ray structure reveals this as a distinct possibility. Fig. 7 shows a view of the putative proton transfer pathway looking from the Q B site toward the protein surface. The contour representation of the surface residues shows a substantial hole through which a small molecule could readily access the putative proton transfer pathway, beginning with Asp L210 . This feature is very little affected by the binding of cadmium (not shown). 5 Thus, it is not surprising that the smaller acids, HN 3 and HF, are not impaired in their ability to rescue the mutants RCs. Small movements of the two surface histidine residues on binding of Cd 2ϩ may indicate some increase in rigidity, which could contribute to the effect seen on formate activity. Further size restrictions are likely encountered deeper in the protein, as the weak acids must, presumably, get past Asp L210 to override the mutational block.
The overall picture, then, is for fast access to a site that is sufficiently occluded that it exhibits a size restriction. However, enzymes commonly exhibit diffusion limited bimolecular rates, as indicated by their k cat /K m values, despite substantially buried active sites. The active site of acetylcholinesterase, for example, is at the bottom of a 20 ϫ 5 Å long "gorge" with a fluctuating gate of average radius only 3 Å. Experimentally, the rate of diffusional encounter of substrate (acetylcholine) is more than 10 9 M Ϫ1 s Ϫ1 , at an ionic strength of 0.15 M (65). This rate is electrostatically enhanced, but even neutral substrate analogs exhibit rates above 10 7 M Ϫ1 s Ϫ1 (66). Brownian dynamics modeling yields similar conclusions, that acetylcholine binding is diffusion-limited and is electrostatically enhanced to reach values in excess of 10 9 M Ϫ1 s Ϫ1 , but if the electrostatic potential is turned off the rate is still close to 10 8 M Ϫ1 s Ϫ1 (67). Catalase provides an even more dramatic and more appropriate example, with a small, neutral substrate (H 2 O 2 ) that must access a buried active site through channels that are Ͼ30 Å in length, with severe restrictions in the last 10 -12 Å proximal to the heme; yet the effective encounter rate is in excess of 10 9 M Ϫ1 s Ϫ1 (68). Structural comparison of several mutant variants of catalase suggests that an internal electrical potential contributes to the orientation and guidance of the dipolar substrate through the final and most restrictive part of the channel (69). Thus, the facility with which small neutral acids access the Q B proton channel, which is only 10 -12 Å to Asp L213 , is remarkable insofar as it is not part of the original, wild-type design.