Modified ligands to FA and FB in photosystem I. Proposed chemical rescue of a [4Fe-4S] cluster with an external thiolate in alanine, glycine, and serine mutants of PsaC.

The FB and FA electron acceptors in Photosystem I (PS I) are [4Fe-4S] clusters ligated by cysteines provided by PsaC. In a previous study (Mehari, T., Qiao, F., Scott, M. P., Nellis, D., Zhao, J., Bryant, D., and Golbeck, J. H. (1995) J. Biol. Chem. 270, 28108-28117), we showed that when cysteines 14 and 51 were replaced with serine or alanine, the free proteins contained a S = 1/2, [4Fe-4S] cluster at the unmodified site and a mixed population of S = 1/2, [3Fe-4S] and S = 3/2, [4Fe-4S] clusters at the modified site. We show here that these mutant PsaC proteins can be rebound to P700-FX cores, resulting in fully functional PS I complexes. The low temperature EPR spectra of the C14XPsaC·PS I complexes (where X = S, A, or G) show the photoreduction of a wild-type FA cluster and a modified FB′ cluster, the latter with g values of 2.115, 1.899, and 1.852 and linewidths of 110, 70, and 85 MHz. Since neither alanine nor glycine contains a suitable side group, an external thiolate provided by β-mercaptoethanol has likely been recruited to supply the requisite ligand to the [4Fe-4S] cluster. The EPR spectrum of the C51SPsaC·PS I complex differs from that of the C51APsaC·PS I or C51GPsaC·PS I complexes by the presence of an additional set of resonances, which may be derived from the serine oxygen-ligated cluster. In all other mutant PS I complexes, a wild-type spin-coupled interaction spectrum appears when FA and FB are simultaneously reduced. Single turnover flash studies indicate ∼50% efficient electron transfer to FA/FB in the C14SPsaC·PS I, C51SPsaC·PS I, C14GPsaC·PS I, and C51GPsaC·PS I mutants and less than 40% in the C14APsaC·PS I and C51APsaC·PS I mutants, compared with ∼76% in the PS I core reconstructed with wild-type PsaC. These data are consistent with the measurements of the rates of cytochrome c6-NADP+ reductase activity, indicating lower rates in the alanine mutants. It is proposed that the chemical rescue of a [4Fe-4S] cluster with a recruited external thiolate at the modified site allows the mutant PsaC proteins to rebind to PS I and to function in forward electron transfer.

PsaC is a ferredoxin-like Photosystem I (PS I) 1 protein that ligates two [4Fe-4S] clusters, F A and F B , which function as intermediates in electron transfer from F X to soluble ferredoxin or flavodoxin. The sequence and the kinetics of electron transfer between the F X , F B , and F A iron-sulfur clusters and to the [2Fe-2S] soluble ferredoxin are not well understood. One of the major experimental difficulties is the near-identity of the optical signatures of F X , F B , F A , and ferredoxin, and the attendant problem of correlating a given acceptor with observed electron transfer kinetics. Although F X , F B , F A , and ferredoxin are distinguishable by low temperature EPR spectroscopy, electron exchange among these redox carriers typically occurs within the rise time of the spectrometer. One approach to differentiating between the iron-sulfur clusters is to modify the kinetic, thermodynamic, and/or spectroscopic properties by altering a cysteine ligand to one of the cubane irons. The most radical change would be to convert a [4Fe-4S] cluster to a [3Fe-4S] cluser, with the consequence that the reduction potential of the cluster should be driven more electropositive (9,10). Based on precedent with proteins of known structure (8,11), this change should occur when the second cysteine of each CXXCXX-CXXXCP motif is changed to aspartic acid.
The amino acid sequence of PsaC has sufficient similarity with the 2[4Fe-4S] ferredoxins from Peptococcus aerogenes and Clostridium pasteuranium to predict a similar three-dimensional backbone structure, especially in the region surrounding the iron-sulfur clusters (12,13). PsaC can be removed from the PS I reaction center with chaotropic agents and replaced with mutant proteins overproduced in Escherichia coli, thereby providing a facile means to alter the ligands to the cubane iron atoms. When cysteines 14 (C14D PsaC ) and 51 (C51D PsaC ) are individually replaced with aspartic acid and the unbound PsaC is chemically reduced, S ϭ 1/2, [4Fe-4S] 1ϩ clusters are found at the unmodified site, and a mixed population of S ϭ 2, [3Fe-4S] 0 and S ϭ 3/2, [4Fe-4S] 1ϩ clusters are tentatively identified at the modified site (14,15). The mutant C14D PsaC and C51D PsaC proteins containing two [4Fe-4S] clusters, but not those containing one [3Fe-4S] and one [4Fe-4S] cluster, could be rebound to P700-F X cores to reconstitute a functional PS I complex. The mixed-ligand [4Fe-4S] clusters are capable of low temperature electron transfer from P700 to F A /F B and room temperature electron transfer to NADP ϩ with ferredoxin or flavodoxin. The ligand at the modified sites in the mutants was not identified; candidates include water, hydroxide ion, the carboxylate from aspartate, and the ␤-mercaptoethanol used in the iron-sulfur insertion protocol.
Further studies with serine and alanine mutations at cysteines 14 and 51 (16) showed that in addition to [3Fe-4S] clusters, S ϭ 3/2, [4Fe-4S] clusters were likely present at the modified sites in the unbound PsaC proteins. While serine, an amino acid with a potential oxygen ligand, might be expected to support a cubane cluster, this result was quite unexpected for alanine, an amino acid that contains a methyl side group. Further complicating the issue was that the C14A PsaC and C51A PsaC as well as C14S PsaC and C51S PsaC could be rebound to P700-F X cores, thereby reestablishing long-lived electron transfer to either F A and/or F B . These studies led to the following generalizations for  (17)(18)(19) is a notable exception in which the heavy metal may cross-link the cysteine thiolates in the F B site thereby retaining three-dimensional structure.) Significantly, S ϭ 3/2, [4Fe-4S] clusters were tentatively identified in unbound C14S PsaC , C51S PsaC , C14A PsaC , and C51A PsaC proteins at the modified site.
The goals of this study are three-fold: 1) to test the hypothesis set forth in Ref. 16 that two intact [4Fe-4S] clusters must be present in mutant PsaC proteins to bind to P700-F X cores, 2) to probe the identity of the ligand to the cubane iron in the modified sites of the mutant PsaC proteins and the spin state of the mixed-ligand [4Fe-4S] clusters, and 3) to determine whether the modified [4Fe-4S] clusters are capable of functioning in electron transfer at cryogenic and room temperatures. We detail the structural, functional, and dynamic properties of mutant PS I complexes, which have been reconstituted with mutant PsaC proteins that contain serine, alanine, or glycine at positions 14 and/or 51.

MATERIALS AND METHODS
Biochemical Protocols-The construction of mutant PS I complexes containing serine, alanine, or glycine in positions 14 and 51 of PsaC was accomplished using a three-step procedure: (i) biochemical resolution of a P700-F A /F B complex into P700-F X cores; (ii) production of the C14X PsaC , and C51X PsaC (X ϭ S, A, or G) holoproteins by overexpressing the site-modified psaC genes in E. coli, purifying the apoproteins, and in vitro reinsertion of the iron-sulfur clusters with iron, sulfide, and ␤-mercaptoethanol (or dithiothreitol); and (iii) rebinding of the PsaC holoproteins to P700-F X cores in the presence of PsaD and PsaE (summarized in Ref. 20).
Site-directed mutagenesis was performed as described previously (21,22); overproduction and purification of wild-type PsaC and mutant C14X PsaC and C51X PsaC (X ϭ S, A, or G) proteins were performed as described (16). Mutations were verified by subjecting apoproteins (or tryptic peptides as appropriate) to amino acid sequence analysis as described (16). Nostoc sp. strain PCC 8009 PsaD was purified from solubilized inclusion bodies isolated from E. coli cells (23). The refolded protein was purified by chromatography on CM-Sepharose CL-6B using a linear gradient of sodium chloride (50 -1000 mM). Synechococcus sp. PCC 7002 PsaE was purified from the solubilized inclusion body by chromatography on DEAE-Sepharose CL-6B with a linear gradient of sodium chloride (10 -200 mM) at pH 8.0. The reassembly of the Photosystem I complex was accomplished simultaneously with the insertion of the iron-sulfur clusters into the wild-type and mutant PsaC apoproteins as described in Ref. 24. All PS I complexes were reconstituted in the presence of ␤-mercaptoethanol unless otherwise noted.
Photoreductase activities of the wild-type and mutant PS I complexes were measured according to the methods described in Ref. 19. Rates of flavodoxin photoreduction were made by monitoring the rate of change in the absorption of flavodoxin at 467 nm. Flavodoxin was purified by DEAE-Sepharose and Sephadex G-75 chromatography from a strain of E. coli containing the Synechococcus sp. strain PCC 7002 isiB gene (25). Rates of flavodoxin-mediated or ferredoxin-mediated NADP ϩ photoreduction were measured as described (15,26).
Time-resolved Optical Absorption Spectroscopy-Transient absorbance changes of P700 at 820 nm (⌬A 820 ) were measured from the microseconds to tens-of-seconds time domain as described in Ref. 29. Each sample was excited at flash energies of 250 J and 53 mJ, the latter being saturating to P700 but generating a substantial contribution of 3 Chl decay to the ⌬A 820 kinetics. The former flash energy brought about 80% saturation of tens-of-milliseconds time domain components in studied mutants, but was sufficiently low in intensity to induce little 3 Chl decay. The curves were fitted to "sum of several exponentials with base-line" using the Marquardt algorithm in Igor Pro (Lake Oswego, OR). The user-defined fit function enabled a fit up to 7 exponentials with all amplitudes and rate constants set free during the fit. In most cases the fit comprised a base-line component accounting for long term phases and/or possible drift of signal zero during long time scale acquisition. The quality of the fit was estimated using standard techniques including analyses of the residuals plots and comparison of the ␥-square values and standard errors of the fit parameters between different fits. Depending on the number of electron acceptors and on their relative contribution to the overall backreaction, the kinetics can be accurately fit by a sum of 4 -6 exponentials. Using this approach to deconvoluting the kinetics, the PS I backreactions can be compared visually on a log time scale, and the ratio between the different pathways that contribute to the backreactions with P700 ϩ can be estimated.
Electron Paramagnetic Resonance Spectroscopy-Electron paramagnetic resonance (EPR) studies were performed using a Bruker ECS-106 X-band spectrometer equipped with either a dual-mode (DM/4116) or standard perpendicular mode (ER/4102ST) resonator. The experimental protocol involves either freezing the sample in darkness and illuminating at low temperature to transfer one electron to the acceptor system, or freezing under continuous illumination to photoaccumulate two or more electrons in the acceptor system. Samples contained either 0.5 mg ml Ϫ1 (wild-type) or 1 mg ml Ϫ1 (mutants) Chl, 1 mM sodium ascorbate, 30 M DCPIP in 50 mM Tris-HCl, pH 8.3. For chemical reduction of the F A and F B clusters, samples were suspended at a chlorophyll concentration of either 0.5 mg ml Ϫ1 (wild-type) or 1 mg ml Ϫ1 (mutants) in 250 mM glycine, pH 10 with 50 mM sodium dithionite. All data manipulations and graphics were performed using IGOR Pro 3.0. Simulations of the S ϭ 1/2 EPR spectra were performed using the EPRSim XOP (provided by Dr. John Boswell, Oregon Graduate Institute), an adaptation of the program QPOW (30 -32).

EPR Studies
Oxidized and Reduced C14X PsaC ⅐PS I and C51X PsaC ⅐PS I (where X ϭ A, G, or S) Complexes-When frozen in darkness in the presence of oxygen or K 3 Fe(CN) 6 Ϫ3 , the C14X PsaC ⅐PS I and C51X PsaC ⅐PS I complexes (where X ϭ A, G, or S) failed to show resonances around g ϭ 2.02 that could be derived from an oxidized S ϭ 1/2 ground state, [3Fe-4S] 1ϩ cluster. Similar to the C14D PsaC ⅐PS I complex (14) and the C51D PsaC ⅐PS I complex (15), a search in the region of g ϭ 10 in dithionite-reduced PS I complexes with the resonator tuned to parallel mode to detect integer spin systems did not reveal any resonances that could be attributed to a reduced S ϭ 2, [3Fe-4S] 0 cluster (data not shown). Rather, two [4Fe-4S] clusters were found in all C14X PsaC ⅐PS I and C51X PsaC ⅐PS I complexes. The EPR characteristics of the individual mutant PS I complexes are described in detail below.
C14A PsaC ⅐PS I Complex-Illumination of a dark-frozen C14A PsaC ⅐PS I complex at 15 K leads to the appearance of resonances characteristic of F A with g values of 2.045, 1.943, and 1.856 (Fig. 1A). Similar to the wild-type cyanobacterial PS I complex, for which microwave power saturation renders the F A resonances maximum at ϳ9 K (Ϫ10 dB microwave power; data not shown), the amplitude of the F A resonances steadily increases in the mutant PS I complex as the temperature is lowered from 12 K (Fig. 1B), becoming maximal between 9 K (Fig. 1C) and 6 K (Fig. 1D), and declining at lower temperatures (data not shown). At 12 K, a second set of resonances become prominent (Fig. 1B, arrows), with a low field g value of 2.115, and unresolved mid-and high field resonances falling between 355 and 375 mT. This signal, termed F B Ј reaches a maximum amplitude at 9 K (Fig. 1C, arrows) and decreases in amplitude at 6 K ( Fig. 1D) and lower temperatures (data not shown). When the C14A PsaC ⅐PS I complex is frozen during illumination, a spectrum appears with g values of 2.047, 1.937, 1.919, and 1.884 ( Fig. 1, E-H), which is nearly identical to the interaction spectrum seen in the wild-type PS I complex. Unlike the latter, where the temperature optimum is ϳ18 K (Ϫ10 dB microwave power), the interaction spectrum derived from reduced F A and F B Ј increases in amplitude as the temperature is lowered from 18 K to 6 K ( Fig. 1, F-H) and declines at lower temperatures (data not shown). The minor resonance at g ϭ 1.756 ( Fig. 1H) indicates that a small amount of F X becomes reduced under conditions of photoaccumulation.
When dithiothreitol was used in the reconstitution protocol in lieu of ␤-mercaptoethanol, the EPR spectral results were identical to that shown in Fig. 1

(A-H) (data not shown).
C14G PsaC ⅐PS I Complex-Like alanine, glycine is incapable of providing a ligand to an iron-sulfur cluster. Yet, the g values, linewidths, and temperature dependence of F A and F B Ј in the C14G PsaC ⅐PS I complex (see Fig. 3D for the 9 K spectrum) were identical to those in the C14A PsaC ⅐PS I complex. Photoaccumulation of the C14G PsaC ⅐PS I complex (data not shown) leads to an interaction spectrum from reduced F A and F B Ј identical to the wild-type PS I complex.
C14S PsaC ⅐PS I Complex-In contrast, serine is capable of providing an oxygen ligand to an iron-sulfur cluster, and based on precedent with similar site-directed mutations (33,34), could support the presence of a [4Fe-4S] cluster. When the dark-frozen C14S PsaC ⅐PS I complex is illuminated at 15 K, a set of resonances appear at g ϭ 2.045, 1.943, and 1.856 characteristic of F A ( Fig. 2A). The temperature dependence (Fig. 2, B-D) is identical to that of F A in the C14A PsaC ⅐PS I complex (see Fig.  1). Similarly, a low field resonance at g ϭ 2.115, and mid-and high field resonances falling between 355 and 375 mT (Fig. 2, B and C, arrows) are present, which have the same spectral appearances and temperature dependences as F B Ј in the C14A PsaC ⅐PS I complex. An interaction spectrum produced by photoaccumulation (Fig. 2, E-H) is also similar to that for the C14A PsaC ⅐PS I and wild-type PS I complexes. The only difference from the EPR properties of C14A PsaC ⅐PS I is the relatively intense set of resonances at g ϭ 2.119, 1.858, and 1.756 due to reduced F X (Fig. 2H).
An identical set of resonances were found in the dithiothreitol-reconstituted C14S PsaC ⅐PS I complex (data not shown). The amount reduced F X is smaller under conditions of photoaccumulation, probably indicating a higher degree of overall reconstitution with dithiothreitol.
EPR Spectrum of F B Ј-The absence of magnetic interaction between redox centers when PS I complexes are frozen in the dark and illuminated at low temperatures allows the EPR spectrum of F B Ј to be extracted from the admixture of F A and F B Ј. The F B Ј spectrum, obtained by subtracting a simulated This was the expected result, since dithiothreitol is the equivalent of two tail-to-tail ␤-mercaptoethanol molecules. The accuracy of the simulations is illustrated in Fig. 3D, which compares an additive composite of F A and F B Ј (Fig. 3D, solid line) with the experimental spectrum of the C14G PsaC ⅐PS I complex (Fig. 3D, dotted line). Taking into account the different spin populations at their respective temperature optima of 6 K and at 9 K, double integration of the simulated F A and F B Ј spectra show that they are photoreduced in a ratio of ϳ1:2. In contrast, F A and F B are reduced at a ratio of ϳ3.2:1 in wild-type PS I complexes under similar conditions (data not shown).
C51A PsaC ⅐PS I Complex-The C51A PsaC ⅐PS I complex shows a set of EPR signals, which are not as easily interpreted as those in the C14A PsaC ⅐PS I complex. Illumination of a darkfrozen C51A PsaC ⅐PS I complex at 15 K leads to the appearance of resonances at g ϭ 2.069, 1.930, and 1.880 characteristic of F B (Fig. 4A). As the temperature is lowered from 12 K to 6 K (Fig.  4, B-D), the F B resonances diminish in amplitude due to the onset of microwave power saturation, behavior that is similar to the F B cluster in the wild-type and the C51D PsaC ⅐PS I complex (15). A second set of resonances derived from an ironsulfur cluster is visible at 15 K (Fig. 4A) with g values of 2.045, 1.943, and 1.856 characteristic of F A . As the temperature is lowered from 12 K to 6 K (Fig. 4, B-D), the amplitude of these resonances increases, consistent with the behavior of F A in the wild-type cyanobacterial PS I complex. When the C51A PsaC ⅐PS I complex is frozen during illumination, a complex set of resonances is found, which can be separated into two temperaturedependent sets. The first set is best observed at 6 K (Fig. 4G) and below (data not shown) and resembles the spin-coupled, interaction spectrum of F A and F B seen in wild-type PS I complexes. The second set is superimposed on these resonances and is best seen at 15 K (Fig. 4E). The extracted g values of 2.062, 1.929, and 1.900, obtained by subtracting the 6 K spectrum from the 15 K spectrum after suitable scaling, show that this set of resonances is derived from a single, non-interacting F B cluster.
When the C51A PsaC ⅐PS I complex is reconstituted using dithiothreitol, a smaller spin population of F B is found when the sample is illuminated at 15 K. A spin-coupled, interaction spectrum is present with resonances at g ϭ 2.047, 1.937, 1.919, and 1.884, similar to that in the wild-type PS I complex, and there is no evidence for the presence of a non-interacting F B FIG. 3. EPR spectrum of the F B iron-sulfur cluster. A, F B Ј in the C14A PsaC ⅐PS I complex. The extracted F B Ј spectrum was obtained by subtracting a simulated spectrum of F A from the 9 K spectrum (Fig. 1C). B, F B Ј in the C14S PsaC ⅐PS I complex. The extracted F B Ј spectrum was obtained by subtracting a simulated spectrum of F A from the 9 K spectrum (Fig. 2C) cluster (data not shown). Hence, the non-interacting population of F B under conditions of photoaccumulation correlates with the amount of F B that is visible under conditions of illumination at 15 K (Fig. 4A).
C51G PsaC ⅐PS I Complex-The EPR spectra and the temperature dependence of F A and F B in the C51G PsaC ⅐PS I complex are similar to those of the C51A PsaC ⅐PS I complex. One minor difference is that when the sample is illuminated at low temperature, the spin concentration of the two clusters and that of P700 ϩ are lower than expected (data not shown). The EPR spectrum of the C51G PsaC ⅐PS I complex shows an interaction spectrum similar to the wild-type (data not shown).
C51S PsaC ⅐PS I Complex-When the dark-frozen C51S PsaC ⅐PS I complex is illuminated at 15 K (Fig. 5A), there is evidence for two independent spin systems, represented by F A at g ϭ 2.045, 1.943, and 1.856 and F B at g ϭ 2.069, 1.930, and 1.880. As the temperature is lowered from 15 K to 6 K (Fig. 5, A-D), the amplitude of the F B resonances declines in amplitude due to the onset of microwave power saturation. The F A resonances are weak, but like the wild-type PS I complex, they increase in amplitude as the temperature is lowered from 15 K to 6 K (Fig.  5, A-D). Based on the amplitude of the resonances under conditions of photoaccumulation (Fig. 5, E-H), and on the relative ratio of the resonances in the C51A PsaC ⅐PS I complex (Fig. 4, A  versus B), the number of spins represented by F A is still lower than expected. When the C51S PsaC ⅐PS I complex is illuminated during freezing, two distinct spin systems are present, which can be distinguished on the basis of their temperature depend-ence. At 15 K (Fig. 5E) and 12 K (Fig. 5F), a set of resonances are found with g values of 2.047, 1.937, 1.919, and 1.884, which can be interpreted as derived from spin coupling between reduced F A and F B (Fig. 5E). As the temperature is lowered to 9 K (Fig. 5G, arrows) and 6 K (Fig. 5H, arrows)

Optical Studies
Optical Kinetic Spectroscopy of the Reconstituted Photosystem I Complexes-Because charge separation between P700 and [F A /F B ] is irreversible at cryogenic temperatures, the EPR data obtained by continuous illumination provide little information on the ability of the mixed-ligand iron-sulfur clusters to function as efficient electron carriers. The latter can be inferred by measuring flash-induced absorbance changes at 820 nm, where the time constants derived from the kinetic transients identify backreactions of the various electron acceptors with P700 ϩ . A complicating feature is that the backreaction kinetics of most electron acceptors in PS I are multiphasic. To accom- plish a meaningful analysis that encompasses all of the electron acceptors, the kinetic decomposition is carried out globally over 6 orders of magnitude of time. This allows monitoring of a continuum of kinetic phases corresponding to P700 ϩ backreactions, as well as those arising from reactions of P700 ϩ with redox agents in the media. The approximate lifetimes of charge recombination from different components of the PS I acceptor side to P700 ϩ range from 35 ns for A 0 (35); 10 and 110 s for A 1 (36); 500 s and 3 ms for F X (26); and 15 and 80 ms for F A /F B (29).
Wild-type PS I Complex and P700-F X Core-In wild-type PS I complexes, the quantum yield of electron transfer approaches 1.0; hence, the majority of the backreactions should be derived from [F A /F B ] Ϫ . In n-dodecyl-␤-D-maltoside PS I particles from Synechococcus sp. PCC 6301, 41% of the recombination kinetics are derived from the 17-ms and 65-ms components attributed to P700 ϩ [F A /F B ] Ϫ recombination (data not shown; see Ref. 29). An additional 35% are derived from slower phases with lifetimes of 282 ms and 2.8 s, resulting in a 76% efficient transfer to F A /F B (in thylakoid membranes the efficiency of electron transfer to F A /F B is 94%). The slowest kinetic phases are due to exogenous donors undergoing redox reactions with P700 ϩ and come about in reaction centers where [F A /F B ] Ϫ has become oxidized by one or more exogenous electron acceptors in the medium. The sum contribution of earlier acceptors, including F X Ϫ and A 1 Ϫ , is 24% of the total absorption change. In P700-F X cores, the [F A /F B ] Ϫ backreaction is replaced with one or more faster components arising from earlier electron acceptors (Fig.  6A).

The 19-ms transient representing the unresolved [F A /F B ]
Ϫ backreaction contributes only about 3%, and those with slower phases representing exogenous donors to P700 ϩ contribute an additional 16% of the total absorption change, resulting in a 19% efficient transfer to residual F A /F B . Over 40% of the absorption change derives from F X Ϫ , with lifetime components of 670 s and 2.1 ms (37). An additional 40% of the absorption change is contributed by A 1 Ϫ , with lifetime components of 15 s and 110 s (38). In a PS I complex reconstituted with wild-type PsaC, much of the wild-type kinetics is recovered (Fig. 6B). The [FA/F B ] Ϫ backreaction, with lifetime components of 19 and 73 ms contributes 41% to the total absorption change; an additional 35% is contributed by the slower donation to P700 ϩ by external donors, resulting in an overall transfer efficiency of 76% to F A /F B . The relative contribution of the F X Ϫ backreaction, represented by the 441-s and 3.2-ms kinetic phases, has declined to about 14% of the total absorption change, and 10% of the absorption change occurs with a lifetime of 35 s, which may represent an unresolved A 1 Ϫ backreaction. Hence, the yield of reconstitution of PsaC onto P700-F X cores is reflected in the large contribution of long-lived charge separation.
C14A PsaC ⅐PS I and C51A PsaC ⅐PS I Complexes-The kinetics of the reconstituted C14A PsaC ⅐PS I complex show that only a fraction of the wild-type kinetics is recovered (Fig. 6C). The [F A /F B ] Ϫ backreaction, with lifetime components of 8 and 68 ms, contributes 10% to the total absorption change; an additional 11% is contributed by the slower donation to P700 ϩ by external donors, leading to a transfer efficiency of 21% to F A / F B . The backreaction occurs primarily from components with lifetimes of 459 s (17%) and 1.9 ms (7%) derived from F X , and over 55% occurs from a component with a lifetime of 38 s. The C51A PsaC ⅐PS I mutant complex shows similar kinetic behavior (Fig. 6D) of 7.2 and 78 ms contributes 15% to the total absorption change; an additional 13% is contributed by the slower donation to P700 ϩ by external donors, resulting in a transfer efficiency of 28% to F A /F B . Some of the backreaction occurs from components with lifetimes of 413 s (22%) and 1.3 ms (10.5%) derived from F X , but ϳ40% occurs from a component or components with a lifetime of 39 s. C14G PsaC ⅐PS I and C51G PsaC ⅐PS I Complexes-The reconstituted C14G PsaC ⅐PS I and C51G PsaC ⅐PS I complexes show a greater degree of recovery of wild-type kinetics. In the C14G PsaC ⅐PS I complex, the [F A /F B ] Ϫ backreaction, with lifetime components of 34 and 209 ms, contributes 56% to the total absorption change; an additional 11% is contributed by the slower donation to P700 ϩ by external donors, resulting in a 67% efficient transfer to F A /F B (Fig. 7A). A smaller percentage of the backreaction occurs from components with lifetimes of 586 s (11%) and 6.2 ms (9%) derived from F X , and only 15% occurs from a component with a lifetime of 63 s. In the C51G PsaC ⅐PS I complex, the [F A /F B ] Ϫ backreaction, with lifetime components of 26 and 119 ms, contributes 36% to the total absorption change (Fig. 7B). An additional 15% is contributed by the slower donation to P700 ϩ by external donors, leading to a 51% efficient transfer to F A /F B . Some of the backreaction occurs from components with lifetimes of 783 s (12%) and 4.6 ms (9%) derived from F X , and 17% occurs from a component or components with a lifetime of 106 s.
C14S PsaC ⅐PS I and C51S PsaC ⅐PS I Complexes-When the reconstituted C14S PsaC ⅐PS I complex is analyzed (Fig. 7C), 31% of the total absorption change is derived from 16-ms and 58-ms lifetime components attributed to [F A /F B ] Ϫ . An additional 19% is due to slower donation to P700 ϩ from exogenous electron donors, leading to a 50% efficient transfer to F A /F B . The relative contribution of the F X Ϫ backreaction, judged by the 550-s and 2.1-ms kinetic phases, equals 22% of the total absorption change, a value higher than that found with wild-type PsaC. About 28% of the total absorption change is derived from the 33-s component related to the backreaction from A 1 Ϫ . Electron transfer in the C51SPsaC⅐PS I complex (Fig. 7D) appears slightly more efficient, with 37% of the absorption change contributed by the 17-ms and 51-ms components derived from [F A /F B ] Ϫ and an additional 19% derived from slower donation to P700 ϩ from exogenous electron donors, leading to a 56% efficient transfer to F A /F B . About 23% of the total absorption change is due to F X Ϫ , with lifetime components of 507 s and 5.2 ms, and about 21% is due to the 31-s component discussed above.
Cytochrome c 6 -NADP ϩ Reductase Activity of the Reconstituted PS I Complexes-While single turnover flashes provide data on the efficiency of electron transfer to the mixed-ligand iron-sulfur clusters, it does not indicate whether donation of electrons occurs from the clusters to the physiologically relevant electron acceptors ferredoxin and flavodoxin. Table I shows rates of NADP ϩ photoreduction mediated by flavodoxin and ferredoxin in a wild-type PS I complex, a P700-F X core, and in PS I complexes reconstituted with PsaC, C14A PsaC , C51A PsaC , C14S PsaC , and C51S PsaC . In wild-type complexes, the reduction of flavodoxin (i) and the reduction of NADP ϩ mediated by flavodoxin (ii) or ferredoxin (iii), occurs at rates of 900, 930, and 820 mol (mg Chl) Ϫ1 h Ϫ1 , respectively. The comparable rates in P700-F X cores are only 7% (i), 17% (ii), and 6% (iii), respectively, of the wild-type rates. The residual rates are probably derived from a small amount of retained PsaC, indicating that F X is unable to transfer electrons efficiently to either flavodoxin or ferredoxin. When the P700-F X cores are reconstituted with PsaC, the comparable rates are 53% (i), 62% (ii), and 60% (iii), respectively, of the control rates. The mutant C14A PsaC ⅐PS I complex supports rates that are 17% (i), 29% (ii), and 24% (iii), respectively, of the control rates, and the mutant C51A PsaC ⅐PS I complex supports rates that are 25% (i), 34% (ii), and 35% (iii) of the control rates, respectively. These low steady-state rates are consistent with the relatively inefficient electron transfer rates on a single flash for the alanine mutants. When C14S PsaC ⅐PS I is used in the reconstitution protocol, the recovery is 62% (i), 56% (ii), and 61% (iii) of the wildtype rates, and when C51S PsaC ⅐PS I is used in the reconstitution protocol, the recovery is 82% (i), 70% (ii), and 76% (iii) of the wild-type rates. Note that when compared with the PsaC-reconstituted complex, both C14S PsaC ⅐PS I and C51S PsaC ⅐PS I support electron transfer rates to flavodoxin and ferredoxin that are equivalent to those rates. The rates for the C14D PsaC ⅐PS I and C51D PsaC ⅐PS I complexes have been published elsewhere (14,15) and are reproduced in Table I to show that they fall between the alanine and serine sets (the glycine mutants were not studied). The most noteworthy features of these results are that a mixed-ligand iron-sulfur clusters in the C14X PsaC ⅐PS I and C51X PsaC ⅐PS I complexes (where X ϭ D or S) can be as effective as all-cysteine [4Fe-4S] clusters in the PsaC⅐PS I complexes in establishing electron throughput from P700 to NADP ϩ .

Two intact [4Fe-4S] Clusters Are Required for PsaC
Binding to P700-F X Cores-Similar to the aspartate series of PS I mutants (14,15), neither S ϭ 1/2, [3Fe-4S] 1ϩ clusters (under oxidizing conditions), nor S ϭ 2, [3Fe-4S] 0 clusters (under reducing conditions) could be found in the reconstituted C14X PsaC ⅐PS I and C51X PsaC ⅐PS I complexes (where X ϭ A, G, or S). In the unbound C14X PsaC and C51X PsaC proteins, the altered site is presumed to be occupied by a mixed population of [3Fe-4S] and S ϭ 3/2, [4Fe-4S] clusters (16). The data show that only those proteins containing two [4Fe-4S] clusters are capable of binding to P700-F X cores. The presence of two intact [4Fe-4S] clusters in the mutant complexes is inferred in the C14X PsaC ⅐PS I and C51X PsaC ⅐PS I complexes by the presence of a spin-coupled EPR spectrum observed when F A and F B are simultaneously reduced within the same reaction center. The g values and linewidths are strikingly similar to the F A /F B interaction spectrum in wild-type PS I complexes. The presence of occupied F A and F B sites in in vitro C14D PsaC ⅐PS I (14) and C51D PsaC ⅐PS I (15) complexes was demonstrated by successiveflash experiments in which two electrons must be promoted before the backreaction proceeds from F X Ϫ . The multiple flash study shows that there is no missing cluster in the modified site of the in vitro aspartate mutants as has been intimated in an in vivo cysteine 13 (equivalent to cysteine 14 in this study) to aspartate change in PsaC of Anabaena variabilis (39). In support of this conclusion, a study of in vivo mutations to introduce C14X PsaC and C51X PsaC (where X ϭ D, S, or A) into Synechocystis sp. PCC 6803 (40)  However, when these mutations were introduced in vivo in Synechocystis sp. PCC 6803, the alanine substitution did not lead to the accumulation of a PsaC on the PS I complexes, and the serine and aspartate substitutions led to a presumed population of S ϭ 3/2 ground state [4Fe-4S] clusters which were not detectable in the g ϭ 2 region (40). Hence, the presence of an oxygen-ligated, iron-sulfur cluster at the C14 position is not supported by the evidence. A second candidate is the ␤-mercaptoethanol used in the iron-sulfur insertion protocol of the mutant PsaC proteins. Unlike the presumed S ϭ 3/2 spin state of an oxygen-ligated cluster at position 14, the spin state of a thiolate-ligated cluster is S ϭ 1/2 when the PsaC protein is rebound to P700-F X cores. Since the spin state of the thiolate-ligated cluster is S ϭ 3/2 in the unbound mutant PsaC proteins, a cross-over must occur to S ϭ 1/2 when the mutant PsaC protein is rebound. The factors that lead to a change of spin state have not been systematically studied in iron-sulfur proteins and the energetics of the cross-over are likely to be subtle and hence difficult to predict.
Proposed Identity of the Ligand to the Rescued [4Fe-4S] Cluster in C51X PsaC ⅐PS I complexes-In contrast to the uniformity of the C14X PsaC ⅐PS I complexes, there are spectral differences among the various C51X PsaC ⅐PS I complexes, which make the interpretation of the EPR spectra challenging. First, most complexes show wild-type resonances, which resemble those of wild-type F A and F B , but the relative intensities of these resonances are low compared with those in the C14X PsaC ⅐PS I complexes. The F B cluster occupies the non-mutant site in the C51X PsaC ⅐PS I complex and by analogy with the wild-type might be expected to show a low level of photoreduction at 15 K. The low spin concentration of F A is therefore surprising, especially since the spin concentration of irreversible P700 ϩ is identical to that for the wild-type and to the C14X PsaC ⅐PS I complexes. We suggest that this is due to the presence of a [4Fe-4S] cluster in the mutant site that cannot be observed in the g ϭ 2 region. The missing spins are likely represented by a higher spin state, presumably S ϭ 3/2, for the F A cluster. The proposed existence of S ϭ 3/2, [4Fe-4S] clusters in the mixedligand site of the unbound C14A PsaC , C14S PsaC , C51A PsaC , and C51S PsaC proteins is consistent with this suggestion (16). While no evidence for signals derived from a S ϭ 3/2 spin system was found in the region of g ϭ 5-6, these resonances would have been difficult, if not impossible, to detect at the sample concentrations employed in this study.
The presence of two [4Fe-4S] clusters in the mutant C51X PsaC ⅐PS I complexes was demonstrated by the presence of a spin-coupled spectrum observed when both F A and F B are simultaneously reduced within the same reaction center. How-  (40) . Electron Transfer Inefficiency Versus Sample Heterogeneity-Using optical kinetic spectroscopy as a probe of electron transfer into the modified clusters, we found that the reconstituted C14X PsaC ⅐PS I and C51X PsaC ⅐PS I complexes function as relatively efficient electron carriers at room temperature. The results of room temperature measurements of P700 ϩ decay kinetics indicate a high efficiency of F A /F B photoreduction comparable with that of a PS I complex reconstructed with wildtype PsaC in all but the alanine mutants. It is suggested that the presence of hydrophobic methyl group in the alanine mutants prevents a high degree of external thiolate access to the iron atom, thus leading to a lower yield of reconstitution with the mutant PsaC proteins. It is also likely that the alanine mutant samples are heterogeneous, containing a subpopulation of the non-reconstructed P700-F X cores. However, a complication arises because sample heterogeneity due to incomplete reconstitution gives rise to the same kinetics in single turnover experiments as does forward electron transfer inefficiency. In performing the optical measurements, an interesting difference in the C14S PsaC ⅐PS I and C51S PsaC ⅐PS I complexes was noted; of the two kinetic phases that are attributed to the backreaction from the F A and F B clusters, the shorter phase (ϳ15-18 ms) is predominant in the reconstituted C51S PsaC ⅐PS I complex, whereas in the wild-type, the PsaC-reconstituted and the C14S PsaC ⅐PS I complex, the longer kinetic phase (ϳ50 -80 ms) is predominant. However, the significance of the two kinetic phases derived from the [F A /F B ] Ϫ backreaction is not clear; both phases are present in Hg-treated PS I complexes, which lack F B , implying that neither can be assigned uniquely to either terminal electron acceptor (29).
Comparison with Other Cysteine to Serine Mutations in [4Fe-4S] Ferredoxins-It is interesting that mixed-ligand iron-sulfur clusters are very rarely found in naturally occurring soluble ferredoxins or in other complex iron-sulfur proteins. It may be that sulfur is a much better ligand than oxygen in supporting biogenesis of iron-sulfur proteins in vivo, and that assembly considerations are more important than electron-transport considerations in dictating the selection of ligands during evolution of iron-sulfur proteins. With the exception of the nitrogenase cluster (3), the few examples of cubane clusters supported by 3 cysteines and 1 serine have been engineered by site-directed mutagenesis. Fumarate reductase with Cys 204 3 Ser, Cys 210 3 Ser, and Cys 214 3 Ser mutations result in enzymes with negligible activity and with subunits that have dissociated from the membrane (43,44). One of the [4Fe-4S] clusters of dimethyl sulfoxide reductase (45) has been changed to a [3Fe-4S] cluster by converting cysteine 102 to serine. There is one example of a modified protein containing a [4Fe-4S] cluster supported by 3 cysteines and 1 serine: the genetically engineered C565S PsaB and C556S PsaB mutants in the F X cluster of Synechocystis sp. PCC 6803. This is a rare example of a interpolypeptide [4Fe-4S] cluster with cysteine ligands contributed by two separate polypeptides (there are several homologs of NifH such as the dark protochlorophyllide reductase of plants, which, on the basis of sequence homology, are also presumed to have interpolypeptide iron-sulfur clusters). The C565S PsaB and C556S PsaB mutants were capable of functioning as an electron transfer carrier from the intermediate electron acceptor, A 1 Ϫ , to the terminal iron-sulfur clusters, F A and F B (37); however, the efficiency of forward electron transfer was lower than the wild-type. The working hypothesis, yet to be tested, is that the reduction potential of the F X cluster is driven more electronegative by the change from cysteine to serine. In the present work we show that such a mixed-ligand [4Fe-4S] cluster in PsaC can be sufficiently functional to mediate electron transfer from F X to ferredoxin and flavodoxin.
Conclusions-Mutant PS I complexes reconstituted in vitro contain two [4Fe-4S] clusters regardless of whether the mutation at C14 or C51 is aspartate (14,15), alanine, glycine, or serine (this work). The identical EPR spectra of the F B Ј cluster in the C14D PsaC ⅐PS I, C14A PsaC ⅐PS I, C14G PsaC ⅐PS I, and C14S PsaC ⅐PS I complexes, along with the inability of alanine and glycine to supply a suitable ligand, suggest that the fourth coordination site is occupied by an external thiolate. The thiolate would likely be supplied by a ␤-mercaptoethanol or dithiothreitol retained from the cysteine exchange reaction between the inorganic [4Fe-4S] clusters and the mutant PsaC apoproteins. The EPR spectrum of the F A cluster in the C51S PsaC ⅐PS I complex differs from the C51A PsaC ⅐PS I or C51G PsaC ⅐PS I complexes, implying that serine has displaced the ␤-mercaptoethanol at this site. This may be a consequence of steric hindrance at this site that comes about when the holoprotein is bound to the P700-F X core. One generalization that follows from this work is that the mixed-ligand iron-sulfur cluster in the free mutant PsaC exists in a high spin state (probably S ϭ 3/2) when either oxygen from serine (or aspartate) or the thiolate from ␤-mercaptoethanol (or dithiothrietol) provides the ligand to the open coordination site. The spin state changes to S ϭ 1/2 when the thiolate-ligated cluster, but not the oxygenligated cluster, is bound to the P700-F X core. A spin-coupled, interaction spectrum identical to that seen in a wild-type PS I complex is observed for all but the PS I-C51S complex; the latter shows an additional set of resonances that may be derived from the serine oxygen-ligated [4Fe-4S] cluster. The proposed chemical rescue of a [4Fe-4S] cluster in the site-modified proteins using external thiolates thus appears to be the critical feature that allows some mutant PsaC proteins to rebind to PS I. These studies indicate that mixed-ligand (3 cysteine, 1 serine) and chemically rescued (3 cysteine, 1 presumed external thiolate) [4Fe-4S] clusters are capable of participating as an electron carrier in a physiologically relevant setting, i.e. within a functional PS I complex.